File: TitlePageBook-for-WE3_2022-v01.

fm July 6, 2022

Water Distribution Systems Engineering

Book for Courses CEME3005 Advanced Civil Engineering Hydraulics &

CEME7305 Advanced Civil Engineering Hydraulics

Chapters 1 to 22, Appendix A, B, C and D

Semester 2, 2022

Prof. Emeritus Angus Simpson

[email protected]

School of Civil, Environmental & Mining Engineering

University of Adelaide, Australia 5005
© Copyright 2022Jul. Prof. Emeritus Angus Ross Simpson, University of Adelaide, Adelaide.
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

CHAPTER 1 INTRODUCTION
1.1 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Overview of Water Collection and Distribution . . . . . . . . . . . 2
1.3 Design Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3.1 Planning of water distribution systems . . . . . . . . . . . . . . . 2
1.3.2 Supply standards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.3 Design requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Computer Analysis of Water Distribution Systems . . . . . . . . . 3
1.5 Steady State Analysis versus Extended Period Simulation . . . 4
1.6 Defining a Water Distribution System . . . . . . . . . . . . . . . . . . . 4
1.7 Trial and Error Design of Water Distribution Systems . . . . . . 6
1.8 Optimization of the Design of Water Distribution Systems . . 7
1.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

CHAPTER 2 PROPERTIES OF WATER

2.1 Fluids, Solids and Gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Types of stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Definition of a fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.3 Behavior of solids and fluids under the action of stresses 10
2.2 Units of Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.1 Newton’s second law of motion . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Force in SI units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Force in US customary units . . . . . . . . . . . . . . . . . . . . . . 13
2.2.4 Summary of properties of water . . . . . . . . . . . . . . . . . . . . 13
2.2.5 Common water quantities . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.6 Conversion of units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1 Definition of absolute viscosity . . . . . . . . . . . . . . . . . . . . 16
2.3.2 Units of absolute viscosity . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.3 Absolute viscosities of substances . . . . . . . . . . . . . . . . . . 19
2.3.4 Kinematic viscosity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.5 When is viscosity important? (Definition of Reynolds
number) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4 Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.4.1 Density of water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.2 When is the density of water important? . . . . . . . . . . . . . 21
2.4.3 Density of gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5 Specific Weight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6 Specific Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.7 Vapor Pressure of Water . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7.1 When is the vapor pressure of water important? . . . . . . . 26
2.8 Bulk Modulus of Elasticity of a Fluid . . . . . . . . . . . . . . . . . . 27
2.8.1 When is the bulk modulus of elasticity K important? . . . 28

i
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

2.9 Other Fluid Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.10 Pressure Variation in a Fluid . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.10.1 Relation between pressure and column height . . . . . . . . . 31
2.11 Absolute and Gage Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.12 Measurement of Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.12.1 Piezometers and pressure gages . . . . . . . . . . . . . . . . . . . . 35
2.12.2 Measuring atmospheric pressure . . . . . . . . . . . . . . . . . . . 35
2.12.3 Manometers or open piezometers for measuring pressure 36
2.12.4 Basic principles for analysis of manometer problems . . . 36
2.12.5 Procedure for solving manometer problems . . . . . . . . . . 37
2.12.6 Solving for the unknown pressure in the pipeline using a
manometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.13 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

CHAPTER 3 EQUATIONS FOR FLUID

MOTION IN PIPES
3.1 Types of Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1.1 Steady versus unsteady flow . . . . . . . . . . . . . . . . . . . . . . 39
3.1.2 Uniform versus non – uniform flow . . . . . . . . . . . . . . . . . 40
3.1.3 Laminar versus turbulent flow . . . . . . . . . . . . . . . . . . . . . 41
3.1.4 An ideal fluid versus a real fluid . . . . . . . . . . . . . . . . . . . 42
3.1.5 One – , two – and three – dimensional flow . . . . . . . . . . . 42
3.2 Dimensionless Parameters for Pipe Flow . . . . . . . . . . . . . . . 43
3.2.1 Newton’s second law of motion . . . . . . . . . . . . . . . . . . . . 43
3.2.2 The Buckingham theorem . . . . . . . . . . . . . . . . . . . . . . . . 45
3.3 The Continuity Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3.1 Assumptions for the analysis of pipe flow . . . . . . . . . . . . 47
3.3.2 Streamlines and streamtubes . . . . . . . . . . . . . . . . . . . . . . 47
3.3.3 Control volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3.4 Mass conservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.4 The Momentum Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4.1 Momentum in solid mechanics versus momentum flux in
hydraulics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.4.2 Force – linear momentum flux equation . . . . . . . . . . . . . . 51
3.4.3 Derivation of the force – momentum flux equation . . . . . 52
3.4.4 Force – momentum flux equation for more than one inflow
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.5 Energy Conservation in Pipe Flow . . . . . . . . . . . . . . . . . . . . 55
3.5.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.5.2 Forms of energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.5.3 Units of head . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.5.4 Potential energy or elevation head . . . . . . . . . . . . . . . . . . 56
3.5.5 Kinetic energy or velocity head . . . . . . . . . . . . . . . . . . . . 57
3.5.6 Pressure energy or pressure head . . . . . . . . . . . . . . . . . . . 58
3.5.7 Mechanical energy and heat energy . . . . . . . . . . . . . . . . . 59

ii
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

3.5.8 Euler’s equation of motion along a streamline . . . . . . . . 59

3.5.9 The Bernoulli equation (excludes losses) . . . . . . . . . . . . 60
3.5.10 Application of the Bernoulli equation to a pitot tube . . . . 62
3.5.11 Head or pressure losses in pipes . . . . . . . . . . . . . . . . . . . . 63
3.5.12 The principle of conservation of energy . . . . . . . . . . . . . 64
3.5.13 The energy equation (includes losses) . . . . . . . . . . . . . . . 65
3.5.14 Pump and turbine power . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5.15 The energy equation without shaft work or minor losses 66
3.6 The Energy Grade Line (EGL) . . . . . . . . . . . . . . . . . . . . . . . 67
3.7 The Hydraulic Grade Line (HGL) . . . . . . . . . . . . . . . . . . . . . 67
3.8 Kinetic Energy Correction Factor . . . . . . . . . . . . . . . . . . . . . 68
3.8.1 Continuity of flow for a non – uniform velocity profile . . 70
3.9 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

CHAPTER 4 PIPE HEAD LOSS

4.1 Features of Pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.1.1 Diameter of a pipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.1.2 Wall thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.1.3 Internal pipe roughness . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2 Pipe Wall Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2.1 Cast iron pipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2.2 Ductile iron cement mortar lined pipe . . . . . . . . . . . . . . . 75
4.2.3 Steel pipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2.4 Plastic pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2.5 Asbestos cement pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2.6 Concrete pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3 Darcy – Weisbach Pipeline Fluid Friction Loss Equation . . . 76
4.3.1 A summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.2 Historical development . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.3.3 Shear stress at the wall of the pipe . . . . . . . . . . . . . . . . . . 81
4.3.4 Dimensional analysis of flow in pipelines . . . . . . . . . . . . 82
4.3.5 Derivation of the Darcy – Weisbach fluid friction equation
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3.6 Non – circular conduits . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.3.7 Variation of head loss with velocity . . . . . . . . . . . . . . . . . 87
4.4 Laminar Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.4.1 Minimum possible Darcy – Weisbach f – value for laminar
flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.4.2 Velocity variation across the pipe for laminar flow . . . . . 88
4.4.3 The head loss for laminar flow . . . . . . . . . . . . . . . . . . . . . 90
4.4.4 Friction factor for laminar pipeline flow . . . . . . . . . . . . . 90
4.5 Changing from Laminar Flow to Turbulent Flow . . . . . . . . . 91
4.6 Losses in Turbulent Flow in Pipelines . . . . . . . . . . . . . . . . . . 92
4.6.1 The boundary layer and the viscous sublayer . . . . . . . . . 92
4.6.2 The Swamee and Jain equation for the Darcy – Weisbach
friction factor f . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

iii
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

4.6.3 Smooth pipe turbulent flow . . . . . . . . . . . . . . . . . . . . . . . 93

4.6.4 Transition between smooth pipe turbulent flow and fully
rough turbulent flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.6.5 Fully rough turbulent flow . . . . . . . . . . . . . . . . . . . . . . . 101
4.6.6 Aging of pipes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.7 Hazen – Williams Equation . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.7.1 The Hazen – Williams head loss equation for SI units . . 107
4.7.2 The Hazen – Williams head loss equation for US customary
units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.7.3 The equivalent pipe diameter for two pipes in parallel for
Hazen – Williams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.7.4 Typical values of the Hazen – Williams coefficient . . . . 112
4.7.5 Comparison of Darcy – Weisbach f and Hazen – Williams C
in SI units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.7.6 Comparison of Darcy – Weisbach f and Hazen – Williams C
in US customary units . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.7.7 Assessment of accuracy of the Hazen – Williams equation .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.8 Generalized Form of Head Loss in a Pipe Segment . . . . . . 124
4.8.1 The resistance factor for the Darcy – Weisbach head loss
equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.8.2 The resistance factor for the Hazen – Williams head loss
equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.9 Converting from Hazen – Williams C to Roughness Height e . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.10 The Manning Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.11 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

CHAPTER 5 MINOR LOSSES IN PIPE

NETWORKS
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.2 Similitude and Dimensional Analysis for Minor Losses . . . 132
5.3 Minor Loss for a Sudden Expansion . . . . . . . . . . . . . . . . . . 134
5.3.1 Minor loss for an exit from a pipe to a downstream reservoir
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.4 Minor Losses for Sudden Contractions . . . . . . . . . . . . . . . . 138
5.5 Minor Losses for Entrances . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.5.1 Square edged entrance . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.5.2 A re – entrant entrance . . . . . . . . . . . . . . . . . . . . . . . . . . 141
5.6 Minor Losses Due to Gradual Changes in Cross – section . . 142
5.6.1 A gradual contraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
5.6.2 A gradual expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
5.7 Bends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
5.8 Valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
5.8.1 Globe valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
iv
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

5.8.2 Plug valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

5.8.3 Gate or slide valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
5.8.4 Butterfly valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
5.8.5 Check valves or non – return valves . . . . . . . . . . . . . . . . 144
5.9 Typical Head Loss Coefficient Values . . . . . . . . . . . . . . . . 144
5.10 Equivalent Pipe Lengths . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.11 When May Minor Losses be Ignored? . . . . . . . . . . . . . . . . . 147
5.12 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

CHAPTER 6 ANALYSIS OF SIMPLE

PIPELINES
6.1 Unknown Pipe Head Loss — an Explicit Expression . . . . . . 150
6.2 Unknown Pipe Discharge – Ignoring Minor Losses . . . . . . 152
6.2.1 Explicit expression for Q assuming smooth pipe turbulent
flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
6.2.2 Explicit expression for Q assuming transitional turbulent
pipe flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
6.2.3 Explicit expression for Q assuming fully rough turbulent
pipe flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
6.2.4 Trial and error determination of Q — including minor losses
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
6.2.5 A comparison of the discharges for different turbulent flow
regimes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
6.3 Which Pipe Carries the Greater Discharge? . . . . . . . . . . . . . 169
6.4 Unknown Pipe Diameter . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
6.4.1 Explicit expression for D assuming smooth pipe turbulent
flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
6.4.2 Explicit expression for D assuming transitional turbulent
flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
6.4.3 Explicit expression for D assuming fully rough turbulent
flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6.4.4 Determining D by trial and error — including the minor
losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6.4.5 Computing the diameter directly when minor losses are
included . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
6.5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

CHAPTER 7 SERIES AND PARALLEL PIPE

SYSTEMS
7.1 Equivalent Uniform Main (EUM) . . . . . . . . . . . . . . . . . . . . 180
7.2 Pipes in Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
7.2.1 Two Darcy – Weisbach pipes in series . . . . . . . . . . . . . . 180
7.2.2 Multiple Darcy – Weisbach pipes in series . . . . . . . . . . 184
7.2.3 Two Hazen – Williams pipes in series (for both SI units and
US customary units) 185

v
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

7.2.4 Multiple Hazen – Williams pipes in series (for both SI units

and US customary units) 188
7.3 Pipes in Parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
7.3.1 Darcy – Weisbach head loss equation for two pipes in
parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
7.3.2 Darcy – Weisbach head loss equation for multiple pipes in
parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
7.3.3 Hazen – Williams head loss equation for pipes in parallel (for
both SI units and US customary units) . . . . . . . . . . . . . . 192
7.3.4 Hazen – Williams head loss equation for multiple pipes in
parallel (for both SI units and US customary units) . . . . 195
7.4 Three Interconnected Reservoirs . . . . . . . . . . . . . . . . . . . . . 196

CHAPTER 8 HARDY CROSS METHOD

8.1 Sign Convention for Flows in Pipes . . . . . . . . . . . . . . . . . . . 202
8.2 Basis for the Hardy Cross Method . . . . . . . . . . . . . . . . . . . . 202
8.3 The Hardy Cross Iterative Solution Technique . . . . . . . . . . 203
8.4 The Correction Term in the Hardy Cross Method . . . . . . . . 203
8.5 Procedure for Solving Water Distribution Systems by the Hardy
Cross Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
8.6 Desirable Features of the Hardy Cross Method . . . . . . . . . . 217
8.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

CHAPTER 9 FORMULATIONS OF THE WATER

DISTRIBUTION SYSTEM EQUATIONS
9.1 The Unknowns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
9.1.1 The unknown flows vector . . . . . . . . . . . . . . . . . . . . . . . 219
9.1.2 The unknown heads vector . . . . . . . . . . . . . . . . . . . . . . . 220
9.2 Network Variable Relationships . . . . . . . . . . . . . . . . . . . . . 222
9.2.1 Theory of graphs and networks . . . . . . . . . . . . . . . . . . . 222
9.2.2 Relationships between variables for water distribution
systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
9.2.3 Treed or determinate networks . . . . . . . . . . . . . . . . . . . . 223
9.3 Topology Matrices for Networks . . . . . . . . . . . . . . . . . . . . . 224
9.3.1 The unknown head node incidence matrix . . . . . . . . . . 224
9.3.2 The fixed head node incidence matrix . . . . . . . . . . . . . 226
9.4 Continuity of Flow at Nodes . . . . . . . . . . . . . . . . . . . . . . . . 227
9.4.1 The continuity equation at a node . . . . . . . . . . . . . . . . . 228
9.4.2 The continuity equations in matrix form for the ten pipe
network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
9.5 Head Loss for a Pipe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
9.6 Various Formulations of the Governing Equations . . . . . . . 230
9.7 Q – equations Formulation for a Network . . . . . . . . . . . . . . 231
9.7.1 The continuity equations . . . . . . . . . . . . . . . . . . . . . . . . 231

vi
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

9.7.2 The energy equations for closed simple loops and required
independent paths for the ten pipe network . . . . . . . . . . 232
9.7.3 The energy equations for the Darcy – Weisbach head loss
equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
9.7.4 The energy equations for the Hazen – Williams head loss
equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
9.7.5 The flow or Q – equations for the ten pipe network . . . . 236
9.8 Nodal Head Equations or the H – equations formulation . . . 237
9.8.1 The H – equations formulation for the ten pipe network 237
9.8.2 The head or H – equations for the ten pipe network system .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
9.8.3 The H – equations formulation for Darcy – Weisbach head
loss equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
9.8.4 The H – equations formulation for Hazen – Williams head
loss equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
9.9 Loop Flows or LF – equations . . . . . . . . . . . . . . . . . . . . . . . 240
9.9.1 The loop flows vector . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
9.9.2 The loop flow equations . . . . . . . . . . . . . . . . . . . . . . . . . 241
9.9.3 The initial flows must satisfy continuity . . . . . . . . . . . . 242
9.9.4 Detailed development of the loop flows equations . . . . 242
9.9.5 The loop flow equations for the ten pipe network . . . . . 243
9.10 The Q+H Equations Formulation . . . . . . . . . . . . . . . . . . . . . 245
9.10.1 The pipe head loss equations for the ten – pipe network 245
9.10.2 The continuity equations for the ten – pipe network . . . 246
9.10.3 The Q+H equations for the ten – pipe network . . . . . . . 246
9.10.4 Q+H equations formulation for Darcy – Weisbach head loss
equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
9.10.5 The Q+H equations formulation for Hazen – Williams head
loss equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
9.11 Global Gradient Algorithm Formulation . . . . . . . . . . . . . . . 248
9.11.1 Background to the Global Gradient Algorithm . . . . . . . 248
9.11.2 Specification of parameters and definition of required
matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
9.11.3 The pipe head loss equations in matrix form – the Global
Gradient Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
9.12 Alternative Formulations of the Network Equations . . . . . . 254
9.13 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255

CHAPTER 10 NUMERICAL TECHNIQUES FOR

ANALYSIS OF WATER
DISTRIBUTION SYSTEMS
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
10.2 Newton Technique for a Set of Nonlinear Equations . . . . . 259
10.2.1 A linear set of equations . . . . . . . . . . . . . . . . . . . . . . . . . 259
10.2.2 A general form of nonlinear equations . . . . . . . . . . . . . . 260

vii
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

10.2.3 Testing for convergence . . . . . . . . . . . . . . . . . . . . . . . . . 262

10.2.4 Assessing the solution accuracy based on the Condition
Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
10.3 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267

CHAPTER 11 COMPUTER ANALYSIS FOR

HYDRAULIC
SIMULATION OF WATER
DISTRIBUTION
SYSTEMS
11.1 The Capability of Simulation Models such as EPANET 2.0 269
11.2 Setting the Environment of EPANET 2.0 (version 2.00.12) 270
11.2.1 Choosing default units and pipe head loss equation . . . 271
11.2.2 Choosing default properties . . . . . . . . . . . . . . . . . . . . . . 271
11.2.3 Choosing the number of decimal places for output . . . . 271
11.2.4 Specifying map options . . . . . . . . . . . . . . . . . . . . . . . . . 271
11.3 Modeling Flows in Individual Pipes . . . . . . . . . . . . . . . . . . 271
11.3.1 Friction head loss formula . . . . . . . . . . . . . . . . . . . . . . . 271
11.3.2 EPANET input for junctions in water distribution systems .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
11.3.3 Storage reservoirs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
11.3.4 EPANET input data for pipes — a smooth pipe example 274
11.3.5 A descriptive title for the network to be analyzed by
EPANET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
11.3.6 Options for the EPANET run . . . . . . . . . . . . . . . . . . . . . 274
11.3.7 The end of an EPANET input data file . . . . . . . . . . . . . 275
11.3.8 Computer solution of Example 11.1 using EPANET 2.0 275
11.3.9 EPANET 2.0 output results for Example 11.1 . . . . . . . . 276
11.4 Modeling a 6 – Pipe Water Distribution System . . . . . . . . . 276
11.4.1 The network and input data file . . . . . . . . . . . . . . . . . . . 276
11.4.2 Results for 6 – pipe water distribution system . . . . . . . . 278
11.5 EPANET Modeling of a 10 – Pipe Water Distribution System .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
11.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284

CHAPTER 12 SOLVING THE Q – EQUATIONS

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
12.2 Solving the Q – equations for a Water Distribution System . 285
12.2.1 The set of Q – equations . . . . . . . . . . . . . . . . . . . . . . . . . 285
12.2.2 The Jacobian matrix for the unknown flow equations . . 286
12.2.3 Derivatives of terms incorporating a modulus of flow . 288
12.2.4 Jacobian values and plots of nonlinear flow equation terms
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

viii
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

12.3 Convergence Stopping Criteria for the Q – equations

Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
12.4 The Q – equations for the Ten Pipe Network . . . . . . . . . . . . 293
12.4.1 Unknowns and initial guesses for the Newton solution . 293
12.4.2 Computation of the pipe resistance factors . . . . . . . . . . 293
12.4.3 Jacobian matrix for the flow equations for the ten pipe
network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
12.4.4 Solving for the flow corrections for the ten pipe network . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
12.4.5 Solving the flow equations for the ten pipe network . . . 298
12.5 A Two Pipe Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
12.5.1 The network and the formulations . . . . . . . . . . . . . . . . . 298
12.5.2 The EPANET results . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
12.6 Solving the Q – equations for the Two Pipe Network . . . . . 300
12.6.1 The unknowns and initial guesses . . . . . . . . . . . . . . . . . 300
12.6.2 The convergence stopping test for the Q – equations (two
pipe example) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
12.6.3 Two pipe example based on Newton method (Q – equations
formulation) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
12.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308

CHAPTER 13 SOLVING THE H – EQUATIONS

13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
13.2 Solving the H – equations for a Water Distribution System . 309
13.2.1 The set of H – equations . . . . . . . . . . . . . . . . . . . . . . . . . 309
13.2.2 The Jacobian matrix for the H – equations formulation . 310
13.2.3 Derivatives of terms incorporating a modulus of heads . 312
13.2.4 Jacobian values and plots of nonlinear head equation terms
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
13.3 Solving the H – equations for the Ten Pipe Network . . . . . . 321
13.3.1 Unknowns and functions for the Newton method . . . . . 321
13.3.2 Jacobian matrix for the head equations for the ten pipe
network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
13.3.3 Solving for the head corrections for the ten pipe network . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
13.4 The Two Pipe Network Solving for Unknown Heads . . . . . 326
13.4.1 Unknowns and initial guesses for the Newton solution . 326
13.4.2 Two pipe example based on Newton method (H – equations
formulation) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
13.5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

CHAPTER 14 SOLVING THE LOOP FLOW

EQUATIONS
14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
14.2 The Loop Flow (LF) Equations for a Water Distribution System
ix
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
14.2.1 Set of loop flow equations . . . . . . . . . . . . . . . . . . . . . . . 334
14.2.2 The Jacobian matrix for the LF – equations formulation 335
14.2.3 Derivatives of terms incorporating a modulus term . . . . 338
14.2.4 Updating the Darcy – Weisbach friction factors . . . . . . 338
14.3 Convergence Stopping Criteria for the LF – equations
Formulation 338
14.4 Solving the LF – equations for the Ten Pipe Network . . . . . 339
14.4.1 Unknowns and functions for the Newton method . . . . . 339
14.4.2 Solving for the LF – corrections for the ten pipe network . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
14.5 Solving the LF – Equations (Loop Flow) for Two Pipe Network
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
14.5.1 The unknowns and initial guesses . . . . . . . . . . . . . . . . . 343
14.5.2 Two pipe example based on Newton method (LF – equations
formulation) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
14.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350

CHAPTER 15 SOLVING THE Q+H

FORMULATION
15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
15.2 Solving the Q+H equations for a Water Distribution System . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
15.2.1 The set of Q+H equations . . . . . . . . . . . . . . . . . . . . . . . 351
15.2.2 The Jacobian matrix for the unknown flow and head
equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
15.3 The Q+H equations Formulation for the Ten Pipe Network 356
15.3.1 Unknowns and functions for the Newton solution of Q+H
equations 356
15.3.2 Jacobian matrix for the Q+H equations for the ten pipe
network 357
15.3.3 Solving for the flow corrections for the ten pipe network . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
15.4 Solving the Q+H equations for the Two Pipe Network . . . . 360
15.4.1 The unknowns and initial guesses . . . . . . . . . . . . . . . . . 360
15.4.2 Two pipe example based on Newton method (Q+H equations
formulation) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
15.5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368

CHAPTER 16 SOLVING THE GLOBAL

GRADIENT
ALGORITHM FORMULATION
16.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369

x
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

16.2 Numerical Solution for the Global Gradient Algorithm . . . 369

16.2.1 Redefinition of quantities in the Global Gradient Algorithm
formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
16.2.2 The Jacobian matrix for Global Gradient Algorithm
formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
16.2.3 Calculation of the Jacobian matrix elements for the Global
Gradient Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
16.2.4 Distinguishing between laminar and turbulent flow . . . 377
16.2.5 Definition of the Schur Complement . . . . . . . . . . . . . . . 379
16.2.6 The Global Gradient Algorithm two – step update vectors .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
16.3 Solving the Global Gradient Algorithm Equations for the Two
Pipe Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
16.3.1 Unknowns and initial guesses . . . . . . . . . . . . . . . . . . . . 384
16.3.2 Example based on the Global Gradient Algorithm . . . . 386
16.4 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394

CHAPTER 17 DATA REQUIREMENTS FOR

WATER
DISTRIBUTION SYSTEM
ANALYSIS
17.1 Basic Data Requirements for Computer Analysis of Networks .
396
17.2 Overview of data requirements . . . . . . . . . . . . . . . . . . . . . . 396
17.2.1 Sources of supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
17.2.2 The existing system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397
17.2.3 Demands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398
17.2.4 Design criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
17.2.5 Cost of options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
17.3 Static versus extended period simulation . . . . . . . . . . . . . . . 400
17.4 Demands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
17.5 Design Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
17.5.1 Design periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
17.5.2 Levels of service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
17.6 Controlling demand growth . . . . . . . . . . . . . . . . . . . . . . . . . 403
17.7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403

CHAPTER 18 PUMPS
18.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
18.2 Useful Definitions for Pumps (alphabetized) . . . . . . . . . . . . 404
18.3 Main Features of a Pump . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
18.3.1 Types of pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
18.4 Energy Equation Applied to a Pumping System . . . . . . . . . 405

xi
Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

18.5 Rotodynamic or Impeller Pumps . . . . . . . . . . . . . . . . . . . . . 407

18.5.1 Classification of impeller pumps . . . . . . . . . . . . . . . . . . 407
18.5.2 Centrifugal pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
18.5.3 Mixed flow pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
18.5.4 Axial flow pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
18.5.5 Pump size specification . . . . . . . . . . . . . . . . . . . . . . . . . 409
18.6 Pump Performance Characteristics . . . . . . . . . . . . . . . . . . . 409
18.6.1 Pump curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409
18.6.2 Definition of water or hydraulic power . . . . . . . . . . . . . 410
18.6.3 Pump efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
18.6.4 Pump shaft power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
18.6.5 Pump energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
18.6.6 Electrical energy input . . . . . . . . . . . . . . . . . . . . . . . . . . 414
18.6.7 Cost of pump operation . . . . . . . . . . . . . . . . . . . . . . . . . 414
18.6.8 Pump speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
18.6.9 Power – torque – speed relationship. . . . . . . . . . . . . . . . . 416
18.7 Pump System Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
18.8 Pump Operating Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418
18.9 Pump Specific Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420
18.10 Types of Pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
18.10.1 Centrifugal pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
18.10.2 Mixed flow pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
18.10.3 Axial flow pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
18.11 Colt Industries Specific Speed Categorization for Pumps . . 426
18.12 Pump Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
18.12.1 Momentum and energy changes in a pump . . . . . . . . . . 426
18.12.2 Requirements for similarity of pumps . . . . . . . . . . . . . . 426
18.12.3 Similarity of velocity vector diagrams for pumps . . . . . 427
18.12.4 Similar velocity diagrams . . . . . . . . . . . . . . . . . . . . . . . 427
18.12.5 Pump flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
18.12.6 Pump speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
18.13 Dimensionless Numbers for Pumps . . . . . . . . . . . . . . . . . . . 429
18.13.1 Dimensionless pump flow . . . . . . . . . . . . . . . . . . . . . . . 429
18.13.2 Dimensionless pump speed . . . . . . . . . . . . . . . . . . . . . . 430
18.13.3 Dimensionless pump head . . . . . . . . . . . . . . . . . . . . . . . 431
18.13.4 Dimensionless pump power . . . . . . . . . . . . . . . . . . . . . . 432
18.13.5 Dimensionless pump torque . . . . . . . . . . . . . . . . . . . . . . 433
18.13.6 A summary of the dimensionless pump parameters . . . 435
18.14 Dimensionless Pump Curves . . . . . . . . . . . . . . . . . . . . . . . . 435
18.14.1 Dimensionless Head – Flow Curve . . . . . . . . . . . . . . . . 435
18.14.2 Dimensionless efficiency curve  versus CQ . . . . . . . . 437
18.14.3 Power – flow curve in dimensionless form . . . . . . . . . . 438
18.15 Pump Affinity Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 438
18.15.1 Pump affinity law number 1 — a change in pump speed 439
18.15.2 Pump affinity law number 2 – a change in pump impeller
diameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440

xii
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

18.16 Multi – Impeller Machines . . . . . . . . . . . . . . . . . . . . . . . . . . 442

18.16.1 A multi – stage pump . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
18.16.2 A multi – flow pump . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
18.17 Effect of Reynolds Number on Pump Behavior . . . . . . . . . 443
18.17.1 Low Reynolds number flows . . . . . . . . . . . . . . . . . . . . . 443
18.17.2 High Reynolds number flows . . . . . . . . . . . . . . . . . . . . 443
18.18 Pumps in Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
18.19 Pumps in Parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 446
18.20 Pump Cavitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448
18.20.1 Net positive suction head required (NPSHR) . . . . . . . . 448
18.20.2 Net positive suction head available (NPSHA) . . . . . . . . 449
18.20.3 Preventing cavitation in a pump . . . . . . . . . . . . . . . . . . . 450
18.21 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452

CHAPTER 19 VALVES
19.1 Isolating Valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
19.2 Pressure Reducing Valves (PRV) . . . . . . . . . . . . . . . . . . . . 454
19.2.1 Operating modes of pressure reducing valves (PRVs) . 455
19.2.2 Elements of pressure reducing valves . . . . . . . . . . . . . . 455
19.2.3 Modeling of PRVs in computer hydraulic analysis software
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
19.2.4 Advantages of the use of PRVs . . . . . . . . . . . . . . . . . . . 466
19.2.5 EPANET approach to modeling of PRVs . . . . . . . . . . . 466
19.3 Flow Control Valves (FCV) . . . . . . . . . . . . . . . . . . . . . . . . . 467
19.4 Control Valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 471
19.4.1 Cavitation in control valves . . . . . . . . . . . . . . . . . . . . . . 471
19.5 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472

CHAPTER 20 OPTIMIZATION OF WATER

DISTRIBUTION SYSTEMS
20.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
20.2 An Overview of the Genetic Algorithm Process . . . . . . . . . 475
20.3 Simulation of Water Distribution Systems . . . . . . . . . . . . . 476
20.4 Trial – and – Error Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
20.5 Optimization of the Design of Water Distribution Systems 477
20.5.1 Previous related genetic algorithm applications . . . . . . 478
20.5.2 The New York City water tunnels network . . . . . . . . . . 478
20.5.3 Combinatorially large design networks . . . . . . . . . . . . . 481
20.6 The Genetic Algorithm Optimization Process . . . . . . . . . . . 481
20.6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
20.7 Genetic Algorithm Optimization . . . . . . . . . . . . . . . . . . . . . 482
20.7.1 What is it? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
20.7.2 Formulation of the optimization problem . . . . . . . . . . . 483

xiii
Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

20.7.3 Steps in using genetic algorithms for water distribution

system optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
20.8 Coding Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
20.9 Selecting the Population Size . . . . . . . . . . . . . . . . . . . . . . . . 486
20.10 The Randomly Generated Initial Population . . . . . . . . . . . . 486
20.11 The Fitness Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486
20.12 An Overview of Genetic Algorithm Operators . . . . . . . . . . 488
20.13 The Selection Operator or the Reproduction Operator . . . . 488
20.13.1 Proportionate or weighted roulette wheel selection . . . . 489
20.13.2 Tournament selection . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
20.14 Crossover Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490
20.14.1 One – point and two – point crossover . . . . . . . . . . . . . . 490
20.14.2 Uniform crossover . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
20.15 Mutation Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
20.15.1 Bit – wise mutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
20.15.2 Adjacency mutation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
20.16 Successive Generations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
20.17 Application of Genetic Algorithm Optimization to Water
Distribution System Design . . . . . . . . . . . . . . . . . . . . . . . . . 493
20.17.1 Layout of a new system . . . . . . . . . . . . . . . . . . . . . . . . . 494
20.17.2 Sizing of components in a new system . . . . . . . . . . . . . 494
20.17.3 Expansion of an existing system . . . . . . . . . . . . . . . . . . 494
20.17.4 Rehabilitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494
20.17.5 Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495
20.17.6 Tender Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495
20.18 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495
20.18.1 Optimization of a 14 – pipe network . . . . . . . . . . . . . . . 495
20.18.2 Other case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
20.19 Altering the Current Role of the Designer . . . . . . . . . . . . . . 503
20.20 Advantages of Genetic Algorithm Optimization . . . . . . . . . 503
20.20.1 Solutions are based on commercially available pipe sizes . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
20.20.2 A range of solutions result . . . . . . . . . . . . . . . . . . . . . . . 504
20.20.3 Full utilization of simulation modeling capability . . . . . 504
20.20.4 GAs deal with a population of solutions . . . . . . . . . . . . 504
20.20.5 GAs only require fitness function information . . . . . . . 504
20.21 Other Varieties of Genetic Algorithm Optimization
Formulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504
20.22 Other Optimization Methods . . . . . . . . . . . . . . . . . . . . . . . . 504
20.22.1 Complete enumeration . . . . . . . . . . . . . . . . . . . . . . . . . . 505
20.22.2 Partial or pruned enumeration of the search space . . . . . 505
20.22.3 Linear programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
20.22.4 Nonlinear programming . . . . . . . . . . . . . . . . . . . . . . . . . 506
20.22.5 Heuristic methods for optimization . . . . . . . . . . . . . . . . 508
20.22.6 Why hasn’t optimization been used in the past? . . . . . . 508
xiv
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

20.23 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508

CHAPTER 21 WATER HAMMER ANALYSIS

21.1 Concepts of Unsteady Flow . . . . . . . . . . . . . . . . . . . . . . . . . 510
21.2 Terms for Water Hammer . . . . . . . . . . . . . . . . . . . . . . . . . . 511
21.3 Transmission of Water Hammer Waves at a Finite Speed . 511
21.4 Fluid Compressibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
21.5 Wave Speed of a Simple Elastic Wave in an Infinite
Compressible Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513
21.5.1 Example of wave speed computation in an infinite fluid 514
21.6 The Joukowsky Head Change — Basic Water Hammer Equation
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515
21.6.1 Time of travel for a pipeline . . . . . . . . . . . . . . . . . . . . . . 515
21.6.2 Rapid valve closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
21.6.3 Slow valve closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
21.6.4 The Joukowsky head change at the upstream side of a valve
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
21.6.5 The Joukowsky head change at the downstream side of a
valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
21.7 Sequence of Events After Sudden Closure of a Valve . . . . . 518
21.7.1 Steady state flow with an initial steady state velocity of Vo
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
21.7.2 Immediately following the valve closure . . . . . . . . . . . . 518
21.7.3 The first L/a seconds, 0 < t < L/a . . . . . . . . . . . . . . . . . . 519
21.7.4 At L/a seconds the positive water hammer wave reaches the
reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 519
21.7.5 The second L/a seconds, L/a < t < 2L/a . . . . . . . . . . . . . 520
21.7.6 At 2 L/a seconds the reflected negative wave reaches the
valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
21.7.7 The third L/a seconds, 2L/a < t < 3L/a . . . . . . . . . . . . . . 522
21.7.8 At 3L/a seconds the negative wave reaches the reservoir . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
21.7.9 The fourth L/a seconds, 3L/a < t < 4L/a . . . . . . . . . . . . 523
21.7.10 At 4L/a seconds the positive reflected wave reaches the valve
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524
21.8 The Joukowsky Head Rise Equation from the Unsteady
Momentum Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
21.8.1 For a partial valve closure where the velocity change is not
DV = – Vo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
21.9 Wave Speed for the Fluid in a Pipeline . . . . . . . . . . . . . . . . 528
21.9.1 Wave speed for a thin – walled pipe . . . . . . . . . . . . . . . . 530
21.9.2 Wave speed for a cement mortar lined pipe . . . . . . . . . . 531
21.9.3 Wave speed for a thick – walled pipe . . . . . . . . . . . . . . . 533
21.9.4 Wave speed for a completely rigid pipe . . . . . . . . . . . . . 534
21.9.5 Wave speed for a highly flexible pipe . . . . . . . . . . . . . . 535
21.10 Hoop Stress and Strain in a Thin Walled Pipe . . . . . . . . . . . 535
xv
 Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

21.11 Design Pressure Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536

21.12 Unsteady Continuity Equation for Pipe Flow . . . . . . . . . . . 537
21.12.1 Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
21.12.2 The continuity equation . . . . . . . . . . . . . . . . . . . . . . . . . 537
21.13 Unsteady Momentum Equation for Pipe Flow . . . . . . . . . . . 538
21.14 The Method of Characteristics Transformation . . . . . . . . . . 540
21.15 Integration of the MOC Compatibility Equations . . . . . . . . 541
21.15.1 The characteristic grid on the x – t plane . . . . . . . . . . . . 541
21.15.2 Integrating the C+ compatibility equation . . . . . . . . . . . 542
21.15.3 Integrating the C – compatibility equation . . . . . . . . . . . 545
21.15.4 Computational examples . . . . . . . . . . . . . . . . . . . . . . . . 547
21.15.5 An MOC example calculating conditions at an interior
computational
section at L/2a seconds (Point P2 on the x-t plane) . . . . 548
21.16 The Closed Valve or Dead End Boundary Condition . . . . . 552
21.16.1 The governing equations . . . . . . . . . . . . . . . . . . . . . . . . 552
21.16.2 An MOC example for the head rise due to instantaneous
valve closure at zero seconds (Point P1 on the x-t plane)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 554
21.17 The Reservoir Boundary Condition . . . . . . . . . . . . . . . . . . . 556
21.17.1 The governing equations . . . . . . . . . . . . . . . . . . . . . . . . 556
21.17.2 An MOC example for steady state conditions at the reservoir
at L/2a seconds (Point P3 on the x-t plane) . . . . . . . . . . 558
21.18 Further MOC Computations . . . . . . . . . . . . . . . . . . . . . . . . 561
21.18.1 An MOC example for conditions at the closed valve at L/2a
seconds (Point P4 on the x-t plane) . . . . . . . . . . . . . . . . 561
21.18.2 A MOC example of flow reversal at the reservoir at L/a
seconds (Point P5 on the x-t plane) . . . . . . . . . . . . . . . . 563
21.18.3 An MOC example for conditions at the closed valve at L/a
seconds (Point P6 on the x-t plane) . . . . . . . . . . . . . . . . 566
21.19 Further Examples of Applying the Method of Characteristics .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
21.19.1 An MOC example for conditions at the mid – point of the
pipeline at 3L/2a seconds (Point P7 on the x-t plane) . . 568
21.19.2 An MOC example of inversion of head at the closed valve at
2L/a seconds (Point P8 on the x-t plane) . . . . . . . . . . . . 571
21.19.3 The MOC calculation sequence in a computer program 574
21.20 Water hammer control methods . . . . . . . . . . . . . . . . . . . . . . 574
21.20.1 Common water hammer control measures . . . . . . . . . . . 574
21.20.2 Overview of surge tanks . . . . . . . . . . . . . . . . . . . . . . . . . 576
21.20.3 Open surge tanks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 576
21.21 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577

CHAPTER 22. HYDROPOWER AND TURBINES

22.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 578

xvi
Table of Contents - File: WDSv3TOC.fm July 6, 2022 Version 1

22.2 The Snowy Hydro Scheme . . . . . . . . . . . . . . . . . . . . . . . . . 579

22.3 Useful Definitions for Turbines (alphabetized) . . . . . . . . . 579
22.4 Components of Hydropower Generating Plants . . . . . . . . . 580
22.5 Classification of Hydropower Generating Plants by Size . . 581
22.6 Turbine Specific Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581
22.7 Turbine Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 582
22.7.1 Pelton wheel turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
22.7.2 Turgo turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
22.7.3 Francis turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584
22.7.4 Kaplan turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585
22.7.5 Ranges of parameters for turbines . . . . . . . . . . . . . . . . . 587
22.7.6 Turbine rotational speeds . . . . . . . . . . . . . . . . . . . . . . . 589
22.8 Energy Equation Applied to a Hydropower Turbine System 590
22.9 Turbine Performance Characteristics . . . . . . . . . . . . . . . . . . 592
22.9.1 Turbine curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 592
22.9.2 Definition of water or hydraulic power for a turbine . . . 592
22.9.3 Turbine efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593
22.9.4 Turbine shaft power output . . . . . . . . . . . . . . . . . . . . . . 593
22.9.5 Turbine energy output . . . . . . . . . . . . . . . . . . . . . . . . . . 594
22.9.6 Electrical energy output . . . . . . . . . . . . . . . . . . . . . . . . . 594
22.9.7 Monetary benefits from turbine operation . . . . . . . . . . . 595
22.10 Pumped hydroelectricity (or pumped hydro) . . . . . . . . . . . . 596
22.10.1 The Snowy 2.0 pumped hydro project . . . . . . . . . . . . . . 596
22.11 Water hammer control for hydropower systems . . . . . . . . . 596
22.12 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
A.1 Density, Specific Weight, Viscosity, Bulk Modulus of Elasticity,
Vapor Pressure and Surface Tension . . . . . . . . . . . . . . . . . . 599
A.2 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 602
B.1 Units of derived quantities and conversions . . . . . . . . . . . . 603
C.1 Roughness Height Values for Darcy – Weisbach Calculations .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
C.2 Hazen – Williams C Values . . . . . . . . . . . . . . . . . . . . . . . . . 609
D.1 MOODY DIAGRAM FOR COMPUTER SIMULATION OF WATER DISTRIBU-
TION SYSTEMS613

xvii
 File: Finding-Errors-6jul2022.fm July 6, 2022 Version 1

1. Free Lunch for Finding Errors

Editing and finding errors in information that you read is an important skill for your engineering career.

I (Professor Emeritus Angus Simpson) would encourage you to read the textbook carefully and criti-
cally.

If you find errors - please let Prof. Simpson know by emailining hin on

[email protected]

and then changes can be made for the next year so that the material can be made as error-free as possi-
ble.

The three people who identify the most errors will be eligible for the free lunch. In the event of a tie, the
three winners will be drawn out of a hat.

The information to be included in the email is:

1. A heading of the subject name and the words “WDS book error found on page xx”.
2. A description of the exact location of error.
3. An adequate description of the error.

ii
 Chapter 1 - File: Chap1.fm July 6, 2022 Version 7

CHAPTER 1 INTRODUCTION1

1.1 Scope
Municipal water distribution systems or networks are the focus of this book. Elements of a network
include the collection headworks, water treatment plants (WTPs) or purification works, transmission
works and distribution works. Computer modeling for the simulation of the water networks is one of the
main aspects considered in this book. Many of the same principles also apply to piped transmission sys-
tems and piped irrigation systems. Transmission systems transfer water within the headworks system or
between tanks in the distribution system. The interaction between the transmission system, storage tanks
and the network often needs to be carefully considered. For example, the transmission system usually
delivers water from the headworks system of storages and dams to tanks or service reservoirs. Thus the
sizing of pipelines in the transmission system is a critical factor when evaluating the refilling of the
tanks after a peak day demand or a critical demand period. Tanks must be refilled in preparation, in case,
another high demand period occurs. For the network downstream of the tanks that supplies water to
industry, homes and parks, the sizing of these pipes usually depends on expected conditions during the
peak hour of water demand or during a fire fighting occurrence.

Users of water are varied. In a city, water use is provided for households (for personal use, cooking,
cleaning and garden watering), industrial use, commercial use, government and university building use,
parkland and oval watering and fire fighting. Estimation of demands for water is a critical part of com-
puter modeling of water distribution systems. The accurate estimation of demands is a specialized topic
in itself. A comprehensive analysis of demand estimation including techniques for estimating population
growth are beyond the scope of this book. For the computer analysis of a particular network, the issue of
various types of demand loading cases such as peak hour, fire loadings, and peak day demands are con-
sidered in this book.

Computer modeling or simulation of networks has reached a mature level of practice. There are many
excellent commercial and government computer programs available. The input and output capabilities
of such programs are highly sophisticated. Many water utilities now have geographical information sys-
tems (GIS) that keep a large database of information about their network. Examples of items stored in
the database include

• pipe diameters (both nominal and actual), material and condition of pipes
• coordinates of nodes at the ends of pipes
• elevations of nodes
• valve locations and types
• pump locations and types
• tanks
Many software vendors are now developing links between their hydraulic solver software programs and
various GIS platforms.

1. © Copyright 2022Jul. Prof. Emeritus Angus Ross Simpson, University of Adelaide, Adelaide,
Australia. Part of the book entitled Water Distribution Systems Engineering.

1
 Chapter 1 - File: Chap1.fm July 6, 2022 Version 7

The underlying governing equations of pipe flow in water distribution systems are well understood.
Numerous solution techniques have been proposed since the first trial and error techniques of Hardy
Cross at the University of Illinois in 1936 (Cross 1936). Various formulations of the flow and head equa-
tions are possible for a network. These are presented in detail in Chapter 9. Numerical solutions of the
governing equations are discussed in Chapter 10. Other components that may be modeled include
pumps (see Chapter 18) and valves (see Chapter 19).

The first 16 Chapters of this book are concerned with steady state analysis of water distribution systems
including extended period simulation (EPS). Water is assumed to be incompressible for the steady state
analysis of networks. Chapter 20 focuses on optimization of water distribution systems using genetic
algorithms. Chapter 21 introduces water hammer analysis in pipeline systems where the compressibility
of water is taken into account.

1.2 Overview of Water Collection and Distribution

The process of obtaining water for human use in cities and towns begins in catchments that harvest rain-
fall. Water collection facilities take on various forms. For example, water may be stored in a reservoir
created by damming a river or alternatively water may be pumped directly from a river to a WTP. The
collection headworks includes

• catchments
• rivers
• streams
• creeks
• reservoirs
• dams
The main focus of this book is the water delivery systems downstream of these collection facilities.

1.3 Design Objectives

Water distribution system infrastructure has deteriorated significantly in many industrialized and third
world countries around the world and there is a great need for renovation and rehabilitation of these sys-
tems. The cost – effective design of renovations and extensions of existing networks poses a problem for
water system managers. The issue of water distribution systems has been receiving increasing attention.

There are a number of objectives that a water utility or authority has in providing water to the commu-
nity. These include provision of

• an adequate quantity of water on – demand for a variety of uses

• water at a pressure equal to or exceeding a minimum allowable pressure
• an adequate water service at the lowest possible cost
Some examples of supply standards and design requirements that specify these design objectives are
given below. Clearly, these depend on location in the world and the climate.

1.3.1 Planning of water distribution systems

The expected life of water distribution systems often exceeds 100 years. It is impossible to imagine all
the changes that will occur during the life of the system. Planning must allow for changes to occur.

Often a water supply system is designed to supply adequate quantities of water on the peak hour of the
year. Thus for regions of hot dry climates, the maximum capacity of a network is only used to its full
potential for less than six weeks a year. For the remainder of the year, the water supply system has con-
siderable excess capacity.

Typical design situations for a network include the design of

• a new network for an existing town or district

• a new network for a new subdivision based on the number of blocks or land area to be
developed

2
 Chapter 1 - File: Chap1.fm July 6, 2022 Version 7

• major headworks for upgrading existing water supply systems (these include pumping sta-
tions, major transmission pipelines and balancing storages)
• revisions needed to upgrade existing substandard networks
Often, the annual water consumption is estimated or based upon consumer meter data. Population pro-
jections and land use proposals are used to predict future water demands. Climatic conditions and annual
average rainfall affect these water demand estimates.

1.3.2 Supply standards

Target minimum supply standards are usually adopted for metropolitan areas. Typical values that permit
the use of garden sprinklers at the same time as washing facilities in the house during periods of peak
flow include

• minimum flow rate of 0.45 L/s per dwelling

• minimum pressure in large mains of 20 m head of water
• minimum operating pressure at services of 17 m head of water
The maximum pressure should not exceed around 90 m of head of water. Otherwise special provision is
required to protect household plumbing, dishwashers and washing machines.

The recommended minimum size for distribution mains is as follows

• residential areas: 100 mm diameter

• industrial and commercial centers: 150 mm diameter
1.3.3 Design requirements
Urban networks tend to have a large number of branches, with pipes often interconnected every 500 m
(Brebbia and Ferrante 1983). The network must be designed such that water flows in all pipes without it
becoming stagnant in any one pipe. Looped networks are common in water supply networks so that sup-
ply is not interrupted in the event of repairs or failure in any pipe.

There are three main criteria in the design of networks

• the capability to supply demands (demand sources are residential, industrial or urban park-
land watering)
• the ability to provide a minimum allowable pressure under peak demand conditions (low
pressures cause problems)
• to not exceed maximum pressure constraints (a typical desirable pressure range is 20 to
90 m of head of water)
1.4 Computer Analysis of Water Distribution Systems
Computer hydraulic analysis of networks during the 1990s reached a mature level of practice. Many
sophisticated computer packages are available. The computing power available to the water distribution
design engineer has improved markedly over the last 15 years and will continue to do so. Computer
analysis of looped networks was pioneered in part by Shamir and Howard (1968) and Epp and Fowler
(1970). The research findings presented by these authors represent the beginning of a move from hand
methods (most commonly the Hardy Cross method) to the use of detailed computer analysis including
extended period simulation analysis of networks.

Selecting pipe and component sizes for elements within a network remains to a large extent a process
based on rules of thumb and trial and error based on engineering experience (Walski et al. 1988). The
engineer must input all of the sizes of components into the computer simulation model, run the program,
evaluate the results, make any necessary adjustments to sizes, and then repeat the process.

Many hydraulic simulation computer programs are available. Most packages analyze a user pre –
defined system (i.e. diameters and pump sizes need to be specified). In general, the engineer uses the
computer simulation model to determine the following by trial and error

• pipe diameters of new pipes

3
 Chapter 1 - File: Chap1.fm July 6, 2022 Version 7

• pipe diameters of pipes to be constructed in parallel to existing pipes

• pump sizes and locations (both booster pumps and source pumps)
• storage sizes, location and normal water surface operating levels
• pressure reducing valve locations and pressure settings
Experienced engineers achieve a reasonably low cost design, although, it has been shown that optimiza-
tion usually improves the design by a reduction in cost of between 15 and 25% (see Chapter 21 on
genetic algorithm optimization).

1.5 Steady State Analysis versus Extended Period Simulation

Two types of steady state computer analysis of water distribution systems are commonly performed. The
first is steady state analysis or static analysis for a particular point in time. The demand pattern selected
is commonly the peak hourly demand (or instantaneous demand) for an extremely high peak demand
day. Examples of peak events include a 1 in 7 year annual recurrence interval (ARI) event or the highest
peak demand day ever recorded adjusted for population growth. The demand used depends on the
design criteria used by a particular water utility or authority. Neither the variation in nodal demands nor
the reservoir water levels remain constant over a period of time.

The second form of analysis is extended period simulation or dynamic analysis (this should not be con-
fused with water hammer analysis) of networks. The performance of storages in balancing supply and
distribution is checked by using an extended period simulation. Extended period simulation ensures an
adequate level of service to consumers under varying conditions of demands and reservoir water levels
(Bhave 1988). Adequate flow rates and pressures (residual heads) must be maintained at all times. Anal-
ysis of the network is usually carried out over a period of 24 to 48 hours under varying nodal demands
and reservoir water levels (Bhave 1988).

1.6 Defining a Water Distribution System

To illustrate a typical gravity feed network, a ten pipe network as shown in Figure 1.1 and Figure 1.2 is
considered. Node numbers, pipe numbers and nodal demands (DMi) are shown. The network in
Figure 1.1 has two sources of supply and 10 pipes. Details of the characteristics of the network and
demand patterns are given in Table 1.1, Table 1.2 and Table 1.3. There are three demand loading pat-
terns: one peak hour and two fire loading cases (at nodes 6 and 9 respectively).

Table 1.1 Elevations for the ten pipe network

Node Elevation
(m)
1 Reservoir WSEL* 365
2 320
3 326
4 332
5 Tank WSEL 371
6 298
7 295
8 292
9 289

*WSEL = water surface elevation

4
Chapter 1 - File: Chap1.fm July 6, 2022 Version 7

Figure 1.1 A ten pipe network – pipe and node numbers

Figure 1.2 A ten pipe network – demands

5
 Chapter 1 - File: Chap1.fm July 6, 2022 Version 7

Table 1.2 Pipe data for the ten pipe network

Pipe Diameter Length Roughness height (  ) for
(mm) (m) Darcy – Weisbach friction
factor (mm*)
1 361 4800 0.25
2 254 1600 0.25
3 254 1600 0.25
4 308 6400 0.25
5 308 1600 0.10
6 254 1600 0.10
7 203 1600 0.10
8 254 1600 0.05
9 254 1600 0.25
10 102 1600 0.10

* Also referred to as a resistance factor

1.7 Trial and Error Design of Water Distribution Systems

Selecting the lowest cost combination of components (pipes, pumps, tanks and valves) in a network is
difficult. In a looped system, the diameter of every pipe affects the flow rate in every other pipe (Walski
et al. 1988). It does not take many decision variables, with comparatively few choices for each one, to
result in a search space size that is combinatorially very, very large. Thus, it is difficult for even the most
experienced network design engineer to be able to guarantee that they have found the lowest possible
cost design.

Table 1.3 Demands and minimum allowable pressures for nodes in the ten pipe network
Loading case no. 1 Loading case no. 2 Loading case no. 3
Node Demand Minimum Demand Minimum Demand Minimum
no. (L/s) allowable (L/s) allowable (L/s) allowable
pressure pressure pressure
(m) (m) (m)
2 13 28.0 13 14.0 13 14.0
3 13 18.0 13 14.0 13 14.0
4 0 18.0 0 14.0 0 14.0
6 19 35.0 82 14.0 19 14.0
7 19 35.0 9 10.5 19 14.0
8 19 35.0 19 14.0 19 14.0
9 19 35.0 13 14.0 51 14.0

The following rules of thumb have traditionally been used for designing networks (Walski et al. 1988)

• velocities of about 0.6 m/s (2.0 ft/sec) at average flow

• velocities < 2.4 m/s (8 ft/sec) at peak flow
• pipe diameters at least 150 mm (6 in.) for systems with fire protection
• no dead – end mains

6
 Chapter 1 - File: Chap1.fm July 6, 2022 Version 7

• enough pumping capacity to provide peak day flow with one pump out of service
• pressures roughly 410 to 550 kPa (42 to 56 m or 60 to 80 psi) under normal conditions
• pressure at least 140 kPa (14 m or 20 psi) under emergency conditions
These rules of thumb are guidelines only. In some systems, the actual velocities may exceed the values
given above. These values should not be interpreted as being values that should never be exceeded but
rather values that are aimed for.

In general, the engineer must use trial and error with computer simulation analysis to arrive at a combi-
nation of pipes and elements such that all the specified design criteria (e.g. minimum allowable pressure
at nodes, maximum velocities in pipelines) are met for the lowest possible cost.

1.8 Optimization of the Design of Water Distribution Systems

Optimization of network design has been the focus of research from the 1970s to the present, although
these techniques are not commonly used in industry at the current time. The question is posed as to why
there are no user – friendly programs for the optimal design of networks. Schaake and Lai (1969) were
the first to develop a program capable of solving for the optimum combination of pipe diameter to mini-
mize the cost in the New York City water tunnels problem. About 100 papers — up to 1988 (Walski
1988) — have been published on the topic of optimization of water distribution systems.

Traditional optimization methods that have been applied to network optimization in the past include

• linear programming
• nonlinear programming (GINO, MINOS computer packages)
• enumeration of a severely pruned search space (WADISO)
• dynamic programming
Since 1990, a new optimization technique for the design of networks based on genetic algorithm optimi-
zation has been developed (Murphy and Simpson 1992, Simpson et al. 1994, Dandy et al. 1996).
Genetic algorithm optimization utilizes all of the power available in hydraulic simulation models and
offers the promise of significant capital and cost savings. The use of optimization will probably change
over the next 20 years as the pressure to squeeze more effective use from scarce resources available for
water supply infrastructure increases. Optimization of even relatively simple systems often lead to dra-
matic savings. Chapter 20 discusses genetic algorithm optimization in detail.

1.9 References
Bhave, P. R. (1988). “Extended period simulation of water systems - direct solution.” Journal of Envi-
ronmental Engineering, 114 (5), October, 1146-1159.

Brebbia, C. A. and A. J. Ferrante (1983). Computational Hydraulics, Pipe Networks, Chapter 4. Compu-
tational hydraulics: 59-121.

Cross, H. (1936). Analysis of flow in networks of conduits or conductors. Bulletin No. 286, University of
Illinois Engineering Experimental Station, Urbana, Illinois.

Dandy, G.C., Simpson, A.R. and Murphy, L.J. (1996). “An improved genetic algorithm for pipe network
optimisation.” Water Resources Research, 32 (2), Feb., 449-458.

Epp, R. and A. G. Fowler (1970). “Efficient code for steady-state flows in networks.” Journal of the
Hydraulics Division, Proceedings of the ASCE January, Vol. 96, No. HY1: 43-56.

Murphy, L.J. and Simpson, A.R. (1992). Pipe optimisation using genetic algorithms, Research Report
No. R93, Department of Civil Engineering, The University of Adelaide, June, 53pp.

Schaake, J. C., and Lai, D. (1969). “Linear programming and dynamic programming applications to
water distribution network design.” MIT Hydrodynamics Laboratory Report No. 116, July.

Shamir, U. and D. D. Howard (1968). “Water distribution systems analysis.” Journal of the Hydraulics
Division, Proceedings of the ASCE, Vol. 94, No. HY1, January, 219-235.

7
 Chapter 1 - File: Chap1.fm July 6, 2022 Version 7

Simpson, A.R., Dandy, G.C. and Murphy, L.J. (1994). “Genetic algorithms compared to other tech-
niques for pipe optimization”, Journal of Water Resources Planning and Management, Vol. 120, No. 4,
July-August, 423-443.

Walski, T. M., Gessler, J., and Siostrom, J. W. (1988). “Selecting optimal pipe sizes for water distribu-
tion systems.” Journal of the American Water Works Association, February, 35-39.

8
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

CHAPTER 2 PROPERTIES OF WATER1

The objective of providing a water distribution system is to transport water from sources to consumers.
Ordinary water is not a simple substance but rather it is a mixture of isotopes and possibly several poly-
mers (Skeat and Dangerfield 1969). This Chapter deals with the important properties of water that are
required for the analysis of pipeline flow. The definition of a fluid is given and its relation to stress, and
in particular, shear stress. Prior to describing properties of water, it is necessary to discuss units and
measurement systems. Various units of measurement have been used in the past. The most common sys-
tems of measurement are SI units and US customary units (or Imperial units). Both of these systems are
described in this Chapter. In addition, both systems are used throughout the book.

There are only a few properties of water that are of any practical importance in pipe flow analysis. The
important properties of water described in this Chapter include viscosity, density, specific weight, spe-
cific gravity, pressure and compressibility of water. The properties generally depend on temperature and
pressure. Usually, the range of temperature and pressure variation of concern is quite small. When water
properties are given it is usual to specify these for pure, air – free water at atmospheric pressure. Fresh
water normally contains impurities and dissolved gases. Their presence modifies the properties only
slightly. It is assumed that the effect on the properties of water due to these impurities is negligible.

The effect of gravity on water results in a pressure variation within the fluid. This concept is extremely
important. The concept of pressure variation within a fluid and measurement of pressure is also intro-
duced in this Chapter. Absolute pressure and gage pressure are defined. Manometers for determination
of pressure in pipe flow systems are also considered.

2.1 Fluids, Solids and Gases

Pure matter is either a pure solid or a pure fluid. Pure fluids are subdivided into liquids and gases. Liq-
uids and gases are assumed to be a continuum (Streeter et al. 1998). Intermolecular forces are large in a
solid, therefore solids have a rigid compact form. These intermolecular forces are smaller in a fluid.
Within the class of fluids, a liquid has a finite volume and a well – defined free surface. Intermolecular
forces are very small in a gas and thus a gas expands to fill the container and consequently has no free
surface.

2.1.1 Types of stresses

A normal force is the force component perpendicular to the surface. A shear force is the force compo-
nent parallel to the surface. Both of these types of forces are shown in Figure 2.1

Normal stress on a surface of area A as a result of a force perpendicular to the surface is

FN
 = -------
- (2.1)
A

DEFINITION OF NORMAL STRESS

where

•  = normal stress (N/m2 or lbf/ft2)

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.

9
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

• F N = force normal to the surface (N or lbf)

• A = area on which the force is acting (m2 or ft2)

Figure 2.1 Normal force and shear force

A shear stress on a surface of area A as a result of a shear force is

F
 = ------T- (2.2)
A

DEFINITION OF SHEAR STRESS

where

•  = shear stress (N/m2 or lbf/ft2)

• F T = force tangential to the surface (N or lbf)

2.1.2 Definition of a fluid

A fluid is defined as a substance that deforms continuously when subjected to a shear stress, no matter
how small the value of shear stress. See Figure 2.2.

Figure 2.2 A fluid subject to shear or tangential stress

The major difference between a fluid and a solid occurs when a fluid is subjected to an infinitesimally
small shear stress. A solid does not move when subjected to a small shear stress while a fluid does begin
to move.

2.1.3 Behavior of solids and fluids under the action of stresses

When solids are subjected to tensile, compressive or shear stresses the result is elastic deformation (ini-
tially) and eventually permanent distortion. When fluids are subjected to tensile or compressive stresses
the result is also elastic deformation but no permanent deformation occurs. Figure 2.3 shows a static
fluid subject to compressive stress as a result of a compressive pressure p, due to say atmospheric pres-
sure, on the surface. The pressure increases by an amount h at a distance h below the surface. The

10
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

pressure increase with depth in the liquid shown in Figure 2.3 is due to gravity. The pressure at a point is
supporting the weight of fluid above it to the surface. A detailed description of the pressure variation
with depth is given in Section 2.10. We will see that the pressure increases linearly with vertical depth
below the water surface. In Figure 2.3 the amount of increase at the bottom of the tank is:

p = gh (2.3)

PRESSURE INCREASE WITH DEPTH

where

• p = change in pressure for an increase in depth of h below the water surface (Pa or psi)
•  = density of the fluid (kg/m3 or slugs/ft3)
• g = gravitational acceleration (m/s2 or ft/s2)
• h = vertical depth below the water surface (m or ft)
The pressure on the water surface is defined as a force per unit area as

FN
p = -------
- (2.4)
A

DEFINITION OF PRESSURE

where

• p = pressure (N/m2 or lbf/ft2)

• F N = force normal to the surface (N or lbf)
• A = area on which the force is acting (m2 or ft2)

*
Figure 2.3 A fluid subject to a compressive stress (atmospheric pressure of p atm
in absolute units is acting on the surface)

2.2 Units of Measurement

There are two commonly used sets of units of measurement. The first is SI units (Systeme d’Internatio-
nale) and the second is the US customary units system. Table 2.1 shows the basic units for both systems.

11
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Table 2.1 Basic quantities for units of measurement systems

Quantity SI units US customary units
length meter (m) foot (ft)
mass kilogram (kg) slug or slugs*
time second (s) second (s)

*sometimes pound – mass or lbm is used.

Some typical units of derived quantities that are important in hydraulics of water distribution networks
are given in Appendix B. Note that where a unit is named after someone (for example, Sir Isaac Newton
for force) — the abbreviation is capitalized (that is, N), whereas when written in full — lower case letters
are used (that is, newton).

It is very important to clearly distinguish between mass and force in both sets of units. The following
sections deal with this difference.

2.2.1 Newton’s second law of motion

Newton’s second law of motion is given as

F = ma (2.5)

NEWTON’S SECOND LAW OF MOTION

where

• F = force (as a vector quantity) (N or lbf)

• m = mass (kg or slugs)
• a = acceleration (as a vector quantity) (m/s2 or ft/s2)
A very important acceleration parameter in the study of hydraulics and water distribution systems is the
acceleration due to gravity. Typical values of gravitational acceleration are given in Table 2.2. The value
of gravitational acceleration g varies slightly with the Earth’s latitude and longitude.

Table 2.2 Acceleration due to gravity

Constant Symbol SI units US customary units
Value Units Value Units
Gravitational acceleration g 9.81 (9.806) m/s2 32.2 (32.174) ft/s2

2.2.2 Force in SI units

The definition of a force in SI units using Newton’s second law of motion (Equation 2.5) is

A force F of one newton accelerates a mass m of one kilogram at an acceleration a of

1 m/s2.

In the SI units system, mass is commonly used rather than force when defining the property of the object
on earth (in contrast to usage of weight in US customary units as discussed below). To convert a force or
weight W involving gravity to a mass m it is necessary to divide by the gravitational acceleration g as

12
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

W
m = ------ (2.6)
g

RELATIONSHIP BETWEEN MASS AND WEIGHT

where

• m = mass (kg or slugs)

• W = weight (N or lbf)
• g = gravitational acceleration (m/s2 or ft/s2)

Thus if a 1 kg mass is falling freely, accelerated by gravity, the force W or F acting on the object is

W = ma = mg = 1  9.81  = 9.81 newtons

2.2.3 Force in US customary units

The definition of a force in US customary units using Newton’s second law of motion (Equation 2.5) is

A force F of one pound (lbf) accelerates a mass m of one slug at an acceleration a of

1 ft/s2.

In the US customary system of units, mass (in slugs) is rarely used while reference to the force in lbf is
commonly used when referring to objects subject to the action of gravity on earth.

Thus if a 1 lbf weight is falling freely, accelerated by gravity, the mass m of the object must be

W 1
m = ------ = ---------- slugs
g 32.2

where

• W = the weight = 1 lbf

• g = gravitational acceleration = 32.2 ft/s2
Sometimes the units of pound – mass or lbm is used for mass in US Customary units such that

1 lb m = 32.2 slugs

2.2.4 Summary of properties of water

The commonly used values of the important properties of water and also the density of air (for compari-
son) are given in Table 2.3. These properties are each discussed later in the Chapter.

Table 2.3 Commonly used values of properties of water

Name Symbol SI units at 15°C US customary units at 60°F
(Celsius) (Fahrenheit)
Value Units Value Units
Absolute viscosity of water  –3 Pa–s –5 slugs/s – ft
1.138 10 2.356 10
Kinematic viscosity of water  –6 2 –5 2
1.140 10 m /s 1.217 10 ft /s

13
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Table 2.3 Commonly used values of properties of water

Density of water (at 15°C)  999.1 kg/m3 1.938 slugs/ft3
Specific weight of water  9801 N/m3 62.37 lbf/ft3
Vapor pressure — absolute p v 1707 Pa 0.59 ft
units
Bulk modulus of elasticity K 2.149 GPa 311,000 lbf/in2
of water
Surface tension of water**  0.0735 N/m 0.00503 lbf/ft

Table 2.4 Properties of water and air at standard conditions (STP)

Name Symbol SI units US customary units
Value* Units Value Units
Density of water at standard condi-
*  1000 kg/m3 1.94 slugs/ft3
tions (4°C or 39.2°F and 760 mm Hg)
Density of air at standard air 1.275 kg/m3 0.002473 slugs/ft3
conditions — for comparison with
water (4°C or 39.2°F and
760 mm Hg)

*The maximum possible density of water at 1000 kg/m3. ** Source https://www.engineeringtoolbox.com/

water-density-specific-weight-d_595.html (accessed 12 November 2019)

2.2.5 Common water quantities

There are a number of units used when referring to quantities of water. Some of these are shown in
Table 2.5.
.
Table 2.5 Common units and abbreviations when referring to water
Unit Equals Abbre- Volume it occupies Example
viation
liter 1 - m 3 of water L A box of A liter carton of milk
----------- 10 cm by 10 cm by 10 cm
1000
kiloliter 1000 liters or 1000 L kL A box of The average household
3 1 m by 1 m by 1 m water use in Adelaide,
( 1  10 L) or
South Australia in 2019
3
1 m of water was 184 kL per annum
megaliter 1 million liters ML A box of Kangaroo Creek Dam
6 10 m by 10 m by 10 m reservoir to the east of
( 1  10 L) of water
Adelaide, South Australia
or 1000 m3 of water has a capacity of 18,700
ML of water
gigaliter 1 thousand million GL A box of The desalination plant in
9 100 m by 100 m by 100 m Adelaide, South Australia
liters ( 1  10 L) of
has a capacity of 100 GL
water or 1 million m3 per annum
of water

2.2.6 Conversion of units

In SI units, common abbreviations of units are shown in Table 2.6.

14
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Table 2.6 Common abbreviations and other defined units in the SI units system
Prefix Value Abbreviation Common examples of usage
micro – 10 – 6  m (micrometer)
milli – 10 – 3 m mm (millimeter)
centi – 10 – 2 c cm (centimeter)
hecto – 102 h hectare (ha) (a measure of area = 10,000 m2 or 100 m by 100 m)
kilo – 103 k km, kg, kPa (kilometer, kilogram, kilopascal)
mega – 106 M ML, MPa (megaliter, megapascal)
giga – 109 G GL (gigaliter)

The conversion of units between the SI units system and the US customary units system, or within each
system, often leads to confusion and needs to be carried out carefully. Table 2.7 shows some examples
of conversions. Detailed lists of conversions are given in Appendix B (Table B.1). Some examples of
abbreviations are given in Table 2.8. Additional abbreviations are given in Table B.2 in Appendix B.

An example of conversion of units is given below.

Example 2.1 Convert 30 psi in US customary units to SI units of pascals.

Answer – Conversion of US Customary units to SI units.

We want a quantity with units of N/m2. It is very important to take a methodical approach to units conversion otherwise
it is easy to make errors. The following procedure is recommended for all units conversions. Using the conversions in
Table B.1 from Appendix B it follows that

2 2 2
lb-  4.44822 N-  -----------------
12 in -  --------------------------------
1 ft
30 psi = 30 ------ -------------------------
in
2 1 lb 1 ft
2
 0.3048  m
2 2

On the right hand side, the second term specifies the number of newtons (4.44822 in the numerator) per pound (1 in the
denominator) – thus the pound (lbf) in the numerator of the first term on the right hand side cancels with the pound
(lbf) in the second term in the denominator. The third term on the right hand side converts square inches to square feet
and finally the fourth term on the right hand side converts square feet to square meters. Once the units on the right hand
side are cancelled out then the remaining units are N/m2 or pascal.

It is always best to write out the conversion in the form shown above. This prevents errors from being made in units
conversion.

Simplifying leads to

30  4.44822   144  N
30 psi = --------------------------------------------- -------- = 206843 Pa
2 2
 0.3048  m
An alternative method for conversion, again using the conversions in Table B.1, is

3 2
lb 6.89476 10 N/m
-  -----------------------------------------------
30 psi = 30 ------
2 2
in 1 lb/in
or

3 2
30 psi = 30  6.89476 10  N/m = 206 843 Pa = 206.8 kPa

___________________________________________________________________________________

15
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Table 2.7 Examples of units conversions

Quantity Symbol SI units US Conversion

and generic customary
units units or
non – SI units
Length L m (meter) ft (feet) 1 ft = 0.304800 m
[L] mm in. (inch) 1 in. = 25.40 mm
km mile 1 mile = 1.609 km
Velocity V m/s ft/s (fps) 1 ft/s = 0.304800 m/s
[LT – 1]
Force F N lb or lbf 1 lbf = 4.44822 N
[MLT – 2] (newton) (pounds force), 1 kgf = 9.80665 N
kgf (kilograms
force)
Pressure p Pa (or N/m2) psi (lbf/in2), 1 Pa = 1 N/m2
Normal stress  (pascal) atm 3
Shear stress 1 lbf/in2 = 6.89476 10 Pa
 kPa (atmospheres),
MPa 1 ft water = 2.988 kPa
[ML – 1T – 2] bars, lbf/ft2 1 atm = 101,325 Pa
GPa
1 atm = 14.6960 psi
1 bar = 0.98692 atm

2.3 Viscosity
2.3.1 Definition of absolute viscosity
Viscosity is that property of a fluid that offers resistance to shear forces that develop in the fluid. Con-
sider two closely spaced plates to examine the viscosity of a fluid as shown in Figure 2.4.

Figure 2.4 Flow between two closely spaced parallel plates

16
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Table 2.8 Examples of abbreviations

Term Abbre- SI units or Term Abbre- SI units or

viation US customary viation US customary
units units
absolute abs both megawatt MW SI
cubic feet cfs US megaliters per day ML/day SI
per second
foot – pound ft – lbf US newton N SI
kilowatt kW SI watt W SI

A liquid is placed between the plates separated by a distance t. Assume the plates are very large and of
area A, such that the edge effects are neglected. Assume the lower plate is fixed and that the upper plate
has a force of F applied. When the force F causes the upper plate to move, no matter how small the mag-
nitude of F, the substance between the two plates is concluded to be a fluid. At the boundary of the upper
moving plate, the fluid and boundary must have the same velocity. This is the no slip condition. Experi-
ments show that the magnitude of the force F to move the upper plate in Figure 2.4 depends on the fol-
lowing quantities

• plate area (A)

FA

• plate velocity (U)

FU

• distance between plates (t)

F  1---
t

Thus combining these relations gives

AU
F  --------- (2.7)
t

Equation 2.7 may be expressed as

AU
F =  --------- (2.8)
t

where

•  proportionality factor called the absolute viscosity or viscosity of the fluid (Pa – s or
slugs/s – ft)
Expressing Equation 2.8 in terms of shear stress (  ) gives

F U
 = ---- =  ---- (2.9)
A t

where

• the ratio U/t is the rate of angular deformation

17
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

In addition, the ratio U/t is may also be written as an angular velocity, du/dy as shown in Figure 2.5.
Note that in Figure 2.4 and Figure 2.5 the velocity distribution is exactly linear.

Figure 2.5 Variation of velocity between two plates

Thus the expression for shear stress in Equation 2.9 becomes

du
 =  ------ (2.10)
dy

NEWTON’S LAW OF VISCOSITY

where

•  = shear stress (N/m2 or lbf/ft2)

•  absolute viscosity or viscosity of the fluid (Pa – s or slugs/s – ft)

du
• the ratio ------ is the rate of angular deformation
dy
For a stationary fluid, the value of du/dy is zero as the velocity u is zero in the fluid. Consequently the
shear stress  is zero in the stationary fluid. Thus in Equation 2.10 it follows that

du
 =  ------ =   0  = 0
dy

To contrast to the behavior of water, consider the shear stress if a vacuum replaced the water. A vacuum
between the plates would have an absolute viscosity of zero (, thus the top plate would be acceler-
ated continuously when a force F is applied to the upper plate. It follows that the shear stress in a vac-
uum is also zero as follows

du du
 =  ------ = 0 ------ = 0
dy dy

2.3.2 Units of absolute viscosity

Rearranging Newton’s law of viscosity in Equation 2.10 gives


 = ------
du
------
dy

The units of absolute viscosity is expressed in general as

FA
   = ------------ (2.11)
Uy

18
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

For SI units this becomes

2
N/m - = N–s
   = ------------------- --------- (2.12)
 m/s /m m
2

But from Newton’s second law of motion in Equation 2.5 as

F = ma

The units of force using SI units are

m kg–m
 F  = N =  m   a  =  kg   ------- = -------------- (2.13)
 2 2
s s

Now as an alternative form the units of absolute viscosity from Equation 2.11 and Equation 2.13
become

 kg–m
--------------  m
2
 2 
FA s kg
   = ------------ = ------------------------------- = ----------
Uy  m/s /m m–s

Thus the units of absolute viscosity are

kg- = ------
M-
   = ---------
m–s LT

where

• M = mass (kg or slugs)

• L = length (m or ft)
• T = time (s)
2.3.3 Absolute viscosities of substances
Water and air have relatively small absolute viscosities. Tar is a highly viscous liquid. For water,
Table 2.3 gives typical values of viscosities. A full list of absolute and kinematic viscosities for water at
various temperatures is given in Appendix A.

–3
For crude oil the absolute viscosity is  = 8 10 N – s/m2.

2.3.4 Kinematic viscosity

The kinematic viscosity  is a derived quantity based on the absolute viscosity and the density of the
fluid. It is defined as

 = 
--- (2.14)


KINEMATIC VISCOSITY OF A FLUID

where

•  = kinematic viscosity of the fluid (m2/s or ft2/s)

•  absolute viscosity of the fluid (Pa – s or slugs/s – ft)

19
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

•  = density of the fluid (kg/m3 or slugs/ft3)

Thus the units of kinematic viscosity from Equation 2.14 are

2 2
   kg/(m–s) m L
   = -------- = ---------------------- = ------- = ------
 kg/m
3 s T

Values of absolute and kinematic viscosities for water are given in Table 2.3.

A full list of kinematic viscosities for water at various temperatures is given in Appendix A. The Inter-
national Standard - IEC (1999) gives the following formula for the kinematic viscosity of water at a Cel-
sius temperature T as:

 – 16.921 + 396.13   107.41 + T  

 = e

2.3.5 When is viscosity important? (Definition of Reynolds number)

Viscosity is important in the analysis of fluid flow in pipes. Later in the book (see Section 3.1.3) a
dimensionless number for pipe flow called the Reynolds number or Re is discussed in detail. The Reyn-
olds number is defined as

VD
Re = -------- (2.15)


where

• Re = Reynolds number (dimensionless)

• V = average flow velocity for fluid in a pipe (m/s or ft/s)
• D = diameter of the pipe (m or ft)
•  = kinematic viscosity of the fluid (m2/s or ft2/s)
Equation 2.15 is the ratio of an inertial force and a viscous force. The units of Reynolds number are
dimensionless as follows

 V   D   m/s   m 
 Re  = ------------------ = ----------------------- = 1
  m /s 
2

and therefore the Reynolds number is a dimensionless quantity. If the Reynolds number is small
(< 2300) then in pipe flow — the flow type is laminar and the viscous effects (the  term) dominate the
inertial effects (the VD term).

2.4 Density
Density is the mass per unit volume in the limit as the size of the volume approaches an infinitely small
size  . Thus

m
 = lim -------- (2.16)
V   V

where

• m = mass of a small volume of fluid (kg or slugs)

• V = volume (m3 or ft3)
An important reference point for specifying conditions in liquids and gases is referred to as standard
conditions — these are summarized in Table 2.9.

20
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Table 2.9 Standard conditions for temperature and pressure (STP)

State Symbol SI units US customary units
Value Units Value Units
Temperature T 4 °C 39.2 °F
Pressure (absolute units) p* 101,325 Pa 14.7 psi
Pressure (absolute units) p* 2116.8 lbf/ft2
Pressure — as a column of p* 760 mm Hg 760 mm Hg
mercury height (absolute units)
Pressure — as a column of water p* 10.34 m 33.91 ft
height (absolute units)

2.4.1 Density of water

The density of water changes slightly with temperature. Water is most dense at 4°C (39.2°F) and
becomes less dense above and below this temperature. Commonly used values of the density of water
are given in Table 2.3. Other values of density for water at different temperatures are given in
Appendix A.

A change in pressure changes the volume occupied by a substance. As a result, density of a liquid varies
slightly with pressure. Thus density increases as pressure increases and density decreases as pressure
decreases. The effect of pressure on density of solids and liquids is very small and the effect is usually
ignored.

The specific volume vs is the reciprocal of density

1
v s = --- (2.17)


DEFINITION OF SPECIFIC VOLUME

The units of specific volume are

3 3
m ft
 v s  = ------- or ---------- (2.18)
kg slug

2.4.2 When is the density of water important?

In a straight section of pipe upstream of the valve, consider a situation where the flow is uniform and the
velocity is constant (see Section 3.1.2). Mass density or more commonly density is important in non –
uniform flow where the velocity of flow changes magnitude or direction along the pipe. An example of
non – uniform flow is the flow through a partially opened valve. As the flow accelerates through the
valve both the magnitude and direction of the velocity vector changes. Once through the valve, the flow
decelerates and resumes as uniform flow in the straight pipe downstream of the valve.

2.4.3 Density of gases

The effect of both temperature and pressure is much more important for gases than liquids. When the
density of a gas is given — usually the temperature and pressure are also given. First let us define the
universal gas constant as

R = 8314.472 J/(kg . K.mol) (2.19)

The gas constant for a particular gas is

21
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

R
R = ----- (2.20)
M

where

• M = molecular weight (for example for air M = 29)

Thus for air:

8314.472
R = ---------------------- = 286.71 (2.21)
29

Now the perfect gas law is

pv s = MRT (2.22)

PERFECT GAS LAW

where

• p = pressure in absolute units (N/m2 or lbf/ft2)

• vs = specific volume (m3/kg or ft3/slugs)
• R = gas constant (m – N/(kg – K) or ft – lbf/(slugs – °R)); K is degrees Kelvin while °R is
degrees Rankine (both in absolute units)
• T = temperature in absolute units (K or °R)
Typical values of the gas constant R are given in Table 2.10.

Table 2.10 Gas constants

Constant Symbol SI units US customary units
Value Units Value Units
Gas constant for air R 287 J/(kg . K) 1715 ft – lbf/(slugs . °R)
Gas constant for air R 53.3 ft – lbf/(lbm – °R)
(US inconsistent units)*
Gas constant for water vapor R 462 J/(kg – K) 2762 ft – lbf/(slugs – °R)

Relationships between the various temperature scales is given in Table 2.11.

Table 2.11 Temperature relationships+

Kelvin (K) Celsius (°C) Fahrenheit (°F) Rankine (°R)
Absolute zero 0K -273.15 °R -459.67 °F 0 °R
(by definition)
Freezing point of water 273.15 K 0 32 °F 491.67 °R
Relationship equation K = °C +273.15 °R = °F +459.67

+ https://en.wikipedia.org/wiki/Rankine_scale (accessed 4 November 2019)

The gas constant is obtained from the molecular weight of a gas. Thus in SI units (Streeter et al. 1998)

8312 m–N
R = ------------ --------------- (2.23)
M kg–K

THE GAS CONSTANT (SI UNITS)

22
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

where

• R = gas constant (m-N)/(kg-K)

• M = molecular weight of gas (for example, for air M = 29)
Alternatively in US customary units the gas constant is

 709- -------------------
R = 49 ft–lb - = 1545 ft–lb-
----------------- ------------ -------------- (2.24)
M slug–R M lb m R

where

• 1 slug = 32.174 lbm

Molecular weights are given in Table 2.12.

Table 2.12 Molecular weights of gases

Type of gas Molecular weight
Air 29.0
Water vapor 18.0

Two examples involving the perfect gas law are now given below.

Example 2.2 For air at standard conditions (standard temperature and pressure = STP) compute its density in SI
units.

Answer – Computation of air density at STP.

The density is computed using the perfect or ideal gas law Equation 2.22 using values from Table 2.9 (T = 4°C,
p = 101, 325 Pa absolute) as

pv s = RT

Thus in SI units and using the gas constant from Table 2.10 gives

p - = ------------------------------------
1- = ------- 101325 - = 1.275 kg/m 3
 = ---
vs RT  287   4 + 273 

Example 2.3 For air at standard conditions compute its density in US customary units.

Answer – Computation of air density at STP.

The density is computed using the perfect or ideal gas law Equation 2.22 and values from Table 2.9 (T = 39.2°F,
p = 14.7 psi = 2116.8 lbf/ft2 absolute) as

pv s = RT

Thus in US customary units and using the gas constant from Table 2.10 gives

1 p 2116.8 3
 = ---- = -------- = ------------------------------------------------ = 0.002473 slug/ft
vs RT  1715   39.2 + 460 

or

1 p - = ---------------------------------------------
2116.8 3
 = ---
- = ------- - = 0.07956 lb m /ft
vs RT  53.3   39.2 + 460 

___________________________________________________________________________________

23
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

2.5 Specific Weight

The specific weight of a fluid is defined as the weight per unit volume. The specific weight is the prod-
uct of the density of the fluid and gravitational acceleration as

 = g (2.25)

SPECIFIC WEIGHT OF A FLUID

where

•  = specific weight of the fluid (N/m3 or lbf/ft3)

•  = density of the fluid (kg/m3 or slugs/ft3)
• g = gravitational acceleration (m/s2 or ft/s2)

Typical units of specific weight are N/m3 or lbf/ft3. Specific weight is temperature and pressure depen-
dent (as for density).

The units of specific weight in SI units are

3 2 kg
   =     g  =  kg/m   m/s  = ------------
2 2
m s

Using the units of force F from Equation 2.13 of

kg–m
 F  = -------------- = N
2
s

It follows that the units of specific weight in SI units are

N- = N/m 3
   = ------
3
m

The units of specific weight in US customary units are

3 2 slug
   =     g  =  slugs/ft   ft/s  = -----------
2 2
ft s

Using the units of force F of

ft  = slug–ft
 F  = lb =  m   a  =  slug   ---- ---------------- (2.26)
 2 2
s s

It follows that the units of specific weight in US customary units from Equation 2.26 are

2
F slug–ft  s lb
   = ------ = ------------------------- = ------
3 3 3
L ft ft

Common values of the specific weight of water are given in Table 2.3.

Other values of specific weight for water are given in Appendix A. Specific weight or unit weight is
important in stratified flow.

24
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

2.6 Specific Gravity

Specific gravity is defined as the ratio of specific weight of the material to the specific weight of water
w as follows


S = ----- (2.27)
w

SPECIFIC GRAVITY OF A FLUID

where

• S = specific gravity of the fluid (dimensionless)

•  = specific weight of the fluid (N/m3 or lbf/ft3)
•  w = specific weight of water (N/m3 or lbf/ft3)

Examples of specific gravities are shown in Table 2.13.

The specific gravity of a gas is usually defined by comparing with the specific gravity of air at a tem-
perature of 20°C and a pressure of 101.325 kPa absolute. The density of air under these conditions is
1.205 kg/m3 or 0.07637 lbm/ft3, where 32.174 lbm = 1 slug.

From Equation 2.22

pv s = RT (2.28)

Thus in SI units the density of air is

1 p 101325 3
 = ---- = -------- = ---------------------------------------- = 1.205 kg/m
vs RT  287   20 + 273 

Thus the specific weight of air is

3
 = g = 1.205  9.81  = 11.821 N/m

Table 2.13 Specific gravities of various materials

Material S
Aluminium (20°C) 2.7
Water (20°C) 1.0
Steel (20°C) 7.8
Copper (20°C) 8.9
Mercury 13.6

In contrast, the density of air in US customary units at 60°F is

1 p 2116.8 3
 = ---- = -------- = ----------------------------------------- = 0.07637 lb m /ft
vs RT  53.3   60 + 460 

Thus the specific weight of air in US customary units is

3
 = g = 0.07637  32.2  = 2.459 lb/ft

25
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

2.7 Vapor Pressure of Water

The vapor pressure is a fluid property giving the pressure at which a liquid will boil. It depends on tem-
perature and barometric pressure. Liquid evaporates at the surface because of molecules escaping from
the liquid surface. The vapor molecules exert a partial pressure in the space, and this is known as the
vapor pressure. Vapor pressure is usually expressed in absolute units rather than gage units (see
Section 2.11). Values of vapor pressure for water are given in Table 2.3. Vapor pressure values as tem-
perature varies are given in Table 2.14.
°
Table 2.14 Vapor pressure (in absolute units) for water for variation in temperature

p v p v

°C Pa (abs) °F psi (abs)

0 611 32 0.0886
5 872 41 0.1265
10 1,228 50 0.1781
12 1,403 54 0.2034
14 1,599 57 0.2318
16 1,817 61 0.2636
17 1,937 63 0.2810
18 2,064 64 0.2993
19 2,197 66 0.3187
20 2,338 68 0.3392
21 2,486 70 0.3606
22 2,644 72 0.3834
23 2,809 73 0.4074
24 2,984 75 0.4328
25 3,168 77 0.4594

Reference: http://en.wikipedia.org/wiki/Properties_of_water#Density_of_water_and_ice (accessed 11 Nov 2019)

The vapor pressure of water at 20°C is p v

p v = 2.338 kPa absolute

or in gage units (assuming the local barometric pressure is equal to standard atmospheric pressure)

p v = p v – p b = 2.338 – 101.325 = – 98.987 kPa

2.7.1 When is the vapor pressure of water important?

In cavitation, liquid flashes to vapor. The cavitation bubbles are swept to a zone of higher pressure
where they collapse and erosion of the surface may occur (for example, on high head spillways). Vapor
pressure is important whenever flow velocities are high. If surface corrosion occurs as a result of cavita-
tion, this may cause considerable damage to interior surfaces of a pump or within hydraulic structures.
An example of severe cavitation in a multijet sleeve valve is shown in Figure 2.6. Appendix A gives
vapor pressures and vapor pressure heads in one degree increments from zero to 100 degrees Celsius.

26
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Figure 2.6 Severe cavitation in a multijet sleeve valve

2.8 Bulk Modulus of Elasticity of a Fluid

The bulk modulus of elasticity is a fluid property relating changes of volume to applied pressure
changes (Crowe et al. 2008).

Often there is a misconception with regard to the compressibility of a number of common materials. It is
clear that gases, such as air, are compressible. When the question as to whether or not steel and water are
compressible is considered, it sometimes leads to confusion. A common misconception is that steel and
water are incompressible. This is not correct. All gases, liquids and solids change volume with every
change in compressive state. Thus both water and steel are compressible, although to different degrees.
Despite the occurrence of the compressibility of water as the pressure changes, incompressibility (or a
constant fluid density) is often assumed in order to simplify the solution of hydraulics problems.

Liquid compressibility is expressed by the bulk modulus of elasticity K. For liquids, the bulk modulus
of elasticity K is defined based on Figure 2.7 as

stress p
K = ---------------------------------------- = ---------------------------- (2.29)
volumetric strain – V M  V M

BULK MODULUS OF ELASTICITY

where

• K = bulk modulus of elasticity (N/m2 or lbf/ft2)

• p = change in pressure (N/m2 or lbf/ft2)
• V M = change in volume — note that the value of V M is negative if the volume of the box
decreases due to the application of the positive pressure (m3 or ft3)
• V M = original volume (m3 or ft3)

27
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

For solids, the modulus of elasticity is defined as the coefficient of elasticity of stretching. For a
stretched wire, Young's modulus of elasticity is the ratio of the stretching force per unit cross – sectional
area to the elongation per unit length. Young’s modulus of elasticity E is also related to stress and strain
as follows

stress  
E = ------------- = --- = -------------- (2.30)
strain  L  L

YOUNG’S MODULUS OF ELASTICITY

where

• E = Young’s modulus of elasticity (N/m2 or lbf/ft2)

•  = normal stress (N/m2 or lbf/ft2)
•  = longitudinal strain
• L = change in length or elongation due to the applied stress (m or ft)
• L = length (m or ft)
For steel, the Young’s modulus of elasticity is E = 200 GPa. Thus water is approximately 100 times
more compressible than steel.

Figure 2.7 Definition of bulk modulus of elasticity K

Table 2.15 gives typical values of the bulk modulus of water.

Table 2.15 Common values of bulk modulus of elasticity of water

Name Symbol SI units at 15°C US customary units at
60°F
Value Units Value Units
Bulk modulus of elasticity K 2.14 GPa 311,000 lbf/in2
Bulk modulus of elasticity K 2.14 x 109 Pa

2.8.1 When is the bulk modulus of elasticity K important?

Compressibility is an important property in water hammer or fluid transient flow. Water hammer is the
unsteady flow in pipelines such that velocity and pressure at a point change with time (see Chapter 21).
Water hammer waves propagate at about 1486 m/s through water (for example, in a lake) due to the
compressibility of water. Water hammer depends on fluid compressibility where there are sudden
changes in pressure. For example, a sudden closure of a valve at the end of a reservoir pipeline system
results in a water hammer wave propagating in the pipe.

28
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

2.9 Other Fluid Properties

Another fluid property is surface tension. Surface tension is not so important in pipeline systems. Sur-
face tension is important in physical model experiments especially where open channel flow is involved.

2.10 Pressure Variation in a Fluid

Pressure variation in a pipeline system is one of the most important considerations in the design of a
water distribution network. As a first step to understanding the way pressure in a fluid works, it is useful
to consider the pressure variation in a stationary liquid. The height of water is the most important factor
in determining pressures.

Consider what happens to the pressure when you dive into water and go deeper and deeper. The pressure
increases linearly with an increase in depth. There is a direct relationship between the pressure in N/m2
(or lbf/ft3) at any point in the water and the height of water in meters or feet above that point. The pres-
sure is expressed in meters or feet of water and is called head (see Section 2.10.1 and specifically
Equation 2.39). Consider the coordinate directions defined in Figure 2.8.

Figure 2.8 Definition of a set of coordinate directions

The axis directions are defined as

• h is measured vertically downwards from the free surface

• y is measured vertically upwards from the bottom
• x is measured along the page to the right
• z is measured perpendicularly into the page
There is no variation in pressure p in the horizontal plane in Figure 2.8. Thus

p = 0 (2.31)
x h, z fixed

and

p = 0 (2.32)
z h, x fixed

Thus in Figure 2.9 it follows for A and B on a horizontal plane at the same depth below the horizontal
water surface that

p A = pB (2.33)

Figure 2.9 Variation of pressure on a horizontal plane

29
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

In contrast, there is a linear variation in pressure with depth in the h – direction downwards from the sur-
face of the liquid as follows

p =  = g (2.34)
h x, z fixed

where

p
• = change in fluid pressure with depth
h
•  = specific weight of the fluid (N/m3 or lbf/ft2)
•  = density of the fluid (kg/m3 or slugs/ft3)
• g = gravitational acceleration (m/s2 or ft/s2)
The linear variation of pressure with depth is shown in Figure 2.10.

Figure 2.10 The variation of pressure with depth

Since p is a function of h only, the partial derivative in Equation 2.34 is replaced by a total derivative as

dp
=  (2.35)
dh

For a homogeneous and incompressible fluid, the specific weight of the liquid  is constant, thus integra-
tion of Equation 2.35 leads to

dp = dh

 dp =  dh =   dh

and thus

p = h + c (2.36)

The pressure at the surface of the liquid is zero in gage units (that is: p = 0 at h = 0 in Figure 2.10) — see
Section 2.11. Substituting these values into Equation 2.36 results in

c = 0 (2.37)

30
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Substituting Equation 2.37 into Equation 2.36 gives

p = h (2.38)

VARIATION OF PRESSURE WITH DEPTH

where

• p = pressure (N/m2 or lbf/ft2)

•  = specific weight of the fluid (N/m3 or lbf/ft3) = g
•  = density of the fluid (kg/m3 or lbf/ft2)
• g = acceleration due to gravity = 9.81 m2/s
• h = vertical depth below the surface (m or ft)

Example 2.4 Compute the pressure 2 m beneath a static water surface as shown in Figure 2.10.

Answer – Computation of pressure.

Assume the water temperature is 20°C. The specific weight is 9,789 N/m3. Using Equation 2.38 gives the pressure at a
depth of 2 meters is

p = h = 9789  2.0  = 19,578 N/m2 = 19, 578 Pa = 19.6 kPa

___________________________________________________________________________________

2.10.1 Relation between pressure and column height

Pressure is often expressed as a head or as a height of a column of water. Pressure at a point below the
surface occurs because of the weight of the fluid above that point. The pressure head is taken as the pres-
sure divided by the specific weight of the fluid as given as follows

p
--- = h (2.39)


DEFINITION OF PRESSURE HEAD

where

• p = pressure (N/m2 [or Pa] or lbf/ft2)

•  = specific weight of the fluid (N/m3 or lbf/ft3)= g
•  = density of the fluid (kg/m3 or slugs/ft3)
• g = acceleration due to gravity = 9.81 m2/s = 32.2 ft2/s)
• h = vertical depth below the surface (m or ft)
To interpret a head of 3 m of water:

This is equivalent to the pressure (force per unit area) at the base of a column of water
of 3 m in height.

The pressure in a liquid column does not depend on the shape of the container. As long as the height of
the water stays the same, changing the shape of the container does not change the pressure at the bottom
of the container. Pressures are also identical for the same depth in any of the containers. Figure 2.11
shows that the pressure is identical at points A, B, C, D and E.

31
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Figure 2.11 Pressure at same height in connected vessels

Thus the pressure head depends only on the height of the column of fluid and not on the size of the base
of the column. The weight of the fluid column above it causes pressure.

Consider the following example that illustrates the concept of the relationship between the height of a
fluid column and the pressure at the base of the fluid column.

Example 2.5 Consider a rectangular column, 3 m high and 0.5 m by 0.5 m square filled with water as shown in
Figure 2.12. What is the pressure head at the base? Assume the water temperature is 20°C.

Figure 2.12 Pressure at bottom of a 3 m high column of fluid

Answer – Computation of pressure at base.

The bottom is supporting the weight of the column of water.

The area of the supporting bottom area is

A = 0.5  0.5 = (0.5)2 m2

The total force on the base is

F = V M = 9789  0.5   0.5   3  N

where

• VM = volume of water in the column (m3 or ft3)

Thus the pressure at the base of the column of water is given by Equation 2.38

2
F  0.5   3   9789 
p = ---- = ----------------------------------------
A  0.5 
2

32
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

or cancelling the (0.5)2 terms gives

2
p = h = 3  9789  = 29 367 N/m

The pressure head from Equation 2.39 is

p 3  9789 
pressure head = --- = -------------------- = 3 m of water
 9789

___________________________________________________________________________________

Now for another example.

Example 2.6 What is the pressure at the base of a water column of 14 ft in height?

Answer – Computation of the pressure at the base of a water column.

Assume the specific weight of water is 62.4 lbf/ft3 then the conversion becomes

2
62.4 lb lb 1 ft
p g = 14 ft = 14 ft  ----------------- = 14  62.4 -------  ------------------ = 6.06 psi
3 2 2
1 ft ft 144 in

where psi = lbf/in2 in gage units (denoted by the subscript “g”)

___________________________________________________________________________________

2.11 Absolute and Gage Pressure

When the water pressure in a main, pipe or container is measured by a gage, the pressure is referred to as
gage pressure. Gage pressure is measured above a zero level specified by local atmospheric pressure. A
pressure gage does not show the total or absolute pressure. Gage pressure does not show the pressure of
the atmosphere. Atmospheric pressure is exerted everywhere (against the outside of the main as well as
inside the main). In many calculations, the effect of atmospheric pressure may be neglected.

Absolute pressure is measured above a zero specified by a perfect vacuum (absolute zero). The relation-
ship between gage and absolute pressure is demonstrated in Figure 2.13. Thus the relationship between
absolute pressure and gage pressure is as follows

p a = p g + p b (2.40)

ABSOLUTE PRESSURE AND GAGE PRESSURE RELATIONSHIP

where

• p a = absolute pressure (N/m2 or Pa); the subscript “a” refers to absolute units; the asterisk
or * as a superscript on the p refers to a pressure in absolute units
• p g = gage pressure (N/m2 or Pa); the subscript “g” refers to gage units; an unsuperscripted
p refers to a pressure in gage units (except in the perfect gas law of Equation 2.22)
• p b = barometric pressure (N/m2 or Pa) in absolute units; the subscript “b” refers to baro-
metric pressure of the atmosphere
The relationship between absolute and gage pressure is illustrated in Figure 2.13 and further described
in Table 2.16.

33
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Figure 2.13 Relationship between absolute, gage, barometric and local atmospheric pressure

Table 2.16 Comparing absolute and gage pressures

Quantity Absolute pressure Gage pressure Pressure head as a
p a pg height of mercury
(absolute units)
Standard atmospheric pressure 101,325 Pa abs. 1,325 Pa 760 mm Hg
* 100,000 Pa abs. 0 Pa 750.1 mm Hg
Local atmospheric pressure ( p b )

For a pressure above zero gage 120,000 Pa abs. 20,000 Pa 900.1 mm Hg

pressure
For a pressure below zero gage 40,000 Pa abs. – 60,000 Pa 300.0 mm Hg
pressure
Full vacuum 0 Pa abs. – 100,000 Pa 0.0 mm Hg

Standard atmospheric pressure (at 4°C or 39.2°F) is defined as

• 760 mm mercury absolute

• 101, 325 Pa absolute or N/m2 absolute
• 101.325 kPa absolute
• 10.34 m water height absolute
If a pressure is less than the local atmospheric pressure (assumed to be 750.1 mm Hg as shown in
Table 2.16) it is called negative, suction or vacuum. Thus 460 mm Hg absolute may be expressed as

• – 290.1 mm Hg gage (= 460 – 750.1)

• 290.1 mm Hg suction
• 290.1 mm Hg vacuum
These are shown in Figure 2.14.

34
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Figure 2.14 Relationship between negative, suction and vacuum pressure

2.12 Measurement of Pressure

2.12.1 Piezometers and pressure gages
Piezometers or vertical open tubes (an example is shown in Figure 2.15) attached at the top of or on the
side of a pipeline may be used to accurately measure the pressure in pipes. However, in many situations
the use of piezometers for pressure measurement is impractical. The height to which the water rises in
the piezometer is a measure of the pressure head (in meters or ft) and hence the pressure (in kPa or psi)
at the centerline of the pipeline. In some cases, for example, piezometers tens of, or even hundreds of
meters high would be required. Thus in most practical situations pressure gage readings are used. A
Bourdon Gage is typically used to measure gage pressures.

In Figure 2.15 the water column in the vertical open manometer rises to the level of the hydraulic grade-
line. Thus the level in the manometer is a measure of pressure head (h1) with respect to the centerline of
the pipe. Thus the pressure at the point A in the pipeline is

p A = h 1 (2.41)

Figure 2.15 An open piezometer tells you the pressure in the pipeline

2.12.2 Measuring atmospheric pressure

As shown in Figure 2.16, a mercury barometer or aneroid barometer is used to measure local atmo-
spheric pressure. The space above a mercury column in a mercury barometer contains mercury vapor
that is at the vapor pressure of mercury at the ambient temperature. The vapor pressure of mercury is
0.173 Pa.

35
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

Figure 2.16 Mercury barometer

2.12.3 Manometers or open piezometers for measuring pressure

Manometers or open piezometers use liquid column lengths to determine differences in pressure as
shown in Figure 2.17. Often liquids used in the piezometer tube are different from the liquid flowing
(for example: mercury or air). If a manometer fluid is different from the flowing fluid then the fluids
must be immiscible.

Figure 2.17 Manometer system connected to a pipeline

2.12.4 Basic principles for analysis of manometer problems

The basic principles of the variation of pressure in a manometer are as follows

• moving horizontally in the same fluid, the pressure remains the same
• moving vertically downwards in the same fluid a vertical distance h1, the pressure increases
by an amount

 fluid h 1

• moving vertically upwards in the same fluid by a vertical distance h2, the pressure
decreases by

 fluid h 2

36
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

where

• h1 and h2 = vertical heights of the fluids being traversed in the manometer (m or ft)
•  fluid = specific weight of the particular fluid in the manometer that is being traversed (N/

2.12.5 Procedure for solving manometer problems

We use the following procedure to solve problems involving a manometer system such as that exhibited
in Figure 2.17

1. Start at one end of the system and assign a pressure. For example, the pressure pA in
Figure 2.17 is unknown.
2. In Figure 2.17 move along the manometer from the centerline of the pipe at A to location R
at the water/mercury interface. A vertical fall (that is, moving downwards through the same
fluid) to the next meniscus (that is, the mercury meniscus in Figure 2.17 at R) increases the
pressure. Thus add to the value pA, the change of pressure to the next meniscus. In this case
the pressure at the next meniscus at R is

p A + w h1

as the fluid that is being traversed is water.

3. In Figure 2.17 move in the manometer fluid from position R to position S. There is no
change in pressure. Moving to the identical horizontal level in the same fluid results in the
same pressure. Thus the pressure at S is the same as the pressure at R as

p A + w h1

In contrast, moving directly from point A to point C' — the pressure cannot be the same
because there are two different fluids connecting the points via the manometer.
4. Continue working your way through to the other end of the manometer at point B.
5. In Figure 2.17 there will be a decrease in pressure moving upwards in the mercury from
point S to point B. A vertical rise (that is, moving upwards through the same fluid) to the
next meniscus decreases the pressure. Thus the pressure at B is
p A +  w h 1 –  Hg  h 1 + h 2 

6. Equate this final expression for pressure to the pressure at this point. For example, pB = 0 in
Figure 2.17.
2.12.6 Solving for the unknown pressure in the pipeline using a manometer
The equation is developed based on the principles of variation of pressure for a manometer as listed
above for the manometer in Figure 2.17 as

p A +  w h 1 –  Hg h 1 –  Hg h 2 = p B (2.42)

where

• p A = pressure at point A (N/m2 [or Pa] or lbf/ft2)

• w = specific weight of water (N/m3 or lbf/ft3)
• h1 and h2 = vertical heights of the fluids being traversed in the manometer
• Hg = specific weight of mercury (N/m3 or lbf/ft3)
• p B = pressure at point B (N/m2 [or Pa] or lbf/ft2)

Using the relationship between specific weight of a fluid and its specific gravity

 Hg = S Hg  w

37
 Chapter 2 - File: Chap2.fm July 6, 2022 Version 6

where

• Hg = specific weight of mercury (N/m3 or lbf/ft3)

• SHg = specific gravity of mercury (= 13.6) (dimensionless)
• w = specific weight of water (N/m3 or lbf/ft3)
Then the pressure at A, pA is determined from

p A +  w  h 1 – S Hg  h 1 + h 2   = p B (2.43)

or

p A = p B –  w  h 1 – S Hg  h 1 + h 2   (2.44)

But at the open surface at the top of the piezometer tube, pB = 0 and thus

p A =  w  S Hg  h 1 + h 2  – h 1  (2.45)

Thus the application of manometer principles have led to a way of measuring the pressure in the pipeline
based on the deflection of the mercury column in the manometer.

2.13 References
Skeat, W.O. and Dangerfield, B. J. ed. (1969). Manual of British water engineering practice. (4th ed.).
Cambridge: Published for the Institution of Water Engineers by Heffer.

Streeter, V.L., Wylie, E.B. and Bedford, K. (1998). Fluid mechanics. 9th Edition. McGraw – Hill.

38
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

CHAPTER 3 EQUATIONS FOR FLUID MOTION IN PIPES1

There are a number of objectives in the analysis of a pipe network or a water distribution system. Some
examples include

• assessing the performance of an existing system to determine zones of low pressure or

pipes with excessive head loss
• determining sizes of pipes and pumps
• simulating the operation of an existing network
The term hydraulics is often used in association with the analysis of piped networks. Hydraulics deals
with applied mechanics in relation to the behavior of water at rest and in motion (Skeat and Dangerfield
1969). Various external forces influence the motion of a fluid. These forces include gravity, mechanical
forces and the characteristics of the boundary (material and condition).

There are three fundamental conservation laws of physics that are essential to the analysis of fluid flow
in pipelines. The three equations include

• law of conservation of mass (often referred to as the continuity equation)

• energy equation
• momentum equation
These equations are considered in detail in this Chapter. Prior to introducing the equations governing
pipeline flow, the different types of flow regimes are described. The concepts of streamlines,
streamtubes and control volumes are also introduced.

As a summary the following tools are used to analyze pipe flow in this Chapter

• similitude and dimensional analysis (discussed in detail in Section 3.2.2)

• equation of continuity (Section 3.3)
• force – momentum flux equation (Section 3.4.2)
• energy principle including the Bernoulli equation (Section 3.5)
• equations of pipe friction loss or fluid resistance (see Chapter 4)
There are a number of dimensionless parameters that are important in pipe flow. These dimensionless
parameters are derived in this Chapter. For pipe flow, the Reynolds number is the most important dimen-
sionless parameter. The concepts of hydraulic grade line (HGL) and energy grade line (EGL) are intro-
duced in this Chapter. They provide a useful insight into how the energy of a fluid particle changes as it
moves through a system. This Chapter concludes with a discussion of the kinetic energy correction fac-
tor that is applied to the kinetic energy in the energy equation.

3.1 Types of Flow

3.1.1 Steady versus unsteady flow
Consider flow along a straight horizontal pipeline as shown in Figure 3.1. Assume the direction along
the centerline of the pipe is the x – coordinate direction.

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Adelaide,
Australia. Part of the book entitled Water Distribution Systems Engineering.

39
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

For steady flow at a given point (that is, x = constant) the velocity does not vary with time as given by

V
= 0 (3.1)
t

DEFINITION OF STEADY FLOW

or

V
= 0 (3.2)
t x = constant

For example, a velocity of 3 m/s (or say 10 ft/s) at a single point or cross – section in the pipeline in
Figure 3.1 at x = 35 m (or say 100 ft) along the pipe remains the same for an indefinite amount of time
and is thus defined as steady flow. In contrast, unsteady flow does vary with time.

V
0 (3.3)
t

DEFINITION OF UNSTEADY FLOW

Figure 3.1 Steady flow in a pipeline

3.1.2 Uniform versus non – uniform flow

In uniform flow, at a given instant in time (that is, t = constant), the velocity does not vary or change
with distance along the pipeline. Thus

V
= 0 (3.4)
x

DEFINITION OF UNIFORM FLOW

or

V = 0 (3.5)
x t = constant

An example of uniform flow is shown in Figure 3.2. Liquid being pumped into a straight constant diam-
eter pipe is described as uniform flow because the velocity vector is the same at all points along the pipe
– same magnitude and same direction).

40
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.2 Uniform flow in a pipeline

In contrast, for non – uniform flow the velocity varies with distance along the pipeline.

V
0 (3.6)
x

DEFINITION OF NON – UNIFORM FLOW

Liquid flowing through a reducing section such as a Venturi flow meter as shown in Figure 3.3 is non –
uniform flow. The velocity vector changes in both magnitude (an increase in V as the cross – section
reduces) and direction.

Figure 3.3 Non – uniform flow through a venturi

3.1.3 Laminar versus turbulent flow

In laminar flow the particles move over smooth paths. One layer glides smoothly over another. The
streamlines are orderly and undisturbed. All the shear stresses are developed purely by molecular inter-
action. Newton’s law of viscosity from Equation 2.10 applies to laminar flow

du
 =  ------ (3.7)
dy

where

•  = shear stress (N/m2 or lbf/ft2)

•  = absolute fluid viscosity ( Pa–s or slugs/s – ft)
du
• ------ = gradient of the longitudinal velocity (u) perpendicular (y) to the flow direction
dy
In turbulent flow, fluid particles move in very irregular paths in the form of random eddies of varying
sizes. Momentum is exchanged from one portion of the fluid to another, across streamlines due to the
irregular movement of finite masses of fluid. The turbulence sets up greater shear stresses than are
attributed to the absolute viscosity via Newton’s law of viscosity for laminar flow. An eddy viscosity is
added to the equation to account for the turbulent flow case (see Equation 3.8) to describe the additional
losses due to turbulence.

du
 =   +   ------ (3.8)
dy

41
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

where

•  = shear stress (N/m2 or lbf/ft2)

•  = absolute fluid viscosity ( Pa–s or slugs/s – ft)
•  = eddy viscosity (Pa – s or slugs/s – ft)

du
• ------ = gradient of the longitudinal velocity (u) perpendicular (y) to the flow direction
dy
Turbulent flow in pipes is the most common flow regime in water engineering. Laminar flow rarely
occurs. However, it is important to fully understand the differences between laminar and turbulent flow.
Osborne Reynolds, in his famous fluid flow experiments in Manchester, England (1880s) showed that
laminar flow was streamlined. Experiments by Reynolds showed the nature of the flow depended on a
dimensionless coefficient, that has come to be known as the Reynolds number. The Reynolds number
was defined in Chapter 2 (Equation 2.15). A detailed discussion of the Reynolds number is given later in
this Chapter (Section 3.2).

Reynolds found evidence of the transition from laminar to turbulent flow

• for Re > 2300, laminar conditions broke down and the flow was unstable
• for Re > 2800, the flow was stable but turbulent (rather than laminar)
3.1.4 An ideal fluid versus a real fluid
An ideal fluid is assumed to have no viscosity (thus both kinematic viscosity  = 0 and absolute viscos-
ity  = 0) and therefore the fluid flow has no shear stresses. In addition, an ideal fluid is also assumed to
be incompressible.

For a real fluid the effects of shear stresses in the fluid are considered. Fluid friction forces caused by
viscosity result from shear stresses that exist between individual streamlines, where they are flowing at
different velocities. These are called boundary layer effects.

3.1.5 One – , two – and three – dimensional flow

Consider the horizontal flow from left to right in Figure 3.4. One – dimensional flow occurs when the
variations in velocity and pressure transverse to the main flow direction are neglected (that is, both ver-
tically and horizontally into the page).

Figure 3.4 One – dimensional flow

For example in one – dimensional flow only the velocity u in the x – direction is considered to be a func-
tion of time

u = u(x, t)

while velocity in the y – direction is constant

v = constant

42
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

and velocity in the z – direction is constant

w = constant

The average velocity V as shown on the right – hand side of Figure 3.4 is used to designate the one –
dimensional flow velocity.

For two – dimensional flow, all particles are assumed to flow in parallel planes along identical paths in
each of these planes. Thus

x – direction: u = u(x, y, t)

y – direction: v = v(x, y, t)

z – direction: w = constant

In three – dimensional flow, all 3 velocity components are functions of x, y, z and t.

x – direction: u = u(x, y, z, t)

y – direction: v = v(x, y, z, t)

z – direction: w = w(x, y, z, t)

3.2 Dimensionless Parameters for Pipe Flow

3.2.1 Newton’s second law of motion
Newton’s second law of motion (Equation 2.5) applied to a fluid particle is

F = ma (3.9)

where

• F = sum of the forces acting on mass m (N or lbf)

• m = mass (kg or slugs)
• a = acceleration (m/s2 or ft/s2)
The forces include

• gravity – Fg
• pressure – Fp
• elasticity – Fe
• surface tension – Fst
• shear force (viscous force) – Fsf
The sum of the forces on the left – hand side of Equation 3.9 is balanced by the inertial force where the

inertial force = ma (3.10)

Thus it follows from Equation 3.10 that

F g + F p + F e + F st + F sf = ma (3.11)

The division of each term in this equation by one of the forces makes the equation dimensionless and
provides dimensionless parameters. For example, divide each term in Equation 3.11 by the viscous force
of Fsf as follows

43
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

F g F p F e F st F sf ma
------ + -------- + -------- + -------- + -------- = ------ (3.12)
F sf F sf F sf F sf F sf F sf

or

F g F p F e F st ma
------ + -------- + -------- + -------- + 1 = -------- = fn  Re (3.13)
F sf F sf F sf F sf F sf

where the Reynolds number Re in pipe flow on the right – hand side of Equation 3.13 is defined by
Equation 2.15.

The relative size of each dimensionless parameter relative to unity indicates its importance in
Equation 3.13. Many different dimensionless parameters may result. Consider pipe flow where the vis-
cosity is the most significant force.

The inertial force of a parcel of fluid is expressed as

3V 2 L
ma = L ---- = L  --- V = L V
2 2
(3.14)
T  T

where

• m = mass (kg or slugs)

• a = acceleration (m/s2 or ft/s2)
•  = fluid density (kg/m3 or slugs/ft3)
3
• L represents the volume of the fluid parcel (m3 or ft3)
• V = represents the velocity of fluid (m/s or ft/s)
• L represents a characteristic length (m or ft)
• T = represents time (s)
The viscous force is expressed as follows by using Newton’s law of viscosity

du V 2
F sf = A =    A =  ---- L = V L (3.15)
 d y L

where

• F sf  shear force (N or lbf)

•  = shear stress (N/m2 or lbf/ft2)

• A = area on which shear force acts (m2 or ft2)
•  absolute viscosity of the fluid (Pa – s or slugs/s – ft)

du
• the ratio ------ is the rate of angular deformation
dy
• V = represents the velocity of fluid (m/s or ft/s)
• L represents a characteristic length (m or ft)
2
• L represents the area on which the force acts (m2 or ft2)
Thus the ratio of the inertial force given by Equation 3.14 and the viscous force given by Equation 3.15
is

44
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

2 2
Inertia force- = V V L- = V
L - = ----------- L-
-------------------------------- ---------------- ------- (3.16)
Viscous force V L  

where

•  = fluid density (kg/m3 or slugs/ft3)

• V = represents the velocity of fluid (m/s or ft/s)
2
• L represents the area on which the force acts (m2 or ft2)
•  absolute viscosity of the fluid (Pa – s or slugs/s – ft)
• L represents a characteristic length (m or ft)
•   kinematic viscosity of the fluid (m2/ s or ft2/s )
This is a form of the Reynolds number as given by Equation 2.15, and is the significant parameter for
pipe flow. In pipe flow it is usual to take the characteristic length L to be the diameter of the pipe D.

3.2.2 The Buckingham  theorem

The purpose of the Buckingham  theorem is to be able to develop a force equation for a complex flow
situation based on knowledge of the pertinent quantities. The equation should be expressed in terms of
the least number of variables or dimensionless groups of variables possible.

The Buckingham theorem is stated as follows

If a general function of n independent variables a 1 a 2  a n can be represented by

f 1  a 1 a 2  a n  = 0

and if those n variables can be expressed in terms of m dimensional units (for exam-
ple, M, L, T for mass, length and time respectively), then the general equation above
can be expressed as (n – m) dimensionless  terms as follows

f 2   1  2   n – m  = 0

Each such  term contains m + 1 variables of which only one need be changed from
term to term.

Generic dimensions for variables are required to apply the Buckingham  theorem and some of the
more common units are given in Table 3.1.

An example of the application of the Buckingham  theorem to the determination of the relationship
between pipe flow variables is given in the next Chapter in Section 4.3.4. As an introduction to the deri-
vation in Chapter 4, let us briefly introduce the problem here. For flow in a pipe, it is required to deter-
mine a relationship between the boundary shear stress on the inside of the pipe and significant fluid
properties and pipe dimensions. For pipe flow, assume the following variables are important and that the
form of the following relationship is required

F 1   o   V  D   = 0 (3.17)

where

•  o = shear stress at the boundary of the inside of the pipe (N/m2 or lb/ft2)
•  = density of the fluid (kg/m3 or slugs/ft3)
•  = absolute viscosity of the fluid (Pa – s or slugs/s – ft)

45
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

• V = flow velocity (m/s or ft/s)

• D = pipe diameter (m or ft)
•  = roughness height for the inside of the pipe (m or ft – usually mm or in. are used – be
careful to make sure the units are consistent)
Thus expressing Equation 3.17 in terms of the shear stress gives

 o = F 2    V  D   (3.18)

The Buckingham  theorem is applied to a detailed example in Section 4.3.4. Dimensional analysis
using the Buckingham  theorem cannot produce definitive mathematical relationships (that is, the
actual form of F1 or F2 above) between the variables, but it does indicate significant combinations of
variables.

In Section 4.3.4, the dimensionless numbers derived are

VL o 
----------- or Re , ---------
- , and ---- .
 V
2 D

The relationship in Equation 3.18, is expressed as


 o =   Re ----  V
2
(3.19)
 D

where

•  o = shear stress at the boundary of the inside of the pipe (N/m2 or lbf/ft2)

- = a function that depends on the two parameters in parentheses

•   Re ---
 D
• Re = Reynolds number (dimensionless)

• - = relative roughness (dimensionless)

---
D
•  = density of fluid (N/m3 or slugs/ft3)
• V = flow velocity (m/s or ft/s)
Thus dimensional analysis may be used to determine the relationships and dimensionless variables that
are important in pipe flow in the example outlined above and also for other flow situations.

Table 3.1 Dimensions of various quantities

Characteristic Symbol Dimension
length L L
area A L2
volume V L3
Quantity Symbol Units
time t T
velocity V L/T
acceleration a L/T2
flow or discharge Q L3/T
absolute viscosity  M/(LT)

46
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Table 3.1 Dimensions of various quantities

kinematic viscosity  L2/T
mass m M
force F ML/T2
density  M/L3
specific weight  M/(L2T2)
surface tension  M/T2
modulus of elasticity E M/(LT2)
bulk modulus of elasticity K M/(LT2)
pressure p M/(LT2)
momentum and impulse mv, I ML/T
energy and work E, W ML2/T2
power P ML2/T3

3.3 The Continuity Equation

3.3.1 Assumptions for the analysis of pipe flow
The following assumptions are made in deriving the continuity equation for flow in pipes. These include

• the pipe is flowing full, that is, there must not be any free surface in the pipeline
• the flow is assumed to be one – dimensional flow for application of the continuity equation.
As a consequence, the average flow velocity at the section and the center – line pressure is
used
• the flow is treated as incompressible and thus the fluid density  is assumed to be a constant
To consider the continuity equation for pipe flow the concepts of a streamline, a streamtube, a control
volume and a control surface must be introduced firstly.

3.3.2 Streamlines and streamtubes

A streamline is a continuous line drawn through the fluid so that it has the direction of the velocity vec-
tor at every point. No flow may cross a streamline. Solid boundaries are streamlines. A streamtube is the
tube made by all streamlines passing through a small, closed curve as shown in Figure 3.5.

Figure 3.5 Streamtube in a fluid (bundle of streamlines)

3.3.3 Control volumes

A control volume is a region in space and its boundary is a control surface. The dashed line in Figure 3.6
is the control surface. The transfer across the boundaries of the control volume by mass and momentum
flux is important in the development of the continuity and force – momentum flux equations respec-
tively.

47
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.6 Control volume for the derivation of the continuity equation

3.3.4 Mass conservation

For the control volume as shown in Figure 3.6, the law of conservation of mass is stated as follows

dm
= 0 (3.20)
dt

Thus there is conservation of mass for the system and no change in mass within the control volume with
time. The mass conservation equation is written for the control volume as follows

mass entering – mass leaving = gain or loss of mass inside (3.21)

This is the continuity equation for unsteady flow. For steady flow there is no accumulation or degrada-
tion of mass within the control volume with time and thus

mass in = mass out (3.22)

Apply this principle to a streamtube in Figure 3.7 as follows

mass in per unit time =  1 V 1 A 1

mass out per unit time =  2 V 2 A 2

Figure 3.7 Streamtube for the derivation of the continuity equation

Thus

 1 V 1 A 1 =  2 V 2 A 2 = constant (3.23)

Now the mean velocity of flow over the entire pipe at a section is used. This is referred to as one –
dimensional flow as was seen in an earlier section (Section 3.1.5), and as shown in Figure 3.8.

48
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.8 Mean velocity across a pipe

Thus the conservation of mass for steady state flow in a pipeline becomes

1 V 1 A1 = 2 V 2 A2 (3.24)

CONSERVATION OF MASS FLOW RATE

where

•  1 ,  2 = densities of the fluid at sections 1 and 2 (kg/m3 or slugs/ft3)

• V 1 , V 2 = average velocities at section 1 and 2 (m/s or ft/s)

• A 1 , A 2 = cross – sectional areas of flow normal to the flow direction at section 1 and 2 (m2
or ft2)
For incompressible flow assume the density is constant or

1 = 2 (3.25)

Thus the volumetric flow or discharge Q is

Q = V 1 A1 = V 2 A2 (3.26)

CONTINUITY EQUATION FOR FLUID FLOW

where

• Q = flow or volume per second (m3/s or ft3/s)

• V 1 ,V 2 = average velocities at section 1 and 2 (m/s or ft/s)
• A 1 , A 2 = cross – sectional areas of flow normal to the flow direction at section 1 and 2 (m2
or ft2)
The continuity of flow equation of Equation 3.26 is expressed in terms of a product of the average
velocity and the cross – sectional area normal to the direction of the flow. In later Chapters, velocities are
assumed to refer to a mean velocity over the cross – section of the pipe.

Thus if the discharge is given then the velocity of flow can be calculated as:

Q
V = ---- (3.27)
A

VELOCITY OF FLUID COMPUTED FROM THE FLOW

49
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

where

• V = average velocity (m/s or ft/s)

• Q = flow or volume per second (m3/s or ft3/s)
• A = cross – sectional area of flow normal to the flow direction (m2 or ft2)

Example 3.1 Consider the flow situation in Figure 3.9. The upstream diameter is D1 = 0.2 m with flow velocity of
V1 = 2.0 m/s. The downstream pipe diameter is D2 = 0.06 m.

(i) Compute the velocity of flow in pipe 2

(ii) Compute the flow

Figure 3.9 Abrupt pipe contraction for applying the continuity equation

Answer – (i) Compute the velocity of flow in pipe 2

The continuity equation from Equation 3.26 is applied between sections 1 and 2

Q = V 1 A1 = V 2 A2

Rearranging gives

  D 1 2
A  -------------  D1 
2
4 -
V2 =  -----1- V 1 =  ------------ V 1 =  --------
- V 1
A   2
 D 2 2
2  D2 
 ------------
4 
-

Thus

 0.2 2   0.2 2 
V 2 =  ------------ V 1 =  ------------  2.0  = 22.22 m/s
2
 0.06   0.06 2

___________________________________________________________________________________
Answer – (ii) Compute the flow.

The flow is

2
 D1 2
  0.2  - = 0.0628 m 3 /s = 62.8 L/s
Q = V 1 A 1 =  2.0  ------------
- =  2.0  -----------------
4 4

___________________________________________________________________________________

3.4 The Momentum Equation

The momentum equation is introduced at this stage. It is needed for consideration of head losses due to
frictional losses in the pipeline in the energy equation in Section 3.5.

3.4.1 Momentum in solid mechanics versus momentum flux in hydraulics

In solid mechanics, the conservation of momentum is defined as follows

50
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

;Now as an example, consider a block hitting a wall. It is stopped in a finite time and only a single object
is involved. Applying the impulse – momentum change equation leads to

FT = mv (3.28)

IMPULSE – MOMENTUM CHANGE EQUATION

where

• F = force vector required to stop the block when it hits the wall (N or lbf)
• T = time duration to decelerate the block to a velocity of zero (s)
• m = mass of block (kg or slugs)
• v = velocity vector of block just prior to impact (m/s or ft/s)

Figure 3.10 Block hitting a wall, impulse – momentum relationship

In fluid mechanics, the focus is not on a finite mass. Instead there is a continuous stream of fluid. Take
Newton’s second law of motion in Equation 2.5 and rewrite in terms of velocity and time

F = ma (3.29)

v gives
and thus the acceleration a with the ratio ---
T

v = m
F = m --- ---- v (3.30)
T T

v
The --- term is the mass/unit time of fluid flowing. The right – hand side of Equation 3.30 becomes
T

 mass per unit time v = QV = momentum flux (3.31)

where the volumetric flow rate as defined by Equation 3.26 is

volume
Q = ------------------ (3.32)
time

Now, if a jet of water hits a wall, the action is a continuous stream and it is not completed in a given
period of time, as it is in solid mechanics.

3.4.2 Force – linear momentum flux equation

In Newton’s second law of motion, the resultant force F causes an acceleration a of the mass m in the
direction of the force. Another way to express Newton’s second law for fluid motion is

d
F =  mv  (3.33)
dt

51
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

where

• F = the resultant of all external forces acting on the system, including body forces such as
gravity (N or lbf)
• m = constant mass of the system (kg or slugs)
• v = velocity of the center of mass of the system (m/s or ft/s)
Surface forces include pressure forces and shear forces (due to friction losses).

In words, the principle of conservation of momentum states

“the resultant force acting on a control volume is equal to the time rate of increase of linear momentum
within the control volume plus the net efflux of linear momentum from the control volume”

3.4.3 Derivation of the force – momentum flux equation

Consider a streamtube as shown in Figure 3.11. Applying Newton’s second law in the x – direction gives

F x = d  mV x  (3.34)
dt

Consider a flow of fluid during a specified time t . The mass of fluid flowing from section 1 to
section 2 is

m = Qt (3.35)

Figure 3.11 Streamtube for the derivation of the momentum equation

The force equation now becomes

V
 F x = Qt  ----------
x
= QV x (3.36)
t 

and

 F x = Q  V x2 – V x1  = QV x2 – QV x1 (3.37)

FORCE – MOMENTUM FLUX EQUATION FOR FLUID FLOW

Equation 3.37 assumes that there is only one inflow stream into the control volume and only one out-
flow stream from the control volume. Another way of writing Equation 3.37 is

 F x pressure +  F x gravity +  F x shear + R x = QV x2 – QV x1 (3.38)

where

52
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

• F x pressure = sum of pressure forces resolved in the x – direction acting on the control
volume (N or lbf)
• F x gravity = sum of gravity forces resolved in the x – direction acting on the control vol-
ume (N or lbf)
• F x shear = sum of shear forces resolved in the x – direction acting on the control volume
(N or lbf)
• R x = reaction on the control volume to keep it in place (N or lbf)
•  = density of the fluid (kg/m3 or slugs/ft3)
• V x , V x = average velocities at section 1 and 2 resolved into the x-direction (m/s or ft/s)
1 2

• Q = discharge (m3/s or ft3/s)

The force – momentum flux equation may be stated in words as follows

sum of forces = momentum flux leaving the control volume (the QV x term)
2
– momentum flux entering the control volume (the QV x term 
1

Similarly for the other coordinate directions the force – momentum flux equation are as follows. In the
y – direction

F y = Q  V y – V y  = QV y – QV y (3.39)

2 1 2

or

F y pressure + F y gravity + F y shear + R y = QV y – QV y (3.40)

2

In the z – direction

F z = Q  V z – V z  = QV z – QV z (3.41)

2 1 2 1

or

F z pressure + F z gravity + F z shear + R z = QV z – QV z (3.42)

2 1

When applying the momentum equation, it is necessary to take account of the following aspects

• the direction for positive forces and positive velocities must be selected in advance
• signs must be observed very carefully
• the term Q is a scalar and it is the entire Q that reaches the body that supplies the force
• the force – momentum flux equation is a vector equation
In the momentum equation, the direction of the mass flow rates (Q)s is not important. The value Q is
essentially a scalar value and represents a rate of mass transfer from one region to another.

3.4.4 Force – momentum flux equation for more than one inflow
Example 3.2 Consider a 600 mm diameter pipe carrying water flowing at 900 L/s that has a 90° bend in the hori-
zontal plane. Assume the loss of head in the bend is negligible and the pressure at the entrance is 300 kPa. Determine
the resultant force R by the water on the bend. Assume the water flowing is at a temperature of 15°C.

Answer – Computation of the resultant force on a bend

The free body diagram in Figure 3.12 indicates the static and dynamic forces acting on the bend. Let the total resultant
force R to hold the bend in place be made up of two components including R x and R y as shown in Figure 3.12.

53
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.12 Force on a bend – momentum flux example

At 15°C the density of water is 999.1 kg/m3.

The area of pipe is

2 2
D   0.6  2
A = ---------- = ------------------ = 0.2827 m
4 4

Thus

2
A 1 = A 2 = 0.2827 m

The velocity of flow from the continuity equation is

Q 0.9
V = ---- = ---------------- = 3.18 m/s
A 0.2827

Thus

V 1 = 3.18 m/s

and

V 2 = – 3.18 m/s

taking into account the direction of velocities relative to the assumed positive x and y – directions.

Apply the force – momentum flux of Equation 3.37 to the bend in Figure 3.13 in the x – direction as follows

 F x = QV x – QV x
2 1

or from Equation 3.38

F x pressure + F x gravity + F x shear + R x = QV x – QV x (3.43)

2 1

At section 1, the pressure force is only in the x – direction so it follows that p 1 x A 1 = p 1 A 1. The pressure force at section 2 has no
component in the x – direction and is thus zero or p 2 x A 2 = 0. Recall that the bend is in the horizontal plane and hence the gravity

force is zero, thus  F x gravity = 0.

The shear stress is also zero, thus  F x shear = 0.0 as it is assumed that there is negligible head loss. Finally at section 1, all of
the velocity is in the x – direction and thus V 1 x = V 1 . In addition there is no velocity component in the x – direction at section 2,
thus V 2 x = 0. Thus substituting these values into Equation 3.43 gives

p 1 A 1 – R x = 0 – QV 1

Thus

R x = p 1 A 1 + QV 1

R x = 300 000  0.2827  + 999.1  0.9   3.18  = 84 810 + 2859 = 87 669 N = 87.7 kN

R x = 87.7 kN

The magnitudes of the velocity and pressure are the same at section 2 and at section 1.

Now apply the force – momentum flux of Equation 3.40 to the bend in Figure 3.13 in the y – direction as follows

54
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Fy = QV y – QV y

2 1

or

F y pressure + F y gravity + F y shear + R y = QV y – QV y (3.44)

2 1

At section 1, the pressure force has no component in the y – direction and is thus zero or p 1 y A 1 = 0. The pressure
force at section 2 is only in the y – direction so it follows that p 2 y A 2 = p 2 A 2. Recall that the bend is in the horizontal
plane and hence the gravity force is zero, thus F y gravity = 0. The shear stress is also zero, thus F y shear = 0 as it
is assumed that there is negligible head loss. Finally at section 1, there is no velocity component in the y – direction,
thus V 1 y = 0. In addition, all of the velocity is in the y – direction at section 2 and thus V 2 y = V 2. Thus substituting
these values into Equation 3.44 gives

p 2 A 2 – R y = QV 2 – 0

or

R y = p 2 A 2 – QV 2

Note that V 2 = – 3.18 m/s because it is in the negative y – direction.

R y = 300 000  0.2827  – 999.1  0.9   – 3.18  = 84 810 + 2859 = 87 669 N = 87 669 N = 87.7 kN

R y = 87.7 kN

The resultant force on the bend is obtained from the Pythagoras theorem

2 2 2
R = Rx + R y

2 2 2 2
R = Rx + R y =  87.7  +  87.7  = 124.0 kN

R = 124.0 kN

Thus the resultant force to hold the bend in place is 124 kN.

___________________________________________________________________________________

A more general form of the force – momentum flux equation is needed if there is more than one inflow
to or outflow from the control volume. The more general form of the equation is

NOUT NIN
F x =   QV x2  i –   QV x1  j (3.45)
i=1 j=1

GENERAL FORM OF FORCE – MOMENTUM FLUX EQUATION

where

• F x = sum of forces resolved in the x – direction acting on the control volume (N or lbf)
• NOUT = number of outflow streams from the control volume
• QV x , QV x = momentum fluxes resolved in the x-direction at sections j and i respec-
1 2

tively (N or lbf)
• NIN = number of inflow streams into the control volume
3.5 Energy Conservation in Pipe Flow
3.5.1 Assumptions
The following assumptions are made in deriving the energy equation for flow in pipes. These include

• The flow is steady. This is represented by

55
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

V
= 0
t

• the pipe is flowing full (that is — there is no free surface in the pipeline)
• one – dimensional flow is assumed (constant velocity and pressure across a cross – section)
• the flow is frictionless for the Bernoulli equation (no losses are included)
• the energy equation (that includes losses) is applied along a streamline
• the flow is incompressible (that is — the density is constant or  = constant)
• the kinetic energy correction factor is taken as  = 1.0 for turbulent flow (a more detailed
discussion of the kinetic energy correction factor is given later on in Section 3.8)
• the velocity head is usually small and thus is often neglected (defined in Section 3.5.2)
3.5.2 Forms of energy
A fluid particle flowing in a pipeline has three forms of energy

• potential energy
• pressure energy
• kinetic energy
The amount of energy possessed by water at any location in a water system may be measured in terms of
head.

3.5.3 Units of head

In civil engineering applications, energy head is expressed as a vertical distance of water height that is
measured in units of meters or feet. Units of head are

• [L] or length
• [F – L/F] — in terms of energy per unit weight where F stands for force
Note that the numerator of energy is the product of a force times a distance, while the denominator of
weight has the units of force. By convention, the force is cancelled in the numerator and denominator
leaving [L] as the units of head.

3.5.4 Potential energy or elevation head

Potential energy measures the amount of energy that water possesses because of its elevation. The
potential energy of a fluid particle is referred to as the elevation head. The elevation head (m or ft) is
represented by the variable

z (3.46)

ELEVATION HEAD

It is measured as a vertical distance from some horizontal reference line (datum) or bench mark eleva-
tion (such as sea level) to the point of interest in the water system.

Consider an element of weight of fluid of mg as shown in Figure 3.13 (Douglas et al. 1979). The poten-
tial energy is

potential energy = mgz

This may also be interpreted as the work required to lift the weight of mg vertically through a distance of
z. The potential energy per unit weight is

mgz
potential energy per unit weight = ---------- = z (3.47)
mg

56
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.13 Energy of a flowing fluid

mgz
The units of elevation head based on Equation 3.47 - the ---------- term are
mg

N–m
 z  = ------------ = m
N

or

ft–lb
 z  = ----------- = ft
lb

3.5.5 Kinetic energy or velocity head

Kinetic energy measures the amount of energy in water due to its velocity or motion. The kinetic energy
of a fluid particle is called the velocity head. The velocity head (m or ft) is represented by the quantity

2
V
------ (3.48)
2g

VELOCITY HEAD

The kinetic energy is the second form of energy associated with a flowing fluid and is a result of the
fluid’s velocity. Consider the fluid element of weight mg as shown in Figure 3.13. The kinetic energy is

kinetic energy = 1--- mV

2
2

The kinetic energy per unit weight is

1--- mV 2
2
2 V
kinetic energy per unit weight = --------------- = ------ (3.49)
mg 2g

1--- mV 2
2
The units of the velocity head based on Equation 3.49 - the --------------- term and are identical to those for
mg
elevation head

2 2
V N–m V ft–lb
------ = ------------ = m or ------ = ----------- = ft (3.50)
2g N 2g lb

57
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

3.5.6 Pressure energy or pressure head

Pressure energy measures the amount of energy in water due to the water pressure. The pressure energy
of a fluid particle is referred to as the pressure head. The pressure head (m or ft) is represented by the
variable

--p- (3.51)


PRESSURE HEAD

It is measured by the vertical distance above the measurement point (usually the centerline of a pipe-
line). Water rises in an open piezometer tube to the level of the pressure head. This level is also called
the hydraulic grade line (defined in Section 3.7).

The pressure energy of the fluid mass is the third form of energy considered. This pressure energy is the
energy of a fluid when flowing under pressure as part of a continuously maintained stream. The pressure
energy is sometimes referred to as potential energy in transit. Fluid flow is now compared to motion that
occurs solid mechanics. As a solid body falls, there is free conversion between potential energy and
kinetic energy. In contrast, in fluid flow in a pipe (of constant diameter) the velocity remains constant.
This is shown by the constant velocity head in Figure 3.14. Thus it is not possible for the potential
energy to be converted into additional kinetic energy as fluid flows along a pipe. Instead, the surplus
energy appears as a pressure increase or decrease (depending on the slope of the pipeline).

Figure 3.14 EGL and HGL for constant velocity head along a pipe

To understand pressure energy, consider a steadily flowing stream of fluid that does work because of its
pressure (see Figure 3.13). The pressure generates a force pA at a section. A volume of liquid V M passes
a section in a set amount of time. Density is defined as the ratio of mass to volume

m
 = ---------
VM

The volume is expressed as

weight mg m
V M = ------------------------------------ = ------- = ---- (3.52)
specific weight g 

The distance traveled by a fluid particle in Figure 3.13 is

volume V M m
distance = ------------------ = --------
- = -------
area A A

The work done is

58
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

m- = -------
pm-
work done = force  distance = p A  ------
A 

This term is the flow or displacement energy content of the fluid — that is the energy or work required to
push a mass of fluid across a boundary of a system (Olson and Wright 1990). Thus

pm
--------
 p
work done per unit weight = -------- = ------ (3.53)
mg g

or

p
work done per unit weight = ---


pm
--------

Again the units of pressure head based on Equation 3.53 - the -------- term and are consistent with those
mg
for elevation head and velocity head as

p N–m p ft–lb
--- = ------------ = m or --- = ----------- = ft
 N  lb

3.5.7 Mechanical energy and heat energy

Two other forms of energy may be 'added to' or 'subtracted from' the flow

• mechanical energy Em is added or subtracted from the flow system by equipment such as
pumps or turbines
• heat energy EH is added or subtracted from the flow system by equipment such as heat
exchangers or energy dissipation (for example, friction loss at pipe walls). These forms of
energy are dealt with later on when the energy equation is considered.
These two forms of energy are neglected in the next section, when the Bernoulli equation is considered.

3.5.8 Euler’s equation of motion along a streamline

Euler’s equation of motion is derived in differential form along a streamline (Streeter et al. 1998)

dp
g dz + ------ + v dv = 0 (3.54)


where

• g = gravitational acceleration (m/s2 or ft/s2)

• z = elevation head of a fluid particle (m or ft)
• p = pressure (N/m2 of lbf/ft2)
•  = density of the fluid (kg/m3 or slugs/ft3)
• v = velocity of a fluid particle measured tangentially at a point on the streamline (m/s or
ft/s)
Integration of Equation 3.54 for a constant density gives the Bernoulli equation

2
p v
gz + --- + ----- = constant (3.55)
 2

59
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

The constant of integration varies from one streamline to another but remains constant

• along a streamline
• for steady flow
• for frictionless flow
• for incompressible flow
These four assumptions must be satisfied for Equation 3.55 to apply.

3.5.9 The Bernoulli equation (excludes losses)

The Bernoulli equation is a special form of the law of conservation of energy. It assumes the flow is
one – dimensional, steady, incompressible and frictionless (no losses). Dividing Equation 3.55 by gravi-
tational acceleration g gives another form of the Bernoulli equation.

2
p- + -----
v - = constant
z + ----- (3.56)
g 2g

These three forms of energy are referred to as available energy. Now if Equation 3.56 is applied to a
fluid particle at section 1 and a fluid particle at section 2 in a pipeline system such as shown in
Figure 3.15 (assuming no losses along the section of pipeline), the result is

2 2
p V1 p V2
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- (3.57)
 2g  2g

THE BERNOULLI EQUATION (LOSSES ARE EXCLUDED)

where

• z 1 = elevation of the fluid particle at point 1 (m or ft)

• p 1 = pressure at point 1 (N/m2 [or Pa] or lbf/ft2)
•  = specific weight of fluid (N/m3 or lbf/ft3)
• V 1 = velocity of flow at point 1 (m/s or ft/s)
• g = gravitational acceleration (m/s2 or ft/s2)
• z 2 = elevation of the fluid particle at point 2 (m or ft)
• p 2 = pressure at point 2 (N/m2 [or Pa] or lbf/ft2)
• V 2 = velocity of flow at point 2 (m/s or ft/s)

Figure 3.15 Bernoulli flow in a pipeline (no losses). Note the pipeline could be sloping.

Another form of Equation 3.57 is

60
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

H1 = H2 (3.58)

where the value H is referred to as the Bernoulli constant (Brater and King 1976) with

2
p V1
H 1 = z 1 + -----1- + --------
- (3.59)
 2g

2
p V2
H2 = z 2 + -----2- + --------
- (3.60)
 2g

In the following example, the Bernoulli equation is applied for determining the velocity of flow from a
reservoir.

Example 3.3 Calculate the velocity of flow from a very short flow pipe leading from a reservoir. Friction in the pipe
may be ignored. The reservoir has a water surface elevation of 4.0 m above the centerline of the pipe. Assume a kine-
–6
matic viscosity of  = 1.004 10 m2/s and a density of 998.2 kg/m3 for water at 20°C.

Answer – Computation of the velocity.

In the reservoir shown in Figure 3.16, the energy content for a fluid particle is the same everywhere irrespective of the
vertical locationThus the selection of point 1 in the reservoir, from which to commence writing the energy equation
from, is arbitrary. Normally, the Bernoulli equation must be applied along a streamline. This example violates this
assumption, when the reservoir is elected as the starting point.

The sum of elevation head, pressure head and velocity head is a constant for the Bernoulli equation in Equation 3.57.
Apply the Bernoulli equation between a fluid particle on the surface of the reservoir (point 1) and a fluid particle at the
level of the centerline of the pipe but still in the reservoir (point 1")

2 2
p V1 p 1'' V 1''
z 1 + -----1- + --------
- = z 1'' + -------
- + -----------
 2g  2g

In the reservoir the velocity is zero or V = 0 and it follows that

p p 1''
z 1 + -----1- + 0 = z 1'' + -------
-+0
 

Thus the sum of the elevation head and the pressure head is a constant in the reservoir

p p 1''
z 1 + -----1- = z 1'' + -------
-
 

Moving vertically downward from the surface of the reservoir a fluid particle has less potential energy per unit weight
z but correspondingly more pressure head energy per unit weight p/

Figure 3.16 Energy for a reservoir-pipeline system

The Bernoulli equation is now applied between two sections (that is, a fluid particle at 1 at the surface of the reservoir
and a fluid particle at 2 at the centerline of the pipeline just outside the exit of the pipeline)

2 2
p V1 p V2
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- (3.61)
 2g  2g

Both the pressure head and velocity head are zero at the surface of the reservoir. If the horizontal datum is taken as the
centerline of the pipe, then the elevation head and pressure head at section 2 are also zero.

61
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Hence

z 1 = 4.0 m

p
-----1- = 0.0


V 1 = 0.0

z 2 = 0.0

p
-----2- = 0.0


Thus substituting into Equation 3.61 gives

2
V2
4.0 + 0.0 + 0.0 = 0.0 + 0.0 + --------
-
2g

Solving for the velocity gives

V2 = 2 g  4.0  = 2  9.81 4 = 8.86 m/s

This is Torricelli’s Theorem.

V2 = 2 gH (3.62)

What if 1" was selected in the reservoir at the entrance to the pipeline rather than at the surface of the reservoir for
application of the Bernoulli equation?

Then

z 1'' = 0.0

p 1''
-------
- = 4.0 m


2
p 1'' V 1''
z 1'' + -------
- + ----------- = 0.0 + 4.0 + 0.0 = 4.0 m
 2g

Thus the velocity is identical to the first calculation above

V2 = 2  9.81 4 = 8.86 m/s

___________________________________________________________________________________

3.5.10 Application of the Bernoulli equation to a pitot tube

As shown in Figure 3.17, the velocity head is measured by insertion of a pitot tube in the flow. The flow
stagnates at the entrance to the pitot tube such that the fluid velocity is zero.

Figure 3.17 Pitot tube for measuring velocity head

Applying the Bernoulli equation (Equation 3.57) between point 1 and point 2 in Figure 3.17 (Streeter et
al. 1998) gives

2 2
p V1 p V2
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- (3.63)
 2g  2g

62
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

But z 1 = z 2 = 0 if the reference datum is taken at the centerline of the entrance to the pitot tube and
V 2 = 0 due to stagnation of the flow at the entrance of the pitot tube. The pressure head at section 1
equals the height of water above it

p
-----1- = h 0


The pressure head at the entrance to the pitot tube is the height of the water column in the pitot tube

p
-----2- = h 0 + h


By substituting these values into Equation 3.63 it follows that

2
V1
- = 0 +  h 0 + h  + 0
0 + h 0 + --------
2g

Thus the velocity in the flow is equal to the height the fluid rises in the pitot tube above the water surface
level

2
V1
--------- = h
2g

Practically, it is very difficult to read the height h above the free water surface.

3.5.11 Head or pressure losses in pipes

There are two types of losses in pipeline flow. These include boundary friction losses and minor losses.
Boundary resistance losses result from the shear force between the fluid and boundary materials as
shown in Figure 3.18. Details of head loss in pipes are given in Chapter 4.

For real fluid flow in a closed pipeline, energy is lost overcoming fluid frictional forces in the form of
viscous drag at the boundary and within the fluid flow. This energy loss is dissipated as heat. Friction
losses are distributed uniformly over long reaches of pipes (of the same diameter and material) and are
caused by fluid viscosity and the resulting shear stresses generated in the fluid. Pipeline flow with
boundary losses is analyzed as steady uniform flow. Uniform flow occurs in a pipe of constant diameter.

The shape of the boundary usually has an effect on losses as well. Minor losses result from recirculating
eddies. These are produced by localized geometry effects of fittings such as valves, bends, expansions,
contractions or transitions. Pipe fittings dissipate energy by producing localized large – scale turbulence.
These losses are called minor losses and are analyzed as steady non – uniform flow. Minor losses are dis-
cussed in Chapter 5.

63
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.18 Head loss in a pipe due to friction

3.5.12 The principle of conservation of energy

Energy may neither be created nor destroyed. The sum of the various forms of energy is a constant.
Energy may be converted or transformed from one form to another. Consider the control volume in
Figure 3.19.

Figure 3.19 Heat flow from a system and work done on a system

There are a number of different forms of energy in fluid flow, some of which have previously been intro-
duced

• mechanical energy
• potential energy per unit weight of fluid flowing or elevation head z —
Section 3.5.2
2
V
• kinetic energy per unit weight of fluid flowing or velocity head ------ — Section 3.5.2
2g
• thermal energy — these become the losses
• internal energy per unit weight of fluid flowing u — this is the form in which energy
is stored within a substance
• heat transferred from the fluid to the surroundings per unit weight of fluid flowing
qH as shown in Figure 3.19
• work
• shaft work being done on the system per unit weight of fluid flowing wS — either by
the energy transferred to the system from a pump (as shown in Figure 3.20) or
energy extracted from the system by a turbine
• flow or displacement work per unit weight of fluid flowing or the pressure head
p   — the energy or work associated with the pressure in the fluid to push a unit
weight of fluid across the boundary of a system — Section 3.5.2.

64
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.20 Head, HP, added to the fluid by a pump

The losses and shaft work terms are incorporated into the energy equation as follows

2 2
p V p V
z 1 + -----1- + ------1 + H P = z 2 + -----2- + ------2 + H T + Losses 1-2 (3.64)
 2g  2g

3.5.13 The energy equation (includes losses)

The energy equation is concerned with the mechanics of internal energy changes and the velocity and
pressure variations within the flow. The energy equation contains scalar quantities, for example, pres-
sure and magnitude of velocity. In contrast, both the continuity and momentum equations consider the
bulk characteristics of the flow or net results of the flow. The momentum equation contains vector quan-
tities. From Equation 3.64 the energy equation is

2 2
p V p V
z 1 + -----1- + ------1 + H = z 2 + -----2- + ------2 + H T + Losses 1-2 (3.65)
 2g P  2g

The full energy equation for full pipe flow including pumps, turbines, friction loss and minor losses
again applied between sections 1 and 2 is

2 2
p1 V 1 p V
- + ------ + H = z 2 + -----2- + ------2 + H T +  h f +  h L
1 + ----- (3.66)
 2g P  2g

where

• z 1 = elevation of the fluid particle at point 1 (m or ft)

• p 1 = pressure at point 1 (N/m2 [or Pa] or lbf/ft2)
•  = specific weight of fluid (N/m3 or lbf/ft3)
• V 1 = velocity of flow at point 1 (m/s or ft/s)
• g = gravitational acceleration (m/s2 or ft/s2)
• H P = pump head added (m or ft)
• z 2 = elevation of the fluid particle at point 2 (m or ft)
• p 2 = pressure at point 2 (N/m2 [or Pa] or lbf/ft2)
• V 2 = velocity of flow at point 2 (m/s or ft/s)
• H T = turbine head extracted (m or ft)

•  h f = sum of the friction losses along pipelines (m or ft)

•  h L = sum of the minor losses in the system due to bends, valves, etc. (m or ft)

65
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Now by substituting for Darcy – Weisbach fluid friction loss (see Chapter 4 – page 76 – Equation 4.3)
and minor losses (see Chapter 5 – page 132 – Equation 5.1) then Equation 3.66 becomes

2 2 2
p1 V 1 p2 V 2 fL V V
1 + -----

- + ------
2g
- + H = z 2 + -----

- + ------
2g
- + H T  D 2g +  K -----
+ -----
- ------
2g
(3.67)
P

ENERGY EQUATION FOR PIPELINE FLOW

where

• z = elevation head (m or ft)

• p   = pressure head (m or ft)
• V = velocity (m/s or ft/s)
• g = gravitational acceleration (m/s2 or ft/s2)
• HP = pump head added to flow (m or ft)
• HT = turbine head extracted from flow (m or ft)
• f = Darcy – Weisbach friction factor (dimensionless)
• L = length of pipe (m or ft)
• D = diameter of pipe (m or ft)
• K = minor loss coefficient (dimensionless)a
Note that the term friction factor can also be referred to as a resistance factor.

3.5.14 Pump and turbine power

Pump power is defined in Section 18.6.4 and turbine power is defined in Section 22.9.4.

3.5.15 The energy equation without shaft work or minor losses

The energy equation for no pump ( H P = 0), turbine ( H T = 0) or minor losses (  h L = 0) becomes

2 2
p V p V
z 1 + -----1- + ------1 = z 2 + -----2- + ------2 +  h f (3.68)
 2g  2g

The variation of energy along a pipeline is shown in Figure 3.21. Because the pipeline cross – section is
converging, the velocity head increases going from section A to section B. The additional velocity head
comes from a decrease in pressure head. The energy grade line (EGL) and hydraulic grade line (HGL)
are shown in Figure 3.21. These are defined in the next two sections.

The different types of energy described above are shown in a converging pipe example in Figure 3.21.
Elevation head, pressure head and velocity head are illustrated at two sections A and B. The following
changes have occurred

2 2
VA VB
• the velocity head ---------
- has increased to ---------
- as the conduit converges
2g 2g
p p
• the pressure head has reduced from ------A- to ------B
 
The HGL refers to the hydraulic grade line (see Section 3.7) and the EGL refers to the energy grade
line. The fact that the EGL drops between A and B indicates that there has been an energy loss due to
friction resistance of the walls of the pipe on the fluid. More details of head loss in pipes are given in
Chapter 4.

66
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.21 Variation of energy along a pipe of changing diameter

3.6 The Energy Grade Line (EGL)

The energy grade line is measured as a vertical distance of the sum of three forms of energy as defined
in Equation 3.57 and is measured vertically from an arbitrarily selected horizontal datum as

2
p V
EGL = z + --- + -------- (3.69)
 2g

DEFINITION OF THE ENERGY GRADE LINE

This is assumed to apply to a fluid particle located at the centerline of the pipe. The locus of the sum of
2
p V
the values of z + --- + -------- along the pipeline gives the EGL.
 2g

Losses or irreversibilities cause the EGL to drop in the direction of flow (Streeter et al. 1998) along the
pipeline. An example of both the energy grade line and the hydraulic grade line are shown in
Figure 3.22. The elevation head z is measured from the horizontal datum to the centerline of the pipe.
The head loss in the pipe in Section 3.5.12 is h f and is dealt with in detail in Section 3.5.13.

3.7 The Hydraulic Grade Line (HGL)

The hydraulic grade line (HGL) is defined as the pressure head p/height above the centerline of the
pipe where

• p = pressure in the pipe (N/m2 or lbf/ft2)

•  = specific weight of water (typically 9789 N/m3 for water at 20 C or 62.4 lbf/ft3)
If z = elevation head of a fluid particle at the center of the pipe, then the elevation of a point on the
hydraulic grade line above a horizontal datum is

HGL = z + --p- (3.70)



DEFINITION OF THE HYDRAULIC GRADE LINE

67
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Figure 3.22 EGL and HGL for flow between two reservoirs - velocity accounted for

The HGL is also referred to as the piezometric head. The piezometric surface is the equivalent height of
water rise that occurs if a piezometer (an open tube) was attached to the side of a pipe. The locus of the
sum of the values of piezometric heads of z + --p- along the pipelines gives the HGL.


For static water systems the HGL is always horizontal and is always the same height as the free water
surface.

Losses or irreversibilities cause the HGL to drop in the direction of flow (Streeter et al. 1998) although
the HGL may actually increase across an expansion if sufficient velocity head is converted to pressure
energy despite the lower energy in the direction of the flow.

Figure 3.18 shows the hydraulic grade line and the energy grade line based on Equation 3.66.

3.8 Kinetic Energy Correction Factor

A one – dimensional form of analysis is usually used for computing pipeline or closed conduit flow. The
average velocity designated as V or V is used for expressing the velocity head. The velocity at a particu-
2 2
V V
lar point on the velocity profile is designated as u It is necessary to correct -------  ------- to obtain the cor-
2g 2g
rect amount of kinetic energy of the flow.

The revised energy equation including kinetic energy correction factors becomes

2 2
p1 V p V
- +  1 --------1 + H = z 2 + -----2- +  2 ------2 + H T + h f + h L
1 + ----- (3.71)
 2g P  2g

For the derivation of an expression for  consider the fluid mass per unit time passing a cross – section
in Figure 3.23

mass = VA (3.72)

Figure 3.23 Mass flow through a pipeline

Consider the kinetic energy per unit time passing the cross – section

68
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

2
u
KE =   -----u d A (3.73)
A
2

where
• KE = kinetic energy of the fluid (N-m or lbf-ft)
•  = density of the fluid (kg/m3 or slugs/ft3)
• u = velocity in the velocity profile across the pipe (m/s or ft/s)
• A = area of pipe (m2 or ft2)

The true kinetic energy per unit time passing a cross – section based on an average velocity V is defined
as follows

1
KE =   --- mV 
2
2 

where
• KE = kinetic energy of the fluid (N-m or lbf-ft)
•  = kinetic energy correction coefficient (dimensionless)
1 2
• --- mV = kinetic energy based on the average velocity (N-m or lbf-ft)
2
But

m· = VA (3.74)

Equating the two expressions (in Equation 3.73 and Equation 3.74) for kinetic energy leads to

2
------- = 
  VA  V
3
---  u d A (3.75)
2 2A

or

3 3
V A =  u dA (3.76)
A

Thus for a constant density it follows that the kinetic energy correction factor is

1 - u3 d A
 = ---------- (3.77)
3
AV A

or

u- 3 d A
1  ---
 = ---  (3.78)
A AV

For a number of strips of area of equal width in a velocity profile, the integral is approximated (Olson
and Wright 1990) as

n
1
 =  ----------   u i
3
(3.79)
 3 
V n i=1

69
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

where

•  = kinetic energy correction coefficient (dimensionless)

• V = average velocity across the pipe cross-section (m2 or ft2)
• n = number of incremental areas making up the total area A
• ui = average velocity in each sub – area (m/s or ft/s)
For laminar flow the velocity distribution is parabolic (see Chapter 4) and from Equation 3.78 it follows
 = 2.0 (3.80)

KINETIC ENERGY CORRECTION FACTOR — LAMINAR FLOW

Figure 3.24 Parabolic velocity distribution for laminar flow

For turbulent flow the velocity profile is more uniform and thus

 = 1.01 to 1.10 (3.81)

KINETIC ENERGY CORRECTION FACTOR — TURBULENT FLOW

The kinetic energy correction factor for turbulent flow is usually neglected and  is assumed to be 1.0.
For Prandtl’s one – seventh power law for the velocity distribution in a pipe

y 1  7
u = u max  --- (3.82)
 R

for the velocity at a distance y from the wall in a pipe of radius R. The value for  in this case is 1.058
(Douglas et al. 1979).

3.8.1 Continuity of flow for a non – uniform velocity profile

For a non – uniform velocity distribution the flow is obtained by integrating the velocity variation over
the cross – sectional area as

Ao
Q =  udA (3.83)
0

where

• Q = discharge or flow (m3/s or ft3/s)

• u = instantaneous average velocity at a distance r from the center of the pipe (m/s or ft/s)
• Ao = total cross – sectional area of flow (m2 or ft2)

3.9 References
Brater, E. F. & King, H.W. (1976). Handbook of hydraulics for the solution of hydraulic engineering
problems. New York: McGraw-Hill.

Douglas, J.F., Gasiorek, J.M. and Swaffield, J.A. (1979) Engineering fluid mechanics. Pitman, London.

Olson, R. M., and Wright, S. J. (1990). Essentials of engineering fluid mechanics, Harper & Row.

70
 Chapter 3 - File: Chap3newB.fm July 6, 2022 Version 6

Skeat, W.O. and Dangerfield, B. J. ed. (1969). Manual of British water engineering practice. (4th ed.).
Cambridge: Published for the Institution of Water Engineers by Heffer.

Streeter, V.L., Wylie, E.B. and Bedford, K. (1998). Fluid mechanics. 9th Edition. McGraw – Hill.

71
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

CHAPTER 4 PIPE HEAD LOSS1

One of the most important parts of this book is the computation of fluid friction loss in pipes for turbu-
lent flow. This Chapter starts out by discussing features of pipes including pipe materials, wall thickness
of a pipe and the internal roughness of a pipe. Details of flow in pipelines are given, and in particular,
the velocity profiles and computation of friction factors. Two common methods for computing head loss
or pressure drop in water distribution systems are the Darcy – Weisbach head loss equation and the
Hazen – Williams equation. Both methods are described in this Chapter. A derivation of the Darcy –
Weisbach equation is presented based on using the momentum equation and dimensional analysis. The
concept of the viscous sublayer is also described and its importance to each of the flow regimes. A
dimensionless parameter is introduced that distinguishes between the three turbulent regimes. The
Moody diagram for determining the Darcy – Weisbach friction factor (or resitance factor) is discussed in
great detail in relation to the different possible flow regimes. These flow regimes include laminar flow,
smooth pipe turbulent flow, transitional turbulent flow and fully rough turbulent flow. In the opinion of
the author, the preferred method for computing the head loss in pipelines is the Darcy – Weisbach head
loss equation and fluid friction factor or resistance factor f. While information is provided in this Chap-
ter for the Hazen – Williams equation, engineers are cautioned to use the Hazen – Williams head loss
equation only within its range of validity. An assessment of the accuracy of the Hazen – Williams for-
mula over a range of flow conditions is discussed. Formula are derived to convert between the Hazen –
Williams C value and both the Darcy – Weisbach friction factor f and the roughness height  . Tables of
roughness height values  (for the Darcy – Weisbach head loss equation) for pipes of different materials
and condition are given as well as tables of Hazen – Williams C values for different material and pipe
condition — in Appendix C. The change in roughness height as a pipe ages is also discussed. The Chap-
ter concludes with a brief discussion of the Manning equation that is often used for head loss calculation
in penstocks and tunnels associated with hydroelectric conduits. Localized losses are called minor
losses. Details of these types of losses are presented in Chapter 5.

4.1 Features of Pipes

4.1.1 Diameter of a pipe
When referring to a pipe size, engineers often discuss a number of different diameters. The nominal
diameter of a pipe is usually considered to be the name of a particular pipe size. It often bears little
resemblance to the actual size of the pipe. The nominal pipe size is usually the approximate size of the
pipe; and it may be smaller or larger than the true internal diameter. The internal diameter of a pipe or
the ID is the measured average internal diameter of the pipe. It is an extremely important physical prop-
erty for modeling of flows in networks. The ID of a pipe is illustrated in Figure 4.1. Another term, refer-
ring to pipe size, is the outside diameter of the pipe or OD. This is defined as the internal diameter plus
twice the wall thickness.

An example of a commercially available pipe is Unplasticized PVC (uPVC) pipe available from the
Hardie Iplex Company. One product is HARDITUBE AS1477 (Australian Standard AS1477 —
Unplasticized PVC (uPVC) Pipes and Fittings for Pressure Applications) pressure pipe that is available
with either rubber ring or solvent weld joints. Each pipe segment has an effective length of 6 m.
Table 4.1 shows typical properties of Class 9 (0.9 MPa or 90 m head) uPVC pipe that complies with
AS1477.

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Adelaide,
Australia. Part of the book entitled Water Distribution Systems Engineering.

72
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Table 4.1 Pipe properties for uPVC HARDITUBE AS1477* pressure pipe

Nominal Mean outside Mean wall Mean inside D/e ratio based
pipe diameter thickness e diameter (ID) D on mean inside
diameter (OD) (mm) diameter
(mm) (mm)
(mm)
80 88.9 3.8 81.3 21.4
100 114.3 4.8 104.7 21.8
150 160.3 6.7 146.9 21.9
200 225.3 8.4 208.5 24.8
225 250.4 9.3 231.8 24.9
250 280.4 10.5 259.4 24.7
300 315.5 11.7 292.1 25.0
375 400.5 14.9 370.7 24.9

*Australian Standard AS1477 — Unplasticized PVC (uPVC) Pipes and Fittings for
Pressure Applications

4.1.2 Wall thickness

Traditionally, engineers designate the wall thickness of a pipe as e. It is important in determining the
pressure rating of the pipe, that is, what is the maximum permissible pressure in the pipeline. The wall
thickness of a pipe also determines the wave speed in the pipeline. The wave speed is important for
water hammer computations.

Figure 4.1 Wall thickness e of a pipe

A thin walled pipe is one where the ratio of the diameter to the wall thickness is greater than 25. This is
defined as

D-  25
--- (4.1)
e

DEFINITION OF A THIN – WALLED PIPE

where

• D = the pipe inside diameter (mm or in.)

• e = pipe wall thickness (mm or in.)
D- ratio is less than 25, then the pipe is classified as a thick walled pipe. This is defined
In contrast, if the ---
e
as

73
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

D-  25
---
e

Table 4.1 shows that all of the listed uPVC pipes of different diameters are thick – walled, although the
pipe diameters above 150 mm diameter are very close to being classified as thin – walled.

4.1.3 Internal pipe roughness

Often, the internal roughness of a pipe governs the resistance to flow and the subsequent pressure drop
of head loss along the pipe. The roughness of the pipe may be characterized by the average height of the
roughness projections on the inside of a pipe. Traditionally to represent the roughness height for a pipe,
engineers use the symbol epsilon  . The roughness height for the inside of commercially available pipes
is usually non – uniform. The average roughness height of the inside of the pipe is either calculated from

• laboratory testing to back – calculate roughness height values for the pipe
• by measuring the root mean square of the roughness elements by moving a small probe
across the interior of the pipes surface
The relative roughness is defined as the average roughness height of the inside of the pipe divided by the
internal diameter of the pipe as follows


---- (4.2)
D

RELATIVE ROUGHNESS

where

•  = average roughness height of the inside of the pipe (mm or in.)

• D = internal diameter of pipe (mm or in.)
Later in this Chapter, the Darcy – Weisbach fluid friction factor or resistance factor is introduced. It is
denoted by the symbol f. For hydraulicians, the important dimensionless numbers for pipe flow are the
Reynolds number Re and the relative roughness   D . These dimensionless numbers are derived using
a dimensional analysis. Both these dimensionless numbers are used for the determination of the Darcy –
Weisbach fluid friction factor f.

4.2 Pipe Wall Materials

Over the last 200 years, for the construction of water distribution systems, engineers have used many
different types of pipe materials. Examples are cast iron, wood stave, ductile iron, steel, PVC, uPVC and
polyethylene. The next sections give descriptions of some of these materials.

4.2.1 Cast iron pipe

In circa 1455, German engineers made use of a cast iron main for the first time in Siegerland, Germany,
where iron pipe was installed to provide water service to Dillenberg Castle (Carpenter 2010). This is
earliest authentic recorded use of cast iron. Records also indicate that in 1664, Louis XIV ordered the
construction of a 24-km (15-mile) pipeline to serve the city and fountains of Versailles, France (Carpen-
ter 2010). This is an outstanding example of an early cast iron main and this iron pipeline has provided
continuous service for more than 330 years.

Today, cast iron pipes are found in many networks. Cast iron is more corrosion resistant than steel but is
more expensive and more rigid. In Australia, cast iron was first used in the 1800s. By 1880 within Syd-
ney, engineers had constructed an extensive system of cast iron pipes. While in 1857, a cast iron main
was first laid from Yan Yean reservoir in Melbourne.

74
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

In the past, manufacturers made cast iron pipes using sand castings. All early pipes were cast horizon-
tally in sand moulds. In 1846, producers began to use an improved technique of vertical casting. How-
ever today, for the production of plain pipe, sand castings are rarely used. In 1914, manufacturers began
producing cast iron pipe using a process of centrifugal casting (Tubemakers Australia 1995). This pro-
cess enabled the production of pipes of long lengths. In 1921, engineers began to produce spun iron
pipes using water cooled steel moulds. In this process, molten metal is poured into a revolving water
cooled steel mould. The metal is held against the sides of the mould by centrifugal force. In Australia,
from 1929 to 1976, manufacturers used this technique to produce cast iron pipe. In 1976, manufacturers
converted to ductile iron pipe. Today, most pipes are centrifugally spun. Predominately, manufacturers
make ductile iron pipe and grey iron pipe. Sand castings are still used for “specials” such as bends,
tapers and flanges (Stephenson 1984).

4.2.2 Ductile iron cement mortar lined pipe

As mentioned above — in the 1960s, manufacturers switched from the production of cast iron pipes to
ductile iron pipes. The production process also introduced a cement mortar lining on the inside of the
pipe. The cement mortar lining protects the ductile iron from corrosion. To place the cement mortar into
the pipe, manufacturers use a process of spinning to place the cement mortar lining uniformly against
the inside of the pipe. Typically, the thickness of the cement mortar lining is usually about 10 mm.

4.2.3 Steel pipe

Steel is one of the most versatile materials available for pipe installation (Stephenson 1984). It is partic-
ularly useful for large diameter pipes. Steel walls are ductile yet have a high strength. Steel pipe is man-
ufactured in lengths of up to 10 m. Although it is heavy, fabricators can work with steel pipe relatively
easily. Usually for larger steel pipes, construction workers join steel pipe sections using welded joints.
These are extremely common and lead to the strongest possible type of joint.

For pipelines subject to higher pressures, engineers need to select a higher grade of steel. In contrast, for
lower pressures then lower grades of steel are more economical and are also easier to weld. Extra wall
thickness should be provided if resistance to external loads needs to be provided.

If the diameter of a steel pipe is less than 450 mm, it is defined as being a small bore steel pipe. For these
sizes, manufacturers are able to roll the steel without seams. A large bore steel pipe is 450 mm in diam-
eter or greater. Construction workers weld large bore pipes from steel plate. The welds are either hori-
zontal or spiral. Steel pipes are usually more expensive than concrete and plastic pipe for low pressure
applications.

4.2.4 Plastic pipes

In the 1960s there were many advances in plastic pipes (Stephenson 1984). The main types of plastic
pipe include

• uPVC (unplasticized polyvinylchloride) pipe

• Polyethylene. Polyethylene pipes are used for small bore pipes. It is usually more expensive
than uPVC for larger diameters.
• HDPE (high density polyethylene) pipe. Polyethylene pipe has better properties and is eas-
ier to work with than uPVC pipe. As a result, it is preferred for gas piping. The three cate-
gories of pipes above are classed as thermoplastics and are relatively easy to extrude when
heated.
• Glass reinforced plastic pipe (GRP or glass fibre and resin). Manufacturers developed GRP
to replace asbestos cement pipe. In the manufacturing process, the pipe composition uses an
inert filler such as sand. GRP has good corrosion resistance and is rigid, light and strong.
• ABS pipe. These pipes are rubber toughened plastics and are useful for flexible pipework.
Plastic pipe has a maximum working stress of 140 MPa. As a result, it cannot be used for larger diameter
pressure mains. Plastic pipes are usually jointed by solvents for fusion welding. This process is more
difficult for larger pipe diameters. Some plastic pipes are glued. Fittings are still usually restricted to
steel or cast iron. It is difficult to manufacture molded bends, tees and branches in uPVC and even more
difficult in polyethylene.

75
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Advantages of plastic pipe include

• plastic pipe is highly resistant to corrosive fluids. As a result, it may be used in difficult
locations including undersea outfalls and gas pipes.
• the internal roughness height  of plastic pipe is usually considerably smaller than for other
materials. Thus the inside of the pipe is smoother and there is less head loss for the same
flow.
• plastic pipe is less likely to be subject to encrustation than other pipe materials
• plastic pipe is easier to lay due to the fact that it is lighter than other materials
• other pipes may be lined with plastic pipe during pipe rehabilitation
• ground movement is accommodated more easily by plastic pipe
Disadvantages of plastic pipe include

• plastic pipe has limited strength and deteriorates with time

• flexibility of plastic pipe may cause distortion, buckling and collapse. Thus plastic pipe is
susceptible to damage.
• plastic pipe has a high coefficient of thermal expansion (although polyethylene pipe has a
lower coefficient than uPVC pipe). Upon significant cooling, the joints may be loosened.
• PVC pipe exhibits creep and thus the unplasticized form or uPVC is used for pipes to
reduce this occurrence
4.2.5 Asbestos cement pipes
In the past, water authorities commonly used asbestos cement pipes. Due to discovery of the dangers of
asbestos fibers, the use of asbestos as a material has essentially ceased. As mentioned previously, in
Australia, and in fact, worldwide, AC pipe has been replaced by glass reinforced plastic pipe (GRP) or
Hobas. Asbestos cement pipes were constructed from concrete and asbestos fibers (Stephenson 1984).
Joints are made with a sleeve fitted with rubber sealing rings. Asbestos cement pipe had the advantage
of being cheap, strong and corrosion resistant. Although they are able to resist relatively high tensile
stresses they are susceptible to shock damage. In a similar way to plastic pipe, asbestos cement pipe
could not be used for “special” fittings — instead cast iron was usually used.

4.2.6 Concrete pipes

Reinforced concrete (RC) or prestressed concrete is economical for larger pipe diameters. An advantage
over steel pipe is the extra wall thickness of concrete pipe that resists external buckling loads more eas-
ily. It is also corrosion resistant. The main weakness of concrete pipe is the joining and the difficulty in
making connections at a later time.

4.3 Darcy – Weisbach Pipeline Fluid Friction Loss Equation

We now consider the computation of the head loss in pipelines using the Darcy – Weisbach fluid friction
loss equation. Prior to the presentation of detailed explanations, first an overview or summary is pro-
vided.

4.3.1 A summary
The Darcy – Weisbach head loss equation expresses the head loss or drop in energy grade line (EGL) or
hydraulic grade line (HGL) along a pipe. These losses are generated by the shear stresses in the fluid and
the interaction of the boundary layer at the wall of the pipe.

In current practice, engineers commonly use the term friction to refer to loss of energy in pipe flow.
When I studied at Colorado State University in 1980, Professor Maurice Albertson suggested an alterna-
tive term of resistance. He argued strongly that the term fluid resistance should always be used rather
than fluid friction when referring to fluid flow; saying that the term friction should only be applied to
solids sliding against one another. By convention, the term friction is used in this book.

76
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

The head loss is

2
LV
h f = f ---- ------ (4.3)
D 2g

THE DARCY – WEISBACH HEAD LOSS EQUATION

where

• h f = the head loss along the pipe (m or ft)

• f = the Darcy – Weisbach friction factor or resistance factor (a function of Reynolds number
Re = VD/ and relative roughness /D – both dimensionless)
• L = pipe length (m or ft)
• D = pipe internal diameter (m/mm or ft/in. – depends on usage, for example in the /D
computation)
• V = the average velocity of flow in the pipe (m/s or ft/s)
• g = gravitational acceleration (m/s2 or ft/s2)
The head loss h f has the dimension of length and is expressed in terms of an energy loss per unit weight
or N – m/N (Streeter et al. 1998). The friction factor f is a dimensionless factor and its variation with Re

and ---- has been measured experimentally for turbulent flow.
D

An approximation of the schematic of a Moody diagram shown in Figure 4.2 defines the variation of the
fluid friction factor f with Reynolds number Re. There are four distinct regions of different types of flow
regimes including

• laminar flow
• smooth pipe turbulent flow
• transitional turbulent flow
• fully rough turbulent flow (complete turbulence regime)
If the flow pictures are geometrically and dynamically the same in every detail for two different pipe-
lines, then the fluid friction factors are the same. Recall from Equation 2.15 the definition of the Reyn-
olds number.

The fluid friction factor or resistance factor f depends on the flow regime as shown in Table 4.2.

A summary of the four regions is given.

1. Laminar flow region. This region is defined by a straight line on the left – hand side of the Moody dia-
gram in Figure 4.2 running diagonally from the top left corner down towards a Reynolds Re of 2,300.
Computation of the fluid friction factor or resistance factor f is analytic for a circular pipe as shown in
Section 4.4.7 and is given by

64-
f = ------ (4.4)
Re

In the next three regions the flow is turbulent. When the flow changes from laminar to turbulent the flow
may be unstable in the Reynolds number range of 2300 to 2800.

2. Smooth pipe turbulent flow region. This region is actually a single line and represents a lower bound
of the Moody diagram in Figure 4.2 running diagonally in a curve from the middle of the diagram (mov-
ing from top to bottom) about one – third the way in from the left side to the bottom right – hand corner
of the diagram. Sometimes this region is also referred to as hydraulically smooth flow. The analysis for
developing an expression for the fluid friction factor or resistance factor f is almost analytic, however, a
velocity distribution needs to be assumed. An outcome of the analysis is that the fluid friction factor is
only a function of the Reynolds number Re and is independent of the roughness height  as follows

77
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

f = Fn  Re  (4.5)

Equation 4.5 applies when the relative roughness approaches zero   D  0 .

Table 4.2 Different types of flow regime

Flow regime Reynolds number Fluid friction factor is a
range function of
Laminar flow Re  2300 64
f = Fn  Re  = -------
Re
Unstable flow Re = 2000 to 4000
Smooth pipe turbulent flow (or hydraulically Re  2800 f = Fn  Re 
smooth) — as long as roughness projections are
sufficiently buried in the viscous sublayer**
Transitional turbulent flow* 200D - Re
3000  Re  ------------- f = Fn  ---
 f D 

Fully rough turbulent flow (or hydraulically 200D 

rough) — the roughness projections must break up Re  ------------- f = Fn  ----
 f  D
the viscous sublayer*

*Also called the laminar sublayer ** See Equation 4.91.

3. Transitional turbulent flow region. This region is the set of curved lines in the turbulent region to the
left of the diagonal dashed line in the middle of the Moody diagram as indicated in Figure 4.2. In this
region the fluid friction factor depends both on the Reynolds number Re and the relative roughness
  D.


f = Fn  ---- Re (4.6)
D 

For a particular   D value, the friction factor f increases as the Reynolds number decreases in the tran-
sitional turbulent flow region as one moves from right to left across the Moody diagram.

4. Fully rough turbulent flow region. This region is to the right – hand side of the dashed line running
approximately diagonally across the Moody diagram (Figure 4.2). In this region the fluid friction factor
only depends on the relative roughness height   D . The approximately horizontal lines are a character-
istic of this region on the Moody diagram (Figure 4.2). The fluid friction factor f is independent of
Reynolds number Re. Thus

-
f = Fn  --- (4.7)
 D

In the fully rough turbulent flow region of the Moody diagram for a pipe of a particular diameter, the
size of the fluid friction factor f increases with an increase in relative roughness D for the pipe.

Typical values of equivalent roughness height  are given in Table 4.3 for different pipe materials and
conditions.

78
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Figure 4.2 The Moody diagram — the Darcy – Weisbach f versus Reynolds number

Table 4.3 Pipe roughness heights (Streeter et al. 1998; Jeppson 1976; SAE Manual 1969)
Material  
mm) (in.)
Riveted steel 0.9 to 9.0 0.036 to 0.36
Concrete 0.3 to 3.0 0.012 to 0.12
Wood stave 0.18 to 0.9 0.008 to 0.04
Cast iron 0.25 0.010
Galvanized iron 0.15 0.006
Asphalted cast iron 0.12 0.0048
Commercial steel or wrought iron 0.046 0.0018

79
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Table 4.3 Pipe roughness heights (Streeter et al. 1998; Jeppson 1976; SAE Manual 1969)
Drawn tubing 0.0015 0.00006
Smooth tubing 0.0013 0.00005
PVC 0.00021 0.000084

Comprehensive information on pipe interior surface roughnesses are given in Appendix C. Part of
Table C.2 in Appendix C is shown in Table 4.4.

Table 4.4 Examples of values of C in Hazen – Williams formula (based on Skeat and Dangerfield —
Manual of British Water Engineering Practice 1969, p.163 and 164) — entire list is given in
Table C.2

Values of C for pipes of diameter

Type of pipe
75 mm 150 mm 300 mm 600 mm 1200
(3 in.) (6 in.) (12 in.) (24 in.) mm
(48 in.)
Uncoated cast iron: smooth and new 121 125 130 132 134
Coated cast iron:
Smooth and new 129 133 138 140 141
30 years old:
Trend 1: slight attack 100 106 112 117 120

4.3.2 Historical development

Historically, Reynolds, Nikuradse, von Karman, Prandtl, Colebrook, White and Moody have been the
important researchers in the field of pipe flow. They have provided significant contributions to the eval-
uation of fluid friction factors. Since about the 1930s, there has been a coherent theory of pipe fluid fric-
tion.

Early experiments (circa 1850) on water flow in long straight cylindrical pipes showed the head loss h f
varied directly with the following quantities

• velocity head
2
V
h f  ------- (4.8)
2g

• pipe length
hf  L (4.9)

and varied inversely with

• pipe diameter

1-
h f  --- (4.10)
D

A proportionality constant referred to as the Darcy – Weisbach friction factor f was introduced as shown
in Equation 4.3.

In the early 20th century researchers observed that f also depended on

• roughness height of the pipe wall 

80
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

• diameter of the pipe D

• velocity of fluid V
• viscosity of the fluid flowing 
As mentioned earlier, the Moody diagram (Figure 4.2) is a plot of Darcy – Weisbach friction factor f ver-
- values.
sus Reynolds number Re for various relative roughness ---
D

4.3.3 Shear stress at the wall of the pipe

Consider flow in a horizontal prismatic pipe. For the pipe in Figure 4.3, the force – momentum flux
equation from Equation 3.37 is applied to a control volume of length L in the x – direction along the
centerline of the pipe as

 F x = QV 2 x – QV 1 x (4.11)

Figure 4.3 Momentum flux equation for steady uniform flow

Considering the control volume in Figure 4.3 (the dashed rectangular box), the significant forces for a
horizontal pipe are the pressure and shear forces. The gravity force for a horizontal pipe is zero in the x –
direction. The momentum influx and efflux are equal, and thus the right – hand side of Equation 4.11
becomes zero. The force – momentum flux equation becomes

 F x = p1 A1 – p2 A2 – 0 P = 0 (4.12)

where

• F x = sum of forces resolved in the x – direction acting on the control volume (N or lbf)
• p1 = pressure at section 1 (N/m2 or lbf/ft2)
• A1 = area of entry to the control volume at section 1 (m2 or ft2)
• p2 = pressure at section 2 (N/m2 or lbf/ft2)
• A2 = area of exit from the control volume at section 2 (m2 or ft2)
•  o = shear stress at the boundary (N/m2 or lbf/ft2)
• P = wetted perimeter or the area of the curved cylindrical part of the control volume (m2 or
ft2)

81
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Substituting for both the area and the wetted perimeter in terms of the diameter of the pipe in
Equation 4.12 gives

2 2
D D
p 1 ---------1- – p 2 ---------2- –  o  DL  = 0
4 4

Thus the shear stress at the wall becomes

D
 o =  p 1 – p 2  ---------- (4.13)
4L

Writing the energy equation between sections 1 and 2 in the pipeline in Figure 4.3 gives

2 2
p V p V2
z 1 + -----1- + ------1- = z 2 + -----2- + -------
- + Losses 1-2 (4.14)
 2g  2g

Recalling that for fluid friction loss only in the pipe (i.e. no minor losses)

Losses 1-2 = h f

The pipe is of constant diameter thus

V1 = V2

and the pipe is horizontal, thus

z1 = z2

Substituting into the energy equation of Equation 4.14 leads to

p p
-----1- = -----2- + h f (4.15)
 

Simplifying and solving for the head loss in Equation 4.15 along the pipe gives

p p
h f = -----1- – -----2- (4.16)
 

Thus combining Equation 4.13 and Equation 4.16 gives an expression for the boundary shear stress

D
 o = h f ---------- (4.17)
4L

4.3.4 Dimensional analysis of flow in pipelines

In this section, dimensional analysis using the Buckingham  theorem (presented in the previous Chap-
ter in Section 3.2.2) is used to determine the relationships and dimensionless variables that are important
in pipe flow. Specifically, a relationship between the boundary shear stress on the inside of the pipe and
significant fluid properties is determined. For a pipe, assume the following variables are important and
that the form of the following relationship is required

F 1   o   V  D L   = 0 (4.18)

82
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

where

•  o = shear stress at the boundary of the inside of the pipe (N/m2 or lbf/ft2)
•  = density of the fluid (kg/m3 or slugs/ft3)
•  = absolute viscosity of the fluid (Pa – s or slugs/s – ft)
• V = flow velocity (m/s or ft/s)
• D = pipe diameter (m or ft)
• L = length of pipe (m or ft)
•  = roughness height for the inside of the pipe (mm or in.- convert to m or ft for this appli-
cation)
The rearrangement of Equation 4.18 in terms of the shear stress gives

 o = F 2    V  D  L  (4.19)

Now the Buckingham  theorem is applied with n = 6 and m = 3, and thus n – m = 3. The result is
dimensionless groups 1, 2 and 3. The procedure for applying the Buckingham  theorem is to first
select the repeating variables. It is usual to select one variable associated with each of the following to
be repeating variables

• mass or M e.g.  
• length or L, e.g. L, D
• time or T, e.g. V, a

For this case, choose the repeating variables to be density of fluid, length of pipe and velocity of flow (,
L and V). The first dimensionless group 1 is obtained by combining the three repeating variables (each
to an unknown exponent) with one of the other variables. For example, for the dynamic or absolute vis-
cosity

a1 b1 c1
1 = V L   (4.20)

The dimensions of are

a1 b1 c1
L M M
  1  = --- L ------ ------- (4.21)
T L
3 LT

Now solving for a1, b1, and c1 to make dimensionless gives

0 0 0
 1  =  M   L   T  (4.22)

Equate the exponents of like terms in Equation 4.21 and Equation 4.22 as follows

• for T

– a 1 – 1 = 0 or a 1 = – 1

• for M

c 1 + 1 = 0 or c 1 = – 1

• for L

83
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

a 1 + b 1 – 3c 1 – 1 = 0

or

b 1 = 1 + 3c 1 – a 1 = 1 + 3  – 1  –  – 1  = – 1

Thus a 1 = – 1, b 1 = – 1, c 1 = – 1 and substituting into Equation 4.20 gives the first dimensionless
group as

–1 –1 –1 -
1 = V L   = ----------
VL

Another dimensionless group  is obtained by combining the three repeating variables (again each to an
unknown exponent) with the other as yet unused variable of boundary shear stress  o as

a2 b2 c2
2 = V L  o (4.23)

The dimensions of 2 are

a2 b2 c2
L M ML
  2  = --- L ------ ------------ (4.24)
T L
3
T L
2 2

and

0 0 0
 2  =  M   L   T  (4.25)

The units of o (force per unit area) are determined from Newton’s second law of motion as

F = ma

L
 ma  =  M  ------
2
T

The units for shear stress are now be determined by dividing the units of force by the area L2 as follows

F  ML 1 M
  o  = --------- = --------- ------ = ----------
2 2 2 2
L T L LT

Now solve for a2, b2, and c2 in Equation 4.24 and Equation 4.25 to make the second dimensionless group
2 dimensionless.
• for T

– a 2 – 2 = 0 or a 1 = – 2

• for M

c 2 + 1 = 0 or c 2 = – 1

84
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

• for L

a 2 + b 2 – 3c 2 – 1 = 0

or

b 2 = 1 + 3c 2 – a 2 = 1 + 3  – 1  –  – 2  = 0

Thus a 2 = – 2, b 2 = 0, c 2 = – 1 and it follows by substituting into Equation 4.23 that

–2 0 –1
2 = V L  o (4.26)

o
 2 = ---------
- (4.27)
2
V

The final dimensionless group  3 is taken as the following ratio (the relative roughness):


 3 = ---- (4.28)
D

A dimensionless number may be factored in a number of ways to provide various other dimensionless
numbers that are equally valid. Thus a new dimensionless number may be created as

1
 4 = -----
1

Thus

VL
 4 = -----------


which is a form of the Reynolds number Re.

The outcome of applying the Buckingham  theorem is three terms as follows

 o  
F 3   4  2  3  = F 3  Re ----------
- ---- = 0 (4.29)
V D
 2

where the 3 quantities inside the brackets are the dimensionless quantities.

Solving for the shear stress in Equation 4.29 gives


 o =   Re ---- V
2
(4.30)
 D

where

-
- is some function of the Reynolds number Re and the relative roughness ---
•   Re ---
 D D

85
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Thus  is the form of the Moody diagram (Figure 4.2). The fluid friction factor f depends primarily on
the pipe roughness (   D — the relative roughness) and to a lesser degree on

• velocity V
• pipe diameter D
• fluid viscosity 
4.3.5 Derivation of the Darcy – Weisbach fluid friction equation
Now, by equating the two above expressions for ofrom Equation 4.17 and Equation 4.30 gives

- =  h ---------
D
V   Re ---
2
 D f 4L- (4.31)

Thus rearranging for the head loss h f gives

2
 L V
h f = 8  Re ---- ------ -------- (4.32)
 D D 2g


By tradition, the quantity 8  Re ---- is put equal to a variable f and is denoted as the Darcy – Weisbach
 D
friction factor as follows.


f = 8  Re ---- (4.33)
 D

The shear stress at the pipe wall as given by Equation 4.30 is expressed as

f 2
 o = ---V (4.34)
8

SHEAR STRESS AT THE PIPE WALL

Now replace the length L with the length L in Equation 4.32. Thus combining Equation 4.32 and
Equation 4.33 gives the Darcy – Weisbach head loss equation in Equation 4.3.

To obtain the slope of the EGL or friction slope divide the left – hand side of Equation 4.3 by the length
and thus

h 2
f V = F  V  D   
f = -----f- = ------------- 1 (4.35)
L D2g

2
Thus the EGL slope is related directly as the velocity V , directly as the fluid friction factor f and
inversely as the diameter (1/D).

4.3.6 Non – circular conduits

For pipes of non – circular cross – section the following expression should be used for the equivalent
diameter

4A
D e = ------- (4.36)
P

86
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

where

• D e = equivalent diameter for a non-circular conduit (m or ft)

• A = cross – sectional area of flow (m2 or ft2)
• P = wetted perimeter of flow (m or ft)
4.3.7 Variation of head loss with velocity
Consider a situation where the flow in a pipeline is gradually increased from a velocity of zero. The
head loss h f between the two ends of the pipe is proportional to the velocity raised to an increasing
exponent starting at 1 and moving to 2 as shown in Figure 4.4

• h f  V for laminar flow, Re < 2300

h 
• f V for flow changing from laminar to turbulent flow (unstable with pockets of turbu-
lence), 2300 Re where the exponent varies between 1.0 and 1.7
1.7 2
• hf  V to V for transitional turbulent flow, Re 
2
• h f  V for fully rough turbulent flow

Figure 4.4 Variation of h f with velocity (L = 300 m, /D = 0.002)

87
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

4.4 Laminar Flow

For laminar flow, the Darcy – Weisbach fluid friction factor is

64 64
f = ------- = -------------- (4.37)
Re VD
 --------
  

FRICTION FACTOR FOR LAMINAR PIPE FLOW

For laminar flow the friction factor only depends on flow velocity V, diameter of pipe D and the kine-
matic viscosity of the fluid  . The friction factor for laminar flow is independent of the average rough-
ness height of the inside of the pipe A full derivation of Equation 4.37 is given in this section.

A significant advantage in the analysis of laminar flow (in contrast to turbulent flow) is that an analyti-
cal analysis is available. The velocity distribution for laminar pipeline flow is parabolic as shown below
in Figure 4.5.

4.4.1 Minimum possible Darcy – Weisbach f – value for laminar flow

The minimum Darcy – Weisbach friction factor in the normal range of laminar flow occurs when the
Reynolds number is 2300. Thus from Equation 4.37

64- = -----------
64 - = 0.028
f = ------
Re 2300

Consider the laminar friction factor for a Reynolds number of 1000. Thus

64- = -----------
64 - = 0.064
f = ------
Re 1000

4.4.2 Velocity variation across the pipe for laminar flow

The average velocity at the sections 1 and 2 in Figure 4.5 is such that

V1 = V2 (4.38)

with the pipe being of constant diameter.

The force – momentum flux equation applied to the control volume in Figure 4.5 gives Equation 4.12.

Substituting for the area in Equation 4.12 gives

2 2
 D   D 
p 1  ----------1  – p 2  ----------2  –  o  DL  = 0 (4.39)
 4   4 

Substituting for the diameter in terms of the radius in Equation 4.39 gives

2 2
p 1 r 1 – p 2 r 2 –  o  2rL  = 0 (4.40)

where

• p 1 , p 2 = pressure at points 1 and 2 respectively (N/m2 [or Pa] or lbf/ft2)

• r 1 , r 2 = radius of pipe at sections 1 and 2 respectively (m or ft)

88
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

•  o = shear stress at the boundary of the inside of the pipe (N/m2 or lbf/ft2)
• r = radial direction in pipe towards the edge of the cylindrical control volume in Figure 4.5
as measured from the centerline of the pipe (m or ft)
• L = length of the control volume (m or ft)
Recalling Newton’s law of viscosity from Equation 2.10 as

 =  du = –  dv (4.41)
dy dr

Figure 4.5 Control volume for momentum equation applied to laminar flow

Now combining Equation 4.40 and Equation 4.41 gives

2 dv
 p 1 – p 2 r +   2rL  = 0 (4.42)
dr

Separating variables and integrating gives

p1 – p2
dv = – -----------------
- rdr
2L

Integrating gives:

 p1 – p2 
 dv = –  ----------------------
- rdr
2L

 p1 – p2 
v = – ----------------------
- r dr
2L 

 p1 – p2  2
v = – ----------------------
-r + C (4.43)
4L

89
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Thus the velocity distribution for laminar flow in a pipeline is parabolic as shown in Figure 4.6.

Figure 4.6 A parabolic velocity distribution for laminar flow

Let the velocity at the centerline be  = max at r = 0 leading to determination of the integration constant
in Equation 4.43 as

C = v max (4.44)

Thus from Equation 4.43 and Equation 4.44 it follows that the velocity distribution for laminar flow is

 p1 – p2  2
v = v max – ----------------------
-r (4.45)
4L

4.4.3 The maximum velocity at the centerline of the pipe for laminar flow
Thus at the pipe wall v = 0 for the no – slip condition and r = Dthus in Equation 4.45 the centerline
(also the maximum) velocity is

 p1 – p2  2
v max = ----------------------
-D (4.46)
16L

Thus substituting back into Equation 4.45 gives

 p1 – p2  D2 2
v = ----------------------
- ------- – r (4.47)
4L 4

VELOCITY DISTRIBUTION FOR LAMINAR PIPE FLOW

4.4.4 Discharge for laminar flow

From Streeter and Wylie (p. 192):

D2
Q = 0 2rv dr (4.48)

Thus substituting Equation 4.47 into Equation 4.48 and carrying out the integration gives:

4
 p 1 – p 2 D
Q = ---------------------------------
- (horizontal pipe only) (4.49)
128L

4.4.5 The mean velocity for laminar flow

The mean velocity V for the parabolic laminar flow is

90
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

2
v max D
----------------------
Q 8 -
V = ---- = ---------------------- (4.50)
A D
2
----------
4

Thus

v max  p1 – p2  2
- = -----------------------
V = ---------- -D (4.51)
2 32L

4.4.6 The head loss for laminar flow

Writing the energy equation between sections 1 and 2 in the pipeline in Figure 4.5 gives

2 2
p V p V
z 1 + -----1- + ------1- = z 2 + -----2- + ------2- + h f (4.52)
 2g  2g

Using z1 = z2 and V1 = V2 it follows that for a horizontal pipeline that the head loss is

p p
h f = -----1- – -----2- (4.53)
 

Thus from Equation 4.51 and Equation 4.53 eliminating p 1 – p 2 gives a head loss of

32V L
h f = --------------------- (4.54)
2
g D

for a horizontal pipe. Thus the head loss is directly proportional to the magnitude of the average velocity
of V . This is called Poiseuille flow.

4.4.7 Friction factor for laminar pipeline flow

2
Taking Equation 4.54 and expressing in terms of the mean velocity squared V gives

2
64 L V
h f =  ------------ ------ -------- (4.55)
 V D D 2g

Compare Equation 4.55 with the Darcy – Weisbach head loss equation (Equation 4.3, page 77). Thus the
friction factor for laminar flow (see Equation 4.37 for definitions of variables) given as the term in
parentheses in Equation 4.55 is

64- = -------------
64 -
f = ------ (4.56)
Re  VD
--------
  

91
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

4.5 Changing from Laminar Flow to Turbulent Flow

Laminar flow exists when a thread of dye in the flow remains unbroken. Osborne Reynolds at the Uni-
versity of Manchester in the 19th century conducted experiments with laminar and turbulent flow and
observed there was a critical velocity Vc at which the dye thread breaks up indicating a change from
laminar to turbulent flow. This flow regime change point is also called the critical zone (McGhee 1991).
From dimensional analysis Reynolds showed the critical velocity at which the dye threads break up
depends on fluid viscosity , fluid density and pipe diameter D as follows


V c = F  ------- (4.57)
 D


V c = F  ------- (4.58)
 D

From experiments, Reynolds found the function in Equation 4.58 was linear and the constant relating Vc

and ------- was approximately 2000. Later experiments showed the value of the constant to be 2300.
D

Thus


V c = 2300  ------- (4.59)
 D

This leads to the critical Reynolds number that separates laminar flow behavior from turbulent flow
defined as

V c D V cD
Re c = --------------
- = ----------
- = 2300 (4.60)
 

The accuracy of this dividing point between laminar and turbulent flow is now addressed. Below a Re of
approximately 2300 it is impossible to have turbulent flow. The corresponding velocity is called the
lower critical velocity.

For a pipe protected from vibration and with a suitably shaped entrance such that turbulence of the
inflow is extremely low, laminar flow is possible up to a Reynolds number of 50,000. Thus, the upper
critical velocity is not well defined, but in practice when Re > 2800 the flow in a circular pipe is usually
turbulent. Generally, the flow of air and water in conduits is turbulent (however, the Reynolds number
Re should be checked, of course). Oil flow is frequently laminar because of the higher viscosity and
lower density. Usually velocities for water flow in pipelines must be very low for laminar flow. Low
velocities do occur in lubrication systems and dashpots.

In the transition region the velocity profile changes from laminar (parabolic) up to Re = 2300 to turbu-
lent (uniform) for a Reynolds number on the order of a couple of hundred thousand. A turbulent flow
velocity profile is shown in Figure 4.7.

Figure 4.7 A turbulent flow velocity profile

92
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

4.6 Losses in Turbulent Flow in Pipelines

von Karman and Prandtl made a very significant contribution to the understanding of boundary fluid
resistance in turbulent pipe flow. They developed a theoretical analysis of pipe flow in smooth and
rough pipes.

There are three distinct types of turbulent flow that are considered in this section. These include smooth
pipe turbulent flow, transitional turbulent flow and fully rough turbulent flow. The behavior of the
boundary layer at the wall and the viscous sublayer (that is in the immediate vicinity of the wall) is the
most important determining factor on which type of flow occurs.

4.6.1 The boundary layer and the viscous sublayer

When a real fluid is in motion along in a pipe, the fluid particles near the wall are retarded by frictional
forces due to the shear stresses in the fluid (SAE Manual 1969). These retarding forces cause a velocity
gradient near the wall. The zone of flow near the wall within the region of the velocity gradient is
referred to as the boundary layer. The thickness of the boundary layer is usually defined by the two
extremes of velocity — zero at the wall and 99% or 99.5% of the local free stream velocity at the outer
edge. Within the boundary layer the velocity profile depends on whether the flow is laminar or turbulent.
Thus the velocity gradient at the wall and hence the boundary shear force or drag depends on whether
the flow is laminar or turbulent.

The viscous sublayer (or laminar sublayer) is part of the boundary layer that grows in the flow at a solid
boundary. The viscous sublayer is very thin and is close to the boundary as shown in Figure 4.8. The vis-
cous sublayer thickness decreases with an increase in Reynolds number.

Figure 4.8 The viscous sublayer

The viscous sublayer is important for smooth pipe turbulent flow. In fully rough turbulent flow, the vis-
cous sublayer is non – existent having been broken up by the flow and the roughness elements at the
boundary. When a pipe has a boundary that is hydraulically smooth, the wall roughness projections are
buried in the viscous sublayer and they do not influence the flow. The average roughness height of the
inside of the pipe  must be less than approximately 1/3 of the viscous sublayer thickness. If the rough-
ness projections protrude through the viscous sublayer the boundary is hydraulically rough. The average
roughness height of the inside of the pipe  needs to be greater than 6 times the viscous sublayer thick-
ness for the flow to be fully rough turbulent flow.

4.6.2 The Swamee and Jain equation for the Darcy – Weisbach friction fac-
tor f
Swamee and Jain (1976) developed an expression for the Darcy – Weisbach friction factor to represent
the Moody diagram (Figure 4.2). The formula predicts f in the smooth pipe turbulent flow region, transi-
tional turbulent flow region and the fully rough turbulent flow region within the bounds on relative
roughness  /D and Reynolds number Re as specified below.

0.25
f = ----------------------------------------------------------- (4.61)
 5.74- 2
log 10  ------------ + ------------
 3.7D 0.9
Re

93
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

The Swamee and Jain equation given by Equation 4.61 is valid the following ranges of the relative
roughness and for the Reynolds number

–6 -  10 – 2
10  ---
D

8
5000  Re  10

The Swamee and Jain equation applies to smooth pipe turbulent flow, transitional turbulent flow and
fully rough turbulent flow.

Equation 4.61 is a representation of the experimental measurement of f. Typical friction factors for pipes
are in the range 0.010 to 0.035.

A more up-to-date estimate by Swamee (1993) of the Darcy-Weisbach friction factor over a broader
8
range of Reynolds number 0  Re  10 covering the following possible flow regimes of laminar, tran-
sition turbulent and fully turbulent flow is given by:

 64 8  5.74 2500 6 – 16 0.125

f =  ------- + 9.5 ln  ------------ + ------------- –  ------------  (4.62)
 3.7D 0.9  Re 
 Re Re 

4.6.3 Smooth pipe turbulent flow

For smooth pipes where the average roughness height  is very small as shown in Figure 4.9 then the
fluid friction factor only depends on the Reynolds number Re. The main characteristic of smooth pipe
turbulent flow is that the irregularities or roughness projections are small compared with the thickness of
the viscous sublayer. Smooth pipe turbulent flow is represented by one line on the Moody diagram.
Experimental measurements showed a scatter of  5% along the smooth pipe turbulent flow curve on
the Moody diagram (Figure 4.2). For a hydraulically smooth pipe the relative roughness /D approaches
zero and then the Darcy – Weisbach friction factor is only a function of Reynolds number as represented
by

f = F 1  Re  (4.63)

Figure 4.9 Viscous sublayer for smooth pipe turbulent flow

In 1911, Blasius found a relationship for the friction factor for smooth pipe turbulent flow

0.316
f = --------------- (4.64)
0.25
Re

FRICTION FACTOR FOR SMOOTH PIPE TURBULENT FLOW

The range of validity for Equation 4.64 is quite limited and is only for Reynolds numbers in the range
2800 < Re < 105.

94
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

For a hydraulically smooth pipe, the von Karman fluid resistance formula (Daily and Harleman 1966, p.
271) is

1
---------- = 2.0log 10  Re f  – 0.8 (4.65)
f

FRICTION FACTOR FOR SMOOTH PIPE TURBULENT FLOW

Smooth pipe turbulent flow type is represented by the lower curved line that extends to the right – hand
side of the Moody diagram (3/4 of diagram) as shown in Figure 4.2. For a particular Reynolds number
the smooth pipe turbulent flow friction factor is the minimum possible value. Equation 4.65 shows that
for smooth pipe turbulent flow the friction factor only depends on the Reynolds number and does not
depend on the average roughness height  on the inside of the pipe. For smooth pipe turbulent flow the
roughness projections are completely buried in the viscous sublayer. As a result, the flow is not at all
influenced by the pipe wall. The roughness projections are so small that they do not have any affect on
the flow. The viscous sublayer dominates the losses in smooth pipe turbulent flow.

Prior to considering the requirements for smooth pipe turbulent flow it is necessary to define a quantity
referred to as the shear velocity. The shear or friction velocity u* is defined as the square root of the ratio
of the shear stress at the wall of the pipe to the fluid density

o
u* = ----- (4.66)


DEFINITION OF THE SHEAR VELOCITY

where

• u * = shear velocity (m/s or ft/s)

• o = shear stress at wall of pipe (N/m2 or lbf/ft2)
•  = density of flowing fluid (kg/m3 or slugs/ft3)
The shear stress at the wall boundary from Equation 4.34 is

2
V
 o = f  -------- (4.67)
8

Thus the shear or friction velocity is expressed from Equation 4.66 and Equation 4.67 as

o
u* = ----- = V ---f (4.68)
 8

An important dimensionless number when considering the type of turbulent pipe flow is as follows

u *
 - = -------
----------- - (4.69)
  u* 

where

•  = average height of the roughness projections on the inside of the pipe (m or ft)
• = kinematic viscosity of the fluid (m2/s or ft2/s)
• u* = shear or friction velocity (m/s or ft/s)

95
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

The numerator on the left – hand side of Equation 4.69 is the average height of the roughness projections
 while the denominator is referred to as a characteristic length or height of the viscous sublayer or

- = --------
-
---- (4.70)
u* o
-----


CHARACTERISTIC LENGTH OR HEIGHT OF VISCOUS SUBLAYER

The dimensionless number on the right – hand side of Equation 4.69 has the form of a Reynolds number.
The size of this dimensionless number indicates the type of turbulent flow regime that exists in the
pipe — either smooth pipe turbulent, transitional turbulent or fully rough turbulent flow. For low values
of the dimensionless expression in Equation 4.69, the flow is smooth pipe turbulent flow and for large
values the flow is fully rough turbulent flow. For moderate values the flow is classified as in the transi-
tional turbulent flow region.

A useful way to consider Equation 4.69 is to interpret it as a length ratio reflecting a measure of the
average roughness height to a characteristic length or height of the viscous sublayer.

For flow to be classified as smooth pipe turbulent flow requires the following expression to be valid


------------  5 (4.71)
  u*

REQUIREMENT FOR SMOOTH PIPE TURBULENT FLOW

Another way of defining the occurrence of smooth pipe turbulent flow by substituting for the shear
velocity in Equation 4.71 from Equation 4.68 is as follows

  V f
------------ = --------------------------- = -----------------  5 (4.72)
  u* f  8
   V --- 
 8

Thus Equation 4.72 becomes

V f-  5
---------------- (4.73)
8

REQUIREMENT FOR SMOOTH PIPE TURBULENT FLOW*

* Note that units of  must be in meters (or feet) and not mm.

An alternative form is as follows:

 f
------------- Re  5 (4.74)
D 8

REQUIREMENT FOR SMOOTH PIPE TURBULENT FLOW*

* Note that units of  and D must be consistent, both in meters (or feet) or both in mm (or in.).

If the flow is given, then one approach to determining if the flow is in the smooth pipe turbulent flow
region is to

• compute the Reynolds number

96
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

• read the friction factor from the Moody diagram (Figure 4.2) or compute it from the
Swamee and Jain equation (Equation 4.61)
• compute the expression in Equation 4.73 and determine if it is less than 5 to verify the flow
is in the smooth pipe turbulent flow region
The following example computes the friction factor for smooth pipe turbulent flow.

Example 4.1 (i) Calculate the fluid friction factor f for flow in a new plastic pipe (uPVC) used for a network in a
subdivision development. Assume a flow velocity of V = 1.5 m/s, a pipe internal diameter of D = 142.7 mm
(150 mm nominal diameter, Class 12 — 120 m head, 8.8 mm wall thickness, density of pipe material = 36.9 kg/m3),
an internal average roughness height of the inside of the pipe of  = 0.003 mm (Hardie Iplex Pipeline Systems), a
–6
kinematic viscosity of  = 1.007 10 m2/s and a density of 998.2 kg/m3 for water at 20°C. (ii) Finally check
if the Swamee and Jain approximation of Equation 4.61 would be valid for this example.

Answer – (i) Computation of the Darcy – Weisbach friction factor.

The roughness height for the uPVC is very small and it is likely that the flow is smooth pipe turbulent flow. Make an
assumption of smooth pipe turbulent flow to commence the calculations and then check if the assumption is valid.

The relative roughness is

 0.003 –5
---- = ------------- = 2.1 10
D 142.7

The Reynolds number is

VD  1.5   0.1427 
Re = -------- = ---------------------------------- = 212600
 1.007 10
–6

The value of the Reynolds number is clearly greater than 2300 and thus the flow is turbulent rather than laminar. The
turbulent flow may be smooth pipe turbulent flow, transitional turbulent flow or fully rough turbulent flow. As the
average roughness height is so small then firstly assume the flow is smooth pipe turbulent flow. Check to verify that
this assumption is valid.

The von Karman expression for the friction factor f for smooth pipe flow from Equation 4.65 requires an iterative
solution to find f

1
---------- = 2.0log 10  Re f  – 0.8
f

Thus

1
---------- = 2.0log 10  212600 f  – 0.8
f
The iterative solution to the above equation is f = 0.0154.

Thus the computed Darcy – Weisbach friction factor is

f = 0.0154 (4.75)

Now check that the flow is actually in the hydraulically smooth zone on the Moody diagram (Figure 4.2). For the flow
to be classified as smooth pipe turbulent flow Equation 4.71 for the length ratio must hold


------------  5
  u*

The shear stress at the boundary (from Equation 4.34) of the inside of the pipe is

2 2
V  1.5  2
 o = f -------- = 0.0154  998.2  -------------- = 4.33 N/m
8 8
The shear velocity from Equation 4.68 is

o 4.33
u* = ----- = ------------- = 0.066 m/s
 998.2

The characteristic length from Equation 4.70 for the viscous sublayer is

97
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

–6
  1.007 10 -4
----- = ---------------- = --------------------------- = 0.15257 10 m = 0.0153 mm
u* o   0.066

The length ratio for the viscous sublayer from Equation 4.72 is

 0.003
------------- = ---------------- = 0.196

 ---- 0.0153
-
u 
*

Recall that for hydraulically smooth flow the length ratio in Equation 4.71 needs to be less than 5. For the length ratio
above


------------- = 0.196  5

 ---- -
u 
*

Clearly the flow is smooth pipe turbulent flow with the roughness projections completely buried in the viscous sub-
layer.

Alternatively from Equation 4.73 the condition for smooth pipe turbulent flow is

V f
---------------  5
8

The left – hand side of Equation is

–3
V f  0.003 10   1.5  0.0154
--------------- = ------------------------------------------------------------------ = 0.196  5
–6
8 8  1.007 10 

Thus leading to the same result as Equation 4.71.

(ii) Check the validity of the Swamee and Jain formula

Previously we have computed above

- = 0.003 –5
--- ------------- = 2.1 10
D 142.7

but the range of validity for the Swamee and Jain equation is:

–6  –2
10  ----  10
D
Thus the Swamee and Jain equation is valid in this situation. From Equation 4.61 then it follows that:

0.25 0.25
f = ---------------------------------------------------------- = ---------------------------------------------------------------------------------------- = 0.0153
log 10  ----------- - + ------------
5.74  2 5.74 - 2
0.003 - + ---------------------------
log 10  -------------------------
 3.7 D 0.9  3.7  142.7  0.9
Re  213200 

which is a very good approximation compared with Equation 4.75 of f = 0.0154 .

___________________________________________________________________________________

4.6.4 Transition between smooth pipe turbulent flow and fully rough turbu-
lent flow
Colebrook (1938 –39) and White developed a formula for determining the Darcy – Weisbach fluid fric-
tion factor f for the transitional turbulent region between smooth pipe turbulent flow and fully rough tur-
bulent flow. In this transitional turbulent region, the fluid friction factor f depends on both the Reynolds
number and the relative roughness. Colebrook and White showed that the fluid resistance to flow for
uniform sand grains is different for the equivalent non – uniform roughness, such as in commercial
pipes. The coarser projections disturb the viscous sublayer before the smaller irregularities become
effective.

In 1933, experiments by Nikuradse proved that the fluid friction factor in transitional turbulent flow
depends on relative roughness /D and Reynolds number Re as

98
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6


f = F 2  Re ----
 D

where

• F2 represents a function determined from experimentation

The Colebrook – White formula for the Darcy – Weisbach fluid friction factor for transitional turbulent
flow is

1  2.51
---------- = – 2.0 log 10  ------------ + -----------------  (4.76)
f  3.7D Re f 

COLEBROOK – WHITE FORMULA FOR f IN TRANSITIONAL

TURBULENT FLOW

Using logarithmic relations of

C = Alog 10 B

C = DlnB

The relation between coefficients A and D is

A A
D = -------------- = ---------------- = 0.4343 A (4.77)
ln(10) 2.3026

Thus the Colebrook – White formula of Equation 4.76 is expressed in terms of the natural logarithm for
transitional turbulent flow as

1  - + -----------------
2.51 
---------- = – 0.8686 ln  ----------- (4.78)
f  3.7D Re f 

COLEBROOK – WHITE FORMULA FOR TRANSITIONAL TURBULENT

FLOW

Note that a coefficient of – 0.86 has been commonly used in the past rather than the slightly larger value
as shown in Equation 4.78.

99
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

The following dimensionless length ratio needs to be satisfied for the flow to be classified as transitional
turbulent


5  ----------
-  70 (4.79)

 -----
u 
*

or (4.80)


5  ----------------------  70

   ----o-
 

or

 f Re
5  ---------------------  70 (4.81)
D 8

REQUIREMENTS FOR TRANSITIONAL TURBULENT FLOW*

Example 4.2 (i) For the following flow situation determine the Darcy – Weisbach friction factor and compute which
region of the Moody diagram the flow is in Figure 4.2. Flow is occurring through a pipe connecting two reservoirs
with an upstream water surface elevation of 50.0 m and a downstream water surface of 44.46 m. Consider a flow
velocity of V = 0.9 m/s, a length of pipe of 2000 m, an internal pipe diameter of D = 300 mm = 0.3 m, an average
–6
roughness height for the inside of the pipe of  = 0.25 mm, a kinematic viscosity of  = 1.007 10 m2/s and a fluid
density of  = 998.2 kg/m3 for water at 20°C. Use the Swamee and Jain (1976) equation to compute the Darcy-Weis-
bach friction factor. (ii) Now use the Swamee (1993) equation of Equation 4.61 to compute the Darcy-Weisbach fric-
tion factor.

Answer – (i) Computation of the Darcy – Weisbach friction factor.

It is difficult to know what type of turbulent flow is occurring here. Either the Moody diagram (Figure 4.2) or the Swa-
mee and Jain equation (Equation 4.61) is used to determine the Darcy – Weisbach friction factor as a first approxima-
tion. The length ratio for the viscous sublayer is then be determined to ascertain which flow regime is occurring.

The relative roughness is

 0.25
---- = ---------- = 0.000833
D 300

The Reynolds number is

0.9  0.3  = 268100

Re = VD
-------- = ---------------------------
 1.007 10
–6

From the Moody diagram (Figure 4.2) the Darcy – Weisbach f value is 0.020. Alternatively from the Swamee and Jain
equation for the Darcy – Weisbach friction factor (Equation 4.61) may be used

0.25
f = ---------------------------------------------------------
-
 5.74 2
log 10  ------------ + ------------
 3.7 D 0.9
Re

0.25
f = ----------------------------------------------------------------------------------- = 0.0201
0.25 5.74 2
log 10 --------------------- + ----------------------------

 3.7  300  0.9
 268100 

f = 0.0201 (4.82)

Thus the Darcy – Weisbach friction factor is

f = 0.0201

100
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

The shear stress at the boundary from Equation 4.34 is

2
=  0.0201 998.2 --------------
f 2 0.9  = 2
 o = ---V 2.02 N/m
8 8
The shear velocity from Equation 4.66 is

 2.02- = 0.0450 m/s

u* = ----o- = ------------
 998.2

The characteristic length for the viscous sublayer from Equation 4.70 is

–6
 1.007 10 = 0.224 10-4 m = 0.0224 mm
- = ---------------------------
---------- = ----
o u* 0.045
-----


The length ratio for the viscous sublayer from Equation 4.72 is

 0.25
------------ = ---------------- = 11.2
  u* 0.0224

Recall that for smooth pipe turbulent flow this ratio needs to be less than 5 and for fully rough turbulent flow the ratio
needs to be greater than 70. Thus transitional turbulent flow is in the range from Equation 4.79 of

 -  70
5  -----------
  u*

Clearly the flow is in the transitional turbulent zone between smooth pipe turbulent flow and fully rough turbulent
flow. Thus our assumption of transitional turbulent flow is valid.

The head loss in the pipe is given as

2 2
L- V
------ = 0.0201  2000 0.9
h f = f --- ------------ ------------------ = 5.532 m
D 2g  0.3  2  9.81 

The difference in the levels between the reservoirs is

z 1 – z 2 = 50 – 44.46 = 5.54 m

that agrees with the head loss to within 8 mm.

(i) Computation of the Darcy – Weisbach friction factor by Swamee (1993) for comparison

From Equation 4.62 then

 64 8  - + ------------
5.74  –  2500 6 – 16 0.125
f =  ------
- + 9.5 ln  ----------- ------------ 
 3.7 D 0.9  Re 
 Re Re 

Thus

 64 - 8 0.25 + --------------------------- 2500 - 6

5.74 - –  ----------------- – 16 0.125
f =  ----------------- + 9.5 ln  ---------------------  = 0.201
 3.7  300  0.9  268100
 268100  268100  

This matches Equation 4.82 of f = 0.0201 .

___________________________________________________________________________________

4.6.5 Fully rough turbulent flow

Moving vertically upwards on the Moody diagram (Figure 4.2) away from the smooth pipe turbulent
flow curve takes one into the transitional turbulent region as described in the section above. The region
on the right – hand side of the dashed line in Figure 4.2 of the Moody diagram is referred to as the fully
rough turbulent flow regime. This flow regime region is considered in this section.

In 1932 and 1933, Nikuradse under the direction of Prandtl and von Karman carried out a series of now
famous artificially roughened pipe experiments. Uniform sized sand grains were glued to the inside of
pipes. Many different pipe flow velocities were tested. A plot of the fluid friction coefficient or Darcy –
Weisbach fluid resistance factor was plotted against Reynolds number for various values of relative

101
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

roughness. The relative roughness was defined as the average diameter k of sand grains divided by the
pipe radius ro. The Nikuradse roughness height k value is not the same as the average roughness height
 of the inside of the pipe as defined by more recent writers. Commercially available pipes have non –
uniform roughness. The average roughness height of the inside of the pipe for commercially available
pipes is sometimes referred to as the rugosity.

.For fully rough turbulent flow in pipes, the height of the protrusions  need to be at least six times
greater than the thickness of the viscous sublayer. A depiction of the height of the roughness projections
relative to the viscous sublayer height for fully rough turbulent flow is shown in Figure 4.10.

Figure 4.10 Viscous sublayer for fully rough turbulent flow

In fully rough turbulent flow in a pipe, the irregularities or roughness projections on the inside of the
pipe break up the viscous sublayer such that fully rough turbulent flow is developed. In the fully rough
turbulent flow zone, the fluid friction factor does not depend on the Reynolds number. The functional
form of the relationship is

-
f = F 3  ---
 D

The value of f is effectively a constant value in this zone and only depends on the relative roughness
height /D. For a hydraulically rough pipe the Prandtl – Nikuradse fluid resistance coefficient formula is

1 D
---------- = 2.0log 10  ---- + 1.14 (4.83)
f  

FRICTION FACTOR FOR FULLY ROUGH TURBULENT FLOW

Thus if the pipe diameter and roughness height are given, the Darcy – Weisbach friction factor for fully
rough turbulent flow can be computed by rearranging Equation 4.83 as follows.

2
1
f = ----------------------------------------------------- (4.84)
 2.0log  --- D 
 10  - + 1.14

The transition to fully rough turbulent flow in a pipe occurs when the following dimensionless number
exceeds the value of 70 to 80.


----------------------  70 (4.85)

   ----o-
 

102
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

or

 -  70
----------- (4.86)
  u*

REQUIREMENT FOR FULLY ROUGH TURBULENT FLOW IN A PIPE*

* Note that units of  must be in meters (or feet) and not mm (or in.).

Recalling that the shear stress at the wall of the pipe is given by Equation 4.34 as

f 2
 o = ---V
8

The shear velocity is expressed from Equation 4.66 as

o f
u* = ----- = V --- (4.87)
 8

From the definition of the Reynolds number it follows that

 Re 
V = --------------- (4.88)
D

Substituting into for the shear velocity from Equation 4.87 and velocity in terms of Reynolds number
from Equation 4.88 from gives the length ratio for the viscous sublayer as

     Re f
---------- = ---------------------------- = --------------------------------------- = -------------------------------------------- = -------------------- (4.89)
 u*  D 8    Re f 
   ---------------
Re  - ---f  D 8
   V ---f 
 8  D 8

Thus the requirement for fully rough turbulent flow becomes

 Re f
----------------------  70 (4.90)
D 8

Now in terms of the Reynolds number from Equation 4.90 it follows that

70D 8
Re  -------------------- (4.91)
 f

Re  200D
------------- (4.92)
 f

REQUIREMENT FOR FULLY ROUGH TURBULENT PIPE FLOW

Thus if the Reynolds number Re exceeds the value on the right – hand side of Equation 4.92 the flow is
fully rough turbulent flow. This is the equation of the dashed line on the Moody diagram of Figure 4.2
(Zipparro and Hansen 1993) that separates the transitional turbulent zone from the fully rough turbulent
flow zone. The following example computes the friction factor for a fully rough turbulent flow regime.

103
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Example 4.3 Consider the flow in a large head works transmission pipeline with a flow velocity of V = 3.0 m/s, a
length of 2000 m, an internal diameter of D = 2 m = 2000 mm, an average roughness height for the inside of the pipe
of  = 1 mm (steel), a kinematic viscosity  = 1.007 x 10-6 m2/s and a density of  = 998.2 kg/m3 for water at 20°C.
Flow is occurring through the pipe that connects two reservoirs with an upstream water surface elevation of 50.0 m
and a downstream water surface of 42.34 m.

(i) Calculate the fluid friction factor.

(ii) Verify the pipe flow and head loss along the pipe.

Answer – (i) Computation of the friction factor.

The roughness height on the inside of the pipe is relatively large. Assume the flow in the pipe is fully rough turbulent
and then check if this assumption is valid.

The relative roughness is

 1
---- = -----------
- = 0.0005
D 2000

The Reynolds number is

VD 3.0  2.0  6
Re = -------- = --------------------------- = 5.976 x 10
v 1.007 10
–6

Using the fluid friction factor f for fully rough turbulent flow from Equation 4.84 (Prandtl – Nikuradse formula) gives

2 2
1 1 1 2
f = ------------------------------------------------------- = ------------------------------------------------------------ =  -------------  = 0.0167
D  7.742 
 2.0log  ---   2.0log  2000  
 10  -  + 1.14  10  ------------ + 1.14
 1

Now check that the flow is actually in the fully rough turbulent flow zone on the Moody diagram (Figure 4.2). For the
flow to be classified as fully rough turbulent flow the following needs to hold (from Equation 4.86)

 
------------ = ------------------------  70
  u* 
   ----o-
 

The shear stress at the boundary from Equation 4.34 is

2
f 2  3.0  2
 o = ---V =  0.0167 998.2 -------------- = 18.8 N/m
8 8
The shear velocity from Equation 4.66 is

o 18.8
u* = ----- = ------------- = 0.137 m/s
 998.2

The characteristic length for the viscous sublayer from Equation 4.70 is

–6
  1.007 10 –6 –3
----------- = ----- = --------------------------- = 7.33 10 m = 7.33 10 mm
o u* 0.137
-----


The length ratio for the viscous sublayer from Equation 4.89 is

 1
------------- = ------------------------ = 136

 ---- 7.33 10
–3
-
u 
*

Recall that for a fully turbulent flow the length ratio needs to be greater than 70. For the length ratio above


------------- = 136  70

 ---- -
u 
*

Clearly the flow is fully turbulent. Thus our assumption of fully rough turbulent flow is valid.

______________________________________________________________________________________________

Answer – (ii) Computation of the pipe flow and head loss.

104
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

The area of flow is

2 2
D   2.0  2
A = ---------- = ------------------ = 3.142 m
4 4

Thus the flow in the pipeline between the two reservoirs is

3
Q = VA =  3.0 3.142 = 9.426 m /s

The head loss in the pipe is given as

2 2
LV 2000 3.0
h f = f ---- -------- = 0.0167  ------------  ------------------ = 7.661 m
D 2g  2.0  2  9.81 

The difference in the levels between the reservoirs is

z 1 – z 2 = 50 – 42.34 = 7.66 m

that agrees with the head loss.

___________________________________________________________________________________

4.6.6 Aging of pipes

Physical and chemical characteristics may change the interior of a pipeline after installation thereby
affecting the interior roughness of the pipe and hence the head losses along the pipe. The actual diameter
may also be reduced — although this is generally not taken into account directly in computer
modeling — the roughness height value is adjusted instead. The condition of a pipe depends on (AWWA
1984)

• type of lining
• age of the pipe
• slime growths
• growths on the inside of the pipe as described below
Conditions that may adversely affect the pipe include (ASCE 1975)

• sedimentation (sand or salt settles out during low velocities)

• organic growths (may be chemically controlled, for example, by chlorine)
• mineral deposits
• turberculation
• corrosion (unlined metallic mains). Lining of a pipe with cement mortar or coal tar inhibits
corrosion. Ductile iron cement mortar lined (DICL) pipes are used by many water authori-
ties.
As pipes grow older, the roughness of the inside of the pipe usually increases with time. An encrustation
referred to as tuberculation causes a growth on the inside of the pipe. The result is a change, in fact an
increase, in the average roughness height of the inside of the pipe with time. Roughness characteristics
of old pipelines may be determined by flow testing on a section of straight pipeline. In addition, inspec-
tion of pipe sections may also provide information on the interior wall conditions.

Colebrook and White (1937) developed a formula for the growth of roughness particles on the wall of
coated pipes. The effective height of roughness particles after T years is given by

 =  o + T (4.93)

where

• = effective height of roughness particles after T years

• o = initial effective height of roughness particles (mm or in.)
•  = the annual rate of growth of roughness particles (dimensionless)

105
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Skeat and Dangerfield (1969) reported that over 100 British studies (Lamont 1952; 1954; 1955) and
other records of cast iron and other pipes confirmed that the uniform growth relationship of Colebrook
and White was valid. The same principle also seems to apply to the growth of slime in spun – lined
pipes.

Skeat and Dangerfield (1969) reported on the growth rate of particles on the inside of coated cast iron
pipes. The growth rate depends on the Langlier Saturation Index of the water (Langlier 1936). A defini-
tive formula for the Langlier Saturation Index of the water is still not available. It is a function of pH,
calcium content and alkalinity of the water. In addition to the influence of the chemical characteristics of
the water, iron bacteria also plays a role in the development of turberculation on coated cast iron pipe.
Skeat and Dangerfield (1969) suggested that it is possible to determine the approximate upper and lower
limits for the growth rate with British waters and to associate particular types of water with particular
growth rates. British waters are classified into four types based on the degree of corrosive attack of
water. This also determines the growth rate. The classification includes slight, moderate, appreciable
and severe degree of corrosive attack of water.

Table 4.5 shows growth rates in coated cast iron pipes.

Table 4.5 Growth rates in coated cast iron pipes

Degree of Approxi- Growth rate

corrosive mate mean mm (in.) per Equivalent surface roughness height in mm (in.)
attack of value of annum
water Langlier
saturation
index
After 30 years After 60years After 100
years
Slight 0 0.025 (0.001) 0.89 (0.035) 1.65 (0.065) 2.67 (0.105)
Moderate – 1.3 0.076 (0.003) 2.41 (0.095) 4.7 (0.185) 7.75 (0.305)
Appreciable – 2.6 0.254 (0.010) 7.75 (0.305) 15.4 (0.605) 25.5 (1.005)
Severe – 3.9 0.762 (0.030) 23.0 (0.905) 45.8 (1.805) 76.3 (3.005)

The range of growth rates for slime formation is similar to that of cast iron pipes (Skeat and Dangerfield
1969). In some cases the thickness of the slime film is so great that the consequent reduction in the inter-
nal diameter of the pipe must be taken into account. In general, slime does not damage pipe walls. The
removal of slime from the interior of the pipe restores the capacity of the pipe to its original value.

4.7 Hazen – Williams Equation

The Hazen – Williams (H-W) equation is an empirical equation that is widely used for pipe flow compu-
tations in the water industry (Liou 1998). Its origins date back to 02 (Liou 1998). Empirical formulas
such as the Hazen – Williams equation for the relationship between head loss and flow only apply to
clean, slime – free water at normal temperatures (Skeat and Dangerfield 1969). In contrast to the Darcy –
Weisbach head loss equation, empirical formula such as the Hazen – Williams equation are not appro-
priate when applied to liquids other than water. The Hazen – Williams equation is in common use in the
water industry due to historical reasons, despite authors cautioning against the indiscriminate use of
equation (Swamee and Jain 1976). Streeter et al. (1998) and Jeppson (1976) have presented comparisons
of the range of validity of the Hazen – Williams equation and the Darcy – Weisbach equation. Liou
(1998) presented a paper where he outlined the limitations and proper use of the Hazen-Williams equa-
tion. The conclusions are that the Hazen – Williams equation approximates the head loss only over a
limited range of Reynolds number (or velocity) even if an appropriate value of the Hazen – Williams
coefficient C is selected for a given Reynolds number (Jeppson 1976). Kamand (1988) also stated “a
constant value of friction coefficient in the H-W equation is accurate only for a certain range of veloci-
ties for a given pipe diameter.” Liou (1998) stated “The level of error when the Hazen-Williams equation

106
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

is used outside its data ranges is significant.” Additionally Christensen (2000) stated that “a range does
exist where this formula (the Hazen-Williams) may be applied, but most practical pipe flows are outside
that range that will be defined by the analysis. Most hydraulic engineers are applying the Hazen-Wil-
liams formula in pipe flow ranges where it is not valid at all.” Christensen (2000) concluded “that usage
of the Hazen-Williams formula should be strongly discouraged.” He computed that for the Hazen-Williams
equation to be valid that the minimum pipe size needs to be at least 1.441 meters for a Darcy-Weisbach rough-
ness height of  = 1 mm.

The Hazen – Williams C value is assumed to depend only on the pipe material and the condition of
inside of the conduit (the Reynolds number dependence is ignored) whereas the Darcy – Weisbach fluid
friction factor f depends on conduit material and condition (reflected through the roughness height),
velocity, internal pipe diameter and kinematic viscosity of the flowing fluid. Temperature variations for
the fluid and the consequent effect on fluid viscosity and density are ignored in the Hazen – Williams
head loss equation. The Hazen – Williams equation is only applicable at high Reynolds numbers (fairly
highly turbulent flow). The Hazen – Williams equation is valid for pipes greater than 50 mm and veloci-
ties less than 3 m/s. The Hazen – Williams equation is well suited to hand computations because C is not
dependent on the Reynolds number Re of the flow. With the availability of computers the justification
for the use of Hazen – Williams formula no longer exists. The Darcy – Weisbach head loss equation has
a better theoretical basis than the Hazen – Williams head loss equation. The Darcy – Weisbach formula
better accounts for the variation of losses with changes in flow velocity and also variations in viscosity
of the fluid.

The Hazen – Williams equation will only properly model new pipes that operate in the hydraulically
smooth turbulent flow range (Christensen et al. 2000). There is always the possibility the formula is
applied outside its range of validity.

Since most pipeline flows occur in the transitional turbulent region (between smooth pipe turbulent flow
and fully rough turbulent flow) the Hazen – Williams equation is not the best available option. For best
engineering practice, it is recommended that the Darcy – Weisbach fluid friction loss equation and the
Colebrook – White formula or the Swamee and Jain formula for fluid friction factor (Equation 4.61) be
used for calculating pipe fluid friction loss (Wunderlich and Giles 1986). With the knowledge that many
water utilities will continue to use the Hazen – Williams formula, a detailed description is included
below. A detailed examination of the range of errors that occur when the Hazen – Williams equation is
used rather than the Darcy – Weisbach fluid friction equation is given in Section 4.7.7.

In US customary units the Hazen – Williams equation from Brater and King (1976, p. 6 – 17)

0.63 0.54
V = 1.318C R Sf (4.94)

THE HAZEN – WILLIAMS EQUATION (US CUSTOMARY UNITS)

where

• V = average flow velocity in the pipe (ft/s)

• C = Hazen – Williams coefficient (typically 80 for very rough, 145 for very smooth)
• R = hydraulic radius of the pipe (ft) = A/P; A = area of flow; P = wetted perimeter (see
Equation 4.99)
• Sf = friction slope or slope of EGL = h f /L (dimensionless); h f = head loss, L = length of
pipe.
In SI units, the Hazen – Williams flow equation is given with V in m/s and R in meters (m) by using the
conversion of 1 m = 3.28 ft and from Equation 4.94 as

0.63 0.54
 V *3.28  = 1.318C  R*3.28  Sf (4.95)

Thus the coefficient on the right – hand side of Equation 4.95 for the SI units equation becomes

107
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

0.63
1.318  3.28  1.318 - = 0.84926
coefficient = ------------------------------------- = ------------------ (4.96)
3.28 3.28
0.37

Thus the Hazen – Williams equation in SI units is

0.63 0.54
V = 0.84926C R Sf (4.97)

THE HAZEN – WILLIAMS EQUATION (SI UNITS)

where

• V = average flow velocity in the pipe (m/s)

• C = Hazen – Williams coefficient (typically 80 for very rough, 145 for very smooth)
• R = hydraulic radius of the pipe (m = A/P; A = area of flow; P = wetted perimeter (see
Equation 4.99)
• Sf = friction slope or slope of EGL = h f /L (dimensionless), h f = head loss, L = length of
pipe.
Five significant figures in Equation 4.97 are retained to ensure consistency between both sets of units
and for derivation of head loss expressions.

4.7.1 The Hazen – Williams head loss equation for SI units

The friction slope is given as

h
S f = -----f- (4.98)
L

FRICTION SLOPE

where

• S f = friction slope of the HGL (dimensionless)

• h f = head loss along the pipe (m or ft)

• L = length of pipe (m or ft)
The hydraulic radius for a circular pipe defined as the area of the cross – section of flow divided by the
wetted perimeter is

2
D 
 ----------
A  4  D-
R = --- = ---------------- = --- (4.99)
P D 4

where

• R = hydraulic radius (m or ft)

• A = flow cross – section (m2 or ft2)
• P = wetted perimeter (m or ft)
• D = internal diameter of the pipe (m or ft)
Substituting for the hydraulic radius and the friction slope from Equation 4.98 and Equation 4.99 into
the SI units version of the Hazen – Williams equation Equation 4.97 gives

108
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

D 0.63  h f  0.54
= 0.84926C  ----
0.63 0.54
V = 0.84926C R Sf ------- (4.100)
 4  L

Solving for the head loss h f gives

0.63 0.54
0.54 4 VL
hf = -------------------------------------
0.63
0.84926C D

1-
---------
 2.8200V L 0.54 0.54
h f =  --------------------------------- (4.101)
 C D 0.63 

Care now needs to be taken with the exponents to ensure enough significant figures are carried so that
results obtained from the US customary units and SI units version of the Hazen – Williams equation are
consistent.

1
---------- = 1.851852  1.852 (4.102)
0.54

0.63
---------- = 1.16666  1.167 (4.103)
0.54

Substituting the exponents defined in Equation 4.102 and Equation 4.103 into Equation 4.101 gives

1.852 L 1.852
h f =  2.8200  ------------------------------- V
1.852 1.167
C D

1.852
LV
h f = 6.8214 ------------------------------- (4.104)
1.852 1.167
C D

HEAD LOSS FOR HAZEN – WILLIAMS HEAD LOSS EQUATION (SI UNITS)

Equation 4.104 shows that

• head loss h f is proportional to the pipe length L

• head loss h f is proportional to velocity of flow V or flow Q to a power that is a fractional
exponent (1.852) — rather than a power of 2 for the Darcy – Weisbach equation
• head loss h f is proportional to internal diameter of pipe D to a power that is a fractional
exponent 1.167 — rather than a power of 1.0 for the Darcy – Weisbach equation
The following example illustrates computation of the velocity of flow in a pipeline based on the Hazen –
Williams head loss equation.

Example 4.4 Consider a pipe of diameter D = 300 mm where the friction slope Sf = 2 m of head loss per 1000 m
length and the Hazen – Williams coefficient is C = 120. Compute the velocity of flow in the pipe.

Answer – Computation of the velocity.

Use the Hazen – Williams equation (in SI units) to compute the velocity.

The friction slope from Equation 4.98 is

109
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

2
S f = ------------ = 0.002
1000

The hydraulic radius from Equation 4.99 is

D 0.3
R = ---- = ------- = 0.075 m
4 4

The velocity from the Hazen – Williams equation in SI units of Equation 4.97 is

0.63 0.54 0.63 0.54

V = 0.84926C R S f = 0.84926  120   0.075   0.002 

V = 0.6951 m/s

Now check the head loss by using Equation 4.104

1.852 1.852
LV  1000   0.6951 
h f = 6.8214 ----------------------------- = 6.8214 -------------------------------------------------
1.852 1.167 1.852 1.167
C D 120 0.3

h f = 1.9996 m  2.0 m

Thus the head loss is confirmed.

___________________________________________________________________________________

Now express the Hazen – Williams head loss equation in terms of flow from Equation 4.104. The result
is

1.852 1.852
LQ LQ
h f = 6.8214 ---------------------------------------------
- = 6.8214 ----------------------------------------------------------
-
1.852 1.167 1.852 2 1.852
C D A 1.852 1.167  D 
C D ----------
 4 

Thus it follows that

1.852 1.852
6.8214  4  LQ
h f = ----------------------------------- -----------------------------------------------
1.852 1.852 1.167 3.704
 C D D

L 1.852
h f = 10.670 ------------------------------- Q (4.105)
1.852 4.871
C D

HEAD LOSS FOR HAZEN – WILLIAMS HEAD LOSS EQUATION (SI UNITS)

Example 4.5 For the pipe described in Example 4.4 (diameter D = 300 mm where the friction slope Sf = 2 m of head
loss per 1000 m length and the Hazen – Williams coefficient is C = 120) compute the flow in the pipe.

Answer – Computation of pipe flow.

From Equation 4.97 the velocity was calculated in Example 4.4 as

V = 0.6951 m/s

The flow is

2 2
Q = VA = 0.6951   0.3 - = 0.04913 m 3 /s
D - = 0.6951  ------------
---------
 4   4 

Now checking the head loss using Equation 4.105 gives

L 1.852 L 1.852
h f = 10.670 -----------------------------Q -  0.04914 
= 10.670 -------------------------------------
1.852 4.871 1.852 4.871
C D 120 0.3

h f = 1.9996 m  2.0 m

110
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Thus the head loss is again confirmed.

___________________________________________________________________________________

4.7.2 The Hazen – Williams head loss equation for US customary units
Substituting for the hydraulic radius and the friction slope from Equation 4.98 and Equation 4.99 into
Equation 4.94 gives

 h------f-
0.63 0.54
D
= 1.318C  ----
0.63 0.54
V = 1.318C R Sf
 4  L

Solving for the head loss h f gives

0.63 0.54
0.54 4 VL
hf = -------------------------------
0.63
1.318CD

1
----------
 1.8171V L 0.54 0.54
hf =  --------------------------------- (4.106)
 C D 0.63 

Using exponents defined by Equation 4.102 and Equation 4.103 in Equation 4.106 gives

1.852
1.852 LV
h f =  1.8171  ------------------------------
-
1.852 1.167
C D

1.852
LV
h f = 3.0226 ------------------------------- (4.107)
1.852 1.167
C D

HEAD LOSS FOR HAZEN – WILLIAMS (US CUSTOMARY UNITS)

1.852
6.0452g L V
h f = -------------------- --------------- --------------- (4.108)
1.852 1.167 2g
C D

In terms of flow from Equation 4.107 gives

1.852 1.852
LQ LQ
h f = 3.0226 ---------------------------------------------- = 3.0226 -----------------------------------------------------------
1.852 1.167 1.852 2 1.852
C D A 1.852 1.167  D 
C D ----------
 4 

Thus

1.852 1.852
3.0226  4  LQ
h f = ----------------------------------- -----------------------------------------------
1.852 1.852 1.167 3.704
 C D D

111
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

L 1.852
h f = 4.7279 ------------------------------
-Q (4.109)
1.852 4.871
C D

HEAD LOSS FOR HAZEN – WILLIAMS (US CUSTOMARY UNITS)

The following example computes the velocity and flow in a pipe based on the Hazen – Williams equa-
tion in US customary units.

Example 4.6 Consider a pipe of diameter D = 0.984 ft (exactly same sized diameter of 300 mm as for Example 4.4)
where the friction slope Sf = 2 ft of head loss per 1000 ft length and the Hazen – Williams coefficient is C = 120.

(i) Compute the velocity of flow and the flow in the pipe.
(ii) Verify the head loss.

Answer – (i) Computation of the velocity.

Use the Hazen – Williams equation (in US customary units) to compute the velocity.

The friction slope from Equation 4.98 is

2
S f = ------------ = 0.002
1000

The hydraulic radius from Equation 4.99 is

D- = 0.984
R = --- ------------- = 0.246 ft
4 4

The velocity from the Hazen – Williams equation in US customary units of Equation 4.94 is

0.63 0.54 0.63 0.54

V = 1.318C R S f = 1.318  120   0.246   0.002 

V = 2.280 fps (identical to the velocity of 0.6951 m/s obtained in Example 4.4)

___________________________________________________________________________________
(ii) Verification of head loss.

Now check the head loss using Equation 4.107 gives

1.852 1.852
LV  1000   0.6591 
h f = 3.0226 ----------------------------- = 3.0226 -------------------------------------------------
1.852 1.167 1.852 1.167
C D 120 0.3

h f = 1.9990 ft  2.0 ft

Thus the head loss is confirmed.

The flow is

2 2
D 0.984
Q = VA = 2.280  ---------- = 2.280  ------------------- = 1.734 cfs
 4   4 

Now checking the head loss using Equation gives

L 1.852 L 1.852
h f = 4.7279 -----------------------------Q = 4.7279 -------------------------------------------  1.734 
1.852 4.871 1.852 4.871
C D 120 0.984

h f = 1.9990 ft  2.0 ft

Thus the head loss is again confirmed.

___________________________________________________________________________________

112
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

4.7.3 The equivalent pipe diameter for two pipes in parallel for Hazen –
Williams
In Chapter 7, an expression (Equation 7.49) is derived for the equivalent uniform main of diameter DE
with a selected Hazen – Williams coefficient CE to replace two pipes in parallel of diameters D1 and D2,
lengths L1 and L2 and Hazen – Williams C coefficients of C1 and C2. It is assumed that the lengths of all
three pipes are identical such that

L E = L1 = L2 (4.110)

Now the equivalent diameter of two pipes in parallel is given by

2.63 2.63 0.3802

 C 1 D1 + C 2 D2 
DE =  ---------------------------------------------------
- (4.111)
 CE 

EQUIVALENT DIAMETER FOR TWO PIPES IN PARALLEL

4.7.4 Typical values of the Hazen – Williams coefficient

Skeat and Dangerfield (1969) gives tables of Hazen – Williams C values that are based on the examina-
tion of over 200 new pipes and over 100 records of old cast iron and other pipes. Comprehensive infor-
mation on pipe interior Hazen – Williams C values is given in Appendix C. Some typical Hazen –
Williams coefficients are presented in Table 4.6 for pipes of different material and condition.

Table 4.6 Typical Hazen – Williams coefficients

Typical Occurrence C value
PVC 150
Extremely smooth, straight pipes; asbestos cement 140
Very smooth pipes; concrete; new cast iron 130
Wood stave; new welded steel 120
Vitrified clay; new riveted steel 110
Cast iron after years of use; brick 100
Riveted steel after years of use 95
Old pipes in bad condition 60 to 80

The Hazen – Williams formula is not suitable for values of coefficient C appreciably below 100. Correc-
tions to Hazen – Williams C are necessary if the velocity of flow in the pipeline varies appreciably from
0.9 m/s. For each halving of the velocity below 0.9 m/s the Hazen – Williams C value should be
increased by 5%. For losses above 0.9 m/s with each doubling of the flow velocity the size Hazen –
Williams C value should be reduced by 5%. Corrections for other C value levels are shown in Table 4.7.
These corrections are in effect attempting to take into account the Reynolds number effect. Computer
simulation models of networks usually only allow for a fixed value of Hazen – Williams coefficient for
each pipe. Thus it is not possible in commercially available computer simulation models to easily allow
for the adjustment of the C value based on flow conditions in the pipe. These arguments all point to rea-
sons why it is preferable to use the Darcy – Weisbach head loss equation rather than the Hazen –
Williams head loss equation. This is especially important in the pipe roughness calibration process. It is
desirable to have as accurate as possible the modeling of the flow and pressure conditions in the net-
work. The example shown in Section 4.7.7 compares results from using the Darcy – Weisbach equation
versus using the Hazen – Williams equation and clearly shows that significant errors may occur.

113
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Table 4.7 Adjustments to Hazen – Williams coefficients

Values of Hazen – Williams C at Velocities Velocities
0.9 m/s (3 ft/s) below 0.9 m/s (3 ft/s) above 0.9 m/s (3 ft/s)
For each halving, re – halving For each doubling, re –
etc. of velocity relative to doubling etc. of velocity
0.9 m/s (3 ft/s) relative to 0.9 m/s (3 ft/s)
C below 100 Add 5% to C Subtract 5% from C
C from 100 to 130 Add 3% to C Subtract 3% from C
C from 130 to 140 Add 1% to C Subtract 1% from C
C above 140 Subtract 1% from C Add 1% to C

4.7.5 Comparison of Darcy – Weisbach f and Hazen – Williams C in SI units

To enable comparison of the Hazen – Williams head loss equation in SI units with the Darcy – Weisbach
formula include a 2g term in both numerator and denominator in Equation 4.104 as follows

1.852
2gL V
h f = 6.8214 ------------------------------- ---------------
1.852 1.167 2g
C D

1.852
13.643g L V
h f = -------------------- --------------- --------------- (4.112)
1.852 1.167 2g
C D

The differences between the exponents in the head loss equations for Darcy – Weisbach and Hazen –
Williams resistance equations of Equation 4.3 and Equation 4.112 are shown in Table 4.8.

Table 4.8 Differences between Darcy – Weisbach and Hazen – Williams equations
Feature Darcy – Weisbach Hazen – Williams
Velocity V exponent n=2 n = 1.852
Diameter D exponent 5.0 4.871
Inverse loss coefficient relation f 1
---------------
1.852
C

A comparison of the Hazen – Williams equation for head loss and the Darcy – Weisbach equation of
Equation 4.3 shows that the variation of Hazen – Williams coefficient C is inversely related to the
Darcy – Weisbach friction factor f as follows

13.643g 0.148
------------------------------- = V f (4.113)
0.167 1.852
D C

Thus the Darcy – Weisbach fluid friction factor (for SI units) in terms of the Hazen – Williams coeffi-
cient

13.643g
f = ------------------------------------------------ (4.114)
0.167 1.852 0.148
D C V

DARCY – WEISBACH f – HAZEN – WILLIAMS C RELATIONSHIP (SI UNITS)

and the Hazen – Williams coefficient (SI units) in terms of the Darcy – Weisbach friction factor is
expressed as

114
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

1
-------------
 13.643g  1.852  13.643g  0.54
C =  --------------------------------------- =  --------------------------------------- (4.115)
 D 0.167 V 0.148 f   D 0.167 V 0.148 f 

The following example computes the flow and head loss for flow in a pipe based on the Hazen –
Williams equation.

Example 4.7 Consider a flow in a pipe of diameter D = 250 mm, length L = 1000 m and roughness height
 = 0.25 mm (a concrete pipe as per Appendix C). Assume the velocity of flow is V = 0.75 m/s and the kinematic
–6
viscosity of water is  = 1.140 10 m2/s at a temperature of 15 degrees Celsius.

(i) Compute the Darcy – Weisbach friction factor.

(ii) Compute the equivalent Hazen – Williams C coefficient.

Answer – (i) Computation of the Darcy – Weisbach friction factor

The Reynolds number is

VD 0.75  0.25  5
Re = -------- = --------------------------- = 1.646 10
v 1.140 10
–6

Compute the Darcy – Weisbach friction factor from the Swamee and Jain equation (Equation 4.61)

0.25 0.25
f = ----------------------------------------------------------- = ------------------------------------------------------------------------------------------- = 0.0215
  5.74  2  0.25 5.74  2
log 10 ------------ + ------------ log 10  --------------------- + ------------------------------------
 3.7 D 0.9
3.7  250  0.9
Re   1.646 10  
5

___________________________________________________________________________________
(ii) Computation of the equivalent Hazen – Williams value.

The equivalent Hazen – Williams C value from Equation 4.115 is

13.643 g 0.54 13.643  9.81  0.54

C = -------------------------------------- = ------------------------------------------------------------------- = 129.84  130
0.167 0.148 0.167 0.148
D V f 0.25 0.75  0.0215 
Now compute the corresponding head losses, first based on the Darcy – Weisbach equation.

2 2
LV 1000 0.75
h f = f ---- --------- =  0.0215  ------------ ------------------ = 2.461 m
D 2g 0.25 2  9.81 
Second, for the Hazen – Williams head loss from Equation 4.115

2 2
 D - = 0.75  0.25
Q = VA = 0.75  --------- ------------------  = 0.03682 m /s
3
 4   4 

The Hazen – Williams head loss using Equation 4.105 gives

L 1.852 1000 1.852

h f = 10.670 -----------------------------Q = 10.670 ------------------------------------------------  0.03682 
1.852 4.871 1.852 4.871
C D 129.84 0.25

h f = 2.461 m

Thus the head loss values as computed from both methods are very similar (that is – within 2 mm).

Example 4.8 Consider the flow conditions in the Example 4.7 and now compute the equivalent Hazen – Williams C
value when the velocity of flow varies. Consider the following velocities of 0.25, 0.5, 0.75, 1.0, 1.25, 1.5, 1.75 and
2.0 m/s.

Answer – Computation of equivalent Hazen – Williams C values.

The following spreadsheet shows the computation of the equivalent Hazen – Williams C values. Clearly the C value
varies by almost 10%.

115
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Table 4.9 Comparison of f and C for a range of different velocities

Table 4.9 highlights the inconsistency in using a constant Hazen – Williams C value and supports the use
of the more theoretically based Darcy – Weisbach head loss equation.

4.7.6 Comparison of Darcy – Weisbach f and Hazen – Williams C in US

customary units
A comparison of the Hazen – Williams equation (Equation 4.107, page 111) in US customary units for
head loss and the Darcy – Weisbach equation (Equation 4.3, page 77) leads to

6.0452g 0.148
------------------------------- = V f (4.116)
0.167 1.852
D C

Thus in terms of the Darcy – Weisbach fluid friction factor (US customary units)

6.0452g
f = ----------------------------------------------- (4.117)
0.167 1.852 0.148
D C V

and in terms of the Hazen – Williams coefficient (US customary units) it follows that

1 -
------------
6.0452g
C =  -------------------------------------- 1.852 = --------------------------------------
6.0452g - 0.54
(4.118)
 0.167 0.148  0.167 0.148
D V f D V f

4.7.7 Assessment of accuracy of the Hazen – Williams equation

The accuracy of the Hazen – Williams equation is assessed in this section through the consideration of a
simple example. It is assumed that the Darcy – Weisbach head loss equation provides the correct solu-
tion for computation of the flow between the two reservoirs as shown in Figure 4.11. Minor losses are
ignored.

Figure 4.11 Flow between two reservoirs

116
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Smooth pipe turbulent flow case

Example 4.9 Consider the modeling of flow between two reservoirs with different elevation differences as shown in
Figure 4.11. The internal diameter D of the pipe connecting the two reservoirs is assumed to be 300 mm, 1000 m
long with a roughness height of  = 0.0015 mm (smooth pipe). Consider the first line in Table 4.10 for a head differ-
ence between the reservoirs of 1.0 m. Minor losses are ignored. Assume the kinematic viscosity of water is
–6
1.140 10 m2/s for water at 15°C. Verify the flow and velocity computations in Table 4.10.

Answer – Verification of flow and velocity calculations.

The reservoir difference is 1.0 m. Ignoring minor losses then the application of the energy equation results in the fluid
friction loss being equal to

h f = z 1 – z 2 = 1.0

The relative roughness for the pipe is

- 0.0015 -5
--- = ---------------- = 0.5 10
D 300

The flow is computed from the Swamee and Jain (1976) explicit expression for flow in a smooth pipe turbulent flow
regime based on the Darcy – Weisbach equation as derived in Chapter 6 as Equation 6.9 is

5 3
 gh f D   2 gh f D  
-  2.0log 10  --------------------
Q = ------------------------  – 0.8 (4.119)
2
8L    L  

  9.81   1.0   0.3  -  2.0log  ---------------------------------------------------

2  9.81   1.0   0.3  - – 0.8
5 3
Q = --------------------------------------------------  10   
8  1000     1.140 10–6  2  1000  

3
Q = 0.04236 m /s

This agrees with Column (2) of Table 4.10.

The cross – sectional area of the pipe is

2 2
A = 
---------- =   0.3  -
D 2
----------------- = 0.07069 m
4 4

The velocity is

V = Q
---- = 0.04235
------------------- = 0.5993 m/s
A 0.07069

This agrees with Column (3) of Table 4.10.

Table 4.10 Comparison of Darcy – Weisbach and Hazen – Williams flows for a smooth pipe

*A negative value indicates that the Hazen – Williams equation is underestimating the flow.

The Reynolds number is

117
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

VD 0.5991  0.3 
Re = -------- = ----------------------------- = 157800
 1.140 10
–6

This agrees with Column (3) of Table 4.11.

The fluid friction factor f using the von Karman expression for the smooth pipe turbulent flow friction factor from
Equation 4.65 is

1
-------- = 2.0log 10 Re f – 0.8
f

Using the Goal Seek function in Microsoft Excel to solve for the friction factor gives

f = 0.0164

This agrees with Column (4) of Table 4.11.

Table 4.11 Comparison of Darcy – Weisbach and Hazen – Williams head losses for a smooth pipe

The fluid friction factor f using the Swamee and Jain (1976) expression (Equation 4.61) for comparison is

0.25 0.25
f = ---------------------------------------------------------- = ------------------------------------------------------------------------------------------- = 0.0163
  5.74  2  0.0015 5.74  2
log 10 ------------ + ------------ log 10  --------------------- + ----------------------------------- -
 3.7 D 0.9
Re  3.7  300   1.578 105   0.9

Now checking the head loss using the von Karman smooth pipe turbulent flow friction factor

2 2
fL V 0.0163  1000   0.5991 
h f = ------ ------ = --------------------------------- ----------------------- = 1.0004 m  1.0 m
D 2g 0.3 2  9.81 
This agrees with the original assumed difference between the reservoir water surface elevations. It also agrees with
Column (5) of Table 4.11

For the Hazen – Williams formula determine the hydraulic radius and friction slope

A = 0.3
R = --- ------- = 0.075 m
P 4

h 1.0
S f = -----f- = ------------ = 0.001
L 1000

The equivalent Hazen – Williams C coefficient value for the above Darcy – Weisbach fluid friction factor is obtained
from Equation 4.115 as

1 1
------------- -------------
13.643 g 1 1.852 13.643  9.81  1 1.852
C =  -------------------- --------------------  =  -------------------------------- -------------------------------------------------- = 150.4
 0.167 0.148   0.167 0.148 
D V f  0.3   0.5991  0.0163
This agrees with Column (6) of Table 4.11.

Now solving for the velocity based on the Hazen – Williams value of 151.92 (and not the equivalent C value of 150.4
from the equation above) at a velocity of approximately 0.9 m/s (corresponds to a 2 m head difference in Table 4.10)

0.63 0.54 0.63 0.54

V = 0.84926C R S f = 0.84926  151.92   0.075   0.001  = 0.6053 m/s (4.120)

This agrees with Column (4) of Table 4.10. The flow is

3
Q = VA = 0.6053  0.07069  = 0.04278 m /s = 42.78 L/s

The percentage error in the prediction by Hazen – Williams is

118
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Q HW – Q DW 42.78 – 42.36
- = ---------------------------------  100  = 1.00 %
% error = ------------------------------
Q DW 42.36

This agrees with Column (7) of Table 4.10. Thus the Hazen – Williams formula overestimates the flow by about 1% for
this example.

___________________________________________________________________________________

Table 4.10 shows that for smooth pipes the Hazen – Williams formula is fairly accurate, only differing
from the Darcy – Weisbach computed flow by an underestimation of up to 2.6%.

Transitional turbulent pipe flow case

Example 4.10 Consider the modeling of flow between two reservoirs with different elevation differences as shown
in Figure 4.11. The internal diameter D of the pipe connecting the two reservoirs is assumed to be 300 mm, 1000 m
long with a roughness height of  = 0.25 mm (transitional turbulent pipe flow). Consider the last line of computed
results in Table 4.12 for a head difference between the reservoirs of 25.0 m. Minor losses are ignored. Assume the
–6
kinematic viscosity of water is 1.140 10 m2/s for water at 15°C. Verify the flow and velocity computations in
Table 4.12.

Answer – Verification of the flow and velocity computations in Table 4.12.

The reservoir difference is 25.0 m. Ignoring minor losses then the application of the energy equation results in the fluid
friction loss being equal to

h f = z 1 – z 2 = 25.0m

The relative roughness for the pipe is

- 0.25 –4
--- = ---------- = 8.333 10
D 300

The flow is computed from the Swamee and Jain (1976) explicit expression for flow for flow in a transitional turbulent
regime based on the Darcy – Weisbach equation as derived in Chapter 6 as Equation 6.16 is

 
gh f D   
5
1.7748 
Q = – 0.9646 ------------------ln ------------ + ----------------------
L  3.7 D 3 
 gh f D 
 ------------------ 
L

 
5  –6 
 9.81   25.0   0.3   0.25 1.7748  1.140 10  
Q = – 0.9646 ---------------------------------------------- ln --------------------- + --------------------------------------------------
1000  3.7  300  3 
  9.81   25.0   0.3  
 ---------------------------------------------- 
1000

-4
Q =  – 0.9646   0.02441 ln  0.0002252 + 0.24885632 10 

3
Q = 0.1953 m /s

This agrees with the last value in Column (2) of Table 4.12.

119
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Table 4.12 Comparison of Darcy – Weisbach and Hazen – Williams flows for a transitional turbulent pipe flow

*
A negative value indicates that the Hazen – Williams equations is underestimating the flow.

The cross – sectional area of the pipe is

2 2
D   0.3  2
A = ---------- = ------------------ = 0.07069 m
4 4

The velocity is

Q 0.1953
V = ---- = ------------------- = 2.763 m/s
A 0.07069

This agrees with Column (3) of Table 4.12.

The Reynolds number is

2.763  0.3 
Re = VD
-------- = --------------------------- = 726470
 1.140 10
–6

This agrees with Column (3) of Table 4.13.

The fluid friction factor f using the Colebrook – White expression for the transitional turbulent zone pipe friction fac-
tor from Equation 4.76 is

 - + -----------------
1 - = – 2.0 log  ----------- 2.51 - 
-------- 10 
f 3.7 D Re f 

Using the Goal Seek function in Microsoft Excel to solve for the friction factor gives

f = 0.0193

This agrees with Column (4) of Table 4.13.

Table 4.13 Comparison of Darcy – Weisbach and Hazen – Williams head losses for a transitional turbulent pipe flow

The fluid friction factor f using the Swamee and Jain (1976) expression (Equation 4.61) for comparison is

120
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

0.25 0.25
f = ---------------------------------------------------------- = ------------------------------------------------------------------------------------ = 0.0194
  5.74  2 0.25 5.74 2
log 10 ------------ + ------------ log 10 --------------------- + -----------------------------

 3.7 D 0.9  3.7  300  0.9
Re  726470 

Now checking the head loss using the von Karman smooth pipe turbulent flow friction factor

2 2
fL- V = 0.0193  1000   2.763  =
h f = ----- -------- --------------------------------- -------------------- 25.03 m  25.0 m
D 2g 0.3 2  9.81 
This agrees with the original assumed difference between the reservoir water surface elevations. It also agrees with
Column (5) of Table 4.13

For the Hazen – Williams formula determine the hydraulic radius and friction slope

A = 0.3
R = --- ------- = 0.075 m
P 4

h 25.0- = 0.025
S f = -----f- = -----------
L 1000

The equivalent Hazen – Williams C coefficient value for the above Darcy – Weisbach fluid friction factor is obtained
from Equation 4.115 as

1 - 1 -
------------ ------------
13.643 g 1 1.852 13.643  9.81  1 1.852
C =  -------------------- ---------------------- =  -------------------------------- -------------------------------------------------- = 121.91
 0.167 0.148   0.167 0.148 
D V f  0.3   2.763   0.0193 
This agrees with Column (6) of Table 4.12.

Now solving for the velocity based on the Hazen – Williams value of 130 at a velocity of approximately 0.9 m/s (corre-
sponds to a 3 m head difference in Table 4.12)

0.63 0.54 0.63 0.54

V = 0.84926C R S f = 0.84926  130   0.075   0.025  = 2.946 m/s

This agrees with Column (4) of Table 4.12.

The flow based on Hazen – Williams velocity is

3
Q = VA = 2.946  0.07069  = 0.2083 m /s = 208.3 L/s

This agrees with Column (5) of Table 4.12.

The percentage error in the prediction by Hazen – Williams is

208.3 – 195.3
% error = ------------------------------  100  = 6.66 %
195.3
This agrees with Column (7) of Table 4.12.

___________________________________________________________________________________

Thus Table 4.12 shows that for transitional turbulent pipe flow the Hazen – Williams formula is increas-
ingly inaccurate, differing from the Darcy – Weisbach computed flow by an overestimation of up to
6.6%.

Fully rough turbulent flow case

Example 4.11 Consider the modeling of flow between two reservoirs with different elevation differences as shown
in Figure 4.11. The internal diameter D of the pipe connecting the two reservoirs is assumed to be 300 mm, 1000 m
long with a roughness height of  = 5 mm (fully rough turbulent flow). Consider the first line in Table 4.14 for a
head difference between the reservoirs of 6.0 m. Minor losses are ignored. Assume the kinematic viscosity of water
–6
is 1.140 10 m2/s for water at 15°C. Verify the flow and velocity computations in Table 4.14.

Answer – Verification of the flow and velocity computations in Table 4.14.

The reservoir difference is 6.0 m. Ignoring minor losses then the application of the energy equation results in the fluid
friction loss being equal to

h f = z 1 – z 2 = 6.0

The relative roughness for the pipe is

121
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

 5
---- = --------- = 0.01667
D 300

The flow is computed from the Swamee and Jain (1976) explicit expression for flow in the fully rough turbulent flow
regime based on the Darcy – Weisbach equation as derived in Chapter 6 as Equation 6.23 is

2
 D 
 gh f D 2.0  log 10  ----  + 1.14
2 5
 
 
Q = ---------------------------------------------------------------------------------------
8L

2
 
  9.81   6.0   0.3  2.0  log 10  300 --------- + 1.14
2 5
 5 
 
Q = ---------------------------------------------------------------------------------------------------------------------
8  1000 

3
Q = 0.06239 m /s

This agrees with Column (2) of Table 4.14.

Table 4.14 Comparison of Darcy – Weisbach and Hazen – Williams flows for fully rough turbulent flow in a pipe

*A negative value indicates that the Hazen – Williams equations is underestimating the flow.

The cross – sectional area of the pipe is

2 2
D   0.3  2
A = ---------- = ------------------ = 0.07069 m
4 4

The velocity is

Q 0.06239
V = ---- = ------------------- = 0.8826 m/s
A 0.07069

This agrees with Column (3) of Table 4.12.

The Reynolds number is

0.8826  0.3 
Re = VD
-------- = ----------------------------- = 232060
 1.140 10
–6

This agrees with Column (4) of Table 4.15.

The fluid friction factor f using the Colebrook – White expression for the transitional turbulent zone pipe friction fac-
tor from Equation 4.83 is

1 - = 2.0log  ---
D
------- 10  - + 1.14
f 

Solving for the friction factor gives

1
f = ------------------------------------------------------------ = 0.0453
D 2
 2.0 log 10  ---- + 1.14
 

122
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

The fluid friction factor f using the Swamee and Jain (1976) expression (Equation 4.61) for comparison is

0.25 0.25
f = ---------------------------------------------------------- = ----------------------------------------------------------------------------------- = 0.0457
  5.74  2 5 5.74 2
log 10 ------------ + ------------ log 10 --------------------- + ----------------------------

 3.7 D 0.9  3.7  300  0.9
Re  232060 

Now checking the head loss using the von Karman smooth pipe turbulent flow friction factor

2 2
fL V 0.0453  1000   0.8826 
h f = ------ -------- = --------------------------------- ----------------------- = 5.995 m  6.0 m
D 2g 0.3 2  9.81 

This agrees with the original assumed difference between the reservoir water surface elevations. It also agrees with
Column (6) of Table 4.15

Table 4.15 Comparison of Darcy – Weisbach and Hazen – Williams head losses for fully rough
turbulent flow in a pipe

For the Hazen – Williams formula determine the hydraulic radius and friction slope

A 0.3
R = --- = ------- = 0.075 m
P 4

h 6.0
S f = -----f- = ------------ = 0.006
L 1000

The equivalent Hazen – Williams C coefficient value for the above Darcy – Weisbach fluid friction factor is obtained
from Equation 4.115 as

1 - 1 -
------------ ------------
1 - 1.852 =  13.643  9.81 - ---------------------------------------------------- 1.852
C =  13.643
-------------------g- --------------------- 1 -
------------------------------- = 84.24
 0.167 0.148   0.167 0.148 
D V f  0.3   0.8826   0.0453 
This agrees with Column (5) of Table 4.15.

Now solving for the velocity based on the Hazen – Williams value of 84.18 at a velocity of approximately 0.9 m/s (cor-
responds to a 6 m head difference in Table 4.16)

0.63 0.54 0.63 0.54

V = 0.84926C R S f = 0.84926  84.18   0.075   0.006  = 0.8828 m/s

This agrees with Column (4) of Table 4.14.

The flow based on Hazen – Williams velocity is

3
Q = VA = 0.8828  0.07069  = 0.06241 m /s = 62.41 L/s

This agrees with Column (5) of Table 4.14.

The percentage error in the prediction by Hazen – Williams is

62.41 – 62.38
% error = ------------------------------  100  = 0.05 %
62.38
This agrees with Column (7) of Table 4.14.

___________________________________________________________________________________

123
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Table 4.14 shows that for transitional turbulent pipe flow the Hazen – Williams formula is increasingly
inaccurate, differing from the Darcy – Weisbach computed flow by an overestimation of up to 5.9%.
Thus it is observed in Table 4.14 that the error in the flow from the Hazen – Williams formula ranges
from an underestimation of the flow below a velocity of 0.9 m/s (up to 6.9% error) and an overestima-
tion above a velocity of 0.9 m/s by up to 5.9%. Finally, it is useful to observe the difference in flow esti-
mates from the different formulas for smooth pipe turbulent flow, transitional turbulent flow and fully
rough turbulent flow as shown in Table 4.16.

Table 4.16 Comparison of Darcy – Weisbach computed flows for different flow regimes

For a direct comparison to the Darcy – Weisbach flows in Table 4.16 the corresponding flows as com-
puted from the Hazen – Williams formula are shown in Table 4.17

Table 4.17 Comparison of Hazen – Williams flows for different flow regimes

A comparison of the Darcy – Weisbach friction factors is shown in Table 4.18.

Table 4.18 Comparison of Darcy – Weisbach friction factors for different flow regimes

124
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Finally, the actual variation of Hazen – Williams coefficients corresponding to the friction factors in
Table 4.18 are shown in Table 4.19. Note that a constant Hazen – Williams C value is assumed in com-
puter simulation models. This leads to errors in comparison to the use of the Darcy – Weisbach formula.

Table 4.19 Comparison of actual Hazen – Williams coefficients for different flow regimes
xxxx SPELLING ERROR NEEDS CHANGING IN EXCEL FILE AND RECOPIED IN

4.8 Generalized Form of Head Loss in a Pipe Segment

The general form of the head loss equation of pipe j between nodes i and k as shown in Figure 4.12 is
given as

n–
H i – H k = h f = r jQ j Q j (4.121)
j

where

• Hi = HGL at upstream node i (m or ft)

• Hk = HGL at downstream node k (m or ft)
• h f = head loss along the section of pipe j between nodes i and k (m or ft)
j

• rj = resistance for pipe j that depends on the form of the head loss equation that is used
• Qj = flow assumed to be positive from node i to node k (m3/s or ft3/s)
• n = exponent that depends on the form of the head loss equation

125
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Figure 4.12 Nomenclature for defining head loss in a pipe

The absolute sign in Equation 4.121 assures head loss occurs in the direction of flow and avoids raising
a negative exponent when the flow direction is negative (Bhave 1988). Various head loss equations may
be used for pipe flow. The Darcy – Weisbach and Hazen – Williams head – loss relations have already
been considered. It has also been noted earlier that Darcy – Weisbach head loss equation is recom-
mended for use (rather than the Hazen – Williams head loss equation) as it has a more sound theoretical
basis and better accounts for variation of losses with changes in flow velocity and also variations in vis-
cosity of the fluid.

4.8.1 The resistance term for the Darcy – Weisbach head loss equation
Expressing the Darcy – Weisbach equation of Equation 4.3 with n = 2 in terms of flow yields

fL 2
h f = ------------------ Q (4.122)
2
2gD A

Thus in terms of the general head loss expression of Equation 4.121, the resistance term or “r” value for
the Darcy – Weisbach head loss equation becomes

fL
r f = ------------------ (4.123)
2
2gD A

or

fL 16 fL 8 fL
r f = ------------------------------ = -------------------- = ----------------- (4.124)
2 2 2 5 2 5
D 2 g D  gD
2gD  ----------
 4 

or

8 fL
r f = ----------------- (4.125)
2 5
 gD

Thus Equation 4.122 for the Darcy – Weisbach head loss equation becomes

126
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

h f = r jQ j Q j (4.126)
j

4.8.2 The resistance term for the Hazen – Williams head loss equation
Equation 4.104 for the Hazen – Williams equation formulation is expressed in terms of flow (SI units
using g = 9.81 m/s2) as

L 1.852
h f = 10.670 ------------------------------
-Q (4.127)
1.852 4.871
C D

Thus in terms of the general head loss equation Equation 4.121 then for SI units the resistance term for
the Hazen – Williams equation is

L
r HW = 10.670 ------------------------------
- (4.128)
1.852 4.871
C D

In Equation 4.121 the value of n is

n = 1.852 (4.129)

In US customary units (D and L in feet, Q in ft3/s or cfs using g = 32.2 ft/s2)

L 1.852
h f = 4.7279 ------------------------------
-Q (4.130)
1.852 4.871
C D

such that the resistance term for the Hazen – Williams equation in US Customary units is

L
r HW = 4.7279 ------------------------------- (4.131)
1.852 4.871
C D

Thus Equation 4.121 for the Hazen – Williams head loss equation becomes for pipe j

0.852
h f = rH W Q j Q j (4.132)
j

4.9 Converting from Hazen – Williams C to Roughness Height 

Consider the Swamee and Jain equation from Equation 4.61 for the Darcy – Weisbach friction factor.

Thus rearranging with the negative root being selected rather than the positive root

 - + ------------
5.74- = – 0.25
log 10  ----------- ----------
 3.7D 0.9 f
Re

Now solving for the roughness height gives

 – 0.25
----------
5.74-
 = 3.7D  10
f 
– ------------
  0.9
  Re

127
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Now by substituting in the first expression for f in terms of C from Equation 4.114 gives the required
expression

 – 0.25  D
0.167
C
1.852
V 
0.148
------------------------------------------------------------
  13.643 g  5.74-
 = 3.7D  10  – ------------ (4.133)
  Re 0.9
 

ROUGHNESS HEIGHT  – HAZEN – WILLIAMS C RELATION (SI UNITS)

Example 4.12 Now consider an example where V = 1.0 m/s, D = 300 mm and C = 130. Compute the pipe roughness
height and Darcy – Weisbach friction factor corresponding to the Hazen – Williams C value. Assume the water is at
–6
20°C. The kinematic viscosity of water is 1.007 10 m2/s.

Answer – Computation of the equivalent roughness height.

The Reynolds number is

 1.0   0.3 - = 297 900

Re = VD
-------- = ---------------------------
 1.007 10
–6

The roughness height corresponding to the Hazen – Williams C is

 0.25  0.3
0.167
130
1.852
1.0
0.148

 – ---------------------------------------------------------------------------
 13.643  9.81   5.74
 = 3.7  0.3   10  – ----------------------- = 0.000242 m =0.242 mm
  297900 0.9
 
The Darcy – Weisbach friction factor f from the Swamee and Jain equation (Equation 4.61)

0.25 0.25
f = ---------------------------------------------------------- = ----------------------------------------------------------------------------------- = 0.0199
 
log 10 ------------ + ------------ 5.74  2
 0.242 5.74 - 2
log 10 --------------------- + ---------------------------
 3.7 D 0.9  3.7  300  0.9
Re  297900 

Example 4.13 Consider a water pipeline with an internal diameter D = 1000 mm, and velocity of V = 1.0 m/s.
–6
Assume a water temperature of 15°C and a corresponding kinematic viscosity of 1.140 10 m2/s. Table 4.20
shows the correspondence between Hazen – Williams C and Darcy – Weisbach fluid friction factor f (and the associ-
ated average roughness height value ) for an aged pipe (C = 70), a moderately aged pipe (C = 100) and a new pipe
(C = 140). Confirm the first row of Table 4.20 corresponding to C = 70.

Answer – Confirmation of the first row of Table 4.20.

The Reynolds number is

1.0  1.0 
Re = VD
-------- = --------------------------- = 876400
 1.140 10
–6

From Equation 4.114 solving for the Darcy – Weisbach friction factor f gives

13.643 g
f = ---------------------------------------------
-
0.167 1.852 0.148
D C V

13.643  9.81 
f = -------------------------------------------------------------- = 0.052
0.167 1.852 0.148
 1.0   70  1.0

For a Darcy – Weisbach fluid friction factor of 0.052 and a Reynolds number of 876,400 the corresponding value of
relative roughness from either the Moody diagram (Figure 4.2)or the Swamee and Jain equation (Equation 4.61) is


---- = 0.021
D

The average roughness height for the pipe wall is

 = 0.021 D = 0.021  1000  = 21 mm

One observes that for a halving of the Hazen – Williams coefficient C from 140 to 70 the equivalent roughness height 

128
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

increases by a factor of 150 from 0.14 mm to 21 mm.

___________________________________________________________________________________

Table 4.20 Correspondence between C and f

C D V VD f Corresponding Average
Re = --------
 from Swamee relative roughness
(mm) (m/s) and Jain - height  
Equation 4.61 roughness ---
D (mm)
70 1000 1.0 876,400 0.052 0.021 21
100 1000 1.0 876,400 0.027 0.0029 2.9
140 1000 1.0 876, 400 0.014 0.00014 0.14

4.10 The Manning Equation

In hydro – power the conduit roughness is sometimes given in terms of the Manning n coefficient, how-
ever, it is not common to apply the Manning equation to pipe flow in the analysis of networks. The Man-
ning equation is usually applied to open channel flow situations. In SI units

1.0 2  3 1  2
V = ------- R S f (4.134)
n

THE MANNING EQUATION (SI UNITS)

where

• V = velocity (m/s - for SI Units Equation 4.134 only)

• n = Manning n coefficient
• R = hydraulic radius (m only)
• Sf = friction slope (dimensionless)
For a circular pipe the hydraulic radius is

D
R = ---- (4.135)
4

Substituting Equation 4.135 into Equation 4.134 gives

1.0 D 2  3 1  2
V = -------  ----  Sf (4.136)
n 4

In US customary units the Manning equation becomes

1.49 2  3 1  2
V = ---------- R S f (4.137)
n

THE MANNING EQUATION (US CUSTOMARY UNITS)

where

• V = velocity (ft/s - for US Customary units Equation 4.137 only)

129
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

• n = Manning n coefficient
• R = hydraulic radius (ft only)
• Sf = friction slope (dimensionless)
Again substituting for the hydraulic radius from Equation 4.135 into Equation 4.136 gives

D- 2  3 S 1  2
V = 1.49
----------  --- (4.138)
n 4  f

This equation is only applicable when the Reynolds number is high. Some typical values of the Manning
n coefficient for various materials are shown in Table 4.21.

Table 4.21 Typical Manning n coefficients

Material n value
Planed wood 0.012
Finished concrete 0.012 to 0.015
Steel pipe 0.008 to 0.015

The relationship between Manning's n and the Darcy – Weisbach friction factor f in SI units is

13 13 13

f R fD fD
n = ----------------- = ----------------------------- = ---------------- (4.139)
78.48 13 124.58
78.48  4 

4.11 References
ASCE (1975). Pipeline design for hydrocarbon gases and liquids. Report of the Task Committee on
Engineering Practice in the Design of Pipelines, Committee on Pipeline Planning, Pipeline Division,
American Society of Civil Engineers, New York, 1-79.

Bhave, P. R. (1988). “Extended period simulation of water systems - direct solution." Journal of Envi-
ronmental Engineering, Vol.114, No.5, October, 1146-1159.

Brater, E. F. & King, H.W. (1976). Handbook of hydraulics for the solution of hydraulic engineering
problems. New York: McGraw-Hill.

Carpenter, R. (2010). “Gray iron and ductile iron pipe-some historical benchmarks impacting condition
assessment.” World Environmental and Water Resources Congress, ASCE, May 16-20, Rhode Island.

Christensen, B. (2000). Discussion of “Limitations and proper use of the Hazen-Williams equation.” by
C.P. Liou 124(9), Journal of Hydraulic Engineering, 126(10), 1510-1512.

Colebrook C. F. and White C. M. (1937). “Experiments with fluid friction in roughened pipes.” Pro-
ceedings Royal Society, London A, 161.

Daily, J. W. and Harleman, D. R. F. (1966). Fluid dynamics, Addison.

Jeppson, R. W. (1976). Analysis of flow in pipe networks, Ann Arbor Science, Ann Arbor, Michigan.

Kamand, F. Z. (1988). “Hydraulic friction factors for pipe flow.” Journal of Irrigation and Drainage
Engineering, 114(2), 311-323.

Langelier, W. (1936). “Analytical control of anti-corrosion water treatment.” Journal of American Water
Works Association, 28, 1500.

Liou, C.P. (1998). “Limitations and proper use of the Hazen-Williams equation.” Journal of Hydraulics
Engineering, 124(9), 951-954.

130
 Chapter 4 - File: Chap4newCombined.fm July 6, 2022 Version 6

Skeat, W.O. and Dangerfield, B. J. ed. (1969). Manual of British water engineering practice. (4th ed.).
Cambridge: Published for the Institution of Water Engineers by Heffer.

Stephenson, D. (1984). Pipeflow analysis. Developments in Water Science, No., Elsevier, Amsterdam,
84.

Streeter, V.L., Wylie, E.B. and Bedford, K. (1998). Fluid mechanics. 9th Edition. McGraw – Hill.

Swamee, P. K. (1993). “Design of a submarine oil pipeline.” Journal of Transportation Engineering,

ASCE, 1(1), 159-170.

Swamee, P. K. and Jain, A.K. (1976). “Explicit equations for pipe-flow problems.” Journal of the
Hydraulics Division, 102(5), 657-664.

Tubemakers Australia (1995). Design Manual: Ductile Iron Pipeline Systems, Tubemakers Australia
Limited, Water, Oil & Gas Industries Group.

Wunderlich, W. O., and Giles, J. E. (1986). Review of pipe network analysis. Tennessee Valley Author-
ity, Office of Natural Resources and Economic Development, Division of Air & Water Resources, Engi-
neering Laboratory, September, Report # WR28-2-900-0, 1-98.

Zipparro, V. J., and H. Hansen, H. (1993). Davis' handbook of applied hydraulics, 4th ed., McGraw-Hill,
New York.

131
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

CHAPTER 5 MINOR LOSSES IN PIPE NETWORKS1

5.1 Introduction
In Chapter 4, it was seen that viscosity caused a velocity gradient at fluid boundaries that gave rise to
friction drag and pressure loss. Any change in flow velocity profile causes a loss of energy and result in
a corresponding pressure loss. In addition, any change in direction of flow, cross – sectional area, or
shape causes a drop in energy grade line. Minor losses (also referred to as form losses) are a localized
loss produced by a geometry or shape configuration change. Minor losses are additional to the frictional
loss in a pipeline system. Causes of minor losses include valves (either fully open or in a semi – open
position), branches or bifurcations, T – junctions, changes in pipe sizes, entrances, exits, flow meters,
orifice plates, bends, expansions, contractions, transitions (e.g. a change from a rectangular section to a
circular section or vice versa), slots and partial obstructions.

A minor loss coefficient K is introduced to enable the computation of the localized head loss due to the
fitting. K coefficients usually assume the flow is fully rough turbulent flow.

The head loss associated with non – uniform flow conditions is generally expressed as

2
V
h L = K ------ (5.1)
2g

HEAD LOSS FOR A MINOR LOSS

In general, K is a function of the Reynolds number of the flow.

Note the similarity of Equation 5.1 to the Darcy – Weisbach head loss equation of Equation 4.3.

In long, straight pipes of constant diameter the flow is uniform. In practice, many conditions may cause
non – uniform flow to occur. An abrupt change in magnitude and direction of the velocity characterizes
the occurrence of non – uniform flow. The losses are concentrated and are sometimes called minor
losses. Sometimes, the 'minor' losses may exceed the straight pipe head loss in a particular piping sys-
tem. A typical example occurs in a water treatment plant, where the straight sections of pipe are short
but there are numerous minor loss components in the system.

Deceleration of flow produces large scale turbulence and large head loss, for example, an abrupt expan-
sion. Acceleration of the flow results in relatively very little head loss, for example, a gradual contrac-
tion. Both a sudden expansion and a sudden contraction are shown in Figure 5.1.

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.

132
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Figure 5.1 A sudden expansion and a sudden contraction

Generally the values of K in Equation 5.1 must be determined experimentally. However, the case of a
pipeline expansion and at the exit from a pipeline to a reservoir is an exception where an analytical solu-
tion for K is possible.

The section at which the velocity head is taken must always be specified when computing the head loss
due to the minor loss fitting based on Equation 5.1. The head loss across a valve is shown in Figure 5.2.
Note how the HGL drops across the valve — it is a concentrated loss.

Figure 5.2 HGL variation for a minor loss

5.2 Similitude and Dimensional Analysis for Minor Losses

A relationship is developed between the localized pressure drop p due to a fitting or minor loss in a
pipeline and the significant fluid properties. The important variables are identified to be fluid viscosity
, fluid density , flow velocity V and pipe diameter D. For a fitting in a pipe assume the form of the fol-
lowing relationship reflected via the function F 1 is required

F 1  p   V  D  = 0 (5.2)

where

• p = pressure drop due to a fitting (N/m2 [or Pa] or lbf/ft2)

•  = absolute viscosity of the fluid (Pa – s or slugs/s – ft)
•  = density of the fluid (kg/m3 or slugs/ft3)
• V = velocity (m/s or ft/s)
• D = diameter of pipe (m or ft)
Application of the Buckingham theorem from Section 3.3.2 leads to

n = 5, m = 3, n – m = 2

Thus two dimensionless parameters result and the following function holds

133
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

F 2   1  2  = 0 (5.3)

The repeating variables are usually selected to be associated with the following

• D = a length variable
•  = a variable associated with mass
• V = a kinematic variable associated rate of change with time
For this example, use D, , and V as repeating variables. Solving for the first  term

a1 b1 c1
 1 = D  V p (5.4)

The units of the first  term must be dimensionless. Thus

0 0 0
 1  =  M   L   T 

Solving the exponent equations gives a1 = 0, b1 = – 1 and c1 = – 2. Thus

p
 1 = ---------- (5.5)
2
V

Now, solving for the second  term gives

a2 b2 c2
2 = D  V  (5.6)

Solving the exponents equations gives a2 = – 1, b2 = – 1 and c2 = – 1.

Thus


 2 = ------------ (5.7)
VD

that is the inverse of the Reynolds number, Re.

Thus substituting into Equation 5.3 with the values of dimensionless  terms

F 2   1  2  = 0

gives

p 
F 2  ---------- ------------ = 0 (5.8)
 2 VD
V

or

p
F 3  ---------- Re = 0 (5.9)
 2 
V

where

• Re = Reynolds number (dimensionless)

134
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Thus the pressure drop across a minor loss fitting is a function of Reynolds number, the density of fluid
flowing and of the velocity of flow

2
p = F 4  Re V (5.10)

The head loss hL across the fitting is

p
h L = -------


assuming the velocity head is the same both upstream and downstream of the fitting and from
Equation 5.10 it follows that

2
p V
------- = F 4  Re  ----------
 g

or

2
p V
------- = 2F 4  Re  ------ (5.11)
 2g

2
V
h L = F 5  Re  ------ (5.12)
2g

Replace the function of Reynolds number F5(Re) with the variable K. It should be kept in mind that K
depends on the Reynolds number. Thus

2
hL = K V
-------- (5.13)
2g

5.3 Minor Loss for a Sudden Expansion

Figure 5.3 shows an abrupt expansion. In this section, the force – momentum flux equation is applied to
the control volume in Figure 5.3 to derive an exact analytic expression for the head loss coefficient K for
an abrupt expansion.

From the force – momentum flux equation from Equation 3.37 applied along the center – line of the
pipes and the expansion

 F x = Q  V x2 – V x1  (5.14)

Thus

p 1 A 2 – p 2 A 2 = QV x – QV x (5.15)

2 1

where shear forces are ignored. In addition, for a horizontal sudden expansion gravitational effects are
zero. Note that A 2 , the area of section 2, is applied with the pressure at section 1. Thus

p 1 A 2 – p 2 A 2 =   V x A 2 V x –   V x A 1 V (5.16)
2 2 1 x

From Equation 5.16 it follows that

135
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

2 2
p2 – p1 V 1 A1 V 2
-----------------
- = --------
- ------ – --------- (5.17)
 g A2 g

where the x – subscript has been dropped from the velocity terms.

Figure 5.3 Derivation of loss coefficient for a sudden expansion

Applying the energy equation between sections 1 and 2 leads to

2 2
p V p V
z 1 + -----1- + ------1 = z 2 + -----2- + ------2 + h f + h L (5.18)
 2g  2g

where

z1 = z2

because the sudden expansion is horizontal.

Because the section between 1 and 2 is short, the friction loss is assumed to be zero

h f = 0.0

Thus in Equation 5.18 it follows

2 2
p2 – p1 V1 V2
-----------------
- = --------
- – --------- – h L (5.19)
 2g 2g

Equating both of the equations in Equation 5.17 and Equation 5.19 above gives

2 2 2 2
V 1 A1 V 2 V1 V2
--------- ------ – --------- = --------
- – --------- – h L (5.20)
g A2 g 2g 2g

136
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

2 2 2
V 1 V 2 2V 2 1 A 1 2V 2
hL = --------
- – --------- – ------------ ------ + ------------ (5.21)
2g 2g 2g A 2 2g

But in the third term on the right – hand side

V 1 A1 = Q

Thus

2 2 2
V 1 V 2 2V 1 Q 2V 2
hL = --------
- – --------- – ---------- ------ + ------------ (5.22)
2g 2g 2g A 2 2g

Now substituting for

Q
V 2 = ------
A2

Then it follows that

2 2
V 2 2V 1 V 2 V 1
h L = --------
- – ----------------- + --------- (5.23)
2g 2g 2g

Thus

2
V 2 – V 1
h L = --------------------------
- (5.24)
2g

Now using the continuity equation applied between section 1 and section 2 in Figure 5.3 gives

V 1 A1 = V 2 A2

V2 is now eliminated and it follows that

A
V 2 = V 1 -----1-
A2

2 2
V 2 – V 1 A 2V 1
- =  1 – -----1- --------
h L = -------------------------- - (5.25)
2g  A 2 2g

Thus the head loss for an abrupt expansion is

2 2
V1 A 2V 1
- =  1 – -----1-  --------
h L = K -------- - (5.26)
2g  A 2  2g

HEAD LOSS FOR A SUDDEN EXPANSION

137
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Thus for a sudden expansion the loss coefficient is

A 2
K =  1 – -----1-  (5.27)
 A  2

HEAD LOSS COEFFICIENT FOR A SUDDEN EXPANSION

Note the loss coefficient is based on the velocity head in the upstream pipe with the smaller diameter
(i.e. the faster velocity).

Example 5.1 For an abrupt expansion of D1 = 500 mm and D2 = 1000 mm for a flow rate of 392.7 L/s.

(i) Compute the head loss coefficient

(ii) Compute the head loss.

Answer – (i) Computation of the head loss coefficient for an abrupt expansion.

The upstream area (smaller) is

2
D   0.5 
2
2
A 1 = ---------1- = ------------------ = 0.19635 m
4 4

The upstream velocity (faster) is

Q 0.3927
V 1 = ------ = ------------------- = 2.0 m/s
A1 0.19635

The downstream area (larger) is

2
D   1.0 
2
2
A 2 = ---------2- = ------------------ = 0.7854 m
4 4

The downstream velocity (slower) is

Q 0.3927
V 2 = ------ = ---------------- = 0.5 m/s
A2 0.7854

The abrupt expansion head loss coefficient from Equation 5.27 is

2 2
A 2  D  500 2 2 1 2
K =  1 – -----1- =  1 – ------1- =  1 –  ------------  =  1 – --- = 0.5625
 A2   2   1000   4
D 2

___________________________________________________________________________________
Answer – (ii) Computation of the abrupt expansion head loss.

The head loss is based on the faster upstream velocity from Equation 5.26 as

2 2 2
V1 A 2V 1  2.0 
- =  1 – -----1- --------
h L = K -------- - = 0.5625 ------------------ = 0.115 m
2g  A2 2 g  2  9.81 

___________________________________________________________________________________

5.3.1 Minor loss for an exit from a pipe to a downstream reservoir

For an exit from a pipeline to a reservoir the ratio

D
------1-  0
D2

Thus for the sudden expansion in the previous section in Equation 5.27 it follows that

138
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

A
-----1-  0
A2

and

A 2
K =  1 – -----1-  = 1.0
 A  2

Thus for an exit from a pipe to a downstream reservoir

2
V1
h L = 1.0 --------
- (5.28)
2g

5.4 Minor Losses for Sudden Contractions

The losses associated with a contraction are much smaller than for a sudden expansion. This is due to
the fact that converting pressure head into velocity head is very efficient (accelerating flow). In fact, the
jet actually contracts to a Vena contracta and then expands again slightly as shown in Figure 5.4. It is
during this expansion coming out of the vena contracta that the energy loss occurs. The analysis pro-
ceeds in the same way as for the sudden expansion as shown above by applying both the force –
momentum flux equation and the energy equation.

Figure 5.4 Loss coefficient for a sudden contraction

From an analytical analysis the head loss for the contraction is shown to be

2
A 2V o
h L =  1 – -----o- --------
- (5.29)
 A  2g2

Now apply the continuity equation between the vena contracta at section 0 and section 2 downstream of
the contraction in Figure 5.4.

Thus

V o Ao = V 2 A2 (5.30)

Now define the contraction coefficient Cc as the ratio of the area of the narrow part of the flow at the
vena contracta to the area of flow in the pipe downstream of the contraction as

A
C c = -----o- (5.31)
A2

139
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Experimental values of the contraction coefficient Cc for different area ratios are given in Table 5.1.

Table 5.1 Contraction coefficients

A Cc
------2
A1
0.1 0.624
0.2 0.632
0.3 0.643
0.4 0.659
0.5 0.681
0.6 0.712
0.7 0.755
0.8 0.813
0.9 0.892
1.0 1.0

Solving in Equation 5.31 for the area A o gives

Ao = C c A2 (5.32)

Substituting Equation 5.32 into Equation 5.30 gives

V o C c A2 = V 2 A2

or

V oCc = V 2

V
V o = ------2 (5.33)
Cc

Combining Equation 5.29, Equation 5.31 and Equation 5.33 above leads to a head loss given by

2 2 2 2
A 2V o 2 V 2  1 1 – C 2V 2 1 2 V 2 
hL =  1 – -----o- -------- - ------ =  --------------c- --------
- =  1 – C c   -------- - =  ------ – 1  --------
-
 A 2 2g C  2 2g  C  2g  C   2g 
c c
c

2 2
V2 1 2 V 2 
- =  ------ – 1  --------
  2g 
h L = K -------- - (5.34)
2g C
c

HEAD LOSS FOR AN ABRUPT CONTRACTION

Note the velocity head is based on the downstream pipe velocity (i.e. the faster velocity). This contrasts
to the expression for the abrupt expansion that is based on the upstream pipe velocity (however, the
faster velocity was used in that case as well).

140
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Thus for a sudden contraction the loss coefficient is

1 2
K =  ------ – 1 (5.35)
C 
c

HEAD LOSS COEFFICIENT FOR AN ABRUPT CONTRACTION

The contraction coefficient Cc and thus the K value for a contraction is a function of Reynolds number

Example 5.2 For an abrupt contraction of D1 = 1000 mm and D2 = 500 mm for a flow rate of 392.7 L/s.

(i) Compute the head loss coefficient

(ii) Compute the head loss.

Answer – (i) Computation of the head loss coefficient for a contraction.

The upstream area (larger) is

2
D 2
  1.0  - = 0.7854 m 2
A 1 = ---------1- = -----------------
4 4

The upstream velocity (slower) is

Q- = 0.3927
V 1 = ----- ---------------- = 0.5 m/s
A1 0.7854

The downstream area (smaller) is

2
D 2
  0.5  - = 0.19635 m 2
A 2 = ---------2- = -----------------
4 4

The downstream velocity (faster) is

Q 0.3927
V 2 = ------ = ------------------- = 2.0 m/s
A2 0.19635

The area ratio for the abrupt contraction is

A 2
500
-----2- = -------------- = 0.25
A1 1000
2

The contraction coefficient from Table 5.1 gives

C c = 0.637

The abrupt contraction head loss coefficient from Equation 5.35 gives

1
K =  -----
2 1 - – 1 2 =  1.57 – 1  2 = 0.325
- – 1 =  ------------
C   0.637 
c

This compares to a value of K = 0.5625 for the abrupt expansion as computed in Example 5.1. Thus for the same area
ratio the head loss coefficient for an abrupt expansion is greater than that for an abrupt contraction.

___________________________________________________________________________________
(ii) Computation of the head loss for an abrupt contraction.

The abrupt contraction head loss from Equation 5.34 (based on the faster downstream velocity) is

2 2
V2 1 2V 2   2.0 
2
- =  ------ – 1  --------
  2 g 
h L = K -------- - = 0.325 ------------------ = 0.066 m
2g C 2  9.81 
c

This compares to a value of h L = 0.115 m for the abrupt expansion as computed in Example 5.1.

___________________________________________________________________________________

141
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

5.5 Minor Losses for Entrances

5.5.1 Square edged entrance

Figure 5.5 A square edged entrance leading out of a reservoir

The coefficient of contraction Cc is approximately 0.6.

Example 5.3 Compute the head loss coefficient for the entrance to a pipe from a reservoir.

Answer – Computation of head loss coefficient for an abrupt contraction into a pipe from a reservoir.

The abrupt contraction head loss coefficient from Equation 5.35 gives

1 2 1 2
K =  ------ – 1 =  ------- – 1 =  1.67 – 1  = 0.45
2
C   0.6 
c

It is assumed that the entrance is square – edged.

___________________________________________________________________________________

Thus as an approximation the energy loss due to the entrance to a pipe is given by

2 2
V2 V2
hL = K --------
- = 0.5 --------
- (5.36)
2g 2g

where K is taken to be 0.5.

5.5.2 A re – entrant entrance

Figure 5.6 Figure 3.34 A re – entrant entrance

The loss coefficient is taken as K = 1.0 for a re – entrant pipe with thin walls such that

2
V
h L = 1.0 ------ (5.37)
2g

A well – rounded or smooth entrance is shown in Figure 5.7.

142
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Figure 5.7 A smooth entrance

Head loss coefficients were determined experimentally as K = 0.01 to 0.05 while the head loss is in the
range shown below.

2 2
V V
h L = 0.01 ------ to 0.05 ------ (5.38)
2g 2g

5.6 Minor Losses Due to Gradual Changes in Cross – section

5.6.1 A gradual contraction

Figure 5.8 Gradual contraction

For a carefully formed contraction the loss of energy is in most practical case is negligible. Thus

K = 0.0 (5.39)

The amount of loss clearly depends on the angle of the gradual contraction. Table 5.2 gives the approach
to computing the loss coefficient as the angle increases.

Table 5.2 Loss coefficient for varying angles of contraction

Angle between two contracting K
sides (degrees)
0 to 30 0
30 to 45 0.05
45 to 180 Compute from sudden contraction formula (Equation 5.35)

5.6.2 A gradual expansion

This is an important flow case in which there may be quite large energy losses. Gibson's experimental
work for values of K versus angle of expansion are given in Figure 5.9.

The head loss is computed from

2 2
V 2 – V 1 A 2V 1
- =  1 – -----1- --------
h L = -------------------------- - (5.40)
2g  A 2 2g

143
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Figure 5.9 A gradual expansion

5.7 Bends
Head losses in pipe bends are caused by the combined effect of separation, wall friction and twin – eddy
secondary flow. The change in flow direction due to the bend results in an increase in pressure around
the outside of the bend and a decrease in pressure around the inside of the bend.

Figure 5.10 Losses in bends

At higher velocities separation occurs at the inner boundary. This pressure difference may be utilized for
flow measurement. The centrifugal acceleration causes a secondary circulation at the bend. For large
radius bends both wall friction and secondary flow are significant. For small radius bends both separa-
tion and secondary flow are significant. Coefficients for bends are experimental and often conflicting. A
researcher called Ito carried out some experimental work in the 1960s. Details of bend loss coefficients
for various types of bends is given in Table 5.4.

Figure 5.11 Secondary circulation in bends

144
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

5.8 Valves
Discrete or localized head losses occur at fittings such as valves, elbows, T – junctions, flow meters etc.
Normally the loss values K are provided by the manufacturer. Many different types of valves are manu-
factured. A summary of sizes of various types of valves are given in Table 5.3. Loss coefficients for
valves are given at the end of Table 5.4.

Table 5.3 Size range for various valves

Valve type Size range
(mm)
Gate valve 150 to 750
Butterfly valve 150 to 1200
Check valve 150 to 750
Globe valve 150 to 900
Angle valve 150 to 900

5.8.1 Globe valve

A globe valve is a stop cock valve that usually comes only in smaller sizes. The flow is turned through
90 degrees so there are considerable losses as the flow path is tortuous even when the valve is fully
opened. For a globe valve, fully opened K = 10.0.

5.8.2 Plug valve

A plug valve is used for rain water outlet valves. For a plug valve fully open K = 0.8.

5.8.3 Gate or slide valve

A slide gate has a clear passage when fully open with a similar shape and size as the pipe, but with
recesses where the gate slides down and also seats that cause some loss. A fully open gate valve offers
very little resistance thus the K value is quite low. This type of valve is prone to cavitation at high heads
above 50 m. For a gate valve fully open K = 0.2. For a gate valve half open K = 5.0.

5.8.4 Butterfly valve

Often used as an emergency or guard gate valve. A fully opened butterfly valve has a K = 0.3.

5.8.5 Check valves or non – return valves

A swing check valve has a head loss coefficient of K = 0.6 to 2.5.

5.9 Typical Head Loss Coefficient Values

Head loss coefficients for a range of fittings are given in Table 5.4.

Table 5.4 Pipeline minor loss coefficients (based on Skeat and Dangerfield 1969, p. 165
and ASCE 1975, p.21)
Type of Fitting K L/D
approx.
(f = 0.023)
Entry Losses
Sharp – edged entrance 0.5 22
Re – entrant entrance 0.8 35

145
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Table 5.4 Pipeline minor loss coefficients (based on Skeat and Dangerfield 1969, p. 165
and ASCE 1975, p.21)
Slightly – rounded entrance 0.25 11
Bellmouthed or well – rounded entrance 0.05 2
Footvalve and strainer 2.5 109
Intermediate Losses
Elbows
45o standard 0.5 21

90o standard 0.7 29

Elbows (R/D = 1/2 approximately)
22.5o 0.2 9

45o 0.4 17

90o 1.0 43
Close Radius Bends (R/D = 1 approximately)
22.5o 0.15 7

45o 0.3 13

90o 0.75 33
Long Radius Bends (R/D = 2 to 7)
22.5o 0.1 4

45o 0.2 9

90o 0.4 17
Sweeps (R/D = 8 to 50)
22.5o 0.05 2

45o 0.1 4

90o 0.2 9
Miter Elbows
22.5o – 2 piece 0.15 7

30o – 2 piece 0.20 9

45o – 2 or 3 piece 0.30 13

Type of Fitting K L/D
approx.
(f = 0.023)
60o – 2 piece 0.65 28

60o – 3 piece 0.25 11

90 o – 2 piece 1.25 54

90 o – 3 piece 0.50 21

90 o – 4 piece 0.30 13
Tees
Flow in line 0.35 15

146
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Table 5.4 Pipeline minor loss coefficients (based on Skeat and Dangerfield 1969, p. 165
and ASCE 1975, p.21)
Line to branch or branch to line:
– sharp – edged 1.2 to 1.8 52 to 78
– radiused 0.8 35
Angle Branches
Flow in line 0.35 15
Line to branch or branch to line:
– 30o angle 0.40 17

– 45o angle 0.60 26

– 90o angle 0.80 35

Sudden expansions or enlargements
Inlet diameter: Outlet diameter Ratio
– for a ratio of 4:5 0.15 7
– for a ratio of 3:4 0.20 9
– for a ratio of 2:3 0.35 15
– for a ratio of 1:2 0.60 26
– for a ratio of 1:3 0.80 35
– for a ratio of 1:4 0.90 39
– for a ratio of 1:5 and over 1.00 44
Sudden Contractions*
Inlet diameter: Outlet diameter ratio
– for a ratio of 5:4 0.15 7
– for a ratio of 4:3 0.20 9
– for a ratio of 3:2 0.30 13
– for a ratio of 2:1 0.35 15
– for a ratio of 3:1 0.45 20
– for a ratio of 5:1 and over 0.50 21
British Standard (B.S.) Tapers*
Flow to small end negligible
Flow to large end
Inlet diameter: Outlet diameter ratio
– for a ratio of 4:5 0.03 1.3

Type of Fitting K L/D

approx.
(f=0.023)
– for a ratio of 3:4 0.04 1.7
– for a ratio of 1:2 0.12 5.2
Valves
Gate Valve – fully open 0.12 to 0.2 5.2 to 9
Gate Valve – ¼ closed 1.0 to 1.2 44 to 52
Gate Valve – ½ closed 5.6 to 6.0 243 to 260
Gate Valve – ¾ closed 24.0 1040

147
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

Table 5.4 Pipeline minor loss coefficients (based on Skeat and Dangerfield 1969, p. 165
and ASCE 1975, p.21)
Globe valve (fully open) 10.0 435
Right angle valve (fully open) 5.0 217
Reflux valve or swing check valve or 0.6 to 2.5 26 to 110
non – return valve
Angle valve 2.5 110
Butterfly valve 0.3 13
Exit losses
Sudden expansion or enlargement 1.0 44
Bellmouthed outlet 0.2 9

* Figures for expansions or enlargements, contractions and B.S. Tapers apply to velocity head for smaller diameter.

5.10 Equivalent Pipe Lengths

The expression for minor losses is given by Equation 5.1. Minor losses may be included in the analysis
of pipeline systems by using the concept of equivalent pipe length. Mathematically, a minor loss caused
by a fitting or configuration change may be considered as an equivalent length of pipe that would have
the same losses at a given flow. An approximation is made in the analysis if an equivalent pipe length is
used to replace a fitting.

Equate the friction loss equation of Equation 4.3 to the head loss given by Equation 5.1 as

2
LV
h f = f ---- ------
D 2g

Thus it then follows that

L V2 V
2
f -----e -------- = K --------
D 2g 2g

Thus the equivalent length is given by

D
L e = K ---- (5.41)
f

EQUIVALENT LENGTH OF PIPE FOR A MINOR LOSS

Typically an assumed value of f must be used. Usually f = 0.020 is a good first estimate. This should be
checked and another iteration performed if the actual value of friction factor or resitance factor in the
pipe is very different from the assumed value.

5.11 When May Minor Losses be Ignored?

The minor losses due to valves, bends, fittings etc. become relatively small (less than 10% of the total)
when long pipes are involved. That is when

L
----  150 (5.42)
D

where

• L = distance between fittings (m or ft)

148
 Chapter 5 - File: Chap5new.fm July 6, 2022 Version 6

• D = diameter of pipe (m or ft)

For a 100 mm pipe from Equation 5.42 this is only 15 m.

Therefore minor losses may be ignored for all practical purposes at least as a first approximation pro-
vided the pipeline is long.

5.12 References
ASCE (1975). Pipeline design for hydrocarbon gases and liquids. Report of the Task Committee on
Engineering Practice in the Design of Pipelines, Committee on Pipeline Planning, Pipeline Division,
American Society of Civil Engineers, New York, 1-79.

Skeat, W.O. and Dangerfield, B. J. ed. (1969). Manual of British water engineering practice. (4th ed.).
Cambridge, Published for the Institution of Water Engineers by Heffer.

149
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

CHAPTER 6 ANALYSIS OF SIMPLE PIPELINES1

This Chapter deals with pipe problems in terms of simple single pipelines. Simple pipe problems
involve the analysis or design of a single pipeline with one diameter. It is assumed that one of three val-
ues is unknown for the pipe either

• head loss h f occurring along the pipe connecting two reservoirs (or z 1 – z 2 where z 1 and
z 2 are the elevations of the upstream and downstream reservoirs respectively)
• discharge Q (or flow) in the pipe
• diameter D of the pipe
The continuity and energy equations (with a pipe friction loss method) are used to solve simple pipe
problems. Steady flow is assumed such that Q in = Q out

where

• Q in = flow into the pipeline

• Q out = flow out of the pipeline

Pipe friction is often the dominant loss considered. Both minor losses due to fittings and the velocity
head are usually ignored. This Chapter does consider problems that both ignore and consider minor
losses.

There are six variables associated with solving simple pipe problems including

• Q = discharge or flow in the pipe (m3/s or ft3/s)

• h f = the head loss along the pipe (or the difference between upstream and downstream res-
ervoir elevations z 1 – z 2 where z 1 and z 2 are the elevations of the upstream and down-
stream reservoirs respectively) (m or ft)
• L = the length of the pipe (m or ft)
• D = the diameter of the pipe (m or ft)
•  = the average roughness height on the inside of the pipe (or Hazen – Williams coeffi-
cient) (mm or ft)
•  = the kinematic viscosity of the fluid flowing (m2/s or ft2/s)

Usually the pipe length, roughness height and kinematic viscosity (L  and  are given for a simple
pipe problem.

The first two of the simple pipe problems mentioned above are referred to as analysis problems where
an existing or assumed pipeline of known diameter is being considered and the unknown to be solved
for is either h f or Q. In the unknown head loss h f problem the solution is determined explicitly. The
Reynolds number and the relative roughness are determined and in turn the Darcy – Weisbach friction
factor or resistance factor are determined directly. For the computation of the discharge Q in a single

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Ade-
laide, Australia. Part of the book entitled Water Distribution Systems Engineering.

150
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

pipeline there are two approaches — an explicit approach based on an assumed turbulent flow regime
type and a trial and error approach for the case where minor losses are considered. Analytic or explicit
expressions are derived in this Chapter for three cases of different types of turbulent flow regime includ-
ing

• smooth pipe turbulent flow

• transitional turbulent flow
• fully turbulent pipe flow
Examples are also given. The third problem is referred to as a design or synthesis problem where the
diameter D of the pipe is unknown. Again there are two possible solution approaches — an explicit solu-
tion procedure and a trial and error approach.

6.1 Unknown Pipe Head Loss — an Explicit Expression

Assume the following are given including: Q, L, D,  and  .

Solve for the head loss h f as shown in Figure 6.1 This is an analysis problem and is straight forward
where the head loss h f (or difference between the two reservoirs) are solved for directly. Because the
solution technique is explicit, it is possible to easily incorporate minor losses in the computation.

Figure 6.1 Solving for head loss (assuming minor losses are negligible)

Use the continuity equation, the energy equation and the Darcy – Weisbach head loss equation to solve
for the unknown head loss. An assumption of steady state flow is made such that

Q in = Q out

The procedure to solve for the unknown head loss is as follows:

1. Draw the HGL between the reservoirs.

2. Write the energy equation for the system from a point on the surface of reservoir 1 to a point
on the surface of reservoir 2. This is a crucial step. In Equation 6.1 minor losses are
included.

2 2 2
p V p V LV
z 1 + -----1- + ------1 = z 2 + -----2- + ------2 + f ---- ------ +  h L (6.1)
 2g  2g D 2g

At the surfaces of the reservoirs, the velocities are V1 = V2 = 0, and the pressures at the
water surfaces are p1 = p2 = 0.0 and thus it follows that

2
LV
z 1 = z 2 + f ---- -------- +  h L (6.2)
D 2g

151
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

3. Compute the Reynolds number. Recall the Reynolds number from Equation 2.15 for the
known or specified discharge is defined as

VD QD 4Q
Re = -------- = --------------------- = ----------- (6.3)
 2 D
 D
----------  
 4 

In this case, the Reynolds number is computed immediately from the known Q,Dand  .

4. Compute the friction factor or resistance factor. Once the Reynolds number is known, then

along with the relative roughness ---- the Moody diagram (Figure 4.2) or the Swamee and
D
Jain equation (Equation 4.61) are used to compute the Darcy – Weisbach friction factor f.

5. Compute the velocity of the pipe flow as

Q
V = ----
A

6. Compute the unknown head loss: the unknown elevation difference between the reservoirs
from Equation 6.2 is

2
LV
z 1 – z 2 = f ---- -------- +  h L (6.4)
D 2g

UNKNOWN HEAD LOSS FOR A PIPE

If minor losses for an entrance and an exit loss are included it follows that
V V 2 2
 h L = K entr -----
2g
- + K exit ------
2g

Typical applications of this simple problem include

• how high one reservoir needs to be above another in order to deliver a particular flow
• how big a pump needs to be
Example 6.1 For discharge between two reservoirs of Q = 125 L/s of water at 15C through commercial steel pipe
300 m in length and 300 mm diameter. Include minor losses. Assume a sharp edged inlet. (i) Calculate the head loss.

Answer – (i) Computation of head loss between the two reservoirs.

Write the energy equation between a fluid particle on the surface of the two reservoirs as shown in Figure 6.1

2 2 2
p V p V LV
z 1 + -----1- + ------1 = z 2 + -----2- + ------1 + f ---- ------ +  h L
 2g  2g D2g

Now, at the surfaces of the reservoirs

p 1 = 0.0, V 1 = 0.0, p 2 = 0.0, V 2 = 0.0

and it follows that

2
fL- ------
V + h
D 2g  L
z 1 = z 2 + -----

The pipe area is

2 2
A =  D- = 
---------  0.3  - = 0.07069 m 2
-----------------
4 4

152
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

The velocity of flow is

Q 0.125
V = ---- = ------------------- = 1.77 m/s
A 0.07069

–6
At 15C the kinematic viscosity of water is  = 1.140 10 m2/s, thus the Reynolds number for the flow is

VD 1.77  0.3  5
Re = -------- = --------------------------- = 464960 = 4.65 x 10
 1.140 10
–6

For commercial steel   = 0.046 mm and the relative roughness is

- = 0.046
--- ------------- = 0.00015
D 300

From the Moody diagram (Figure 4.2) f = 0.015 for the values of Re and /D.

For comparison, the Darcy – Weisbach friction factor or resistance factor from the Swamee and Jain equation
(Equation 4.61) is

0.25 0.25
f = ---------------------------------------------------------
- = ------------------------------------------------------------------------------------------
- = 0.0152
  5.74  2  0.046 5.74  2
log 10 ------------ + ------------ log  --------------------- + -----------------------------------
- 
 3.7 D 0.9 10
Re  3.7  300   4.65 x 10 5  0.9

Thus the Moody diagram value is confirmed.

The head loss in the pipeline is

2 2
LV 0.015  300   1.77 
h f = f ---- ------ = --------------------------- ------------------ = 2.40 m
D2g 0.3 2  9.81 
Thus the resultant pipe friction loss is

h f = 2.40 m

Minor losses for the entrance to pipe and the exit from the pipe are computed from

V- 2 + K -----
V2
 hL = K entr -----
2g exit -
2g

 1.77  2
 hL =  0.5 + 1.0  ------------------ = 0.24 m
2  9.81 
Thus the total losses between the two reservoirs (that is also equal to the difference in elevation between the reservoirs)
is

h f +  h L = 2.40 + 0.24 = 2.64 m

___________________________________________________________________________________

6.2 Unknown Pipe Discharge – Ignoring Minor Losses

Figure 6.2 Solving for the unknown discharge (assuming minor losses are negligible)

153
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

In the unknown discharge problem for a simple pipe, assume the following are given including h f , L, D,
 and  .

Again, as for the unknown head loss simple pipe problem, the unknown discharge problem is an analy-
sis problem because the diameter of the pipe is known. Two approaches to the unknown discharge prob-
lem are presented in the following sections. In the first, one of the three turbulent flow regime types
(smooth pipe turbulent, transitional turbulent or fully rough turbulent) is assumed and then an explicit
formula for the discharge is used. Each of the appropriate formulas is derived in this Chapter. A check
then needs to be made as to whether the assumption about the turbulent flow regime was correct. It is
recommended to assume the flow is transitional turbulent flow initially. In the second approach, a trial
and error approach is used if minor losses are included. The Reynolds number cannot be computed
directly. Thus both Q and the Darcy – Weisbach f are unknown. It is necessary to iterate between Q and f
until there is negligible change.

Explicit expressions are derived for the discharge for each of the three turbulent flow regimes — smooth
pipe turbulent flow, transitional turbulent flow and fully rough turbulent flow.

6.2.1 Explicit expression for Q assuming smooth pipe turbulent flow

In this section, the expression for Q for smooth pipe turbulent flow is derived.

Take the Darcy – Weisbach head loss expression given by Equation 4.3 and rearrange in terms of dis-
charge Q as follows

2 2 2
L Q L Q 16 f LQ 8 f LQ
f = f ---- ------------- = f ---- -------------------------- = ------------------------- = --------- ------------
D 2g A 2 D D
2 2
D 2g D
2 4
 gD
2 5
2g  ----------
 4 

Solving for the friction factor or resitance factor f

2 5
 gh f D
f = ------------------------ (6.5)
2
8LQ

Taking the square root of both sides of Equation 6.5 gives

5
 gh f D
f = -------------------------
- (6.6)
8LQ

Inverting Equation 6.6 gives

1 8LQ
---------- = -------------------------- (6.7)
f  gh f D
5

Taking the von Karman expression for smooth pipe turbulent flow friction factor from Equation 4.65

1
---------- = 2.0log 10  Re f  – 0.8
f

Now expressing the Reynolds number in terms of discharge Q gives

154
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

 
1  QD 
---------- = 2.0log 10  ------------------- f  – 0.8 (6.8)
f   D 2 
   ---------- 
4

Substituting Equation 6.6 and Equation 6.7 into Equation 6.8 to remove the Darcy – Weisbach fluid fric-
tion factor gives

 5
8LQ  QD  gh f D 
-------------------------- = 2.0log 10  ------------------- -------------------------- – 0.8
 gh f D
5   D 2 8LQ 
   ---------- 
4

and solving for Q yields an expression for discharge in a pipe when conditions are smooth pipe turbulent
flow as

5 5
 gh f D   4 gh f D  
-  2.0log 10  ------- ---------------------
Q = ------------------------- - – 0.8
8L   D 8L  

Simplifying gives

 gh f D   16gh D 5 
5
- 2.0log 10  ----------------------------
Q = ------------------------- f  – 0.8
8L  
 8L D 
2 2  


and thus the expression for the unknown discharge in a pipe that has a smooth pipe turbulent flow
regime is

5 3
gh f D   2gh f D  
Q =  -----------------
-  2.0log 10  --------------------
- – 0.8 (6.9)
8L   2
 L  

DISCHARGE FOR SMOOTH PIPE TURBULENT FLOW

Upon using Equation 6.9, be very careful to make sure that you square the kinematic viscosity value
when carrying out the computations.

Recall from Chapter 4 (Equation 4.71), that the Equation 6.9 is valid for

V f
-----------------  5 (6.10)
8

REQUIREMENT FOR SMOOTH PIPE TURBULENT FLOW*

* Note that units of  must be in meters (or feet) and not mm.

155
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

An alternative form is as follows.

 f
------------- Re  5 (6.11)
D 8

REQUIREMENT FOR SMOOTH PIPE TURBULENT FLOW*

* Note that units of  and D must be consistent, either both in meters (or feet) or both in mm (or in.).

To apply the explicit solution procedure for Q (where the flow is assumed to be smooth pipe turbulent
flow) the following steps apply

1. Draw the HGL.

2. Write the energy equation for the system as shown in Equation 6.1.
3. Compute the discharge from the explicit equation derived above (Equation 6.9).
4. Check assumption. Check whether the assumption about the flow regime is correct by com-
puting the dimensionless number for the viscous sublayer. For the flow to be smooth pipe
turbulent flow the following must hold


------------  5 (6.12)
  u*

or

V f
---------------  5 (6.13)
 8

5. Recompute discharge if necessary. If the dimensionless number in Equation 6.12 is not less
than 5 then recompute the discharge by using the explicit expression for the appropriate tur-
bulent flow regime.
Example 6.2 For a reservoir – pipe – reservoir system shown in Figure 6.3 with D = 300 mm, L = 500 m, z1 = 50 m,
–6
z2 = 45 m and v = 1.007 10 m2/s (water at 20C ). Assume  = 0.003 mm for uPVC pipe. Ignore minor
losses. Based on the value of the roughness height for the pipe assume the type of turbulent flow to be smooth pipe
turbulent.

(i) Compute the discharge.

(ii) Compute the Darcy – Weisbach friction factor or resistance factor using the Swamee and Jain formula.
(iii) Confirm the friction factor for smooth turbulent pipe flow from the von Karman formula.
(iv) Check that the flow regime is in fact – smooth turbulent pipe flow.

Answer – (i) Computation of the discharge between the two reservoirs – smooth turbulent pipe flow.

The two – reservoir system is shown in Figure 6.2.

Apply the energy equation between points 1 and 2 in Figure 6.2

2 2 2
p V1 p V2 LV
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- + f ---- ------
 2g  2g D2g

At the surfaces of the reservoirs the velocity is zero or V = 0 and the pressure p = 0 hence it follows that

2
L- ------
V
z 1 + 0 + 0 = z 2 + 0 + 0 + f ---
D2g
Thus it follows that

2
LV
f ---- ------ = z 1 – z 2 = 50 – 45 = 5 m;
D2g
Thus the head loss is

156
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

hf = 5 m

The pipe area is

2 2
A = 
---------- =   0.3  - = 0.07069 m 2
D
-----------------
4 4

The discharge is computed from Equation 6.9 as

5 3
 gh f D   2 gh f D    9.81  5.0   0.3  -  2.0log  ------------------------------------------------
5
2  9.81   5.0   0.3  
3 
-  2.0log 10  --------------------
Q = ------------------------- - – 0.8 = ---------------------------------------------  10  - – 0.8
2 2
8L    L   8  500     1.007 10–6   500  

A spreadsheet given in Figure 6.3 is presented to implement the solution of this problem. From the spreadsheet, the dis-
charge in the pipeline is

3
Q = 0.153 m  s

___________________________________________________________________________________
Answer – (ii) Computation of the Darcy – Weisbach friction factor or resistance factor using the Swamee and
Jain formula (Equation 4.61).

The velocity is

V = Q 0.153 - = 2.164 m/s

---- = ------------------
A 0.07069

The Reynolds number is

2.164  0.30 
Re = VD
-------- = ----------------------------
- = 644569
 1.007 10
–6

An estimate of the friction factor using the Swamee and Jain formula (Equation 4.61) is

0.25 0.25
f = ---------------------------------------------------------- = ------------------------------------------------------------------------------------------- = 0.0127
  5.74  2  0.003 5.74  2
log 10 ------------ + ------------ log 10  --------------------- + ----------------------------------- -
 3.7 D 0.9
Re  3.7  300   6.45 x 10 5  0.9

___________________________________________________________________________________
Answer – (iii) Confirmation of the friction factor for smooth pipe turbulent flow from the von Karman for-
mula.

The turbulent smooth pipe friction factor f is given by Equation 4.65 as

1
-------- = 2.0log 10  Re f  – 0.8 (6.14)
f

The use of Goal Seek in Microsoft Excel to solve Equation 6.14 for the Darcy – Weisbach friction factor is shown in
Figure 6.4. To initiate the Goal Seek function set up the spreadsheet shown in Figure 6.4 and then complete the follow-
ing steps

1. Put a guess of the friction factor or resistance factor in cell E9 (for example, 0.02)
2. Go to “Tools” on the Microsoft Excel toolbar and select “Goal Seek ...”
3. Set up the three values in the Goal Seek box shown in Figure 6.4.
4. Click on the OK box
5. Cell E9 changes to 0.01257 representing the solution to friction factor in the von Karman equation in cell E11
(Equation 6.14). Note that the values E9 and E11 change in the Goal Seek box in Figure 6.4 depending on the
actual reference of the cells in the spreadsheet.

Using Goal Seek in Microsoft Excel to solve for f in Equation 6.14 gives

f = 0.0126

157
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

Figure 6.3 Spreadsheet for computation of smooth pipe turbulent flow (unknown Q) by explicit formula

Figure 6.4 Spreadsheet for using Goal Seek in Microsoft Excel to compute smooth pipe friction factor

This confirms the value of the Darcy – Weisbach equation from the Swamee and Jain equation (Equation 4.61). The
computations shown in the spreadsheet in Figure 6.3 have been verified by checking the equality in Equation 6.14.

___________________________________________________________________________________
Answer – (iv) Checking that the flow regime is in fact smooth turbulent pipe flow.

Compute the dimensionless quantity associated with the viscous sublayer

 0.003
------------- 2.164 0.0126
V f  1000 
--------------- = ------------------------------------------------------ = 0.26  5
–6
 8 1.007 10 8

Thus the assumption of smooth pipe turbulent flow is OK.

158
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

___________________________________________________________________________________

6.2.2 Explicit expression for Q assuming transitional turbulent pipe flow

An explicit expression for the discharge is computed in this section assuming the flow is in the transi-
tional turbulent regime. The equation is derived by commencing with Equation 4.76 for the Colebrook –
White formula for the friction factor or resistance factor.

Recalling that rearranging the Darcy – Weisbach head loss equation in terms of discharge from
Equation 6.6 leads to

1 8LQ
---------- = -------------------------- (6.15)
f  gh f D
5

Combining Equation 4.76 and Equation 6.15 to eliminate the friction factor f leads to

8LQ   2.51 8LQ 

-------------------------- = – 2.0 log 10  ------------ + ---------- --------------------------
 gh f D
5  3.7D Re  gh D 5 
f

The Reynolds number from Equation 2.15 is expressed as

Re = VD
-------- = QD QD 4Q-
--------- = ------------------- = ----------
 A D
2 D
  ----------
 4 

Combining the two equations above, simplifying and solving for the discharge gives

– 2.0-  gh D 5 log  -----------

 8LQ - 
2.51  D - -------------------------
Q = ---------- f 10  3.7D- + --------------------------
4Q
8L   gh D 
5
f

 
gh D
5 
– 2.0  2.51 8 
f
- log  ------------ + ------------------ ----------------------
Q = -------------- -----------------
8 L 10  3.7D 4 3
 gh f D
 ------------------
L

 
gh f D
5 
 1.7748
Q = – 2.221 ------------------ log  ------------ + ---------------------- (6.16)
L 10  3.7D 3
 gh f D
 ------------------
L

DISCHARGE FOR TRANSITIONAL TURBULENT FLOW (log BASE 10)

Now using Equation 4.77 to convert Equation 6.16 to a natural logarithm form gives

159
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

 
gh f D  
5
1.7748
Q = – 0.4343*2.221 ------------------ ln ------------ + ----------------------
L  3.7D 3
 gh f D
 ------------------ 
L

or

 
gh f D  
5
1.7748-
Q = – 0.9646 ------------------ ln ------------ + --------------------- (6.17)
L  3.7D 3
 gh f D
 -
-----------------
L

DISCHARGE FOR TRANSITIONAL TURBULENT FLOW (NATURAL LOG)

Recall from Chapter 4 (Equation 4.81), that the Equation 6.17 is valid for

5  --------------------
f Re-  70 (6.18)
D 8

* Note that units of  and D must be consistent, either both in meters (or feet) or both in mm (or in.).

Note the coefficient in Equation 6.17 of – 0.9646 differs from those given by Streeter et al. (1998) of a
value of – 0.955 based on the work by Swamee and Jain (1976) due to the difference in the coefficient
of the Colebrook – White formula.

To apply the explicit solution procedure for Q (where the flow is assumed to be transitional turbulent)
the following steps apply

1. Draw the HGL.

2. Write the energy equation for the system as shown in Equation 6.1.
3. Compute the discharge from the explicit equation derived above (Equation 6.16 or
Equation 6.17).
4. Check assumption. Check whether the assumption about the turbulent flow regime is cor-
rect by computing the dimensionless number for the viscous sublayer. For transitional tur-
bulent flow the following must hold

 -  70
5  ----------- (6.19)
  u*

or

V f
5  ---------------  70 (6.20)
 8

5. Recompute discharge if necessary. If the dimensionless number from Equation 6.19 or

Equation 6.20 is not in this range then recompute the discharge using the explicit expres-
sion for the appropriate flow regime.

160
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

Example 6.3 For a reservoir – pipe – reservoir system as shown in Figure 6.3 with a roughness height of
–6
 = 0.25 mm. Other pertinent quantities are D = 300 mm, L = 500 m, z1 = 50 m, z2 = 45 m, v = 1.007 10 m2/s.
Based on the value of the roughness height for the pipe assume the type of turbulent flow to be transitional turbulent.

(i) Compute the discharge.

(ii) Compute the Darcy – Weisbach friction factor or resistance factor using the Swamee and Jain formula.
(iii) Compute the friction factor from the Colebrook – White formula.
(iv) Check that the flow regime is in fact – transitional turbulent pipe flow.
This example is very similar to the smooth pipe turbulent flow example of Example 6.2.

Answer – (i) Computation of the discharge between the two reservoirs – transitional turbulent flow.

Apply the energy equation between points 1 and 2 in Figure 6.3

2 2 2
p V1 p V2 LV
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- + f ---- ------
 2g  2g D2g

At the surface of the reservoir the velocity is zero or V = 0 and the pressure p = 0 hence it follows that

2
L- V
z 1 + 0 + 0 = z 2 + 0 + 0 + f --- ------
D2g
Thus it follows that

2
L- V
f --- ------ = z 1 – z 2 = 50 – 45 = 5 m
D2g
Thus the head loss is

hf = 5 m

The discharge is computed from Equation 6.17 as

 
gh f D   
5
1.7748 
Q = – 0.9646 ------------------ ln ------------ + ----------------------
L  3.7 D 3
 gh f D 
 ------------------
L

Thus

 
5 –6 
 5.0   0.3  -ln  --------------------- 1.7748  1.007 10 
Q = – 0.9646 9.81 0.25 + -------------------------------------------------
-------------------------------------
500  3.7  300  3 
 9.81  5.0   0.3  - 
 ------------------------------------- 
500

From the spreadsheet in Figure 6.5 the discharge in the pipeline is

3
Q = 0.123 m /s

___________________________________________________________________________________
Answer – (ii) Compute the Darcy – Weisbach friction factor or resistance factorvusing the Swamee and Jain
formula (Equation 4.61).

The pipe area is

2 2
A =  D- = 
---------  0.3  - = 0.07069 m 2
-----------------
4 4

The velocity is

V = Q 0.123 - = 1.740 m/s

---- = ------------------
A 0.07069

The Reynolds number is

161
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

VD 1.740  0.30 
Re = -------- = ----------------------------- = 519,191
 1.007 10
–6

Figure 6.5 Spreadsheet for computation of transitional turbulent flow (unknown Q) by explicit formula

An estimate of the friction factor using the Swamee and Jain formula (Equation 4.61) is

0.25 0.25
f = ---------------------------------------------------------- = ------------------------------------------------------------------------------------------- = 0.0196
  5.74  2  0.25 5.74  2
log 10 ------------ + ------------ log 10  --------------------- + ----------------------------------- -
 3.7 D 0.9
Re  3.7  300   5.18 x 10 5   0.9

___________________________________________________________________________________
Answer – (iii) Computation of the transitional turbulent Darcy – Weisbach friction factor or rsistance factor using
the Colebrook – White formula

The Colebrook – White friction factor f is given by Equation 4.76 as

1  2.51
-------- =  – 2.0 log 10  ------------ + --------------- 
f  3.7 D Re f 

Using Goal Seek in Microsoft Excel as shown in Figure 6.6 to solve for f in Equation gives

f = 0.0195

Figure 6.6 Spreadsheet for using Goal Seek in Microsoft Excel to compute transitional turbulent friction factor

Thus the value from the Swamee and Jain equation (Equation 4.61) is confirmed by the spreadsheet. The computations

162
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

shown in the spreadsheet in Figure 6.5 have been verified by checking the equality of Equation 6.21.

___________________________________________________________________________________
Answer – (iv) Checking that the flow regime is in fact transitional turbulent pipe flow.

Now check if the discharge is in the transitional turbulent flow regime by computing

0.25- 1.740 0.0195

 -----------
V f  1000
------------------ = ---------------------------------------------------- = 21.3  5 but  70
–6
 8 1.007 10 8

Thus the assumption of transitional turbulent flow is OK.

___________________________________________________________________________________

6.2.3 Explicit expression for Q assuming fully rough turbulent pipe flow
An explicit expression for the discharge is computed in this section assuming the flow is in the fully
rough turbulent regime. The equation is derived by commencing with Equation 4.83 for the Prandtl –
Nikuradse fluid friction coefficient formula for the friction factor or resistance factor of

1 - = 2.0log  ---
D
--------- 10  - + 1.14
f

Solving for f gives

1
f = -------------------------------------------------------------- (6.21)
2
 D 
2.0  log 10  ----  + 1.14
 
 

Take the Darcy – Weisbach head loss expression given by Equation 4.3 and rearrange in terms of dis-
charge Q to obtain

2 2 2
L Q L Q 16 fLQ 8 f LQ
f = f ---- ------------- = f ---- -------------------------- = ------------------------- = --------- -----------
D 2g A 2 D D
2 2
D 2g D
2 4 2
 g D
5
2g  ----------
 4 

Solving for the discharge Q

2 5
 gh f D
Q = ------------------------ (6.22)
8Lf

Substituting Equation 6.21 into Equation 6.22 gives

2
 D-  + 1.14
 gh f D 2.0  log 10  ---
2 5
  
 
Q = --------------------------------------------------------------------------------------- (6.23)
8L

DISCHARGE FOR FULLY ROUGH TURBULENT FLOW

Recall from Chapter 4 (Equation 4.92), that the Equation 6.23 is valid for

163
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

Re  200D
------------- (6.24)
 f

* Note that units of  and D must be consistent, either both in meters (or feet) or both in mm (or in.).

To apply the explicit solution procedure for Q (where the flow is assumed to be fully rough turbulent)
the following steps apply

1. Draw the HGL.

2. Write the energy equation for the system as shown in Equation 6.1.
3. Compute the discharge from the explicit equation derived above (Equation 6.23).
4. Check assumption. Check whether the assumption about the turbulent flow regime is cor-
rect by computing the dimensionless number for the viscous sublayer. For fully rough tur-
bulent flow the following must hold


------------  70 (6.25)
  u*

or

V f
---------------  70 (6.26)
 8

5. Recompute discharge if necessary. If the dimensionless number from Equation 6.25 or

Equation 6.26 is not in this range then recompute the discharge using the explicit expres-
sion for the appropriate flow regime.
Example 6.4 For a reservoir – pipe – reservoir system shown in Figure 6.8 with D = 300 mm, L = 500 m, z1 = 50 m,
–6
z2 = 45.0 m,  = 1 mm, and for water at 20C v = 1.007 10 m2/s. Assume the minor losses are negligible.
Based on the value of the roughness height for the pipe assume the type of turbulent flow to be fully rough turbulent.

(i) Compute the discharge.

(ii) Compute the Darcy – Weisbach friction factor or rsistance factor using the Swamee and Jain formula.
(iii) Compute the friction factor from the Prandtl – Nikuradse formula.
(iv) Check that the flow regime is in fact fully rough turbulent pipe flow.

Answer – (i) Computation of the discharge between the two reservoirs – fully rough turbulent flow.

Apply the energy equation between points 1 and 2 in Figure 6.2

2 2 2
p V1 p V2 LV
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- + f ---- ------
 2g  2g D2g

At the surfaces of the reservoirs the velocity is zero or V = 0 and the pressure p = 0 hence it follows that

2
L- ------
V
z 1 + 0 + 0 = z 2 + 0 + 0 + f ---
D2g
Thus it follows that

2
L- V
f --- ------ = z 1 – z 2 = 50 – 45.0 = 5.0 m
D2g
Thus the head loss is

h f = 5.0 m

The fully rough turbulent flow is computed from the analytic expression for the discharge of Equation 6.23 as

164
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

2
 D 
 gh f D 2.0  log 10  ----  + 1.14
2 5
 
 
Q = ---------------------------------------------------------------------------------------
8L

2
 300 
  9.81   5.0   0.3  2.0  log 10  ---------  + 1.14
2 5
 1 
 
Q = ---------------------------------------------------------------------------------------------------------------------
8  500 

3
Q = 0.1045 m /s

___________________________________________________________________________________
Answer – (ii) Computation of the Darcy – Weisbach friction factor or resistance factor using the Swamee and
Jain formula (Equation 4.61).

The pipe area is

2 2
D   0.3  2
A = ---------- = ------------------ = 0.07069 m
4 4

It follows that a check on the velocity of flow gives

Q 0.1045
V = ---- = ------------------- = 1.478 m/s
A 0.07069

The Reynolds number is

VD 1.478  0.30 
Re = -------- = ----------------------------- = 440475
 1.007 10
–6

An estimate of the friction factor using the Swamee and Jain formula (Equation 4.61) is

0.25 0.25
f = ---------------------------------------------------------- = ------------------------------------------------------------------------------------------- = 0.0273
  5.74  2  1.0 5.74  2
log 10 ------------ + ------------ log 10  --------------------- + ------------------------------------
 3.7 D 0.9
Re  3.7  300   4.40 x 10 5   0.9

The solution to this problem is also given by spreadsheet in Figure 6.7.

165
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

Figure 6.7 Spreadsheet for computation of fully rough turbulent flow (unknown Q) by explicit formula

___________________________________________________________________________________
Answer – (iii) Confirmation of the friction factor or resistance factor for fully rough turbulent flow in a pipe
from the Prandtl – Nikuradse formula.

The fully rough turbulent friction factor f is given by Equation 6.21 (Prandtl – Nikuradse formula) as

1 1
f = -------------------------------------------------------------
- = ------------------------------------------------------------------
- = 0.0269
2 2
 D   300 
2.0  log 10  ----  + 1.14 2.0  log 10  ---------  + 1.14
   1.0 
   

confirming the value from the Swamee and Jain equation.

___________________________________________________________________________________
Answer – (iv) Checking that the flow regime is in fact fully rough turbulent pipe flow.

Now check if the discharge is in the fully rough turbulent flow regime by computing

1.0 
 ------------ 1.478 0.0269
V f  1000
--------------- = ---------------------------------------------------- = 85.1  70
–6
 8 1.007 10 8

Thus the assumption of fully turbulent flow is OK.

___________________________________________________________________________________

166
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

6.2.4 Trial and error determination of Q — including minor losses

To include minor losses in the calculation of the discharge it is not possible to derive explicit expres-
sions as shown in the previous sections. In computing the unknown discharge problem for a simple pipe
using a trial and error approach, the following quantities are assumed to be given — z1, z2, L, D, , and
minor loss coefficients K1, K2,... etc. A trial and error approach is used. The discharge Q is unknown and
it follows that neither the Reynolds number nor the friction factor or resistance factor can be computed
directly. Thus both Q and the Darcy – Weisbach f are unknown. It is necessary to iterate between Q and f
until there is negligible change.

The procedure below may also be applied to the unknown discharge problem while ignoring minor
losses. The method then is an alternative to using the explicit equations derived in the previous sections,
however, this approach is less attractive than the explicit formula approach.

Consider a simple pipeline system between two reservoirs that has three minor losses. Write the energy
equation for the system from a point on reservoir 1 to a point on reservoir 2.

2 2 2
p V p V L- V
z 1 + -----1- + ------1 = z 2 + -----2- + ------2 + f --- ------ +  h L
 2g  2g D 2g

At the surfaces of the reservoirs V1 = V2 = 0.0, and p1 = p2 = 0.0 and thus for say three minor losses in
the single pipeline

2
fL V
z 1 = z 2 +  ------ + K 1 + K 2 + K 3 ------
D  2g

Thus

2
fL V
z 1 – z 2 =  ------ + K 1 + K 2 + K 3 ------
D  2g

Now substitute for the velocity in terms of the discharge

2
fL- + K + K + K  -------------------------
Q
z 1 – z 2 =  ----- -
D 1 2 3 2 2
 D 
2g ----------
 4 

2
8 fL Q
z 1 – z 2 =  ---------  ------ + K 1 + K 2 + K 3 -------
 2  D  4
 g D

2 4
 g  z 1 – z 2 D
Q = -------------------------------------------------------
- (6.27)
fL
8  ------ + K 1 + K 2 + K 3
D 

The Reynolds number is also unknown during the iterative process. Write the Reynolds number from
Equation 2.15 in terms of discharge.

VD QD 4Q
Re = -------- = ------------------- = ----------- (6.28)
 2 D
 D
---------- 
 4 

167
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

The steps for applying the trial and error approach for determining an unknown discharge in a simple
pipeline while including minor losses are as follows. A trial and error method is used to iterate between
f and Q and the procedure is

1. Draw the HGL.

2. Compute an estimate of the friction factor or resistance factor. As a first estimate, compute
the relative roughness D and determine f from the flat portion of the Moody diagram
(Figure 4.2) assuming fully rough turbulent flow or from the Prandtl – Nikuradse friction
factor formula for fully rough turbulent flow. In the fully rough turbulent flow regime the
friction factor only depends on the relative roughness and is independent of the Reynolds
number. The formula from Equation 4.83 is

1 D
---------- = 2.0log 10  ---- + 1.14
f  

or by solving for the friction factor as in Equation 6.21 gives

1
f = -------------------------------------------------------------- (6.29)
2
 D 
2.0  log 10  ----  + 1.14
 
 

3. Compute the discharge from Equation 6.27 for the current value of the Darcy – Weisbach
friction factor f.

2 4
 g  z 1 – z 2 D
Q = -------------------------------------------------------
- (6.30)
 fL 
8 ------ + K 1 + K 2 + K 3
D 

4. Compute the Reynolds number from Equation 6.28.

4Q
Re = ----------- (6.31)
D

5. Recompute friction factor. Recompute a revised estimate of the friction factor f ' from the
Swamee and Jain equation (Equation 4.61) or the Moody diagram (Figure 4.2) based on the
improved estimate of Reynolds number and the known value of relative roughness /D.
6. Continue the iterative procedure until convergence. Repeat the procedure (steps 3 to 5)
until there is no appreciable change in Q and f. At this point there should be convergence to
a satisfactory tolerance between f and f’.
It does not take many iterations to converge on an adequate solution. The friction factor f
should only be found to an accuracy of 2 or 3 significant figures (e.g. 0.018 or 0.0183).
Example 6.5 For a reservoir – pipe – reservoir system as shown in Figure 6.9 for pipe with a roughness height of
 = 0.25 mm. Other pertinent quantities are D = 300 mm, L = 500 m, z1 = 50 m, z2 = 45 m and
–6
v = 1.007 10 m2/s (water at 20C ).

(i) Compute the discharge taking into account minor losses in the system.

Answer – (i) Computation of discharge between the two reservoirs – minor losses accounted for.

Write the energy equation between the two reservoirs including minor losses

2 2 2
p V1 p V2 fL V
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- + ------ ------ +  h L
 2g  2g D 2g

168
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

But at the surfaces of the reservoirs p1 = 0.0, V1 = 0.0, p2 = 0.0, V2 = 0.0, and z1 – z2 = 5.0 m and thus it follows that

2 2 2
fL V V V
5.0 = ------ ------ + K entrance ------ + K exit ------
D 2g 2g 2g

Only two minor losses are considered here. It is assumed that the entrance to the pipe is sharp edged ( K entrance = 0.5 )
and at the exit a full velocity head is lost ( K exit = 1.0 ). Thus

2
V fL
5.0 = ------ ------ + 0.5 + 1.0 (6.32)
2g D

Figure 6.8 Computation of discharge taking into account minor losses

The velocity is expressed in terms of the discharge

2 2 2 2
= Q 16Q - = 16Q ------------- = 1.6211Q
2
V ------ = ------------ ----------------------- (6.33)
2 2 4 2 4 4
A  D  D D

Substituting Equation 6.33 into Equation 6.32 gives

2
1.6211Q 1 fL
5.0 = ----------------------- ------ ------ + 1.5
D
4 2g D

Substituting for the length and diameter of the pipe and gravitational acceleration gives

2
1.6211Q 1 500
5.0 = ----------------------- ------------------ f --------- + 1.5
 0.3  2  9.81  0.3
4

2
5.0 = 10.201Q  1666.7 f + 1.5 

5.0 2
-------------------------------------------------------- = Q
10.201  1666.7 f + 1.5 

2 0.4901
Q = ---------------------------------------
 1666.7 f + 1.5 

0.4901
Q= --------------------------------------- (6.34)
 1666.7 f + 1.5 

Iteration No. 1

Estimate the friction factor or resistance factor for the pipe from Equation 6.29 based on the assumption of fully tur-
bulent flow from the Prandtl – Nikuradse formula gives

1 1
f = -------------------------------------------------------------- = -------------------------------------------------------------------- = 0.0188
2 2
 D   300 
2.0  log 10  ----  + 1.14 2.0  log 10  ----------  + 1.14
   0.25
   

This is equivalent to the value of friction factor for the flat portion of the Moody diagram (Figure 4.2) for the corre-
sponding relative roughness value.

Now the discharge from Equation 6.34 is

169
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

0.4901 3
Q= --------------------------------------------------------
- = 0.122 m /s
 1666.7  0.0188  + 1.5 

The pipe area of flow is

2 2
A =  D- =   0.3  - = 0.07069 m 2
--------- -----------------
4 4

The velocity of flow is

V = Q 0.122 - = 1.726 m/s

---- = ------------------
A 0.07069

The Reynolds number from Equation 6.31 is

-------- = ------------------------------
1.726   0.3 
Re = VD - = 514201
 1.007 10
–6

From the Swamee and Jain equation for the Darcy – Weisbach friction factor from Equation 4.61 it follows that

' 0.25 0.25

f = ---------------------------------------------------------- = ----------------------------------------------------------------------------------- = 0.0196
  5.74  2 0.25 5.74 2
log 10 ------------ + ------------ log 10 --------------------- + ----------------------------

 3.7 D 0.9  3.7  300  0.9
Re  514201 

The next estimate of the discharge for the revised estimate of the friction factor is given from Equation 6.34 as

0.4901 3
Q= --------------------------------------------------------- = 0.1198 m /s
 1666.7  0.0196  + 1.5 

The velocity of flow is

Q 0.1198
V = ---- = ------------------- = 1.695 m/s
A 0.07069

The Reynolds number from Equation 6.31 is

VD  1.695   0.3 
Re = -------- = ------------------------------- = 504965
 1.007 10
–6

From the Swamee and Jain equation for the Darcy – Weisbach friction factor from Equation 4.61 is

' 0.25 0.25

f = ---------------------------------------------------------- = ----------------------------------------------------------------------------------- = 0.0196
  5.74  2 0.25 5.74 2
log 10 ------------ + ------------ log 10 --------------------- + ----------------------------

 3.7 D 0.9  3.7  300  0.9
Re  504965 

Thus the friction factor has changed negligibly. Thus the discharge in the pipe taking into account minor losses is

3
Q = 0.120 m /s

This compares with Q = 0.123 m3/s from Example 6.3 where the minor losses were ignored. The difference is negligi-
ble and illustrates the reason that minor losses can usually be neglected.

___________________________________________________________________________________

6.2.5 A comparison of the discharges for different turbulent flow regimes

In this section the differences that changes in roughness height have on the discharge are highlighted.
Recall the details of the reservoir – pipeline – reservoir system considered in Example 6.2, Example 6.3,
Example 6.4 and Example — D = 300 mm, L = 500 m, z1 = 50 m, z2 = 45 m.

–6
The kinematic viscosity is v = 1.007 10 m2/s (water at20°C). Discharges and friction factors or
resistance factors are compared in Table 6.1.

Table 6.1 Comparison of discharges (minor losses are neglected unless otherwise noted)
Roughness height Turbulent flow regime Example f Computed
 discharge
(mm) (m3/s)

170
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

Table 6.1 Comparison of discharges (minor losses are neglected unless otherwise noted)
0.003 Smooth pipe turbulent Example 6.2 0.0126 0.153
0.25 Transitional turbulent Example 6.3 0.05 0.123
1.0 Fully rough turbulent Example 6.4 0.0269 0.105
0.25 Transitional turbulent Example 6.5 0.06 0.120
plus minor losses

6.3 Which Pipe Carries the Greater Discharge?

Consider the pipe system in Figure 6.9 where three identical single pipes are in parallel. The pipes are at
different elevations. Which pipe carries more discharge?

Example 6.6 Determine which pipe carries more discharge in Figure 6.9. Assume that the pipes are or have

1. The same pipe material.

2. The same diameter.
3. The same length.
4. The same minor losses.

Figure 6.9 Reservoirs with three connecting pipes

Answer – Computation and comparison of discharge in three different pipes.

Write the energy equation between reservoir 1 and reservoir 2 for pipe [1]

2 2 2
p V p V L 1  V  1 
z 1 + -----1- + ------1 = z 2 + -----2- + ------1 + f  1  ---------
- ------------ +  h L
 2g  2g D 1  2 g

Now, at the surfaces of the reservoirs p 1 = 0.0, V 1 = 0.0, p 2 = 0.0, V 2 = 0.0and thus it follows that

2 2
L 1  V  1  V 1
- ------------ +  K i ------------
z 1 – z 2 = f  1  ---------
D 1  2 g i 2g

A unique combination of pipe velocity V[1] and friction factor f[1] satisfy Equation and applies equally to pipes 1, 2
and 3 so that it follows that all pipes have exactly the same discharge.

Q1 = Q2 = Q3

Thus it is concluded that the elevation of the pipe has no bearing on the discharge through each of the three pipes. The
discharge only depends on the elevation difference between the reservoirs and the pipe friction and minor losses.

___________________________________________________________________________________

6.4 Unknown Pipe Diameter

Unlike the unknown head loss simple pipe problem and the unknown discharge problem already dis-
cussed in previous sections, the unknown pipe diameter problem is a design problem because the diam-
eter of the pipe is unknown. In this problem for a simple pipe, assume the following are given including
h f , L, Q,  and  .

171
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

Two approaches to the unknown pipe diameter problem are presented in the following sections. In the
first, an explicit formula for the pipe diameter is used. This method is applicable when the minor losses
are negligible. In the second approach, a trial and error approach is used that is applied when minor
losses are included in the analysis. Neither the Reynolds number nor the relative roughness may be com-
puted directly, as the diameter is unknown. Thus both D and the Darcy – Weisbach f are unknown. In the
trial and error approach it is necessary to iterate between D and f until there is negligible change.

Figure 6.10 Solving for diameter

For the system in Figure 6.10 write the energy equation between point 1 at the upper reservoir and point
2 on the lower reservoir assuming the minor losses are negligible. From Equation 6.2 it follows that

2
f LV
z 1 = z 2 + ------ ------
D 2g

Now substitute for the velocity in terms of the discharge

2
fL Q
z 1 – z 2 = ------ --------------------------
D D
2 2
2g  ----------
 4 

Thus the friction loss is equal to the elevation difference between the two reservoirs

z1 – z2 = h f

2
8 - ------
f L ------
Q-
h f =  --------
 2  D 4
 g D

2
 g- h = ------
--------- f L-
2 f 5
8Q D

2
 g-h
f - = -------------
------ (6.35)
5 2 f
D 8LQ

The terms on the right – hand side of Equation 6.35 are known while the terms on the left – hand side of
Equation 6.35 are unknown. The diameter could be solved from Equation 6.35 by trial and error or by an
explicit formula as given in the next section.

172
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

6.4.1 Explicit expression for D assuming smooth pipe turbulent flow

Streeter and Wylie (1983) developed an explicit formula for the diameter D based on the work of Swa-
mee and Jain (1976). There are different formulas for the different turbulent flow regimes.

For smooth pipe turbulent flow the diameter (Swamee and Jain 1976; Streeter and Wylie 1983) is given
by

9.4  L - 5.2 0.04

D = 0.66 Q ---------- (6.36)
 gh 
f

PIPE DIAMETER FOR SMOOTH PIPE TURBULENT FLOW

Note that in Equation 6.36, the units of meters and seconds must be used for all variables. Assume
smooth pipe turbulent flow.

Example 6.7 For the single pipe in a reservoir – pipe – reservoir system as shown in Figure 6.10 for a pipe with a
roughness height of  = 0.003 mm. Other pertinent quantities are Q = 153 L/s, L = 500 m, z1 = 50 m, z2 = 45 m
–6
and v = 1.007 10 m2/s (water at 20C ). Assume smooth pipe turbulent flow is occurring.

Compute the diameter required to carry the specified flow, ignoring minor losses in the system.

Answer – Computation of the diameter of a pipe between the two reservoirs – smooth pipe turbulent flow.

Apply the energy equation between points 1 and 2 in Figure 6.10

2 2 2
p V1 p V2 LV
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- + f ---- ------
 2g  2g D2g

At the surface of the reservoir the velocity is zero or V = 0 and p = 0 hence it follows that

2
L- V
z 1 + 0 + 0 = z 2 + 0 + 0 + f --- ------
D 2g
Thus the head loss is

2
LV
h f = f ---- ------ = z 1 – z 2 = 50 – 45 = 5 m
D2g

Convert units of Q to m3/s

3
Q = 0.153 m /s

Assume the flow is smooth pipe turbulent flow.

The diameter is computed from Equation 6.36 for smooth pipe turbulent flow as follows

9.4  L - 5.2 0.04

D = 0.66 Q ----------
 gh 
f

Assume gravitational acceleration is 9.81 m/s2

9.4  500 5.2 0.04

-----------------------
–6
D = 0.66 1.007 10  0.153  = 0.304 m
 9.81  5.0 

The result is close (a slight overestimate) to the value used in Example 6.3 of 300 mm diameter pipe that computed the
discharge of 153 L/s using the explicit formula for discharge based on the smooth pipe turbulent flow formula of von
Karman for the Darcy – Weisbach friction factor.

___________________________________________________________________________________

173
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

6.4.2 Explicit expression for D assuming transitional turbulent flow

For transitional turbulent flow the pipe diameter (Swamee and Jain 1976; Streeter and Wylie 1983) is
given as follows

2 4.75 0.04
1.25  LQ 9.4  L - 5.2
D = 0.66   ---------- + Q ---------- (6.37)
 gh f   gh 
f

PIPE DIAMETER FOR TRANSITIONAL TURBULENT FLOW

Note that in Equation 6.37, the units of meters and seconds must be used for all variables (i.e.  in
meters and not mm, L in meters, Q in m3/s, g in m/s2, h f in meters,  in m2/s).

Example 6.8 For the single pipe for a reservoir – pipe – reservoir system as shown in Figure 6.10 for pipe with a
roughness height of  = 0.25 mm. Other pertinent quantities are Q = 123 L/s, L = 500 m, z1 = 50 m, z2 = 45 m and
–6
v = 1.007 10 m2/s (water at 20C ). Assume transitional turbulent flow. Compute the diameter for the specified
flow ignoring minor losses in the system

Answer – Computation of the diameter of a pipe between the two reservoirs – transitional turbulent flow.

Apply the energy equation between points 1 and 2 in Figure 6.10

2 2 2
p V1 p V2 L- V
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- + f --- ------
 2g  2g D2g

At the surface of the reservoir the velocity is zero or V = 0 and the pressure p = 0 hence it follows that

2
LV
z 1 + 0 + 0 = z 2 + 0 + 0 + f ---- ------
D2g
Thus the head loss is

2
LV
h f = f ---- ------ = z 1 – z 2 = 50 – 45 = 5 m
D2g
Assume the flow is transitional turbulent.

The diameter is computed from Equation 6.37 for transitional turbulent flow is as follows

4.75 0.04
1.25  LQ
2
9.4  L 5.2
D = 0.66   ---------- + Q -----------
 gh f   gh 
f

2 4.75 5.2 0.04

0.25 1.25  500  0.123   500
D = 0.66  ------------
9.4 
-----------------------
–6
------------------------------ + 1.007 10  0.123  = 0.307 m
 1000  9.81  5.0    9.81  5.0 

Note that the roughness height  = 0.25 mm has been converted from mm to meters. The result agrees with
Example 6.3 that computed the discharge of 123 L/s for a 300 mm diameter pipe using the explicit formula for dis-
charge based on the transitional turbulent flow formula of Colebrook – White for the Darcy – Weisbach friction factor
or resistance factor.

___________________________________________________________________________________

174
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

6.4.3 Explicit expression for D assuming fully rough turbulent flow

For fully rough turbulent pipe flow, the pipe diameter (Swamee and Jain 1976; Streeter and Wylie 1983)
is given as follows

2 4.75 0.04
1.25  LQ
D = 0.66   ---------- (6.38)
 gh f 

PIPE DIAMETER FOR FULLY ROUGH TURBULENT FLOW

Note that in Equation 6.38, that in the units of meters and seconds must be used for all variables (i.e. 
in meters and not mm, L in meters, Q in m3/s, g in m/s2 and h f in meters).

Example 6.9 For the single pipe in a reservoir – pipe – reservoir system as shown in Figure 6.10 with a pipe of
roughness height of  = 1 mm. Other pertinent quantities are Q = 105 L/s, L = 500 m, z1 = 50 m and z2 = 45 m
(water at 20C ). Assume fully rough turbulent flow.

Compute the diameter for the specified flow, ignoring minor losses in the system.

Answer – Computation of the diameter of a pipe between the two reservoirs – fully rough turbulent flow.

Apply the energy equation between points 1 and 2 in Figure 6.10

2 2 2
p V1 p V2 L- V
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- + f --- ------
 2g  2g D2g

At the surface of the reservoir the velocity is zero or V = 0 and the pressure p = 0 hence it follows that

2
LV
z 1 + 0 + 0 = z 2 + 0 + 0 + f ---- ------
D2g
Thus the head loss is

2
LV
h f = f ---- ------ = z 1 – z 2 = 50 – 45 = 5 m
D2g
Assume the flow is fully rough turbulent.

The diameter is computed from Equation 6.37 for fully rough turbulent flow as follows

4.75 0.04
1.25  LQ 
2
D = 0.66   ---------- 
 gh f 

2 4.75 0.04
1 1.25  500  0.105  
D = 0.66  ------------ ------------------------------ = 0.308 m
 1000  9.81  5.0  

Note that the roughness height  has been converted from mm to m. The result is similar (a slight overestimate) of the
value used in Example 6.3 of 300 mm diameter pipe that computed the discharge as 105 L/s using the explicit formula
for discharge based on the fully rough turbulent flow formula of Prandtl – Nikuradse for the Darcy – Weisbach friction
factor.

___________________________________________________________________________________

6.4.4 Determining D by trial and error — including the minor losses

In the previous section the diameter was found using an explicit equation. When minor losses are
included, a trial and error solution procedure must be developed. The following variables are assumed to
be known — Q, z1, z2, L, and minor loss coefficients (K1, K2,... etc.)

For the system in Figure 6.10 write the energy equation between point 1 at the upper reservoir and point
2 on the lower reservoir as shown in Equation 6.1.

175
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

Figure 6.11 Computation of unknown diameter taking into account minor losses

At the surfaces of the reservoirs V1 = V2 = 0, and p1 = p2 = 0.0 and thus for say three minor losses

2
fL V
z 1 = z 2 +  ------ + K 1 + K 2 + K 3 ------
D  2g

Now substitute for the velocity in terms of the discharge

2
fL Q
z 1 – z 2 =  ------ + K 1 + K 2 + K 3 --------------------------
D  2 2
2g  D ----------
 4 

2
8   fL + Q
z 1 – z 2 =  --------
- ------ K 1 + K 2 + K 3 -------
 2  D  4
 g D

2
 g 1 fL
----------  z 1 – z 2  = -------  ------ + K 1 + K 2 + K 3
4 
D D
2
8Q

2
1 fL  g
-------  ------ + K 1 + K 2 + K 3 = ----------  z 1 – z 2  (6.39)
4 D  2
D 8Q

or in a more general form for any number of minor losses

2
1 fL  g
-------  ------ +  K  = ----------  z 1 – z 2  (6.40)
4 D  2
D 8Q

Consider the Reynolds number. Eliminate the velocity V using the continuity equation of

Q = VA

The Reynolds number from Equation 2.15 becomes

VD QD 4Q
Re = -------- = ------------------- = ----------- (6.41)
 D  
2 D
 ----------
 4 

176
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

There is only one unknown, the diameter D, in the right – hand side of the Reynolds number equation. In
addition the relative roughness D is also unknown. Thus determination of the Darcy – Weisbach fric-
tion factor or resistance factor in Equation 6.39 is not straightforward.

The procedure for solving for the diameter by trial and error is as follows

1. Draw the HGL.

2. Estimate the diameter of the pipe. For a first approximation a velocity of 0.6 m/s is
assumed. Thus from the continuity equation

Q = VA

2
A = D Q -
---------- = ----------------------
4 V assumed

Thus solving for the diameter based on an assumed velocity of flow gives

D = 4Q -
-------------------------- (6.42)
V assumed

As an alternative the explicit formula developed in the previous section (Equation 6.37) is
used as an estimate, by taking say 85% of h f to take into account the effect of the minor
losses. A flow regime of transitional turbulent flow is assumed.
3. Compute the Reynolds number. Based on the current estimate of the pipe diameter D solve
for the Reynolds number from Equation 6.41.

4Q
Re = ----------- (6.43)
D

4. Compute the relative roughness for the estimate of diameter D


---- (6.44)
D

5. Compute the friction factor. Either use the Moody diagram (Figure 4.2) to determine the
Darcy – Weisbach friction factor or solve for the friction factor from the Swamee and Jain
equation (Equation 4.61).
6. Compute the unknown diameter. Solve by trial and error (or the Goal Seek function in
Microsoft Excel) for the diameter D from Equation 6.40. All terms on the right – hand side
are known.

2
1 fL  g
-------  ------ +  K  = ----------  z 1 – z 2  (6.45)
4 D  2
D 8Q

7. Continue the iterative procedure until convergence. Repeat steps 3 to 6. When the value of
f does not change in the first two significant figures, all equations are satisfied and the prob-
lem for the unknown diameter is solved.
6.4.5 Computing the diameter directly when minor losses are included
Alternatively the unknown diameter D is determined directly. Recall the Swamee and Jain equation
from Equation 4.61 with the Reynolds number substituted as

177
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

0.25
f = --------------------------------------------------------------------- (6.46)
  2
  5.74 
log 10  ------------ + -----------------------
 3.7D  4Q  0.9
 -----------
 D 

Substituting Equation 6.46 into Equation 6.45 gives

 
 
 
 
NML  2
1  0.25 L  g
-------  --------------------------------------------------------------------- ---- +  K  = ----------  z 1 – z 2  (6.47)
D 
4   2D k = 1  8Q
2
   5.74  
 log 10  ------------ + ---------------------- - 
  3.7D  4Q  0.9 
 ----------
-
  D  

where NML = number of minor losses in the system. Equation 6.47 can be solved using Goal Seek in
Microsoft Excel.

Example 6.10 Consider a commercial steel pipe needed to convey 200 L/s water at 20C a distance of 5 km with a
head drop of 3.77 m. The pipeline connects two reservoirs, has a re – entrant entrance (K = 0.8), a submerged outlet,
four standard elbows (K = 0.9), and a globe valve (K = 10). Include minor losses in the computation.

What is the diameter of the pipe required to covey the specified flow.

Answer – Computation of the diameter of a pipe between the two reservoirs – considering minor losses.

For commercial steel pipe use a roughness height = 0.046 mm.

Write the energy equation between the 2 reservoirs

2 2
p V p V LV
2
z 1 + -----1 + ------1 = z 2 + -----2 + ------2 + f ---- ------ +  h L
 2g  2g D2g

But at the surfaces of the reservoirs p1 = 0.0, V1 = 0.0, p2 = 0.0, V2 = 0.0, and z1 – z2 = 3.77 m and thus it follows that

2 2
LV V
3.77 = f ---- ------ + ------  K entr + K exit + 4 K elbow + K valve  (6.48)
D2g 2g

The sum of the minor loss coefficients is

K = K entr + K exit + 4 K elbow + K valve = 0.8 + 1.0 + 4  0.9  + 10.0 = 15.4

Using Equation 6.45 to rearrange Equation 6.48 gives

2
1 fL  g
-------  ------ +  K  = ----------  z 1 – z 2 
4 
D D
2
8Q
Thus

2
1 f  5000    9.81 
-------  --------------------- + 15.4 = ---------------------  3.77  = 1140.7
4 D  2
D 8  0.2 

1 5000 f
-------  ---------------- + 15.4 = 1140.7 (6.49)
4 D 
D

Iteration no. 1

Estimate the diameter of the pipe from Equation 6.42 based on a velocity of 0.6 m/s as

178
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

4Q- = 4  0.2 - =
D = ------- --------------- 0.4244 = 0.651 m
V   0.6 

The pipe area is

2 2
A =  D- =   0.651  - = 0.3329 m 2
--------- -----------------------
4 4

The velocity of flow is

V = Q 0.2 - = 0.601 m/s

---- = ---------------
A 0.3329

The Reynolds number from Equation 6.43 is

 0.601   0.651  =
Re = VD
5
-------- = ------------------------------------
- 388500 = 3.89 10
 1.007 10
–6

From the Swamee and Jain equation for the Darcy – Weisbach friction factor or resistance factor from Equation 4.61 is

0.25 0.25
f = ---------------------------------------------------------- = ----------------------------------------------------------------------------------- = 0.0146
  5.74  2 0.046 5.74 2
log 10 ------------ + ------------ log 10 --------------------- + ----------------------------

 3.7 D 0.9  3.7  651  0.9
Re  388500 

The next estimate of the diameter is given from Equation 6.45 as

1 5000 f
-------  ---------------- + 15.4 = 1140.7
4 D 
D

1 5000  0.0146 
-------  --------------------------------- + 15.4 = 1140.7
4 D 
D
Using Goal Seek in Microsoft Excel to solve the above equation results in D = 0.591 m as shown in Figure 6.12. The
steps for using Goal Seek are given in Example 6.2.

Figure 6.12 Spreadsheet for computation of diameter of a pipe by Goal Seek for a fixed friction factor value — iteration 1

Iteration no. 2 — for an estimate of diameter of 591 mm.

The pipe area is

2 2
A =  D- =   0.591  - = 0.2743 m 2
--------- -----------------------
4 4

The velocity of flow is

Q 0.2
V = ---- = ---------------- = 0.729 m/s
A 0.2743
The Reynolds number from Equation 6.43 is

VD  0.729   0.591  5
Re = -------- = ------------------------------------- = 427800 = 4.28 10
 1.007 10
–6

From the Swamee and Jain equation for the Darcy – Weisbach friction factor from Equation 4.61 it follows that:

0.25 0.25
f = ---------------------------------------------------------- = ----------------------------------------------------------------------------------- = 0.0145
  5.74  2 0.046 5.74 2
log 10 ------------ + ------------ log 10 --------------------- + ----------------------------

 3.7 D 0.9  3.7  591  0.9
Re  427800 

Note the friction factor has only changed very slightly — thus no further iterations are necessary. The final estimate of

179
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

the diameter is given from Equation 6.45 as

 0.0145  + 15.4 = 1140.7

-------  5000
1
---------------------------------
4 D 
D
Using Goal Seek in Microsoft Excel to solve the above equation results in D = 0.590 m as shown in Figure 6.13. The
steps for using Goal Seek are given in Example 6.2.

Figure 6.13 Spreadsheet for computation of diameter of a pipe by Goal Seek for a fixed friction factor value — iteration 2

A commercial pipe size of 600 mm would be adopted.

As an alternative to the step – by – step iterative solution procedure as described above it is possible to solve directly for
D as shown in Equation 6.47

 
 
 
 

1- ------------------------------------------------------------
1.325 L- + K  = ---------
2
 g-  z – z 
------
4

 
- ---
2 D   2 1 2
D   8Q
   5.74  
 ln  ------------ + ---------------------- - 
  3.7 D  ---------- 4Q  0.9
-  
   D 

 
 
 
 

1- -----------------------------------------------------------------------------------------------------------
1.325 
------  - 5000
------------ + 15.4 = 1140.7 (6.50)
D 
4   2 D 
  0.046 5.74  
 ln  ----------------------------- + ----------------------------------------------------  
 3.7 D  1000  4  0.2  0.9


 -----------------------------------------  
  –6  
D  1.007 10 

Equation 6.50 may be solved directly for D using Goal Seek in Microsoft Excel as shown in Figure 6.14. The steps for
using Goal Seek are given in Example 6.2.

180
 Chapter 6 - File: Chap6new_b.fm July 6, 2022 Version 6

Figure 6.14 Spreadsheet for direct computation of diameter of a pipe by Goal Seek

___________________________________________________________________________________

6.5 References
Streeter, V.L. and Wylie, E.B. (1983). Fluid mechanics. 1st SI Edition. McGraw – Hill.

Streeter, V.L., Wylie, E.B. and Bedford, K. (1998). Fluid mechanics. 9th Edition. McGraw–Hill.

Swamee, P. K., and Jain, A.K. (1976). “Explicit equations for pipe-flow problems.” Journal of the
Hydraulics Division, 102(5), 657-664.

181
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

CHAPTER 7 SERIES AND PARALLEL PIPE SYSTEMS1

This Chapter introduces the concepts of flow in simple pipe systems. The topics covered here mainly
deal with the analysis of pipes in series and pipes in parallel. The concept of an equivalent uniform main
(EUM) is introduced. An EUM may be used to replace pipes in series or in parallel such that the equiva-
lent single pipe has the same flow characteristics for analysis of steady flow situations. An EUM is use-
ful in computer modeling when a parallel pipe is added into an existing system, for example. Many
numerical examples are given in this Chapter using the formulas developed. Both the Darcy – Weisbach
head loss equation and the Hazen – Williams equation formulations are considered in the derivations. In
addition a three reservoir problem is also analyzed in this Chapter.

7.1 Equivalent Uniform Main (EUM)

Two or more pipes in either series or in parallel of different diameter may be mathematically replaced
with a single pipe of constant diameter that would have the same head loss at a given flow. An equiva-
lent uniform main (EUM — designated by the subscript “E”) is defined in terms of friction loss and only
applies for steady state computations. An EUM does not apply for water hammer calculations. The use
of the EUM concept may simplify the calculations for series and parallel pipe systems. The flow is usu-
ally assumed to be fully rough turbulent flow such that the friction factor or resistance factor is constant.

7.2 Pipes in Series

Consider two or more pipes in series (up to a total of NP) with diameters D1, D2, D3,..., DNP; lengths L1,
L2, L3,..., LNP; roughness heights 1,  2,  3,...,  NP (or Hazen – Williams coefficients C1, C2, C3,..., CNP).
The outcome that is desired is the determination of an equivalent pipe diameter DE of a specified length
or an equivalent pipe length LE of a specified diameter such that the total head loss in both systems is the
same.

Now consider exactly two pipes in series as shown in Figure 7.1. Multiple series pipes will be consid-
ered in Section 7.2.2.

7.2.1 Two Darcy – Weisbach pipes in series

Figure 7.1 Pipes in series

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Ade-
laide, Australia. Part of the book entitled Water Distribution Systems Engineering.

182
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

In Figure 7.1, the diameter of the upstream pipe is larger and thus by continuity, the velocity will be
slower in the upstream pipe compared to the downstream pipe. In turn, the losses per unit length in the
upstream pipe will be smaller and hence both the EGL and HGL for the upstream pipe have a smaller
slope (losses are proportional to the square of the velocity for the Darcy – Weisbach formula and
inversely proportional to the diameter). The vertical separation between the HGL and the EGL (the
velocity head) is smaller in the upstream pipe because the velocity is lower.

The basic equations for the series system in Figure 7.1 for the flow and head loss are

Q1 = Q2 (7.1)

Head Loss = h f + h f (7.2)

1 2

Now, the series pipeline system in Figure 7.1 is replaced by an “equivalent uniform main” (EUM). Let
the EUM be of diameter DE, length LE and flow velocity VE such that the flow QE in the EUM is

2
D
QE = V E AE = V E ----------E- (7.3)
4

The flow or discharge must be the same in both of the series pipes (see Equation 7.1) and also in the
EUM and thus

2 2 2
D D D
Q 1 = V 1 ----------1 = Q 2 = V 2 ----------2 = Q E = V E ----------E- (7.4)
4 4 4

Hence solving for both V1 and V2 in terms of VE from Equation 7.4 gives

2
DE
V 1 = -------V (7.5)
2 E
D1

and

2
DE
V2 = -------V (7.6)
2 E
D2

The total head loss along the series pipeline made up of two pipes is the same as the head loss in the
EUM. From Equation 7.2, it follows that

hf = hf + hf (7.7)
E 1 2

or based on the Darcy – Weisbach head loss equation of Equation 4.3

2
L V L V
fE = f 1 ------1- ------1- + f 2 ------2- -----2 (7.8)
D 1 2g D 2 2g

and thus

2 2
LE V E L V L V
f E ------- ------- = f 1 ------1- ------1- + f 2 ------2- ---- (7.9)
D E 2g D 1 2g D2 2

183
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Substituting for the velocities on the right – hand side of Equation 7.9 from Equation 7.5 and
Equation 7.6 gives

2 2
LE V E L D E 4 V E L D E 4 V
f E ------- ------- = f 1 ------1-  ------- ------- + f 2 ------2-  ------- ---- (7.10)
D E 2g 
D 1 D 1 2g  D2  D2  2

Canceling out the velocity head VE2/2g in all terms in Equation 7.10 and rearranging gives

LE L L
- = f 1 ------1- + f 2 -----2
f E --------- (7.11)
5 5
DE D1 D

Thus the equivalent diameter DE is found for a chosen length of pipe LE (usually taken as the sum of the
two series pipeline lengths L1 + L2) and a chosen friction factor or resistance factor f E as

5 f E LE
D E = ---------------------------------- (7.12)
L L
f 1 ------1- + f 2 ------2-
5 5
D1 D2

or

 f E LE  0.2
 --------------------------------
DE =  f 1 L 1 f 2 L 2 (7.13)
 ------------- + -------------
 D5 D2 
5
 1 

EQUIVALENT PIPE DIAMETER FOR TWO PIPES IN SERIES (DARCY –

WEISBACH)

Note that the chosen friction factor f E for the EUM is arbitrary but is usually selected based on the
likely friction factors in the two series pipelines.

Alternatively to Equation 7.13, the equivalent length of pipe LE (usually not equal to L1 + L2) of a cho-
sen diameter DE and a chosen friction factor f E is found from rearrangement of Equation 7.11 as

5
D E  L1 L2 
L E = -------  f 1 ------5- + f 2 ------5- (7.14)
f E D D 2
1

EQUIVALENT PIPE LENGTH FOR TWO PIPES IN SERIES (DARCY –

WEISBACH)

The length of the EUM varies with flow because in actual fact the friction factor or resistance factor
depends on Reynolds number. Examples demonstrating both the computation of (1) the diameter of the
EUM and (2) the length of the EUM of a specified diameter are now given.

184
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Example 7.1 Consider two pipes in series – one with length 500 m, diameter 500 mm and roughness height
0.25 mm (pipe 1), the other with length 700 m, diameter 300 mm, roughness height 0.10 mm (pipe 2).

(i) Calculate the diameter of an EUM for a 1200 m long pipe. Select a suitable friction factor or resistance factor for
the EUM.
(ii) Compute the head loss in each system for a flow of 200 L/s.

Answer – (i) Computation of the diameter of an equivalent uniform main – pipes in series/Darcy – Weisbach.

Assume a velocity of 1.5 m/s – now compute the friction factor in each pipe. Assume a kinematic viscosity of
–6 2
1.140 10 m /s corresponding to a water temperature of 15 degrees Celsius.

The Reynolds numbers for pipe 1 and pipe 2 for the assumed velocity of 1.5 m/s are:

V 1 D1 1.5  0.5  = 658000

Re 1 = -------------- = ---------------------------
 1.140 10
–6

V 2 D2 1.5  0.3 
Re 2 = -------------- = --------------------------- = 395000
 1.140 10
–6

The corresponding Darcy – Weisbach friction factors using the Swamee and Jain equation (Equation 4.61) are

0.25 0.25
f 1 = ------------------------------------------------------------- = ----------------------------------------------------------------------------------- = 0.0176
  5.74 
2 0.25 5.74 2
log 10  -------------- + ------------ log 10 --------------------- + ----------------------------

3.7 D  3.7  500  0.9
 1 Re 1 
0.9  658000 

0.25 0.25
f 2 = ------------------------------------------------------------
- = ----------------------------------------------------------------------------------
- = 0.0169
  5.74 
2 0.10
 --------------------- 5.74 2
log 10  -------------- + ------------ log 10  + ----------------------------
3.7  300   395000  0.9 
 3.7 D 2 Re 20.9

Choose the EUM friction factor f E as 0.0170 and the EUM length L E as 1200 m (L1 + L2).

The EUM diameter from Equation 7.13 is

 f E LE  0.2 0.0170  1200 

 -------------------------------------------------------------------  0.2
 -------------------------------
-  0.0176  500  0.0169  700 -
 f 1 L 1 f 2 L 2
DE = ------------ + ------------ =  ------------------------------ + ------------------------------ = 331 mm
 5 5    0.5  5 5

 D1 D2   0.3 
   

Note that the EUM diameter lies between the diameters of pipe 1 and pipe 2 of 500 mm and 300 mm respectively.

___________________________________________________________________________________
Answer – (ii) The head loss for 200 L/s in the series system will depend on the velocity.

First compute the area of each pipe.

2
D   0.5 
2
2
A 1 = ---------1- = ------------------ = 0.1964 m
4 4

2
D   0.3 
2
2
A 2 = ---------2- = ------------------ = 0.07069 m
4 4

2
D 2
A E = ----------E- =   0.331  - = 0.08595 m 2
-----------------------
4 4

The corresponding velocities are

Q- = ---------------
0.2 - = 1.018 m/s
V 1 = -----
A1 0.1964

Q- = ------------------
V 2 = ----- 0.2 - = 2.829 m/s
A2 0.07069

185
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Q 0.2
V E = ------- = ------------------- = 2.327 m/s
AE 0.08595

The head loss in the series system is

2 2 2 2
L V L V 500  1.018  700  2.829 
h f = f 1 ------1- ------1- + f 2 ------2- ------2- = 0.0176 --------- -------------------- + 0.0169 --------- --------------------
D1 2 g D2 2 g 0.5 2  9.81  0.3 2  9.81 

h f = 0.93 + 16.08 = 17.02 m

It is clear that the majority of the head loss is occurring in pipe 2 of diameter 300 mm.

The head loss in the EUM of diameter 331 m is

2 2
L V 1200  2.327 
hf = f E ------E- ------E- = 0.0170 ------------- -------------------- = 17.02 m
E DE 2 g 0.331 2  9.81 

Thus the head loss is the same through the series system and the EUM, as expected.

Example 7.2 Consider two pipes in series – one with length 500 m, diameter 500 mm and roughness height 0.25
mm (pipe 1), the other with length 700 m, diameter 300 mm, roughness height 0.10 mm (pipe 2).

(i) Calculate the length of an EUM for a specified diameter of 400 mm.
(ii) Check the head loss for a flow of 200 L/s.

Answer – (i) Computation of the length of an equivalent uniform main – pipes in series/Darcy – Weisbach.

Use the friction factors or resistance factors computed in Example 7.1 where f 1 = 0.0176 and f 2 = 0.0169.

The EUM length of pipe from Equation 7.14 for a diameter DE of 400 mm and a friction factor fE of 0.0170 is

5
D  L L  5
L E = ------E-  f 1 ------1- + f 2 ------2- = --------------
0.4   0.0176 --------------
500 + 0.0169 --------------
700  = 3102 m
f E D 5 5
 0.017  5 5
1 D2  0.5   0.3 

___________________________________________________________________________________
Answer – (ii) Checking the head loss for the EUM at a flow of 200 L/s.

Note that the head loss for the series system from Example 7.1 is 17.02 m.

First compute the area and velocity in the EUM.

2
D   0.4 
2
2
A E = ----------E- = ------------------ = 0.1257 m
4 4

Q- = ---------------
0.2 - = 1.592 m/s
V E = ------
AE 0.1257

The head loss for the EUM of diameter 400 mm and length of 3102 m is

2 2
L V 3102  1.592 
h f = f E ------E- ------E- = 0.0170 ------------ -------------------- = 17.02 m
DE 2 g 0.4 2  9.81 

This value agrees with the result obtained in Example 7.1.

___________________________________________________________________________________

7.2.2 Multiple Darcy – Weisbach pipes in series

Now consider three or more (up to NP in number) multiple pipes in series. Let the “equivalent uniform
main” (EUM) be of diameter DE, length LE and flow velocity VE. The flow must be the same in all of the
series pipes and the EUM and thus

Q 1 = Q 2 =  = Q NP = Q E (7.15)

The total head loss along the series pipeline made up of two pipes is

186
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

fE = hf + hf +  + hf (7.16)
1 2 N

Using a similar procedure as that for two pipes in series as outlined in Section 7.2.1, the equivalent
diameter DE is found for a given length of pipe LE (usually taken as L E = L 1 + L 1 +  + L NP ) and a
chosen friction factor or resistance factor f E as

 f E LE  0.
 ------------------------------------------------------------------------
-
DE =  f 1 L1 f 2 L2 f NP L NP (7.17)
 ------------- + ------------- +  + -------------------- -
 D5 D
5
D
5 
 1 2 NP 

EQUIVALENT PIPE DIAMETER FOR MULTIPLE PIPES IN SERIES

(DARCY – WEISBACH)

Alternatively, the equivalent length of pipe LE (usually not equal to L 1 + L 1 +  + L NP ) of diameter DE

is found from rearrangement of Equation 7.17 as

5
D E  L1 L2 L NP 
E = -------  f 1 ------5- + f 2 ------5- +  + f NP ----------
- (7.18)
f E D D D
5 
1 2 NP

EQUIVALENT PIPE LENGTH FOR MULTIPLE PIPES IN SERIES (DARCY –

WEISBACH)

7.2.3 Two Hazen – Williams pipes in series (for both SI units and US cus-
tomary units)
The total head loss for the EUM and along the series pipeline of pipes 1 and 2 using Equation 7.7 and
the Hazen – Williams equation for head loss in SI units from Equation 4.104 is

LE L1 1.852 6.8214 L 2
= 6.8214 = 6.8214
1.852 1.852
hf ---------------- --------------
-V ---------------- --------------
-V + ---------------- --------------- V 2 (7.19)
E 1.852 1.167 E 1.852 1.167 1 1.852 1.167
CE DE C1 D1 C2 D2

In Equation 7.19, the SI units form of the Hazen – Williams equation has been used. The constant of
6.8214 actually cancels out in all terms so it would not have mattered if the US customary units form of
the equation had been used instead — the result would have been identical.

From Equation 7.5 and Equation 7.6 it follows that

3.704
1.852 DE 1.852
V1 = --------------
-V (7.20)
3.704 E
D1

and

3.704
1.852 DE 1.852
V2 = --------------
-V (7.21)
3.704 E
D2

Substituting Equation 7.20 and Equation 7.21 and cancelling out common terms in Equation 7.19 gives

187
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

1.852 3.704 1.852 3.704 1.852

LE V E DE L1 V E DE L2 V E
------------------------------
- = --------------
- --------------
- --------------- + --------------
- --------------
- ---------------
1.852 1.167 3.704 1.167 1.852 3.704 1.167 1.852
C E DE D1 D1 C1 D2 D2 C2

Note that the coefficient of 6.8214 associated with the SI units version of the Hazen – Williams has can-
celled out throughout the equation. As a result, the following now applies to both sets of units — both SI
units and US customary units.

Combining diameter terms and cancelling the velocity terms gives

LE L1 L2
------------------------------
- = ------------------------------
- + ------------------------------
-
1.852 4.871 4.871 1.852 4.871 1.852
C E DE D1 C 1 D2 C 2

Inverting gives

1.852 4.871
C E DE 1
------------------------------- = --------------------------------------------------------------------- (7.22)
LE L1 L2
------------------------------- + ------------------------------ -
4.871 1.852 4.871 1.852
D1 C 1 D2 C 2

Thus solving for the equivalent diameter DE of a pipe of specified length LE (logically taken as L1 + L2)
gives

4.871 LE 1
DE = --------------
- ---------------------------------------------------------------------
CE
1.852 L 1 L2
------------------------------
- + ------------------------------ -
4.871 1.852 4.871 1.852
D1 C 1 D2 C 2

or

1
  ------------
-
 L  4.871
1
DE =  --------------
E
- ---------------------------------------------------------------------  (7.23)
 1.852 L L2 
CE 1
------------------------------
- + ------------------------------ - 
 4.871 1.852
D1 C 1
4.871 1.852 
D2 C 2

EQUIVALENT PIPE DIAMETER FOR TWO PIPES IN SERIES (HAZEN –

WILLIAMS)

Alternatively, the length LE of an equivalent uniform main of specified diameter DE from rearrangement
of Equation 7.22 is

1.852 4.871 L1 L2
LE = C E DE ------------------------------
- + ------------------------------
- (7.24)
4.871 1.852 4.871 1.852
D1 C 1 D2 C 2

EQUIVALENT PIPE LENGTH FOR TWO PIPES IN SERIES (HAZEN –

WILLIAMS)

188
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Example 7.3 Consider two pipes in series – one with length 500 m, diameter 500 mm and Hazen – Williams C of
131, the other with length 700 m, diameter 300 mm and Hazen – Williams C of 133. The values of C1 and C2 have
been computed to correspond to the friction factors or resistance factors in Example 7.1 based on Equation 4.115 in
Chapter 4.

(i) Calculate the diameter of an EUM for a 1200 m long pipe.

(ii) Confirm that the head loss in each system is similar.

Answer – (i) Computation of the diameter of an equivalent uniform main – pipes in series/Hazen – Williams.

Select the Hazen – Williams C value of 132 for the EUM and the EUM length as 1200 m (L1 + L2).

The EUM diameter from Equation 7.25 is

1
  ------------
-
 L  4.871
 E 1 
DE = -------------- ------------------------------------------------------------------
 1.852 L1 L2 
CE ----------------------------- + ----------------------------- 
 4.871 1.852
D1 C 1
4.871 1.852 
D2 C 2

1
  ------------
4.871
-
 1200 1 
DE =  ----------------------- - 
- ---------------------------------------------------------------------------------------------------- = 332 mm
500
  132  1.852 ---------------------------------------------- 700
- + ----------------------------------------------- 
 4.871 1.852 4.871 1.852 
 0.5   131   0.3   133 

Note that in comparison the EUM in Example 7.1 for the Darcy – Weisbach equation gave a diameter of 331 mm.

_________________________________________________________________________________
Answer – (ii) Confirm the head loss in the EUM.

Use the area of each pipe as computed in Example 7.1.

The area for a diameter of 500 mm is

2
A 1 = 0.1964 m

The area for a diameter of 300 mm is

2
A 2 = 0.07069 m

The area of the EUM for a diameter of 332 mm is

2
D   0.332 
2
2
A E = ----------E- = ------------------------ = 0.08657 m
4 4

The corresponding velocities are (repeated from Example 7.1 for completeness)

Q- = ---------------
V 1 = ----- 0.2 - = 1.018 m/s
A1 0.1964

Q- = ------------------
0.2 - = 2.829 m/s
V 2 = -----
A2 0.07069

Q- = ------------------
0.2 - = 2.310 m/s
V E = ------
AE 0.08657

The head loss in the series system is

6.8214 L 1 1.852 6.8214 L 2 1.852

h f = ---------------- --------------V + ---------------- --------------V 2
1.852 1.167 1 1.852 1.167
C1 D1 C2 D2

6.8214 500 1.852 6.8214 700 1.852

h f = ------------------------ ----------------------  1.018  + ------------------------ ----------------------  2.829 
1.852 1.167 1.852 1.167
 131   0.5   133   0.3 
or

h f = 0.95 + 15.57 = 16.52 m

The head loss in the EUM of diameter 332 mm is

189
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

LE
= 6.8214
1.852 6.8214 - ---------------------------
1200 -  2.310  1.852 = 16.52 m
hf ---------------- --------------V = -----------------------
E 1.852 1.167 E 1.852 1.167
CE DE  132   0.332 

Thus the head loss is the same through the series system and the EUM, as expected. Note though that the head loss is
less than that calculated by the Darcy – Weisbach equation of 17.02 m as computed in Example 7.1.

___________________________________________________________________________________

7.2.4 Multiple Hazen – Williams pipes in series (for both SI units and US
customary units)
The equations for the two pipe case in Section 7.2.3 can be generalized to NP pipes in series to give the
equivalent diameter DE for a given length of pipe LE (usually taken as L E = L 1 + L 2 +  + L NP ) and a
chosen equivalent Hazen – Williams coefficient of C E as

  ------------
1
-
  4.871
 LE 
DE =  -------------------------------------------------------------------------------------------------------------------------------------------
- (7.25)
 1.852 L1 L2 L NP 
CE ------------------------------- + ------------------------------ - +  + ------------------------------ -
4.871 1.852 4.871 1.852 4.871 1.852 
 D1 C 1 D2 C 2 D NP C NP 

EQUIVALENT PIPE DIAMETER FOR MULTIPLE PIPES IN SERIES

(HAZEN – WILLIAMS)

Alternatively, the length LE of an equivalent uniform main of specified diameter DE from rearrangement
of Equation 7.25 is

1.852 4.871 L1 L2 L NP
LE = C E DE ------------------------------ - +  + ------------------------------
- + ------------------------------ - (7.26)
4.871 1.852 4.871 1.852 4.871 1.852
D1 C 1 D2 C 2 D NP C NP

EQUIVALENT PIPE LENGTH FOR MULTIPLE PIPES IN SERIES (HAZEN –

WILLIAMS)

7.3 Pipes in Parallel

Consider two or more pipes in parallel (up to a total of NP) with diameters D1, D2, D3,..., DNP; lengths L1,
L2, L3,..., LNP; roughness heights 1,  2,  3,...,  NP (or Hazen – Williams coefficients C1, C2, C3,..., CNP).
The desired outcome in this case is an equivalent pipe diameter DE of a specified length or an equivalent
pipe length LE of a specified diameter such that the total head loss in both systems (between points A
and B) is the same.

7.3.1 Darcy – Weisbach head loss equation for two pipes in parallel
Consider two pipes in parallel with diameters D1 and D2; lengths L1 and L2; and roughness heights 1 and
 2.

Figure 7.2 Pipes in parallel

190
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Let the “equivalent uniform main” EUM be of diameter DE, length LE and flow velocity VE such that

2
D
Q E = V E A E = V E ----------E- = Q 3 = Q 4 (7.27)
4

The head loss between nodes A and B in Figure 7.2 must be the same for each pipe in parallel and for
the equivalent uniform main thus

hf = hf = hf (7.28)
1 2 E

2 2
L V L V LE V
= f 1 ------1- ------1- = h f = f 2 ------2- ------2- = h f = f E ------- -- (7.29)
1 D 1 2g 2 D 2 2g E DE 2

Thus solving for V1 and V2 in terms of VE from Equation 7.29 gives

D1 f E L E
V 1 = V E ------------
- ------------- (7.30)
f 1 L1 D E

and

D2 f E L E
V 2 = V E ------------
- ------------- (7.31)
f 2 L2 D E

The total flow in the EUM must be the sum of the individual flows in the two parallel pipes

Q E = Q1 + Q2 (7.32)

Now

2 2 2
D D D
V E ----------E- = V 1 ----------1 + V 2 ----------2 (7.33)
4 4 4

Cancelling the   4 in allterms and substituting Equation 7.30 and Equation 7.31 into Equation 7.33
leads to

2 2 2
V E D E = V 1 D1 + V 2 D2 (7.34)

2 2 D1 f E L E 2 D2 f E L
E DE = V E D 1 ------------
f 1 L1 D E
- -------------- + V E D 2 ------------
f 2 L2 D E
- ---------

Cancelling VE in all terms and rearranging gives

5 5
2 DE D1 D2
DE -------------
- = ------------
- + -----------
f E LE f 1 L1 f 2L

and finally

191
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

5 5 5
DE D1 D2
-------------
- = - + -----------
------------ (7.35)
f E LE f 1 L1 f 2L

Solving for the diameter DE of the EUM of length LE gives

 5 5 
5  D1 D2 
D E = f E L E  ------------
- + ------------
- (7.36)
 f 1 L1 f 2 L2 
 
or

 5 5 
0.2 0.2  D1 D2 
DE = f E LE  ------------- + -------------  (7.37)
 f 1 L1 f 2 L2 
 
EQUIVALENT DIAMETER FOR TWO PARALLEL PIPES (DARCY –
WEISBACH)

Alternatively, the equivalent length LE of a pipe of diameter DE is found from Equation 7.36 as

5
DE
L E = -------------------------------------------------------- (7.38)
 5 5 
 D1 D2 
f E  ------------- + ------------- 
 f 1 L1 f 2 L2 

EQUIVALENT LENGTH FOR TWO PIPES IN PARALLEL (DARCY –

WEISBACH)

Example 7.4 Consider two pipes in parallel – one with length 500 m, diameter 500 mm and roughness height
0.25 mm, the other with length 500 m, diameter 300 mm, roughness height 0.10 mm.

(i) Calculate the diameter of an EUM for a 500 m long pipe. Select a suitable friction factor or resistance factor for
the EUM.
(ii) Confirm the head loss for the EUM.

Answer – (i) Computation of the diameter of an equivalent uniform main – pipes in parallel/Darcy – Weisbach.

Assume a velocity of 1.5 m/s in pipe 1 – now compute the friction factor in each pipe. Assume a kinematic viscosity
–6 2
of 1.140 10 m /s .

The Reynolds numbers for pipe 1 and pipe 2 are

VD 1.5  0.5 
Re 1 = -------- = --------------------------- = 658000
 1.140 10
–6

VD 1.5  0.3 
Re 2 = -------- = --------------------------- = 395000
 1.140 10
–6

The Darcy – Weisbach friction factors using the Swamee and Jain equation (Equation 4.61) are

0.25 0.25
f 1 = ---------------------------------------------------------
- = ----------------------------------------------------------------------------------
- = 0.0176
  5.74  2 0.25 5.74 2
log 10 ------------ + ------------ log 10 --------------------- + ----------------------------

 3.7 D 0.9  3.7  500  0.9
Re  658000 

0.25 0.25
f 2 = ---------------------------------------------------------
- = ----------------------------------------------------------------------------------
- = 0.0169
 5.74 2 0.10 5.74 2
log 10  ------------ + ------------ log 10  --------------------- + ----------------------------
 3.7 D 0.9   3.7  300   395000  0.9 
Re

192
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

The area for pipe 1 is

2
D 2
  0.5  - = 0.1964 m 2
A 1 = ---------1- = -----------------
4 4

The flow for pipe 1

3
Q 1 = V 1 A 1 = 1.5  0.1964  = 0.2946 m /s

The head loss for pipe 1 is

2 2
L V 500  1.5 
h f = f 1 ------1- ------1- = 0.0176 --------- ------------------ = 2.018 m
D1 2 g 0.5 2  9.81 

The velocity for pipe 2 to give exactly the same head loss as pipe 1 is

D 2 2 gh 0.3  2   9.81   2.018 

V2 = -------------------f- = -------------------------------------------------- = 1.4057 = 1.186 m/s
f 2 L2 0.0169  500 

The area for pipe 2 is

2
D 2
  0.3  - = 0.07069 m 2
A 2 = ---------2- = -----------------
4 4

The corresponding flow for pipe 2 is

3
Q 2 = V 2 A 2 = 1.186  0.07069  = 0.0838 m /s

The equivalent flow for the EUM will be

3
Q E = Q 1 + Q 2 = 0.2946 + 0.0838 = 0.3784 m /s

Select the EUM friction factor as 0.0170 and the EUM length as 500 m ( L E = L 1 = L 2 ). Thus

f E = 0.0170

and

L E = 500 m

The EUM diameter for the two pipes in parallel from Equation 7.37 is

5 5 0.4 0.
0.2  
5 5
0.2 0.2  D1 D2  0.2  0.5   0.3 
D E = f E L E  ------------ + ------------  =  0.0170   500   ------------------------------ + ------------------------------
 f 1 L1 f 2 L2   0.0176  500  0.0169  500 

Thus the diameter is

D E = 549 mm

___________________________________________________________________________________
Answer – (ii) Confirmation of the head loss in the EUM.

Check the head loss for the EUM at a flow of 378.4 L/s (the combined flow for both pipes 1 and 2) as computed in part
(i).

First compute the area and velocity for the EUM of diameter 549 mm.

2
D   0.549 
2
2
A E = ----------E- = ------------------------ = 0.2367 m
4 4

Q 0.3784
V E = ------E- = ---------------- = 1.598 m/s
AE 0.2367

The head loss for the EUM of diameter 549 mm and length of 500 m is

2 2
L V 500  1.598 
h f = f E ------E- ------E- = 0.0170 ------------- -------------------- = 2.015 m
DE 2 g 0.549 2  9.81 

This value agrees with the result of the head loss in pipe 1 of 2.018 m as computed in part (i).

193
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

___________________________________________________________________________________

7.3.2 Darcy – Weisbach head loss equation for multiple pipes in parallel
Consider multiple pipes (up to NP) in parallel. The head loss between nodes A and B in Figure 7.2 must
be the same for each pipe in parallel and for the equivalent uniform main thus

f1 = hf =  = hf = hf (7.39)
2 NP

The total flow in the EUM must be the sum of the individual flows in the multiple parallel pipes

Q E = Q 1 + Q 2 +  + Q NP (7.40)

Solving for the diameter DE of the EUM of length LE gives

 5 5 5 
5  D1 D2 D NP 
E = f E L E  ------------- + ------------- +  + ---------------------  (7.41)
 f 1 L1 f 2 L2 f NP L NP 
 
or

 5 5 5 
0.2 0.2  D1 D2 D NP 
E = f E L E  ------------ - +  + --------------------
- + ------------ - (7.42)
 f 1 L1 f 2 L2 f NP L NP 
 
EQUIVALENT DIAMETER FOR MULTIPLE PARALLEL PIPES (DARCY –
WEISBACH)

Alternatively, the equivalent length LE of a pipe of diameter DE is found by rearranging Equation 7.42 as

5
DE
E = ---------------------------------------------------------------------------------------------------- (7.43)
 5 5 5 
 D1 D2 D NP 
f E  ------------- + ------------- +  + --------------------- 
 f 1 L1 f 2 L2 f NP L NP 

EQUIVALENT LENGTH FOR MULTIPLE PIPES IN PARALLEL (DARCY –

WEISBACH)

7.3.3 Hazen – Williams head loss equation for pipes in parallel (for both SI
units and US customary units)
The head loss along each parallel pipe must be the same.

hf = hf = hf (7.44)
E 1 2

From Equation 4.104 it follows that for SI units that

6.8214 L E 1.852 6.8214 L 1 1.852 6.8214 L 2 1.852

---------------- --------------- V E = ---------------- --------------
-V = ---------------- --------------
-V (7.45)
1.852 1.167 1.852 1.167 1 1.852 1.167 2
CE DE C1 D1 C2 D2

194
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Again in Equation 7.45, the SI units form of the Hazen – Williams equation has been used. As discussed
earlier for the case of pipes in series, the constant of 6.8214 actually cancels out in all terms so it would
not have mattered if the US customary units form of the equation had been used instead — the result
would have been identical.

Cancelling common terms leads to

LE
1 - -------------- 1.852 L1
1 - -------------- 1.852 L2
1 - -------------- 1.852
-------------- -V E = -------------- -V 1 = -------------- -V (7.46)
1.852 1.167 1.852 1.167 1.852 1.167 2
C E DE C1 D1 C2 D2

Solving for V1 and V2 in terms of VE from Equation 7.46 gives

1.852 1.167
1.852 C 1 D1 LE 1.852
V1 = ------------------------------
- ------------------------------
-V
L1 1.852 1.167 E
C E D E

or

1 -
1.852 1.167 ------------
 C 1 D1 1.852
LE 1.852
V1 =  ------------------------------ -V E 
- ------------------------------ (7.47)
 L1 1.852 1.167
C E DE 

and

1 -
1.852 1.167 ------------
 C 2 D2 1.852
LE 1.852
V2 =  ------------------------------ -V E 
- ------------------------------ (7.48)
 L2 1.852 1.167
C E DE 

Substituting Equation 7.47 and Equation 7.48 into Equation 7.34 gives

1 1
1.852 1.167 ------------- 1.852 1.167 -------------
 C 1 D1 LE  1.852  C 2 D2 LE  1.852
2 2 2
V E D E = D 1 V E  ------------------------------- ------------------------------
- + D 2 V E  ------------------------------- ------------------------------
-
 L 1 C
1.852
D
1.167  L 2 C
1.852
D
1.167
E E E E

Cancelling VE and rearranging gives

0.63 0.54 0.63 0.54

2 2  C 1 D1 LE  2  C 2 D2 LE 
DE = D 1  -------------------- --------------------- + D 2  -------------------- ---------------------
 L 0.54 0.63 0.54 0.63
1 C E DE   L
2 C E DE 

Solving for the diameter DE for a specified length LE (usually taken as equal to L1 or L2) gives

0.54 2.63 2.63

2.63 L E  C 1 D1 C 2 D2 
DE -  -------------------- + -------------------
= ----------- - (7.49)
C E  L 0.54 L2
0.54 
1

1
0.54 2.63 2.63 ---------
2.63
-
L E  C 1 D1 C 2 D2 
DE = ------------  -------------------- + -------------------- (7.50)
C E  L 0.54 L2
0.54 
1

The exponents become

195
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

0.54
---------- = 0.20532  0.2053
2.63

1
---------- = 0.38023  0.3802
2.63

Thus Equation 7.50 becomes

0.2053 2.63 2.63 0.3802

LE  C 1 D1 C 2 D2 
-  -------------------
D E = ---------------- -
- + ------------------- (7.51)
0.3802  0.54 0.54 
C E L 1 L 2

EQUIVALENT DIAMETER FOR TWO PIPES IN PARALLEL (HAZEN –

WILLIAMS)

Now if all pipe lengths are the same, that is, L1 = L2 = LE then

2.63 2.63 0.3802

 C 1 D1 + C 2 D2 
D E =  ----------------------------------------------
- (7.52)
 CE 

EQUIVALENT DIAMETER FOR TWO PIPES IN PARALLEL AND OF

EQUAL LENGTH (HAZEN – WILLIAMS)

Alternatively, the equivalent length LE of EUM of diameter DE for parallel pipes from Equation 7.49 is
given by

1.852 4.871
C E DE
L E = ---------------------------------------------------------------
- (7.53)
2.63 2.63 1.852
C 1 D1 C 2 D2
-------------------- + --------------------
0.54 0.54
L1 L2

or

1.852
2.63
C E DE
LE = ----------------------------------------------- (7.54)
2.63 2.63
C 1 D1 C 2 D2
-------------------- + ------------------- -
0.54 0.54
L1 L2

EQUIVALENT LENGTH FOR TWO PIPES IN PARALLEL (HAZEN –

WILLIAMS)

Example 7.5 Two pipes are connected in parallel between two reservoirs A and B. The pipes have the following
characteristics — L1 = 2500 m, D1 = 1.2 m (diameter of old cast – iron pipe), C1 = 100; L2 = 2500 m, D2 = 1.0 m,
C2 = 90. For a difference in water surface elevations between the reservoirs of 3.6 m determine the total flow of
water at 20C. Base your computations on an EUM of length 2500 m with a C value of 100. Ignore minor losses.

Answer – Computation of the flow in parallel pipes between two reservoirs using an equivalent uniform main.

Write the energy equation between the 2 reservoirs

196
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

2 2
p V p V
z A + ------A + ------A- = z B + -----B- + ------B- + h f
 2g  2g

pA = 0.0, VA = 0.0, pB = 0.0, VB = 0.0, zA – zB = 3.6 m

z A = zB + h f

Thus

h f = 3.6 m

For the Hazen – Williams equation from Equation 4.104

1.852
LV
h f = 6.8214 ----------------------------- (7.55)
1.852 1.167
C D

Compute diameter of the EUM for L E = 2500 m , and C E = 100 from Equation 7.51 as

0.2053 2.63 2.63 0.3802

LE  C 1 D1 C 2 D2 
D E = ----------------
0.3802
 -------------------
0.54
-
+ ------------------
0.54
CE  L1 L2 

Thus

0.2053  0.3802
90  1.0  
2.63 2.63
2500 100  1.2 
D E = ------------------------  ----------------------------- + --------------------------
0.3802 0.54 0.54
100  2500 2500 

Thus the diameter of an equivalent uniform main is

D E = 1.418 m

Solving for the velocity V in Equation 7.55 gives

1 - 1 -
------------ ------------
1.852 1.167
 h f C E D E  1.852  3.6   100 
1.852 1.167
 1.418  - 1.852 = 1.291 m/s
VE =  -----------------------------------
- =  -----------------------------------------------------------------
 6.8214 L E   6.8214  2500  

The area is

2 2
D   1.418  2
A = ---------- = ------------------------ = 1.579 m
4 4

The flow is

3
Q = VA = 1.291  1.579  = 2.04 m /s

___________________________________________________________________________________

7.3.4 Hazen – Williams head loss equation for multiple pipes in parallel (for
both SI units and US customary units)
The head loss along each of the NP parallel pipes must be the same.

fE = hf = hf =  = hf (7.56)
1 2 N

Recalling that for a parallel pipe system, the total flow must be the sum of the individual flows

Q E = Q 1 + Q 2 +  + Q NP (7.57)

197
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Solving for the diameter of the EUM, DE for a specified length LE based on a similar procedure as shown
in Section 7.3.3 gives

0.2053 2.63 2.63 2.63 0.3802

LE  C 1 D1 C 2 D2 C NP D NP 
DE -  -------------------
= ---------------- - +  + -----------------------
- + ------------------- - (7.58)
0.3802  0.54 0.54 0.54 
C E L 1 L 2 L NP

EQUIVALENT DIAMETER FOR MULTIPLE PIPES IN PARALLEL (HAZEN –

WILLIAMS)

Now if all pipe lengths are the same, that is, L 1 = L 2 =  = L NP = L E then

2.63 2.63 2.63 0.3802

 C 1 D 1 + C 2 D 2 +  + C NP D NP 
DE =  -----------------------------------------------------------------------------------------
- (7.59)
 CE 

EQUIVALENT DIAMETER FOR MULTIPLE PIPES IN PARALLEL OF

EQUAL LENGTH (HAZEN – WILLIAMS)

Alternatively, the length LE of EUM of diameter DE for parallel pipes from Equation 7.59 is given by

1.852
2.63
C E DE
LE = ------------------------------------------------------------------------------------------- (7.60)
2.63 2.63 2.63
C 1 D1 C 2 D2 C NP D NP
-------------------- + -------------------- +  + ------------------------
0.54 0.54 0.54
L1 L2 L NP

EQUIVALENT LENGTH FOR MULTIPLE PIPES IN PARALLEL (HAZEN –

WILLIAMS)

s
7.4 Three Interconnected Reservoirs
In the reservoir flow situation shown in Figure 7.3, the flow through each pipe is to be calculated when
the reservoir levels are given.

Figure 7.3 Three interconnected reservoirs

Assume the elevations of the reservoirs are given as

z A = E L A , z B = E L B and z C = E L C

198
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

The Darcy – Weisbach friction loss equation for each of the three pipes must be satisfied. The continuity
equation must also be satisfied at the junction of the three pipes.
Question – What are the two different flow situations that could exist?

Answer – Flow could be either into reservoir B or out of reservoir B depending on the level of the HGL at the node J.

Question – What would the HGL look like for the two different cases?

Answer – In the case of flow into reservoir B, the HGL at junction J must be above the level of reservoir B. For flow
out of reservoir B, the HGL at junction J must be below the level of reservoir B.

Introduce a sign convention such that flow out of a junction is positive and flow into a junction is nega-
tive. Thus in applying the continuity equation to Figure 7.3 then Q1 is negative and Q3 is positive. Q2
may be either positive or negative. The continuity equation at the junction J for the case where flow Q2
is assumed to be into reservoir B and thus positive is

–Q1 + Q2 + Q3 = 0 (7.61)

For the second case (corresponding to the flow directions as shown in Figure 7.3) the continuity equa-
tion at junction J where flow Q2 is assumed to be from reservoir B towards the junction and thus nega-
tive is

–Q1 – Q2 + Q3 = 0 (7.62)

The solution procedure is as follows

1. Assume a sign convention for flow either

(a) out of reservoir B or
(b) into reservoir B
and that in turn determines whether Equation 7.61 or Equation 7.62 applies. For example,
choose the flow directions as depicted by Equation 7.62 above.
2. Assume an elevation of the hydraulic grade line at the junction J
Question – What would the best first estimate be?

Answer – Assume the HGL to be at the level of reservoir B, then Q2 is zero.

3. Compute the flows Q1, Q2 and Q3 from the known head loss (ignore velocity heads and
minor losses as a first approximation) as calculated from the reservoir elevations and the
guessed level of the HGL at J from step 2. Write the energy equation for a fluid particle on
the surface of reservoir A and a fluid particle at the centerline of pipe 1 just upstream of the
junction J

2 2
pA V A p V f 1 L1 V
- + ------- = z J + -----J- + ------1- + ------------
A + ------ - ----- (7.63)
 2g  2g D 1 2g

The elevation of reservoir A is

zA = E LA

The velocity and pressure at the surface of the reservoir are both zero and thus V A = 0 and
pA = 0 .
The hydraulic grade line at J is

p
z J + -----J- = HGL J


199
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Neglecting the velocity head in pipe 1

2
V1
-------  0
2g

Thus for pipe 1 it follows that

2 2
L V L Q1
f1 = z A – HGL J = f 1 ------1- ------1- = f 1 ------1- ------------- (7.64)
D 1 2g D 1 2g A
1

Solving for Q1 in Equation 7.64 gives

2
 z A – HGL J 2g A 1 D 1
Q1 = -----------------------------------------------------
- (7.65)
f 1 L1

Similarly, assuming flow from reservoir B to the junction J and solving for Q2 and Q3

2
 z B – HGL J 2g A 2 D 2
Q2 = ------------------------------------------------------ (7.66)
f 2 L2

This flow need not be computed when the elevation of the HGL at J is taken to be equal to
the elevation of the water surface for the reservoir at B. That is Q 2 = 0 . Finally the flow
from the junction to reservoir C is given by

2
 HGL J – z C 2g A 3 D 3
Q3 = -----------------------------------------------------
- (7.67)
f 3 L3

4. Substitute the values obtained for Q1, Q2 and Q3 above into the continuity equation
(Equation 7.2) to obtain the continuity imbalance at junction J
Q = – Q 1 – Q 2 + Q 3 (7.68)

5. If the imbalance in flow at the junction is

• negative (i.e. a negative Q value from Equation 7.68 such that the flow into the
junction exceeds the flow out of the junction), a higher grade line elevation at J
reduces the flow into the junction and increases the flow out of the junction OR
• positive (i.e. a positive Q value from Equation 7.68 such that the flow into the
junction is less than the flow out of the junction), a lower grade line elevation at J
increases the flow into the junction and decreases the flow out of the junction
6. To estimate the next guess of the HGL at J, plot the HGL at J versus Q as shown in
Figure 7.4 (with the Q computed from either Equation 7.61 or Equation 7.62 depending
on which one was selected in step 1). Point AA in Figure 7.4 represents a pair of values
(HGL at J, resulting Q) for the first guess of the HGL at J, while point BB represents a
corresponding pair for the next guess of the HGL at J. The actual positions of AA and BB
for a particular three reservoir problem clearly depend on the values and the initial guess of
the HGL at junction J.

200
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Figure 7.4 HGL at J versus Q

7. Use a linear interpolation between points AA and BB through the Q = 0 line (the vertical
axis in Figure 7.4 to point CC) or a linear extrapolation to the Q = 0 line to determine the
next estimate of HGL at J.
8. Repeat steps 3 to 7 as necessary until an insignificant change in HGL at J results.
Example 7.6 Three reservoirs A, B and C have the following elevations in Figure 7.3 — zA = 30.0 m, zB = 27.0 m
and zC = 17.0 m. The pipes have the following characteristics — pipe 1: L1 = 1000 m, D1 = 200 mm,  1 = 1.0 mm;
pipe 2: L2 = 300 m, D2 = 200 mm,  2 = 1.0 mm, and pipe 3: L3 = 600 m, D3 = 300 mm,  3 = 3.0 mm. Determine the
flows in each pipe.

Answer – Computation of the flows in a three – reservoir water distribution system.

Computing friction factors in each pipe assuming fully rough turbulent flow using Equation 4.84 (Prandtl – Nikuradse
formula) gives

2 2
1 1
f1 = f2 = ------------------------------------------------------- = --------------------------------------------------------- = 0.0303
 2.0log  ------ D 1   2.0log  200  
 10  - + 1.14  10  --------- + 1.14
 1 1

2 2
1 1
f3 = ------------------------------------------------------- = --------------------------------------------------------- = 0.0379
 2.0log  ------ D 3   2.0log  300  
 10  - + 1.14  10  --------- + 1.14
 3 3

A summary of pipe and node properties is now given.

Table 7.1 Pipe data for the three – reservoir network (completely turbulent flow assumed)

Pipe Length Diameter  f rf

(m) (mm) (mm)
1 1000 200 1.0 0.0303 7824

2 300 200 1.0 0.0303 2347

3 600 300 3.0 0.039 796

Table 7.2 Node data for the three – reservoir network

Node Elevation
(m)
A 30.0

B 27.0

C 17.0

Compute the area of each pipe.

2
D   0.2 
2
2
A 1 = A 2 = ---------1- = ------------------ = 0.0314 m
4 4

2
D   0.3 
2
2
A 3 = ---------2- = ------------------ = 0.07069 m
4 4

201
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Assume the hydraulic grade line at node J is at 27 m (equal to the elevation of reservoir B such that the flow
Q 2 = 0.0 ). The flows (from Equation 7.65 and Equation 7.67) in the pipes are

2
 z A – HG L J 2 g A 1 D 1 2
 30 – 27 2  9.81   0.0314   0.2 - = 0.0196 m 3 /s
Q1 = -----------------------------------------------------
- = ------------------------------------------------------------------------------
f 1 L1 0.0303  1000 

and

2
 HG L J – z C 2 g A 3 D 3 2
---------------------------------------------------------------------------------
27 – 17 2  9.81   0.07069   0.3 - = 0.114 m 3 /s
Q3 = ------------------------------------------------------ =
f 3 L3 0.0379  600 

Substituting into the continuity balance equation from Equation 7.68

3
Q = – Q 1 – Q 2 + Q 3 = – 0.0196 – 0.0 + 0.114 = 0.0944 m /s

Note that the flow in pipe 3 out of junction J is much greater than the flow in pipe 1 into the junction. Thus Q is pos-
itive and this suggests that the hydraulic grade line at node J needs to be lowered to increase Q 1 and decrease Q 3 . Try
a hydraulic grade line at node J of 20 m.

The flows in the pipes are

2
 z A – HG L J 2 g A 1 D 1 2
 30 – 20 2  9.81   0.0314   0.2 - = 0.0357 m 3 /s
Q1 = -----------------------------------------------------
- = ------------------------------------------------------------------------------
f 1 L1 0.0303  1000 

and from Equation 7.66

2
 z B – HG L J 2 g A 2 D 2 2
 27 – 20 2  9.81   0.0314   0.2 - = 0.0545 m 3 /s
Q2 = ------------------------------------------------------ = ------------------------------------------------------------------------------
f 2 L2 0.0303  300 

and

2
 HG L J – z C 2 g A 3 D 3  20 – 17 2  9.81   0.07069   0.3 
2
3
Q3 = -----------------------------------------------------
- = ---------------------------------------------------------------------------------- = 0.0623 m /s
f 3 L3 0.0379  600 

Substituting into the continuity balance equation from Equation 7.68

3
Q = – Q 1 – Q 2 + Q 3 = – 0.0357 – 0.0545 + 0.0623 = – 0.0279 m /s

Note that the flow in pipe 3 out of junction J is now less than the sum of the flows in pipes 1 and 2 into the junction.
Thus Q is negative and this suggests that the hydraulic grade line at node J needs to be increased to decrease Q 1 and
Q 2 and to increase Q 3 .

Interpolating using a straight line between the two previous guesses as shown in the Figure 7.5 provides the next guess
of the hydraulic grade line of 21.7 m.

The flows in the pipes are

2
 z A – HG L J 2 g A 1 D 1 2
 30 – 21.7 2  9.81   0.0314   0.2 - = 0.0326 m 3 /s
Q1 = -----------------------------------------------------
- = ----------------------------------------------------------------------------------
f 1 L1 0.0303  1000 

and

2
 z B – HG L J 2 g A 2 D 2  27 – 21.7 2  9.81   0.0314   0.2 
2
3
Q2 = ------------------------------------------------------ = ----------------------------------------------------------------------------------- = 0.0475 m /s
f 2 L2 0.0303  300 

and

2
 HG L J – z C 2 g A 3 D 3  21.7 – 17 2  9.81   0.07069   0.3 
2
3
Q3 = -----------------------------------------------------
- = -------------------------------------------------------------------------------------- = 0.0780 m /s
f 3 L3 0.0379  600 

202
 Chapter 7 - File: Chap7new.fm July 6, 2022 Version 6

Figure 7.5 Interpolation of the next guess for the HGL at J

Substituting into the continuity balance equation from Equation 7.68

3
Q = – Q 1 – Q 2 + Q 3 = – 0.0326 – 0.0475 + 0.0780 = – 0.0021 m /s

The hydraulic grade line at node J needs to be just a little higher. Try a hydraulic grade line at node J of 21.8 m.

The flows in the pipes are

2
 z A – HG L J 2 g A 1 D 1 2
 30 – 21.8 2  9.81   0.0314   0.2 - = 0.0324 m 3 /s
Q1 = -----------------------------------------------------
- = ----------------------------------------------------------------------------------
f 1 L1 0.0303  1000 

and

2
 z B – HG L J 2 g A 2 D 2  27 – 21.8 2  9.81   0.0314   0.2 
2
3
Q2 = ------------------------------------------------------ = ----------------------------------------------------------------------------------- = 0.0470 m /s
f 2 L2 0.0303  300 

and

2
 HG L J – z C 2 g A 3 D 3  21.8 – 17 2  9.81   0.07069   0.3 
2
3
Q3 = -----------------------------------------------------
- = -------------------------------------------------------------------------------------- = 0.0788 m /s
f 3 L3 0.0379  600 

Substituting into the continuity balance equation from Equation 7.68

3
Q = – Q 1 – Q 2 + Q 3 = – 0.0324 – 0.0470 + 0.0788 = – 0.0006 m /s

This value is sufficiently small (< 1 L/s) and thus the HGL is 21.8 m for this three – reservoir problem.

A summary table of the unknowns that have been found is given below.

Table 7.3 Pipe flows solution for the three – reservoir network

Pipe Flow Flow Head Loss

(m3/s) (L/s) (m)

1 0.0324 32.4 8.2

2 0.0470 47.0 5.2

3 0.0788 78.8 4.8

Table 7.4 Node junction HGL solution for the three – reservoir network

Node HGL(m) or Elevation(m)

J 21.8
For Comparison the Reservoir Elevations

A 30.0

B 27.0

C 17.0

___________________________________________________________________________________

203
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

CHAPTER 8 HARDY CROSS METHOD1

Complex water distribution systems are characterized by numerous pipes being connected. Usually
there are several circuits or loops as well as multiple reservoirs. There may be many entry and with-
drawal points. An example of a complex pipe network is a network for a small town or new subdivision.
Another example of a network is an off – farm irrigation delivery system. Flows in a network may be
computed if the length, diameter, pipe roughness and external flows leaving and entering the network
are known. Because the simultaneous set of governing equations are nonlinear (some may be linear
equations), an iterative or trial and error solution technique is required to solve for the flows and heads
in the network. There are different ways to formulate the governing equations in terms of the unknown
variables. The different formulations are explored in Chapter 9. In this Chapter, the Hardy Cross itera-
tive technique of the governing equations is considered (Cross 1936). It is a successive relaxation tech-
nique. Other more modern iterative techniques suited to computer simulation such as the Newton –
Raphson technique are considered in Chapter 10. Today, most networks are solved using commercial or
Government computer programs.

The Hardy Cross method is a hand iterative solution technique that was commonly used up to the 1970s
prior to the availability of computer programs. The Hardy Cross technique was developed in 1936 at the
University of Illinois. Although it is not used today in iterative computer algorithms, it is useful to
become familiar with the application of the Hardy Cross technique to understand the iterative nature of
the solution technique. Details of the various ways of formulation of the pipe network equations are
given in Chapter 9. Numerical methods for the iterative solution of these equations are given in
Chapter 10 to Chapter 16. Nielsen (1989) described the original Hardy Cross method as corresponding
to only solving the system of equations by one sweep of point relaxation, combined with successive
updating of the coefficient matrix.

8.1 Sign Convention for Flows in Pipes

A sign convention for flow direction around loops is needed to apply the Hardy Cross method. Firstly
each pipe needs to be assigned a number. An arrow denoting the flow direction in each pipe is then
selected and labeled onto the network diagram as shown in Figure 8.1. In addition a loop arrow (usually
in the clockwise direction) is placed in each loop in the network. Upon considering a loop or a circuit,
the following sign convention as shown in Figure 8.1 is adopted

• if the assumed flow direction for the pipe is in the clockwise direction around the loop the
flow is initially treated as positive
• if the assumed flow direction for the pipe is in the anti – clockwise direction around the loop
the flow is initially treated as a negative
As a result of the above sign convention assumed for the flows the following occurs

• head losses in the clockwise direction of flow are considered positive

• head losses in the anti – clockwise (or the counter – clockwise) direction of flow are consid-
ered to be negative
8.2 Basis for the Hardy Cross Method
The method is based on the following

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Ade-
laide, Australia. Part of the book entitled Water Distribution Systems Engineering.

204
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

• the total flow entering a junction must be equal to the total flow leaving the junction (conti-
nuity)
• the algebraic sum of the head losses around each circuit or loop must be zero. Or expressed
in another way the pressure drop between two points in the network must be the same irre-
spective of the path taken
• the head loss equation must be satisfied for each pipe

Figure 8.1 Sign convention for head losses

8.3 The Hardy Cross Iterative Solution Technique

The Hardy Cross method is more or less a method of successive approximations to find the flow rate in
each pipe. A systematic approach dealing with one loop at a time is used. Flows for each pipe are ini-
tially estimated or assumed so that continuity is satisfied at all junctions. A correction term LFk to the
flows in each circuit or loop is then computed in turn and applied to bring the circuits or loops into
closer balance. Minor losses may be included as equivalent pipe lengths. Some pipes may be common to
more than one loop or circuit and have their flow adjusted each time an adjacent loop is considered. The
continuity equations do not form part of the LFk – equations directly as long as continuity of the initial
estimates of flows satisfy continuity at every junction. Nielsen (1989) suggested a method for an initial
vector Q0 that satisfies continuity. These initial flow estimates, while they satisfy continuity, do not gen-
erally satisfy the head loss or LFk – equations (Jeppson 1976). Thus the flows need to be corrected
until a point where the LFk – corrections are insignificantly small. At this point both continuity and
energy equations of the network are satisfied.

8.4 The Correction Term in the Hardy Cross Method

For the Darcy – Weisbach equation of Equation 4.3

fL 2 2
h f = ------------------ Q = r f Q
2
2Dg A

or in completely general terms for the Darcy – Weisbach equation

2
hf = rf Q (8.1)

where

fL fL 8 fL 0.8106 fL
r f = ------------------ = ------------------------------ = --------------------- = ----------------------- (8.2)
2 2 2 2 5 5
2Dg A D g D  gD
2Dg  ----------
 4 

such that

205
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

0.8106 fL
r f = ----------------------- (8.3)
5
gD

This factor is computed for each pipe. The Darcy – Weisbach friction factor or resistance factor f is
determined from the Moody diagram (Figure 4.2) or the Swamee and Jain equation (Equation 4.61).

For the Hazen – Williams equation in SI units (Equation 4.104) then

1.852 1.852 1.852

LV 13.643g L V 13.643g L Q
h f = 6.8214 ------------------------------- = -------------------- --------------- --------------- = -------------------- --------------- ---------------------- (8.4)
1.852 1.167 1.852 1.167 2g 1.852 1.167 1.852
C D C D C D 2g A

Thus

13.643g L 1.852
h f = ---------------------- ---------------------------------------------
-Q (8.5)
1.852 2 1.852
2gC 1.167  D 
D ----------
 4 

1.852
h f = r HW Q (8.6)

where

1.852
13.643  4  L L
r HW = -------------------------------------- --------------- = 10.670 ------------------------------- (8.7)
1.852 1.852 D 4.871 C
1.852 4.871
D
2 C

and finally

10.670L -
r HW = ------------------------------ (8.8)
1.852 4.871
C D

From above and as shown in Chapter 4 (Equation 4.105), the Hazen – Williams equation in SI units is

L 1.852 1.852
h f = 10.670 ------------------------------- Q = r HW Q (8.9)
1.852 4.871
C D

The general form of Equation 8.1 and Equation 8.6 is

n n–1
h f = rQ = rQ Q (8.10)

Table 8.1 Head loss equation r and n values

Formulation r n SI units
Darcy – Weisbach 0.8106 fL 2 L in m, D in m
r f = -----------------------
5
gD
Hazen – Williams 10.670L 1.852 L in m, D in m
r HW = -------------------------------
1.852 4.871
C D

To determine the loop flow correction term in the Hardy Cross method assume that

206
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

0
• Qj is the assumed initial flow for a particular pipe j that satisfies continuity at each junc-
tion.
k 
• Qj is the updated flow in pipe j at iteration k after the application of the loop flow correc-
tion term. Note that this is quite different from the loop flows formulation discussed in
init
Section 9.7 where the initially selected flows Q j remain the same for all iterations.
*
• Q j is the correct flow that is required to be found for pipe j such that the head loss equa-
*
tions are satisfied for each loop and path. These Q j values are unknown at this stage.
• LF s is the loop flow correction term for loop s or a required independent path s. Adjust-
k 
ment of all of the flows in the pipes in the loop by this term brings the flow Q j closer to
the correct flow Q *j .

Equation 8.10 for the correct flows Q *j becomes

n n–1
h f = r  Q *j  = rQ *j Q *j (8.11)

Consider the case where only one more step is required to alter the current estimates of flows to be the
k  *
correct flows. Thus to move the current flow Q j at iteration k to the correct flow Q j , the loop flow
correction term LF s is applied to every pipe in a particular loop or required independent path s that
satisfies

* k 
Q j = Q j + LF s (8.12)

Then for each pipe the head loss is

* n k  n
h f = r  Q j  = r  Q j + LF s  (8.13)

By the binomial expansion applied to the term in brackets in Equation 8.13 then

k  n n k  n – 1 nn – 1 k  n – 2 2
h f = r  Q j  + -----  Q j  LF s + --------------------  Q j   LF s  + 
1! 2!
n  n – 1 2.1  k  0 n
+ ----------------------------------  Q j   LF s 
n!

or

k  n k  n – 1 nn – 1 k  n – 2 2 k  0 n
h f = r Q j  + nQ j  LF s + --------------------  Q j   LF s  +  +  Q j   LF s 
2!

k  2
If LF s is small compared with Q j then all terms  LF s  and higher order are negligible.

Thus an approximation of the above equation is

k  n k  n – 1 k + 1
h f = rQ j  + nQ j  LF s  (8.14)

207
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

k + 1
Note that the iteration count of (k+1) has been added to the term LF s . Now for a circuit or loop
with NPLs pipes the head loss around the circuit should be zero

NP L s
 hf j = 0 (8.15)
j=1

where NPLs = number of pipes in the loop s under consideration. For a path between two reservoirs the
right – hand side of Equation 8.15 should be replaced by the difference in reservoir elevations
E L m – E L q where ELm = HGL or water surface elevation at the reservoir m and E L q = HGL or water
surface elevation at the reservoir q (m or ft).

Thus from Equation 8.11 it follows that

NP L s
* n
 r jQ j  = 0 (8.16)
j=1

To ensure that the losses are in the same direction as the flow direction let

* n * * n–1
Q j  = Q j Q j (8.17)

Now Equation 8.16 becomes

NP L s
* * n–1
 r jQ j Q j = 0 (8.18)
j=1

Substituting Equation 8.12 of

* k  k + 1
Q j = Q j + LF s

into Equation 8.18 gives for each loop s = 1,....., NL gives

NP L s
k  k + 1 k  k + 1 n–1
 r j  Q j + LF s  Q j + LF s = 0 (8.19)
j=1

Recall that LFs is the correction term for the s – th loop or path. Using the binomial expansion and
dropping the second order and higher order terms in LFs gives

NP L s
k  k n n–1 k + 1
 r jQ j  + nQ j  LF s  = 0 (8.20)
j=1

Thus again introducing the absolute value to ensure head losses are in the direction of flow gives

NP L s NP L s n–1
k  k  n–1 k  k + 1
 r jQ j Q j +  nr j Q j LF s = 0 (8.21)
j=1 j=1

208
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

Solving for the correction of flows for a loop LFs in each pipe gives

NP L s
k  k  n–1
–  r jQ j Q j
k + 1 j=1
LF s = ------------------------------------------------------- (8.22)
NP L s n–1
k 
 nr j Q j
j=1

LOOP FLOW CORRECTION FOR HARDY CROSS METHOD

Another way of viewing the loop correction term in Equation 8.22 is as

k + 1 –  net head loss around loop for assumed flows 

LF s = ------------------------------------------------------------------------------------------------------------------- (8.23)
NP L s
 hf 
  n --------- k 
- for the assumed flows
j = 1 Qj 

k + 1
The correction term LF s is added to each flow in the loop or circuit s.

8.5 Procedure for Solving Water Distribution Systems by the Hardy

Cross Method
The following outlines a general procedure for solving network problems by using the Hardy Cross
method as applied to the looped pipe network shown in Figure 8.2.

1. Identify all the loops in the network and number each loop. There is one loop in Figure 8.2.
2. Put a clockwise circular arrow in each loop with a numbered LFs where s = loop or circuit
number (e.g. LF1, LF2,......, LFNL up to the total number of loops NL). Only one loop
flow correction of LF1 applies to Figure 8.2.
3. Mark each junction as nodes A, B, C,......... etc.
5
4. Compute the resistance term r – values (for example r f = 0.8106 fL   g D  ) for the
pipes in Figure 8.2 as shown in Table 8.2.
.
Table 8.2 Data for the one – loop network
Pipe Length (m) Diameter (mm)  f rj
(mm)
AB 800 745 0.25 0.0133 3.840
BC 800 507 0.25 0.0133 26.30
CD 800 457 0.25 0.0121 40.19
DE 800 398 0.25 0.0148 97.86
AE 800 211 0.25 0.0155 2452

5. Assume a distribution of flow in each pipe of the network that satisfies continuity at every
junction or node. Do not assume any pipes have zero flow initially. Flows in pipes in the
same direction as the assumed LFs direction (often clockwise) are assumed to be positive
for the computations below. Flows in pipes opposite to the assumed LFs direction are
assigned to be negative flows for the computations below.

209
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

The magnitudes of the flows at nodes can be based approximately on the size of the resist-
ance terms in each pipe that is connected to the junction under consideration. From
Equation 8.1 solve for the discharge as

hf
Q = ------ (8.24)
rf

Thus if the friction head loss was the same in two pipes leading away from a junction then
the flow in a pipe would be larger in the pipe for a smaller r f value and conversely smaller
in the pipe with the larger r f value.
Note that for junction A in Figure 8.2 the inflow is 300 L/s. The r f value for pipe AB is
small with a value of 5.10 (from Table 8.2) and for pipe AE is much larger with a value of
2452. Thus it may be expected that the flow in pipe AB will be substantially larger than in
pipe AE. An initial flow of 280 L/s has been assumed for pipe AB and an assumed flow of
20 L/s has been assumed for pipe AE. Continuity is satisfied at junction A as required.
Computation of flows in the other pipes is now an easier task. Consider junction B. The
flow in AB has already been set to 280 L/s and the outflow demand is 85 L/s thus for conti-
nuity to be satisfied at junction B the flow in pipe BC must be 195 L/s.

All of the assumed flows for the one loop network are shown in Figure 8.2.

6. Set up a table as shown in Table 8.3 for a loop with 5 pipes to analyze each closed loop
semi – independently. For each pipe in the loop compute the head loss (column 4 of
Table 8.3)

k  k  n – 1
h f = r jQ j Q j where k = 0
j

7. For each loop, the sum of the net head loss around the loop or for the path is

NP L s NP L s
k  k  n–1
 hf =
j
 r jQ j Q j (8.25)
j=1 j=1

k 
around the loop. For each pipe Q j is positive for an assumed clockwise flow in a loop.
Thus the head loss around loop 1 for the assumed flows shown in column 4 of Table 8.3 is
7.158 m.

210
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

Example 8.1 Consider the network shown in Figure 8.3 with demands at the three nodes shown and inflows from
two reservoir nodes. Determine the loop flow corrections for the first iteration of the Hardy Cross method.

Figure 8.2 One loop – network for Hardy Cross method

Answer – The values are given in Table 8.3.

Table 8.3 A Hardy Cross iteration (Darcy – Weisbach equation, n = 2)

(1) (2) (3) (4) (5)

Assumed Head loss for each Hardy Cross
flow for pipe j (m) denominator term for each pipe j
pipe j
(m3/s)
Pipe rj 0 0 0 n – 1 0 n – 1
Q j k=0 r jQ j Q j nr j Q j

AB 5.10 +0.280 0.400 2.856

BC 37.69 +0.195 1.433 14.699

CD 64.33 +0.295 5.598 37.955

DE 135.7 +0.095 1.225 25.781

AE 3746 – 0.020 – 1.498 149.833

NP L s NP L s n–1
0 0 n – 1 0
 r jQ j Q j = 7.158  nr j Q j = 231.124
j=1 j=1

___________________________________________________________________________________

8. For each pipe in the loop compute (column 5 of Table 8.3)

k  n – 1
nr j Q j (8.26)

9. For each loop compute (last entry of column 5 of Table 8.3)

NP L s n–1
k 
 nr j Q j (8.27)
j=1

This value is 231.124.

10. Determine the general flow correction for the loop from the following:

211
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

NP L s
k  k  n–1
–  r jQ j Q j
k + 1 j=1
LF s = ------------------------------------------------------- (8.28)
NP L s n–1
k 
 nr j Q j
j=1

For the example in Table 8.3 the value of the loop flow correction for the first iteration is
given by

5 2–1
0 0
–  r jQ j Q j
1 j=1 7.158 = – 0.03097 = – 0.031 m 3 /s = – 31 L/s
LF 1 = ------------------------------------------------------ = – ------------------
-
5 2–1 231.124
0
 2r j Q j
j=1

Thus

1
LF 1 = – 31 L/s (8.29)

k + 1 k + 1
11. Apply the loop flow correction LF s to all the pipe flows by adding the LF s
to flows used to determine items 7 to 10 above as shown in Table 8.4. Both positive and
negative flows are adjusted.

Table 8.4 Corrected flows for a Hardy Cross iteration (Darcy – Weisbach equation, n = 2)

(1) (2) (3) (4)

Pipe Assumed flow for Correction for Adjusted flow for pipe j
pipe j at iteration loop 1 at iteration (m3/s)
zero (m3/s) one (m3/s)
0 1 1 0 1
Qj LF 1 Qj = Q j + LF 1

AB +0.280 – 0.031 0.249

BC +0.195 – 0.031 0.164
CD +0.295 – 0.031 0.264
DE +0.095 – 0.031 0.064
AE – 0.020 – 0.031 – 0.051

12. For a pipe that is common in another loop, corrections from both loops are made. The cor-
rections to pipes in the first loop are made first before doing calculations for the next loop.
13. Indicate the corrected flows on the network diagram.
14. Check continuity at each junction to determine if any errors have been made.
15. Repeat steps 6 to 13 until either
• the head loss for a loop is balanced within desired limits or
• the loop flow corrections are made as small as desirable
A detailed example of the application of the Hardy Cross method follows (based on the example 12.10
from Streeter et al. (1998).

212
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

Example 8.2 For the network shown in Figure 8.3 compute the flows in each of the pipes by the Hardy Cross
method.

Figure 8.3 Two – loop network for Hardy Cross method

Answer – Compute the flows in each of the pipes by the Hardy Cross method.

First compute the rf values for pipes.

Assume the Darcy – Weisbach friction factor or resistance factor is used such that n = 2. The head loss relationship for
each pipe is given by Equation 8.10 and Table 8.1 as

2 2–1
hf = rf Q = rf Q Q = rf Q Q

The data for the network is given in Table 8.5 with the rf value being computed from Equation 4.123 as given by

fL
r f = ------------------
2
2 gD A

• f = Darcy – Weisbach friction factor or resistance factor (dimensionless)

• L = length of pipe (m or ft)
• g = gravitational acceleration (9.81 m2/s or 32.2 ft2/s)
• D = pipe diameter (m or ft)
• A = pipe area (m2 or ft2)
In the iterative solution below the r f values have been rounded to the nearest integer.

Table 8.5 Data for 5 – pipe network

Pipe Label L D  f rf Initial Q*

(m) (mm) (mm) (m3/s)
1 AB 20 353 0.25 0.020 6.03 +0.070
2 BC 50 440 0.25 0.020 5.01 – 0.030
3 AC 54 495 0.25 0.020 3.00 +0.035 or
– 0.035
4 AD 50 607 0.25 0.020 1.00 +0.015
5 CD 20 440 0.25 0.020 2.00 – 0.035

* Sign of flow depends on assumed loop flow directions as shown in Figure 8.4.

A Microsoft Excel spreadsheet has been created below to show the solution iteration by iteration.

SET OF INITIAL FLOWS THAT SATISFY CONTINUITY

The set of initially assumed flows for the network are shown in Figure 8.4. These initial estimates must satisfy continu-
k + 1
ity of flow at each node. There are two loops. The loop flow correction values for each the two loops of  LF 1
k + 1
and  LF 2 are designated by clockwise arrows in Figure 8.4. When dealing with each loop in turn flows in
Figure 8.4 are designated as positive if the flow is clockwise in the loop and negative if it is anti – clockwise or counter

213
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

clockwise in the loop. For example in loop 1, the flow in pipe BC is taken as negative 30 L/s.

Figure 8.4 Assumed initial flows in network that satisfy continuity at all nodes

ITERATION 1 FOR LOOP 1

Table 8.6 Two loop network – Hardy Cross iteration – Iteration 1 for Loop 1

(1) (2) (3) (4) (5)

Assumed Head loss for each Hardy Cross denominator
flow for pipe j term for each pipe j
pipe j (m3/s) (m)

Pipe rj 0 0 0 n – 1 0 n – 1

Q j k=0 r jQ j Q j nr j Q j

AB 6 +0.070 0.0029400 0.8400

BC 5 – 0.030 – 0.004500 0.3000

AC 3 +0.035 0.003675 0.2100

NP L s NP L s n–1
0 0 n – 1 0
 r jQ j Q j = 0.028575  nr j Q j = 1.3500
j=1 j=1

3
0 0 2 – 1
–  r jQ j Q j
- = – 0.028575
1 j=1 3 1
 LF 1 = ---------------------------------------------------
3
---------------------- = – 0.02117 m /s = – 21.2 L/s or  LF 1 = – 21.2 L/s
0
2–1 1.3500
 j j 2r Q
j=1

Table 8.7 Corrected flows for a Hardy Cross iteration – Iteration 1 for Loop 1

(1) (2) (3) (4)

Pipe Assumed flow for pipe Correction for loop 1 Adjusted flow for pipe
j at iteration zero at iteration one (m3/s) j
(m3/s) (m3/s)
0 1 1 0 1
Qj  LF 1 Qj = Q j +  LF 1

AB +0.070 – 0.0212 0.0488

BC – 0.030 – 0.0212 – 0.0512
AC +0.035 – 0.0212 0.0138

214
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

Figure 8.5 Flow distribution after Iteration 1 (for Loop1)

ITERATION 2 FOR LOOP 2

Table 8.8 Two loop network – Hardy Cross iteration – Iteration 2 for Loop 2

(1) (2) (3) (4) (5)

Assumed Head loss for each Hardy Cross denominator
flow for pipe j term for each pipe j
pipe j
(m3/s) (m)

Pipe rj 1 1 1 n – 1 1 n – 1

Q j k=1 r jQ j Q j nr j Q j

AD 1 +0.015 0.000225 0.0300

CD 2 – 0.035 – 0.002450 0.1400

AC* 3 – 0.0138 – 0.000574 0.082980

NP L s NP L s n–1
1 1 n – 1 1
 r jQ j Q j = – 0.00278  nr j Q j = 0.253
j=1 j=1

* Note the sign of the flow in pipe AC in now NEGATIVE for this iteration in Loop 2

3
1 1 2 – 1
–  r jQ j Q j
2 j=1 – 0.00278 3 2
 LF 2 = ---------------------------------------------------
3
- = – ---------------------- = 0.01099 m /s = 11.0 L/s or  LF 2 = 11.0 L/s
1
2–1 0.253
 2r j Q j
j=1

Table 8.9 Corrected flows for a Hardy Cross – Iteration 2 for Loop 2

(1) (2) (3) (4)

Pipe Assumed flow for pipe j Correction for loop 2 Adjusted flow for
at iteration zero (m3/s) at iteration one (m3/s) pipe j (m3/s)
1 2 1 1 2
Qj  LF 2 Qj = Q j +  LF 2

AD +0.015 0.0111 0.0261

CD – 0.035 0.0111 – 0.0239
AC – 0.0138 0.0111 – 0.00277

215
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

Figure 8.6 Flow distribution after Iteration 2 (for Loop 2)

* Note the sign of the flow in pipe AC in now NEGATIVE for this iteration in Loop 2

ITERATION 3 FOR LOOP 1

Table 8.10 Two loop network – Hardy Cross iteration – Iteration 3 for Loop 1

(1) (2) (3) (4) (5)

Assumed Head loss for each Hardy Cross denominator
flow for pipe j term for each pipe j
pipe j (m3/s) (m)

Pipe rj 2 2 2 n – 1 2 n – 1

Q j k=2 r jQ j Q j nr j Q j

AB 6 0.0488 0.014306 0.58596

BC 5 – 0.05117 – 0.0013092 0.511700

AC* 3 +0.00277 0.000023 0.001662

NP L s NP L s n–1
2 2 n – 1 2
 r jQ j Q j = 0.001237  nr j Q j = 1.114
j=1 j=1

* Note the sign of the flow in pipe AC in now POSITIVE for this iteration in Loop 1

3
2 2 2 – 1
–  r jQ j Q j
- = – 0.001237
3 j=1 3 3
 LF 1 = ---------------------------------------------------
3
---------------------- = – 0.00111 m /s = – 1.11 L/s or  LF 1 = – 1.11 L/s
2
2–1 1.114
 2r j Q j
j=1

Table 8.11 Corrected flows for a Hardy Cross – Iteration 3 for Loop 1

(1) (2) (3) (4)

Pipe Assumed flow for pipe Correction for loop 1 Adjusted flow for pipe
j (m3/s) (m3/s) j (m3/s)
2 3 3 2 3
Qj  LF 1 Qj = Q j +  LF 1

AB 0.0488 – 0.0011 0.0477

BC – 0.0512 – 0.0011 – 0.0523
AC +0.00277 – 0.0011 0.00166

ITERATION 4 FOR LOOP 2

216
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

Table 8.12 Two loop network – Hardy Cross iteration – Iteration 4 for Loop 2

(1) (2) (3) (4) (5)

Assumed Head loss for each Hardy Cross denominator
flow for pipe j term for each pipe j
pipe j (m3/s) (m)

Pipe rj 3 3 3 n – 1 3 n – 1

Q j k=3 r jQ j Q j nr j Q j

AD 1 0.02606 0.000679 0.052120

CD 2 – 0.02394 – 0.001146 0.095760

AC* 3 – 0.00166 – 0.000008 0.009960

NP L s NP L s n–1
3 3 n – 1 3
 r jQ j Q j = – 0.000475  nr j Q j = 0.1578
j=1 j=1

* Note the sign of the flow in pipe AC in now NEGATIVE for this iteration in Loop 2

3
3 3 2 – 1
–  r jQ j Q j
4 0.000475 = 0.0030 m 3 /s = 3.0 L/s or  LF  4  = – 3.0 L/s
- = – –------------------------
j=1
 LF 2 = ---------------------------------------------------
3
- 2
2–1 0.1578
3
 2r j Q j
j=1

Table 8.13 Corrected flows for a Hardy Cross – Iteration 4 for Loop 2

(1) (2) (3) (4)

Pipe Assumed flow for pipe Correction for loop 2 Adjusted flow for
j (m3/s) (m3/s) pipe j (m3/s)
3 4 4 3 4
Qj  LF 2 Qj = Q j +  LF 2

AD 0.0261 +0.0030 0.0291

CD – 0.0239 +0.0030 – 0.0209
AC – 0.00166 +0.0030 0.00135

ITERATION 5 FOR LOOP 1

Table 8.14 Two loop network – Hardy Cross iteration – Iteration 5 for Loop 1

(1) (2) (3) (4) (5)

Assumed Head loss for each Hardy Cross denominator
flow for pipe j term for each pipe j
pipe j (m3/s) (m)

Pipe rj 4 4 4 n – 1 4 n – 1

Q j k=4 r jQ j Q j nr j Q j

AB 6 0.04772 0.013663 0.57264

BC 5 – 0.05228 – 0.0013666 0.522800

AC* 3 – 0.00135 0.000005 0.008100

3 3 n–1
4 4 n – 1 4
 r jQ j Q j = – 0.000008  nr j Q j = 1.1035
j=1 j=1

* Note the sign of the flow in pipe AC in now NEGATIVE for this iteration in Loop 2

217
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

3
4 4 2 – 1
–  r jQ j Q j
5 j=1 – 0.000008 -5 3 5
 LF 1 - = – ------------------------- = 0.72496602 10 m /s = 0.007 L/s or  LF 1 = – 0.007 L/s
= ---------------------------------------------------
3
4
2–1 1.1035
 2r j Q j
j=1

Table 8.15 Corrected flows for a Hardy Cross – Iteration 5 for Loop 1

(1) (2) (3) (4)

Pipe Assumed flow for pipe Correction for loop 1 Adjusted flow for pipe
j (m3/s) (m3/s) j (m3/s)
4 5 5 4 5
Qj LF 1 Qj = Q j +  LF 1

AB 0.04772 0.000007 0.04773

BC – 0.05228 0.000007 – 0.05227
AC – 0.00135 0.000007 – 0.00134

Note that the sign of the flow in pipe AC has switched from positive to negative.

ITERATION 6 FOR LOOP 2

Table 8.16 Two loop network – Hardy Cross iteration – Iteration 6 for Loop 2

(1) (2) (3) (4) (5)

Assumed Head loss for each Hardy Cross denominator
flow for pipe j term for each pipe j
pipe j (m3/s) (m)

Pipe rj 5 5 5 n – 1 5 n – 1

Qj r jQ j Q j nr j Q j

AD 1 0.02907 0.000845 0.058140

CD 2 – 0.02093 – 0.000876 0.08372

AC* 3 0.00134 – 0.000005 0.008040

NP L s NP L s n–1
5 5 n – 1 5
 r jQ j Q j = – 0.000026  nr j Q j = 0.1578
j=1 j=1

* Note the sign of the flow in pipe AC in now NEGATIVE for this iteration in Loop 2

3
5 5 2 – 1
–  r jQ j Q j
6 0.000026- = 0.0001713 m 3 /s = 1.71 L/s or  LF  6  = – 1.71 L/s
- = – –------------------------
j=1
 LF 2 = ---------------------------------------------------
3 2
2–1 0.1499
5
 2r j Q j
j=1

Table 8.17 Corrected flows for a Hardy Cross – Iteration 4 for Loop 2

(1) (2) (3) (4)

Pipe Assumed flow for pipe Correction for loop 2 Adjusted flow for
j (m3/s) (m3/s) pipe j (m3/s)
3 4 4 3 4
Qj  LF 2 Qj = Q j +  LF 2

AD 0.02907 +0.0017 0.02924

CD – 0.02093 +0.0017 – 0.02076
AC 0.00134 +0.0017 0.00151

CONVERGED SOLUTION AFTER 6 ITERATIONS

218
 Chapter 8 - File: Chap8new.fm July 6, 2022 Version 7

Figure 8.7 Flow distribution after 6 iterations

8.6 Desirable Features of the Hardy Cross Method

Errors in calculating the loop flow corrections LFk values are eliminated in the next balance.

8.7 References
Cross, H. (1936). Analysis of flow in networks of conduits or conductors. Bulletin No. 286, University of
Illinois Engineering Experimental Station, Urbana, Illinois.

Jeppson, R. W. (1976). Analysis of flow in pipe networks, Ann Arbor Science, Ann Arbor, Michigan

Nielsen, H. B. (1989). “Methods for analyzing pipe networks.” Journal of Hydraulic Engineering,
Vol.115, No.2, February, 139-157.

Streeter, V.L., Wylie, E.B. and Bedford, K. (1998). Fluid mechanics. 9th Edition. McGraw– Hill.

219
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

CHAPTER 9 FORMULATIONS OF THE WATER

DISTRIBUTION SYSTEM EQUATIONS1

Pipes, pumps, tanks and valves are connected in a water distribution system, often in a complex manner.
A network for a small town or new subdivision is an example. Several loops or circuits are usually pres-
ent in the network. There are usually entry points to the network (for example – tanks, storages and res-
ervoirs) and also many withdrawal points (for example – homes, industries, parks and gardens,
commercial buildings, etc.).

Nodes are defined as the end points of links (pipes, pumps or valves) and are identified as either junction
nodes or fixed head nodes (for example – reservoirs). Flow may enter or leave the network at junction
nodes. Simple loops within a network include all closed pipe circuits that have no additional closed pipe
circuits in them (Wood and Rayes 1981).

There are three types of governing equations for flow and head (HGL or pressure) in a network of pipes.
These include

• continuity of flow at each node

• head loss – flow relationships for each individual pipe
• conservation of energy for each closed simple loop and the required independent paths
(each loop or path comprising a number of links). A path is taken between two reservoirs or
nodes of fixed head – it is like an open loop.
There are a number of alternative formulations of the equations describing a water distribution system
(Wood and Rayes (1981) – for the first three formulations; Q+H equations for the fourth formulation;
Todini and Pilati (1988) for the fifth formulation). The fifth formulation is referred to as the Global Gra-
dient Algorithm (Todini 2006). Five formulations that are presented in detail in the Chapter include:

• flow equations or Q – equations formulation in terms of unknown flows (Qs) in each link
(see Section 9.7)
• head equations or H – equations formulation in terms of unknown heads or HGLs (Hs) at
each node (see Section 9.8)
• loop flow equations formulation or LF – equations in terms of unknown loop flows (LFs)
(see Section 9.9)
• flow and head equations or Q+H equations formulation in terms of unknown flows and
unknown heads (see Section 9.10)
• Global Gradient Algorithm formulation in terms of unknown flows and unknown heads
(Todini and Pilati 1988) (see Section 9.11)
For networks that contain loops, the number of unknown loop flows (LFs) are always less than or equal
to the number of unknown heads (Hs) that in turn are always less than the number of unknown link
flows (Qs). A smaller set of equations may be more concise but the solution may take more iterations
and/or time and thus be less desirable.

Assumptions in the analysis of steady state flow in networks are usually made including

• minor losses (e.g. entrance, exit losses, losses at junctions) are assumed to be small and are
neglected (care needs to be taken in complex systems where they may be important)

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Ade-
laide, Australia. Part of the book entitled Water Distribution Systems Engineering.

220
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

• velocity heads are assumed to be small and are neglected

• the hydraulic grade line (HGL) and energy grade line (EGL) are assumed to be equivalent –
this follows directly from the two previous assumptions
9.1 The Unknowns
Consider a typical water distribution network as shown in Figure 9.1 with demands DMi at each of the
non – reservoir nodes.

Figure 9.1 A ten pipe network with demands (where DM i = demand at node i
and E L m = elevation of reservoir m; these are assumed to be known)

9.1.1 The unknown flows vector

The general column vector (shown as the transpose of a row vector) of flows in a water distribution sys-
tem is

T
q =  Q 1 Q 2  Q NP  (9.1)

where

• q = vector of pipe or link flows (m3/s or ft3/s)

• Q j = flow in pipe or link j (m3/s or ft3/s)
• NP = number of pipes or other links (pumps, valves, orifices etc.)
By convention, lower case bold letters are used to designate vectors while upper case bold letters are
used to designate matrices. In some cases, upper case bold letters are used to designate elements of vec-
tors where by tradition these symbols have been used historically in hydraulics, for example, Q for flow
and H for head or HGL.

For the ten pipe network in Figure 9.2 the flows are shown. The vector of unknown flows for the 10 pipe
network is given as Equation 9.2.

221
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Figure 9.2 A ten pipe network showing the primary unknown Qs and Hs (note that demands are
also shown but these are assumed to be known)

Q1
Q2
Q3
Q4
T Q5
q =  Q 1 Q 2  Q 10  = (9.2)
Q6
Q7
Q8
Q9
Q 10

9.1.2 The unknown heads vector

The general column vector of heads is

T
h =  H 1 H 2  H NJ  (9.3)

where

• h = vector of nodal heads (m3/s or ft3/s)

• H i = nodal head at node i (m or ft)
• NJ = number of unknown head nodes (excluding fixed head nodes such as reservoirs)
The column vector of unknown heads for the seven nodes in the ten pipe network in Figure 9.2 is

222
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

H2
H3
H4
T
h =  H 2 H 3  H 9  = H (9.4)
6
H7
H8
H9

Note that in the vector in Equation 9.4, H 1 and H 5 do not appear because they are known reservoir or
fixed head elevations and are referred to by EL 1 and EL 5 .

The general column vector of unknown pressure heads in a network is

T
p =  p 1 p 2  p NJ  (9.5)

where

p
--- = h – z for each node i = 1,...., NJ (9.6)


where

• p = vector of nodal pressures (N/m2 or lb/ft2)

• p i = nodal pressure head at i (m or ft)
•  = specific weight of water (N/m3 or lb/ft3)
• h = vector of nodal heads (m or ft)
T
• z = vector of nodal elevations =  z 1 , z 2 , z NJ 
• NJ = number of unknown head nodes (excluding fixed head nodes such as reservoirs)
The known column vector of demands for NJ nodes in a general network is

T
d m =  DM 1 DM 2  DM NJ  (9.7)

In addition, define the vector of known demands for the ten pipe network in Figure 9.1 (at each of the
nodes where the nodal head is unknown) as

DM 2
DM 3
DM 4
T
m =  DM 2 DM 3 DM 4 DM 6 DM 7 DM 8 DM 9  = DM (9.8)
6
DM 7
DM 8
DM 9

excluding the demands DM 1 and DM 5 as these nodes are reservoirs.

223
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

9.2 Network Variable Relationships

9.2.1 Theory of graphs and networks
Graph and network theory forms a large subdiscipline of mathematics (Templeman and Yates 1984).
Water distribution networks form only a small subset of general graphs and networks. A graph is defined
by Templeman and Yates (1984) as

“A set of nodes together with a set of arcs that link the nodes.”

Arcs have several associated quantities in a network including

• flow in a pipe or pump or valve

• physical constants and characteristics — internal pipe diameter, pipe length and roughness,
pump characteristics, valve characteristics
A network is defined by Templeman and Yates (1984) as

“A graph whose arcs carry flows that may (in the case of a directed network) or may
not (in the case of an undirected network) have a defined direction along arcs.”

General network theory deals with problems such as assigning flows to arcs and to paths within a net-
work so as to optimize an objective (for example, maximum throughput or minimum flow path length).
A water distribution system is an example of a potentiated network. A potentiated network is defined by
Templeman and Yates (1984) as

“A network that, in addition to arc flows, has nodal potentials that are functionally
related to the arcs and their flows.”

The existence of the nodal potentials and the functional dependence upon arc flows delineates potenti-
ated networks as a special subset of general networks. Graph theoretic models depend on direct utiliza-
tion of continuity equations (Kesavan and Chandrashekar 1972). The general network – theoretic
principles are not applicable to potentiated networks such as water distribution systems (Templeman and
Yates 1984). Specialized solution techniques are required.

9.2.2 Relationships between variables for water distribution systems

Consider a network in the most general terms (based on the nomenclature of both Wood and Rayes 1981
and Boulos and Altman 1991). From graph theory it is shown that the network variables are related as
follows (Boulos and Altman 1991)

NP = NJ + NL +  NF – NC  (9.9)

where

• NP = number of links (including pipes, pumps and valves)

• NJ = number of nodes (excluding reservoirs or fixed head nodes)
• NL = number of closed simple loops (loops that have no interior crossing links — also
called non – overlapping or natural or primary loops)
• NF = number of reservoirs or fixed head nodes
• NC = number of separate disconnected subnetworks (components in graph theory).
• (NF – NC) = NPL = number of required independent paths

Usually NC = 1 for a network but if a link closes in a network that results in separated parts it is possible
to have NC > 1.
A network is subjected to various external actions applied at nodes only (Templeman and Yates 1984).
The external actions are either demands or outflow of volumetric flow or inflows. A sign convention is
adopted where outflows from nodes are assumed to be positive. Nodal potentials are defined for each
network node and for a network are the heads or hydraulic grade line (HGL) elevations at the nodes
(both junction and fixed head nodes).

224
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

For networks containing two or more reservoirs (or other form of fixed head source), it is necessary to
consider  NF – NC  required independent paths between nodes of fixed head in the analysis of a net-
work. A path is defined as a non – intersecting series of links between any two fixed head nodes (e.g.
reservoirs) that does not contain a closed simple loop. Consider an example network that is fully con-
nected such that NC = 1. For two reservoirs ( NF = 2 ) there is only one path between the two reser-
voirs (thus NF – NC = 2 – 1 = 1 ). For three reservoirs ( NF = 3 ) and consequentially three possible
paths between the reservoirs, however, there are two independent paths required
( NF – NC = 3 – 1 = 2 ) although there are a total of 3 possible paths. It does not matter which paths
between the reservoirs are selected in the three reservoir case, as one path between two of the reservoirs
is redundant and is not required.

It generally holds that there are more links NP than nodes NJ (as indicated by Equation 9.9) such that

NP  NJ

In addition there are usually more links NP or nodes NJ than closed simple loops NL plus required inde-
pendent paths  NF – NC  such that

NP  NL +  NF – NC 

For networks with at least one closed simple loop number of nodes (NJ) exceeds the number of closed
simple loops plus required independent paths

NJ  NL +  NF – NC 

The primary unknowns in a network are assumed to be

• the NP link flows (pipes, pumps or valves) Q1, Q2,...., QNP where the flow in link j is desig-
nated as Qj
• the NJ heads or HGLs H1, H2,...., HNJ where the head at node it is designated as Hi
Note that the term link refers to either a pipe, a pump or a valve. For the network in Figure 9.2 the net-
work characteristics are shown in Table 9.1.

From inspection of the network in Figure 9.2, it follows that the number of links is

NP = 10

The right – hand side of Equation 9.9 based on the values in Table 9.1 becomes

NJ + NL +  NF – NC  = 7 + 2 +  2 – 1  = 10

Thus the equality of Equation 9.9 is verified for the network shown in Figure 9.2.The values for the
characteristics of the ten pipe network as defined above are given in Table 9.1.

9.2.3 Treed or determinate networks

A tree – like water distribution system has no closed simple loops. This case is special in that the net-
work is determinate (Templeman and Yates 1984). In a treed network, the following relation holds

NP = NJ +  NF – NC  (9.10)

225
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Thus a treed and connected network has one less link than the number of total nodes. This is the least
number of arcs that can adequately carry the externally applied nodal actions (Templeman and Yates
1984). A network that satisfies Equation 9.10 is not necessarily a connected network. Thus
Equation 9.10 is not a sufficient condition to result in a treed network.
Table 9.1 Network characteristics for the ten pipe network
Quantity Symbol Value
Number of links (pipes in this case) NP 10
Number of nodes* NJ 7
Number of closed simple loops NL 2
Number of fixed head nodes NF 2
Number of separate disconnected subnetworks (or components) NC 1
Number of required independent paths  NF – NC  = NPL 1
Number of total loops and required paths NTL = NL + NPL 3

*excludes fixed head nodes (nodes 1 and 5)

For a treed network, the number of unknown flows equals the number of continuity equations. This lin-
ear system may be solved directly for the unknown Qs without iteration. Link flow rates can thus be
determined solely from the set of continuity equations. The nonlinearity of the energy equations presents
no special difficulties in calculations for a treed network as the flows are easily determined. Substituting
the flows into the head loss expressions for each link yields all the other unknown quantities (HGLs and
pressures).

9.3 Topology Matrices for Networks

A number of matrices are required to describe the topology of a network (Todini and Pilati 1988) includ-
ing:

• The unknown head node incidence matrix = A 1 (the size of the matrix is NP x NJ or the
number of links (NP rows) by number of nodes (NJ columns)). Details are given in
Table 9.2. (This is matrix A12 in Todini and Pilati (1988).)
• The fixed head node incidence matrix = A 2 (the size of matrix is NP x NF or the number of
links by number of fixed head nodes – usually reservoirs or tanks). Details are given in
Section 9.3.2. (This is matrix A10 in Todini and Pilati (1988).)
9.3.1 The unknown head node incidence matrix A 1

The unknown head node incidence matrix A 1 stores information about the connectivity of the nodes
and the links in the network. Elements in the unknown head node incidence matrix A 1 are defined
according to Table 9.2, that considers link (in row j of the matrix) and a node (in column i of the matrix).
It is assumed that each link has a designated flow direction. Elements in the unknown head node inci-
dence matrix A 1 are assigned as:

226
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Table 9.2 Elements in the unknown head incidence matrix

Element Value Explanation
a ji –1 If the link (in row j of the A 1 matrix) enters the node in the designated
flow direction (in column i of the A 1 matrix)

a ji 0 If the link (in row j of the A 1 matrix) is not connected to the node (in
column i of the A 1 matrix)

a ji 1 If the link (in row j of the A 1 matrix) leaves the node in the designated
flow direction (in column i of A 1 matrix)

Note that the unknown head node incidence matrix is not a square matrix.

The full A 1 matrix for the ten pipe network in Figure 9.2 is given below in Equation 9.11. In Table 9.3
the opposite sign convention to that used by Todini and Pilati (1988) is used in order to be consistent
with the sign convention used in the other equation formulations presented earlier in the Chapter.

Node = 02 03 04 06 07 08 09

–1 0 0 0 0 0 0 pipe 1
1 –1 0 0 0 0 0 pipe 2
0 1 –1 0 0 0 0 pipe 3
0 0 –1 0 0 0 0 pipe 4
A1 = 1 0 0 –1 0 0 0 pipe 5 (9.11)
0 1 0 0 –1 0 0 pipe 6
0 0 1 0 0 –1 0 pipe 7
0 0 0 1 –1 0 0 pipe 8
0 0 0 0 1 –1 0 pipe 9
0 0 0 0 0 1 –1 pipe 10

UNKNOWN HEAD NODE INCIDENCE MATRIX (10 x 7)

In a row in Equation 9.11 corresponding to a particular link, if there are two non – zero elements (for
example, row 2) one must have a negative value and the other must have a positive value. Also for this
case, the two end nodes are non – reservoir nodes. In contrast, if a row corresponding to a particular link
only has one non – zero entry (for example, row 1 and row 4) then the link is connected to a reservoir or
fixed head node at one end.

227
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

T
The transpose of A 1 or A 1 for the ten pipe network from Equation 9.11 is

Pipe= 01 02 03 04 05 06 07 08 09 10

–1 1 0 0 1 0 0 0 0 0 node 2
0 –1 1 0 0 1 0 0 0 0 node 3
0 0 –1 –1 0 0 1 0 0 0 node 4
T
A1 = 0 0 0 0 –1 0 0 1 0 0 node 6 (9.12)
0 0 0 0 0 –1 0 –1 1 0 node 7
0 0 0 0 0 0 –1 0 –1 1 node 8
0 0 0 0 0 0 0 0 0 –1 node 9

TRANSPOSE OF UNKNOWN HEAD NODE INCIDENCE MATRIX (7 x 10)

Now consider a detailed example for node 2 in Figure 9.2. The values of the first column (for node 2) of
the unknown head node incidence matrix A 1 are shown in Table 9.3. Node 2 has links 1, 2 and 5 con-
nected to it.
Table 9.3 Unknown head node incidence A1 entries for node 2 (column 1) of the ten pipe network
Element Value Explanation
a j1 –1 Link 1 (row 1 or j of matrix A 1 ) enters node 2 in the designated flow direc-
tion (column 1 or i of matrix A 1 – the negative signs denotes flow entering
the node)
a j1 0 Links 3, 4, 6, 7, 8, 9 and 10 (row j values of matrix A 1 ) are not connected to
node 2 (column 1 or i of matrix A 1 )

a j1 1 Links 2 and 5 (row j values, j = 2 and j = 5 of matrix A 1 ) leave node 2 in the

designated flow direction (column 1 or i of matrix A 1 )

9.3.2 The fixed head node incidence matrix A 2

The fixed head node incidence matrix A 2 stores information about the connectivity of the fixed head
nodes (often reservoirs) and the links in the network. The dimensions of the fixed head node incidence
matrix A 2 are NP x NF (number of links – NP rows by number of reservoirs or fixed head nodes – NF
columns). Elements in the fixed head node incidence matrix A 2 are defined according to Table 9.4, that
considers a fixed head node (in column f of the matrix) and a link (in row j of the matrix).

Table 9.4 Elements in the fixed head node incidence matrix A 2

Element Value Explanation

a jf –1 If the link (in row j of matrix A 2 ) enters the fixed head node in the
designated flow direction (in column f of matrix A 2 )

a jf 0 If the link (in row j of matrix A 2 ) is not connected to the fixed head
node (in column f of matrix A 2 )

228
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Table 9.4 Elements in the fixed head node incidence matrix A 2

a jf 1 If the link (in row j of matrix A 2 ) leaves the fixed head node in the
designated flow direction (in column f of matrix A 2 )

The fixed head loop incidence matrix A 2 for the ten pipe network in Figure 9.3 is defined in
Equation 9.13 based on Table 9.4. Note that the fixed head node incidence matrix is not a square matrix.

Reservoir = 01 05

pipe 1
1 0
pipe 2
0 0
pipe 3
0 0
pipe 4
0 1
pipe 5
A2 = 0 0
pipe 5 (9.13)
0 0
pipe 6
0 0
pipe 7
0 0
pipe 8
0 0
pipe 9
0 0
pipe 10

FIXED HEAD NODE INCIDENCE MATRIX (10 x 2)

There are two fixed head nodes (1 and 5) in the two columns in Equation 9.13. Each row refers to pipe 1,
2, etc. up to 10. In a row corresponding to a particular pipe if there is a non – zero entry of 1 then the pipe
is connected to a reservoir.

Consider the elements in the fixed head incidence matrix for reservoir 1. The values of the first column
of the fixed head node incidence matrix are shown in Table 9.5.

Table 9.5 Fixed head node incidence A 2 entries for reservoir 1 (column 1) of the ten pipe network

Element Value Explanation

a j1 –1 There are no – 1 entries in column 1 of matrix A 2 (for reservoir 1, f = 1) as
there are no links (all row j values) that enter reservoir 1
a j1 0 For all other pipes 2 to 10 (row j values) these entries are 0 in matrix A 2
because they are not connected to reservoir 1 (column 1 or f = 1)
a 11 1 This entry is +1 in matrix A 2 as the link 1 (row 1 value) leaves reservoir 1
(column 1 value or f = 1)

9.4 Continuity of Flow at Nodes

Conservation of mass at nodes or junctions in a water distribution system yields a set of linear algebraic
equations in terms of flows. Flow continuity at each node is expressed as

Flow In = Flow Out + Demand (or Withdrawal Flow) (9.14)

where

• Flow In = total flow in all the links leading towards the junction (m3/s or ft3/s)

229
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

• Flow Out = total flow in all the links leading away from the junction (m3/s or ft3/s)
• Demand = demand at the node (m3/s or ft3/s)
9.4.1 The continuity equation at a node
A general expression for the continuity of flow at node i with a demand designated as DMi (also referred
to as an offtake, delivery or outflow from a network) is given by Equation 9.15 as

NP
  a ji Q j  + DM i = 0 for nodes i = 1,2,..., NJ (9.15)
j=1

CONTINUITY EQUATION AT A NODE OR JUNCTION

where

• NP = number of links
• a ji = element of the A 1 unknown head node incidence matrix (see Equation 9.12 with val-
ues zero, – 1 or +1), thus only Qj values for pipes connected to the node are included.
• Qj = flow in link j (m3/s or ft3/s)
• DMi = demand at the node i (m3/s or ft3/s)
• NJ = total number of nodes or junctions (excluding fixed head nodes such as reservoirs)
The sign convention adopted is such that a flow away from the node is considered to be positive. Thus
the demand DMi at a node is also positive in sign.

The total inflow to the system must be equal to the total outflow. As a result there must be a linear
dependence between the NJ + NF equations written for all nodes (including fixed head nodes). Thus
only NJ +  NF – NC  of the continuity equations are required and a total of NC of the continuity equa-
tions are superfluous where NC = number of separate disconnected networks. Usually NC = 1 as men-
tioned earlier.

9.4.2 The continuity equations in matrix form for the ten pipe network
The continuity equations for the ten pipe network in Figure 9.2 based on Equation 9.15 are given in
Table 9.6.
Table 9.6 The seven continuity equations for the ten pipe network
Node Continuity equation Node
2 – Q 1 + Q 2 + Q 5 + DM 2 = 0 7 – Q 6 – Q 8 + Q 9 + DM 7 = 0

3 – Q 2 + Q 3 + Q 6 + DM 3 = 0 8 – Q 7 – Q 9 + Q 10 + DM 8 = 0

4 – Q 3 – Q 4 + Q 7 + DM 4 = 0 9 – Q 10 + DM 9 = 0

6 – Q 5 + Q 8 + DM 6 = 0

Putting the continuity equations in Table 9.6 into matrix form using the transpose of the unknown head
node incidence matrix A 1 (Equation 9.12, page 228), the vector of unknown flows q (Equation 9.2,
page 222) and the vector of known nodal demands d m leads to

230
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Q1
Q2
DM 2
–1 1 0 0 1 0 0 0 0 0 Q3
DM 3
0 –1 1 0 0 1 0 0 0 0 Q4
0 0 –1 –1 0 0 1 0 0 0 DM 4
Q5
0 0 0 0 –1 0 0 1 0 0 + DM = 0 (9.16)
6
Q6
0 0 0 0 0 –1 0 –1 1 0 DM 7
0 0 0 0 0 0 –1 0 –1 1 Q7
DM 8
0 0 0 0 0 0 0 0 0 –1 Q8
DM 9
Q9
Q 10

or

T
A1 q + dm = 0 (9.17)

Note that each row of Equation 9.16 represents the continuity equation for each node where the head is
unknown.

9.5 Head Loss for a Pipe

The head loss in a pipe in the water distribution system may be computed from a number of empirically
obtained equations. The two commonly used equations are the Darcy – Weisbach head loss equation and
the Hazen – Williams flow equation as discussed in Chapter 4. The general expression for the head loss
in a pipe j located between nodes i and k is given by

n–1
h f = r jQ j Q j for j = 1,......,NP (9.18)
j

n–1
H i – H k = r jQ j Q j for j = 1,......,NP (9.19)

HEAD LOSS EQUATION FOR A PIPE

where

• h f = head loss in pipe j (m or ft)

j

• rj = resistance term for the pipe j depending on the head loss relationship assumed for the
pipe (for example, Darcy – Weisbach or Hazen – Williams: see Section 4.8 for formulas)
• Qj = flow in the pipe (m3/s or ft3/s)
• n = exponent of the flow in the head loss equation (Darcy – Weisbach n = 2 or Hazen –
Williams equation n = 1.852)
• Hi = nodal head or HGL at one end of the pipe at node i (m or ft)
• Hk = nodal head or HGL at the other end of the pipe at node k (m or ft)
• NP = number of pipes
Usually the values of head at nodes and flows in pipes or links (Hi, Hk and Qj) are all unknown. Recall
that the absolute sign in both Equation 9.18 and Equation 9.19 ensures that the head loss occurs in the
direction of the flow and avoids raising a negative number to an exponent when the flow direction is
negative (Bhave 1988). To obtain a unique solution one of the heads at a node somewhere in the network
must be known.

231
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

For the Darcy – Weisbach head loss equation with n = 2 then Equation 9.19 becomes

2–1
H i – H k = r jQ j Q j = r jQ j Q j for j = 1,......,NP (9.20)

For the Hazen – Williams head loss equation with n = 1.852 then Equation 9.19 becomes

1.852 – 1 0.852
H i – H k = r jQ j Q j = r jQ j Q j for j = 1,......,NP (9.21)

9.6 Various Formulations of the Governing Equations

Five different formulations of the network equations are considered in this Chapter including

• Q – equations formulation or flow equations in which the NP link flows (including flows in
pipes, pumps or valves) are treated as unknowns (see Section 9.7). A set of continuity equa-
tions at nodes are written for the network. In addition, energy losses in closed simple loops
and required independent paths are expressed in terms of flows (Ormsbee and Wood 1986).
Once the flow in each link has been determined, the heads or HGLs and thus the pressures
at each node may be calculated. (see Section 9.7)
• H – equations formulation or head equations in which the NJ heads are treated as unknowns
(see Section 9.8). The flow resistance equations for each link are formulated in terms of
heads rather than flows. The flows (Qs) in the continuity equations are substituted for in
terms of the nodal heads (Hs) (Ormsbee and Wood 1986). The resultant set of equations is
smaller in number than for the Q – equations, however, all equations are nonlinear (Jeppson
1976). Shamir and Howard (1968) proposed the simultaneous node method for solving NJ
nodal head equations. Once the heads have been computed the flow rates may be deter-
mined. (see Section 9.8)
• LF – equations formulation or the loop flow equations in which NL +  NF – NC  loop
flows and required independent path flows (LFs) are treated simultaneously as unknowns
(where NL = number of closed simple loops, NF = number of fixed head nodes and
NC = number of disconnected separate networks) (see Section 9.9). Epp and Fowler (1970)
proposed a simultaneous loop (path) computer solution technique to solve the loop flow
and required independent path equations. The loop flows are computed for each of the NL
closed simple loops and each of the  NF – NC  required independent paths between fixed
head nodes (see Section 9.8). A set of flows must be initially assumed in the network that
must satisfy continuity at each node. This method is similar to the hand – computation
Hardy Cross solution technique (see Chapter 8 & Cross 1936) developed in the 1930s
except here all loop flows are solved for simultaneously. In contrast, only one loop flow
correction is computed at a time in the Hardy Cross method.
• Q+H equations formulation or flow and head equations formulation in which both the
flows and heads are solved for simultaneously (see Section 9.10)
• Global Gradient Algorithm formulation (Todini and Pilati 1988) in which both the heads
and flows are solved for simultaneously in a sequential iterative process based on a smart
partitioning of the problem using a block matrix from of the continuity and the head loss
equations (see Section 9.11).

232
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

For the example network in Figure 9.2 the number of simultaneous equations in each of the formulations
is given in Table 9.7.
Table 9.7 Numbers of equations for various formulations for the ten pipe network

Formulation Continuity Simple loop and Link Total Number of

equations required independent head loss equations unknowns
path energy equations equations
Q – equations NJ = 7 NL +  NF – NC  = 3 – 10 10 Qs
H – equations NJ = 7 – – 7 7 Hs
LF – equations – NL +  NF – NC  = 3 – 3 3 LFs
Q+H equations NJ = 7 – NP = 10 17 10 Qs
and 7 Hs
Global Gradient NJ = 7 – NP = 10 7*, 10 7 Hs and
Algorithm equa- 10 Qs *
tions

* In each iteration, the NJ heads are solved for first and then the NP flows are solved for separately in a sequential process.

where for Figure 9.3 and Table 9.7 the variables are defined in Section 9.2.2, page 224.

9.7 Q – equations Formulation for a Network

The flow equations are formed from both the continuity equations at each of the nodes and the energy
equations for simple closed loops and required independent paths in the network. The Q – equations for
the ten pipe network are given as an example in this Section.

The equations for the ten pipe network shown in Figure 9.2 are presented in this section. The combina-
tion of the continuity equations (Table 9.8), the energy equations for the closed simple loops and the
energy equations for the required independent paths (Table 9.11) gives 10 equations in 10 unknowns.
There is no direct solution of this set of nonlinear equations.

The unknown flows Qjs are labeled on Figure 9.3. The direction of flow in each pipe has been selected
based on an expected flow pattern in the network. The direction may have been selected incorrectly in
some pipes but this does not matter. In the final solution, if a pipe flow turns out to be negative, then that
flow was assumed to be in the wrong direction initially.

To implement this formulation the loops need to be defined via the loop matrix (Todini and Pilati 1988).
This may be provided through input data or can be obtained by partitioning the node matrix A 1 (see
Section 9.11.2). Details are given in Todini and Pilati (1988). In addition, required independent paths
between nodes of fixed head also have to be defined.

9.7.1 The continuity equations

Seven continuity equations based on Equation 9.15 for the non – reservoir nodes in Figure 9.3 are shown
in Table 9.8. It is assumed that there is a demand DMi at each of the non – reservoir nodes in Figure 9.3

Table 9.8 The continuity equations for the ten pipe network
Node Function Continuity equation Node Function Continuity equation
2 f 1  q  – Q 1 + Q 2 + Q 5 + DM 2 = 0 7 f 5q – Q 6 – Q 8 + Q 9 + DM 7 = 0

3 f 2q – Q 2 + Q 3 + Q 6 + DM 3 = 0 8 f 6q – Q 7 – Q 9 + Q 10 + DM 8 = 0

4 f 3q – Q 3 – Q 4 + Q 7 + DM 4 = 0 9 f 7q – Q 10 + DM 9 = 0

6 f 4q – Q 5 + Q 8 + DM 6 = 0

233
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Figure 9.3 Unknown Q – equations formulation for a ten pipe network (note demands are also
shown but these are assumed to be known)

In these 7 continuity equations there are 10 unknown Qjs. These 7 equations are independent because
two external flows from each of the reservoirs are unknown. Another three equations based on energy
equations are needed in order to provide a complete set of equations that could then be solved for the 10
unknown Qjs. These are presented below.

9.7.2 The energy equations for closed simple loops and required indepen-
dent paths for the ten pipe network
Energy equations around closed simple loops (also referred to as non – overlapping or natural loops)
(Jeppson 1976) or between fixed head nodes along required independent paths in a network are nonlin-
ear. Upon traversing a closed simple loop the sum of pipe head losses around the loop must be zero. This
is expressed as

 hf j = 0 for each closed simple loop s = 1,..., NL (9.22)

j  Ls

Figure 9.4 shows the head losses in each pipe that coincide with the directions of the flows assumed in
Figure 9.3.

234
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Figure 9.4 Head losses in each of the pipes for the ten pipe network corresponding to the assumed
sign convention adopted for the Qs shown in Figure 9.3 (note demands are also shown but these
are assumed to be known)

The head loss equations for each of the pipes in a closed simple loop are now expressed in terms of
unknown flows in the pipes by using Equation 9.18 as:

n–1
 r jQ j Q j = 0 s = 1,2,..., NL (9.23)
j  Ls

ENERGY EQUATION FOR A CLOSED SIMPLE LOOP

where

• Ls = set of indices of pipes in a closed simple loop s

• rj = resistance term for the pipe j depending on the head loss relationship assumed for the
pipe (for example, Darcy – Weisbach or Hazen – Williams: see Section 4.8 for formulas)
• Qj = flow in the pipe (m3/s or ft3/s)
• NL = number of closed simple loops {= NP – NJ –  NF – NC  }
Note that the each loop and required independent path must be traversed in the same direction as the
loop flow arrow.

For the Darcy – Weisbach head loss equation with n = 2 then Equation 9.23 for a closed simple loop
becomes

2–1
 r jQ j Q j = 0 (9.24)
j  Ls

or

 r jQ j Q j = 0 (9.25)
j  Ls

For the Hazen – Williams head loss equation with n = 1.852 then Equation 9.23 for a closed simple
loop becomes

235
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

1.852 – 1
 r jQ j Q j = 0 (9.26)
j  Ls

or

0.852
 r jQ j Q j = 0 (9.27)
j  Ls

The energy equations for each of the closed simple loops in the example network in Figure 9.4 are now
given in Table 9.9 in terms of head losses in each pipe. Each loop should be traversed in the same direc-
tion as the loop flow direction (usually clockwise – by convention rather than necessity).
Table 9.9 The energy equations for closed simple loops in terms of head losses for the ten pipe
network
Loop Energy equation Loop Energy equation
number number

1 f2 + hf – hf – hf = 0 2 f3 + hf – hf – hf = 0
6 8 5 7 9 6

Note that for the set of indices of pipes in this first closed simple loop, Ls = {2, 6, 8, 5}.

If the assumed flow in a pipe is in a clockwise direction around the loop then the head loss in that pipe is
assumed to be positive (i.e. the HGL decreases as the pipe is traversed from one end of the pipe to the
other). If the flow in the pipe is in an anticlockwise (counter clockwise) direction around the loop then
the head loss in that pipe is assumed to be negative (i.e. the HGL increases as the pipe is traversed from
one end of the pipe to the other).

In addition to closed simple loop energy equations,  NF – NC  required independent paths between
reservoirs or fixed head nodes in the network must also be considered.

Consider a series of NPLs pipes between two reservoirs designated as reservoir m and reservoir q. The
head loss in the pipes between 2 reservoirs must be equal to the difference in elevation or HGL between
the 2 reservoirs. Each path should be traversed in the same direction as the loop flow direction (usually
clockwise). This may be expressed as

 h f j = E Lm – E Lq s = 1,2,...,(NF – NC) (9.28)

j  Ls

where

• Ls = set of indices of pipes in a closed simple loop s

• h f = head loss in individual pipes j between the two reservoirs (m or ft)
j

• ELm = HGL or water surface elevation at the reservoir m (m or ft)

• E L q = HGL or water surface elevation at the reservoir q (m or ft)
•  NF – NC  = number of required independent paths
Again by substituting Equation 9.18 for each pipe in the path into Equation 9.28 gives

n–1
 r jQ j Q j = E Lm – E Lq (9.29)
j  Ls

ENERGY EQUATION FOR A REQUIRED INDEPENDENT PATH

236
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Note that application of Equation 9.29 may be a little tricky. It is best to traverse the loop in the direction
of the loop flow arrow between the reservoirs – then make the right – hand side equal to the head loss
difference between the reservoirs. Finally move the right – hand side term to the left – hand side of the
equation so that the function can be equal to zero.

For the Darcy – Weisbach head loss equation with n = 2 for the required independent paths then
Equation 9.29 becomes

 r j Q j Q j = E Lm – E Lq (9.30)
j  Ls

For the Hazen – Williams head loss equation with n = 1.852 for the required independent paths then
Equation 9.29 becomes

0.852
 r jQ j Q j = E Lm – E Lq (9.31)
j  Ls

For the ten pipe network in Figure 9.3, one required independent path equation is needed between the
reservoir at node 5 and at node 1 as shown in Table 9.10.
Table 9.10 The energy equation for the required independent path in the ten pipe network
Path Energy equation
1 – h f – h f – h f –  E L5 – E L1  =
f4 3 2 1

The same sign convention regarding the sign of the assumed head loss and the assumed flow direction in
the pipe is assumed for required independent path energy equations as for closed simple loop energy
equations. If the closed simple loop and required independent path energy equations in Table 9.9 and
Table 9.10 are now expressed in terms of flows as expressed in Equation 9.19 as the column vector
T
q =  Q 1 Q 2  Q NP  then a set of nonlinear algebraic equations results.

The head loss or energy equations for the two closed simple loops and one required independent path in
the ten pipe network in Figure 9.3, Table 9.9 and Table 9.10 are now expressed in terms of unknown
flows in the pipes by using Equation 9.19.

Substituting for the head losses in each of the pipes in the loops and required independent paths in
Table 9.9 and Table 9.10 leads to Table 9.11.

Table 9.11 The energy equations for simple closed loops and required independent paths in terms
of flows for the ten pipe network

Loop Function Energy equation

1 f 8q r 2 Q2 Q2
n–1
+ r 6 Q6 Q6
n–1
– r 8 Q8 Q8
n–1
–r 5 Q5 Q5
n–1
= 0

2 f 9q r 3 Q3 Q3
n–1
+ r 7 Q7 Q7
n–1
– r 9 Q9 Q9
n–1
– r 6 Q6 Q6
n–1
= 0

Path 1 f 10  q  r 4 Q4 Q4
n–1
–r 3 Q3 Q3
n–1
–r 2 Q2 Q2
n–1
– r 1 Q1 Q1
n–1
–  E L5 – E L1  = 0

237
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

9.7.3 The energy equations for the Darcy – Weisbach head loss equation
For the Darcy – Weisbach head loss equation with n = 2 then Table 9.11 is rewritten as Table 9.12.

Table 9.12 The energy equations for simple closed loops and required independent paths in terms
of flows for the ten pipe network for the Darcy – Weisbach head loss equation

Loop no. Function Energy equation

1 f 8q r 2 Q2 Q2 + r 6 Q6 Q6 – r 8 Q8 Q8 –r 5 Q5 Q5 = 0

2 f 9q r 3 Q3 Q3 + r 7 Q7 Q7 – r 9 Q9 Q9 – r 6 Q6 Q6 = 0

Path 1 f 10  q  r 4 Q4 Q4 –r 3 Q3 Q3 –r 2 Q2 Q2 – r 1 Q1 Q1 –  E L5 – E L1  = 0

9.7.4 The energy equations for the Hazen – Williams head loss equation
For the Hazen – Williams head loss equation with n = 1.852 , then Table 9.11 is rewritten as Table 9.13.

Table 9.13 The energy equations for closed simple closed loops and required independent paths in
terms of flows for the ten pipe network for the Hazen – Williams head loss equation

Loop no. Func- Energy equation

tion
1 f 8  q  r Q Q 0.852 + r Q Q 0.852 – r Q Q 0.852 – r Q Q 0.852 = 0
2 2 2 6 6 6 8 8 8 5 5 5

2 f 9  q  r Q Q 0.852 + r Q Q 0.852 – r Q Q 0.852 – r Q Q 0.852 = 0

3 3 3 7 7 7 9 9 9 6 6 6

Path 1 f 10  q  r Q Q 0.852 – r Q Q 0.852 – r Q Q 0.852 – r Q Q 0.852 –  E L – E L  = 0

4 4 4 3 3 3 2 2 2 1 1 1 5 1

9.7.5 The flow or Q – equations for the ten pipe network

The seven continuity equations in Table 9.8 and the three energy equations in Table 9.11 (also
Table 9.12 and Table 9.13) gives a total of 10 equations in 10 unknown Qjs. These are the flow or Q –
equations and are shown in Table 9.14.
Table 9.14 The flow or Q – equations for the ten pipe network
Func- Continuity equations, energy equations for simple loop or required independent
tion paths
f 1q – Q 1 + Q 2 + Q 5 + DM 2 = 0 f 5q – Q 6 – Q 8 + Q 9 + DM 7 = 0

f 2q – Q 2 + Q 3 + Q 6 + DM 3 = 0 f 6q – Q 7 – Q 9 + Q 10 + DM 8 = 0

f 3q – Q 3 – Q 4 + Q 7 + DM 4 = 0 f 7q – Q 10 + DM 9 = 0

f 4q – Q 5 + Q 8 + DM 6 = 0

f 8q r 2 Q2 Q2
n–1
+ r 6 Q6 Q6
n–1
– r 8 Q8 Q8
n–1
–r 5 Q5 Q5
n–1
= 0

f 9q r 3 Q3 Q3
n–1
+ r 7 Q7 Q7
n–1
– r 9 Q9 Q9
n–1
– r 6 Q6 Q6
n–1
= 0

f 10  q  r Q Q n – 1 – r Q Q n – 1 – r Q Q n – 1 – r Q Q n – 1 –  E L – E L  = 0
4 4 4 3 3 3 2 2 2 1 1 1 5 1

238
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

9.8 Nodal Head Equations or the H – equations formulation

The nodal head equations are formed from the continuity equations (Equation 9.15) and the link flow
equations (Equation 9.19) by eliminating the flow terms between both sets of equations. A formulation
in terms of head or H – equations has two main advantages (Nielsen 1989) including

• the system of NJ equations in the unknowns H1, H2,...., HNJ is unique

• it is easy to implement the solution procedure in a computer code. It is not necessary to
define the loops for the heads formulation and this is a huge advantage of the technique.
Only the link connectivity information is needed including the node numbers at the end of
each link, the length, diameter and roughness of each pipe and the properties of each node
(elevation and demand) (Todini and Pilati 1988).
On the other hand there are significant disadvantages including

• the difficulty of finding a good starting vector

• the possibility of non – convergence or slow convergence due to oscillation
Nielsen (1989) supported the suggestion of Wood and Rayes (1981) that the formulation in terms of
heads should not be used.

9.8.1 The H – equations formulation for the ten pipe network

Figure 9.5 Unknown H – equations formulation for the ten pipe network (note demands are also
shown but these are assumed to be known; in addition, the Qs are solved for as an intermediate
step to update the Darcy-Weisbach friction factors or resistance factors at each iteration)

For the ten pipe network in Figure 9.5 the head loss equations in terms of the seven unknown heads
using Equation 9.19 are given in Table 9.15.

For the 10 equations in Table 9.15 there are 17 unknowns (7 unknown heads His and 10 unknowns flows
Qjs). All 17 unknowns could be solved by solving both the 7 continuity equations in Table 9.8 and the
10 head loss equations in Table 9.15 (this is a variation of the Q+H equations formulation that is
described in Section 9.10 but it is not the Global Gradient Algorithm formulation that is presented in
Section 9.11). The head equations may be determined by rewriting the link head loss equations from
Table 9.15 in terms of flow and then substituting into the continuity equations in Table 9.8.

Table 9.15 The head loss equations for each pipe in the ten pipe network
Pipe no. Head loss equation Pipe no. Head loss equation

239
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Table 9.15 The head loss equations for each pipe in the ten pipe network

1 n–1 6 n–1
E L1 – H 2 = r 1 Q1 Q1 H 3 – H 7 = r 6 Q6 Q6

2 n–1 7 n–1
H 2 – H 3 = r 2 Q2 Q2 H 4 – H 8 = r 7 Q7 Q7

n–1 n–1
3 H 3 – H 4 = r 3 Q3 Q3 8 H 6 – H 7 = r 8 Q8 Q8

n–1 n–1
4 E L5 – H 4 = r 4 Q4 Q4 9 H 7 – H 8 = r 9 Q9 Q9

5 n–1 10 n–1
H 2 – H 6 = r 5 Q5 Q5 H 8 – H 9 = r 10 Q 10 Q 10

From Equation 9.19 the pipe flow in pipe j is expressed (Bhave 1988) as

1  n – 1
Hi – Hk Hi – Hk  1  1  n –
Q j = ------------------------------------------------------------------- =  ----------  H i – H k  H i – H k (9.32)
1n  r 1  n
rj j

Table 9.16 Head loss equations in terms of flows for each pipe in the ten pipe network
Pipe Head loss or pipe energy equation Pipe Head loss or pipe energy equation
no. no.
1 1 -  H – H  H – H 1  n – 1
--- – 1 Q 6 = ---------
1 n 1n 3 7 3 7
1 Q 1 = ----------  EL 1 – H 2  EL 1 – H 2 6 r6
1n
r1

1 1  n – 1 1 1  n – 1
Q 2 = ----------  H 2 – H 3  H 2 – H 3 Q 7 = ----------  H 4 – H 8  H 4 – H 8
2 1n 7 1n
r2 r7

1 -  H – H  H – H 1  n – 1 1 -  H – H  H – H 1  n – 1
Q 3 = --------- 3 4 3 4 Q 8 = --------- 6 7 6 7
3 1n 8 1n
r3 r8
1 1 1  n – 1
--- – 1 Q 9 = ----------  H 7 – H 8  H 7 – H 8
4 1 n 1n
Q 4 = ----------  EL 5 – H 4  EL 5 – H 4 9 r9
1n
r4

1 1  n – 1 1
Q 5 = ----------  H 2 – H 6  H 2 – H 6 --- – 1
5 1n 10 1 -H – H  H – H n
r5 Q 10 = --------- 8 9 8 9
1n
r 10

The modulus sign avoids raising a negative exponent when Hi < Hk. For the Darcy – Weisbach head loss
equation where n = 2, Equation 9.32 becomes

 1  – 0.5
Q j =  ----------  H i – H k  H i – H k (9.33)
 r 1  2
j

For the Hazen – Williams head loss equation where n = 1.852, Equation 9.32 becomes

 1  – 0.46
Q j =  -----------  H i – H k  H i – H k (9.34)
 r 0.54
j

The reformulated pipe head loss equations expressed in terms of flows from Table 9.15 are shown in
Table 9.16.

240
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

9.8.2 The head or H – equations for the ten pipe network system
Each term in Equation 9.34 may be substituted into the continuity equations in Table 9.8 to form 7 non-
linear equations in 7 unknown heads H2, H3, H4, H6,....,H9 as shown in Table 9.17 and Table 9.18.

Table 9.17 The head or H – equations for the ten pipe network
Node Continuity equation at each node in terms of unknown heads
no.
1  n – 1 1  n – 1 1  n – 1
 E L1 – H 2  E L1 – H 2  H 2 – H 3 H 2 – H 3  H 2 – H 6 H 2 – H 6
2 -------------------------------------------------------------------------- + -------------------------------------------------------------------
- + -------------------------------------------------------------------
- + DM 2 =
1n 1n 1n
r1 r2 r5

1  n – 1 1  n – 1 1  n – 1
 H 2 – H 3 H 2 – H 3  H 3 – H 4 H 3 – H 4  H 3 – H 7 H 3 – H 7
3 -------------------------------------------------------------------
- + -------------------------------------------------------------------
- + -------------------------------------------------------------------
- + DM 3 =
1n 1n 1n
r2 r3 r6

1  n – 1 1  n – 1 1  n – 1
 H 3 – H 4 H 3 – H 4  EL 5 – H 4  EL 5 – H 4  H 4 – H 8 H 4 – H 8
4 -------------------------------------------------------------------- – ------------------------------------------------------------------------
1n 1n
- + -------------------------------------------------------------------
1n
- + DM 4 = 0
r3 r4 r7

1  n – 1 1  n – 1
 H 2 – H 6 H 2 – H 6  H 6 – H 7 H 6 – H 7
6 – -------------------------------------------------------------------
- + -------------------------------------------------------------------
- + DM 6 = 0
1n 1n
r5 r8

1  n – 1 1  n – 1 1  n – 1
 H 3 – H 7 H 3 – H 7  H 6 – H 7 H 6 – H 7  H 7 – H 8 H 7 – H 8
7 -------------------------------------------------------------------
- + -------------------------------------------------------------------
- + -------------------------------------------------------------------
- + DM 7 =
1n 1n 1n
r6 r8 r9

1  n – 1 1  n – 1 1  n – 1
 H 4 – H 8 H 4 – H 8  H 7 – H 8 H 7 – H 8  H 8 – H 9 H 8 – H 9
8 -------------------------------------------------------------------
- – -------------------------------------------------------------------
- + -------------------------------------------------------------------
- + DM 8 =
1n 1n 1n
r7 r9 r 10

1  n – 1
 H 8 – H 9 H 8 – H 9
9 – -------------------------------------------------------------------
- + DM 9 = 0
1n
r 10

9.8.3 The H – equations formulation for Darcy – Weisbach head loss equa-
tion
For the Darcy – Weisbach head loss equation with n = 2 then Table 9.17 is rewritten as Table 9.18. The
resistance term r – values  r 1 r 2  r 10  in Table 9.17 would be computed from the formulas in
Section 4.8 for pipe j for either SI units or US customary units. These values would be updated at every
iteration.
Table 9.18 The head or H – equations for the ten pipe network for the Darcy – Weisbach head loss
equation
Node Continuity equation at each node in terms of unknown heads
no.
– 0.5 – 0.5 – 0.5
 E L1 – H 2  E L1 – H 2  H 2 – H 3 H 2 – H 3  H 2 – H 6 H 2 – H 6
2 ---------------------------------------------------------------
- + ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 2 =
0.5 0.5 0.5
r1 r2 r5

– 0.5 – 0.5 – 0.5

 H 2 – H 3 H 2 – H 3  H 3 – H 4 H 3 – H 4  H 3 – H 7 H 3 – H 7
3 ---------------------------------------------------------
- + ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 3 =
0.5 0.5 0.5
r2 r3 r6

– 0.5 – 0.5 – 0.5

 H 3 – H 4 H 3 – H 4  EL 5 – H 4  EL 5 – H 4  H 4 – H 8 H 4 – H 8
4 ---------------------------------------------------------- – --------------------------------------------------------------
- + ---------------------------------------------------------
- + DM 4 = 0
0.5 0.5 0.5
r3 r4 r7

– 0.5 – 0.5
 H 2 – H 6 H 2 – H 6  H 6 – H 7 H 6 – H 7
6 – ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 6 = 0
0.5 0.5
r5 r8

241
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Table 9.18 The head or H – equations for the ten pipe network for the Darcy – Weisbach head loss
equation
– 0.5 – 0.5 – 0.5
 H 3 – H 7 H 3 – H 7  H 6 – H 7 H 6 – H 7  H 7 – H 8 H 7 – H 8
7 ---------------------------------------------------------
- + ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 7 =
0.5 0.5 0.5
r6 r8 r9

– 0.5 – 0.5 – 0.5

 H 4 – H 8 H 4 – H 8  H 7 – H 8 H 7 – H 8  H 8 – H 9 H 8 – H 9
8 ---------------------------------------------------------
- – ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 8 =
0.5 0.5 0.5
r7 r9 r 10

– 0.5
 H 8 – H 9 H 8 – H 9
9 – ---------------------------------------------------------
- + DM 9 = 0
0.5
r 10

9.8.4 The H – equations formulation for Hazen – Williams head loss equa-
tion
For the Hazen – Williams head loss equation with n = 1.852 , then Table 9.17 is rewritten as
Table 9.19.
Table 9.19 The head or H – equations for the ten pipe network for
the Hazen – Williams head loss equation
Node Continuity equation at each node in terms of unknown heads
no.
– 0.46 – 0.46 – 0.46
 E L1 – H 2  E L1 – H 2  H 2 – H 3 H 2 – H 3  H 2 – H 6 H 2 – H 6
2 -----------------------------------------------------------------
- + ------------------------------------------------------------ + ------------------------------------------------------------ + DM 2 =
0.54 0.54 0.54
r1 r2 r5

– 0.46 – 0.46 – 0.46

 H 2 – H 3 H 2 – H 3  H 3 – H 4 H 3 – H 4  H 3 – H 7 H 3 – H 7
3 ------------------------------------------------------------ + ------------------------------------------------------------ + ------------------------------------------------------------ + DM 3 =
0.54 0.54 0.54
r2 r3 r6

– 0.46 – 0.46 – 0.46

 H 3 – H 4 H 3 – H 4  EL 5 – H 4  EL 5 – H 4  H 4 – H 8 H 4 – H 8
4 ------------------------------------------------------------ – ----------------------------------------------------------------- + ------------------------------------------------------------ + DM 4 = 0
0.54 0.54 0.54
r3 r4 r7

– 0.46 – 0.46
 H 2 – H 6 H 2 – H 6  H 6 – H 7 H 6 – H 7
6 – ------------------------------------------------------------ + ------------------------------------------------------------ + DM 6 = 0
0.54 0.54
r5 r8

– 0.46 – 0.46 – 0.46

 H 3 – H 7 H 3 – H 7  H 6 – H 7 H 6 – H 7  H 7 – H 8 H 7 – H 8
7 ------------------------------------------------------------ + ------------------------------------------------------------ + ------------------------------------------------------------ + DM 7 =
0.54 0.54 0.54
r6 r8 r9

– 0.46 – 0.46 – 0.46

 H 4 – H 8 H 4 – H 8  H 7 – H 8 H 7 – H 8  H 8 – H 9 H 8 – H 9
8 ------------------------------------------------------------ – ------------------------------------------------------------ + ------------------------------------------------------------ + DM 8 =
0.54 0.54 0.54
r7 r9 r 10

– 0.46
 H 8 – H 9 H 8 – H 9
9 – ------------------------------------------------------------ + DM 9 = 0
0.54
r 10

9.9 Loop Flows or LF – equations

9.9.1 The loop flows vector
The column vector of unknown loop flows in a network is

T
u =  LF 1 LF 2  LF NTL  for s=1,...., NTL (9.35)

where

• u = vector of loop flows for each of the closed simple loops and required independent paths
(m3/s or ft3/s)

242
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

• LF s = loop flow for each loop s (closed simple loops or required independent paths) (m3/s
or ft3/s)
• NTL = number of independent loops (simple plus path) (= NL + NF – NC); NL = number
of closed simple loops; NF = number of reservoirs or fixed head nodes; NC = number of
separate disconnected networks
The column vector of unknown loop flows that are shown in Figure 9.6 for the ten pipe network is

T
u =  LF 1 LF 2 LF 3  (9.36)

9.9.2 The loop flow equations

The loop flows or LF – equations are obtained by writing the energy equation for each closed simple
loop and each required independent path between nodes of fixed head. As each closed simple loop is tra-
versed the head loss along each link (based on a flow that incorporates the loop flow correction values)
is added (or subtracted) according to the sign convention used. An initial flow must be assumed in each
pipe as described in Section 9.9.3. If the direction of the loop flow LFs for loop s is in the same direction
as the assumed link flow, the LFs value is added to the initial flow assumed for the pipe or link. To the
contrary, if the direction of the loop flow LFs for loop s being considered is opposite to the assumed flow
init
direction in the link, the LFs value is subtracted. The initial link flow Q j is adjusted by the final con-
verged value of LFs computed for each closed simple loop s that the link is a member of.

The minimum number of equations for any of the formulations considered results for the LF – equations
formulation. The set of LF – equations has equations that are all nonlinear. In general the number of
unknowns in the loop flow equations is substantially less than the number of Q – equations or the H –
equations. As for the Q – equations formulation all loops and required independent paths need to be
identified for the LF – equations formulation.

Figure 9.6 Unknown LF – equations formulation for the ten pipe network (note demands are also
init
shown but these are assumed to be known; in addition Q j values that are initially selected to
satisfy continuity stay constant iteration by iteration)

243
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

The LF – equations in terms of the unknown loop flows in the network have in the past been recom-
mended as the preferred formulation of the equations for a network (Epp and Fowler 1970, Wood and
Rayes 1981, Nielsen 1989). Ellis and Simpson (1996) indicated that the process for determining the pre-
ferred formulation is not straight forward. It not only depends on the number of equations but also the
initial set up time prior to the equations being iteratively solved for (e.g. in determining a set of link
flows that satisfy continuity) and also is quite dependent on the starting vector for the set of unknowns
that is selected to commence the iterative solution process.

The advantages of the LF – equations formulation include (Nielsen 1989)

• assured convergence compared to the H – equations formulation

• the reduced number of unknowns to be solved compared to the H – equations formulation
• the quick convergence when linear theory method (LTM) is used for the first iteration step
and the Newton – Raphson method (NR) is used for successive iterations (Nielsen 1989)
The main disadvantage of this technique is

• the need for computing a basis for the complete solution to the continuity equations at every
init
node to obtain an initial starting vector q
The Global Gradient Algorithm formulation is preferred to the loop flows method as will be seen later in
Chapter.

9.9.3 The initial flows must satisfy continuity

To commence the iterative solution process, a set of initial flows for the network must be assumed that
init
satisfy continuity at every node. The column vector of initial flows q is

init init init init T

q =  Q 1  Q 2   Q NP  (9.37)

The continuity equations do not form part of the set of the LF – equations as long as the initial flow in
init
each link Q j satisfies the continuity at each node. Nielsen (1989) suggested a way of an initial col-
init init
umn vector of unknown flows or q that satisfies continuity. The Q j flows for the ten pipe exam-
ple network are shown in Figure 9.6.

9.9.4 Detailed development of the loop flows equations

The LF – equations are developed by firstly assuming that

init
• Qj is the assumed initial flow in each link j that satisfies continuity at each node. These
remain the same for all iterations. Note that this is quite different from the loop flows for-
mulation for the Hardy Cross method discussed in Section 8.4 where the initially selected
k  k + 1
flows Q j are updated to Q j in the loop immediately after the computation of each
loop flow.
*
• Q j is the correct flow (i.e. the flow that is required to be found) in link j such that the LF –
equations or head loss energy equations for each closed simple loop and required inde-
pendent path s are satisfied
• LFs is the loop flow for a closed simple loop s or required independent path s that must be
init *
added (or subtracted) to the Q j values to obtain the actual flow rates Q j in each link
assuming the values have converged

244
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

* init
Qj = Qj +  LF s for each link j = 1,...., NP (9.38)
s  PL j

where

*
• Q j = actual true flows in the network (m3/s or ft3/s)
init
• Qj = assumed initial flows in each link j (m3/s or ft3/s)
• LF s = loop flow for each loop s (closed simple loops or required independent paths) (m3/s
or ft3/s)
• PLj = set of indices of the loops associated with pipe j
• NP = number of links
Each of the LF terms is only included within the summation if link j is a member in the closed simple
loop or required independent path s.

Consider the head loss for pipe j in a closed simple loop or required independent path s by combining
Equation 9.19 and Equation 9.38 as

* n–1 n–
= r j  Q j +  LF s  Q j +  LF s
* init init
= r jQ j Q j (9.39)
f j  s  PL j  s  PL j

For the Darcy – Weisbach head loss equation with n = 2 then Equation 9.39 becomes

= r j Q j Q j = r j  Q j +  LF s  Q j +  LF s
* * init init
(9.40)
f j  s  PL j  s  PL j

For the Hazen – Williams head loss equation with n = 1.852 then Equation 9.39 becomes

* 0.852 0.85
= r j  Q j +  LF s  Q j +  LF s
* init init
= r jQ j Q j (9.41)
f j  s  PL j  s  PL j

Variables are defined beneath Equation 9.39. Flow directions must be assumed for each pipe. A direc-
tion (usually positive for clockwise) for each of the loop flow LFs must also be assumed. Now for the
head loss around a closed simple loop

n–1
 r j  Q j + s P LLF s  Q j + s P LLF s
init init
= 0 for s = 1,2,...,NL (9.42)
L j
j j

It follows that for the head loss along a required independent path between two reservoirs or nodes of
fixed head is

n–1
 r j  Q j + s P LLF s  Q j + s P LLF s
init init
–  E L m – E L q  = 0for s=1,2,...,(NF – NC)(9.43)
L j
j j

9.9.5 The loop flow equations for the ten pipe network
For the ten pipe network in Figure 9.6, Equation 9.42 and Equation 9.43 for the two closed simple loops
and one path are shown in Table 9.20.

245
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Table 9.20 The loop flows or LF – equations for the ten pipe network
Loop no. Energy equation
1 init init n–1
r 2  Q 2 + LF 1 – LF 3  Q 2 + LF 1 – LF 3
init init n–1
+ r 6  Q 6 + LF 1 – LF 2  Q 6 + LF 1 – LF 2
init init n–1 init init n–1
– r 8  Q 8 – LF 1  Q 8 – LF 1 – r 5  Q 5 – LF 1  Q 5 – LF 1 = 0
2 init init n–1 init init n–1
r 3  Q 3 + LF 2 – LF 3  Q 3 + LF 2 – LF 3 + r 7  Q 7 + LF 2  Q 7 + LF 2
init init n–1 init init n–1
– r 9  Q 9 – LF 2  Q 9 – LF 2 – r 6  Q6 + LF 1 – LF 2  Q 6 + LF 1 – LF 2 = 0
3 init init n–1 init init n–1
r 4  Q 4 + LF 3  Q 4 + LF 3 – r 3  Q 3 + LF 2 – LF 3  Q 3 + LF 2 – LF 3
init init n–1 init init n–1
– r 2  Q 2 + LF 1 – LF 3  Q 2 + LF 1 – LF 3 – r 1  Q 1 – LF 3  Q 1 – LF 3
–  E L5 – E L1  = 0

For the Darcy – Weisbach head loss equation with n = 2 then Table 9.20 is rewritten as Table 9.21. The
resistance term r – values  r 1 r 2  r 10  in Table 9.21 would be computed from the formulas in
Section 4.8 for pipe j for either SI units or US customary units.
Table 9.21 The loop flows or LF – equations for the ten pipe network for the Darcy – Weisbach
head loss equation
Loop no. Energy equation
1 init init
r 2  Q 2 + LF 1 – LF 3  Q 2 + LF 1 – LF 3
init init init init
+ r 6  Q 6 + LF 1 – LF 2  Q 6 + LF 1 – LF 2 – r 8  Q 8 – LF 1  Q 8 – LF 1
init init
– r 5  Q 5 – LF 1  Q 5 – LF 1 = 0
2 init init
r 3  Q 3 + LF 2 – LF 3  Q 3 + LF 2 – LF 3
init init init init
+ r 7  Q 7 + LF 2  Q 7 + LF 2 – r 9  Q 9 – LF 2  Q 9 – LF 2
init init
– r 6  Q 6 + LF 1 – LF 2  Q 6 + LF 1 – LF 2 = 0
3 init init init init
r 4  Q 4 + LF 3  Q 4 + LF 3 – r 3  Q 3 + LF 2 – LF 3  Q 3 + LF 2 – LF 3
init init init init
– r 2  Q 2 + LF 1 – LF 3  Q 2 + LF 1 – LF 3 – r 1  Q 1 – LF 3  Q 1 – LF 3
–  E L5 – E L1  = 0

246
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

For the Hazen – Williams head loss equation with n = 1.852 , then Table 9.20 is rewritten as Table 9.22
and Table 9.23.
Table 9.22 The loop flows or LF – equations for the ten pipe network for the Hazen – Williams
head loss equation
Loop Energy equation
no.
1 init init 0.852
r 2  Q 2 + LF 1 – LF 3  Q 2 + LF 1 – LF 3
init init 0.852 init init 0.852
+ r 6  Q 6 + LF 1 – LF 2  Q 6 + LF 1 – LF 2 – r 8  Q 8 – LF 1  Q 8 – LF 1
init init 0.852
– r 5  Q 5 – LF 1  Q 5 – LF 1 = 0
2 init init 0.852 init init 0.852
r 4  Q 4 + LF 3  Q 4 + LF 3 – r 3  Q 3 + LF 2 – LF 3  Q 3 + LF 2 – LF 3
init init 0.852 init init 0.852
– r 2  Q 2 + LF 1 – LF 3  Q 2 + LF 1 – LF 3 – r 1  Q 1 – LF 3  Q 1 – LF 3 =0
3 init init 0.852 init init 0.852
r 4  Q 4 + LF 3  Q 4 + LF 3 – r 3  Q 3 + LF 2 – LF 3  Q 3 + LF 2 – LF 3
init init 0.852 init init 0.852
– r 2  Q 2 + LF 1 – LF 3  Q 2 + LF 1 – LF 3 – r 1  Q 1 – LF 3  Q 1 – LF 3
–  E L5 – E L1  = 0

9.10 The Q+H Equations Formulation

For this formulation the flow and head equations are formed from both the head loss equations for each
pipe (Section 9.5) expressed in terms of nodal heads at each end of the pipe and the continuity equations
(Section 9.4) at each of the nodes. The column vector of unknown flows and heads is denoted as

T
m =  Q 1 Q 2  Q NP H 1 H 2  H NJ  (9.44)

where

• m = vector of unknown pipe flows (or other flow elements – pumps or valves) plus
unknowns heads at each node (m3/s or ft3/s).
• Q j = flow in link j (m3/s or ft3/s)
• NP = number of links
• H i = nodal head at node i (m or ft)
• NJ = number of nodes (excluding fixed head nodes such as reservoirs)
The number of equations in this formulation is clearly larger than the number of equations for the Q –
equations formulation or the H – equations formulation. Thus the size of the matrices that need to be
dealt with in the iterative solution process will consequently be larger. However, the form of the govern-
ing equations is simpler and computation of the Jacobian elements is considerably easier. In addition, it
is not necessary to determine the loops as for the Q – equations formulation.

The Q+H equations formulation for the ten pipe network shown in Figure 9.2 as seen earlier in the
Chapter.

9.10.1 The pipe head loss equations for the ten – pipe network
The head loss equation for a pipe expressed in terms of the unknown heads at the end nodes and the
unknown discharge in the pipe from Equation 9.19 in a functional form is

n–1
H i – H k – r jQ j Q j = 0 for j = 1,......,NP (9.45)

247
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

The head loss equation based on Equation 9.45 for each of the ten pipes in Figure 9.7 are given in
Table 9.23.

Table 9.23 The head loss equations for each of the ten pipes in the ten pipe network
PN* Funct. Head loss equation PN Funct. Head loss equation

1 f 1m E L – H – r Q Q n – 1 = 0 6 f 6m H – H – r Q Q n – 1 = 0
1 2 1 1 1 3 7 6 6 6

2 f 2m H – H – r Q Q n – 1 = 0 7 f 7m H – H – r Q Q n – 1 = 0
2 3 2 2 2 4 8 7 7 7

3 f 3m H – H – r Q Q n – 1 = 0 8 f 8m H – H – r Q Q n – 1 = 0
3 4 3 3 3 6 7 8 8 8

4 f 4m E L – H – r Q Q n – 1 = 0 9 f 9m H – H – r Q Q n – 1 = 0
5 4 4 4 4 7 8 9 9 9

5 f 5m H – H – r Q Q n – 1 = 0 10 f 10  m  H – H – r Q Q n – 1 = 0
2 6 5 5 5 8 9 10 10 10

*PN = pipe number

9.10.2 The continuity equations for the ten – pipe network

The continuity equations at each of the seven nodes based on Equation 9.15 (page 230) for Figure 9.7
are given in Table 9.24.
Table 9.24 The continuity equations at each of the seven nodes for the ten pipe network

Node Function Continuity equation Node Function Continuity equation

2 f 11  m  – Q 1 + Q 2 + Q 5 + DM 2 = 0 7 f 15  m  – Q 6 – Q 8 + Q 9 + DM 7 = 0

3 f 12  m  – Q 2 + Q 3 + Q 6 + DM 3 = 0 8 f 16  m  – Q 7 – Q 9 + Q 10 + DM 8 = 0

4 f 13  m  – Q 3 – Q 4 + Q 7 + DM 4 = 0 9 f 17  m  – Q 10 + DM 9 = 0

6 f 14  m  – Q 5 + Q 8 + DM 6 = 0

9.10.3 The Q+H equations for the ten – pipe network

The combination of the head loss equations (Table 9.23) and the continuity equations (Table 9.24) for
the ten pipe network gives 17 equations in 17 unknowns (10 unknown Qs and 7 unknown Hs). All the
17 equations are given in Table 9.25.

Table 9.25 The Q+H equations for the ten pipe network
Func- Pipe head loss equations and continuity equations
tion
f 1m E L1 – H 2 – r 1 Q1 Q1
n–1
= 0 f 11  m  – Q 6 – Q 8 + Q 10 + DM 7 = 0

f 2m H 2 – H 3 – r 2 Q2 Q2
n–1
= 0 f 12  m  – Q 7 – Q 9 + Q 10 + DM 8 = 0

f 3m H 3 – H 4 – r 3 Q3 Q3
n–1
= 0 f 13  m  – Q 10 + DM 9 = 0

f 4m E L5 – H 4 – r 4 Q4 Q4
n–1
= 0 f 14  m  – Q 1 + Q 2 + Q 5 + DM 2 = 0

248
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Table 9.25 The Q+H equations for the ten pipe network

f 5m H 2 – H 6 – r 5 Q5 Q5
n–1
= 0 f 15  m  – Q 2 + Q 3 + Q 6 + DM 3 = 0

f 6m H 3 – H 7 – r 6 Q6 Q6
n–1
= 0 f 16  m  – Q 3 – Q 4 + Q 7 + DM 4 = 0

f 7m H 4 – H 8 – r 7 Q7 Q7
n–1
= 0 f 17  m  – Q 5 – Q 8 + DM 6 = 0

f 8m H 6 – H 7 – r 8 Q8 Q8
n–1
= 0

f 9m H 7 – H 8 – r 9 Q9 Q9
n–1
= 0

f 10  m  H – H – r Q Q n – 1 = 0
8 9 10 10 10

9.10.4 Q+H equations formulation for Darcy – Weisbach head loss equation
For the Darcy – Weisbach head loss equation with n = 2 then Table 9.25 is rewritten as Table 9.26.
Table 9.26 The Q+H equations with 17 unknowns for the ten pipe network (Darcy – Weisbach)
Func- Continuity equations and pipe head loss equations
tion
f 1m E L1 – H 2 – r 1 Q1 Q1 = 0 f 11  m  – Q 6 – Q 8 + Q 10 + DM 7 = 0

f 2m H 2 – H 3 – r 2 Q2 Q2 = 0 f 12  m  – Q 7 – Q 9 + Q 10 + DM 8 = 0

f 3m H 3 – H 4 – r 3 Q3 Q3 = 0 f 13  m  – Q 10 + DM 9 = 0

f 4m E L5 – H 4 – r 4 Q4 Q4 = 0 f 14  m  – Q 1 + Q 2 + Q 5 + DM 2 = 0

f 5m H 2 – H 6 – r 5 Q5 Q5 = 0 f 15  m  – Q 2 + Q 3 + Q 6 + DM 3 = 0

f 6m H 3 – H 7 – r 6 Q6 Q6 = 0 f 16  m  – Q 3 – Q 4 + Q 7 + DM 4 = 0

f 7m H 4 – H 8 – r 7 Q7 Q7 = 0 f 17  m  – Q 5 – Q 8 + DM 6 = 0

f 8m H 6 – H 7 – r 8 Q8 Q8 = 0

f 9m H 7 – H 8 – r 9 Q9 Q9 = 0

f 10  m  H 8 – H 9 – r 10 Q 10 Q 10 = 0

9.10.5 The Q+H equations formulation for Hazen – Williams head loss equa-
tion
For the Hazen – Williams head loss equation with n = 1.852 , then Table 9.25 is rewritten as Table 9.27

249
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Table 9.27 The Q+H equations with 17 unknowns for the ten pipe network (Hazen – Williams)
Func- Continuity equations and pipe head loss equations
tion
Func- Pipe head loss equations and continuity equations
tion
f 1m E L1 – H 2 – r 1 Q1 Q1
0.852
= 0 f 11  m  – Q 6 – Q 8 + Q 10 + DM 7 = 0

f 2m H 2 – H 3 – r 2 Q2 Q2
0.852
= 0 f 12  m  – Q 7 – Q 9 + Q 10 + DM 8 = 0

f 3m H 3 – H 4 – r 3 Q3 Q3
0.852
= 0 f 13  m  – Q 10 + DM 9 = 0

f 4m E L5 – H 4 – r 4 Q4 Q4
0.852
= 0 f 14  m  – Q 1 + Q 2 + Q 5 + DM 2 = 0

f 5m H 2 – H 6 – r 5 Q5 Q5
0.852
= 0 f 15  m  – Q 2 + Q 3 + Q 6 + DM 3 = 0

f 6m H 3 – H 7 – r 6 Q6 Q6
0.852
= 0 f 16  m  – Q 3 – Q 4 + Q 7 + DM 4 = 0

f 7m H 4 – H 8 – r 7 Q7 Q7
0.852
= 0 f 17  m  – Q 5 – Q 8 + DM 6 = 0

f 8m H 6 – H 7 – r 8 Q8 Q8
0.852
= 0

f 9m H 7 – H 8 – r 9 Q9 Q9
0.852
= 0

f 10  m  H – H – r Q Q 0.852 = 0
8 9 10 10 10

9.11 Global Gradient Algorithm Formulation

The Global Gradient Algorithm formulation by Todini and Pilati (1988) is effectively a type of Q+H
equations formulation. The formulation is based on two sets of equations solving for both all unknown
heads and flows simultaneously in a sequential iterative nature. The authors based their derivation on the
Collins et al. (1978) Content Model. This formulation of the equations for solving the hydraulic condi-
tions in a network is considered to be one of the most computationally efficient algorithms for steady
state computation (Salgado-Castro et al. 1988). The method developed an extremely efficient approach
to the inversion of the Jacobian matrix (see Chapter 10, Section 10.2.2) by partitioning the governing
equations in a smart way. The advantage of the Global Gradient Algorithm is that the Jacobian matrices
are symmetric. The Global Gradient Algorithm formulation is used in EPANET (Rossman 2000) and a
number of other commercially available hydraulic software packages for the simulation of water distri-
bution systems.

9.11.1 Background to the Global Gradient Algorithm

Todini and Pilati (1988) provide a detailed justification of their method. Some of the material from their
paper is summarized in the following section. The alternative methods for solving for conditions in a
water distribution system were given as:

1. Local gradient (an example is the Hardy Cross approach that has already been described in
Chapter 8). These methods solve the original system of equations in terms of a local gradi-
ent (a loop flow correction for one loop in the case of the loop flows formulation of the
Hardy Cross method)
2. Newton – Raphson method (described in detail in Chapter 10 – Warga 1954, Martin and
Peters 1963, Shamir and Howard 1968). Todini and Pilati (1988) viewed this as an exten-
sion of the Hardy Cross method by means of a simultaneous multidimensional correction
algorithm.

250
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

3. Linearisation (the linear theory method of Charles and Wood (1972) described in
Chapter 10).
4. Numerical minimization based on a Content Model (Collins et al. 1978) that searches for a
minimum of a nonlinear convex objective function. Mathematical optimization techniques
are used to solve the system of equations. The advantage is that due to the convexity of the
objective function combined with the linear constraints guarantees the existence and
uniqueness of the problem. Unfortunately the solution algorithms are inefficient particu-
larly for large complex networks.
5. The Global Gradient Algorithm is described as a bridge between optimization and the
Newton – Raphson technique that proves the existence and uniqueness of the solution (the
reader is referred to the Todini and Pilati (1988) paper for details of the proof). A slightly
modified Content Model (Collins et al. 1978) is minimized leading to a system of heads and
flows that are solved for simultaneously. The Newton – Raphson technique is then applied
to the enlarged space of heads and flows where the proof of existence and uniqueness holds.
This is the key to the unconditional convergence of the Global Gradient Algorithm. A
Lagrange multiplier technique is used by Todini and Pilati (1988) to transform a con-
strained minimization problem into an unconstrained one. A proof is given of monotonicity
of one of the terms in the governing equations that leads to the uniqueness and existence of
the solution.
Finally the problem is algebraically rearranged (see Chapter 10 for details) to be
• a recursive solution of a linear system of size NJ equal to number of unknown heads
and
• a matrix projection of the results over the NP unknown link flows
A sparse symmetric, positive definite and Stieltjes type matrix that results allows for the
efficient solution using the Incomplete Choleski Factorization Method/Modified Conjugate
Gradient algorithm (ICF/MCG) of Kershaw (1978). More details are given in Chapter 10.
9.11.2 Specification of parameters and definition of required matrices
A number of quantities that have (mostly) been presented earlier in this Chapter are now considered.
The quantities considered are

• NP = number of links
• NJ = number of nodes
• NF = number of reservoirs or fixed head nodes
• NC = number of separate disconnected subnetworks (or components)
• NL = number of closed simple loops {= NP – NJ –  NF – NC  }
• NPL = number of required independent paths {=  NF – NC  }
• NTL = number of independent loops (simple loops plus required independent paths) {=
 NL + NPL  }
As an example consider the ten pipe network in Figure 9.7. This network has been studied for other for-
mulations earlier in the Chapter.

251
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Figure 9.7 Unknown Global Gradient Algorithm formulation variables for ten pipe network (note
demands are also shown but these are assumed to be known)

The continuity equation has been considered earlier in the Chapter - see Equation 9.15, Equation 9.16
and Equation 9.17

Thus the matrix formulation of the continuity equations for the Global Gradient Algorithm from
Equation 9.17 is

T
A1 q + dm = 0 (9.46)

MATRIX FORM OF THE CONTINUITY EQUATIONS

where

• A 1 = unknown head node incidence matrix

• q = vector of unknown pipe flows (m3/s or ft3/s)
• d m = vector of known demands for each of the nodes (m3/s or ft3/s)

9.11.3 The pipe head loss equations in matrix form – the Global Gradient
Algorithm
The head loss equations in each of the pipe j from Equation 9.19 are

n–1
H i – H k = r jQ j Q j for j = 1,..., NP (9.47)

Rearranging Equation 9.47 for the pipes not attached to reservoirs for the ten pipe network gives

n–1
–r j Q j Q j + Hi – Hk = 0 for j = 1,..., NP except pipes 1 and 4 (9.48)

For pipes 1 and 4 that are attached to fixed head node reservoirs the following equations apply

n–1
–r 1 Q1 Q1 – H + EL 1 = 0 for pipe 1 (9.49)
2

252
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

n–1
–r 4 Q4 Q4 – H + EL 5 = 0 for pipe 4 (9.50)
4

n–1
Consider the first term r j Q j Q j on the left – hand side of Equation 9.48. Introduce the following
matrix form for the ten pipe network in Figure 9.6 as follows

Q1
n–1 Q2
r 1 Q1 0 0  0 0 0
n–1 Q3
0 r 2 Q2 0  0 0 0
n–1 Q4
0 0 r 3 Q3  0 0 0
Q5
       (9.51)
Q6
n–1
0 0 0  r 8 Q8 0 0
Q7
n–1
0 0 0  0 r 9 Q9 0 Q8
n–1
0 0 0  0 0 r 10 Q 10 Q9
Q 10

Denote the 10 by 10 square matrix in Equation 9.51 as G.

Thus the G matrix for the most general form of the head loss equation is

n–1
r 1 Q1 0 0  0 0 0
n–1
0 r 2 Q2 0  0 0 0
n–1
0 0 r 3 Q3  0 0 0
G =        (9.52)
n–1
0 0 0  r 8 Q8 0 0
n–1
0 0 0  0 r 9 Q9 0
n–1
0 0 0  0 0 r 10 Q 10

Thus Equation 9.51 becomes

Q1
Q2
Q3
Q4
Q5
G = Gq (9.53)
Q6
Q7
Q8
Q9
Q 10

Note that the G is a positive diagonal matrix that is dependent on the flow vector q . Another way to
write G is

253
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

 n – 1
G = diag  r j Q j  for j = 1,..., NP (9.54)
 

Now returning to Equation 9.48 for the ten pipe network and considering the second and third unknown
nodal head terms – H i + H k on the left – hand side of

n–1
–r j Q j Q j + Hi – Hk = 0 for j = 1,..., NP except pipes 1 and 4 (9.55)

A matrix form can be introduced by using the unknown head incidence matrix A 1 for the ten pipe net-
work from Equation 9.11 and the vector of unknown heads h from Equation 9.4 as follows

–1 0 0 0 0 0 0
1 –1 0 0 0 0 0 H2
0 1 –1 0 0 0 0 H3
0 0 –1 0 0 0 0 H4
1 0 0 –1 0 0 0
H 6 = A1 h (9.56)
0 1 0 0 –1 0 0
H7
0 0 1 0 0 –1 0
0 0 0 1 –1 0 0 H8
0 0 0 0 1 –1 0 H9
0 0 0 0 0 1 –1

Now consider a pipe that is attached to a reservoir (from Equation 9.49). The third term of – EL 1 on the
left – hand side in

n–1
–r 1 Q1 Q1 – H + EL 1 = 0 for pipe 1 (9.57)
2

may be put in matrix form as follows. Introduce a general column vector of known reservoir elevations
for a total of NF reservoirs

T
e L =  EL 1 EL 2  EL NF  (9.58)

For the ten pipe network the vector of reservoir elevations is

T EL 1
e L =  EL 1 EL 5  = (9.59)
EL 5

The matrix form of the third term in Equation 9.57 using the fixed head node incidence matrix A 2
defined in Section 9.3.2 is

254
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

1 0
0 0
0 0
0 1
0 0 EL 1 = A e (9.60)
2 L
0 0 EL 5
0 0
0 0
0 0
0 0

where

• EL 1m = water surface elevation of reservoir node m

• A 2 = fixed node incidence matrix (Section 9.3.2)
• e L = fixed nodal head or reservoir elevations matrix (Equation 9.59) (m or ft)

Thus the resultant matrix form for the head loss equations for the pipes from Equation 9.47
n–
H
( i – H k = h f = r jQ j Q j ) in a network is
j

– Gq + A 1 h + A 2 e L = 0 (9.61)

MATRIX FORM OF THE PIPE HEAD LOSS EQUATIONS

where

n–1
• G = positive diagonal square matrix of r j Q j values (see Equation 9.52)
• n = exponent of the flow in the head loss equation (Darcy – Weisbach n = 2 or Hazen –
Williams equation n = 1.852)
• q = column vector of unknown pipe flows (Equation 9.1 and Equation 9.2) (m3/s or ft3/s)
• A 1 = unknown head node incidence matrix (Section 9.3.1 and Equation 9.11)
• h = column vector of unknown heads for each of the nodes (m or ft) (Equation 9.4)
• A 2 = fixed node incidence matrix (Section 9.3.2 and Equation 9.13)
• e L = reservoir elevations vector (Equation 9.59) (m or ft)

Together Equation 9.46 and Equation 9.61 form the Global Gradient Algorithm equation formulation
that is the basis of the Global Gradient Algorithm for solving for conditions in a network.

T
A1 q + dm = 0 (9.62)

– Gq + A 1 h + A 2 e L = 0 (9.63)

THE GLOBAL GRADIENT ALGORITHM FORMULATION

Expanding Equation 9.61 into full matrix form gives Equation 9.64.

255
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Q1
n–1 Q2
r 1 Q1 0 0  0 0 0
n–1 Q3
0 r 2 Q2 0  0 0 0
n–1 Q4
0 0 r 3 Q3  0 0 0
Q5
–        (9.64)
Q6
n–1
0 0 0  r 8 Q8 0 0
Q7
n–1
0 0 0  0 r 9 Q9 0 Q8
n–1
0 0 0  0 0 r 10 Q 10 Q9
Q 10

–1 0 0 0 0 0 0 1 0
1 –1 0 0 0 0 0 H2 0 0
0 1 –1 0 0 0 0 H3 0 0
0 0 –1 0 0 0 0 H4 0 1
+ 1 0 0 –1 0 0 0 + 0 0 EL 1 = 0
H6
0 1 0 0 –1 0 0 0 0 EL 5
H7
0 0 1 0 0 –1 0 0 0
0 0 0 1 –1 0 0 H8 0 0
0 0 0 0 1 –1 0 H9 0 0
0 0 0 0 0 1 –1 0 0

A combined partitioned matrix formulation of these two sets of equations (Equation 9.62 and
Equation 9.63) gives

–G | A1 A2 e L
q
––– | ––– –– + ––– = 0 (9.65)
T h dm
A1 | 0

The derivation above differs from that of Todini and Pilati (1988) but arrives at the same result in
Equation 9.65. Todini and Pilati (1988) derived Equation 9.62 and Equation 9.63 based on a Content
Model approach based on the transformation of a constrained minimization into an unconstrained one
by means of Lagrange multipliers. Todini and Pilati (1988) concluded that because the r j values for
each pipe j are positive or the head loss expressions for each pipe f j  Q j  are all monotonically increas-
ing functions then the problem is convex and therefore a solution exists and is unique. However, this is
true only in the space of unknown Q and H (Todini and Pilati 1988). An alternative derivation of these
equations was provided by Ellis (2003) who used a total variation of parameters for the development of
the iterative solution procedure.

The numerical solution of Equation 9.65 using a Newton Raphson approach is given in Chapter 16.

In contrast to the Q – equations formulation and the H – equations formulation, the Global Gradient
Algorithm solves for the heads and flows simultaneously in a sequential iterative manner. The parti-
tioned form of the matrix in Equation 9.65 based on the continuity equations (Equation 9.16 and
Equation 9.46) and the pipe head loss equations is used to solve for both flows and heads.

9.12 Alternative Formulations of the Network Equations

Design information (such as the best combination of pipe diameters and pump sizes) is not yielded
directly when an analysis approach is used (Ormsbee and Wood 1986). For solving the analysis prob-
lem, physical data for the water distribution system must be available including the network layout, con-
nectivity of elements in network, pipe sizes and internal roughness characteristics, lengths, tank

256
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

locations and normal water surface operating levels, pump locations and characteristic operating curves,
pressure reducing valves and pressure settings, constants and external demands (Templeman and Yates
1984). Analysis involves the determination of the internal performance of the network in terms of flows
and pressures at nodes. Total inflow to the system must be equal to the total outflow.

Boulos and Wood (1990) divided parameters associated with a network into three categories

• design parameters such as pipe diameter, pipe length, pump power, pump head, storage
level and valve characteristics
• operating parameters such as pump speed, pressure regulating valve setting, control valve
setting, and flow or pressure specification
• calibration parameters such as pipe roughness and nodal demands
In general, one of the following occurs at all nodes in the network (Brebbia and Ferrante 1983)

• the head is specified (e.g. a reservoir)

• the demand or consumption is specified
• neither the head nor the demand is specified
Both the demand and head cannot be specified simultaneously.

As an alternative to the analysis problem described above, it is also possible to recast the equations to
take on a synthesis or design formulation. For example, the pipe network governing equations may be
formulated in terms of unknown pipe diameters, if flow velocities are specified in each line. Other vari-
ables may also be recast as unknowns including heads, water demands and pipe roughnesses or resis-
tances (Shamir and Howard 1968). Although not explicitly discussed by Shamir and Howard (1968) it is
possible that pipe diameters, pipe length or pipe roughness could be determined from values of pipe
resistance (Ormsbee and Wood 1986; Boulos and Wood 1990; Boulos and Ormsbee 1991). Ormsbee
and Wood in 1986 stated that they thought that direct solution of design parameters had great potential
and provided an alternative to implicit optimization procedures to water distribution systems. They sug-
gested that an optimum hydraulic design may be found in the sense that the designs are carried out to
just meet specified conditions. Whilst it is possible to recast the equations as suggested above, pre –
defining velocities in particular pipes is problematic as it may be that the particular velocity that is
selected is not optimal. Whilst having some use it is the opinion of the author that this method is not
nearly effective as using optimization of the design of networks such as genetic algorithm optimization.
Details of genetic algorithm analysis are shown in Chapter 20.

9.13 References
Bhave, P. R. (1988). “Extended period simulation of water systems - direct solution.” Journal of Envi-
ronmental Engineering, 114(5), October, 1146-1159.

Boulos, P. F., and Altman, T. (1991). “A graph-theoretic approach to explicit nonlinear pipe network
optimization.” Applied Mathematical Modelling, Vol. 15(September), 459-466.

Boulos, P. F., and Ormsbee, L. E. (1991). “Explicit network calibration for multiple loading conditions.”
Civ. Engrg. Syst., 8, 153-160.

Boulos, P. F., and Wood, D. J. (1990). “Explicit calculation of pipe-network parameters.” J. Hydr.
Engrg., 116(11), 1329-1344.

Brebbia, C. A. and A. J. Ferrante (1983). “Computational Hydraulics, Pipe Networks, Chapter 4." Com-
putational hydraulics: 59-121.

Collins, M., Cooper, L., Helgason, R., Kennington, J., and LeBlanc, L. (1978). “Solving the pipe net-
work analysis problem using optimization techniques.” Management Science, 24(7), 747-760.

Cross, H. (1936). Analysis of flow in networks of conduits or conductors. Bulletin No. 286, University of
Illinois Engineering Experimental Station, Urbana, Illinois.

Epp, R. and A. G. Fowler (1970). “Efficient code for steady-state flows in networks.” Journal of the
Hydraulics Division, Proceedings of the ASCE, January, Vol. 96, No. HY1: 43-56.

257
 Chapter 9 - File: Chap9NwOld10-v02.fm July 6, 2022 Version 6

Jeppson, R. W. (1976). Analysis of flow in pipe networks, Ann Arbor Science, Ann Arbor, Michigan.

Kershaw, D. S. (1978). “The incomplete Cholesky - conjugate gradient method for the iterative solution
of systems of linear equations.” Journal of Computational Physics, 26, 43-65.

Kesavan, H. K., and Chandrashekar, M. (1972). “Graph-theoretical models for pipe network analysis.”
Journal of the Hydraulics Division, 98(2), 345-364.

Martin, D. W., and Peters, G. (1963). “The application of Newton's method to network analysis by digi-
tal computer.” Journal of Institute of Water Engineers and Scientists, 17(2), 115-129.

Ormsbee, L. E., and Wood, D. J. (1986). “Hydraulic design algorithms for pipe networks.” Journal of
Hydraulic Engineering, 112(12), 1195-1207.

Rossman, L. A. (2000). EPANET 2 users manual. U.S. Environmental Protection Agency, Washington,
D.C., EPA/600/R-00/057.

Salgado-Castro, R. O. (1988). Computer Modelling of Water Supply Distribution Networks Using Gra-
dient Method, Volume 1 and 2., PhD Thesis, University of Newcastle-Upon-Tyne, Newcastle, UK.

Shamir, U. and D. D. Howard (1968). “Water distribution systems analysis.” Journal of the Hydraulics
Division, Proceedings of the ASCE Vol. 94, No. HY1, January, 219-235.

Templeman, A. B., and Yates, D. F. (1984). “Mathematical similarities in engineering network analy-
sis.” Civil Engineering Systems, Vol. 1(March), 114-122.

Todini, E., Pilati, S. (1988). “A gradient algorithm for the analysis of pipe networks.” Proceedings,
Computer Applications in Water Supply, Research Studies Press, Letchworth, Hertfordshire, UK, 1-20.

Warga, J. (1954). “Determination of the steady state flows and currents in a network.” Proceedings of
Instrument Society of America, Vol. 9, Pt. 5, Paper No. 54-43-4.

Wood, D. J., and Charles, C. O. A. (1972). “Hydraulic network analysis using linear theory.” Journal of
the Hydraulics Division, Vol. 98, No. HY7, 1157-1170.

Wood, D. J., and Rayes, A. G. (1981). “Reliability of algorithms for pipe network analysis.” Journal of
the Hydraulic Division. 107(HY10), 1145-1161.

258
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

CHAPTER 10 NUMERICAL TECHNIQUES FOR ANALYSIS

OF WATER DISTRIBUTION SYSTEMS1

10.1 Introduction
As seen in Chapter 9, a set of simultaneous nonlinear equations may be formed to represent flows and
heads in a network. Five representations were outlined in Chapter 9 including

• the unknown flows

• the unknown heads
• the unknown loop flows
• the unknown flows and heads
• the unknown flows and heads for the Global Gradient Algorithm
Flows and nodal pressures in a network may be computed if the pipe lengths, diameters, pipe rough-
nesses, and all external flows leaving and entering the network are known (although reservoir inflows do
not need to be known). No direct solution technique exists for these various sets of nonlinear equations.
The nonlinear formulations that were presented in Chapter 9 include

• the flow equations or Q – equations formulation where the continuity, closed simple loop
energy and required independent path energy equations are expressed in terms of unknown
flows in the pipes (or other flow elements) in the network. There are NP unknown flows for
this formulation.
• the head equations or H – equations formulation where the equations are formulated in
terms of unknown heads. There are NJ unknown heads in the network.
• the loop flow equations or LF – equations formulation where the equations are formulated
in terms of unknown loop flows. An initial estimate of all the flows in the network such that
continuity is satisfied at every node is required. There are NL +  NF – NC  unknown loop
flows for this formulation.
• the flow and head equations or Q – H equations formulation where both the continuity and
pipe head loss equations are expressed in terms of unknown flows in the pipes (or other
flow elements) and unknown heads at nodes. There are a total of NP + NJ unknown flows
and heads.
• the Global Gradient Algorithm. The formulation begins in a similar way to the Q – H equa-
tions formulation but a smart decomposition is used in developing the solution procedure.
Thus both the Q and H variables are dealt with simultaneously but the size of the matrix
that requires inversion in the iterative process is the size of the number of unknown heads
(NJ) rather than the size of the number of links plus the number of nodes (NP+NJ).
Each formulation for the example network in Equation 9.1 has a different matrix size that needs to be
computed are shown in Table 10.1. These matrices differ in their sparseness and their properties in terms
of symmetry, diagonal dominance and whether the matrix is positive definite and of Stieltjes type. This
will also have an influence on the ease of solution.

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.

259
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

Table 10.1 Nature of matrices required for each formulation

Method Matrix size Matrix size Type of matrix
for network in
Figure 9.2
Q – equations NP  NP 10  10 Non –
symmetric
H – equations NJ  NJ 77 Symmetric
LF – equations* NLT  NLT 33 Symmetric
Q – H equations  NJ + NP    NJ + NP  17  17 Symmetric
Global Gradient Algorithm NJ  NJ 77 Symmetric,
positive definite
* where NLT = total number of loops and required independent paths in the network = NL + (NF – NC) (see
Section 9.2.2 for definition of variables)

Usually the solution techniques of these sets of nonlinear equations are complicated and require iterative
solutions. Linearization of the nonlinear equations is the most common approach (Templeman and Yates
1984). A comprehensive comparison of the various solution techniques has been presented by Wood and
Rayes (1981), Salgado-Castro (1988) and Ellis and Simpson (1996). Often a Taylor series approxima-
tion, truncated to eliminate the nonlinear terms, forms the basis of iterative linearization procedures. The
changes or corrections in the unknown variables are then solved for, using techniques for solving sets of
linear equations. A new set of approximations to the unknown variables are found by adding the incre-
ments onto the values of the unknown variables at the previous iteration. Iterations continue such that
the solution is improved until a specified convergence criterion is met. The most common computer
solution techniques include (Heath 2001)

• the Newton – Raphson (NR) or Newton method (Warga 1954, Martin and Peters 1963)
• the linear theory method (LTM) (Wood and Charles 1972), this is actually fixed point itera-
tion
• the Global Gradient Algorithm (Todini and Pilati 1988), this is actually a form of Newton’s
method but it formulates the equations in a special way
Techniques for solving the set of equations include

• successive relaxation techniques such as the Hardy Cross (1936) method at the University
of Illinois at Urbana – Champaign, Illinois, USA. Elementary circuits or loops are balanced
in turn until all conditions for flow are satisfied.
• iterative methods including (a) successive substitution including both the Gauss – Seidel
and Jacobi methods (b) the Newton method where a nonlinear problem is turned into a lin-
ear problem. There are various variations of the Newton method including the damped
Newton method and where the Jacobian matrix is not recomputed at every iteration. New-
ton's method can also be used with either line search or trust region step control (Wolfram
website 2008).
This Chapter describes in general form the numerical techniques used to solve these sets of governing
equations. Chapter 12 to Chapter 16 then apply the numerical techniques to solve various formulations
of the water distribution system equations. The Hardy Cross method (Cross 1936, Cornish 1939), as dis-
cussed in Chapter 8, was the method of choice when hand computations were common prior to the mid –
1960s. Loop flow corrections are solved and applied to initially assumed flow values for each pipe and
are successively made loop by loop and iteration by iteration. These loop flow corrections are based on
a first order Taylor series expansion of the energy equations. As shown in Chapter 8, the method is suit-
able for hand computation but converges very slowly because each loop flow correction adjustment is
computed independently from every other loop.

260
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

The Newton method linearizes the nonlinear terms in the governing equations and solves a set of simul-
taneous linear equations to solve iteratively for corrections to the unknowns being solved for in the net-
work. Unknown variables are adjusted simultaneously thus achieving a more rapid convergence rate
than the Hardy Cross method.

Expressions for the Jacobian matrix in the Newton method for the ten pipe network (already introduced
in Chapter 9) are derived in Chapter 12 to Chapter 16 for the five formulations of the equations
described above. The choice of the equation formulation has a direct bearing on the behavior of the iter-
ative technique to solve the set of nonlinear equations (Nielsen 1989). Nielsen (1989) drew some con-
clusions from his investigations (prior to general knowledge about the Todini and Pilati 1988 paper)

• it is better to formulate the flow equations in terms of pipe flows (or other flow elements)
than in terms of heads at nodes
• the Newton iterative procedure is the preferred solution method
• the linear theory method results in oscillations when the iteration gets close to the solution,
however, it may provide a good starting point for the Newton method
The linear theory method (LTM) was a method that gained popularity in the early 1970s and was
thought at that time to be the method of choice. It involves linearization of the energy equations and an
iterative solution process of both the continuity equations and the linearized energy equations—a total
of NP equations. Starting solutions for the initial values are not required and they are usually assumed to
be one unit of flow to simplify the computations. This is an outcome of the way the linearization process
is implemented. Convergence problems have been experienced with the linear theory method technique
especially for large networks. The method is really a successive substitution solution method. It has
been consequently shown to be less effective than the Newton method for solving the set of nonlinear
equations.

The Global Gradient Algorithm formulation of the equations solved with the Newton technique appears
to be the most effective solution method. Matrices are pre – manipulated to maximize the efficiency of
the numerical solution. Application of the Global Gradient Algorithm to the solution of flows and heads
for the ten pipe network is given in Chapter 16.

10.2 Newton Technique for a Set of Nonlinear Equations

10.2.1 A linear set of equations
Although this section deals with the solution of non-linear of equations ultimately a set of linear equa-
tions will need to be solved iteratively. So first consider the linear system (based on Baldick 2006, Chap-
ter 5, page 186)

Ax = b (10.1)

where

• A = coefficient matrix
• x = vector of unknowns to be solved for
• b = right hand side vector
To solve the system in Equation 10.1 the inverse of the coefficient matrix needs to be found such that

–1
x = A b (10.2)

For small problems Cramer’s rule can be used to invert the coefficient matrix but for computational
efficiency Gaussian elimination is preferred (Baldick 2006). For large systems, Gaussian elimination
takes much less computational effort than Cramer’s rule. The computational effort to perform Gaussian
elimination is bounded by a cubic polynomial (Baldick 2006). LU factorization involves factorizing
matrix A into the product of a lower triangular matrix L and an upper triangular matrix U. The method
is similar to Gaussian elimination (Baldick 2006). Because symmetry often arises in formulations of the

261
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

governing equations for water distribution systems a variant of LU factorization for symmetric systems
may be used. In addition the matrix A is often sparse. If this is the case, then the calculations in the LU
factorization can be arranged to reduce the computational effort considerably. The techniques discussed
so far are called direct solution techniques. Very large and sparse systems of equations are more effi-
ciently solved using iterative solution techniques (Baldick 2006).

10.2.2 A general form of nonlinear equations

The sets of nonlinear equations presented in Chapter 9 require an iterative solution procedure to locate
the roots or solution values of a set of simultaneous equations. In this section the general form of the
Newton iterative procedure for solving a set of nonlinear equations is described (Holistic Numerical
Methods - NEWTON-RAPHSON METHOD CHAPTER 03.04 Website 2019). This iterative method is
described as having fast quadratic convergence when the iterate column vector x(k) on the kth iteration is
close to the solution row vector x* (see Equation 10.5). The Newton technique has been applied to net-
works by a number of researchers (Warga 1954, Martin and Peters 1963, Shamir and Howard 1968, Epp
and Fowler 1970, Todini and Pilati 1988, Nielsen 1989, Boulos and Wood 1990).

A general set of unknowns is considered where the set of n unknown variables is the column vector x or

T
x = x 1 x 2  x n (10.3)

Consider the set of n functions as

f 1 x 1 x 2  x n  = 0

f 2  x 1 x 2  x n  = 0

f n  x 1 x 2  x n  = 0 (10.4)

where

• x = vector of unknowns
• f j  x 1 x 2  x n  = jth non-linear function
• x i = unknown variable to be solved for

An iterative solution procedure is required to determine the values of xi in Equation 10.4 for n unknown
variables of x 1 x 2  x n . The true solution to this set of equations is assumed to be the following col-
umn vector

T
x * = x *1 x *2  x *n (10.5)

Let a general deviation or correction column vector for the n unknown variables be defined as

T
x = x 1 x 2  x n (10.6)

The following will hold when x is small

262
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

x * = x + x (10.7)

In the neighborhood of the current iteration point each of the functions in Equation 10.4 may be
expanded in a Taylor series (Press et al. 1996) as

n  f x
1 2
f 1 x + x  = f 1  x  + -x i + O  x 
 ----------------
x i
i=1

n  f x
2 2
f 2 x + x  = f 2  x  + -x i + O  x 
 ----------------
x i
i=1

n  f x
n 2
f n x + x  = f n  x  + -x i + O  x 
 ---------------- (10.8)
x i
i=1

where

• f i  x , f j x + x  = ith function
• x = deviation vector of the unknown variables; x i = ith deviation value in the vector
• n = number of functions
2
• O  x  = second order and higher terms of the flow and head corrections
Putting Equation 10.8 into matrix notation (Press et al. 1996) gives

2
f  x + x  = f  x  + J  x x + O  x  (10.9)

where

• J  x  = the Jacobian matrix (size n x n) as defined in Equation 10.12

2 2
By dropping second order terms O  x  and higher order terms and assuming that  x  is small in
Equation 10.9 and by setting

*
f  x + x  = f  x  = 0

when x is small in Equation 10.9 gives

k  k + 1 k 
Jx x = –f  x  (10.10)

GENERAL FORM OF THE NEWTON METHOD SOLUTION

or

263
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

k  k  k 
 f 1x   f 1x   f 1x  k + 1
----------------------- ------------------------  -----------------------
- x 1 k  k  k 
x1 x2 xn f 1  x 1  x 2   x n
k + 1
k  k   k  x 2 k  k  k 
 f 2x   f 2x   f 2x  f 2  x 1  x 2   x n
----------------------- ------------------------  -----------------------
- = – (10.11)
x1 x2 xn  
     
k  k  k  k  k  k 
 f nx   f nx   f nx  k + 1 f n  x1  x2   x n
- x n
----------------------- ------------------------  -----------------------
x1 x2 xn

where the superscripts k and (k + 1) refer to successive iteration numbers. Equation 10.10 will be solved
k + 1
for the unknown vector x to avoid computation of the inverse of the Jacobian (as presented
below in Equation 10.13).

The Jacobian matrix for a general system of equations in Equation 10.11 is a square n by n matrix and
based on the variable values at iteration k is defined as

k  k  k 
 f 1x   f 1x   f 1x 
------------------------ ------------------------  -----------------------
-
x1 x2 xn
k  k  k 
 f 2x   f 2x   f 2x 
k  - ------------------------  -----------------------
x  = -----------------------
x1 x2 xn
- (10.12)

   
k  k  k 
 f nx   f nx   f nx 
------------------------ ------------------------  -----------------------
-
x1 x2 xn

Equation 10.10 can also be rearranged to give an expression for a column vector of unknown corrections
at iteration (k + 1) as

k + 1 –1 k  k 
x = –J  x f  x  (10.13)

where

–1 k 
• J  x  is the inverse of the Jacobian matrix (the Jacobian matrix is defined by
Equation 10.12).
An apparent weakness in the Newton method for solving systems of nonlinear equations is the necessity
to invert the Jacobian matrix (Burden and Faires 1989). In practice, however, explicit inversion of the
Jacobian matrix may be avoided as mentioned previously. Generally, the Jacobian matrix has very few
non – zero entries (usually 2 to 5%) for large networks. Sparse matrix techniques may be used to
decrease the computation time for the solution of large networks (Chandrashekar and Stewart 1974).

10.2.3 Testing for convergence

During the iterative process a stopping test check needs to be made to see if the solution has converged
to a sufficiently accurate solution. It is useful to first consider in computer programs, IEEE standard
double precision arithmetic with precision, measured by machine epsilon (Forsythe & Moler 1967), of

– 16
 mach  2 10 (10.14)

This value will be used later on as part of a stopping test for convergence.

264
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

There are three main norms that are defined – the one – norm, the two – norm and the infinity – norm. For
k + 1
the correction vector x of length n at the (k+1)st iteration during the Newton iterative solution
process, the p – norm (as defined by the MATLAB function reference webpage) is as follows.
th

Definition of the pth-norm

--1-
k + 1  n  k + 1  p
p
x p =   x j  (10.15)
j=1 

Definition of the 1-norm

Thus the 1 – norm for p = 1 from Equation 10.15 is

n
k + 1 k + 1
x 1 =  x j (10.16)
j=1

Definition of the 2-norm

The 2 – norm (p = 2) is

1
---
 n 2  2
k + 1 k + 1
x 2 =   x j  (10.17)
 j=1 

k + 1
This is the Euclidean length of a vector x . The root – mean – square (RMS) value of an n – element
k + 1
vector x at the (k+1)st iteration may be defined (from the MATLAB function reference) as

1---
 n k + 1 2 
2
  x j 
 j=1 
RMS = -------------------------------------------------- (10.18)
n

or is the average squared – deviation.

Definition of the infinity-norm

The infinity – norm (p =  ) is the absolute value of the maximum sized deviation of all changes in the
k + 1
vector x at the (k+1)st iteration

k + 1 max  x  k + 1  
x  = j (10.19)
j=1, n

These norms can be applied to the flow change vector, the head change vector or the loop flow correc-
tion vector depending on what formulation has been selected.

The computer program EPANET for simulating water distribution systems uses the following relative
flow convergence test parameter (Rossman 2000) of

265
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

NP
k + 1
 Q j
k + 1
q j=1
EPANET convergence = ---------------------------1 = ----------------------------------------- (10.20)
k + 1 NP
q 1 k + 1
 Qj
j=1

Head change norms

The test in Equation 10.20 considers only the flows but from a practical standpoint it might be consid-
ered an appropriate alternative to cease iterating when, from one iteration to the next, the heads at each
node in a network differ by a sufficiently small margin (Elhay and Simpson 2011). The infinity – norm
applied to the head changes at all nodes in the network may be used as the check for convergence in all
the formulations presented in Chapter 12 to Chapter 16 as this seems to be the most physically based. In
other words, the engineer can say that they wish to have the maximum deviation to be say, no greater
than 1 mm (0.001 m). It is more difficult to attach a physical meaning to other convergence measures
such as the relative flow measure in Equation 10.20.

The infinity – norm for the head changes for the (k+1)st iteration may be computed as:

k + 1 max k + 1
h  =  H i  (10.21)
i=1, NJ

where

k + 1
• h  = infinity – norm for the head changes at the (k+1)st iteration (m or ft)
k + 1
• h = vector of nodal head changes at the (k+1)st iteration (m or ft)
k + 1
• H i = head change at node i at the (k+1)st iteration (m or ft)
• NJ = number of nodes of unknown head in the network
Thus the stopping test such that the iterative solution process terminates when the following is true

k + 1
h    stop (10.22)

PREFERRED STOPPING TEST BASED ON THE INFINITY-NORM

where

•  stop = stopping test tolerance (m or ft - for example: may be set equal to say 0.001 m or
1 mm).
Equation 10.22 is applicable to three of the five formulations for checking the convergence for stopping
the iterative solution process. These formulations include (i) the H-equations (Chapter 12) (ii) the Q+H
equations (Chapter 14) and (iii) the Global Gradient Algorithm equations.

In contrast, Equation 10.22 is not suitable for the other two formulations of (i) the Q-equations (Chapter
11) and (ii) the loop flow equations (Chapter 13). The heads at the nodes are not computed in either of
these formulations so an alternative approach must be taken. Revised convergence stopping tests will be
introduced for these two formulations in the appropriate Chapters.

266
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

An additional residuals test for solution acceptability

Modelers using an iterative procedure (for the Q+H equation formulation and the Global Gradient Algo-
rithm equations) are sometimes tempted, where the iteration process has signaled a failure to converge
after a certain number of steps based on Equation 10.20 or Equation 10.22, to relax the stopping test by
–5 –2
increasing the stopping tolerance (e.g. from the value of  stop  10 to, say,  stop  10 ). This is a
dangerous practice and so an additional test based on the residuals is introduced below (Elhay and Simp-
son 2011). The residuals of the computed solution for Equation 9.62 and Equation 9.63 should also be
checked, once the iterative solution process has been stopped by Equation 10.20 or Equation 10.22,
since residuals that are larger than expected will indicate an inaccurate solution. More precisely, for any
set of flows and heads q and h the following are defined, respectively, the pipe head loss residual and
the continuity residual by

 h  q h  = – Gq + A 1 h + A 2 e L (10.23)

T
 c  q h  = A 1 q + d m (10.24)

These residuals will be zero at the exact solution. A computed solution which satisfies Equation 10.20 or
Equation 10.22 should be considered unacceptable if for either of the following infinity norms of the
residuals (with norms defined by Equation 10.19)

 h  q h  
 u 1  stop (10.25)

or

 c  q h  
 u 2  mach (10.26)

where  mach is defined in Equation 10.14 and where u1 = 1 and u2 = 100 have been found by Elhay and
Simpson (2011) to be suitable choices for networks with up to 10,000 pipes. For larger networks these
may need to be increased in size. Rejecting any solution for which one or both of Equation 10.25 and
Equation 10.26 holds safeguards against accepting an inaccurate solution.

Flow change norms

An alternative to the infinity – norm for the head changes is the infinity – norm for the flow changes at the
(k+1)st iteration expressed as a fractional change (FC) as follows

k + 1
max  Q j 
FC  =  -----------------------
- (10.27)
j=1, NP  Q k + 1  
j

The difficulty with the infinity – norm for the flow changes is that it is difficult to assess a physically
appropriate stopping criteria. Obviously if there are values in the true solution that give zero flows then
this convergence parameter will cause problems due to the divide by zero or a number close to zero. The
following adjustment can be made to overcome this

k + 1
max  Q j 
FCA  =  -------------------------------
 k + 1
-

(10.28)
j=1, NP  1.0 + Q j 

Now the convergence criterion will not be affected by near zero flows.

267
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

10.2.4 Assessing the solution accuracy based on the Condition Number

The solvability of the set of linear equations of Equation 10.10 is determined by whether the Jacobian
k 
matrix J  x  is singular or non – singular (Heath 2001). If J is singular then there may be either no
solution or an infinite number of solutions. An n by n matrix is singular if:

• J has no inverse
• det( J ) = 0 (the determinant is zero)
• rank( J ) < n (the rank of a matrix is the maximum number of linearly independent rows or
columns it contains)
• the condition number is cond(J) =  (defined in Equation 10.29 below)
If the matrix is nonsingular there will be a unique solution. The condition number is a measure of how
close the matrix is to being nonsingular (Heath 2001). A matrix with a large condition number is nearly
singular (in contrast, the size the determinant is not a good indicator of nearness to singularity – in other
words, the determinant can be large but the matrix may not be near to nonsingular). If the condition
number is large the system of equations is then considered to be ill conditioned. Thus a small change in
the coefficient matrix or a small change in the right – hand side results in a large change in the solution
vector (Holistic Numerical Methods Institute website, 2008).

In order to define the condition number, the norm of a matrix needs to be first defined. The norm of a
matrix is a scalar that gives some measure of the magnitude of the elements of the matrix (Mathworks
Documentation- norm-Vector and matrix norms website - 2019). Two definitions of the norm of a matrix
are the 1 – norm and the infinity – norm (Heath 2001). The 1 – norm is the column sum norm and is the
largest column sum of a matrix A of size n by m defined as:

m
max
A 1 =  a ij (10.29)
1in j=1

The infinity – norm is the row sum norm and is the largest row sum of a matrix A of size n by m defined
as:

n
max
A  =  a ij (10.30)
1im j=1

The norm in Equation 10.30 finds the sum of the absolute of the elements of each row of the matrix A. It
is usual to be dealing with square matrices such that m = n. The condition number of matrix A is defined
as the product of the norm of the matrix and the norm of the inverse of the matrix:

–1
cond(A) = A  A  (10.31)

Because the condition number involves the inverse of the matrix, its computation is a nontrivial task
(Heath 2001). Computing the matrix norm A  is easy while computing the norm of the inverse
–1
A  is the challenge. Heath (2001) presents a method of predicting a bound for the norm of the
inverse matrix so that the estimate of the condition number can be achieved as an inexpensive byproduct
of the solution process.

For a matrix A the condition number is always cond(A)  1 (Heath 2001). By convention
cond(A) =  if matrix A is singular.

For any diagonal matrix D = diag(di), the condition number is the ratio of the maximum diagonal ele-
ment to the minimum diagonal element as (Heath 2001):

268
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

max d
cond(D) = -----------------i- (10.32)
min d i

The usefulness of the condition number is in assessing the accuracy of the solution (Heath 2001). The
size of the condition number gives the number of dependable significant figures in the solution obtained.
The larger the condition number the smaller the number of reliable significant figures. The computed
solution loses about log10(cond(A)) decimal digits of accuracy relative to the accuracy of input (Heath
2001). This may be expressed as the rule of thumb that for every multiple of 10 of the condition number
one significant figure of accuracy is lost1.

An example of the calculation of the norms of both the Jacobian matrix and the inverse of the Jacobian
as well as the condition number is given in Example 12.4 (on page 292).

10.3 References
Baldick, R. (2006). Applied Optimization - Formulation and Algorithms for Engineering Systems. Cam-
bridge University Press, 786pp.

Burden, R. L., and Faires, J. D. (1989). Numerical analysis, PWS-Kent.

Chandrashekar, M., and Stewart, K. H. (1975). “Sparsity oriented analysis of large pipe networks.”
Journal of the Hydraulics Division, Proceedings, Vol. 101, No. HY4, 341-355.

Cross, H. (1936). Analysis of flow in networks of conduits or conductors. Bulletin No. 286, University of
Illinois Engineering Experimental Station, Urbana, Illinois.

Elhay, S. and Simpson, A.R. (2011). “Dealing with zero flows in solving the non-linear equations for
water distribution systems.” Journal of Hydraulic Engineering, Vol. 137, No. 10, Oct., 1216-1224.

Ellis, D. E., and Simpson, A. R. (1996). Convergence of iterative solvers for the simulation of water dis-
tribution pipe networks. Report R138, Department of Civil and Environmental Engineering, The Uni-
versity of Adelaide, 50.

Epp, R. and A. G. Fowler (1970). “Efficient code for steady-state flows in networks.” Journal of the
Hydraulics Division, Proceedings of the ASCE, January, Vol. 96, No. HY1: 43-56.

Forsythe, G. E., and Moler, C. B. (1967). Computer solution of linear algebraic systems.

Heath, M. T. (2001). Scientific Computing. An Introductory Survey, McGraw – Hill, 2nd edition.

Holistic Numerical Methods Website - NEWTON-RAPHSON METHOD CHAPTER 03.04 http://

nm.mathforcollege.com/topics/newton_raphson.html Accessed 12 September 2019

Martin, D. W., and Peters, G. (1963). “The application of Newton's method to network analysis by digi-
tal computer.” Journal of Institute of Water Engineers and Scientists, 17(2), 115-129.

Mathworks Documentation- norm-Vector and matrix norms website - https://au.mathworks.com/help/

matlab/ref/norm.html Accessed 12 September 2019

Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1996) FORTRAN numerical recipes
(Vol. 1), Cambridge England.

Rossman, L A. (2000). EPANET 2 Users Manual. U.S. Environmental Protection Agency, Washington,
D.C., EPA/600/R-00/057.

Salgado-Castro, R. O. (1988). Computer Modelling of Water Supply Distribution Networks Using Gra-
dient Method, Volume 1 and 2., PhD Thesis, University of Newcastle-Upon-Tyne, Newcastle, UK.

1. From Dr. Sylvan Elhay of the School of Computer Science, University of Adelaide, Australia

269
 Chapter 10 - File: Chap10new4-a.fm July 6, 2022 Version 7

Shamir, U. and D. D. Howard (1968). “Water Distribution Systems Analysis.” Journal of the Hydraulics
Division, Proceedings of the ASCE Vol.94, No.HY1, January: 219-235.

Templeman, A. B., and Yates, D. F. (1984). “Mathematical similarities in engineering network analy-
sis.” Civil Engineering Systems, Vol. 1(March), 114-122.

Todini, E., Pilati, S. (1988). “A gradient algorithm for the analysis of pipe networks.” Proc., Computer
Applications in Water Supply, Research Studies Press, Letchworth, Hertfordshire, UK, 1-20.

Warga, J. (1954). “Determination of the steady state flows and currents in a network.” Proceedings of
Instrument Society of America, Vol. 9(Pt. 5, Paper No. 54-43-4).

Wolfram Research website https://reference.wolfram.com/language/tutorial/UnconstrainedOptimiza-

tionNewtonsMethodRoot.html#456855863 accessed 12 September 2019,

Wood, D. J., and Charles, C. O. A. (1972). “Hydraulic network analysis using linear theory.” Journal of
the Hydraulics Division, Vol. 98, No. HY7, 1157-1170.

Wood, D. J., and Rayes, A. G. (1981). “Reliability of algorithms for pipe network analysis.” Journal of
the Hydraulic Division. 107(HY10), 1145-1161.

270
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

CHAPTER 11 COMPUTER ANALYSIS FOR HYDRAULIC

SIMULATION OF WATER DISTRIBUTION
SYSTEMS1

In this Chapter the computer analysis using simulation software of EPANET (Rossman 2000) for the
modeling of water distribution systems is considered in detail. In Chapter 9 and Chapter 10, both the
formulation and solution of the network equations have been presented. Computer analysis is a very
important tool for analyzing networks. This Chapter considers the simulation of networks that are com-
prised of storage tanks and pipes only. Modeling of pumps and pressure regulating valves are consider
later in the book. Details of solving various formulations of the pipe network equations are also given in
later Chapters.

EPANET defines both links in the network as components that have an upstream and downstream node
(pipes, pumps and valves) and nodes including junctions, reservoirs and tanks. In this Chapter, the sim-
ulation model of EPANET 2.0 (Rossman 2000) is used to demonstrate computer simulation of networks.
The first simulation example presented in this Chapter involves the detailed discussion of the modeling
of the flow in a single pipeline connecting two reservoirs. The second example of the application of the
EPANET 2.0 model includes a minor loss valve in the single pipeline. Thirdly a network with 5 – pipes
and two loops is modeled. This network was solved in Chapter 8, using the Hardy Cross (1936) method.
Finally the ten pipe network introduced in Chapter 9 is also modeled with EPANET.

Many other computer programs for hydraulic simulation of networks are available that performs similar
tasks as the EPANET 2.0 examples shown in this book. In EPANET 2.0, as with other available pro-
grams, visualization of the network is possible through a map. Flows, pressures, hydraulic grade line
elevations, the variations in levels of tanks over, for example, a 24 – hour period may all be easily visu-
alized (in both tabular and graphical form) making interpretation of the results much easier.

11.1 The Capability of Simulation Models such as EPANET 2.0

Many different computer simulation models for studying the behavior of networks are available. These
include the commercially available programs, some examples of which are given as follows:

6. Synergi Water (DNV GL Group, Germany; originally Stoner and Associates),

7. Mike Urban-Water Distribution System (DHI, Denmark),
8. WaterCAD and WaterGEMS (Bentley Systems, USA),
9. InfoWater (Innovyze, USA),
10. WATSYS (HCP Software Pty Ltd, Australia), KYPIPE-Pipe 2016 (USA),
11. Sir 3S (3s Consult GmbH, Germany)
12. Along with many others.
EPANET 2.0 was written and released in the early 1990s by Dr. Lewis Rossman of the US Environmen-
tal Protection Agency in Cincinnati, Ohio. EPANET 2.0 is a freely available computer program (from
website https://www.epa.gov/water-research/epanet - accessed 12 Sep 2019). The source code is also
available. Dr. Rossman retired in 2016 and the work of maintaining EPANET and future versions has
been taken over by a group from the EWRI Water Distribution Systems Analysis Committee of the
American Society of Civil Engineering through a website called EPANET: EPANET Open Source -
Open Water Analytics (http://wateranalytics.org/EPANET/ - accessed 12 Sep 2019).

1. © Copyright 2020. Prof. Angus R. Simpson, The University of Adelaide, Adelaide, Aus-
tralia. Part of the book entitled Water Distribution Systems Engineering.

271
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

The performance of EPANET has proven to be excellent, so much so in fact, that many simulation
model software vendors have adopted EPANET as the solution engine for their software.

EPANET 2.0 performs (as do most of the commercially and freely available software programs) the fol-
lowing functions for the analysis of pressurized networks (Rossman 2000)

• steady state analysis for a water demand loading case at a particular instant in time
• extended period simulation of the hydraulic behaviour over a period of time
• water quality behaviour within a network
EPANET 2.0 has an integrated environment for graphically editing network input data, running hydrau-
lic and water quality simulations and for viewing results in a number of different formats. Rossman
(2000) provided a very comprehensive manual for EPANET 2.0.

The focus of the discussion of EPANET 2.0 in this Chapter and later Chapters are on the hydraulic
behavior of water distribution systems. Consideration of water quality modeling aspects is beyond the
scope of this book. EPANET 2.0 enables tracking of chemical species throughout the network to deter-
mine their movement and fate and as well as determination of water age and source tracing (Rossman
2000).

The elements that may be modeled by computer hydraulic simulation models include: pipes, junctions,
pumps, valves and storage tanks or reservoirs. The outcome of running a simulation model for a water
demand loading cases includes: flows and velocities in pipes, flows through pumps and valves in the
system as well as the pressures and hydraulic grade line elevations at junctions. The format for viewing
the results includes a map of the network with numeric values and color coding, as well as, tables and
graphs of results. The visualization of the results using the map of the network in EPANET 2.0 is very
useful. It is possible to see both the numerical values of quantities computed from the hydraulic analysis
as well as colors of both nodes and pipes to indicate ranges of values. Contour plots are also possible.
Most of these features are also available in other computer simulation hydraulic modeling software.

11.2 Setting the Environment of EPANET 2.0 (version 2.00.12)

The window for EPANET 2.0 is shown in Figure 11.1.

Figure 11.1 The EPANET 2.0 window

Once in the EPANET 2.0 integrated environment a number of defaults may be chosen for output of the
results.

272
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

11.2.1 Choosing default units and pipe head loss equation

The following selection may be made to determine output units of the flows or discharges in the system.
Under the EPANET 2.0 menu, select Project >> Defaults >> Hydraulics Tab. Click on the flow units
and then select from the pull down menu on the right – hand side from the following values in
Table 11.1.

Table 11.1 Choices of flow units in EPANET 2.0

Units Type Designator Units
SI units LPS Liters/s
SI units LPM Liters/min
SI units MLD Megaliters/day
SI units CMH cubic meters/hr
SI units CMD cubic meters/day
US customary units CFS cubic feet/s
US customary units GPM gallon/min
US customary units MGD million gal/day
US customary units IMGD Imperial MGD
US customary units AFD acre – feet/day

For the EPANET 2.0 example presented in the next section the units of LPS were selected. Demands at
junctions must be specified in the units as selected from Table 11.1. The Darcy – Weisbach equation (D –
W) was also selected.

11.2.2 Choosing default properties

Select from the EPANET 2.0 menu, Project >> Defaults >> Properties Tab. Fill in various default val-
ues including, for example, pipe lengths, diameters and roughness coefficients.

11.2.3 Choosing the number of decimal places for output

If Darcy – Weisbach head loss equation is used then select 4 decimal places for the friction factor or
resistance factor. This is accomplished by selecting from the EPANET 2.0 menu File >> Preferences
>> Formats Tab.

11.2.4 Specifying map options

The labels and values for nodes and links may be viewed on the map. Select the Map Options icon from
the menu bar (the last colored icon, click on Notation and check the appropriate boxes).

11.3 Modeling Flows in Individual Pipes

To define a pipe in a network for analysis using a computer simulation model two end points of the pipe
need to be firstly defined. The end points may be either of three items: a junction, a reservoir or a storage
tank. Creating a junction node in a network is discussed in Section 11.3.2. Thus to input a pipe within
EPANET 2.0 it is necessary to define the ending points of the pipe first and then create the pipe linking
the end points. To define a pipe the following information is needed

• a pipe number, identifier or name

• length of pipe in meters or ft
• internal diameter of the pipe in mm or inches
• a measure of the pipe roughness of the interior of the pipe as described in Section 11.3.1

273
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

11.3.1 Friction head loss formula

In summary, the alternative values that may be input to the computer simulation model for computation
of head loss include

• average pipe roughness height,  , for computation of the Darcy – Weisbach friction factor
or resistance factor
• Hazen – Williams coefficient, C (typically 80 for very rough, 145 for very smooth)
• Manning coefficient, n
In Chapter 4, three types of head loss formula were discussed including:

Darcy – Weisbach formula (Equation 4.3, page 77).

EPANET 2.0 uses the Swamee and Jain equation (Equation 4.61).

Hazen – Williams formula (Equation 4.100, page 109 – SI units or Equation 4.94, page 107 – US Cus-
tomary units)

There are two forms of the Hazen – Williams equation depending on the system of units, either or US
customary units. Both are shown below.

In EPANET 2.0 all computations are carried out US customary units with the input parameters in SI
units being converted to US customary units. Once the analysis in US customary units is complete, the
results in US customary units are converted to SI units.

The Manning equation (Equation 4.136, page 129 – SI units or Equation 4.138 – US Customary units)

11.3.2 EPANET input for junctions in water distribution systems

Junctions are nodes or points where links (usually pipes) join together. In addition, water enters or
leaves the network at junctions.

The basic input data required for junctions are

• junction number, designator or name

• elevation
• water demand
Junctions are input onto a map in EPANET 2.0 by clicking on the button on the toolbar shown in
Table 11.2.

Table 11.2 Various icons on the toolbar in EPANET 2.0

EPANET 2.0 toolbar but- Purpose
ton
To insert a junction on the
map at a node
To insert a tank on the map
at a node
To insert a reservoir on the
map at a node
To insert a pipe on the map
between two nodes

274
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

A click on the appropriate position on the map results in a junction being inserted on the map. Upon
double clicking on the junction on the map, an input data box appears requesting the input of data asso-
ciated with the junction. In EPANET two reservoirs and one center junction may be inserted on the map
and then two pipes are inserted to join the reservoirs with the central junction. Data is entered for each
reservoir, tank, junction and pipe using the pop – up menus that occur when the element is clicked upon
on the map.

To demonstrate the modeling of flow in a pipe between two reservoirs, Example 6.3 is revisited.

Example 11.1 Compute the flow for a reservoir – pipe – reservoir system as shown in Figure 11.2 for a smooth pipe
with a roughness height of 0.003 mm (new uPVC pipe). Other pertinent quantities are a diameter of D = 300 mm, a
length of L = 500 m, an upstream reservoir elevation of z1 = 50 m, a downstream reservoir elevation of z2 = 45 m,
–6
and a kinematic viscosity of water of v = 1.007 10 m2/s.

Figure 11.2 Reservoir – smooth pipe – reservoir system for Example 11.1

Answer – Computation of the flow using EPANET for flow between two reservoirs – smooth pipe turbulent
flow.

The solution to this pipe flow problem is detailed in Section 11.3.8. First, all of the necessary data definitions for the
EPANET input data file are developed in the next sections. To illustrate the input data required the input for EPANET
Version 1 is used (.INP files) and is shown below.

___________________________________________________________________________________

To demonstrate the EPANET input format, the style of format of input data as presented in EPANET
Version 1 is used here. The text style format as shown below is based on that developed by Rossman
(2000) for the original EPANET Version 1 that required the network to be described in a text input data
file. In EPANET 2.0 the data input may be made graphically via the map so it is not necessary to specify
the input data format as shown below, however, the EPANET Version 2.0 format is presented here and in
later Chapters in this book to illustrate the data needed by the simulation model. An example of the input
data for the junction to be modeled for Example 11.1 in Figure 11.2 in EPANET text input format is
shown below. An EPANET model requires at least one junction or node. In Figure 11.2, the pipe is
divided into two equal – length sections with junction 3 at the connection point for the two pipes. Defini-
tion of the junction is as follows
[JUNCTIONS]
;--------------------
; Elev Demand
; ID m L/s
;--------------------
3 10 0

The keyword [JUNCTIONS] is needed to commence the description of the nodal data in the water dis-
tribution network according to the data requirements of EPANET Version 2.0. The junction number or
identifier is the number 3. Note that this junction has been created in the EPANET 2.0 input data file for
Example 11.1 as it requires at least one junction. The junction at 3 is assumed to have an elevation of
10 m and the demand at the junction is set to 0 L/s.

275
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

11.3.3 Storage reservoirs

In EPANET 2.0 storage reservoirs are defined as supply points where the water level is specified and
assumed to remain constant.

The primary input value for a reservoir is the water surface elevation. A storage reservoir is a boundary
point to a network, and thus its head remains a constant and cannot be affected by what happens within
the network (Rossman 2000).

The input data for the storage reservoirs in Example 11.1 to be modeled in EPANET 2.0 is shown as fol-
lows (in EPANET 1 format)
[RESERVOIRS]
;----------
; Elev
; ID m
;----------
1 50
2 45

The keyword [RESERVOIRS] is needed to commence the description of the storage reservoirs. The
storage reservoirs are identified as the numbers 1 and 2 representing the reservoirs in Figure 11.2. The
water surface of the reservoir at node 1 is at elevation 50.0 m while the water surface of the reservoir at
node 2 is at 45.0 m.

11.3.4 EPANET input data for pipes — a smooth pipe example

The input data for the pipe in Example 11.1 to be modeled in EPANET is shown as follows
[PIPES]
;--------------------------------------------------
; Start End Length Diam Roughness
;ID Node Node m mm Height mm
; (epsilon)
;--------------------------------------------------
1 1 3 250 300 0.003
2 3 2 250 300 0.003

As mentioned previously, two pipes are used to model the single pipe in Example 11.1 as EPANET
requires at least one junction. A junction numbered 3 has been added in the center of the pipeline. The
keyword [PIPES] is needed in EPANET 1 text style input format to commence the description of the
pipes in the water distribution network. The numbers or identifiers (ID labels) of the pipes are the num-
bers 1 and 2 representing the pipe in Figure 11.2. In Version 1 of EPANET, pipe designators had to be
numeric while in EPANET 2.0 designators have been extended to be able to be alphanumeric.

11.3.5 A descriptive title for the network to be analyzed by EPANET

Each EPANET computer simulation run should be given a descriptive title. A keyword of [TITLE] is
required when using text data input. For the Example 11.1 the following description was chosen
[TITLE]
Flow in a single pipeline - Example 11.1 (Example 6.3 revisited)
Smooth pipe turbulent flow

11.3.6 Options for the EPANET run

There are a multitude of possible options that may be specified for an EPANET run. All options have
been provided with default values (Rossman 2000). Some important options include the system of units
to be used, the viscosity of water, the head loss formula to be used and are shown below. In addition, the
map file name may be specified (only needed for EPANET Version 2.0 style of input).
[OPTIONS]
;----------------------------------------------
MAP Ch11Ex1.map; Map coordinates file

276
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

UNITS SI
HEADLOSS D-W
VISCOSITY 1.007E-6

The options section in the data file must commence with the keyword [OPTIONS]. For the keyword
MAP, the map file that contains the coordinates of the junctions, reservoirs and tanks is contained in the
file Ch11Ex1.map (EPANET Version 1 input data format). An example of the map file is shown in
Section 11.3.8. In EPANET 2.0 data the map information may alternatively be placed at the bottom of
the input data file. For the keyword UNITS, the default units are US customary units. In this case, the
keyword UNITS is set to SI and therefore selects SI units. For the keyword HEADLOSS, the default
head loss equation is Hazen – Williams (H – W) and in this case the Darcy – Weisbach equation has been
selected (D – W). Finally, the keyword VISCOSITY allows the user to specify a value of the kinematic
–6
viscosity. For Example 11.1, the kinematic viscosity of water is v = 1.07 10 m2/s.

11.3.7 The end of an EPANET input data file

To complete the input data file (EPANET 1 format) the following keyword is needed.
[END]

11.3.8 Computer solution of Example 11.1 using EPANET 2.0

The various elements of the EPANET input data for Example 11.1 have been combined into a complete
input data file (in EPANET 1 text input data format) as shown below.
Answer (to Example 11.1 as outlined in Section 11.3.4) — The full EPANET 1 data input file in text input format
based on combining the data input information outlined in the above Sections is as follows. The name of the file has
been designated as Ch11Ex1.inp — it is always useful to use the extension of.inp to designate it as an EPANET input
data file.

[TITLE]
Flow in a single pipeline - Example 11.1 (Example 6.3 revisited)
Smooth pipe turbulent flow

[JUNCTIONS]
;--------------------
; Elev Demand
; ID m L/s
;--------------------
3 10 0

[RESERVOIRS]
;----------
; Elev
; ID m
;----------
1 50
2 45

[PIPES]
;--------------------------------------------------
; Start End Length Diam Roughness
;ID Node Node m mm Height mm
; (epsilon)
;--------------------------------------------------
1 1 3 250 300 0.003
1 3 2 250 300 0.003

[REPORT]
FILE Ch11Ex1.out

[OPTIONS]
;----------------------------------------------
MAP Ch11Ex1.map ; Map coordinates file
UNITS SI
HEADLOSS D-W

277
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

VISCOSITY 1.007E-6

[END]

The EPANET co-ordinates file (only for EPANET 1) for the Example 11.1 is the file Ch11Ex1.map as follows

[COORDINATES]
;Nodal coordinates
;----------------------
;ID x-coord. y-coord.
;----------------------

1 0 50
2 500 45
3 250 47.5

[LABELS]
0 60 "Upstream reservoir" 1
450 35 "Downstream reservoir" 2

An alternative to using the map file (for input into EPANET 2.0) is to incorporate the [COORDINATES]
and [LABELS] data in the EPANET input data file directly prior to the [END] statement. The input for
the keyword [COORDINATES] is self – explanatory as shown above. The [LABELS] keyword is used
to specify x and y – coordinate positions for a label name on the map.

11.3.9 EPANET 2.0 output results for Example 11.1

The output results computed from EPANET 2.0 for Example 11.1 for the flow in the 300 mm diameter
pipe connecting two reservoirs are shown in Table 11.3. The results from EPANET 2.0 are very similar
to the results from the explicit formula as computed in Example 6.3.

Table 11.3 EPANET 2.0 tabular results for pipes in Example 11.1
Link Length Diameter Roughness Flow Velocity Friction factor
height or resistance
(m) (mm) (mm) (LPS) (m/s) factor
Pipe 1 250 300 0.003 151.84 2.15 0.0127
Pipe 2 250 300 0.003 151.84 2.15 0.0127

In Example 11.1 the analytic solution for the computation of the flow based on the Colebrook – White
formula for friction factor or resistance factor is 0.152 m3/s (or 152 L/s) that is very close to the value in
Table 11.3.

The output results computed from EPANET 2.0 for junctions and reservoirs are shown in Table 11.4.

Table 11.4 EPANET 2.0 tabular results for junctions in Example 11.1
Node Elevation Total Pressure
(m) head (m)
(m)
Junction 3 10 47.50 37.50
Reservoir 1 50 50.00 0.00
Reservoir 2 45 45.00 0.00

278
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

11.4 Modeling a 6 – Pipe Water Distribution System

11.4.1 The network and input data file
EPANET 2.0 is now used to model a more complicated network with two loops. A similar example has
previously been solved in Example 8.2 using the Hardy Cross method as outlined in Chapter 8. How-
ever, the demands here to demonstrate EPANET have been multiplied by a factor of 10 to increase the
head losses. The new network with the increased demands is shown in Figure 11.3 below. All other
characteristics of the network are the same.

Example 11.2 Solve for the flows and pressures using the computer simulation model EPANET 2.0 for the network
shown in Figure 11.3

Figure 11.3 Layout of a looped example water distribution system

The water distribution system configuration data is given in Table 11.6.

Table 11.5 Node data for 6 – pipe water distribution system (factored up by 10 compared with Chapter 8)

Node Elevation Demand

(L/s)

1 100 0
2 60 200
3 60 500
4 60 300
5 60 0

Table 11.6 Data for 6 – pipe water distribution system

Pipe Start Node End Node L D 

(m) (mm) (mm)
[1] 5 2 20 353 0.25
[2] 5 4 50 440 0.25
[3] 2 4 54 574 0.25
[4] 2 3 50 619 0.25
[5] 4 3 20 449 0.25
[6] 1 5 40 581 0.25

Answer – Computation of the flows and pressures using the computer simulation model EPANET 2.0

The EPANET 2.0 input data file called Ch11Ex2.inp (saved in the .inp format) for the water distribution system
described in Figure 11.3

[TITLE]
Flow in a 6-pipe network - Example 11.2 (Example 8.2 revisited)

279
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

[JUNCTIONS]
;ID Elev Demand (L/s)
;------------------------------
2 60 200
3 60 500
4 60 300
5 60 0

[RESERVOIRS]
;ID Head
;------------------
1 100

[PIPES]
;ID Node1 Node2 Length Diam Roughness
; m mm Height mm
;--------------------------------------------------------------
1 5 2 20 353 0.25
2 5 4 50 440 0.25
3 2 4 54 495 0.25
4 2 3 20 607 0.25
5 4 3 50 440 0.25
6 1 5 40 581 0.25

[OPTIONS]
;-------------------------
Units LPS
Headloss D-W
Viscosity 1.007E-6

[COORDINATES]
;Node X-Coord Y-Coord
;------------------------------------
1 0 30
2 20 20
3 70 20
4 70 0
5 20 0

[END]

11.4.2 Results for 6 – pipe water distribution system

Try this model above yourself to confirm the results below.
Page 1
**********************************************************************
* E P A N E T *
* Hydraulic and Water Quality *
* Analysis for Pipe Networks *
* Version 2.0 *
**********************************************************************

Input File: Ch11Ex2.inp

Flow in a 6-pipe network - Example 11.2 (Example 8.2 revisited)

Link - Node Table:

----------------------------------------------------------------------
Link Start End Length Diameter
ID Node Node m mm
----------------------------------------------------------------------
1 5 2 20 353
2 5 4 50 440
3 2 4 54 495
4 2 3 20 607
5 4 3 50 440
6 1 5 40 581

280
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

Node Results:
----------------------------------------------------------------------
Node Demand Head Pressure Quality
ID LPS m m
----------------------------------------------------------------------
2 200.00 97.93 97.93 0.00
3 500.00 97.89 97.89 0.00
4 300.00 97.96 97.96 0.00
5 0.00 99.18 99.18 0.00
1 -1000.00 100.00 0.00 0.00 Reservoir

Link Results:
----------------------------------------------------------------------
Link Flow Velocity Unit Headloss Status
ID LPS m/s m/km
----------------------------------------------------------------------
1 474.00 4.84 62.26 Open
2 526.00 3.46 24.35 Open
3 -99.20 0.52 0.51 Open
4 373.20 1.29 2.35 Open
5 126.80 0.83 1.49 Open
6 1000.00 3.77 20.57 Open

11.5 EPANET Modeling of a 10 – Pipe Water Distribution System

The pipe and node properties for the ten pipe water distribution system in Figure 9.2 are given in
Table 11.7 and Table 11.8.
Table 11.7 Pipe properties for ten pipe network
Pipe Up – Down – D L 
stream stream (mm) (m) (mm)
node node
1 1 2 356 4800 0.25
2 2 3 254 1600 0.25
3 3 4 254 1600 0.25
4 5 4 356 6400 0.25
5 2 6 254 1600 0.25
6 3 7 254 1600 0.25
7 4 8 254 1600 0.25
8 6 7 203 1600 0.25
9 7 8 203 1600 0.25
10 8 9 203 1600 0.25

Table 11.8 Node properties for ten pipe network

Node ID Elevation (m) Demand (L/s)
1 365.0 Reservoir
2 320.0 17
3 319.0 17
4 322.0 17
5 372.0 Reservoir
6 310.0 50
7 310.0 50

281
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

Table 11.8 Node properties for ten pipe network

8 308.0 50
9 304.0 32

The input data for this ten pipe network to be modeled in EPANET is shown as follows
[TITLE]

[JUNCTIONS]
;ID Elev Demand Pattern
2 320 17 ;
3 319 17 ;
4 322 17 ;
6 310 50 ;
7 310 50 ;
8 308 50 ;
9 304 32 ;

[RESERVOIRS]
;ID Head Pattern
1 365 ;
5 372 ;

[TANKS]
;ID etc

[PIPES]
;ID N1 N2 Length Dia Roughness
; MinorLoss Status
1 1 2 4800 356 0.25 0 Open ;
2 2 3 1600 254 0.25 0 Open ;
3 3 4 1600 254 0.25 0 Open ;
4 5 4 6400 356 0.25 0 Open ;
5 2 6 1600 254 0.25 0 Open ;
6 3 7 1600 254 0.25 0 Open ;
7 4 8 1600 254 0.25 0 Open ;
8 6 7 1600 203 0.25 0 Open ;
9 7 8 1600 203 0.25 0 Open ;
10 8 9 1600 203 0.25 0 Open ;

[PUMPS]
;ID Node1 etc

[VALVES]
;ID Node1 etc

[TAGS]

[DEMANDS]
;Junction etc

[STATUS]
;ID Status/Setting

[PATTERNS]
;ID Multipliers

[CURVES]
;ID X-Value Y-Value

[CONTROLS]

[RULES]

[ENERGY]
Global Efficiency 75

282
Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

Global Price 0
Demand Charge 0

[EMITTERS]
;Junction Coefficient

[QUALITY]
;Node InitQual

[SOURCES]
;Node Type Quality Pattern

[REACTIONS]
;Type Pipe/Tank Coefficient

[REACTIONS]
Order Bulk 1
Order Tank 1
Order Wall 1
Global Bulk 0
Global Wall 0
Limiting Potential 0
Roughness Correlation 0

[MIXING]
;Tank Model

[TIMES]
Duration 0:00
Hydraulic Timestep 1:00
Quality Timestep 0:05
Pattern Timestep 1:00
Pattern Start 0:00
Report Timestep 1:00
Report Start 0:00
Start ClockTime 12 am
Statistic NONE

[REPORT]
Status No
Summary No
Page 0

[OPTIONS]
Units LPS
Headloss D-W
Specific Gravity 1
Viscosity 1
Trials 40
Accuracy 0.001
Unbalanced Continue 10
Pattern 1
Demand Multiplier 1.0
Emitter Exponent 0.5
Quality None mg/L
Diffusivity 1
Tolerance 0.01

[COORDINATES]
;Node X-Coord Y-Coord
2 2400.00 5200.00
3 4000.00 5200.00
4 5600.00 5200.00
6 2400.00 3600.00
7 4000.00 3600.00
8 5600.00 3600.00
9 7200.00 3600.00
1 800.00 8800.00
5 7200.00 10000.00

283
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

[VERTICES]
;Link X-Coord Y-Coord
1 800.00 5200.00
4 7200.00 5200.00

[LABELS]
;X-Coord Y-Coord Label & Anchor Node

[BACKDROP]
DIMENSIONS 0.00 0.00 10000.00 10000.00
UNITS None
FILE
OFFSET 0.00 0.00

[END]

The solution for the heads or HGLs at the nodes and discharges in each pipe (the primary variables)
were obtained using the hydraulic simulation software EPANET. The nodal heads and the discharges in
each pipe solved by EPANET are shown in Figure 11.4 and Figure 11.5 respectively. The minimum
allowable pressure head for this example is 20 m. The pressure heads at each delivery node and the
velocities in each pipe are shown in Figure 11.6 and Figure 11.7.

Table 11.9 EPANET results for pipes in the ten pipe network
Pipe Up – Down – Q (L/s) V (m/s) f
stream stream
node node
1 1 2 114.34 1.149 0.0191
2 2 3 36.30 0.716 0.0213
3 3 4 -33.47 -0.661 0.0214
4 5 4 118.66 1.192 0.0191
5 2 6 61.04 1.205 0.0207
6 3 7 52.77 1.041 0.0208
7 4 8 68.19 1.346 0.0206
8 6 7 11.04 0.341 0.0240
94 7 8 13.81 0.427 0.0234
10 8 9 32.00 0.989 0.0221

Table 11.10 Node properties for ten pipe network

Node ID Head (m) Pressure head Minimum Excess nodal
(m) allowable pressure head
pressure head (m)
(m)
1 365.0 0.0 N/A
2 347.67 27.67 20.0 7.67
3 344.17 25.17 20.0 5.17
4 347.16 25.16 20.0 5.16
5 372.0 0.0 N/A
6 338.04 28.04 20.0 8.04
7 336.92 26.92 20.0 6.92
8 335.2 27.20 20.0 7.2
9 326.54 22.54 20.0 2.54

284
Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

Figure 11.4 Heads or HGLs (m) at nodes from EPANET for 10 pipe network

Figure 11.5 Discharges or flows (L/s) in pipes from EPANET for 10 pipe network

285
 Chapter 11 - File: Chap16-Comp-Anal-v04.fm July 6, 2022 Version 6

Figure 11.6 Pressure heads (m) at nodes from EPANET for 10 pipe network (ph=pressure head)

Figure 11.7 Velocities (m/s) for pipes from EPANET for 10 pipe network

11.6 References
Cross, H. (1936). Analysis of flow in networks of conduits or conductors. Bulletin No. 286, University of
Illinois Engineering Experimental Station, Urbana, Illinois.

286
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

CHAPTER 12 SOLVING THE Q – EQUATIONS1

12.1 Introduction
Solution of the flow equations or Q – equations formulation based on the Newton method is presented in
this Chapter. For the Q – equations, the continuity, closed simple loop energy and required independent
path energy equations are expressed in terms of unknown flows in the pipes (or other flow elements) in
the network. There are NP unknowns for this formulation.

12.2 Solving the Q – equations for a Water Distribution System

12.2.1 The set of Q – equations
Consider the general set of nonlinear equations for a set of unknown flows for the NP pipes in a network
expressed by the following set of NP – functions made up of the continuity equations and the loop
energy equations and required independent path energy equations as described in Chapter 9 as

f 1  Q 1 Q 2  Q NP  = 0

f 2  Q 1 Q 2  Q NP  = 0

f NP  Q 1 Q 2  Q NP  = 0 (12.1)

where

• f j  Q 1 Q 2  Q NP  = jth function in the Q – equations formulation (either a continuity

equation or an energy equation around a loop or along a required path)
• NP = number of pipes in the network
• Q j = flow in pipe j (or other flow element) (m3/s or ft3/s)

The unknown flows that are required to be solved for is the column vector as follows

T
q =  Q 1 Q 2  Q NP  (12.2)

Let the true solution to the set of nonlinear equations in Equation 12.1 be the following column vector

* * * * T
q =  Q 1 Q 2  Q NP  (12.3)

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.

287
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

In order to find a solution by a successive iterative technique, let the deviation or correction column vec-
tor for the unknown flow problem be defined as

T
q =  Q 1 , Q 2 ,, Q NP  (12.4)

This is the quantity that will be solved for in the Newton iterative method.

In the Newton numerical solution technique discussed in the next section the objective is to find a solu-
k + 1 *
tion q at the (k +1)st iteration that is close to q within an acceptable tolerance, in other words
*
when q is very small. Thus the relationship between the true or correct flow vector q , the current
flow vector q , and the flow correction vector q is

*
q = q + q (12.5)

when q is small.

An initial estimate (referred to as the zeroth iteration indicated by the superscript zero (0)) needs to be
made for each of the unknown flows to commence the iterative solution process. The column vector of
initial estimates of the pipe flows is

0 0 0 0 T

q =  Q 1  Q 2   Q NP  (12.6)

A commonly used value for commencing the iterative solution process is to set all velocities to 1.0 fps
(0.3048 m/s) in all pipes and then to compute the corresponding discharges (Rossman 2000).

12.2.2 The Jacobian matrix for the unknown flow equations

The solution for the unknown flows requires linearization of the nonlinear governing equations that
were presented in Section 9.7 and given in Equation 12.1. In the neighborhood of the current iteration
point each of the functions in Equation 12.1 may be expanded in a Taylor series as shown in
Equation 10.8

NP  f q
1
f 1 q + q  = f 1  q  + - Q j + O  Q 
 -----------------
Q j
j=1

NP  f q
2
f 2  q + q  = f 2  q  + - Q j + O  Q 
 -----------------
Q j
j=1

NP f q
NP
f NP  q + q  = f NP  q  + - Q j + O  Q 
 --------------------- (12.7)
Q j
j=1

where

• f j  q , f j  q + q = jth function in the Q – equations formulation (either a continuity or an

energy equation)

288
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

• q = vector of the unknown flows (m3/s or ft3/s)

• q = deviation vector of the unknown flows; Q j = jth deviation flow in the q vector
(m3/s or ft3/s)
• NP = number of pipes in the network
2
• O  Q  = second order and higher terms of the flow corrections
Putting Equation 12.7 into matrix notation using Equation 10.9 (Press et al. 1996) gives

2
f  q + q  = f  q  + J  q q + O  q  (12.8)

where

• J  q  = Jacobian matrix based on the flows

From the set of equations in Equation 12.8, the Jacobian matrix for the flow equations formulation based
on the values of the flows at the iteration k is defined as

k  k  k 
 f 1q   f 1q   f 1q 
------------------------ ------------------------  -----------------------
-
Q 1 Q 2 Q NP
k  k  k 
k 
 f 2q   f 2q   f 2q 
q  = ------------------------ ------------------------  -----------------------
- (12.9)
Q 1 Q 2 Q NP
   
k  k  k 
 f NP  q   f NP  q   f NP  q 
---------------------------- ----------------------------  ---------------------------
-
Q 1 Q 2 Q NP

2
If q is small then second order and higher terms of O  q  can be dropped and also by setting in
Equation 12.8

*
f  q + q  = f  q  = 0

gives

k  k + 1 k 
Jq q = –f  q  (12.10)

EQUATIONS TO BE SOLVED FOR THE Q – EQUATIONS FORMULATION

or

k  k  k 
 f 1q   f 1q   f 1q  k + 1
------------------------ -  -----------------------
----------------------- - Q 1 k  k  k 
Q 1 Q 2 Q NP f 1  Q 1  Q 2   Q NP 
k + 1
k  k  k  Q 2 k  k  k 
 f 2q   f 2q   f 2q  f 2  Q 1  Q 2   Q NP 
------------------------ -  -----------------------
----------------------- - = – (12.11)
Q 1 Q 2 Q NP  
     
k  k  k  k  k  k 
 f NP  q   f NP  q   f NP  q  k + 1 f NP  Q 1  Q 2   Q NP
--------------------------- ----------------------------  ---------------------------- Q NP
Q 1 Q 2 Q NP

289
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Note the subscripts k and (k + 1) have been introduced to represent the successive iteration numbers.
k + 1
The flow correction vector q on the left – hand side of Equation 12.10 is for the (k + 1)st itera-
k  k + 1
tion. The Jacobian matrix J  q  and the function vector q are Q – equations formulation the
k 
flows q at iteration k.

k + 1
Equation 12.11 will be solved for the unknown vector q to avoid computation of the inverse of
the Jacobian.

k + 1
The vector of corrections to the flows q can be expressed from Equation 12.10 as

k + 1 –1 k  k 
q = –J  q f  q  (12.12)

where

–1 k 
• J  q  = inverse of the Jacobian matrix for the Q – equations formulation based on the
flows at iteration k (the Jacobian matrix is defined by Equation 12.9)
As mentioned previously, Equation 12.12 is not usually solved. It is desirable to avoid actually comput-
k + 1
ing the Jacobian explicitly. Instead, Equation 12.10 is solved for the vector q of corrections at
iteration (k + 1). This equation is a set of linear equations and may be solved by an LUD method
(lower – upper decomposition) or a sparse matrix solution technique.

k + 1
The vector of corrections q at the (k + 1)st iteration from Equation 12.10 is then added to the cur-
k 
rent solution vector q to give the next best estimate of the unknowns as

k + 1 k k + 1
q = q + q (12.13)

or

k + 1
k + 1 k  Q 1
Q1 Q1
k + 1
k + 1 k  Q 2
Q2 Q2
= +  (12.14)
 
  
k + 1 k 
Q NP Q NP k + 1
Q NP

k + 1
The process is iterated until the vector q converges to within an acceptable tolerance. During the
iterative process it is possible that a variable may become zero.

12.2.3 Derivatives of terms incorporating a modulus of flow

In Chapter 9, formulations of the network equations involving individual pipe flows incorporated a
modulus of the flow to ensure the sign of the head loss in the pipe is consistent with the sign of flow or
velocity. To compute the Jacobian matrix, derivatives of terms including a modulus term are required.
Consider the function taking the derivative of the function f  Q  in a single unknown variable flow of
Q, then based on Lambert (1994)

290
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

 f  Q  - = ----------------
d----------------------- f  Q - d-----------------
f Q (12.15)
dQ f  Q  dQ

or the derivative of the product is

f Q d f Q d f Q
-  f  Q  f  Q   = f  Q  ----------------- ------------------ + f  Q  ------------- (12.16)
Q f  Q  dQ dQ

12.2.4 Jacobian values and plots of nonlinear flow equation terms

Now consider the head loss function for pipe j between nodes i and k that appears in energy equations in
the Q – equation formulation as

n–1
f Q j = r jQ j Q j (12.17)

The partial derivative of Equation 12.17 using Equation 12.16 is

n–1 n–1
f  n–1 n – 1 Q j Qj  n–1
----- = r j Q j  Qj  + Qj = r j Q j ------------------
- Q j  + Qj  1.0 
Qj Q j Q j Qj
n – 1 Q j

and

n–1 n–1
----- Qj
f - = r Q ------------------ n–1–1 n–1 n–1 n–
j j -  n – 1 Q j + Qj = r jn – 1 Q j + Qj  = nr j Q j
Qj Qj
n–1

Thus the partial derivative of the expression containing a modulus in Equation 12.17 is

f n–1
--------- = nr j Q j (12.18)
Q j

Example 12.1 Consider the case of n = 2 in the Darcy – Weisbach head loss equation formulated for pipe 1 in
Figure 9.3 using Equation 12.17.

n–1 2–1
f  Q1  = r 1 Q1 Q1 = r 1 Q1 Q1 = r 1 Q1 Q1 (12.19)

Determine the derivative of function with respect to Q 1 .

Answer – Computation of the derivative of the Darcy – Weisbach head loss equation including a modulus sign.

The derivative of Equation 12.19 based on Equation with n = 2 is

f n–1 2–1 f
--------- = nr 1 Q j = 2r 1 Q j or --------- = 2r 1 Q 1
Q 1 Q 1

Example 12.2 Consider the case of n = 1.852 in the Hazen – Williams head loss equation. Determine the deriva-
tive with respect to Q 1 of the following function for pipe 1 in Figure 9.3

n–1 1.852 – 1 0.852

f  Q1  = r 1 Q1 Q1 = r 1 Q1 Q1 = r 1 Q1 Q1 (12.20)

Answer – Computation of the derivative of the Hazen – Williams head loss equation including a modulus sign.

The derivative of Equation 12.19 based on Equation is

291
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

f n–1 0.852
= nr 1 Q 1 = 1.852r 1 Q 1
 Q1

or

f 0.852
= 1.852r 1 Q 1
 Q1

Example 12.3 Consider the function from Equation 12.17 for pipe 1 in Figure 9.3 assuming a Darcy – Weisbach
head loss equation is used where n = 2 as follows

n–1
f  Q1  = r 1 Q1 Q1 = r 1 Q1 Q1 (12.21)

Assume r 1 = 1.0 .

(i) Compute the function at Q 1 = – 0.5 (m3/s or ft3/s)

(ii) Plot the function for the range of Q 1 = – 1 to Q 1 = 1 (m3/s or ft3/s)

(iii) What is the formula for the derivative of this function?

(iv) Compute the derivative of the function at Q 1 = – 0.5 (m3/s or ft3/s)

(v) Plot the derivative of the function for the range of Q 1 = – 1 to Q 1 = 1 (m3/s or ft3/s)

Answer – (i) Computation of the Darcy – Weisbach head loss equation for a particular flow.

The function becomes for r 1 = 1.0 and thus

f  Q1  = Q1 Q1

The function value at Q 1 = – 0.5 (m3/s or ft3/s) is

f  Q 1  = r 1 Q 1 Q 1 = 1.0  – 0.5  – 0.5 = – 0.25 (m or ft)

___________________________________________________________________________________

Answer – (ii) Plot of the head loss function for the range of Q 1 = – 1 to Q 1 = 1 (m3/s or ft3/s).

The plot of the function is shown in Figure 12.1 with the values tabulated in Table 12.1.

Note that the gradient of this function at Q 1 = 0 is zero and the slope of this function is always positive.

___________________________________________________________________________________
Answer – (iii) The formula for the derivative of the head loss function.

The derivative using Equation with r 1 = 1.0 and n = 2 for the Darcy – Weisbach head loss equation is

 f  Q1  n–1 2–1
------------------ = nr 1 Q 1 = 2  1.0  Q 1 = 2 Q1 (12.22)
Q 1

___________________________________________________________________________________

Answer – (iv) Computation of the derivative of the head loss function at Q 1 = – 0.5 (m3/s or ft3/s).

The derivative of the function value at Q 1 = – 0.5 (m3/s or ft3/s) from Equation 12.22 is

 f  Q1 
------------------ = 2 Q 1 = 2.0 – 0.5 = 1.0
Q 1

This value is confirmed by the computed value shown in Figure 12.2.

292
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Figure 12.1 Plot of the function f  Q 1  = Q 1 Q 1 for Q 1 =  – 1 1 

___________________________________________________________________________________

Answer – (v) Plot of the derivative of the head loss function for the range of Q 1 = – 1 to Q 1 = 1 (m3/s or ft3/s).
This function is plotted in Figure 12.2. Tabulated values are shown in Table 12.1.

Table 12.1 Function values and derivative values for f  Q 1  = Q 1 Q 1

Q1 Q1 Q1  Q Q Q1 Q1 Q1  Q Q
 Q1 1 1  Q1 1 1

–1 –1 2 0.05 0.0025 0.1

– 0.9 – 0.81 1.8 0.1 0.01 0.2
– 0.8 – 0.64 1.6 0.2 0.04 0.4
– 0.7 – 0.49 1.4 0.3 0.09 0.6
– 0.6 – 0.36 1.2 0.4 0.16 0.8
– 0.5 – 0.25 1 0.5 0.25 1
– 0.4 – 0.16 0.8 0.6 0.36 1.2
– 0.3 – 0.09 0.6 0.7 0.49 1.4
– 0.2 – 0.04 0.4 0.8 0.64 1.6
– 0.1 – 0.01 0.2 0.9 0.81 1.8
– 0.05 – 0.0025 0.1 1 1 2
0.0 0.0 0.0

293
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Figure 12.2 Plot of the derivative of the function f  Q 1  = Q 1 Q 1 for Q 1 =  – 1 1 

The derivative has a slope discontinuity at Q 1 = 0 and the derivative is always a positive value for all Q 1 values.

_______________________________________________________________________________________________________

12.3 Convergence Stopping Criteria for the Q – equations Formula-

tion
As mentioned in Section 10.2.3, Equation 10.22 is not suitable as a convergence stopping criterion
because the nodal heads are not computed at each iteration. As an alternative, a stopping test based on
the head loss around closed loops and along paths between nodes of fixed head is proposed.

Similarly to Equation 10.23, for the Q – equations formulation let the loop equation (for loop s) be desig-
nated respectively as

 Q- L  q  =  r jQ j Q j (12.23)
s
j  P Ls

and for the path equations (for path s)

 Q- P  q  =  r j Q j Q j +  E Lm – E Lq  (12.24)
s
j  P Ls

where P L s =indices of the pipes in loop or path s.

The infinity – norm test (based on Equation 10.19) can be used to stop the iterative convergence process
when for the closed loop equations

k + 1 max k + 1
 Q- L  q  =   Q- L  q     stop (12.25)

s
j=1, NL s

and for the path equations

k + 1 max k + 1
 Q- P  q  =   Q- P  q     stop (12.26)

s
j=1,  NF – NC  s

Both these conditions need to be satisfied.

294
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

The checking of the nodal continuity equations is not required as the continuity equations exactly satisfy
continuity at each node from iteration one onwards. As a result, performing this check would be redun-
dant for the Q – equations formulation.

12.4 The Q – equations for the Ten Pipe Network

In this section a formulation of the Q – equations for a ten pipe network is given while a numerical
example of a two pipe network is given in Section 12.5 and Section 12.6.

12.4.1 Unknowns and initial guesses for the Newton solution

Consider the equations for the network in Figure 12.3 formulated in terms of 10 flows (Q1 to Q10 – thus
NP = 10) in each of the pipes for the network previously considered in Chapter 9. The properties of the
network are given in Table 12.2 and Table 12.3. The column vector of unknown flows from
Equation 12.2 at the kth iteration is

k  k  k  k  T
q =  Q 1  Q 2   Q 10  (12.27)

The column vector of flow corrections in the pipes (from Equation 12.4) that will be solved for itera-
tively in the Newton scheme for the ten pipe network at the (k + 1)st iteration becomes

k + 1 k + 1 k + 1 k + 1 T
q =  Q 1  Q 2   Q 10  (12.28)

A column vector of initial estimates for the flows in each pipe needs to be made in order for the iterative
solution process to begin

0 0 0 0 T

q =  Q 1  Q 2   Q 10  (12.29)

12.4.2 Computation of the pipe resistance terms

Darcy – Weisbach Equation

0
The pipe resistance terms r j can now be computed for the Q – equations formulation based on the ini-
0 0 0 0
tial guesses of the flows Q j s. The resistance term r – values  r 1  r 2   r 10  for the Darcy –
Weisbach head loss equation for each pipe that will be used for computing terms in both Equation 12.17
and Table 12.5 (n = 2) may be computed from Equation 4.122 for pipe j for either SI units or US cus-
tomary units as

8 f jL j
r f = ------------------
- (12.30)
j 2 5
 gD j

The Darcy – Weisbach friction factor or resistance factor fj for each pipe depends on both Reynolds

number Rej and relative roughness ------j and would need to be updated at each iteration. The Swamee and
Dj
Jain equation for the friction factor fj from Equation 4.58 for pipe j may be used as

0.25
f j = --------------------------------------------------------------- (12.31)
 j 5.74 -
2
log 10  ------------- - + -------------- 
 3.7D j Re 0.9
j

295
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

If the flow in a pipe is laminar, Equation 12.30 and Equation 12.31 do not apply. The appropriate
expression for the laminar resistance factor is given in Chapter 16 in Section 16.2.4.

Hazen – Williams Equation

The r – values  r 1 r 2  r 10  for the Hazen – Williams head loss equation for each pipe in network
would be computed for each pipe j from Equation 4.128 for SI units as

Lj
r H W = 10.670 ----------------------------------- (12.32)
j 1.852 4.871
Cj Dj

or from Equation 4.128 for US customary units as

Lj
r H W = 4.7279 ----------------------------------- (12.33)
j 1.852 4.871
Cj Dj

The Hazen – Williams coefficient Cj for each pipe j is assumed to be constant and thus would not need to
be updated at each iteration, in contrast to the Darcy – Weisbach head loss equation.

Figure 12.3 A ten pipe network – unknown Q – equations formulation

The initial estimates of the Darcy – Weisbach friction factors or resistance factors corresponding to the
guessed velocities of 1.0 fps (0.3048 m/s) are also shown in Table 12.2.
Table 12.2 Computation for ten pipe network of friction factors & resistance terms for an assumed
velocity of 1.0 fps (0.3048 m/s) in each pipe
Pipe Up – Down –  V (m/s) f
Re = VD
-------- 8 fL
r f = -----------------
stream stream (mm) = 1.0 fps  j 2 5
node node  gD
1 1 2 0.25 0.3048 95131 0.0214 1484.9
2 2 3 0.25 0.3048 67874 0.0232 2902.3
3 3 4 0.25 0.3048 67874 0.0232 2902.3
4 5 4 0.25 0.3048 95131 0.0214 1979.9
5 2 6 0.25 0.3048 67874 0.0232 2902.3
6 3 7 0.25 0.3048 67874 0.0232 2902.3

296
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Table 12.2 Computation for ten pipe network of friction factors & resistance terms for an assumed
velocity of 1.0 fps (0.3048 m/s) in each pipe
7 4 8 0.25 0.3048 67874 0.0232 2902.3
8 6 7 0.25 0.3048 54246 0.0246 9437.9
9 7 8 0.25 0.3048 54246 0.0246 9437.9
10 8 9 0.25 0.3048 54246 0.0246 9437.9
Note that in Table 12.2, the values of velocity in each pipe are the initial guesses to commence the Newton iterative solution pro-
cess. The r f values are also based on the friction factor or resistance factor for this initial guess of velocity.
j

The updating of the pipe resistance terms r j can now proceed based on the revised Q j s at each itera-
tion.

The 10 equations for flows in the ten pipe network in Figure 12.3 are made up of seven continuity equa-
tions (at each of the non – reservoir nodes), three energy equations around the two closed simple loops
and for the one required independent path energy equation between the two reservoirs. The continuity
equations are linear, while the loop and path equations are nonlinear in terms of the flows.

The set of governing equations for the ten pipe network were presented in Chapter 9. These are repeated
in Table 12.3 for the Darcy – Weisbach head loss equation with n = 2.

Table 12.3 The flow or Q – equations formulation for the ten pipe network using the Darcy –
Weisbach head loss equation (n=2)
Continuity equations, energy equations for closed simple loops or required
independent paths for values of Q j at iteration k

f 1 q  = – Q 1 + Q 2 + Q 5 + DM 2 = 0 f 4 q  = – Q 5 + Q 8 + DM 6 = 0

f 2 q  = – Q 2 + Q 3 + Q 6 + DM 3 = 0 f 5 q  = – Q 6 – Q 8 + Q 9 + DM 7 = 0

f 3 q  = – Q 3 – Q 4 + Q 7 + DM 4 = 0 f 6 q  = – Q 7 – Q 9 + Q 10 + DM 8 = 0

f 7 q  = – Q 10 + DM 9 = 0

f 8  q  = r 2 Q2 Q2 + r 6 Q6 Q6 – r 8 Q8 Q8 –r 5 Q5 Q5 = 0

f 9 q  = r 3 Q 3 Q 3 + r 7 Q 7 Q 7 – r 9 Q 9 Q 9 – r 6 Q 6 Q 6 = 0

f 10 q  = r 4 Q 4 Q 4 – r 3 Q 3 Q 3 – r 2 Q 2 Q 2 – r 1 Q 1 Q 1 –  E L 5 – E L 1  = 0

12.4.3 Jacobian matrix for the flow equations for the ten pipe network
The Jacobian matrix of Equation 12.9 based on values at the kth iteration is a 10 by 10 square matrix and
becomes

297
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

k  k  k 
 f 1q   f 1q   f 1q 
------------------------ ------------------------  -----------------------
-
Q 1 Q 2 Q 10
k  k  k 
k 
 f 2q   f 2q   f 2q 
q  = ------------------------ ------------------------  -----------------------
- (12.34)
Q 1 Q 2 Q 10
   
k  k  k 
 f 10  q   f 10  q   f 10  q 
-------------------------- --------------------------  --------------------------
Q 1 Q 2 Q 10

Now compute the elements of the Jacobian for the pipe network in Figure 12.3. The first function from
Table 12.3 is

k  k  k  k 
f 1 q  = – Q 1 + Q 2 + Q 5 + DM 2 (12.35)

The partial derivatives or the first row of the Jacobian matrix in Equation 12.34 are given in Table 12.4.
Note that the derivatives of each of the terms in the continuity equations can only be -1, +1 or zero as
these terms are linear in Qj.

Table 12.4 The Jacobian matrix elements for function 1 in Table 12.3
Element Jacobian elements for values of Q j at iteration k

k  k  k 
Non – zero ele-  f 1q   f 1q   f 1q 
------------------------ = – 1 ------------------------ = 1 ------------------------ = 1
ments Q 1 Q 2 Q 5

k 
Zero elements  f 1q 
------------------------ = 0.0 for j = 3, 4, 6, ... , 10
Q j

These Jacobian values are constant and do not vary from iteration to iteration. This is true for all conti-
nuity equations in the Q – equations formulation. All of the other Jacobian elements for the continuity
equations in functions two to seven in Table 12.3 are computed in a similar way as shown above and are
shown in Table 12.5.

The eighth function from Table 12.3 based on the first energy equation for loop 1 is

k  k  k  k  k  k  k  k  k  k  k  k 
f 8  q  = r 2 Q2 Q2 + r 6 Q6 Q6 – r 8 Q8 Q8 – r 5 Q5 Q5 = 0 (12.36)

The partial derivatives or the eighth row of the Jacobian matrix in Equation 12.34 are now evaluated.

The first element in eighth row of the Jacobian matrix is

k 
 f 8q 
------------------------ = 0.0 (12.37)
Q 1

k 
The second derivative of the function f 8 q  requires the general form from Equation 12.18 of

 n–1 n–1
r jQ j Q j  = nr j Q j (12.38)
Q j

298
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Thus the derivative of Equation 12.36 with respect to the flow in pipe Q 2 for n = 2 is

k 
 f 8q  k  n – 1 k 
------------------------ = nr 2 Q 2 = 2r 2 Q 2 (12.39)
Q 2

k 
The other derivatives of the function f 8  q  with respect to Q 6 , Q 8 and Q 5 respectively using
Equation 12.38 are similar to Equation 12.39 with a single term each and are shown in Table 12.5. All of
the other Jacobian elements for the loop equation and the required independent path equation in func-
tions nine and ten in Table 12.3 are computed in a similar way as shown above.

Thus the full 10 by 10 square Jacobian matrix of Equation 12.34 for the flow equations for the ten pipe
network based on flows at iteration k is shown in Table 12.5.

Table 12.5 The Jacobian matrix for the Q – equations at iteration k for the ten pipe network (the Q
and r values are based on the values at the kth iteration
1 2 3 4 5 6 7 8 9 10
1 –1 1 0 0 1 0 0 0 0 0
2 0 –1 1 0 0 1 0 0 0 0
3 0 0 –1 –1 0 0 1 0 0 0
4 0 0 0 0 –1 0 0 –1 0 0
5 0 0 0 0 0 –1 0 –1 0 1
6 0 0 0 0 0 0 –1 0 –1 1
7 0 0 0 0 0 0 0 0 0 –1
8 0 2r 2 Q 2 0 0 –2 r 5 Q5 2r 6 Q 6 0 – 2r 8 Q 8 0 0
9 0 0 2r 3 Q 3 0 0 –2 r 6 Q6 2r 7 Q 7 0 –2 r 9 Q9 0
10 –2 r 1 Q1 –2 r 2 Q2 –2 r 3 Q3 2r 4 Q 4 0 0 0 0 0 0

It may be seen that the Jacobian for the Q – equations formulation in Table 12.5 is not a symmetric
matrix. The matrix is also sparse (30% of the elements in the Jacobian matrix are non – zero).

12.4.4 Solving for the flow corrections for the ten pipe network
Once the Jacobian matrix for iteration k has been evaluated the set of linearized equations to be solved
for the corrections to the pipe flows during the iterative process from Equation 12.10 is

k + 1 k 
Q 1 f 1q 
k  k + 1 k 
Jq  Q 2 = – f 2q  (12.40)
 
k + 1 k 
Q 10 f 10  q 

or

k  k + 1 k 
Jq q = –f  q  (12.41)

299
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Once the Jacobian matrix has been determined for a particular iteration as well as the function values on
the right hand side of Equation 12.41 based on the flow values at iteration k, then Equation 12.41 must
k + 1
now be solved for the vector of flow corrections q . During the iterative solution procedure, the
k + 1
corrections q for each iteration are added to the current solution vector to give the next best esti-
mate of the unknowns as

k + 1 k  k + 1
q = q + q (12.42)

12.4.5 Solving the flow equations for the ten pipe network
Assume all the initial guesses for the flows are based on a velocity of 1.0 fps (0.3048
m/s) (Rossman, 2000). Based on Table 12.5 for the Jacobian for the ten pipe network it follows that

(12.43)

12.5 A Two Pipe Network

12.5.1 The network and the formulations
To investigate the details of the Newton iterative solution of the various formulations of the pipe net-
work equations a simple two pipe network as shown in Figure 12.4 is considered (see Table 12.6). This
simpler network is taken for detailed consideration of the iterative steps rather than the ten pipe network
previously considered in Chapter 9. Each of the formulations for the two pipe network is considered in
turn in this Chapter through Chapter 16.

Table 12.6 Nature of matrices required for each formulation for the two pipe network
Method Matrix size Matrix size Type of matrix Chapter
for
network in
Figure 12.4
Q – equations NP  NP 22 Non – 11
symmetric
H – equations NJ  NJ 11 Symmetric 12
LF – equations* NLT  NLT 11 Symmetric 13
Q+H equations  NJ + NP    NJ + NP  33 Non – 14
symmetric
Global Gradient Algo- NJ  NJ 11 Symmetric, 15
rithm formulation (Todini positive definite
and Pilati 1988)
* where NLT = total number of loops and required independent paths in the network= NL + (NF – NC)
(see Section 9.2.2 for definition of variables)

300
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Figure 12.4 A two pipe network – unknown Q – equations formulation

The two pipe network has two reservoirs and one demand node. Assume the temperature of the water is
–6 2
20  C with a kinematic viscosity of  = 1.004 10 m /s. The pipe and node properties of the net-
work are given in Table 12.7 and Table 12.8.

Table 12.7 Pipe properties for the two pipe network

Pipe Upstream Downstream D L 
node node (mm) (m) (mm)
1 2 1 300 1000 0.25
2 1 3 300 1000 0.25

Table 12.8 Node properties for the two pipe network

Node ID Elevation (m) Demand (L/s)
1 40.0 50
2 80.0 Reservoir
3 50.0 Reservoir

12.5.2 The EPANET results

The hydraulic simulation software EPANET results are discussed in this section1. The solution for the
discharges, the velocities and the friction factors or resistance terms in each pipe from EPANET are
shown in Figure 12.5 and Table 12.9.

Table 12.9 Pipe quantities for the two pipe network from solution by EPANET
Pipe Discharge Velocity f r f ** h f **
(L/s) (m/s)
(m)
1 173.606 2.456 0.01937 658.64 19.851
2 123.606 1.749 0.01957 665.44 10.167
** Computed outside EPANET for use in example Example 12.4 below and Example 13.7.

The EPANET results for node 1 are summarized in Figure 12.6 and Table 12.10. The minimum allow-
able pressure head for this example is 20 m.
Table 12.10 Head and pressure values for the node in the two pipe network from EPANET
Node Head Pressure head
(m) (m)
1 60.159 20.159

–5
1. A scaling factor of 0.98539 was used in EPANET to adjust the 15C value of kinematic viscosity 1.14 10 ft2/s to
–6
give the same Reynolds number for a kinematic viscosity 1.004 10 m2/s at 20C with a velocity of 0.6 m/s (also
taking into account that EPANET uses a conversion factor of 0.3048 rather than 1.0/3.28).

301
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Figure 12.5 EPANET pipe results for the two pipe network

Figure 12.6 EPANET node results for the two pipe network

12.6 Solving the Q – equations for the Two Pipe Network

12.6.1 The unknowns and initial guesses
Now consider the flow equations for the network in Figure 12.7. Flows for each of the two pipes need to
be solved for, that is Q1 and Q2. From Equation 12.2 (page 287), the column vector of unknown flows at
the kth iteration for the two pipe network is

k  k  k  T
q =  Q1  Q2  (12.44)

Figure 12.7 Unknown Qs to be solved for in the two pipe network

The functions to be solved include one continuity equation at node 1 and one path energy equation
between the two reservoirs (see Equation 12.49 and Equation 12.50 in Example 12.4 below).

The column vector of flow corrections in the pipes to be solved at each iterative step in the Newton
method (from Equation 12.4, page 288) for the two pipe network at the (k + 1)st iteration becomes

k + 1 k + 1 k + 1 T
q =  Q 1  Q 2  (12.45)

A column vector of initial estimates for the flows in each pipe needs to be made in order for the iterative
solution process to begin

0 0 0 T

q =  Q1  Q2  (12.46)

302
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

0
The initial flows Q j are often based on a flow velocity of 1.0 fps (0.3048 m/s) (as used in EPANET -
Rossman 2000) as shown in Table 12.11. The pipe properties for the network are given in Table 12.7.

Table 12.11 The flow values for an initial guess of velocity of 1.0 fps (0.3048 m/s)
Pipe Diameter Area (m2) V (m/s) V (m/s)* Q (m3/s) Q (L/s)
(mm)
1 300 0.070686 0.3048 0.3048 0.02155 21.55
2 300 0.070686 0.3048 0.3048 0.02155 21.55
* Computed based on 1.0/3.28 for conversion of fps to m.

The pipe resistance terms ( r f =  r 1 r 2  ) can now be computed based on initial flow guesses shown in
2
Table 12.11 for the Darcy – Weisbach head loss equation h f = r f Q (where n = 2) using
j

Equation 12.39. The pipe resistance term values r j are shown in Table 12.12 with the friction factors
computed from the Swamee and Jain equation (Equation 4.58). These pipe resistance values r j are
updated at each iteration based on the latest flow values.

12.6.2 The convergence stopping test for the Q – equations (two pipe exam-
ple)
Based on Equation 12.24, the path head loss equation (for the one and only path as there are no loops) is
given from Equation 12.50 with s = 1 as

k + 1
 Q- P  q  = – r 1 Q 1 Q 1 – r 2 Q 2 Q 2 +  EL 2 – E L 3  (12.47)
1 

Convergence of the iterative solutions process will be stopped when

k + 1
 Q- P  q    stop (12.48)
1 

–6
where  stop = 1.0 10 m, for example

12.6.3 Two pipe example based on Newton method (Q – equations formula-

tion)

Example 12.4 For the two pipe network in Figure 12.4 complete the following:

(i) Develop the Q – equations formulation for the two pipe network system.
(ii) Verify that the EPANET flow values from Table 12.9 satisfy the Q – equations formulation for the two pipe net-
work.
(iii) Develop the Jacobian matrix for the Q – equations formulation and carry out the first two iterations for the two
pipe network.
(iv) Compute the condition number of the Jacobian matrix at the final iteration for the Q – equations formulation.
(v) Tabulate all iterations for the solution process for the Q – equations formulation.
(vi) Plot the convergence characteristics for the iterations for the Q – equations formulation for the two pipe network.
Assume the Darcy – Weisbach head loss equation is used for each pipe.

Answer – (i) Develop the Q – equations formulation for the two pipe network system.

The Q – equations formulation of the governing equations for this two pipe network includes a continuity equation at
node 1 and a required independent path equation between the two reservoirs at nodes 2 and 3 that includes pipes 1 and
2 (in a clockwise direction) as follows:

303
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

f 1  q  = – Q 1 + Q 2 + DM 1 = 0 at node 1 (12.49)

f 2  q  = – r 1 Q 1 Q 1 – r 2 Q 2 Q 2 +  EL 2 – E L 3  = 0 along the path between reservoirs (12.50)

Note that the “constants” for the two terms for rf in Equation 12.50 are dependent on the friction factor or resistance
factor f and hence are required to be updated at each iteration. The values for rf based on the initial flow guesses are
shown in Table 12.12.

Substituting the known quantities into Equation 12.49 and Equation 12.50 for the calculation for the first iteration
gives:

f 1  q  = – Q 1 + Q 2 + 0.050 = 0 at node 1 (12.51)

f 2  q  = – 746.954 Q 1 Q 1 – 746.954 Q 2 Q 2 +  80.0 – 50.0  = 0 along the path between reservoirs (12.52)

(assuming a velocity of 1.0 fps (0.3048 m/s) in each pipe)

___________________________________________________________________________________
Answer – (ii) Verify that the EPANET flow values from Table 12.9 satisfy the Q – equations formulation for the
two pipe network.

The EPANET values for flows from Table 12.9 are substituted into Equation 12.49 and Equation 12.50 (as follows:
3
f 1  q  = – 0.173606 + 0.123606 + 0.050 = 0.00 m /s

f 2  q  =  – 658.64   0.173606  0.173606 –  665.44  0.123606 0.123606 +  80.0 – 50.0  = – 0.0177 m

Note that the first continuity equation is satisfied exactly while in the path equation is in error by 18 mm. This is
acceptable and represents an error of 0.05% in 30.0 meters of head difference between the two reservoirs.
___________________________________________________________________________________
Answer – (iii) Develop the Jacobian matrix for the Q – equations formulation and carry out the first two itera-
tions for the two pipe network.

k + 1
The system to be solved iteratively for a sequence of q values is:

k  k + 1 k 
Jq q = –f  q  (12.53)

The terms in the 2 by 2 Jacobian at the kth iteration can be determined by taking the appropriate partial derivatives of
Equation 12.51 and Equation 12.52 as follows:

k  k 
 f 1q   f 1q 
------------------------ ------------------------
k  Q 1 Q 2 – 1.0 1.0
q  = = k  k 
(12.54)
k 
 f 2q   f 2q 
k  – 2.0r f Q 1 – 2.0r f Q 2
1 2
------------------------ ------------------------
Q 1 Q 2

Note that the Jacobian matrix for the Q – equations formulation is not symmetric.

ITERATION 1, Q – equations
The initial guesses for the flows in the two pipes in this example are based on a standard value of 1.0 fps (0.3048 m/s).
The corresponding velocities in SI units and the flows are given in Table 12.12 and may be represented as:

3
0 m 0 3
Q1 = 0.02155 -------- Q 2 = 0.02155 m /s . (12.55)
s

Thus the vector of unknowns at iteration zero (0) is:

0
= 0.02155 m /s
3
q (12.56)
0.02155

The friction factors or resistance factors (based on the Swamee and Jain equation – Equation 4.58) and corresponding rf

304
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

values (Equation 4.120) for the flows in Equation 12.56 for pipes 1 and 2 in the two pipe network are given in
Table 12.12.

Table 12.12 The rf values for initial flow guesses based on 1.0 fps (0.3048 m/s) for Q – equations formulation

Pipe V D Area (m2) Re**  f L (m) rf value

(m/s)* (m) (mm) (Swamee and Jain) (Equation 4.120)
(Equation 4.58)
1 0.3048 0.3 0.07069 90,828 0.25 0.02197 1000 746.954
2 0.3048 0.3 0.07069 90,828 0.25 0.02197 1000 746.954
–6 2
*Computed based on 1.0/3.28 for conversion of fps to m. **  = 1.007 10 m /s at 20C ;

0.25 fL -
Using in the above Table: f = ---------------------------------------------------------
- ; r f = -----------------
  5.74  2 2 gD A
2
log 10 ------------ + ------------
 3.7 D 0.9
Re

or

f = 0.02197 r f = 746.95
0.02197 746.95

The flows from Equation 12.56 may be substituted into the continuity and the path equations of Equation 12.51 and
Equation 12.52 as follows:
0 3
f 1 q  = – Q 1 + Q 2 + 0.050 = – 0.02155 + 0.02155 + 0.050 = 0.050 m /s

0
f 2q  = – 746.95 Q 1 Q 1 – 746.95 Q 2 Q 2 + 30.0 = – 746.95  0.02155  0.02155 – 746.95  0.02155  0.02155 + 30.0
= – 0.3469 – 0.3469 + 30.0 = 29.306 m

Thus:

0 0.050
f q = (12.57)
29.306

Thus the imbalance in the continuity equation is 50 L/s while the imbalance in the head along the path is 29.306 m and
hence the flows are substantially underestimated by the initial guesses of 1.0 fps (0.3048 m/s) as not enough head loss
is generated to offset the difference of 30.0 m between the two reservoirs.

The Jacobian matrix for the first iteration from Equation 12.54 for the initial flow guesses (Equation 12.56) and the rf
values (Equation 4.120) is:

0 – 1.0 1.0 – 1.0 1.0

Jq  = 0 0
= (12.58)
– 2.0r f Q 1 – 2.0r f Q 2 – 2.0  746.95  0.02155 – 2.0  746.95  0.02155
1 2

or

0 – 1.0 1.0

Jq  = (12.59)
– 32.195 – 32.195

The inverse of the Jacobian matrix is shown below for this simple two pipe network (note that this would never be
actually calculated in practice for a network of normal size – instead a method such as a Gaussian LUD or an alterna-
tive solution method would be used that avoids calculating the inverse explicitly would be employed)

0
 = – 0.5 – 0.015531
–1
J q (12.60)
0.5 – 0.015531

The flow corrections for the first iteration are calculated from Equation 12.12 based on Equation 12.57 and
Equation 12.60 as:

1 0 0

 = – – 0.5 – 0.015531 0.050 = 0.4801
–1
q = –J  q f  q (12.61)
0.5 – 0.015531 29.306 0.4301

305
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

As mentioned above, the inverse of the Jacobian would never be actually calculated in practice for a network of normal
size. It is presented in this way for convenience. The numerical values obtained would be identical to those from a
Gaussian LUD solution procedure.

The new flows after the first iteration are:

1 0 1

= 0.02155 + 0.4801 = 0.5017 m /s
3
q = q + q (12.62)
0.02155 0.4301 0.4517

Note the flows have been increased in size by a factor of 20 to 25 times compared with the initial guesses (and also
these new flows are much larger than the actual solution as obtained from EPANET shown in Table 12.9). The flows
for the first iteration from Equation 12.62 may now be substituted into the continuity equation of Equation 12.51 as fol-
lows:
1 3
f 1 q  = – Q 1 + Q 2 + 0.050 = – 0.5017 + 0.4517 + 0.050 = 0.000 m /s

Note that the continuity equation is exactly satisfied after the first iteration. This remains the case for all subsequent
iterations.

Now values can be substituted into the path equation of Equation 12.52 (without updating the friction factors and the
corresponding rf values)

1
f 2q  = – 746.69 Q 1 Q 1 – 746.69 Q 2 Q 2 + 30.0 = – 746.69  0.5017  0.5017 – 746.69  0.4517  0.4517 + 30.0
= – 310.403 m

(12.63)

Thus the imbalance in the head along the path is now negative 310.4 m (compared to the previous value of positive
29.31 m), thus the flows are now substantially overestimated by the values of flows at the end of the first iteration.

To check on convergence for iteration 1, use Equation 12.25 and Equation 12.47 for the infinity – norm for the govern-
ing path equation as follows

k + 1 max k + 1
 Q- P  q  =   Q- P  q     stop (12.64)

s
j=1, NF – NC s

where NF – NC = 1, s = 1 and k = 0.

Thus for the two pipe network the check is for the path equation only (from Equation 12.63) as there are no closed
loops

1
 Q- P  q  = – r 1 Q 1 Q 1 – r 2 Q 2 Q 2 +  EL 2 – E L 3  = – 310.403 = 310.403 m (12.65)
1 

The following is not satisfied

1 –6
 Q- P  q    stop where  stop = 1.0 10
1 

Thus the iterative process has not converged.

ITERATION 2, Q – equations

The vector of new flows is given by Equation 12.62 above. The friction factors (based on the Swamee and Jain
equation – Equation 4.58) and corresponding rf values (Equation 4.120) for the flows that resulted from iteration 1 in
Equation 12.62 are computed in Table 12.13.

Table 12.13 The rf values for flows from Iteration 2 for the Q – equations formulation

Pipe Q V D Area (m2) Re**  f L (m) rf value

(m3/s) (m/s)* (m) (mm) (Swamee and Jain) (Equation 4.120)
(Equation 4.58)
1 0.5017 7.098 0.3 0.07069 2114448 0.25 0.01901 1000 646.73
2 0.4517 6.3909 0.3 0.07069 1903716 0.25 0.01904 1000 647.48
–6 2
**  = 1.007 10 m /s at 20C ;

306
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

fL 0.25
Using in the above Table: r f = ------------------ ; f = ----------------------------------------------------------
2  5.74 2
2 gD A log 10 ------------ + ------------

 3.7 D 0.9
Re

or

1
f = 0.01901 (12.66)
0.01904

and

1
rf = 646.50 (12.67)
647.25

The flows from Equation 12.62 may be substituted into the continuity and the path equations of Equation 12.49 and
Equation 12.50 with updates for the resistance terms rf as follows:
1 3
f 1 q  = – Q 1 + Q 2 + 0.050 = – 0.5017 + 0.4517 + 0.050 = 0.0 m /s

1
f 2q  = – r 1 Q 1 Q 1 – r 2 Q 2 Q 2 + 30.0 = – 646.50  0.5017  0.5017 – 647.25  0.4517  0.4517 + 30.0 = – 264.777 m

or the vector for the right – hand side becomes:

1 0.0
fq  = (12.68)
– 264.777

The Jacobian matrix on the second iteration from Equation 12.54 for the flows from iteration 1 (Equation 12.62) and rf
values (Equation 12.67) is:

1 – 1.0 1.0 – 1.0 1.0

Jq  = 1 1 1 1
= (12.69)
– 2.0r f Q 1 – 2.0r f Q 2 – 2.0  646.50  0.5017 – 2.0  647.25  0.4517
1 2

or

1 – 1.0 1.0

Jq  = (12.70)
– 648.69 – 584.72

The inverse of the Jacobian matrix for the second iteration

1
 = – 0.4741 – 0.0008108
–1
J q (12.71)
0.5259 – 0.0008108

The flow corrections for the second iteration are calculated from Equation 12.12 by substituting Equation 12.68 and
Equation 12.71 as follows:

2 1 1

 = – – 04741 – 0.0008108 0.0 = – 0.2148
–1
q = –J  q f  q
0.5259 – 0.0008108 – 264.777 – 0.2148

Note the flow corrections for pipes 1 and 2 are identical as continuity was already satisfied by the flows at the end of
the first iteration.

The new flows after the second iteration are:

2 1 2

= 0.5017 + – 0.2148 = 0.2870 m /s
3
q = q + q
0.4517 – 0.2148 0.2370

The corresponding head loss along the path becomes:

2
f 2q  = – r 1 Q 1 Q 1 – r 2 Q 2 Q 2 + 30.0 = – 646.50  0.2870  0.2870 – 647.25  0.2370  0.2370 + 30.0

2
f 2q  = – 53.251558 – 36.355385 + 30.0 = – 59.6069 m (12.72)

307
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

To check on convergence for iteration 2, use Equation 12.25 and Equation 12.47 for the infinity – norm for the govern-
ing path equation as follows

k + 1 max k + 1
 Q- P  q  =   Q- P  q     stop (12.73)

s
j=1, NF – NC s

where NF – NC = 1, s = 1 and k = 1.

Thus for the two pipe network the check is for the path equation only (from Equation 12.72) as there are no closed
loops

2
 Q- P  q  = – r 1 Q 1 Q 1 – r 2 Q 2 Q 2 +  EL 2 – E L 3  = – 59.6069 = 59.6069 m (12.74)
1 

The following is not satisfied

2 –6
 Q- L  q    stop where  stop = 1.0 10 m
1 

Thus the iterative process has not converged.

The iterations continue up to iteration 9 and are summarized below in Answer Part (v). The convergence check is
shown as the infinity – norm for the path equation as shown in the last column of Table 12.15.
___________________________________________________________________________________
Answer – (iv) Compute the condition number of the Jacobian matrix at the final iteration for the Q – equations
formulation.

ITERATION 9 – FINAL ITERATION, Q – equations

The flows from iteration 9 in Table 12.16 are:

9
= 0.17360 m /s
3
q (12.75)
0.12360

The rf values from iteration 9 (see Table 12.14) are:

9
rf = 658.62 (12.76)
665.28

Table 12.14 The rf values for flows from Iteration 9 for the Q – equations formulation

Pipe Q V D Area (m2) Re**  f L (m) rf value

(m3/s) (m/s)* (m) (mm) (Swamee and Jain) (Equation 4.120)
(Equation 4.58)
1 0.1736 2.456 0.3 0.07069 731,660 0.25 0.01937 1000 658.62
2 0.1236 1.749 0.3 0.07069 520,928 0.25 0.01957 1000 647.28
–6 2
**  = 1.007 10 m /s at 20C ;
0.25 fL
Using in the above Table: f = ---------------------------------------------------------- ; r f = ------------------
  5.74  2 2 gD A
2
log 10 ------------ + ------------
 3.7 D 0.9
Re

The Jacobian matrix for the final iteration from Equation 12.54 for the initial flow guesses (Equation 12.56) and the rf
values (Equation 4.120) is:

9 – 1.0 1.0 – 1.0 1.0

Jq  = 9 9 9 9
= (12.77)
– 2.0r f Q1 – 2.0r f Q 2 – 2.0  658.38  0.17360 – 2.0  665.04 0.12360
1 2

or

9 – 1.0 1.0

Jq  = (12.78)
– 228.59 – 164.39

308
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

The infinity – norm (L – inf or the row – sum norm) of the Jacobian matrix of size NP by NP (where the number of
pipes NP=2) from Equation 10.30 is:

NP
max
J  =  a ij = max   – 1.0 + 1.0   – 228.59 + – 164.39   = max  2 392.98  = 392.98
1  i  NP j=1

The inverse of the Jacobian matrix is given as:

9
 = – 0.4183 – 0.002545
–1
J q (12.79)
0.5817 – 0.002545

The infinity – norm (L – inf) of the inverse of the Jacobian matrix (NP=2) from Equation 10.30 is:

NP
–1 max
J  =  a ij = max   – 0.4183 + – 0.002545   0.5817 + – 0.002545  
1  i  NP j=1

= max  0.4208 0.5842  = 0.5842

The condition number of Jacobian matrix in Equation 12.77 from Equation 10.31 based on the infinity – norm is
–1
cond(J) = J  J  = 392.98  0.5842  = 229.58

The size of the condition number is acceptable given that between two and three significant figures would be lost from
say a total of 15 significant figures for computations using double precision calculations.
___________________________________________________________________________________
Answer – (v) Tabulate all iterations for the solution process for the Q – equations formulation

FINAL SOLUTION – ALL ITERATIONS, Q – equations

The iterations for the full solution based on the Q – equations formulation are shown in Table 12.15 and Table 12.16.
Nine iterations are shown. The stopping tolerance based on the infinity – norm of the loop and path head losses (only
the path head loss is needed for this example) that was used in the computer simulation run for producing Table 12.15
–6
was 1.0 10 m. While this is much smaller than is normally required, it is used to demonstrate the quadratic conver-
gence properties. This behavior is clear after iteration 5 in Table 12.15. Realistically the stopping tolerance could have
–3
been 1.0 10 m or 1 mm which means that convergence would have occurred in 6 iterations.

Table 12.15 Iteration sequence of flow corrections for the Newton solution for the Q – equations formulation
–6
(stopping tolerance 1.0 10 m achieved at iteration 7) for the two pipe network

Iteration Q 1 m3/s Q 2 m3/s V 1 m/s V 2 m/s

1 0.480142 0.430142 6.7926 6.0853

2 – 0.214672 – 0.214672 –3.0370 –3.0370
3 – 0.088311 – 0.088311 –1.2494 –1.2494
4 – 0.023026 – 0.023026 –0.3257 –0.3257
5 – 0.002044 – 0.002044 –0.0289 –0.0289
6* – 4.34E – 05 – 4.34E – 05 –0.0006 –0.0006
7 – 6.38E – 07 – 6.38E – 07 –9.03E – 06 –9.03E – 06
–3
* stopping tolerance of 1.0 10 m or 1 mm achieved in iteration 6

Table 12.16 Iteration sequence of flows for the Newton solution for the Q – equations formulation
–6
(stopping tolerance 1.0 10 m) for the two pipe network

Iteration Q 1 m3/s Q 2 m3/s V 1 m/s V 2 m/s k + 1

 Q- P  q 
1 
(from Equation 12.47) **
0 0.02155 0.02155 0.305 0.305 29.306
1 0.50169 0.45169 7.098 6.390 310.403
2 0.28702 0.23702 4.061 3.353 59.621

309
 Chapter 12 - File: Chap11-Q-form.fm July 6, 2022 Version 7

Table 12.16 Iteration sequence of flows for the Newton solution for the Q – equations formulation
–6
(stopping tolerance 1.0 10 m) for the two pipe network

3 0.19871 0.14871 2.811 2.104 10.179

4 0.17568 0.12568 2.485 1.778 0.6984
5 0.17364 0.12364 2.457 1.749 5.53E-03
6* 0.17360 0.12360 2.456 1.749 2.50E-06
7 0.17360 0.12360 2.456 1.749 5.39E-09
–3
* stopping tolerance of 1.0 10 achieved

** Infinity – norm for path head loss (m) used for testing for convergence

max k + 1
=   Q- P  q  
s=1, NF – NC s

___________________________________________________________________________________
Answer – (vi) Plot convergence characteristics for the iterations for the Q – equations formulation for the two
pipe network.

Figure 12.8 Convergence of flows for solution of the two pipe network based on the Q – equations formulation for the New-
ton method

___________________________________________________________________________________

12.7 References
Lambert, M.F. (1994). “Hydraulics of Compound Channels.” PhD Thesis, Department of Civil Engi-
neering and Surveying, University of Newcastle.

Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1996) FORTRAN numerical recipes
(Vol. 1), Cambridge England.

Rossman, L A. (2000). EPANET 2 users manual. U.S. Environmental Protection Agency, Washington,
D.C., EPA/600/R-00/057.

Todini, E., Pilati, S. (1988). “A gradient algorithm for the analysis of pipe networks.” Proc., Computer
Applications in Water Supply, Research Studies Press, Letchworth, Hertfordshire, UK, 1-20.

310
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

CHAPTER 13 SOLVING THE H – EQUATIONS1

13.1 Introduction
In this Chapter, the head equations or H – equations formulation where the equations are formulated in
terms of unknown heads are solved using the Newton method. There are NJ unknowns in the network
which is usually less than the number of unknown flows (NP) in the network that needed to be solved
for in the Q – equations formulation given in Chapter 12.

13.2 Solving the H – equations for a Water Distribution System

13.2.1 The set of H – equations
The second formulation to be considered is the solution of the head equations. Shamir and Howard
(1968) developed a simultaneous solution technique based on the nodal head equations. A linearization
of these head equations is required to iteratively solve the set of nonlinear equations. In the previous sec-
tion, the flow equations were solved using the Newton technique. Consider now the general set of non-
linear equations for a set of NJ unknown heads at nodes or junctions in a network expressed by the
following set of NJ – functions. These functions are made up of the continuity equations with the flows
substituted in terms of the unknown heads as described in Chapter 9. Their general form of H –
equations is

f 1  H 1 H 2  H NJ  = 0

f 2 H 1 H 2  H NJ  = 0

f NJ  H 1 H 2  H NJ  = 0 (13.1)

where

• f i  H 1 H 2  H NJ  = ith function in the H – equations formulation (a continuity equation

with the flows substituted in terms of unknown heads
• NJ = number of nodes with unknown heads in the network
• H i = head at node i (m or ft)

The unknown heads that are required to be solved for may be expressed as the column vector

T
h =  H 1 H 2  H NJ  (13.2)

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.

311
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

The true solution to the set of nonlinear equations in Equation 13.1 is assumed to be the column vector

* * * * T
h =  H 1 H 2  H NJ  (13.3)

The objective, in the Newton numerical solution technique discussed in the next section, is to find a
*
solution h that is acceptably close to h within a specified tolerance. Let the column vector for the
deviations or corrections for the heads be defined as

T
h =  H 1 H 2  H NJ  (13.4)

This the quantity that will be solved for in the iterative Newton method solution.

*
The relationship between the correct nodal head vector h , the current nodal head vector h , and the
nodal head correction vector h is

*
h = h + h (13.5)

To commence the iterative solution process an initial set of heads at iteration 0 (zero) are required. The
column vector of the initial estimates of the unknown heads is designated as

0 0 0 0 T

h =  H 1  H 2   H NJ  (13.6)

13.2.2 The Jacobian matrix for the H – equations formulation

As for the solution to the flow equations in Section 12.2.2, the solution for the unknown heads requires
linearization of the nonlinear governing equations. Consider the following Taylor series expansion about
the vector h as follows

NJ  f h
1
f 1 h + h  = f 1  h  + - H i + O  H 
 -----------------
H i
i=1

NJ  f h
2
f 2  h + h  = f 2  h  + - H i + O  H 
 -----------------
H i
i=1

NJ f h
NJ
f NJ  h + h  = f NJ  h  +  ---------------------H i + O  H  (13.7)
H i
i=1

where

• f i  h , f i  h + h  = ith function in the H – equations formulation (a continuity equation

with the flows substituted in terms of unknown heads)
• h = column vector of unknown heads for each of the nodes (m or ft)

312
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

• h = deviation vector of the unknown nodal heads; H i = ith deviation of head in the h
vector (m or ft)
• NJ = number of nodes with unknown heads in the network
2
• O  H  = second order and higher terms of the head corrections
Putting Equation 13.7 into matrix notation (Press et al. 1996) gives:

2
f  h + h  = f  h  + J  h h + O  h  (13.8)

where

• J  h  = Jacobian matrix based on the heads

From the set of equations in Equation 13.7, the Jacobian matrix for the unknown head equations formu-
lation based on the heads at iteration k is defined as

k  k  k 
 f 1h   f 1h   f 1h 
------------------------ -  -----------------------
----------------------- -
H 1 H 2 H NJ
k  k  k 
k 
 f 2h   f 2h   f 2h 
h  = ------------------------ -  -----------------------
----------------------- - (13.9)
H 1 H 2 H NJ
   
k  k  k 
 f NJ  h   f NJ  h   f NJ  h 
---------------------------- ----------------------------  ---------------------------
-
H 1 H 2 H NJ

To solve for the corrections to the heads the following formulation is made (in a similar way for the
unknown flow formulation as presented in Section 12.2.2)

2
If h is small then second order and higher terms of O  h  can be dropped and also by setting in
Equation 13.8

*
f  h + h  = f  h  = 0

gives

k  k + 1 k 
Jh h = –f  h  (13.10)

EQUATIONS TO BE SOLVED FOR UNKNOWN HEADS FORMULATION

or

k + 1
H 1 k  k  k 
f 1  H 1  H 2   H NJ 
k + 1
H 2 k  k  k 
k 
f 2  H 1  H 2   H NJ 
Jh   = – (13.11)

 
k  k  k 
k + 1 f NJ  H 1  H 2   H NJ 
H NJ

313
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

The subscripts k and (k + 1) are used to represent to the successive iteration numbers. Equation 13.10 is
k + 1
the equation that is used to solve for the unknown head correction vector of h .

k + 1
The vector of unknown nodal head corrections h can be expressed from Equation 13.10 as

k + 1 –1 k  k 
h = –J  h f  h  (13.12)

All of the values on the right – hand side of Equation 13.12 are based on the head values at iteration k.
As mentioned previously it is very uncommon to actually solve explicitly for the inverse of the Jacobian
in Equation 13.12 and thus it is usual to work with Equation 13.10. This equation is a set of linear equa-
tions and may be solved by an LUD method (lower – upper decomposition) or a sparse matrix technique.

k + 1
Once the corrections h have been determined for an iteration they are then added to the current
k 
solution vector h to give the next best estimate of the unknown heads as

k + 1 k  k + 1
h = h + h (13.13)

or

k + 1
k + 1 k  H 1
H1 H1
k + 1
k + 1 k  H 2
H2 H2
= +  (13.14)
 
  
k + 1 k 
H NJ H NJ k + 1
H NJ

k + 1
The iterative solution process is continued until the vector h converges to within a specified tol-
erance.

Almost any assumed set of estimates of values for the NJ unknown heads may be used to initiate the
iterative solution process. It is best to avoid all zeros and/or heads at the two ends of a pipe being equal.
An upper bound for the initial starting estimates is the elevation of the highest reservoir in the system.
During the iterative process for the unknown heads at nodes, if the heads at the nodes at the end of a pipe
become similar serious problems may occur with regard to the singularity of the coefficient matrix
(Nielsen 1989). As a result, convergence is not guaranteed with either the Newton method or for the lin-
ear theory method for a H – equations formulation.

13.2.3 Derivatives of terms incorporating a modulus of heads

Consider the function taking the derivative of the function f  H  in a single unknown variable flow of
H, then based on Lambert (1994)

 f H   f H  d f H 
------------------------- = ------------------- ----------------- (13.15)
dH f H  dH

or for the derivative of a product

f H  d f H  d f H
f  H  f  H   = f  H  ----------------- ------------------- + f  H  ----------- (13.16)
f  H  dH dH

314
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

13.2.4 Jacobian values and plots of nonlinear head equation terms

Now consider a two variable term involving a modulus or absolute value of a nodal head term based on
Equation 9.32 for a pipe j connecting node i and node k as

1 -H – H  H – H 1  n –
f  H i H k  = --------- i k i k (13.17)
1n
rj

The exponent in Equation 13.17 for the Darcy – Weisbach head loss equation is n = 2 or

1 – 0.
f  H i H k  = ----------  H i – H k  H i – H k (13.18)
1 2
rj

For the Hazen – Williams head loss equation the exponent in Equation 13.17 is n = 1.852 or

1 -–1
------------
1 1.852 1 0.54 –
f  H i H k  = -------------------  H i – H k  H i – H k = -----------  H i – H k  H i – H k
1  1.852 0.54
rj rj

1 – 0.4
f  H i H k  = -----------  H i – H k  H i – H k (13.19)
0.54
rj

The derivative of a term involving a modulus or absolute value with respect to the variable H i in
Equation 13.17 is given as

f 1   1  n – 1  1  n – 1 
------- = ----------  H i – H k   Hi – Hk  + Hi – Hk (H – H k)
Hi rj
1 n  Hi H i i

It then follows using Equation 13.16 for the first term on the right – hand side that

 1  n – 1  1  n – 1
f 1 Hi – Hk   1  n – 1 
-
------- = ----------  H i – H k  -------------------------------------------  H i – H k   + Hi – Hk  1  (13.20)
Hi 1  n  1  n  – 1  H i 
rj Hi – Hk

 1  n – 1  1  n – 1
f 1 Hi – Hk  1  n – 2
------- = ----------  H i – H k  -------------------------------------------
-  1  n – 1 H i – H k  + Hi – Hk (13.21)
Hi 1  n  1  n – 1
rj Hi – Hk

 1  n – 1  1  n – 1
f 1 1 1 1  1  n – 1
-------- = ----------   --- – 1 H i – H k + Hi – Hk = --- ---------- H i – H k
H i 1  n   n  nr1  n
rj j

Thus

f 1
1--- ---------  1  n – 1
--------- = - H – Hk (13.22)
H i n r1 n i
j

DERIVATIVE OF f WITH RESPECT TO H i

For the Darcy – Weisbach head loss equation the derivative in Equation 13.22 is

315
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

f 1
1 ---------  1  n – 1 1 1  1  2 – 1 1 – 0.5
-------- = --- - H – Hk = --- ---------- H i – H k = 0.5 -------- H i – H k (13.23)
H i n r1 n i 2r1  2 rj
0.5
j j

For the Hazen – Williams head loss equation the derivative in Equation 13.22 is

f 1
1--- ---------  1  n – 1 1 1  1  1.852  – 1 1 – 0.46
-------- = - H – Hk = ------------- ------------------- H i – H k = 0.54 ----------- H i – H k
H i n r1 n i 1.852 r 1  1.852 rj
0.54
j j

(13.24)

1  1  n –
Another derivative of a term of the form of f  H i H k  = ----------  H i – H k  H i – H k from
1 n
rj
Equation 13.17 with respect to H k is now considered. It again involves a modulus and is given as

------ 1 - H – H    H – H
f - = ---------  1  n – 1  1  n – 1  H
i k H i k  + Hi – Hk ( – H k)
Hk rj
1n
k H k i

It then follows that

 1  n – 1  1  n – 1
f 1 Hi – Hk   1  n – 1 
-
-------- = ----------  H i – H k  -------------------------------------------  H i – H k   + Hi – Hk  –1 
Hk 1n  1  n – 1   H 
rj Hi – Hk 2

(13.25)

1  n – 1 1  n – 1
f 1 Hi – Hk 1 1  n – 1 – 1
-  –  --- – 1   H i – H k 
---------- = ----------  H i – H k  ------------------------------------------- – Hi – Hk
H k 1  n  1  n  – 1   n  
rj Hi – Hk

(13.26)

1  n – 1 1  n – 1
f 1 1 1 1  n – 1
--------- = – ----------  --- – 1 H i – H k + Hi – Hk = – 1--- ---------- H i – H k (13.27)
H k 1  n n  n r1  n
rj j

Thus

f 1
---------- = – 1--- ---------- H i – H k
1  n – 1
(13.28)
H k n r1  n
j

DERIVATIVE OF f WITH RESPECT TO H k

For the Darcy – Weisbach head loss equation the derivative in Equation 13.28 is

f 1 1 1 1
--------- = – 1--- ---------- H i – H k
 1  n – 1  1  2 – 1 – 0.5
= – --- ---------- H i – H k = – 0.5 -------- H i – H k (13.29)
H k n r1 n 2r1  2 rj
0.5
j j

For the Hazen – Williams head loss equation the derivative in Equation 13.28 is

316
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

f 1 1 1 1
--------- = – 1--- ---------- H i – H k
 1  n – 1  1  1.852  – 1 – 0.46
= – ------------- ------------------- H i – H k = – 0.54 ----------- H i – H k
H k n r1 n 1.852 r 1  1.852 rj
0.54
j j

(13.30)

Example 13.1 Now consider the case where n = 2 for the case of the Darcy – Weisbach head loss equation
Equation 13.17 for unknown heads H 2 and H 3 Figure 13.5 becomes

1  1  n – 1 1 –0
 H 2 H 3  = ----------  H i – H k  H i – H k = ----------  H 2 – H 3  H 2 – H 3 (13.31)
1n 12
rj rj

Compute the derivative of Equation 13.31.

Answer – Computation of the derivative of the nodal head formulation equation/Darcy – Weisbach.

The derivative of Equation 13.31 from Equation 13.17 with n = 2 is

f 1 1  1  n – 1 1 1 – 0.5
---------- = --- ---------- H 2 – H 3 = --- ---------- H 2 – H 3 (13.32)
H 2 n r1  n 2 r1  2
j j

or

f 1 – 0.5
---------- = 0.5 ---------- H 2 – H 3 (13.33)
H 2 rj
12

 f - = – 0.5 ---------
1 - H – H – 0.5
--------- 2 3 (13.34)
H 3 r
12
j

Example 13.2 For the Hazen – Williams head loss equation where n = 1.852 then Equation 13.17 for unknown
heads H 2 and H 3 Figure 13.5 becomes

1
------------- – 1
1 -  H – H  H – H  1  n  – 1 = -----------------
1 -  H – H  H – H 1.852 1 -H – H  H – H – 0
 H 2 H 3  = --------- 2 3 2 3 2 3 2 3 = --------- 2 3 2 3
1n 1  1.852 0.54
rj rj rj

1 0.54 – 1 1 – 0.4
f  H 2 H 3  = ------------------  H 2 – H 3  H 2 – H 3 = ----------  H 2 – H 3  H 2 – H 3 (13.35)
1  1.852 0.54
rj rj

Compute the derivative of Equation 13.35.

Answer – Computation of the derivative of the nodal head formulation equation/Hazen – Williams.

The derivative from Equation for n = 1.852 with respect to H 2 is

1 -–1
------------
f 1 1  1  n – 1 1 1 1.852 1 0.54 – 1
---------- = --- ---------- H 2 – H 3 = ------------- ------------------ H 2 – H 3 = 0.54 ---------- H 2 – H 3
H 2 n r1  n 1.852 r 1  1.852 rj
0.54
j j

 f - = 0.54 ---------
1 - H – H – 0.46
--------- 2 3
H 2 rj
0.54

and

 f - = – 0.54 ---------
1 - H – H – 0.46
--------- 2 3
H 3 r
0.54
j

317
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

Example 13.3 Consider a function where the value of n = 1.852 is taken for the Hazen – Williams head loss equa-
1
tion in Equation 13.17 for unknown heads H 2 and H 3 in Figure 13.5 with ---------- = 3.0 . Thus for the head formula-
1n
rj
tion function is

1 -–
------------
1 1  n – 1 1.852
f  H 2 H 3  = ----------  H i – H k  H i – H k = 3.0  H 2 – H 3  H 2 – H 3 (13.36)
1n
rj

0.54 – 1 – 0.4
f  H 2 H 3  = 3.0  H 2 – H 3  H 2 – H 3 = 3.0  H 2 – H 3  H 2 – H 3 (13.37)

(i) Plot the variation of the function with H 2 for a range of 0 to 60.0 (m or ft) given a value of H 3 = 30.0 (m or ft)

(ii) Compute the value of the function with respect at H 2 = 2.5 (m or ft).

(iii) What is the derivative of this function?

(iv) Compute the derivative with respect to H 2 given a value of H 3 = 30.0 (m or ft) at H 2 = 2.5 (m or ft).

(v) Plot of the derivative of the nodal head function for the range of H 2 = 0 to H 2 = 60.0 (m or ft).

Answer – (i) Plot of the function of the nodal head function or the range of H 2 = 0 to H 2 = 60.0 (m or ft) with
H 3 = 30.0

A plot of the function is shown in Figure 13.1.

– 0.46
Figure 13.1 Plot of function f  H 2 H 3  = 3.0  H 2 – 30.0  H 2 – 30.0

Note that at H 2 = 30.0 (m or ft) the derivative is positive infinity. Contrast this to Figure 12.2 for the derivative of the
modulus term in the Q – equations formulation.
___________________________________________________________________________________

Answer – (ii) Computation of the function for a particular value of H 2 = 2.5 (m or ft)

The function value from Equation 13.37 at H 2 = 2.5 (m or ft) with H 3 = 30.0 (m or ft) is

– 0.46 – 0.46
f  2.5 30.0  = 3.0  H 2 – H 3  H 2 – H 3 = 3.0  2.5 – 30.0  2.5 – 30.0
– 0.46
= 3.0  – 27.5   27.5  = – 17.962 m or ft

This value is confirmed by the computed value shown in Table 13.1.

___________________________________________________________________________________

318
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

Answer – (iii) Determination of the derivative of the heads formulation function.

1 1
The derivative using Equation 13.37 with ---------- = ------------------ = 3.0 and n = 1.852 with H 3 = 30.0 (m or ft) is
1n 1  1.852
rj rj

 f - = 1--- ---------
1 1  n – 1
--------- - H – H3
H 2 n r1  n 2
j

1 -–1
------------
f 1 1 1.852 0.54 – 1 – 0.46
---------- = ------------- ------------------ H 2 – 30.0 =  0.54 3.0 H 2 – 30.0 = 1.62 H 2 – 30.0 (13.38)
H 2 1.852 r 1  1.852
j

A plot of the derivative is shown in Figure 13.2 for a range of values of H 2 = 0 to H 2 = 60.0 (m or ft). Note that at
H 2 = 30.0 (m or ft) the derivative is infinity.

___________________________________________________________________________________
Answer – (iv) Computation of the derivative of the nodal head formulation function at a particular value.

The derivative of the function value at H 2 = 2.5 (m or ft) with H 3 = 30.0 (m or ft) is

f – 0.46 – 0.46
---------- = 1.62 2.5 – 30.0 = 1.62  27.5  = 0.3527
H 2

This value is confirmed by the computed value shown in Table 13.1.

___________________________________________________________________________________

Answer – (v) Plot of the derivative of the nodal head function for the range of H 2 = 0 to H 2 = 60.0 (m or ft).

A plot of the derivative of the function in Equation 13.38 is given in Figure 13.2. A table of the function values and
derivative values are given in Table 13.1.

– 0.46
Figure 13.2 Derivative of function f  H 2 H 3  = 3.0  H 2 – 30.0  H 2 – 30.0 with respect to H 2

319
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

– 0.46
Table 13.1 Function values and derivative values for f  H 2 H 3  = 3.0  H 2 – 30.0  H 2 – 30.0

H2 f  H 2 H 3  Derivative of function H2 f  H 2 H 3  Derivative of function

with respect to H 2 with respect to H 2
(Equation 13.38) (Equation 13.38)
0 – 18.8264 0.3389 30.03 0.4516 8.1290
2.5 – 17.9623 0.3527 30.1 0.8652 4.6721
5 – 17.0612 0.3685 31 3.0000 1.6200
7.5 – 16.1176 0.3868 32.5 4.9205 1.0628
10 – 15.1244 0.4084 35 7.1543 0.7727
12.5 – 14.0722 0.4342 37.5 8.9054 0.6412
15 – 12.9482 0.4661 40 10.4021 0.5617
17.5 – 11.7342 0.5069 42.5 11.7342 0.5069
20 – 10.4021 0.5617 45 12.9482 0.4661
22.5 – 8.9054 0.6412 47.5 14.0722 0.4342
25 – 7.1543 0.7727 50 15.1244 0.4084
27.5 – 4.9205 1.0628 52.5 16.1176 0.3868
29 – 3.0000 1.6200 55 17.0612 0.3685
29.9 – 0.8652 4.6721 57.5 17.9623 0.3527
29.92 – 0.7670 5.1772 60 18.8264 0.3389
29.999999 0.0000 932.2127

___________________________________________________________________________________

Example 13.4 Consider the case where n = 2 for the Darcy – Weisbach head loss equation in Equation 13.17 for
unknown heads H 2 and H 3 in Figure 13.5 becomes

1 -  H – H  H – H  1  n  – 1 = ---------
1 -H – H  H – H –0
 H 2 H 3  = --------- 2 3 2 3 2 3 2 3 (13.39)
1n 12
rj rj

Compute the derivative with respect to H 3 of Equation 13.39.

Answer – Computation of the derivative of the nodal head formulation function (Darcy – Weisbach).

The derivative with respect to H 3 of Equation 13.39 from Equation is

 f - = – 1--- ---------
1 - H – H 1  n – 1
---------
H 3 n r1  n 2 3
j

or

f 1 – 0.5
---------- = – 0.5 ---------- H 2 – H 3
H 3 r
12
j

Example 13.5 For the Hazen – Williams head loss equation with n = 1.852 Equation 13.17 for unknown heads
H 2 and H 3 in Figure 13.5 becomes

1 -  H – H  H – H  1  n  – 1 = -----------------
1 -  H – H  H – H 0.54 – 1 = ---------
1 -  H – H  H – H – 0.
 H 2 H 3  = --------- 2 3 2 3 2 3 2 3 2 3 2 3
1n 1  1.852 0.54
rj rj rj

(13.40)

Determine the derivative with respect to H 3 of Equation 13.40.

Answer – Computation of the derivative of the nodal head formulation function (Hazen – Williams).

320
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

The derivative with respect to H 3 of Equation 13.40 with n = 1.852 from Equation 13.28 is

1--- – 1
f 1 1 n 1
1 ----------------- 0.54 – 1 1 – 0.46
---------- = – --- ---------- H 2 – H 3 = – ------------
- - H2 – H3 = – 0.54 ---------- H 2 – H 3
H 3 n r1  n 1.852 r 1  1.852 rj
0.54
j j

or

f 1 – 0.46
---------- = – 0.54 ---------- H 2 – H 3
H 3 r
0.54
j

___________________________________________________________________________________

Example 13.6 Consider the form of the expression for the flow expressed in terms of the heads from Chapter 9 for-
mulated for the Darcy – Weisbach head loss equation (n = 2) as Equation 9.33

1 – 0.5
Q j = ----------  H i – H k  H i – H k (13.41)
12
rj

Consider part of the nodal head formulation function in Equation 13.17 for unknown heads H 2 and H 3 in
1
Figure 13.5where the value of n = 2 and ---------- = 3.0 is
1n
rj

– 0.
f  H 2 H 3  = 3.0  H 2 – H 3  H 2 – H 3

(i) Compute the value of the function for the value of H 3 = 20.0 (m or ft) given a value of H 2 = 50.0 (m or ft).

(ii) Plot the variation of the function given a value of H 2 = 50.0 (m or ft).

(iii) What is the derivative of this function with respect to H 3 ?

(iv) Compute the derivative for the value of H 3 = 20.0 (m or ft).

(v) Plot both the variation of the derivative with respect to H 3 given a value of H 2 = 50.0 (m or ft).

Answer – (i) Computation of the nodal head formulation function for a particular head value.

The function value with H 2 = 50.0 (m or ft) is

– 0.5
f  50.0 H 3  = 3.0  50.0 – H 3  50.0 – H 3 (13.42)

The function value at H 3 = 20.0 (m or ft) is

– 0.5 – 0.5 3.0  30 

f  50.0 20.0  = 3.0  50.0 – H 3  50.0 – H 3 = 3.0  50.0 – 20.0  50.0 – 20 = ------------------ = 16.432
30

This value is verified in Table 13.2

___________________________________________________________________________________
Answer – (ii) Plot of the variation of the nodal head formulation function.

The plot of the function is shown in Figure 13.3. A table of the function values and derivative values are given in
Table 13.2.
___________________________________________________________________________________
Answer – (iii) Determination of the derivative of the nodal head formulation function.

1
The derivative using Equation 13.28 with ---------- = 3.0 and n = 2 is
1n
rj

f 1 1  n – 1
---------- = – 1--- ---------- H 2 – H 3
H 3 n r 1j  n

321
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

f – 0.5
---------- = – 0.5  3.0  H 2 – H 3
H 3

f – 0.5
---------- = – 1.5 H 2 – H 3
H 3

___________________________________________________________________________________
Answer – (iv) Computation of the derivative of the nodal head formulation function at a particular value.

The derivative of the function value at H 3 = 20.0 (m or ft) with H 2 = 50.0 (m or ft) is

 f - = – 1.5 50.0 – 20.0 – 0.5 = – 1.5  30.0  – 0.5 = – 0.274

---------
H 3

This value is verified in Table 13.2.

– 0.5
Figure 13.3 Plot of function f  50.0 H 3  = 3.0  50.0 – H 3  50.0 – H 3

___________________________________________________________________________________
Answer – (v) Plot of the derivative of the nodal head formulation function.

A plot of the derivative of the function in Equation 13.42 is given in Figure 13.4. A table of the function values and
derivative values are given in Table 13.2. Note that the derivative approaches negative infinity at H 3 = 50.0 .

322
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

– 0.5
Figure 13.4 Derivative of the function f  H 2 H 3  = 3.0  50.0 – H 3  50.0 – H 3 with respect to H 3

Note that at H 2 = 50.0 (m or ft) the derivative is negative infinity.

– 0.5
Table 13.2 Function values and derivative values forf  H 2 H 3  = 3.0  50.0 – H 3  50.0 – H 3

H3 f  H 2 H 3  Derivative of function H3 f  H 2 H 3  Derivative of function with

with respect to H 3 respect to H 3

20.0 16.4317 – 0.2739 49.9999 0.0300 – 150.0000

22.5 15.7321 – 0.2860 50.1 – 0.9487 – 4.7434
25.0 15.0000 – 0.3000 51.0 – 3.0000 – 1.5000
27.5 14.2303 – 0.3162 52.5 – 4.7434 – 0.9487
29.0 13.7477 – 0.3273 55.0 – 6.7082 – 0.6708
31.0 13.0767 – 0.3441 57.5 – 8.2158 – 0.5477
32.5 12.5499 – 0.3586 60.0 – 9.4868 – 0.4743
35.0 11.6190 – 0.3873 62.5 – 10.6066 – 0.4243
37.5 10.6066 – 0.4243 65.0 – 11.6190 – 0.3873
40.0 9.4868 – 0.4743 67.5 – 12.5499 – 0.3586
42.5 8.2158 – 0.5477 70.0 – 13.4164 – 0.3354
45.0 6.7082 – 0.6708 72.5 – 14.2303 – 0.3162
47.5 4.7434 – 0.9487 75.0 – 15.0000 – 0.3000
49.0 3.0000 – 1.5000 77.5 – 15.7321 – 0.2860
49.9 0.9487 – 4.7434 80.0 – 16.4317 – 0.2739
___________________________________________________________________________________

13.3 Solving the H – equations for the Ten Pipe Network

13.3.1 Unknowns and functions for the Newton method
As a example of the formulation of the Newton technique for the H – equations, consider the network
shown in Figure 13.5 whereby the equations are formulated in terms of the 7 unknown heads of H2 to H9
or HGLs (thus NJ = 7 excluding the two non – reservoir nodes).

323
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

The column vector of unknown heads with 7 unknown heads is

T
h =  H 2 H 3 H 4 H 6 , H 7 H 8 H 9  (13.43)

For the 7 unknown heads a 7 by 7 symmetric Jacobian matrix results. The column vector of the correc-
tions to the heads that will be solved for in the Newton iterative scheme (of length 7) from Equation 13.4
at the (k+1)st iteration becomes

k + 1 k + 1 k + 1 k + 1 k + 1 k + 1 k + 1 k + 1 T
h =  H 2  H 3 , H 4  H 6 , H 7 , H 7 , H 9 

(13.44)

A column vector of initial estimates for the 7 unknown heads needs to be made to proceed with the iter-
ative solution process

0 0 0 0 0 0 0 0 T

h =  H 2  H 3  H 4  H 6  H 7  H 8  H 9 (13.45)

Figure 13.5 Unknown H – equations formulation (Hs) for the ten pipe network

Seven nonlinear equations in the seven unknown heads of column vector h in Equation 13.43 were
determined in Chapter 9 for the Darcy – Weisbach head loss equation and are repeated in Table 13.3.
Appropriate designation of function numbers have been added to Table 13.3.

324
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

Table 13.3 The head or H – equations formulation for the ten pipe network (n = 2)
Node Func- Continuity equation at each node in terms of unknown heads
no. tion
2  E L1 – H 2  E L1 – H 2
– 0.5
 H 2 – H 3 H 2 – H 3
– 0.5
 H 2 – H 6 H 2 – H 6
– 0.5
f 1 h  ---------------------------------------------------------------
- + ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 2 =
0.5 0.5 0.5
r1 r2 r5

– 0.5 – 0.5 – 0.5

 H 2 – H 3 H 2 – H 3  H 3 – H 4 H 3 – H 4  H 3 – H 7 H 3 – H 7
3 f 2 h  ---------------------------------------------------------
- + ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 3 =
0.5 0.5 0.5
r2 r3 r6

– 0.5 – 0.5 – 0.5

 H 3 – H 4 H 3 – H 4  EL 5 – H 4  EL 5 – H 4  H 4 – H 8 H 4 – H 8
4 f 3 h  ---------------------------------------------------------- – --------------------------------------------------------------
- + ---------------------------------------------------------
- + DM 4 = 0
0.5 0.5 0.5
r3 r4 r7

– 0.5 – 0.5
 H 2 – H 6 H 2 – H 6  H 6 – H 7 H 6 – H 7
6 f 4  h  – ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 6 = 0
0.5 0.5
r5 r8

– 0.5 – 0.5 – 0.5

 H 3 – H 7 H 3 – H 7  H 6 – H 7 H 6 – H 7  H 7 – H 8 H 7 – H 8
7 f 5 h  ---------------------------------------------------------
- – ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 7 =
0.5 0.5 0.5
r6 r8 r9

– 0.5 – 0.5 – 0.5

 H 4 – H 8 H 4 – H 8  H 7 – H 8 H 7 – H 8  H 8 – H 9 H 8 – H 9
8 f 6 h  ---------------------------------------------------------
- – ---------------------------------------------------------
- + ---------------------------------------------------------
- + DM 8 =
0.5 0.5 0.5
r7 r9 r 10

– 0.5
 H 8 – H 9 H 8 – H 9
9 f 7 h  – ---------------------------------------------------------
- + DM 9 = 0
0.5
r 10

Consider node 2 in Figure 13.5 or f 1 h  in Table 13.3. Breakdown the continuity equation for node 2
into the individual flows into and out of the node as follows.

f 1  h  = – Q 1 + Q 2 + Q 5 + DM 2 = 0 (13.46)

where

– 0.5
 E L1 – H 2  E L1 – H 2
Q 1 = ----------------------------------------------------------------
- (13.47)
0.5
r1

– 0.5
 H 2 – H 3 H 2 – H 3
Q 2 = ----------------------------------------------------------
- (13.48)
0.5
r2

– 0.5
 H 2 – H 6 H 2 – H 6
Q 5 = ----------------------------------------------------------
- (13.49)
0.5
r5

It is better to keep the form of the equation for node 2 shown in Equation 13.46 by computing the corre-
sponding flows separately. In this way, friction factors and hence r j values can be computed initially
and updated during the iterative solution process as indicated in Equation 12.30 and Equation 12.31 for
the Darcy – Weisbach head loss equation. The latest resistance r – values  r 1 r 2  r 10  for the
Darcy – Weisbach head loss equation for each pipe that will be used for computing terms in both
Table 13.3 and Table 13.4 (n=2) may be computed as shown in Section 12.4.2.

325
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

The other functions in Table 13.3 for nodes 3 to 9 (excluding node 5) would be computed in a similar
way as shown above in Equation 13.46 by computing individual flows in individual pipes for the updat-
ing of the r j values.

13.3.2 Jacobian matrix for the head equations for the ten pipe network
The Jacobian matrix of size 7 by 7 is symmetric (where the columns are associated with each unknown
head of H2, H3, H4, H6, H7, H8 and H9) and given from Equation 13.9 based on the unknown heads at
iteration k becomes

k  k  k 
 f 1h   f 1h   f 1h 
------------------------ ------------------------  -----------------------
-
H 2 H 3 H 9
k  k  k 
 f 2h   f 2h   f 2h 
k  ------------------------ ------------------------  -----------------------
-
h  = H 2 H 3 H 9 (13.50)

   
k  k  k 
 f 10  h   f 10  h   f 10  h 
-------------------------- --------------------------  --------------------------
H 2 H 3 H 9

The first function in the H – equations formulation for node 2 for the Darcy – Weisbach head loss equa-
tion (n = 2) from Table 13.3 is

– 0.5 – 0.5 – 0.5

 E L1 – H 2  E L1 – H 2  H 2 – H 3 H 2 – H 3  H 2 – H 6 H 2 – H 6
1 h  = – ----------------------------------------------------------------
0.5
- + ----------------------------------------------------------
0.5
- + ----------------------------------------------------------
0.5
- + DM 2 =
r1 r2 r5

(13.51)

The partial derivative of the first term with respect to H 2 of Equation 13.51 is based on the general
expression for the derivative from Equation 13.28 of

 1 1  n – 1 1 – 0.5
----------  H – H k  H i – H k = – ------------ H i – H k (13.52)
 H k r1  n i 2r j
0.5
j

The partial derivative of the second term with respect to H 2 on the right – hand side of Equation 13.51 is
based on the general expression for the derivative from Equation 13.28 of

 1 1  n – 1 1 – 0.5
----------  H – H k  H i – H k = ------------ H i – H k (13.53)
 H i r1  n i 2r j
0.5
j

The first element in row 1 of the Jacobian matrix is obtained by differentiating Equation 13.51 with
respect to H2 using Equation 13.52 and Equation 13.53 based on H i values at iteration k with n = 2 as
follows

k 
f 1h  1  k  – 0.5 1 k   k  – 0.5 1 k   k  – 0.
---------------------
- = ------------ E L 1 – H 2 + ------------ H 2 – H 3 + ------------ H 2 – H 6 (13.54)
H 2 2r 1
0.5
2r 2
0.5
2r 5
0.5

326
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

The other Jacobian elements for node 2 (function 1) in row 1 of the matrix (differentiated with respect to
H 3 and H 6 respectively) and the Jacobian elements based on the other functions in Table 13.3 (func-
tions 2 to 7) are obtained in a similar way to described above. Note that the Jacobian matrix is symmet-
ric for the H – equations formulation and as such, only the elements along the diagonal of the matrix and
above the diagonal in the Jacobian need be calculated. Once again, as pointed out previously, the Jaco-
bian matrix for the H – equations formulation is symmetric. This is in contrast to the Jacobian matrix for
the Q – equations formulation that has a non – symmetric Jacobian matrix.

The non – zero terms in the Jacobian matrix of Equation 13.50 for the head equations for the ten pipe
network at iteration k are shown in Table 13.4.

Table 13.4 The non – zero Jacobian elements for the H – equations for the ten pipe network
Node Deriv- Jacobian element
no. ative
2 f 1 - E L – H  k  –0.5 + ----------
--------- 1 - H  k  – H  k  –0.5 + ----------
1 - H  k  – H  k  –0.5
----------1 0.5 1 2 0.5 2 3 0.5 2 6
H 2 2r 1 2r 2 2r 5

2 f 1 k   k  – 0.5
– ----------- H 2 – H 3 f 1 k   k  – 0.5
– ----------- H 2 – H 6
----------1 0.5 ----------1 0.5
H 3 2r 2 H 6 2r 5

3 f 1
– ----------- H 2 – H 3
– 0.5 f 1
– ----------- H 3 – H 4
– 0.5
----------2 0.5 ----------2 0.5
H 2 2r 2 H 4 2r 3

3 f 1
----------- H 2 – H 3
– 0.5 1 - H – H –0.5 + ----------
+ ---------- 1 - H – H –0.5
----------2 0.5 0.5 3 4 0.5 3 7
H 3 2r 2 2r 3 2r 6

3 f 1
----------- H 3 – H 7
– 0.5
----------2 0.5
H 7 2r 3

4 f 1 - H – H –0.5 f 1 - H – H –0.5
----------3 – ----------
0.5 3 4 ----------3 – ----------
0.5 4 8
H 3 2r 3 H 8 2r 7

4 f 1 - H – H –0.5 + ---------- 1 - H – H –0.5

1 - EL – H –0.5 + ----------
----------3 ----------
0.5 3 4 0.5 5 4 0.5 4 8
H 4 2r 3 2r 4 2r 7

6 f 1 - H – H –0.5
– ---------- f 1
– ----------- H 6 – H 7
– 0.5
----------4 0.5 2 6 ----------4 0.5
H 2 2r 5 H 7 2r 8

6 f 1
----------- H 2 – H 6
– 0.5 1
+ ----------- H 6 – H 7
– 0.5
----------4 0.5 0.5
H 6 2r 5 2r 8

7 f 1 – 0.5 f 1 – 0.5 f 1
– ----------- H 7 – H 8
– 0.5
----------5 – ----------- H 3 – H 7 ----------5 ----------- H 6 – H 7
0.5 ----------5 0.5
H 3 2r 6
0.5 H 6 2r 9 H 8 2r 8

7 f 1
----------- H 3 – H 7
– 0.5 1
+ ----------- H 6 – H 7
– 0.5 1
+ ----------- H 7 – H 8
– 0.5
----------5 0.5 0.5 0.5
H 7 2r 6 2r 8 2r 9

8 f 1 – 0.5 f 1 – 0.5
----------6 – ----------- H 4 – H 8
0.5 ----------6 – ----------- H 7 – H 8
0.5
H 4 2r 7 H 7 2r 9

8 f 1 – 0.5 1 - H – H –0.5 + ----------

1 - H – H –0.5
----------6 ----------- H 4 – H 8
0.5
+ ----------
0.5 7 8 0.5 8 9
H 8 2r 7 2r 9 2r 10

8 f 1 – 0.5
----------6 – ----------- H 8 – H 9
H 9 2r 10
0.5

9 f 1 – 0.5 f 1 – 0.5
----------7 – ----------- H 8 – H 9
0.5 ----------7 ----------- H 8 – H 9
0.5
H 8 2r 10 H 9 2r 10

327
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

13.3.3 Solving for the head corrections for the ten pipe network
k + 1
The set of linearized equations for the nodal head corrections h (a column vector of length 7) to
be solved for during the iterative process is

k + 1 k 
H 2 f 1h 
k + 1 k 
H 3 f 2h 
k + 1 k 
H 4  9  f 3h 
k 
Jh  H  k + 1  = – f  h  k   (13.55)
6 4
k + 1 k 
H 7 f 5h 
k + 1 k 
H 8 f 6h 
k + 1 k 
H 9 f 7h 

or

k  k + 1 k 
Jh h = –f  h  (13.56)

k + 1
The corrections h are solved for Equation 13.56 then added to the current solution vector to give
the next best estimate of the unknowns as

k + 1 k  k + 1
h = h + h (13.57)

13.4 The Two Pipe Network Solving for Unknown Heads

Now consider the head equations for the network in Figure 13.6. Only the head at the central node needs
to be solved for, that is H1.

13.4.1 Unknowns and initial guesses for the Newton solution

From Equation 13.2, the unknown head at the kth iteration for the two pipe network is

k  k  T
h = H1  (13.58)

There is only one unknown head in Figure 13.6 and the one function to be solved. This function is that
includes the continuity equation at node 1 with the flows substituted for in terms of the heads at the ends
of each of the pipes (see Equation 13.62 in Example 13.7 below).

Figure 13.6 Unknown H to be solved for in the two pipe network

The column vector of head corrections in the pipes (from Equation 13.4) to be solved for by the Newton
method for the two pipe network at the (k + 1)st iteration becomes

328
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

k + 1 k + 1 T
h =  H 1  (13.59)

An initial estimate for the head at node 1 needs to be made in order for the iterative solution process to
begin

0 0 T
h = H1  (13.60)

As well as calculating the unknown head, the pipe flows actually need to be calculated so that the fric-
tion factors in the Darcy – Weisbach head loss equation can be updated. As for the Q – equations formu-
lation, the friction factors for the first iteration can be based on initial flows for a flow velocity of 1.0 fps
(0.3048 m/s) as shown in Example 13.7. The pipe resistance terms for the two pipes ( r f =  r 1 r 2  ) can
2
be computed based on initial velocity guesses with the Darcy – Weisbach head loss equationh f = r f Q
j

using Equation 12.30.

13.4.2 Two pipe example based on Newton method (H – equations formula-

tion)

Example 13.7 For the two pipe network in Figure 13.6 do the following:

(i) Develop the H – equations formulation for the two pipe network system.
(ii) Verify that the EPANET head values from Table 13.5 satisfy the H – equations formulation.
(iii) Develop the Jacobian matrix for the H – equations formulation and carry out the first two iterations for the two
pipe network.
(iv) Tabulate all iterations for the solution process for the H – equations for the two pipe network.
(v) Compute the condition number of the Jacobian matrix at the final iteration and hence assess the accuracy of your
result for the H – equations.
(vi) Plot convergence characteristics for the iterations for the H – equations for the two pipe network.
Assume the Darcy – Weisbach head loss equation is used for each pipe.

Answer – (i) Develop the H – equations formulation for the two pipe network system.

The H – equations formulation of the governing equations for this two pipe network includes the continuity equation at
node 1 with the discharges substituted in terms of unknown heads at the ends of the pipes as follows:

f 1  q  = – Q 1 + Q 2 + DM 1 = 0 at node 1 (13.61)

– 0.5 – 0.5
 E L2 – H 1  E L2 – H 1  H 1 – E L3  H 1 – E L3
f 1  h  = – ---------------------------------------------------------------
- + ---------------------------------------------------------------
- + DM 1 = 0 at node 1 (13.62)
0.5 0.5
r1 r2

Note that the “constants” for the two terms for rf in Equation 13.62 are dependent on the friction factor which in turn
requires computation of flows and hence are required to be updated at each iteration. In addition, the structure of these
equations Equation 13.61 and Equation 13.62 are important. The flows are required during the iterative process to
update the resistance terms (when the Darcy – Weisbach head loss equation is used) so it is best to compute
Equation 13.62 in parts so that the flows are remembered. These flows can then be used to update the Darcy –
Weisbach friction factors for the computation of the resistance terms, rj.
___________________________________________________________________________________
Answer – (ii) Verify that the EPANET head values from Table 13.5 satisfy the H – equations formulation.

The EPANET value for head at node 1 from Table 13.5 and the rf values from Table 12.9 are substituted into
Equation 13.62 as follows:

329
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

– 0.5 – 0.5
 80.0 – 60.159  80.0 – 60.159  60.159 – 50.0  60.159 – 50.0
f 1  h  = – ---------------------------------------------------------------------------------- + ---------------------------------------------------------------------------------- + 0.050 (13.63)
0.5 0.5
658.64 665.44
3
= – 0.173563 + 0.123558 + 0.050 = – 0.000005 m /s

Note that the nodal head equation is satisfied to a high degree of accuracy to within 0.005 L/s and represents an error of
0.01% of the demand of 50 L/s at node 1. In addition, the flows have been calculated in Equation 13.63 for pipes 1 and
2 as:
3
Q 1 = 0.1736 m /s

3
Q 2 = 0.1236 m /s

___________________________________________________________________________________
Answer – (iii) Develop the Jacobian matrix for the H – equations formulation and carry out the first two itera-
tions for the two pipe network.

k + 1
The equation to be solved iteratively for a sequence of h values from Equation 13.56 is:

k  k + 1 k 
Jh h = –f  h  (13.64)

The one term in the Jacobian at the kth iteration can be determined by taking the appropriate partial derivative of
Equation 13.63 as follows:

k   k  – 0.5 k  – 0.5
 = -----------------------
f 1  h  = 0.5 80.0 – H 1 0.5 H 1 – 50.0
k 
Jh - ----------------------------------------------- + -----------------------------------------------
- (13.65)
H 1 r1
0.5
r2
0.5

For the right – hand side of Equation 13.64, substituting the known quantities into Equation 13.62 for each iteration
gives:

k   k  – 0.5 k  k  – 0.5
k   80.0 – H 1  80.0 – H 1  H 1 – 50.0  H 1 – 50.0
f 1h  = – ------------------------------------------------------------------------
- + ------------------------------------------------------------------------
- + 0.050 (13.66)
0.5 0.5
r1 r2

ITERATION 1, H – equations

To use Equation 13.65 and Equation 13.66 values of the resistance terms rj are needed. These depend on the Darcy –
Weisbach friction factor, however, the flows and hence the velocities are not known. Thus it is reasonable to make a
guess of the velocities in each pipe. Assume these velocities are initially 1.0 fps (0.3048 m). These velocities will be
updated after each iteration.

The rf values for velocities of 1.0 fps (0.3048 m/s) from Table 12.12 (page 305) are:

r f = 746.95
746.95

Table 13.5 The rf values for initial flow guesses based on 1.0 fps (0.3048 m/s) for H – equations formulation

Pipe V D Area (m2) Re**  f L (m) rf value

(m/s)* (m) (mm) (Swamee and Jain) (Equation 4.123)
(Equation 4.61)
1 0.3048 0.3 0.07069 90,828 0.25 0.02197 1000 746.954
2 0.3048 0.3 0.07069 90,828 0.25 0.02197 1000 746.954
–6 2
*Computed based on 1.0/3.28 for conversion of fps to m. **  = 1.007 10 m /s at 20C ;

fL
Using in the above Table: ; r f = ------------------
2
2 gD A

The initial starting head at node 1 is taken as the average of the two reservoir levels:

0
H1 = 65.0 m (13.67)

330
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

0
This is also the vector of unknowns h at iteration zero (0) as there is only one unknown head.

The head from Equation 13.67 may be substituted into the nodal head equation of Equation 13.66 as follows:
– 0.5 – 0.5
0  80.0 – 65  80.0 – 65 - + ------------------------------------------------------------
 65 – 50.0  65 – 50.0 - + 0.050
f 1h  = – ------------------------------------------------------------
0.5 0.5
746.95 746.95
3
= – 0.1417 + 0.1417 + 0.050 = 0.050 m /s

Thus the imbalance in the nodal head equation is 50.0 L/s. In addition, the flows have been calculated in pipes 1 and 2
as:
3
Q 1 = 0.1417 m /s

3
Q 2 = 0.1417 m /s

Note that unlike for the Q – equations formulation, the continuity of flow at the junction is not automatically satisfied
for all the iterations after the first iteration for this H – equations formulation.

The Jacobian matrix for the first iteration from Equation 13.65 is:

0 – 0.5 – 0.5

Jh 0.5 80.0 – 65.0
 = --------------------------------------------- 0.5 65.0 – 50.0
- + ---------------------------------------------- = 0.009447 (13.68)
0.5 0.5
746.95 746.95

The inverse of the Jacobian matrix is shown below for this simple two pipe network (note that this would never be
actually calculated in practice for a network of normal size) but in this case is very straight forward for a single number,
as it is the reciprocal of the value in Equation 13.68 as follows:

–1 0 1
J h  = ---------------------- = 105.85 (13.69)
0.009447

The head corrections for the first iteration are calculated from Equation 13.12 as:

1 –1 0 0

h = –J h f  h  = – 105.85 0.050 = – 5.2925 m (13.70)

The new head at node 1 after the first iteration is:

1 0 1

h = h + h = 65.0 – 5.2925 = 59.708 m (13.71)

or

1
H1 = 59.708 m (13.72)

To check on convergence for iteration 1, use Equation 10.21 and Equation 10.22 for the infinity – norm for the head
changes as follows

k + 1 max k + 1
h  =  H    stop (13.73)
j=1, NJ

where NJ = 1 and k = 0.

Thus for the two pipe network the check is for the head at node 1 only (from Equation 10.21)

1 1
h  = H1 = – 5.2925 = 5.2925 m (13.74)

The following is not satisfied

1 –6
h    stop where  stop = 1.0 10 m

Thus the iterative process has not converged.

The head for the first iteration from Equation 13.71 may now be substituted into the nodal head equation of
Equation 13.66 (without updating the friction factors and the corresponding rf values) as follows:

331
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

– 0.5 – 0.5
1  80.0 – 59.708  80.0 – 59.708  59.708 – 50.0  59.708 – 50.0
f 1h  = – ---------------------------------------------------------------------------------- + ---------------------------------------------------------------------------------- + 0.050 (13.75)
0.5 0.5
746.95 746.95
3
= – 0.1648 + 0.1140 + 0.050 = – 0.0008 m /s

Thus the imbalance in the flow at node 1 is 0.8 L/s (compared to the previous value of 50 L/s), and the head is now
overestimated at the end of the first iteration. In addition, the flows have been calculated in pipes 1 and 2 as:
3
Q 1 = 0.1648 m /s

3
Q 2 = 0.1140 m /s

ITERATION 2, H – equations

The new head is given by Equation 13.72 and the corresponding flows are given by Equation 13.75. The friction fac-
tors and corresponding rf values for the flows that resulted from iteration 1 can now be computed as shown in
Table 13.5.

Table 13.6 The rf values for flows from Iteration 2 for the Q – equations formulation

Pipe Q V D Area (m2) Re**  f L (m) rf value

(m /s) 3 (m/s)* (m) (mm) (Swamee and Jain) (Equation 4.123)
(Equation 4.61)
1 0.1648 2.331 0.3 0.07069 694,571 0.25 0.01940 1000 659.52
2 0.1140 1.613 0.3 0.07069 480,468 0.25 0.01962 1000 667.15
–6 2
**  = 1.007 10 m /s at 20C ;
fL 0.25
Using in the above Table: r f = ------------------ ; f = ----------------------------------------------------------
2 gD A
2
log 10  ----------- - + ------------
5.74  2
 3.7 D 0.9
Re

Thus

1
f = 0.01940 (13.76)
0.01962

and

1
rf = 659.52 (13.77)
667.15

The head from Equation 13.67 may be substituted into the nodal head equation of Equation 13.66 as follows:
– 0.5 – 0.5
1  80.0 – 59.708  80.0 – 59.708  59.708 – 50.0  59.708 – 50.0
f 1h  = – ---------------------------------------------------------------------------------- + ---------------------------------------------------------------------------------- + 0.050
0.5 0.5
659.52 667.15
3
= – 0.17541 + 0.1206 + 0.050 = – 0.00481 m /s

In addition, the flows have been calculated in pipes 1 and 2 as:

3
Q 1 = 0.1018 m /s

3
Q 2 = 0.1871 m /s

Thus the imbalance in the nodal head equation is 4.7 L/s.

The Jacobian matrix for the second iteration from Equation 13.65 is:

1 – 0.5 – 0.5

Jh 0.5 80.0 – 59.708
 = --------------------------------------------------- 0.5 59.708 – 50.0
- + ---------------------------------------------------- = 0.01054 (13.78)
0.5 0.5
659.52 667.15

The inverse of the Jacobian matrix is shown below for this simple two pipe network as follows:

332
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

–1 1 1
J h  = ------------------- = 94.92 (13.79)
0.01054

The head corrections for the second iteration are calculated from Equation 13.12 as:
2 –1 1 1
h = –J  h f  h  = – 94.92 – 0.004784 = 0.4541 m

To check on convergence for iteration 2, use Equation 10.21 and Equation 10.22 for the infinity – norm for the head
changes as follows

k + 1 max k + 1
h  =  H    stop (13.80)
j=1, NJ

where NJ = 1 and k = 1.

Thus for the two pipe network the check is for the head at node 1 only (from Equation 10.21)

2 2
h  = H1 = 0.4541 = 0.4541 m (13.81)

The following is not satisfied

1 –6
h    stop where  stop = 1.0 10 m

Thus the iterative process has still not converged.

The new head at node 1 after the second iteration is:

2 1 2

h = h + h = 59.708 + 0.4541 = 60.162 m (13.82)

or

2
H1 = 60.162 m (13.83)

The iterations continue up to iteration 5 and are summarized below in part (iv).

___________________________________________________________________________________
Answer – (iv) Tabulate all iterations for the solution process for the H – equations formulation for the two pipe

network.

FINAL SOLUTION – ALL ITERATIONS, H – equations

The iterations for the full solution based on the H – equations formulation are shown in Table 13.7. Five iterations are
shown. The stopping tolerance based on the infinity – norm of the head changes (the maximum allowable size of any
–6
change in heads from one iteration to the next) that was used in the computer simulation run was 1.0 10 m. Again
this value of stopping tolerance is much smaller than is normally required, but it is used to demonstrate the quadratic
–3
convergence properties of the Newton method. Realistically the stopping tolerance could have been 1.0 10 m or 1
mm which means that convergence would have occurred in 4 iterations.

Table 13.7 Iteration sequence for the Newton solution for the H – equations formulation for the two pipe network

Iteration H1 m H 1 m k + 1 Q1  Q2 
h 
(from Equation 10.21) ** (m3/s) (m3/s)
0 65.0 0.1417 0.1417
1 59.708 – 5.2925 5.2925 0.1648 0.1140
2 60.1616 0.4541 0.4541 0.1734 0.1234
3 60.1584 – 3.180E – 03 3.180E – 03 0.1736 0.1236
4* 60.1583 – 8.385E – 05 8.385E – 05 0.1736 0.1236
5 60.1583 – 7.817E – 07 7.817E – 07 0.1736 0.1236
–3
* A stopping tolerance of less than 1.0 10 m or 1 mm achieved at iteration 4

333
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

max k + 1
** Infinity – norm for head changes (m) used for testing for convergence =  Hi 
i=1, NJ

 Flow values from the first iteration onwards are computed as part of the iterative H – equation process.

___________________________________________________________________________________
Answer – (v) Compute the condition number of the Jacobian matrix at the final iteration and hence assess the
accuracy of your result for the H – equations.

ITERATION 5 – FINAL ITERATION, H – equations

The Jacobian and inverse of the Jacobian for the last iteration is:

6
Jh  = 0.010456 (13.84)

and

–1 6
J h  = 95.639 (13.85)

The infinity – norm (L – inf or the row – sum norm) of the Jacobian matrix of size NJ by NJ (where the number of nodes
NJ=1) from Equation 13.84 is:

NJ
max
 a ij = max  0.010456  = 0.010456
1  i  NJ j=1

The infinity – norm (L – inf) of the inverse of the Jacobian matrix from Equation 13.85 is:

NJ
–1 max
J  =  a ij = max  95.639  = 95.639
1  i  NJ j=1

The condition number of Jacobian matrix in Equation 13.85 from Equation 10.25 based on the infinity – norm is:
–1
cond(J) = J  J  = 0.010456  95.639  = 1.0

as would be expected for a Jacobian with one matrix element.

The size of the condition number is clearly acceptable given that one place of accuracy would be lost from say a total of
15 significant figures for computations using double precision calculations.

___________________________________________________________________________________
Answer (vi) Plot convergence characteristics for the iterations for the H – equations for the two pipe network.

Figure 13.7 Convergence of H – equations formulation for solution of the two pipe network

334
 Chapter 13 - File: Chap12-H-form.fm July 6, 2022 Version 7

13.5 References
Nielsen, H. B. (1989). “Methods for analyzing pipe networks.” Journal of Hydraulic Engineering,
Vol.115, No.2, February, 139-157.

Lambert, M.F. (1994). “Hydraulics of Compound Channels.” PhD Thesis, Department of Civil Engi-
neering and Surveying, University of Newcastle.

Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1996) FORTRAN numerical recipes
(Vol. 1), Cambridge England.

Shamir, U. and D. D. Howard (1968). “Water distribution systems analysis.” Journal of the Hydraulics
Division, Proceedings of the ASCE, Vol. 94, No. HY1, January, 219-235.

335
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

CHAPTER 14 SOLVING THE LOOP FLOW EQUATIONS1

14.1 Introduction
In this Chapter, the loop flow equations or LF – equations formulation are solved for based on the New-
ton method.These equations are formulated in terms of unknown loop flows. An initial estimate of all
the flows in the network such that continuity is satisfied at every node is required.

14.2 The Loop Flow (LF) Equations for a Water Distribution System
14.2.1 Set of loop flow equations
There are NL +  NF – NC  unknowns for this formulation where these variables are defined in
Section 9.2.2.

The number of equations for this formulation is usually significantly less than either the number of flow
equations (NP equations or the number of pipes — Q values as unknowns) or the number of nodal equa-
tions (NJ equations or the number of nodes with unknown head — H values as unknowns). Epp and
Fowler (1970) improved the single loop adjustment method of Hardy Cross (1936) by applying a New-
ton solution technique to computing all loop flows and required independent path flows (LFs) simulta-
neously. The convergence characteristics were improved significantly.

The expression for the number of nonlinear equations for loop flows and required independent path
flows was developed in Chapter 9.

The general set of equations to be solved for the NTL loop flows for a network are

f 1  LF 1 LF 2  LF NTL  = 0

f 2  LF 1 LF 2  LF NTL  = 0

f NTL  LF 1 LF 2  LF NTL  = 0 (14.1)

where

• f s  LF 1 LF 2  LF NTL  = sth function in the loop flow equations (an energy equation)
• LF s = loop flow in a loop or required independent path s (m3/s or ft3/s)
• NTL = number of total closed simple loops plus required independent paths
(NTL = NL +  NF – NC  )

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.

336
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

The column vector of unknown loop flows that are required to be solved for is defined as:

T
u =  LF 1 LF 2  LF NTL  (14.2)

The true solution to this set of nonlinear equations in Equation 14.1 is assumed to be the following col-
umn vector

* * * * T
u =  LF 1 LF 2  LF NTL  (14.3)

Let a column vector of the loop flow corrections problem be defined as

T
u =  LF 1 LF 2  LF NTL  (14.4)

This is the quantity that will be solved for successively in the Newton iterative process.

Thus the relationship between the correct loop flow vector u * , the current loop flow vector u , and the
loop flow correction vector u is

u * = u + u (14.5)

An initial estimate (referred to as the zeroth iteration indicated by the superscript zero (0)) needs to be
made for each of the unknown variables. The column vector of initial estimates of the loop flows is

0 0 0 0 T

u =  LF 1  LF 2   LF NTL  (14.6)

A column vector of initial estimates for the pipe flows needs to be selected such that the flows selected
satisfy all the continuity equations at all junctions (see Figure 14.2)

init init init init T

q =  Q 1  Q 2   Q NTL  (14.7)

The starting vector should be selected as a value that is considered to be likely e.g. 0.010 m3/s or 10 L/s.
It is probably desirable to select the initial loop flow to be different e.g. 0.010, 0.015, 0.020 etc. rather
than being identical.

14.2.2 The Jacobian matrix for the LF – equations formulation

Equation 14.1 may be expanded in a Taylor series as shown earlier in Equation 10.8

NTL  f u
1 2
f 1 u + u  = f 1  u  + - LF s + O  LF 
 -----------------
 LF  s
s=1

NTL  f u
2 2
f 2 u + u  = f 2  u  + - LF s + O  LF 
 -----------------
 LF  s
s=1

337
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

NTL  f u
NTL
f NTL u + u  = f NTL  u  + - LF s + O  LF 
 ------------------------ (14.8)
 LF  s
s=1

where

• f s  u , f s u + u  = sth function in the LF – equations formulation (either a loop or a

required independent path equation)
• u = column vector of unknown loop flows m3/s or ft3/s)
• u = deviation vector of the unknown loop flows; u s = sth deviation of loop flow in the
vector m3/s or ft3/s)
• NTL = number of total closed simple loops plus required independent paths
2
• O  LF  = column vector of second order and higher terms of the loop flow corrections
Putting Equation 14.8 into matrix notation (Press et al. 1996) gives

2
f  u + u  = f  u  + J  u u + O  u  (14.9)

where

• J  u  = Jacobian matrix based on the loop flows

The Jacobian matrix for the loop flows at iteration k is defined as

k  k  k 
 f 1u   f 1u   f 1u 
------------------------ ------------------------  ------------------------
-
 LF  1  LF  2  LF  NTL
k  k  k 
k 
 f 2u   f 2u   f 2u 
u  = ------------------------ ------------------------  ------------------------
- (14.10)
 LF  1  LF  2  LF  NTL
   
k  k  k 
 f NTL  u   f NTL  u   f NTL  u 
------------------------------- -------------------------------  ------------------------------
-
 LF  1  LF  2  LF  NTL

When the unknown loop flow correction term of u is small then it follows that both

2
1. second order and higher terms of O  u  can be dropped and
2. also the following can be set in Equation 14.9

f  u + u  = f  u *  = 0

It then follows that

k  k + 1 k 
Ju u = –f  u  (14.11)

EQUATIONS TO BE SOLVED FOR THE LOOP FLOW FORMULATION

or

338
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

k + 1
LF 1 k  k  k 
f 1  LF 1  LF 2   LF NTL 
k + 1
LF 2 k  k  k 
k 
f 2  LF 1  LF 2   LF NTL 
Ju   = – (14.12)

 
k  k  k 
k + 1 f NTL  LF 1  LF 2   LF NTL 
LF NTL

The subscripts k and (k + 1) have been introduced to represent to the successive iteration numbers. The
k + 1
vector of loop flow corrections u on the left – hand side of Equation 14.11 is for the (k + 1)st iter-
k  k  k 
ation. The Jacobian matrix J  u  , the function vector f  u  and the loop flows vector u are all at
iteration k. Once again, as pointed out previously, the Jacobian matrix for the LF – equations formulation
is symmetric. This is in contrast to the Jacobian matrix for the Q – equations formulation.

k + 1
The vector of loop flow corrections u may be expressed from Equation 14.8 as

k + 1 –1 k  k 
u = –J  u f  u  (14.13)

where

–1 k 
• J  u  = inverse of the Jacobian matrix for the LF – equations formulation based on the
loop flows at iteration k (the Jacobian matrix is defined by Equation 14.10)
As mentioned previously, Equation 14.13 is not usually solved. It is desirable to avoid actually comput-
k + 1
ing the Jacobian explicitly. Instead, Equation 14.11 is solved for the vector u of the loop flow
corrections. This equation is a set of linear equations and may be solved by an LUD method (lower –
upper decomposition) or sparse matrix technique.

k + 1 st
The vector of loop flow corrections u at the  k + 1  iteration from Equation 14.11 are then
k 
added to the current solution vector u to give the next best estimate of the unknown loop flows as

k + 1 k  k + 1
u = u + u (14.14)

or

k + 1
k + 1 k  LF 1
LF 1 LF 1
k + 1
k + 1 k  LF 2
LF 2 LF 2
= +  (14.15)
 
  
k + 1 k 
LF NTL LF NTL k + 1
LF NTL

k + 1
The process is iterated until the vector u converges to within an acceptable tolerance.

339
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

14.2.3 Derivatives of terms incorporating a modulus term

Computation of the Jacobian elements in Equation 14.10 require consideration of terms containing a
modulus sign. Consider a pipe j that is part of two loops – loop s and loop t. A general term in the loop
flow governing equations for pipe j for the values at iteration k is

k  n–
g 1 u  = r  Q init + LF  k  – LF  k   Q init + LF  k  – LF  k  (14.16)
j j s t j s t

The partial derivative of Equation 14.17 with respect to LF s (the loop flow for loop s) is based on the
general expression for the derivative from Equation 13.18 adapted to be in terms of LF s as follows

 k  init k  k  n – 1
 g 1 u   = nr j Q j + LF s – LF t (14.17)
  LF  s

The partial derivative of Equation 14.16 with respect to LF t (the loop flow for loop t) is based on the

general expression for the derivative Equation 13.28 adapted to be in terms of LF t as follows

 k  init k  k  n – 1
 g 1 u   = – n r j Q j + LF s – LF t (14.18)
  LF  t

14.2.4 Updating the Darcy – Weisbach friction factors

Once the new vector of LF corrections for an iteration has been computed the flows in each pipe may be
calculated

k + 1 init
Qj = Qj +  b sj LF s (14.19)
s  PL j

where PL j = indices of the loops s associated with pipe j and b sj = – 1 or +1 depending on the sign of
the loop flow direction compared to the direction of the flow assumed in the pipe ( – 1 for opposite signs
and +1 for same signs).The updated friction factors can then be calculated based on these updated dis-
charges in each pipe. For a network that has more than one loop, Equation 14.19 would require that each
initial flow would be adjusted by all loop flows associated with that pipe.

14.3 Convergence Stopping Criteria for the LF – equations Formula-

tion
As mentioned in Section 10.2.4, Equation 10.23 is not suitable as a convergence stopping criterion for
the LF – equations formulation because the nodal heads are not computed at each iteration. As an alter-
native, a stopping test based on the head loss around closed loops and along paths between nodes of
fixed head is proposed (in a similar way as for the Q – equations formulation in Chapter 12).

Similarly to Equation 12.23, for the LF – equations formulation let the loop head loss equation (for loop
s) be designated respectively as

 LF - L  u  =  r jQ j Q j (14.20)
s
j  P Ls

or

340
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

 init  k + 1  init k + 1
 LF - L  u  =
s
 r j  Q j +  b sj LF s 
Q j +  b sj LF s (14.21)
j  PL s  PL j s  PL j
s

where subscript LF – L refers to loop flow equations for individual loops and b sj = – 1 or b sj = 1
(depending on the flow direction in the pipe relative to the assigned loop flow direction) for each of
pipes in loop s.

Now for the path equations (for path s)

 init  k + 1  init k + 1
LF - P s  u  =  r j  Q j +  b sj LF s 
Q j +  b sj LF s +  E Lm – E L p
j  PL s  LP j s  LP j
s

(14.22)

where subscript LF – P refers to loop flow equations for individual paths, P L s =indices of the pipes in
loop or path s, and LP j =indices of the loops associated with pipe j.

The infinity – norm test (based on Equation 10.20) can be used to stop the iterative convergence process
when for the closed loop head loss equations

k + 1 max k + 1
 LF - L  u  =   LF - L  u     stop (14.23)

s
j=1, NL s

and for the path head loss equations

k + 1 max k + 1
 LF - P  u  =   LF - P  u     stop (14.24)

s
j=1,  NF – NC  s

Both these conditions need to be satisfied.

The checking of the nodal continuity equations is not required as the continuity equations are exactly
init
satisfied for the initially selected flows Q j .

14.4 Solving the LF – equations for the Ten Pipe Network

14.4.1 Unknowns and functions for the Newton method
The total number of closed simple loops and required independent paths for a network is

NTL = NL +  NF – NC  = 2 +  2 – 1  = 3 (14.25)

As shown in Chapter 9, the column vector of unknown loop flows to be solved for based on
Equation 9.36 for the ten pipe network in Figure 14.1 has three unknown loop flows as:

k  k  k  k  T
u =  LF 1  LF 2 , LF 3  (14.26)

Note the superscript k has been introduced here to distinguish between successive iterations.

The column vector of the loop flow corrections that will be solved for in the Newton iterative scheme
for the (k + 1)st iteration from the generalized form of Equation 10.6 and Equation 14.6 becomes

341
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

k + 1 k + 1 k + 1 k + 1 T
u =  LF 1  LF 2 , LF 3  (14.27)

A column vector of initial estimates for the loop flows needs to be made to proceed with the iterative
solution process

0 0 0 0 T

u =  LF 1  LF 2 , LF 3  (14.28)

A vector of initial estimates for the pipe flows needs to be selected such that the ten flows in each of the
pipes must satisfy all the continuity equations at all junctions (see Figure 14.2)

init init init init init T

q =  Q 1  Q 2  ,Q 9  Q 10  (14.29)

These values stay constant for all iterations of the Newton method.

Figure 14.1 Unknown LF – equations formulation for the ten pipe network; the initial pipe flows
are selected such that they satisfy continuity but do NOT change from iteration to iteration

For the ten pipe network in Figure 14.1 assuming a Darcy – Weisbach head loss equation where n = 2,
the loop flows equations are repeated from Chapter 9 in Table 14.1.

342
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

Table 14.1 The LF – equations for the ten pipe network

Function Energy equations with all LF s values taken at iteration k

k  n–1
f 1u  r  Q init + LF – LF  Q init + LF – LF
2 2 1 3 2 1 3
init init n–1
+ r 6  Q 6 + LF 1 – LF 2  Q 6 + LF 1 – LF 2
init init n–1 init init n–1
– r 8  Q 8 – LF 1  Q 8 – LF 1 – r 5  Q 5 – LF 1  Q 5 – LF 1 = 0
k  n–1
f 2u  r  Q init + LF – LF  Q init + LF – LF
3 3 2 3 3 2 3
init init n–1 init init n–1
+ r 7  Q 7 + LF 2  Q 7 + LF 2 – r 9  Q 9 – LF 2  Q 9 – LF 2
init init n–1
– r 6  Q 6 + LF 1 – LF 2  Q 6 + LF 1 – LF 2 = 0
k  n–1
f 3u  r  Q init + LF  Q init + LF
4 4 3 1 4 3
init init n–1
– r 3  Q 3 + LF 2 – LF 3  Q 3 + LF 2 – LF 3
init init n–1
– r 2  Q 2 + LF 1 – LF 3  Q 2 + LF 1 – LF 3
init init n–1
– r 1  Q 1 – LF 3  Q 1 – LF 3 –  E L5 – E L1  = 0
k 
Note that  E L 5 – E L 1  has been moved to the left hand side of f 3  u 

init
The superscripted terms of “init” in the equations above are the initial estimates of the flows Q j in
init
each of the pipes such that the flows satisfy continuity of flow at each of the nodes. The values of Q j
do not change at each iteration – they remain the same values for the entire iterative solution process.

The 3 by 3 Jacobian matrix for the loop flows for the ten pipe network at iteration k from the general
form of Equation 10.12 becomes

k  k  k 
 f 1 u   f 1 u   f 1 u 
------------------------ ------------------------ ------------------------
 LF  1  LF  2  LF  3
k  k  k  k 
u  = -----------------------
f 2u   f 2u   f 2u 
- ------------------------ ------------------------ (14.30)
 LF  1  LF  2  LF  3
k  k  k 
 f 3u   f 3u   f 3u 
------------------------ ------------------------ ------------------------
 LF  1  LF  2  LF  3

k 
Note the Jacobian elements are based on the column vector of loop flow values u determined at the
previous kth iteration.

The first element in row 1 of the Jacobian matrix for the ten pipe network is obtained by differentiating
k  k 
f 1 u  in Table 14.1. Consider the first term on the left – hand side of f 1 u  given in
Equation 14.31.

k 
g 2 u  = r  Q init + LF – LF  Q init + LF – LF (14.31)
2 2 1 3 2 1 3

The partial derivative with respect to LF 1 using Equation 14.17 is given as the first row in Table 14.2 as

343
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

k 
 g 2 u  init
------------------------- = 2r 2 Q 2 + LF 1 – LF 3 (14.32)
 LF  1

k 
Now consider the third term on the left – hand side of f 1 u  given in Equation 14.33.

k 
g 3 u  = r  Q init – LF  Q init – LF (14.33)
5 5 1 5 1

The partial derivative of Equation 14.33 with respect to LF 1 using Equation 14.18 is given as

k 
 g 3 u  init
------------------------- = – 2 r 5 Q 5 – LF 1 (14.34)
 LF  1

Note the negative sign on the right – hand side of Equation 14.34.

k 
The differentiation of all other terms in f 1 u  and the other two functions in Table 14.1 proceeds in a
similar way to as described above. The Jacobian elements for all functions are shown in Table 14.2.

Table 14.2 The non – zero Jacobian elements for LF – equations for the ten pipe network
Derivatives Jacobian element with all LF s values taken at iteration k

 f1 init init
-----------------
- 2r 2 Q 2 + LF 1 – LF 3 + 2r 6 Q 6 + LF 1 – LF 2
 LF  1
init init
+ 2r 8 Q 8 – LF 1 + 2r 5 Q 5 – LF 1
 f1 init
-----------------
- – 2r 6 Q 6 + LF 1 – LF 2
 LF  2

 f1 init
-----------------
- – 2 r 2 Q 2 + LF 1 – LF 3
 LF  3

 f2 init
-----------------
- – 2r 6 Q 6 + LF 1 – LF 2
 LF  1

 f2 init init
-----------------
- 2r 3 Q 3 + LF 2 – LF 3 + 2r 7 Q 7 + LF 2
 LF  2
init init
+ 2r 9 Q 9 – LF 2 + 2r 6 Q 6 + LF 1 – LF 2
 f2 init
-----------------
- – 2r 3 Q 3 + LF 2 – LF 3
 LF  3

 f3 init
-----------------
- – 2 r 2 Q 2 + LF 1 – LF 3
 LF  1

 f3 init
-----------------
- – 2 r 3 Q 3 + LF 2 – LF 3
 LF  2

 f3 init init
-----------------
- 2r 4 Q 4 + LF 3 + 2r 3 Q 3 + LF 2 – LF 3
 LF  3
init init
+ 2r 2 Q 2 + LF 1 – LF 3 + 2r 1 Q 1 – LF 3

The partial derivatives are also based on the general expressions for the derivatives from Equation 13.22
and Equation 13.28.

344
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

 f1  f2  f1  f3  f2  f3
The pairs of non – zero Jacobian elements -----------------
- & - , -----------------
----------------- - & -----------------
- and -----------------
- & -----------------
- in
 LF  2  LF  1  LF  3  LF  1  LF  3  LF  2
Table 14.2 are identical — thus the Jacobian matrix is symmetric.

14.4.2 Solving for the LF – corrections for the ten pipe network
k 
The set of linearized equations for the loop flow corrections (with three unknown values in u ) to be
solved for the ten pipe network during the iterative process is

k + 1 k 
LF 1 f 1 u 
k 
Ju  LF  k + 1  = – f  u  k   (14.35)
2 2
k + 1 k 
LF 3 f 3 u 

or

k  k + 1 k 
Ju u = –f  u  (14.36)

k + 1
The vector of the loop flow correction values u is then added to the current solution vector of
k 
loop flows values u to give the next best estimate of the unknowns for the ten pipe network as

k + 1 k  k + 1
u = u + u (14.37)

14.5 Solving the LF – Equations (Loop Flow) for Two Pipe Network
14.5.1 The unknowns and initial guesses
Now consider the loop flow equations for the network in Figure 14.2. Only one loop flow (LF1) needs to
be solved for along the path between the two reservoirs as designated in. Figure 14.2 also shows the pipe
init init
flows Q 1 and Q 2 that must be selected such that they must satisfy continuity at node 1.

Figure 14.2 Unknown LF to be solved for in two pipe network

From Equation 14.2, the vector of unknown loop flows at the kth iteration for the two pipe network is

k  k  T
u =  LF 1  (14.38)

One function needs to be solved to determine the one unknown loop flow. This function includes the
energy equation around the path in Figure 14.2 written in terms of the unknown loop flows. The column
vector of loop flow corrections in the pipes solved by the Newton method (from Equation 10.6) for the
two pipe network at the (k + 1)st iteration becomes

345
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

k + 1 k + 1 T
u =  LF 1  (14.39)

This vector is solved for using the Newton method.

It follows that the loop flows for iteration (k+1) are

k + 1 k  k + 1
u = u + u (14.40)

but

k + 1 k + 1
u =  LF 1  (14.41)

An initial estimate for the loop flow needs to be made in order for the iterative solution process to begin

0 0 T
u =  LF 1  (14.42)

A column vector of initial estimates for the pipe flows needs to be selected such that the flows selected
satisfy all the continuity equations at all junctions (see Figure 14.2)

init init init T

q =  Q1  Q2  (14.43)

During the iterative process at any iteration (k+1) the flows in each of the pipes are updated from the ini-
tial estimates of the flows at each iteration of the Newton solution process using Equation 14.19 that
leads to for the two pipe network for pipes 1 and 2

k + 1 init k + 1 init k + 1
Qj = Qj +  b sj LF s = Q j – LF 1 (14.44)
s  PL j

where b sj = – 1 and each of pipes 1 and 2 only has one associated loop flow correction term.

During the iterative process, the friction factors and pipe resistance terms ( r f =  r 1 r 2  ) for each itera-
2
tion can be updated based on the current flows. For the Darcy – Weisbach head loss equationh f = r f Q
j

would be used.

14.5.2 Two pipe example based on Newton method (LF – equations formu-
lation)

Example 14.1 For the two pipe network in Figure 14.2 do the following:

(i) Develop the LF – equations formulation (loop flow) for the two pipe network system.
(ii) Develop the Jacobian matrix for the LF – equations formulation and carry out the first two iterations.
(iii) Compute the condition number for the last iteration for the LF – equations formulation.
(iv) Tabulate all iterations for the solution process for the LF – equations formulation.
(v) Plot convergence characteristics for the iterations for the LF – equations formulation for the two pipe network.
Assume the Darcy – Weisbach head loss equation is used for each pipe.

Answer – (i) Develop the LF – equations formulation for the two pipe network system.

The LF – equations formulation of the governing equations for this two pipe network includes the energy equation
along the path between the two reservoirs written in terms of unknown loop flow LF1:

346
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

init init init init

f 1  u  = – r 2  Q 2 – L F 1   Q 2 – L F 1  – r 1  Q 1 – L F 1   Q 1 – L F 1  +  EL 2 – E L 3  = 0 for the path (14.45)

Note that the “constants” for the two terms for rf in Equation 14.45 are dependent on the friction factor which in turn
requires computation of flows from Equation 14.44 and hence are required to be updated at each iteration. The
unknown LF1 is assumed to be clockwise in Figure 14.2 thus the energy equation to go around the path begins with
pipe 2 and then moves to pipe 1.
___________________________________________________________________________________
Answer – (ii) Develop the Jacobian matrix for the LF – equations formulation and carry out the first two itera-
tions.

The linear system from Equation 14.36 that need to be solved iteratively for the vector of loop flow corrections
k + 1
u is

k  k + 1 k 
Ju u = –f  u  (14.46)

There is only one term in the Jacobian at the kth iteration and this can be determined by taking the appropriate partial
derivative of Equation 14.45 as follows:

k 
 = -----------------------
f 1  u  = 2.0r  Q init – L F  + 2.0r  Q init – L F 
k 
Ju - 2 2 1 1 1 1 (14.47)
 LF 1

ITERATION 1, LF – equations
Firstly an initial starting guess of the loop flow to be solved for associated with the path in Figure 14.2 is taken as:

0 0 3
u = LF 1 = 0.025 m /s (14.48)

In addition pipe flows that satisfy the continuity equation at node 1 are selected as:

init 3
Q1 m
= 0.100 --------
init
q = (14.49)
Q2
init 0.050 s

The corresponding pipe flows from Equation 14.19 are:

0 init 0 init 0 3

Q1 = Q1 +  b sj L F s = Q1 – L F 1 = 0.100 – 0.025 = 0.075 m /s (14.50)
s  PL 1

0 init 0 init 0 3

Q2 = Q2 +  b sj L F s = Q2 – L F 1 = 0.050 – 0.025 = 0.025 m /s (14.51)
s  PL 1

where b sj = – 1 for both pipes 1 and 2. Thus the loop flow corrections are subtracted from the pipe flows because the
direction of the loop flow correction arrow has been selected to be opposite to the pipe flows around the loop. From
Equation 14.50 and Equation 14.51 it follows that

0
0 Q1
= 0.075
3
q = m /s (14.52)
0 0.025
Q2

The friction factors (based on the Swamee and Jain equation – Equation 4.61) and corresponding rf values
(Equation 4.123) for the flows in Equation 14.49 (assuming the Darcy – Weisbach head loss equation) for pipes 1 and 2
in the two pipe network are given in Table 14.3

347
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

Table 14.3 The rf values for flows from Iteration 2 for the Q – equations formulation

Pipe Q V D Area (m2) Re**  f L (m) rf value

(m3/s) (m/s)* (m) (mm) (Swamee and Jain) (Equation 4.123)
(Equation 4.61)
1 0.075 1.061 0.3 0.07069 316,097 0.25 0.01997 1000 679.06
2 0.025 0.354 0.3 0.07069 105,366 0.25 0.02163 1000 735.49
–6 2
**  = 1.007 10 m /s at 20C ;
0.25 fL -
Using in the above Table: f = ---------------------------------------------------------
- ; r f = -----------------
  5.74  2 2 gD A
2
log 10 ------------ + ------------
 3.7 D 0.9
Re

The value of loop flow from Equation 14.48 and the initial flows in each of the pipes from Equation 14.49 may be sub-
stituted into the path energy equation of Equation 14.45 as follows:
0
f 1 u  =   – 735.49   0.050 – 0.025   0.050 – 0.025  – 679.06  0.100 – 0.025   0.100 – 0.025  +  80.0 – 50.0  
= – 0.4597 – 3.8197 + 30.0 = 25.721 m

Thus the imbalance in the head around the path is 25.721 m. It should be zero.

The Jacobian matrix for the first iteration from Equation 14.47 is:

0
Ju  = 2.0  735.49   0.050 – 0.025  + 2.0  679.06   0.100 – 0.025  = 138.63 (14.53)

The inverse of the Jacobian matrix is shown below for this simple two pipe network (note that the inverse of the Jaco-
bian would never be actually calculated in practice for a network of normal size) but in this case is very straight for-
ward for a single number, as it is the reciprocal of the value in Equation 14.53 as follows:

–1 0
J u  = 0.0072134 (14.54)

The loop flow correction for the first iteration are calculated from Equation 14.13 as:

1 –1 0 0 3

u = –J  u f  u  = – 0.0072134 25.721 = – 0.1855 m /s (14.55)

It follows then that the loop flow for the path after the first iteration is:

1 0 1 3

u = u + u = 0.025 – 0.1855 = – 0.1605 m /s (14.56)

Thus

1 3
LF 1 = – 0.1605 m /s (14.57)

The corresponding pipe flows from Equation 14.19 are:

1 init 1 init 1 3

Q1 = Q1 +  b sj L F s = Q1 – L F 1 = 0.100 – – 0.1605 = 0.2605 m /s (14.58)
s  PL 1

1 init 2 init 1 3

Q2 = Q2 +  b sj L F s = Q2 – L F 1 = 0.050 – – 0.1605 = 0.2105 m /s (14.59)
s  PL 1

where b sj = – 1 for both pipes 1 and 2. Thus it follows that

1
1 Q1
= 0.2605
3
q = m /s (14.60)
1 0.2105
Q2

Note that the new flows still satisfy continuity at node 1. These updated flows are only solved for in order to recompute
the friction factors.
For the Hazen – Williams head loss equation it is not necessary to compute the flows at every iteration as the loss coef-
ficients do not depend on flows.

348
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

For the path energy equation balance with the new flows by substituting into Equation 14.45 is:

0
f 1 u  =  – 735.49   0.050 – – 0.1605   0.050 – – 0.1605  (14.61)
– 679.06  0.100 – – 0.1605   0.100 – – 0.1605 
+  80.0 – 50.0  = – 32.589746 – 46.081181 + 30.0 = – 48.671 m

To check on convergence for iteration 1, use Equation 14.23 for the infinity – norm for the governing path equation in
the LF – equations formulation as follows

k + 1 max k + 1
 LF - P  u  =   LF - P  u     stop (14.62)

s
j=1, NF – NC s

where NF – NC = 1, s = 1 and k = 0.

Thus for the two pipe network the check is for the path equation only (from Equation 14.62) as there are no closed
loops

 k + 1  k + 1
r j  Q j +  b sj L F s
1 init init
 LF - P  q 
1 
=   
Qj +  b sj L F s +  E Lq – E L p 
j  P L1 s  LP j s  LP j

The following is not satisfied

1 –6
 LF - P  q    stop where  stop = 1.0 10 m.
1 

Thus the iterative process has not converged.

ITERATION 2, LF – equations

The friction factors and corresponding rf values for the flows in Equation 14.55 that resulted from iteration 1 can now
be computed as shown in Table 14.6.
Table 14.4 The rf values for flows from Iteration 2 for the Q – equations formulation

Pipe Q V D Area (m2) Re**  f L (m) rf value

(m3/s) (m/s)* (m) (mm) (Swamee and Jain) (Equation 4.123)
(Equation 4.61)
1 0.2605 3.685 0.3 0.07069 1,097,911 0.25 0.01920 1000 652.78
2 0.2105 2.978 0.3 0.07069 887,179 0.25 0.01928 1000 655.60

–6 2
**  = 1.007 10 m /s at 20C ;
0.25 fL
Using in the above Table: f = ---------------------------------------------------------- ; r f = ------------------
 5.74 2 2
log 10  ------------ + ------------ 2 gD A
 3.7 D 0.9
Re

Thus

1
f = 0.01920 (14.63)
0.01928

and

1
rf = 652.78 (14.64)
655.59

The loop flow from Equation 14.56 may be substituted into the path energy equation of Equation 14.45 as follows:

1
f 1 u  =   – 655.59   0.050 –  – 0.1605    0.050 –  – 0.1605   – 652.78  0.100 –  – 0.1605    0.100 –  – 0.1605  
+  80.0 – 50.0  
= – 29.057 – 44.307 + 30.0 = – 43.365 m

Thus the imbalance in the head around the path is now negative 43.365 m compared with the previous value for itera-
tion 1 of positive 25.721 m.

The Jacobian matrix for the second iteration from Equation 14.47 is:

349
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

1
Ju  = 2.0  655.59   0.050 –  – 0.1605   + 2.0  652.78   0.100 –  – 0.1605   = 616.177 (14.65)

The inverse of the Jacobian matrix is the reciprocal of the value in Equation 14.65 as follows:

–1 1
J u  = 0.0016229 (14.66)

The flow corrections for the second iteration are calculated from Equation 14.13 as:
2 –1 1 1 3
u = –J  u f  u  = –  0.0016229   – 43.365  = 0.07038 m /s

The loop flow for the path after the second iteration is:

2 1 2 3

u = u + u = – 0.1605 + 0.07038 = – 0.09015 m /s (14.67)

Thus

2 3
LF 1 = – 0.09015 m /s (14.68)

The corresponding pipe flows from Equation 14.19 are:

2 init 2 init k  3

Q1 = Q1 +  b sj L F s = Q1 – L F 1 = 0.100 –  – 0.09015  = 0.19015 m /s (14.69)
s  PL 1

2 init 2 init 2 3

Q2 = Q2 +  b sj L F s = Q2 – L F 1 = 0.050 –  – 0.09015  = 0.14015 m /s (14.70)
s  PL 1

where b sj = – 1. Thus it follows that

2
2 Q1
= 0.19015
3
q = m /s (14.71)
2 0.14015
Q2

The updated rf values are given in Table 14.5.

Table 14.5 The rf values for flows from Iteration 2 for the Q – equations formulation

Pipe Q V D Area (m2) Re**  f L (m) rf value

(m3/s) (m/s)* (m) (mm) Swamee and Jain (Equation 4.123)
(Equation 4.61)
1 0.2605 3.685 0.3 0.07069 1,097,911 0.25 0.01920 1000 652.78
2 0.2105 2.978 0.3 0.07069 887,179 0.25 0.01928 1000 655.60

–6 2
**  = 1.007 10 m /s at 20C ;
0.25 fL -
Using in the above Table: f = ---------------------------------------------------------
- ; r f = -----------------
 5.74 2 2
log 10  ------------ + ------------ 2 gD A
 3.7 D 0.9
Re

For the path energy equation with the new flows:

0
f 1 u  =   – 655.60   0.050 –  – 0.09015    0.050 –  – 0.09015   (14.72)
– 652.78  0.100 –  – 0.09015    0.100 –  – 0.09015   +  80.0 – 50.0  
= – 12.877 – 23.603 + 30.0 = – 6.480 m

To check on convergence for iteration 2, use Equation 14.23 for the infinity – norm for the governing path equation in
the LF – equations formulation as follows

k + 1 max k + 1
 LF - P  u  =   LF - P  u     stop

s
j=1, NF – NC s

where NF – NC = 1, s = 1 and k = 1.

350
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

Thus for the two pipe network the check is for the path equation only (from Equation 14.72) as there are no closed
loops

1  init k + 1  init k + 1

 LF - P  q  =  r j  Q j +  b sj L F s  Q j +  b sj L F s +  E Lq – E L p 

1
j  P Ls  s  PL j  s  PL j

The following is not satisfied

1 –6
 LF - P  q    stop where  stop = 1.0 10 m
1 

Thus the iterative process has not converged.

The remaining iterations are shown in Table 14.6.

___________________________________________________________________________________
Answer – (iii) Compute the condition number for the last iteration for the LF – equations formulation.

ITERATION 7 – FINAL ITERATION, LF – equations

The condition number of Jacobian matrix from Equation 10.26 based on the infinity – norm at iteration 7 is:

–1 J 
cond(J) = J  J  = ----------
- = 1.0
J 

as would be expected for a Jacobian with one matrix element.

The size of the condition number is acceptable given that one place of accuracy would be lost from say a total of 15 sig-
nificant figures for computations using double precision calculations.

___________________________________________________________________________________
Answer – (iv) Tabulate all iterations for the solution process for the LF – equations formulation.

FINAL SOLUTION – ALL ITERATIONS, LF – equations

The iterations for the full solution based on the LF – equations formulation are shown in Table 14.6. Seven iterations
are shown. The stopping tolerance based on the infinity – norm of the head changes that was used in the computer sim-
–6
ulation run was 1.0 10 m. Again as for the Q – equations formulation (see Table 12.15, page 305) this value of stop-
ping tolerance is much smaller than is normally required, but it is used to demonstrate the quadratic convergence
–3
properties of the Newton method. Realistically the stopping tolerance could have been 1.0 10 m or 1 mm which
means that convergence would have occurred in 5 iterations.

Table 14.6 Iteration sequence for the Newton solution for LF – equations formulation
–6
(stopping tolerance 1.0 10 m achieved at iteration 5) for two pipe network

Iteration LF 1 m3/s Q 1 m3/s Q 2 m3/s k + 1

 LF - L  u 
s 
(from Equation 14.24) **
0 0.02500 0.100 0.050

1 – 0.16053 0.26053 0.21053 43.365

2 – 0.09015 0.01915 0.14015 6.480
3 – 0.07460 0.17460 0.12460 0.0319
4 – 0.07359 0.17359 0.12359 1.36E-03
5* – 0.07357 0.17357 0.12357 4.33E-07
–3
* A stopping tolerance of less than 1.0 10 m or 1 mm achieved at iteration 5
** Infinity – norm for path head losses (m) are used for testing for convergence

max k + 1
=   LF - P  u  
j=1,  NF – NC  s

___________________________________________________________________________________
Answer – (v) Plot convergence characteristics for the iterations for the LF – equations formulation for the two
pipe network.

351
 Chapter 14 - File: Chap13-Loop-Form.fm July 6, 2022 Version 7

Figure 14.3 Convergence of loop flows for two pipe network for the Newton method

___________________________________________________________________________________

14.6 References
Cross, H. (1936). Analysis of flow in networks of conduits or conductors. Bulletin No. 286, University of
Illinois Engineering Experimental Station, Urbana, Illinois.

Epp, R. and A. G. Fowler (1970). “Efficient code for steady-state flows in networks.” Journal of the
Hydraulics Division, Proceedings of the ASCE, January, Vol. 96, No. HY1: 43-56.

Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1996) FORTRAN numerical recipes
(Vol. 1), Cambridge England.

352
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

CHAPTER 15 SOLVING THE Q+H FORMULATION1

15.1 Introduction
The flow and head equations or Q+H equations formulation are solved for by the Newton method in this
Chapter. Both the continuity equations at nodes and the pipe head loss equations are expressed in terms
of unknown flows in the pipes (or other flow elements) and unknown heads at nodes. There are a total of
NP + NJ unknown flows and heads that need to be solved for. The number of unknowns to be solved
for in the formulation is by far the greatest of any of the formulations. Thus larger matrices need to be
dealt with in the solution process. This is a severe disadvantage of this method and as a consequence it
will not be recommended for usage. It will be seen that in Chapter 16 that the Global Gradient Algo-
rithm is a smart variant of the Q+H equations formulation that ends up only requiring to solve a system
of equations of size NJ – the number of unknown heads. The Global Gradient Algorithm formulation is
the preferred formulation.

15.2 Solving the Q+H equations for a Water Distribution System

15.2.1 The set of Q+H equations
Consider the general set of nonlinear equations for a set of both unknown flows for the NP pipes and for
unknown heads for NJ nodes in a network and expressed by the following set of (NP+NJ) – functions
made up of the head loss equations for each link and the continuity equations at each node in the net-
work as described in Chapter 9 (Section 9.10, page 247) as

f 1  Q 1 Q 2  Q NP H 1 H 2  H NJ  = 0

f 2  Q 1 Q 2  Q NP H 1 H 2  H NJ  = 0

f NP  Q 1 Q 2  Q NP H 1 H 2  H NJ  = 0

f NP + 1  Q 1 Q 2  Q NP H 1 H 2  H NJ  = 0

f NP + NJ  Q 1 Q 2  Q NP H 1 H 2  H NJ  = 0 (15.1)

where

• f j  Q 1 Q 2  Q NP H 1 H 2  H NJ = jth function in the Q+H equations formulation

(either a pipe head loss equation or a continuity equation)

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.

353
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

• Q j = flow in pipe j (or other flow element) (m3/s or ft3/s)

• H i = nodal head at node i (m or ft)

The first NP functions in Equation 15.1 are the pipe head loss equations for each pipe while the last NJ
functions are the continuity equations at each of the nodes.

Let the unknown flows and heads that are required to be solved for be expressed as the column vector as
follows

T
m =  Q 1 Q 2  Q NP H 1 H 2  H NJ  (15.2)

Thus the functions from Equation 15.1 become

f 1  m 1 m 2  m NP + NJ  = 0

f 2  m 1 m 2  m NP + NJ  = 0

f NP + NJ  m 1 m 2  m NP + NJ  = 0 (15.3)

where

• f s  m 1 m 2  m NP + NJ  = sth function in the Q+H – equations formulation (either a pipe

head loss or a continuity equation)
• m s = sth head or flow in the m vector
• NP = number of pipes in the network
• NJ = number of nodes with unknown heads in the network
Let the true solution to the set of nonlinear equations in Equation 15.1 be the following column vector

T
m* =  Q *1 Q *2  Q *NP H *1 H *2  H *NJ  (15.4)

In order to find a solution by a successive iterative technique, let the deviation or correction column vec-
tor for the unknown flow and head problem be defined as

T
m =  Q 1 , Q 2 ,, Q NP H 1 H 2  H NJ  (15.5)

This is the quantity that will be solved for in the iterative Newton method.

k + 1
In the Newton numerical solution technique discussed, the objective is to find a solution m at the
(k +1)st iteration that is close to m* within an acceptable tolerance, in other words when m is very
small. Thus the relationship between the true or correct flow vector m* , the current flow vector m, and
the flow correction vector m is

m* = m + m (15.6)

when m is small.

354
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

An initial estimate (referred to as the zeroth iteration indicated by the superscript zero (0)) needs to be
made for each of the unknown flows and heads to commence the iterative solution process. The column
vector of initial estimates of the pipe flows and nodal heads is

0 0 0 0 0 0 0 T

m =  Q 1  Q 2   QNP  H 1  H 2   H NJ  (15.7)

A commonly used value for commencing the iterative solution process for the flows is to set all veloci-
ties to 1.0 fps (0.3048 m/s) in all pipes and then to compute the corresponding discharges (Rossman
2000). In addition, initial guesses of the heads at the nodes are needed.

15.2.2 The Jacobian matrix for the unknown flow and head equations
The solution for the unknown flows and heads requires linearization of the nonlinear governing equa-
tions that were presented in Section 9.10 and given in Equation 15.1. In the neighborhood of the current
iteration point each of the functions in Equation 15.1 may be expanded in a Taylor series as shown ear-
lier in Equation 10.8

NP + NJ  f
1m
f 1 m + m  = f 1  m  +  -m s + O  m 
------------------
m s
s=1

NP + NJ  f
2m
f 2  m + m  = f 2  m  +  -m s + O  m 
------------------
m s
s=1

NP + NJ  f
NP + NJ  m 
f NP + NJ  m + m  = f NP + NJ  m  +  -m s + O  m 
----------------------------------- (15.8)
m s
s=1

where

• f s  m , f s  m + m  = sth function in the Q+H equations formulation (either a pipe head

loss or a continuity equation)
• m = column vector of unknown heads and flows (m or ft) and flows (m3/s or ft3/s)
• m = deviation vector of the unknown heads and flows; m s = sth deviation of heads and
flows in the m vector
• NP = number of pipes in the network
• NJ = number of nodes with unknown heads in the network
2
• O  m  = second order and higher terms of the flow and head corrections
Putting Equation 15.8 into matrix notation using Equation 10.9 (Press et al. 1996) gives

2
f  m + m  = f  m  + J  m m + O  m  (15.9)

where

• J  m  = Jacobian matrix based on the flows and heads

355
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

From the set of equations in Equation 15.9, the Jacobian matrix for the flow equations formulation based
on the values of the flows at the iteration k is defined as

k  k  k 
 f 1m   f 1m   f 1m 
-------------------------- -------------------------
-  --------------------------
m 1 m 2 m NP + NJ
k  k  k 
k 
 f 2m   f 2m   f 2m 
m  = -------------------------- -------------------------
-  -------------------------- (15.10)
m 1 m 2 m NP + NJ
   
k  k  k 
 f NP + NJ  m   f NP + NJ  m   f NP + NJ  m
----------------------------------------- -----------------------------------------  -----------------------------------------
m 1 m 2 m NP + NJ

2
If m is small then second order and higher terms of O  m  can be dropped and also by setting in
Equation 15.9

f  m + m  = f  m*  = 0

gives

k  k + 1 k 
Jm m = –f  m  (15.11)

EQUATIONS TO BE SOLVED FOR THE Q+H EQUATIONS FORMULATION

Equation 15.11 can also be expressed as

k  k  k 
 f 1m   f 1m   f 1m  k + 1
------------------------- ------------------------
-  ------------------------- m 1
m 1 m 2 m NP + NJ f 1  Q 1 Q 2  Q NP H 1 H 2  H NJ 
k + 1
k  k  k  m 2 f 2  Q 1 Q 2  Q NP H 1 H 2  H NJ 
 f 2m   f 2m   f 2m 
------------------------- ------------------------
-  ------------------------- = –
m 1 m 2 m NP + NJ  
    

k  k  k  f NP + NJ  Q 1 Q 2  Q NP H 1 H 2  H NJ
f NP + NJ  m   f NP + NJ  m   f NP + NJ  m  k + 1
-------------------------------------- -----------------------------------------  ----------------------------------------- m NP + NJ
m 1 m 2 m NP + NJ

(15.12)

Note the subscripts k and (k + 1) have been introduced to represent to the successive iteration numbers.
k + 1
The flow and head correction vector m on the left – hand side of Equation 15.12 is for the (k +
k  k 
1)st iteration. The Jacobian matrix J  m  and the flows and heads m are at iteration k.

k + 1
Equation 15.12 will be solved for the unknown vector m to avoid computation of the inverse of
the Jacobian. This equation is a set of linear equations and may be solved by a standard matrix technique
or sparse matrix technique.

k + 1
The vector of corrections to the flows m can be expressed from Equation 15.11 as

k + 1 –1 k  k 
m = –J  m f  m  (15.13)

where

356
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

–1 k 
• J  m  = inverse of the Jacobian matrix for the Q+H equations formulation based on
the flows at iteration k (the Jacobian matrix is defined by Equation 15.10)
k + 1
Once the vector of corrections m at the (k + 1)st iteration is determined it is then added to the
k 
current solution vector of flows and heads m to give the next best estimate of the unknowns as

k + 1 k k + 1
m = m + m (15.14)

or

k + 1
Q 1
k + 1 k 
Q1 Q1 k + 1
Q 2
k + 1 k 
Q2 Q2

 
  
k + 1 k  k + 1
Q NP Q NP Q NP
= + (15.15)
k + 1 k  k + 1
H1 H1 H 1
k + 1 k  k + 1
H2 H2 H 2
  
 
k + 1 k  
Q NJ H NJ
k + 1
H NJ

k + 1
The process is iterated until the vector m converges to within an acceptable tolerance.

357
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

15.3 The Q+H equations Formulation for the Ten Pipe Network
In this section a formulation of the Q+H equations for a ten pipe network as shown in Figure 15.1 is
given. In Section 15.4 a numerical example of a two pipe network is developed.

Figure 15.1 Unknown Q+H equations formulation for the ten pipe network

15.3.1 Unknowns and functions for the Newton solution of Q+H equations
Consider the equations for the network in Figure 15.1 formulated in terms of a total of 17 unknowns
including 10 flows (Q1 to Q10 – thus NP = 10) in each of the pipes and 7 heads (H2 to H9 – thus NJ = 7,
with nodal heads at 1 and 5 are not included) at each of the nodes for the network previously considered
in Chapter 9. The properties of the network are given in Table 12.2 and Table 12.2. The column vector
of unknown flows and heads from Equation 15.2 at the kth iteration is

T
m =  Q 1 Q 2  Q 10 H 2 H 3  H 9  (15.16)

In order to find a solution by a successive iterative technique, let the deviation or correction column vec-
tor for the unknown flow and head problem be defined as

T
m =  Q 1 , Q 2 ,, Q 10 H 2 H 3  H 9  (15.17)

This vector needs to be solved for in the iterative process.

An initial estimate (referred to as the zeroth iteration indicated by the superscript zero (0)) needs to be
made for each of the unknown flows and heads to commence the iterative solution process. The column
vector of initial estimates of the pipe flows is

0 0 0 0 0 0 0 T

m =  Q 1  Q 2   Q 10  H 2  H 3   H 9  (15.18)

358
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

The set of governing equations for the ten pipe network were presented in Chapter 9. These are repeated
in Table 15.1 for the Darcy – Weisbach head loss equation with n = 2.

Table 15.1 The Q+H equations for the ten pipe network based on the pipe head equations and
continuity equations (for the Darcy-Weisbach head loss equation n = 2)
Func- Pipe head loss equation Func- Continuity equation
tion tion
f 1m – r 1 Q1 Q1 +  E L1 – H 2  = 0 f 11  m  – Q 6 – Q 8 + Q 10 + DM 7 = 0

f 2m – r 2 Q2 Q2 +  H 2 – H 3  = 0 f 12  m  – Q 7 – Q 9 + Q 10 + DM 8 = 0

f 3m – r 3 Q3 Q3 +  H 3 – H 4  = 0 f 13  m  – Q 10 + DM 9 = 0

f 4m – r 4 Q4 Q4 +  E L5 – H 4  = 0 f 14  m  – Q 1 + Q 2 + Q 5 + DM 2 = 0

f 5m – r 5 Q5 Q5 +  H 2 – H 6  = 0 f 15  m  – Q 2 + Q 3 + Q 6 + DM 3 = 0

f 6m – r 6 Q6 Q6 +  H 3 – H 7  = 0 f 16  m  – Q 3 – Q 4 + Q 7 + DM 4 = 0

f 7m – r 7 Q7 Q7 +  H 4 – H 8  = 0 f 17  m  – Q 5 – Q 8 + DM 6 = 0

f 8m – r 8 Q8 Q8 +  H 6 – H 7  = 0

f 9m – r 9 Q9 Q9 +  H 7 – H 8  = 0

f 10  m  – r 10 Q 10 Q 10 +  H 8 – H 9  = 0

15.3.2 Jacobian matrix for the Q+H equations for the ten pipe network
The Jacobian matrix of Equation 15.10 based on values at the kth iteration is a 17 by 17 square matrix for
the 10 unknown pipe flows and the 7 unknown heads and becomes

k  k  k 
 f 1m   f 1m   f 1m 
-------------------------- --------------------------  -------------------------
-
m 1 m 2 m 17
k  k  k 
k 
 f 2m   f 2m   f 2m 
m  = -------------------------- --------------------------  -------------------------
- (15.19)
m 1 m 2 m 17
   
k  k  k 
 f 17  m   f 17  m   f 17  m 
---------------------------- ----------------------------  ---------------------------
-
m 1 m 2 m 17

The first function from Table 15.1 based on the first pipe head loss equation is

f 1  m  = – r 1 Q1 Q1 +  E L1 – H 2  = 0 (15.20)

The two non – zero partial derivatives of the first row of the Jacobian matrix for Equation 15.20 are now
evaluated. The 17 by 17 Jacobian matrix is shown in Table 15.3

The first non – zero element of the first row of the Jacobian matrix (the partial derivative with respect to
Q 1 )in the first column of Table 15.3 is

359
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

k  k 
 f 1m   f 1m  k 
------------------------- = -------------------------
- = – 2.0 r 1 Q 1 or element A in Table 15.2. (15.21)
m 1 Q 1

The second non – zero element of the first row of the Jacobian matrix (the partial derivative with respect
to H 2 ) in the eleventh column in Table 15.3 is

k  k 
 f 1m   f 1m 
-------------------------- = -------------------------
- = –1 (15.22)
m 11 H 2

All other partial derivatives in row 1 of the Jacobian matrix in Table 15.3 are zero.

Derivatives for the other pipes are given in Table 15.2 and Table 15.3. The derivatives of the pipe head
loss equations in the top portion of the Jacobian matrix (rows 1 to 10, red colored area) are shown below
in Table 15.3.

Table 15.2 The Jacobian matrix for the Q+H equations at iteration k for the ten pipe network
(for the Darcy-Weisbach head loss equation n = 2)
Element in Non – zero Value of Element in Non – zero Value of
Table 15.3 Jacobian Jacobian Table 15.3 Jacobian ele- Jacobian
element element ment element
A k  k  F k  k 
 f 1m  – 2.0 r 1 Q 1  f 6m  – 2.0 r 6 Q 6
-------------------------- --------------------------
Q 1 Q 6

B k  k  G k  k 
 f 2m  – 2.0 r 2 Q 2  f 7m  – 2.0 r 7 Q 7
-------------------------- --------------------------
Q 2 Q 7

C k  k  H k  k 
 f 3m  – 2.0 r 3 Q 3  f 8m  – 2.0 r 8 Q 8
-------------------------- --------------------------
Q 3 Q 8

D k  k  J k  k 
 f 4m  – 2.0 r 4 Q 4  f 9m  – 2.0 r 9 Q 9
-------------------------- --------------------------
Q 4 Q 9

E k  k  K k  k 
 f 5m  – 2.0 r 5 Q 5  f 10  m  – 2.0 r 10 Q 10
-------------------------- ----------------------------
Q 5 Q 10

Examples of the partial derivatives for the continuity equation at node 2 in the eleventh row of the Jaco-
bian matrix in Equation 15.19 are given for the Q – equations formulation in Table 12.6 (page 300).
Three elements are non – zero in columns 1, 2 and 5. In addition elements 11 through 17 of the eleventh
row will also take on a zero value.The derivatives of the other continuity equations are also shown in
Table 15.3 in the bottom portion of the Jacobian matrix (rows 11 to 17, colored blue).

Four important partitions are evident by the colored areas in Table 15.3 as follows

1. The red partition is the derivative of the flow terms in each pipe (from the head loss equa-
tion) along the diagonal (A, B, C,... etc.) and created a diagonal matrix in that partition.
2. The upper right hand side blue area is the unknown head incidence matrix A 1
(Equation 9.11, page 227).

360
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

T
3. The lower left hand side blue area is the transpose of the unknown head incidence matrix A 1
(Equation 9.12, page 228)
4. The yellow area in the lower right hand corner is all zeros.

Table 15.3 The Jacobian matrix for the Q+H equations at iteration k for the ten pipe network (for
definitions of A, B, C,....etc. see Table 15.2)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Pipe 1 1 A –1
Pipe 2 2 B 1 –1
Pipe 3 3 C 1 –1
Pipe 4 4 D –1
Pipe 5 5 E 1 –1
Pipe 6 6 F 1 –1
Pipe 7 7 G 1 –1
Pipe 8 8 H 1 –1
Pipe 9 9 J 1 –1
Pipe 10 10 K 1 –1
Node 2 11 –1 1 1 0
Node 3 12 –1 1 1 0
Node 4 13 –1 –1 1 0
Node 6 14 –1 1 0
Node 7 15 –1 –1 1 0
Node 8 16 –1 –1 1 0
Node 9 17 –1 0
All empty cells represent a zero value in the Jacobian

As will be seen in the Chapter the form of the Jacobian in Table 15.3 is the same as that derived for the
GGA of

k 
k  –n G A1
Jm  = (15.23)
T
A1 0

where from Chapter the node-arc incidence matrix (from Equation 15.24) is

Node = 02 03 04 06 07 08 09

–1 0 0 0 0 0 0 pipe 1
1 –1 0 0 0 0 0 pipe 2
0 1 –1 0 0 0 0 pipe 3
0 0 –1 0 0 0 0 pipe 4
A1 = 1 0 0 –1 0 0 0 pipe 5 (15.24)
0 1 0 0 –1 0 0 pipe 6
0 0 1 0 0 –1 0 pipe 7
0 0 0 1 –1 0 0 pipe 8
0 0 0 0 1 –1 0 pipe 9
0 0 0 0 0 1 –1 pipe 10

and the matrix G is defined from Equation 9.52 as

361
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

n–1
r 1 Q1 0 0  0 0 0
n–1
0 r 2 Q2 0  0 0 0
n–1
0 0 r 3 Q3  0 0 0
G =        (15.25)
n–1
0 0 0  r 8 Q8 0 0
n–1
0 0 0  0 r 9 Q9 0
n–1
0 0 0  0 0 r 10 Q 10

15.3.3 Solving for the flow corrections for the ten pipe network
Once the Jacobian matrix for iteration k has been evaluated the set of linearized equations to be solved
for the corrections to the pipe flows and heads during the iterative process from Equation 15.12 is

k + 1
Q 1
k + 1
Q 2 k 
f 1m 

k  k 
Jm  Q  k + 1  = – f 2  m  (15.26)
10

k + 1
H 2 k 
f 17  m 

k + 1
H 9

or

k  k + 1 k 
Jm m = –f  m  (15.27)

Once the Jacobian matrix has been determined for a particular iteration, Equation 15.27 must now be
k + 1
solved for the vector of flow corrections m . During the iterative solution procedure, the correc-
k + 1
tions m for each iteration are added to the current solution vector to give the next best estimate of
the unknowns as

k + 1 k  k + 1
m = m + m (15.28)

15.4 Solving the Q+H equations for the Two Pipe Network
15.4.1 The unknowns and initial guesses
Now consider the flow and head equations for the network in Figure 15.2. Flows for each of the two
pipes need to be solved for, that is Q1 and Q2, as well as, the head at node 1, H1.. From Equation 15.2
(page 354), the column vector of unknown flows and at the kth iteration for the two pipe network is

k  k  k  k  T
m =  Q1  Q2  H 1  (15.29)

362
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

Figure 15.2 Unknown Qs and the H to be solved for in the two pipe network

The functions to be solved include two head loss equations for the pipes and one continuity equation at
node 1.

The column vector of flow and head corrections in the pipes to be solved at each iterative step in the
Newton method (from Equation 15.17, page 358) for the two pipe network at the (k + 1)st iteration
becomes

k + 1 k + 1 k + 1 k + 1 T
m =  Q 1  Q 2  H 1  (15.30)

A column vector of initial estimates for the flows in each pipe and the heads at each node needs to be
made in order for the iterative solution process to begin

0 0 0 0 T

m =  Q1  Q2  H 1  (15.31)

0
The initial flows Q j are often based on a flow velocity of 1.0 fps (0.3048 m/s) (as used in EPANET -
Rossman 2000) as shown in Table 12.11 (Table 12.11, for the Q – equations formulation). An estimate
of the unknowns heads also needs to be made. The pipe properties for the network are given in
Table 12.8.

The pipe resistance factors ( r f =  r 1 r 2  ) can now be computed based on initial flow guesses shown in
2
Table 12.11 for the Darcy – Weisbach head loss equation h f = r f Q (where n = 2) using
j

Equation 12.30. The pipe resistance values r j are shown in Table 12.12 (for the Q – equations formula-
tion) with the friction factors computed from the Swamee and Jain equation. These pipe resistance val-
ues r j are updated at each iteration based on the latest flow values.

15.4.2 Two pipe example based on Newton method (Q+H equations formu-
lation)

Example 15.1 For the two pipe network in Figure 15.2 complete the following:

(i) Develop the Q+H equations formulation for the two pipe network.
(ii) Develop the Jacobian matrix for the Q+H equations formulation and carry out the first two iterations for the two
pipe network.
(iii) Tabulate all iterations for the solution process for the Q+H equations formulation.
(iv) Plot the convergence characteristics for the iterations for the Q+H equations formulation for the two pipe net-
work.
Assume the Darcy – Weisbach head loss equation is used for each pipe.

Answer – (i) Develop the Q+H equations formulation for the two pipe network.

The Q+H equations formulation of the governing equations for this two pipe network includes two pipe head loss
equations, one for each pipe, and a continuity equation at node 1 as follows:

363
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

f 1  m  = – r 1 Q 1 Q 1 +  EL 2 – H 1  = 0 for pipe 1 (15.32)

f 2  m  = – r 2 Q 2 Q 2 +  H 1 – EL 3  = 0 for pipe 2 (15.33)

f 3  m  = – Q 1 + Q 2 + DM 1 = 0 at node 1 (15.34)

The “constants” for the two terms for rf in Equation 15.32 and Equation 15.33 are dependent on the friction factor and
hence are required to be updated at each iteration. Values of rf are given in Table 15.4.

Table 15.4 The rf values for velocity guesses based on 1.0 fps (0.3048 m/s) for Q+H – equations formulation

Pipe V D Area (m2) Re** f L (m) rf value

(m/s)* (m) (mm) (Swamee and Jain) (Equation 4.123)
(Equation 4.61)
1 0.3048 0.3 0.07069 90,828 0.25 0.02197 1000 746.954
2 0.3048 0.3 0.07069 90,828 0.25 0.02197 1000 746.954
–6 2
*Computed based on 1.0/3.28 for conversion of fps to m. **  = 1.007 10 m /s at 20C ;
0.25 fL
Using in the above Table: f = ---------------------------------------------------------- ; r f = ------------------
log 10  ----------- - + ------------
5.74  2 2 gD A
2
 3.7 D 0.9
Re

Substituting the known quantities and the quantities from Table 12.12 (from the Q – equations formulation example,
Example 12.4, page 303) into Equation 15.32 to Equation 15.34 for the calculation for the first iteration gives:

f 1  m  = – 746.95Q 1 Q 1 +  80.0 – H 1  = 0 for pipe 1 (15.35)

f 2  m  = – 746.95Q 2 Q 2 +  H 1 – 50.0  = 0 for pipe 2 (15.36)

f 3  m  = – Q 1 + Q 2 + 0.050 = 0 at node 1 (15.37)

(assuming a velocity of 1.0 fps (0.3048 m/s) in each pipe for computation of the rf values).

In matrix form (according to Table 9.2) the unknown node incidence matrix A 1 [having 2 rows (NP = 2 for each of the
two pipes) and 1 column (NJ = 1 for the one node)] for the two pipe network is

A1 = –1 (15.38)
1

The transpose of A 1 is

T
A1 = –1 1 (15.39)

The fixed node incidence matrix A 2 [having 2 rows (NP = 2 for the two pipes) and 2 columns (NF = 2 for the two
fixed – head nodes)] for the two pipe network (according to Table 9.4) is

A2 = 1 0 (15.40)
0 –1

___________________________________________________________________________________
Answer – (ii) Develop the Jacobian matrix for the Q+H equations formulation and carry out the first two itera-
tions for the two pipe network.

The terms in the Jacobian at the kth iteration can be determined by taking the appropriate partial derivatives of
Equation 15.32 to Equation 15.34 as follows:

364
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

k  k  k 
 f 1m   f 1m   f 1m 
------------------------- ------------------------- -------------------------
Q 1 Q 2 H 1 k 
– 2.0r f Q 1 0 –
f 2  m   f 2  m   f 2  m  = – nG A 1 =
k  k  k  1
 = ------------------------
k 
m - ------------------------- ------------------------- k  (15.41)
Q 1 Q 2 H 1 T 0 – 2.0r f Q 2 1
A1 0 2

k  k  k  –1 1 0
 f 3m   f 3m   f 3m 
------------------------- ------------------------- -------------------------
Q 1 Q 2 H 1

Note that the Jacobian matrix for the Q+H equations formulation is symmetric in contrast to the Jacobian for the Q –
equations formulation (Equation 12.54, page 304).

ITERATION 1, Q+H equations

k + 1
The system to be solved iteratively for a sequence of m values is:

k  k + 1 k 
Jm m = –f  m  (15.42)

The initial guesses for the flows in the two pipes and the head at node in this example are based on a standard value of
velocity of 1.0 fps (0.3048 m/s) and converted to a flow (same values as for the Q – equations formulation example,
Table 12.11, page 303) and the same value of head guess to commence the iterative solution for the H – equations for-
mulation (Equation 13.67, page 330). These are:

3 3
0 m   0  = 0.02155 m 0
Q1 = 0.02155 -------
- Q2 --------, H 1 = 65.0 m (15.43)
s s

These initial values have been selected to enable a fair comparison of the iterative solution process to be able to be
made between the Q – equations, the H – equations and the Q+H equations. Thus from Equation 15.34 the vector of
unknowns at iteration zero (0) is:

0
Q1
0.02155
0
m = 0 = 0.02155 (15.44)
Q2
0 65.0
H1

Note that it turns out that the sequence of iterations of the Q and H values here is actually independent of the choice of
0
H1 . In this formulation the Jacobian matrix in Equation 15.41 for all iterations is only dependent on the Q values
and has no dependence on H values.

The friction factors (based on the Swamee and Jain equation – Equation 4.61) and corresponding rf values
(Equation 4.123) for the flows in Equation 15.43 for pipes 1 and 2 in the two pipe network are given in Table 12.12
(page 305) and were calculated for the Q – equations formulation. These are:

f = 0.02197
0.02197

and

0 8 f jL j
rf - = 746.95
= ------------------
2 5
 gDj 746.95

The values from Equation 15.44 may be substituted into the pipe head loss and continuity equations of Equation 15.35
to Equation 15.37 as follows:

0
f 1 m  = – 746.954  0.02155  0.02155 +  80.0 – 65.0  = – 0.3468873 + 15.0 = 14.6531 m (15.45)

0
f 2 m  = – 746.954  0.02155  0.02155 +  65.0 – 50.0  = – 0.3468873 + 15.0 = 14.6531 m (15.46)

0 3
f 3 m  = – Q 1 + Q 2 + 0.050 = – 0.02155 + 0.02155 + 0.050 = 0.050 m /s (15.47)

These functions should be zero. Clearly the initially guessed discharges of 21.55 L/s are too small.

365
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

Thus the right hand side vector of Equation 15.42 is:

14.6531
0
f m  = 14.6531
0.050

The Jacobian matrix for the first iteration from Equation 15.41 is:

k 
– 2.0r f Q 1 0 –1 – 2.0  746.954  0.02155 0 –1
1
0
Jm  = k  = 0 – 2.0  746.954  0.02155 1 (15.48)
0 – 2.0r f Q 2 1
2
–1 1 0
–1 1 0

– 32.1937 0 –1
0
Jm  = 0 – 32.1937 1 (15.49)
–1 1 0

The inverse of the Jacobian matrix is shown below for this simple two pipe network (note that this would never be
actually calculated in practice for a network of normal size – an alternative solution method would be used to solve
Equation 15.27 (page 362) that avoids calculating the inverse explicitly).

– 0.015531 – 0.015531 – 0.5

–1 0
J m  = – 0.015531 – 0.015531 0.5 (15.50)
– 0.5 0.5 16.097

The flow and head corrections for the first iteration as calculated from Equation 15.27 are:

– 0.015531 – 0.015531 – 0.5 14.6531 0.48015

1 –1 0 0
m = –J  m f  m  = – – 0.015531 – 0.015531 0.5 14.6531 = 0.43015 (15.51)
– 0.5 0.5 16.097 0.050 – 0.80486

Thus the correction vector for iteration 1 is

1
Q 1
0.48014
1
m = Q 2
1 = 0.43014 (15.52)
1 – 0.80486
H 1

To check on convergence for iteration 1, use Equation 12.25 and Equation 12.47 for the infinity – norm for the govern-
ing path equation as follows

k + 1 max k + 1
h  =  H    stop (15.53)
j=1, NJ

where NJ = 1, s = 1 and k = 0.

Thus for the two pipe network the check is for the head change at node 1 (from Equation 15.52) as

1 1
h  = H1 = – 0.80486 = 0.80486 (15.54)

The following is not satisfied

1 –3
h    stop where  stop = 1.0 10 m

Thus the iterative process has not converged.

The new flows after the first iteration from Equation 15.28 are:

0.02155 0.48014 0.5017

1 0 1
m = m + m = 0.02155 + 0.43014 = 0.4517 (15.55)
65.0 – 0.80486 64.150

366
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

Note the new flows are exactly the same as the flows obtained for iteration 1 for the Q – equations formulation. The
flows satisfy the continuity equation exactly as follows:

1 3
f 3 m  = – Q 1 + Q 2 + 0.050 = – 0.5017 + 0.4517 + 0.050 = 0.000 m /s (15.56)

The continuity equation at node 1 is also exactly satisfied for all subsequent iterations.

The head at iteration 1 at node 1 is 64.150 m, however, this value is different to the value obtained in the H – equations
formulation of 59.708 m (Equation 13.72, page 331).

ITERATION 2, Q+H equations

The vector of new flows and head is given by Equation 15.55 above. The friction factors (based on the Swamee and
Jain equation – Equation 12.31) and corresponding rf values (Equation 4.123) for the flows that resulted from iteration
1 in Equation 15.55 are computed in Table 15.5.
Table 15.5 The rf values for flows from Iteration 1 for the Q – equations formulation

Pipe Q V D Area (m2) Re** f L (m) rf value

(m3/s) (m/s)* (m) (mm) (Swamee and Jain) (Equation 4.123)
(Equation 4.61)
1 0.5017 7.098 0.3 0.07069 2114448 0.25 0.01901 1000 646.73
2 0.4517 6.3909 0.3 0.07069 1903716 0.25 0.01904 1000 647.48

–6 2
**  = 1.007 10 m /s at 20C ;
0.25 fL
Using in the above Table: f = ---------------------------------------------------------- ; r f = ------------------
  5.74  2 2 gD A
2
log 10 ------------ + ------------
 3.7 D 0.9
Re

Thus

1
f = 0.01901 (15.57)
0.01904

and

1 8 f jL j
rf - = 646.73
= ------------------ (15.58)
2 5
 gDj 647.48

The flows and head from Equation 15.55 may be substituted into the pipe head loss and continuity equations of
Equation 15.35 to Equation 15.37 with updates to the resistance factors rf as follows:

1
f 1 m  = – 646.73  0.5017  0.5017 +  80.0 – 64.150  = – 146.93 m for pipe 1 (15.59)

1
f 2 m  = – 647.48  0.4517  0.4517 +  64.150 – 50.0  = – 117.96 m for pipe 2 (15.60)

1
f 3 m  = – 0.5017 + 0.4517 + 0.050 = 0.0 at node 1 (15.61)

or the vector for the right – hand side of Equation 15.42 becomes:

– 146.93
1
fm  = – 117.96 (15.62)
0.0

The Jacobian matrix for the second iteration from Equation 15.41 is:

k 
– 2.0r f Q 1 0 –1 – 2.0  646.73  0.5017 0 –1
1
1
Jm  = k  = 0 – 2.0  647.48  0.4517 1 (15.63)
0 – 2.0r f Q 2 1
2
–1 1 0
–1 1 0

367
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

– 648.93 0 –1
1
Jm  = 0 – 584.93 1 (15.64)
–1 1 0

The inverse of the Jacobian matrix for the second iteration

– 0.0008105 – 0.0008105 – 0.4741

–1 1
J m  = – 0.0008105 – 0.0008105 0.5259 (15.65)
– 0.4741 0.5259 307.632

The flow corrections for the second iteration are calculated from Equation 15.28 by substituting Equation 15.62 and
Equation 15.65 as follows:

– 0.0008105 – 0.0008105 – 0.4741 – 146.93 – 0.2147

2 –1 1 1
m = –J  m f  m  = – – 0.0008105 – 0.0008105 0.5259 – 117.96 = – 0.2147
– 0.4741 0.5259 307.632 0.0 – 7.664

Thus

2
Q1
– 0.2147
2
m = Q2
2 = – 0.2147 (15.66)
2 – 7.664
H1

To check on convergence for iteration 2, use Equation 12.25 and Equation 12.47 for the infinity – norm for the govern-
ing path equation as follows

k + 1 max k + 1
h  =  H    stop (15.67)
j=1, NJ

where NJ = 1, s = 1 and k = 1.

Thus for the two pipe network the check is for the head change at node 1 only (from Equation 15.66) as

1 1
h  = H1 = – 16.1406 = 16.1406 (15.68)

The following is not satisfied

1 –3
h    stop where  stop = 1.0 10 m

Thus the iterative process has not converged.

Note the flow corrections for pipes 1 and 2 are identical as continuity was already satisfied by the flows at the end of
the first iteration.

The new flows after the second iteration are:

0.5017 – 0.047458 0.454242

2 1 2
m = m + m = 0.4517 + 0.0025423 = 0.4491577
64.150 16.1406 48.0094

The iterations continue up to iteration 8 and are summarized below in Answer Part (iii).
___________________________________________________________________________________
Answer – (iii) Tabulate all iterations for the solution process for the Q+H equations formulation

FINAL SOLUTION – ALL ITERATIONS, Q+H equations

The iterations for the full solution based on the Q+H equations formulation are shown in Table 15.6 and Table 15.7.
Eight iterations are shown. The stopping tolerance based on the infinity – norm of the head changes that was used in the
–6
computer simulation run for producing Table 15.6 was 1.0 10 m. Realistically the stopping tolerance could have
–3
been 1.0 10 m or 1 mm which means that convergence would have occurred in 7 iterations.

368
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

Table 15.6 Iteration sequence of flow and head corrections for the Newton solution for the Q+H equations formulation
–6
(stopping tolerance 1.0 10 m achieved at iteration 8) for the two pipe network

Iteration Q 1 m3/s Q 2 m3/s H1 m k + 1

H 
(from Equation 10.22) **
1 0.480142 0.430142 – 0.80486 0.80486
2 – 0.214680 – 0.214680 – 7.664 7.664
3 – 0.088321 – 0.088321 2.826 2.826
4 – 0.023033 – 0.023033 0.736 0.736
5 – 0.002045 – 0.002045 0.064 0.064
6 – 4.35E – 05 – 4.35E – 05 0.001 0.001
7* – 6.39E – 07 – 6.39E – 07 1.68E – 05 1.68E – 05
–3
* A stopping tolerance of less than 1.0 10 m or 1 mm achieved in iteration 7

max k + 1
** Infinity – norm for Head Changes (m) used for testing for convergence =  Hi 
i=1, NJ

Note that the convergence of the Q+H equations in Table 15.6 is faster than for the Q – equations in Table 12.16. The
sequence of Q iterates are virtually identical for both cases but the H values converge faster for the Q+H equations for-
mulation. One iteration is saved.

Table 15.7 Iteration sequence of flows and head for the Newton solution for the Q+H equations formulation
–6
(stopping tolerance 1.0 10 m) for two pipe network

Iteration Q 1 m3/s Q 2 m3/s H1 m k + 1

H 
(from Equation 10.22) **
0 0.02155 0.02155 65.0
1 0.50169 0.45169 64.195 0.804868

2 0.28701 0.23701 56.531 7.664

3 0.19869 0.14869 59.357 2.826

4 0.17566 0.12566 60.093 0.736

5 0.17361 0.12361 60.157 0.0641

6 0.17357 0.12357 60.158 0.00122

7* 0.17357 0.12357 60.158 1.68E – 05

–3
* stopping tolerance of 1.0 10 achieved
___________________________________________________________________________________
Answer – (vi) Plot convergence characteristics for the iterations for the Q+H equations formulation for the two
pipe network.

369
 Chapter 15 - File: Chap14-QH-form-v03.fm July 6, 2022 Version 7

Figure 15.3 Convergence of flows for solution of the two pipe network using the Q+H equations formulation for the Newton
method

Figure 15.4 Convergence of head for solution of the two pipe network using the Q+H equations formulation for the Newton
method

15.5 References
Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1996) FORTRAN numerical recipes
(Vol. 1), Cambridge England.

Rossman, L A. (2000). EPANET 2 Users Manual. U.S. Environmental Protection Agency, Washington,
D.C., EPA/600/R-00/057.

370
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

CHAPTER 16 SOLVING THE GLOBAL GRADIENT

ALGORITHM FORMULATION1

16.1 Introduction
In this Chapter, the Global Gradient Algorithm formulation is used to solve the set of simultaneous non-
linear equations formed to represent flows and heads in a network. As mentioned in Chapter 9, both the
H and Q variables are dealt with simultaneously in the Global Gradient Algorithm formulation. Recall
that in Chapter 15 for the Q+H equations formulation the size of the matrix that requires inversion in the
iterative process is the size of the number of links plus the number of nodes (NP+NJ). A significant
advantage of the Global Gradient Algorithm formulation is that the matrix that requires inversion in the
iterative solution process is smaller and only the size of the number of unknown heads (NJ). In addition,
the matrix to be inverted is symmetric and positive definite.

The speed with which the GGA executes the Newton iterations to solve the set of non-linear equations
governing flows and heads in a water distribution system has probably contributed most to the method’s
popularity (Elhay et al. 2014). The GGA determines the solution of a nonlinear system by a two-stage
iteration in which the linear solver deals with a matrix of dimension only NJ (the number of nodes).
This, together with the fact that the matrix to be inverted is sparse and symmetric, leads to a very fast
algorithm (Elhay et al. 2014).

16.2 Numerical Solution for the Global Gradient Algorithm

The basis for the Global Gradient Algorithm formulation was outlined in Section 9.10. The numerical
implementation of the Global Gradient Algorithm formulation is given by Todini and Pilati (1988) and
Salgado-Castro (1988). These authors refer to the method as a conjugate gradient algorithm. The equa-
tions are solved using a Newton technique applied to simultaneously solving for the heads and flows in
a sequential iterative manner. Todini and Pilati (1988) used an analytical expression for the inverse of a
partitioned matrix (Ayres 1974). An incomplete Choleski conjugate gradient method was then proposed
to solve the sparse system of equations (Kershaw 1978, Ajiz and Jennings 1984).

A different derivation to the Global Gradient Algorithm is given in this chapter based on Simpson and
Elhay (2009).

The matrix form of Global Gradient Algorithm based on the continuity equations (Equation 9.46) and
the pipe head loss equations (Equation 9.61) developed in Chapter 9 are repeated here.

T
A1 q + dm = 0 (16.1)

–G q + A1 h + A2 e L = 0 (16.2)

where

• A 1 = unknown head node incidence matrix (Equation 9.3 and Equation 9.11) of size NP
pipes by NJ nodes
• q = column vector of pipe flows (or other flow elements) (Equation 9.3) (m3/s or
T
ft3/s) =  Q 1 , Q 2 , Q NP  of length NP pipes
• h = column vector of heads for each of the nodes (m or ft)
T
=  H 1 , H 2 , H NJ  of length NJ nodes

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.

371
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

• A 2 = reservoir loop incidence matrix (Section 9.3.2) of size NP pipes by NF fixed – head
nodes
• e L = column vector of reservoir elevations (Equation 9.59) (m or ft)
T
=  EL 1 , EL 2 , ,EL NF  of length NF fixed – head nodes
• d m = column vector of demands for each of the nodes (m3/s or ft3/s)
T
=  DM 1 , DM 2 , DM NJ  of length NJ nodes

A combined partitioned matrix formulation of the two sets of equations in Equation 16.1 and
Equation 16.2 and from Equation 9.65 gives

–G A1 q A e
+ 2 L = 0 (16.3)
T
A1 0 h dm

Recall that the definition of G (an NP by NP diagonal matrix) from Chapter 9 (Equation 9.52) written
in a slightly more general form is

n–1
r 1 Q1 0 0  0 0 0
n–1
0 r 2 Q2 0  0 0 0
n–1
0 0 r 3 Q3  0 0 0
G =       
n–1
0 0 0  r NP – 2 Q NP – 2 0 0
n–1
0 0 0  0 r NP – 1 Q NP – 1 0
n–1
0 0 0  0 0 r NP Q NP

(16.4)

A very important proviso needs to be made about the form of the above matrix definition. These diago-
nal terms in Equation 16.4 are valid for the Hazen – Williams formulation as well as for the Darcy –
Weisbach formulation as long as the flow is not laminar. The appropriate terms for laminar flow are
given in Section 16.2.4.

Equation 16.3 can be rewritten as

Q1
n–1
r 1 Q1 0  0 | Q2
n–1
0 r 2 Q2  0 | 
Q NP A2 e L
–     | –A1 + = 0 (16.5)
n–1 dm
0 0 0 r NP Q NP | H1
––– ––– ––– ––– –– ––– ––– ––– | –– H2
T
–A1 | 0 
H NJ

372
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

The set of equations in Equation 16.3 has a solution if the inverse of G is not singular (Todini and Pilati
1988). The terms on the diagonal of the inverse of the G matrix are

1
------------------------ (16.6)
n–1
rj Qj

Singularity of the inverse of the G matrix occurs if the flow Q j in any pipe becomes very small or zero.
Todini and Pilati (1988) suggested that a lower bound should be defined for all flows to avoid this prob-
lem. As an alternative, a regularization method to overcome the problem of zero flows was developed
by Elhay and Simpson (2011).

16.2.1 Redefinition of quantities in the Global Gradient Algorithm formula-

tion
The Global Gradient Algorithm expressed as Equation 16.3 has a total number of unknowns that is the
sum of the number of pipes plus the number of nodes as

NTOT = NP + NJ (16.7)

In a similar approach to that in Chapter 15 (Section 15.2.1), define in Equation 16.3 the following quan-
tity to enable a Taylor’s series expansion to be written thus hence a Newton formulation to be derived

q
m = (16.8)
h

or

T
m =  m 1 m 2  m NTOT  (16.9)

or

T
m =  Q 1 Q 2  Q NP H 1 H 2  H NJ  (16.10)

Also define

A2 e L
p = (16.11)
dm

It follows that substituting Equation 16.8 and Equation 16.11 into Equation 16.3 gives

–G A1
m + p = 0 (16.12)
T
A1 0

For the column vector m the following set of NTOT – functions made up of the pipe flow equations and
the nodal continuity equations can be written as

f 1  m 1 m 2  m NTOT  = 0

373
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

f 2  m 1 m 2  m NTOT  = 0

f NTOT  m 1 m 2  m NTOT  = 0 (16.13)

where

• f s  m 1 m 2  m NTOT  = 0 or f s  Q 1 Q 2  Q NP H 1 H 2  H NJ = sth function in

the Global Gradient Algorithm formulation (either a pipe head loss equation or a continuity
equation)
• Q j = flow in pipe j (or other flow element) (m3/s or ft3/s)
• H i = nodal head at node i (m or ft)

The first NP functions are the pipe head loss equations that have the general form

n–1
–r j Q j Q j + Hi – Hk = 0 for pipes j = 1,..., NP (16.14)

where one of the heads may be a reservoir elevation if the pipe is attached to the reservoir. The remain-
ing functions in Equation 16.13 are NJ nodal continuity equations from Equation 9.15 of the form

NP
  a ij Q j  + DM i = 0 for nodes i = 1,2,..., NJ (16.15)
j=1

where

• a ij = element of the unknown head node incidence matrix A 1 (with values zero, – 1 or
+1), thus only Qj values for pipes connected to the node are included.
Let the true solution to the set of nonlinear equations in Equation 16.13 be the following column vector

* * * * T
m =  m 1 m 2  m NTOT  (16.16)

or

* * * * * * * T
m =  Q 1 Q 2  Q NP H 1 H 2  H NJ  (16.17)

Let the deviation or correction column vector for the unknown flows and heads problem be defined as

T
m =  m 1 m 2  m NTOT  (16.18)

or

T
m =  Q 1 Q 2  Q NP H 1 H 2  H NJ  (16.19)

This is the quantity that needs to be solved for in the Newton iterative scheme.

374
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

16.2.2 The Jacobian matrix for Global Gradient Algorithm formulation

The solution for the unknown flows and heads requires linearization of the nonlinear governing equa-
tions in Equation 16.13. In the neighborhood of the current iteration point each of the functions in
Equation 16.13 may be expanded in a Taylor series as shown earlier in Equation 10.8

NTOT  f m
1
f 1  m + m  = f 1  m  + -m j + O  m 
 ------------------
m j
j=1

NTOT  f m
2
f 2  m + m  = f 2  m  + -m j + O  m 
 ------------------
m j
j=1

NTOT  f m
NTOT
f NTOT  m + m  = f NTOT  m  + - m j + O  m 
 ------------------------------ (16.20)
m j
j=1

where

• f s  m , f s  m + m  = sth function in the GGA equation formulation (either a pipe head loss
or a continuity equation)
• m = column vector of unknown heads (m or ft) and flows (m3/s or ft3/s)
• m = deviation vector of the unknown heads and flows; m s = sth deviation of heads and
flows in the m vector
• NP = number of pipes in the network
• NJ = number of nodes with unknown heads in the network
• NTOT = NP +NJ
2
• O  m  = column vector of second order and higher terms of the flows and heads correc-
tions
Rewriting Equation 16.20 in matrix notation using Equation 10.9 (Press et al. 1996) gives

2
f  m + m  = f  m  + J  m m + O  m  (16.21)

where

• J  m  = Jacobian matrix based on the flows and heads.

From the set of equations in Equation 16.20, the Jacobian matrix for the flow equation formulation
based on the values of the flows at the iteration k is defined as

375
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

k  k  k 
 f 1m   f 1m   f 1m 
-------------------------- -------------------------
-  --------------------------
m 1 m 2 m NTOT
k  k  k 
k 
 f 2m   f 2m   f 2q 
m  = -------------------------- -------------------------
-  ------------------------ (16.22)
m 1 m 2 m NTOT
   
k  k  k 
 f NTOT  m   f NTOT  m   f NTOT  m
------------------------------------- -------------------------------------  ------------------------------------
-
m 1 m 2 m NTOT

or in matrix partition form as shown below as

k  k  k  k  k  k 
 f 1m   f 1m   f 1m   f 1m   f 1m   f 1m 
------------------------
- ------------------------
-  ------------------------
- | ------------------------- -------------------------  ------------------------
-
Q 1 Q 2 Q NP H 1 H 2 H NJ
k  k  k  k  k  k 
 f 2m   f 2m   f 2q   f 2m   f 2m   f 2m 
------------------------
- ------------------------
-  ------------------------ | ------------------------- -------------------------  ------------------------
-
Q 1 Q 2 Q NP H 1 H 2 H NJ
    |    
k  k  k  k  k  k 
 f NP  m   f NP  m   f NP  m   f NP  m   f NP  m   f NP  m 
----------------------------
- ----------------------------
-  ----------------------------- | ----------------------------- -----------------------------  -----------------------------
Q 1 Q 2 Q NP H 1 H 2 H NJ
k 
m  = ––– ––– –– ––– ––– –– –– ––– ––– –– | ––– ––– –– ––– ––– –– –– ––– ––– –
k  k  k  k  k  k 
 f NP + 1  m   f NP + 1  m   f NP + 1  m   f NP + 1  m   f NP + 1  m   f NP + 1  m
-----------------------------------
- ------------------------------------  -----------------------------------
- | -----------------------------------
- ------------------------------------  ---------------------------------
Q 1 Q 2 Q NP H 1 H 2 H NJ
k  k  k  k  k  k 
 f NP + 2  m   f NP + 2  m   f NP + 2  m   f NP + 2  m   f NP + 2  m   f NP + 2  m
-----------------------------------
- ------------------------------------  -----------------------------------
- | -----------------------------------
- ------------------------------------  ---------------------------------
Q 1 Q 2 Q NP H 1 H 2 H NJ
    |    
k  k  k  k  k  k
 f NP + NJ  m   f NP + NJ  m   f NP + NJ  m   f NP + NJ  m   f NP + NJ  m   f NP + NJ  m
- -----------------------------------------  ----------------------------------------
---------------------------------------- - | ----------------------------------------- -----------------------------------------  ------------------------------------
Q 1 Q 2 Q NP H 1 H 2 H NJ

(16.23)

2
If m is small then second order and higher terms of O  m  can be dropped and also by setting in
Equation 16.21

*
f  m + m  = f  m  = 0

gives

k  k + 1 k 
Jm m = –f  m  (16.24)

Equation 16.24 can be expressed as:

k + 1
m 1 k  k  k 
f 1  m 1  m 2   m NTOT 
k + 1
m 2 k  k  k 
k 
f 2  m 1  m 2   m NTOT 
Jm   = – (16.25)

 
k  k  k 
k + 1 f NTOT  m 1  m 2   m NTOT 
m NTOT

or

376
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

k + 1
Q 1
k + 1
Q 2
k 
 f 1  Q 1 Q 2  Q NP H 1 H 2  H NJ 
k + 1 k 
k  Q NP f 2  Q 1 Q 2  Q NP H 1 H 2  H NJ 
Jm  = – (16.26)
k + 1 
H 1

k + 1 k 
H 2 f NTOT  Q 1 Q 2  Q NP H 1 H 2  H NJ 

k + 1
H NJ

Note the subscripts k and (k + 1) have been introduced to represent the successive iteration numbers.
k + 1
The flows and heads correction vector m on the left – hand side of Equation 16.24 is for the
k  k 
(k + 1)st iteration. The Jacobian matrix J  m  and the function vector f  m  are the Q+H equations
k 
formulation for the flows and heads m at iteration k.

k + 1
The vector of corrections to the flows m may be expressed from Equation 16.24 as

k + 1 –1 k  k 
m = –J  m f  m  (16.27)

where

–1 k 
• J  m  = inverse of the Jacobian matrix for the Global Gradient Algorithm formulation
at iteration k (the Jacobian matrix is defined by Equation 16.22)
As mentioned previously Equation 16.27 would never normally be used because of the computational
intensity required in obtaining the inverse of the Jacobian. Equation 16.26 would be solved instead.

k + 1
Once the vector of corrections m at the (k + 1)st iteration is determined it is then added to the
k 
current solution vector of flows and heads m to give the next best estimate of the unknowns as

k + 1 k k + 1
m = m + m (16.28)

However, this formulation is different from that in Chapter 15, in that the H – component of the
k + 1
m vector is solved separately from the Q – component. This will be seen later on in the Chapter.

16.2.3 Calculation of the Jacobian matrix elements for the Global Gradient
Algorithm
To evaluate the Jacobian elements the functions given Equation (given below in Equation 16.30) needs
to be evaluated term by term.

377
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

Q1
n–1
–r 1 Q1 0  0 | Q2
n–1
0 –r 2 Q2  0 | 
   | A 1 Q NP
 A2 e L
+ = 0 (16.29)
n–1 H1 dm
0 0 0 – r NP Q NP |
– – – – – – – – – – – – – – – – – – – – – – – | – – H2
T 
A1 | 0
H NJ

Thus in matrix form

–G A1 q A e
+ 2 L = 0 (16.30)
T
A1 0 h dm

or

– Gq A 1 h A2 e L
+ = 0 (16.31)
T
A1 q 0 dm

Consider derivatives of each block in turn whilst keeping in mind the general form of the Jacobian given
in Equation 16.23.

1. Derivative of Gq in the upper left – block in Equation 16.31 with respect to Q j

This block is the pipe equations of the form of Equation 16.14. To determine the Jacobian elements in
Equation 16.23, this block needs to be differentiated with respect to Q j variables. Each diagonal term in
n–1
this matrix block of Gq is r j Q j Q j. Only the diagonal terms of the Gq block are non – zero thus
Jacobian elements along the diagonal are

k 
 f jm  n–1
------------------------ = nr j Q j (16.32)
Q j

All other Jacobian elements in this NP by NP block of Gq are zero. Thus the Jacobian part of this block
is

nG (16.33)

2. Derivative of A 1 h in the upper right – block with respect to H i

The Jacobian elements given by the derivative with respect to H i of this block in Equation 16.23 or
Equation 16.31 or of A 1 h in Equation 16.31 is just

k 
 f jm 
------------------------- = A 1 (16.34)
H i

378
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

T
3. Derivative of A 1 q in the lower left – block with respect to Q j

The Jacobian elements given by the derivative with respect to Q j of this block in Equation 16.23 or
T
Equation 16.31 of A 1 q in Equation 16.31 is just

k 
 f jm  T
------------------------- = A 1 (16.35)
Q i

4. Derivative of zero in the lower right – block with respect to H i

Clearly the derivative of the lower right – hand block will be zero n Equation 16.23 or Equation 16.31.

Thus the final expression for the Jacobian matrix for the Global Gradient Algorithm is

k 
k  –n G A1
Jm  = (16.36)
T
A1 0

where

k 
• Jm  = Jacobian matrix at iteration k based on the flows and heads.
k 
• m = column vector of unknown heads (m or ft) and flows (m3/s or ft3/s) at iteration k
k  n–1
• G is the a diagonal matrix with terms along the diagonal for each pipe j of r j Q j as
defined in Equation 16.4 at iteration k
• A 1 = unknown head node matrix as defined in Section 9.3 and Equation 9.11.

Note that the Jacobian in Equation 16.36 only depends on values of Qj and does not depend on values of
Hi at iteration k.

16.2.4 Distinguishing between laminar and turbulent flow

First define the inverse of G at the k – th iteration as

1
---------------------------- 0 0  0 0 0
k  n – 1
r 1 Q1
1
0 ---------------------------- 0  0 0 0
k  n – 1
r 2 Q2
–1
k  =      (16.37)
G
1
0 0 0  0 -----------------------------------------------
- 0
k  n–1
r NP – 1 Q NP – 1
1
0 0 0  0 0 ---------------------------------
k  n – 1
r NP Q NP

–1
Let each term along the diagonal of the inverse G be designated as

379
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

1
 j = ---------------------------- (16.38)
k  n – 1
rj Qj

Thus

–1 · 
G = diag   1  2  NP – 1  NP  (16.39)

Obviously, zero flows in Equation 16.37 are going to lead to an extremely large condition number and
–1
hence a singularity of the G matrix, particularly for the Hazen – Williams head loss formulation.
However, there it is useful to differentiate between laminar and turbulent flow for the diagonal terms in
–1
G for the Darcy – Weisbach formulation. Traditionally both laminar and turbulent flow are dealt with
in the same way through the Darcy – Weisbach head loss equation from Equation 4.3.

For laminar flow the friction factor is defined as

64 64 64 64 16D
f = ------- = --------------- = ------------------------- = ----------------------------------- = ----------------- (16.40)
Re VD   QD   A  4QD    D  
2 Q

The problem in computing diagonal elements in Equation 16.37 with zero flows for the laminar flow
case can be avoided if the Darcy – Weisbach equation for the head loss is reformulated by combining
Equation 16.40 and Equation 4.3 as follows

2 2
L Q 16D L Q 16  16 LQ 128LQ
h f = f ---- ------------- = ----------------- ---- -------------------------- = -------------------------------- = -------------------- = r L Q (16.41)
D 2g A 2 Q D 2 2 2 4 4
2g  D
2g   D  g D
----------
 4 

or

h f = r LQ (16.42)

where the laminar resistance term r L is defined by

128L
r L = ---------------- (16.43)
4
g D

Note that in Equation 16.41 the head loss is linearly related to the discharge. As a result the term on the
–1
diagonal of the G matrix is the constant r L rather than r f Q . Thus in the matrix inverse G is a pipe
has laminar flow then the diagonal term for that pipe will be

1
----- (16.44)
rL

and because this is independent of flow for laminar conditions (unlike turbulent flow) then this term in
–1
the G matrix will not cause a singularity to occur for normal sizes of values in Equation 16.43.

380
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

16.2.5 Definition of the Schur Complement

Now define the following term as

T –1
U = A1 G A1 (16.45)

THE U MATRIX FOR THE GLOBAL GRADIENT ALGORITHM

T –1
where – A 1 G A 1 is the Schur complement.

The diagonal elements of the U matrix are of the form

NP
 a ij  j for node i (16.46)
j=1

where

• NP = number of pipes in the network

T
• a ij = elements corresponding to node i in the appropriate row of A 1 (the transpose of the
unknown head node incidence matrix which only has values zero, – 1 or +1) that show
which pipes are connected to node i
1 –1
•  j = ---------------------------
- = terms along the diagonal of the G matrix
k  n – 1
rj Qj

k  n – 1
Thus for a particular node i all the  j with a value of 1   r j Q j  or 1  r L for the pipes connected
to node i are added up to give the diagonal elements of the U matrix.

Consider the example of the ten pipe network that has a total of 7 nodes and 10 pipes as shown in
Figure 16.1. Assume the flows in all pipes are not laminar and that the Darcy – Weisbach head loss equa-
tion is used.

Figure 16.1 The ten pipe network

381
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

Node 2 has pipes 1, 2 and 5 connected to it. Thus the diagonal element of U in the first row (assuming
the rows are in ascending order of numeric numbers) of the matrix for node 2 is

NP NP 1 NP 1 1
 a ij  j =  a 2 j  j =  1 +  2 +  5 =  a ij  j = --------------
r Q
- + --------------- + --------------
r Q r
(16.47)
=1 j=1 j=1 1 1 2 2 5 Q5

Denote this sum for node 2 by

1 1 1
 2 = --------------- + --------------- + --------------- (16.48)
r 1 Q1 r 2 Q2 r 5 Q5

The off – diagonal entries in U for a particular node i only occur at positions associated with the nodes k
(that are not fixed head nodes) that have pipes connected to node i. The entry is usually only one value
k  n – 1
being the negative of the value  j or 1   r j Q j  for the connected pipe placed in the column
associated with the connecting node k (except for parallel pipes between the same nodes where there
will be more than one value). For example in the ten pipe network for node 2 nodes 3 and 6 are con-
nected and thus the matrix U will have elements in the columns corresponding to these nodes.

Thus for node 3, pipe 2 is connected thus the entry in the matrix U for the column corresponding to node
2 of the first row is

–2 (16.49)

and for node 6, pipe 5 is connected to node 2 thus the entry in the matrix U for column corresponding to
node 6 of the first row is

–5 (16.50)

T –1
Computation of the U = A 1 G A 1 matrix for the ten pipe network gives

1 0 0 0 0 0 0 0 0 0
0 2 0 0 0 0 0 0 0 0 –1 0 0 0 0 0 0
1 –1 0 0 0 0 0
–1 1 0 0 1 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0
0 1 –1 0 0 0 0
0 –1 1 0 0 1 0 0 0 0 0 0 0 4 0 0 0 0 0 0
0 0 –1 0 0 0 0
0 0 –1 –1 0 0 1 0 0 0 0 0 0 0 5 0 0 0 0 0 1 0 0 –1 0 0 0
0 0 0 0 –1 0 0 1 0 0
0 0 0 0 0 6 0 0 0 0 0 1 0 0 –1 0 0
0 0 0 0 0 –1 0 –1 1 0
0 0 0 0 0 0 7 0 0 0 0 0 1 0 0 –1 0
0 0 0 0 0 0 –1 0 –1 1
0 0 0 1 –1 0 0
0 0 0 0 0 0 0 0 0 –1 0 0 0 0 0 0 0 8 0 0
0 0 0 0 1 –1 0
0 0 0 0 0 0 0 0 9 0
0 0 0 0 0 1 –1
0 0 0 0 0 0 0 0 0  10

Multiplying these matrices out with the diagonal elements defined by Equation 16.48 gives

382
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

1 –2 0 –5 0 0 0
–2 2 –3 0 –6 0 0
0 –3 3 0 0 –7 0
T –1
U = A1 G A1 = – 0 0 4 –8 0 0 (16.51)
5
0 –6 0 –8 5 –9 0
0 0 –  7 0 –  9  6 –  10
0 0 0 0 0 –  10  7

where

NP
• j =  a ij  j
j=1

1
•  j = ------------------------.
n–1
rj Qj

The inverse of the matrix U needs to be dealt with in the Global Gradient Algorithm iterative procedure.
The terms along the diagonal of matrix U in Equation 16.51 are shown in Table 16.1.
Table 16.1 Diagonal terms of matrix U for Darcy – Weisbach head loss equation
Column/  j in Equation 16.51
Row
1 1 1
 2 +  5 = --------------- + ---------------
r 2 Q2 r 5 Q5
2 1 - + --------------
1 - + --------------
1 -
 2 +  3 +  6 = --------------
r 2 Q2 r 3 Q3 r 6 Q6

3 1 - + --------------
1 -
 3 +  7 = --------------
r 3 Q3 r 7 Q7
4 1 1
 5 +  8 = --------------- + ---------------
r 5 Q5 r 8 Q8

5 1 1 1
 6 +  8 +  9 = --------------- + --------------- + ---------------
r 6 Q6 r 8 Q8 r 9 Q9

6 1 1 1
 7 +  9 +  10 = --------------- + --------------- + --------------------
r 7 Q 7 r 9 Q 9 r 10 Q 10

7 1
 10 = --------------------
r 10 Q 10

16.2.6 The Global Gradient Algorithm two – step update vectors

The column vector of flow and head corrections is given from Equation 16.24 as

k + 1
k  k 
J  q  k  h  k   q = –f  q  h  (16.52)
k + 1
h

383
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

k  k 
Consider the function term f  q h  on the right – hand side of Equation 16.24 from Equation 16.3

k  k  k 
k  k  f1  q h  –G A1 q k  A e
fq h  = = + 2 L (16.53)
k  k  T k  dm
f2  q h  A1 0 h

k  k 
–G q + A1 h + A2 e L
=
T k 
A1 q + dm

Substituting Equation 16.36 and Equation 16.53 and introducing the unknowns into Equation 16.52
gives

k  k  k 
–n G A1  q k + 1  –G q + A1 h + A2 e L
= – (16.54)
T k + 1 T k 
A1 0 h A1 q + dm

where the superscripts k and (k+1) refer to the iteration number. This is equivalent to Equation (9) in the
T
Todini and Pilati (1988) paper. Note that matrices A 1 , A 2 and A 1 are all independent of the iteration
number as too are the vectors e L and d m . Thus as a result these matrices and vectors do not require a
iteration counter superscript.

The following derivation is based on that given by Simpson and Elhay (2009)1.

Substituted the following into Equation 16.54

k + 1 k + 1 k 
q =
q –q
(16.55)
k + 1 k + 1 k 
h h –h

giving

k  k  k 
–n G A1 q k + 1  – q k  –G q + A1 h + A2 e L
= – (16.56)
T k + 1 k  T k 
A1 0 h –h A1 q + dm

Thus it follows that

k  k  k  k 
–n G A1 q k + 1  –n G A1 q k  –G q + A1 h + A2 e L
= – (16.57)
T k + 1 T k  T k 
A1 0 h A1 0 h A1 q + dm

k + 1 k + 1 k  k  k  k 
– n Gq + A1 h – n Gq + A1 h –G q + A1 h + A2 e L
= – (16.58)
T k + 1 T k  T k 
A1 q A1 q A1 q + dm

1. Thanks go to Dr. Sylvan Elhay, School of Computer Science, University of Adelaide for his elegant
derivation of the Global Gradient Algorithm equations

384
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

or writing out the individual equations

k + 1 k + 1 k  k  k  k 
– n Gq + A1 h = – n Gq + A1 h –  –G q + A1 h + A2 e L  (16.59)

k + 1 k + 1 k 
– n Gq + A1 h =  1 – n Gq – A2 e L (16.60)

k + 1 k + 1 k 
A1 h = nGq –   n – 1 Gq – A2 e L  (16.61)

and

T k + 1
A1 q = –dm (16.62)

STIIL HAVE TO ALTER THE SECTIONS BELOW WITH NEW SIGN CONVENTION

T –1 T –1
Multiply both sides of Equation 16.61 by A 1 G and use U = A 1 G A 1

T –1 k + 1 T –1 k + 1 k 
A1 G  A1 h  = A 1 G  nGq –  n – 1 Gq + A2 e L  (16.63)

Thus

k + 1 T –1 k + 1 T –1 k 
Uh = nA 1 G Gq – A 1 G   n – 1 Gq + A2 e L  (16.64)

or

k + 1 T k + 1 T –1 k  T –1
Uh = nA 1 q –  n – 1 A 1 G Gq + A1 G A2 e L (16.65)

Substituting from the continuity equation of Equation 16.62 into the first term on right hand side of
Equation 16.65 gives

k + 1 T k  –1
Uh = – n d m + A 1   1 – n q – G A2 e L  (16.66)

Thus

k + 1 –1  T k  –1 
h = U  – n d m + A 1   1 – n q – G A 2 e L   (16.67)
 

GLOBAL GRADIENT ALGORITHM HEAD VECTOR UPDATE EQUATION

–1 –1
Note that in Equation 16.67 the only difficult inverse to compute is U . The other inverse G
–1 k 
involves a diagonal matrix which is very easy to invert. G is known based on the previous q vec-
tor value while matrices A 1 and A 2 as well as vectors e L and d m are all known at the very beginning
of the iterative solution procedure and remain constant throughout.

385
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

Now use Equation 16.61 to solve for the discharge

k + 1 k + 1 k 
A1 h = nGq –  n – 1 Gq + A2 e L (16.68)

k + 1 k  k + 1
nGq =  n – 1 Gq – A2 e L + A1 h (16.69)

k + 1 1 –1  k  k + 1 
q = --- G   n – 1 Gq – A 2 e L + A 1 h  (16.70)
n  

Thus

 k + 1 
= 1---   n – 1 q + G  – A 2 e L + A 1 h
k + 1 k  –1
q  (16.71)
n 

GLOBAL GRADIENT ALGORITHM FLOW VECTOR UPDATE EQUATION

–1
This time the only inverse in Equation 16.71 is for a diagonal matrix G . This makes computation of
k + 1
the updates flows q very easy. Thus Equation 16.71 is very straightforward to evaluate as a series
of matrix multiplications.

Thus the recursive Newton algorithm includes Equation 16.45, Equation 16.67 and Equation 16.71

T –1
U = A1 G A1 (16.72)

k + 1 –1  T k  –1 
h = U  – n d m + A 1   1 – n q – G A 2 e L   (16.73)
 

k + 1 1 k  –1 k + 1 
q = ---   n – 1 q + G  – A 2 e L + A 1 h  (16.74)
n 

GLOBAL GRADIENT ALGORITHM RECURSIVE FORMULATION

16.3 Solving the Global Gradient Algorithm Equations for the Two
Pipe Network
16.3.1 Unknowns and initial guesses
Now consider the flow and head equations for the network in Figure 16.2. The unknowns that need to be
solved for are pipe flows Q1 and Q2 and the head at node 1, H1..

386
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

Figure 16.2 Unknown Global Gradient Algorithm Qs and H to be solved for in the two pipe net-
work

The functions to be solved include two head loss equations for the pipes and one continuity equation at
node 1, however, these are expressed in the block form as proposed by Todini and Pilati.

From Equation 16.10 (page 373), the column vector of unknown flows and at the kth iteration for the two
pipe network is

k 
Q1
k  k 
m = Q2 (16.75)
k 
H1

Thus the heads and flows to be solved successively at each iterative step (from Equation 16.67) and
Equation 16.71 for the two pipe network at the (k + 1)st iteration becomes

k + 1 k + 1
h = H1  and then (16.76)

k + 1
k + 1 Q1
q = (16.77)
k + 1
Q2

A column vector of initial estimates for the flows only in each pipe (initial estimates of the heads are not
needed) is required to be made in order for the iterative solution process to begin

0
0 Q1
q = (16.78)
0
Q2

0
The initial flows Q j are often based on a flow velocity of 1.0 fps (0.3048 m/s) (as used in EPANET
Rossman 2000) as shown in Table 12.11 (page 303, used as initial guesses for the Q – equations and for
the Q+H Q+H equations formulations). The pipe properties for the network in Figure 16.2 are given in
Table 12.8 (page 301) and Table 12.9.

The pipe resistance terms ( r f =  r 1 r 2  ) can now be computed based on initial flow guesses shown in
2
Table 12.11 for the Darcy – Weisbach head loss equation h f = r f Q (where n = 2) using
j

Equation 12.30. It is assumed that the flows are turbulent and not laminar in this example. The pipe
resistance values r j are shown in Table 12.12 (page 305) (for the Q – equations formulation) with the
friction factors computed from the Swamee and Jain equation. These pipe resistance values r j are
updated at each iteration based on the latest flow values.

387
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

16.3.2 Example based on the Global Gradient Algorithm

Example 16.1 For the two pipe network in Figure 16.2 complete the following:

(i) Develop the Global Gradient Algorithm for the two pipe network.
(ii) Carry out the first two iterations of the Global Gradient Algorithm solution for the two pipe network.
(iii) Tabulate all iterations to convergence for the solution process for the Global Gradient Algorithm such that the
infinity – norm is less than 10 – 3 m.
(iv) Plot the convergence characteristics for the iterations for the Global Gradient Algorithm for the two pipe net-
work.
Assume the Darcy – Weisbach head loss equation is used for each pipe.

Answer – (i) Develop the Global Gradient Algorithm for the two pipe network.

The vector of unknowns is

k 
Q1
k 
m = k  (16.79)
Q2
k 
H1

The Global Gradient Algorithm formulation of the governing equations for this two pipe network includes two pipe
head loss equations (Equation 16.14) – one for each pipe, and a continuity equation at node 1 (Equation 16.15) as fol-
lows:

k  k  k  k 
f 1 m  = r 1 Q1 Q1 – EL 2 + H 1 = 0 for pipe 1 (16.80)

k  k  k  k 
f 2 m  = r 2 Q2 Q2 – H1 + EL 3 = 0 for pipe 2 (16.81)

k  k 
f 3 m  = Q1 –Q2 – D M 1 = 0 at node 1 (16.82)

k 
The “constants” for the two terms for r j in Equation 16.80 and Equation 16.81 are dependent on the friction factor
and hence are required to be updated at each iteration.

In matrix form the unknown node incidence matrix A 1 [having 2 rows (NP = 2 for each of the two pipes) and 1 col-
umn (NJ = 1 for the one node)] for the two pipe network is

A1 = –1 (16.83)
1

The transpose of A 1 is

T
A1 = –1 1 (16.84)

The fixed node incidence matrix A 2 [having 2 rows (NP = 2 for the two pipes) and 2 columns (NF = 2 for the two
fixed – head nodes)] for the two pipe network is

A2 = 1 0 (16.85)
0 –1

The demand vector [having 1 row (NJ = 1 for the one node) and one column] is

d m =  DM 1  = 0.050 m3/s (16.86)

The fixed head elevation vector [having 2 rows (NF = 2 for the two reservoirs)] is

388
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

E L1
eL = = 80.0 m (16.87)
E L2 50.0

Thus the set of equations to be solved using the Global Gradient Algorithm from Equation 16.12 is

k 
G | –A1 G | –A1 Q1
A2 e L
– – – | – – – m – p = – – – | – – – Q2
k  – = 0 (16.88)
T T dm
–A1 | 0 –A1 | 0 H1
k 

The diagonal G matrix (Equation 16.4) and the inverse of the G matrix (Equation 16.37) for this two pipe network are

k 
r 1 Q1 0
G = (16.89)
k 
0 r 2 Q2

and the inverse from Equation 16.36 is

1
------------------ 0
k 
–1 r 1 Q1
G = (16.90)
1
0 ------------------
k 
r 2 Q2

Substituting values into Equation 16.88 gives:

k  r 1 Q1 0 k 
G | –A1 Q1 | – –1 Q1 1 0 80.0
k 
A2 e L 0 r 2 Q2 1 k  0 – 1 50.0
––– | ––– Q2 – = Q2 – (16.91)
T dm ––– | ––– –––
–A1 | 0 H1
k 
H1
k 
– –1 1 | 0 0.050

r 1 Q1 0 k 
| – –1 Q1
80.0
0 r 2 Q2 1 k 
= Q2 – – 50.0 = 0
––– | ––– 0.050
k 
H1
– –1 1 | 0

The Global Gradient Algorithm two step solution for each iteration is given from Equation 16.67 as

k + 1 –1 T k  –1
h = U  – n d m + A 1   1 – n q – G A 2 e L   (16.92)

or

k + 1 T –1 –1 T k  –1
h =  A 1 G A 1   – n d m + A 1   1 – n q – G A2 e L   (16.93)

Now substituting values into Equation 16.93 and also n = 2 gives

–1
 1    1 
 ------------------
k 
0    k 
------------------
k 
0 
k + 1  r 1 Q1 – 1    Q r Q 1 0 80.0 
h =  –1 1   – 2 0.050 + – 1 1   1 – 2  1 – 1 1

 1 1    Q2
 k  1 0 – 1 50.0  
 0 ------------------    0 ------------------ 
k  k 
 r 2 Q2    r 2 Q2 

or

389
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

–1
 1    1 
 ----------------------
k  k 
0    k 
----------------------
k  k 
0 
k + 1  r Q1  
– 1   – 0.100 + – 1 1  – Q 1 – r 1 Q 1 1 0 80.0   (16.94)
h =  –1 1 1  
 1 1    Q k  1 0 – 1 50.0  
 0 ---------------------
-    2 0 ---------------------
- 
k  k  k  k 
 r 2 Q2    r 2 Q2 

Once the head vector has been found, the unknown Qs can be solved for from Equation 16.71 as

k + 1 1 k  –1 k + 1 
q = ---   n – 1 q + G  A 2 e L + A 1 h  (16.95)
n 

or by expanding the last bracketed term as

k + 1 1 k  –1 –1 k + 1
q = ---   n – 1 q + G A 2 e L + G A 1 h  (16.96)
n

Again substituting quantities into Equation 16.95 and also n = 2 gives

 1 - 
 ----------------- 0 
k  k 
k + 1 1  Q1 r 1 Q1  1 0 80.0 – 1  
q = ---   2 – 1  +  + H1
 k + 1   (16.97)
2 k  1  0 – 1 50.0 1  
Q2 0 ------------------
 k  
 r 2 Q2 

 1 - 1 - 
 k  ----------------- 0 ----------------- 0 
k  k 
k + 1  Q1 r 1 Q1 r
1 0 80.0 + 1 1 Q –1 
q = 0.5  + k + 1  (16.98)
H1
 Q  k 
0 1 - 0 – 1 50.0
----------------- 0 1 - 1
----------------- 
 2 k  k  
 r 2 Q2 r 2 Q2 

___________________________________________________________________________________
Answer – (ii) Carry out the first iteration of the Global Gradient Algorithm solution for the two pipe network.

ITERATION 1, Global Gradient Algorithm formulation

The initial guesses for the flows in the two pipes in this example are based on a standard value of velocity of 1.0 fps
(0.3048 m/s) and converted to a flow (same values as for the Q – equations formulation example, Table 12.11,
page 303). These are:

3
0 m 0 3
Q1 = 0.02155 -------- Q 2 = 0.02155 m /s (16.99)
s

Thus from Equation 12.55 the vector of initial guesses of the unknowns at iteration zero (0) is:

0
0 Q1
= 0.02155
3
q = m /s (16.100)
0 0.02155
Q2

The friction factors (based on the Swamee and Jain equation – Equation 4.61) and corresponding rf values
(Equation 4.123) for the flows in Equation 16.100 for pipes 1 and 2 in the two pipe network are given in Table 12.12
(page 305) and were calculated for the Q – equations formulation. These are:

0
f = 0.02197
0.02197

and

0
rf = 746.95 (16.101)
746.95

as shown in Table 16.2.

390
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

Table 16.2 The rf values for initial velocity guesses based on 1.0 fps (0.3048 m/s) for Q – equations formulation

Pipe V D Area (m2) Re** f L (m) rf value

(m/s)* (m) (mm) (Swamee and Jain) (Equation 4.123)
(Equation 4.61)
1 0.3048 0.3 0.07069 90,828 0.25 0.02197 1000 746.954
2 0.3048 0.3 0.07069 90,828 0.25 0.02197 1000 746.954
–6 2
*Computed based on 1.0/3.28 for conversion of fps to m. **  = 1.007 10 m /s at 20C ;
0.25 fL -
Using in the above Table: f = ---------------------------------------------------------
- ; r f = -----------------
  5.74  2 2 gD A
2
log 10 ------------ + ------------
 3.7 D 0.9
Re

Determining the Global Gradient Algorithm h vector values for iteration 1:

The values from Equation 16.100 and Equation 16.101 based on estimates at the zeroth iteration may be substituted
into the expression for the inverse of the G matrix in Equation 16.90:

1
---------------------- 0 1
0 0
r 1 Q1 --------------------------------------- 0
= 746.95 0.02155 = 0.062122 0
–1
G = (16.102)
1 1 0 0.062122
0 ---------------------- 0 --------------------------------------
-
0 0 746.95 0.02155
r 2 Q2

or

= 0.062122 0
–1
G (16.103)
0 0.062122

Now solve for the unknown head at node 1 by using Equation 16.94 as
–1
 1    1 
 ----------------------
0 0
0    0
----------------------
0 0
0 
1  r Q1  
– 1   – 0.100 + – 1 1  – Q 1 – r 1 Q 1 1 0 80.0  
h =  –1 1 1  
 1 1    Q 0  1 0 – 1 50.0  
 0 ---------------------
-    2 0 ---------------------
- 
0 0 0 0
 r 2 Q2    r 2 Q2 

Substituting for G – 1 from Equation 16.102 at the zeroth iteration and q(0) gives

–1
1  – 1   – 0.100 + – 1 1  – 0.02155 – 0.062122 1 0 80.0  
h =  – 1 1 0.062122 0

0

 0 0.062122 1    0.02155 0 0.062122 0 – 1 50.0  

(16.104)

Multiplying out matrices gives:

–1
1     80.0  
h =  – 0.062122 0.062122 – 1   – 0.100 + – 1 1  – 0.02155 – 0.062122 0

 1    0.02155 0 – 0.062122 50.0  

A matrix inversion is required for the first term on the right hand side but in this case it is trivial as it is just the inverse
of a single number.

Simplifying further gives:

1 –1   
h =  0.124244   – 0.100 + – 1 1  – 0.02155 – 4.96978   (16.105)
  0.02155 – 3.10612  

with the last term of 4.96978 being G –1 A e that will be used below in computing the value of q  1  . Thus
2 L
– 3.10612

391
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

–1
G A2 e L = 4.96978 (16.106)
– 3.10612

Simplifying further gives

1   
h = 8.04864  – 0.100 + – 1 1  – 4.99133  
  3.08457  

1
h = 8.04864  – 0.100 +  4.99133 + 3.08455  

1
h = 8.04864  – 0.100 + 8.07588 

1
h = 8.04864  7.97588 

1
h = 64.1949 m

and thus

1 = 64.1949 m (16.107)

H1

It is not possible to check for convergence based on the change of heads for iteration 1 - we will have to wait for the
second iteration.

Determining the Global Gradient Algorithm q vector values for iteration 1:

1
The q value can now be obtained from Equation 16.98 of

 1 1 
 0 ---------------------- 0 ---------------------- 0 
0 0 0 0
1  Q1 r 1 Q1 1 0 80.0 + 1 Q 1
r –1 
q = 0.5  + 1  (16.108)
H1
 Q 0  0 1 - 0 – 1 50.0
--------------------- 0 1 - 1
--------------------- 
 2 0 0 0 0 
 r 2 Q2 r 2 Q2 

which is just by a series of matrix multiplications (no matrix inversion is required). Recall the form of the general equa-
tion of

k + 1
= 1---   n – 1 q + G A 2 e L + G A 1 h
k + 1 k  –1 –1
q  (16.109)
n

–1 1
Now substitute for G from Equation 16.104, H 1 from Equation 16.107 and use the computed expression for
–1
G A 2 e L from Equation 16.105 as

1  
q = 0.5  0.02155 + 4.96978 + 0.062122 0 –1
64.1949 
 0.02155 – 3.1061 0 0.062122 1 

1  
q = 0.5  0.02155 + 4.96978 + – 0.062122 64.1949 
 0.02155 – 3.10612 0.062122 

1  
q = 0.5  0.02155 + 4.96978 + – 3.98795 
 0.02155 – 3.10612 3.98795 

1  
q = 0.5  0.02155 + 0.98183 
 0.02155 0.88183 

1  
q = 0.5  0.02155 + 0.98183 
 0.02155 0.88183 

1  
q = 0.5  1.00339 
 0.90339 

392
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

3
1 m
q = 0.50169 -------- (16.110)
0.45169 s

and thus

1
Q1
= 0.50169
3
m /s (16.111)
1 0.45169
Q2

Note these values of pipe flows and the head at node 1 are exactly same as for the values at iteration 1 for the solution
to the Q+H equations formulation, however, to solve the Global Gradient Algorithm a Jacobian matrix only of size
NJ = 1 had to be inverted in comparison to a Jacobian size of NP+NJ = 3 that had to be inverted for the Q+H equa-
tions. In addition the sequence of iterates of Qs for the Q – equations formulation is also identical to the Global Gradi-
ent Algorithm formulation iterates. For the Q – equations formulation a Jacobian matrix of size NP = 2 had to be
inverted.

Note also these values of pipe flows are exactly same as for the values at iteration 1 for the solution to the Q – equations
formulation.

ITERATION 2, Global Gradient Algorithm

The vector of new flows from iteration 1 in Equation 16.110 in now used in the second iteration of the Global Gradient
Algorithm solution method. The friction factors (based on the Swamee and Jain equation – Equation 4.61) and corre-
sponding rf values (Equation 4.123) for the flows that resulted from iteration 1 are computed in Table 12.13 (page 306)
as computed for iteration 1 of the Q – equations formulation, as the flows that were computed are identical.

Thus from Table 12.13

1
rf = 646.50 (16.112)
647.25

as shown in Table 16.3.

Table 16.3 The rf values for flows from Iteration 2 for the Q – equations formulation

Pipe Q V D Area (m2) Re** f L (m) rf value

(m3/s) (m/s)* (m) (mm) (Swamee and Jain) (Equation 4.123)
(Equation 4.61)
1 0.5017 7.098 0.3 0.07069 2114448 0.25 0.01901 1000 646.73
2 0.4517 6.3909 0.3 0.07069 1903716 0.25 0.01904 1000 647.48
–6 2
**  = 1.007 10 m /s at 20C ;
0.25 fL -
Using in the above Table: f = ---------------------------------------------------------
- ; r f = -----------------
  5.74  2 2 gD A
2
log 10 ------------ + ------------
 3.7 D 0.9
Re

The values from Equation 16.111 and Equation 16.112 may be substituted into the expression for the inverse of the G
matrix (from Equation 16.37) for the second iterations as

1
---------------------- 0 1
1 1
r 1 Q1 --------------------------------------- 0
= 646.50 0.50169 = 0.0030821 0
–1
G = (16.113)
1 1 0 0.0034192
0 ---------------------- 0 ---------------------------------------
1 1 647.25 0.45169
r 2 Q2

Determining the Global Gradient Algorithm h vector values:

Now solve for the unknown head at node 1 for iteration 2 into Equation 16.94 as
–1
 1    1 
 ----------------------
1 1
0    1
----------------------
1 1
0 
2  r Q1 – 1   – 0.100 + – 1 1  – Q 1 – r 1 Q 1 1 0 80.0  
h =  –1 1 1  
 1 1    Q 1  1 0 – 1 50.0  
 0 ----------------------    2 0 ---------------------- 
1 1 1 1
 r 2 Q2    r 2 Q2 

and thus by substituting from Equation 16.112 and Equation 16.113 gives

393
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

–1
2  – 1   – 0.100 + – 1 1  – 0.50169 – 0.0030821 1 0 80.0  
h =  – 1 1 0.0030821 0

0

 0 0.0034192 1    0.45169 0 0.0034192 0 – 1 50.0  

Multiplying out matrices gives:

–1
2     80.0  
h =  – 0.0030821 0.0034192 – 1   – 0.100 + – 1 1  – 0.50169 – 0.0030821 0

 1    0.45169 0 – 0.0034192 50.0  

Simplifying further gives:

2 –1   
h =  0.0065013   – 0.100 + – 1 1  – 0.50169 – 0.24656  (16.114)
  0.45169 – 0.17096 

with the last term of 0.24656 being G –1 A e that will be used below in computing the value of q  2  .
2 L
– 0.17096

2 –1   
h =  153.815   – 0.100 + – 1 1  – 0.74825  
  – 0.28073  

2
h = 153.8158  – 0.100 +  0.74825 – 0.28073  

2
h = 153.8158  – 0.100 + 0.46752 

2
h = 153.8158  0.36752 

2
h = 56.5313 m

and thus

2 = 56.5313 m (16.115)

H1

Determining the Global Gradient Algorithm q vector values:

1
The q value can now be obtained from Equation 16.98 of

 1 - 1 - 
 1 --------------------- 0 --------------------- 0 
1 1 1 1
2  Q1 r 1 Q1 1 0 80.0 r 1 Q1 –1 
q = 0.5  + + 2  (16.116)
H1
 Q 1  0
1
----------------------
0 – 1 50.0
0
1
----------------------
1 
 2 1 1 1 1 
 r 2 Q2 r 2 Q2 

which is just by a series of matrix multiplications (no matrix inversion is required). Recall the form of the general equa-
tion of

k + 1 1 k  –1 –1 k + 1
q = ---   n – 1 q + G A 2 e L + G A 1 h  (16.117)
n

–1 –1
Now substitute for G from Equation 16.113 and use the computed expression for G A 2 e L from Equation 16.114
as

2  
q = 0.5  0.50169 + 0.24656 + 0.0030821 0 –1
56.5313 
 0.45169 – 0.17096 0 0.0034192 1 

2  
q = 0.5  0.50169 + 0.24656 + – 0.0030821 56.5313 
 0.45169 – 0.17096 0.0034192 

2  
q = 0.5  0.50169 + 0.24656 + – 0.17423 
 0.45169 – 0.17096 0.19329 

394
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

2  
q = 0.5  0.50169 + 0.07233 
 0.45169 0.02233 

2  
q = 0.5  0.50169 + 0.98183 
 0.45169 0.88183 

2  
q = 0.5  0.57402 
 0.47402 

2
= 0.28701
3
q m /s (16.118)
0.23701

and thus

1
Q1
= 0.28701
3
m /s (16.119)
1 0.23701
Q2

Note again, as for iteration 1, these values of pipe flows and the head at node 1 are exactly same as for the values at
iteration 2 for the solution to the Q+H equations formulation. In addition, the sequence of iterates of Qs for the Q –
equations formulation is also identical to the Global Gradient Algorithm formulation iterates. The iterations continue
up to iteration 8 and are summarized below in Answer Part (iii).
___________________________________________________________________________________
Answer – (iii) Tabulate all iterations to convergence for the solution process for the Global Gradient Algorithm
formulation

FINAL SOLUTION – ALL ITERATIONS, Global Gradient Algorithm

The iterations for the full solution based on the Global Gradient Algorithm formulation are shown in Table 16.4. Eight
iterations are shown. The stopping tolerance based on the infinity – norm of the head changes that was used in the com-
–6
puter simulation run for producing Table 16.4 was 1.0 10 m. Realistically the stopping tolerance could have been
–3
1.0 10 m or 1 mm which means that convergence would have occurred in 7 iterations.

Table 16.4 Iteration sequence of flows and head for the Newton solution for the Global Gradient Algorithm formulation
–6
(stopping tolerance 1.0 10 m achieved at iteration 8) for two pipe network

Iteration Q 1 m3/s Q 2 m3/s H1 m k + 1

H 
(Equation 10.22) **

0 0.02155 0.02155
1 0.50169 0.45169 64.195 15.458

2 0.28701 0.23701 56.531 7.664

3 0.19869 0.14869 59.357 2.826

4 0.17566 0.12566 60.093 0.736

5 0.17361 0.12361 60.157 0.064

6 0.17357 0.12357 60.158 0.001

7* 0.17357 0.12357 60.158 1.68E – 05

8 0.17357 0.12357 60.158 2.44E – 07

–3
* A stopping tolerance of less than 1.0 10 m or 1 mm achieved at iteration 7

max k + 1
** Infinity – norm for Head Changes (m) used for testing for convergence =  Hi 
i=1, NJ

___________________________________________________________________________________

395
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

Answer – (vi) Plot convergence characteristics for the iterations for the Global Gradient Algorithm formulation
for the two pipe network.

Figure 16.3 Convergence of flows for solution of the two pipe network using the Global Gradient Algorithm
formulation

Figure 16.4 Convergence of head at node 1 for solution of the two pipe network using the Global Gradient Algorithm formu-
lation

16.4 References
Ajiz, M. A., and A. Jennings, A. (1984). “A robust incomplete Choleski conjugate gradient algorithm.”
Int. J. Num. Meth. Engng., 20, 949-966

Ayres, F. (1974). Schaum's outline of theory and problems of matrices, SI (metric) ed., McGraw-Hill,
London, England.

Elhay, S. and Simpson, A.R. (2011). “Dealing with zero flows in solving the non-linear equations for
water distribution systems.” Journal of Hydraulic Engineering, Vol. 137, No. 10, Oct., 1216-1224.

396
 Chapter 16 - File: Chap15-TP-QH-form-v02.fm July 6, 2022 Version 7

Elhay, S., Simpson, A.R., Deuerlein, J., Alexander, B. and Schilders, W.H.A., (2014). “Reformulated
Co-tree flows method competitive with the Global Gradient Algorithm for solving the water distribution
system equations.” Journal of Water Resources Planning and Management, Dec., Vol. 140, No. 12.
DOI: 10.1061/(ASCE)WR.1943-5452.0000431.

Kershaw, D. S. (1978). “The incomplete Cholesky - conjugate gradient method for the iterative solution
of systems of linear equations.” Journal of Computational Physics, 26, 43-65.

Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. (1996) FORTRAN numerical recipes
(Vol. 1), Cambridge England.

Rossman, L A. (2000). EPANET 2 Users Manual. U.S. Environmental Protection Agency, Washington,
D.C., EPA/600/R-00/057.

Salgado-Castro, R.O. (1988). “Computer modelling of water supply distribution networks using the gra-
dient method.” PhD Thesis, University of Newcastle-Upon-Tyne, UK.

Simpson, A. R., and Elhay, S. (2009). “A framework for alternative formulations of the pipe network
equations.” 11th Annual Symposium on Water Distribution Systems Analysis, American Society of Civil
Engineers, Kansas City, USA.

Todini, E., and Pilati, S. (1988). “A gradient algorithm for the analysis of pipe networks,” B. Coulbeck
and O. Chun-Hou, eds., Wiley, London, 1–20.

397
 Chapter 17 - File: Chap17-Data-needs.fm July 6, 2022 Version 6

CHAPTER 17 DATA REQUIREMENTS FOR WATER

DISTRIBUTION SYSTEM ANALYSIS1

To analyze the hydraulic performance of a water distribution system a considerable amount of data is
required. Computer simulation of water distribution systems was introduced in Chapter 11. Most of the
data listed in this Chapter would be required to be available for either a hand – analysis or a computer
analysis. Much of the data describes the physical components in the water distribution, the topology of
the network and components (that is — the connectivity), details of the various components and how the
system is operated. At the end of the Chapter some specific data requirements are discussed relating to
convergence of the hydraulic solver.

17.1 Basic Data Requirements for Computer Analysis of Networks

To analyze a water distribution system or pipe network to meet water demand loadings (e.g. peak instan-
taneous or peak hour, fire and extended period simulation) the following information is required

• pipe information — lengths, diameters, material, condition of existing pipes (roughness),

cost information for new pipes
• flows drawn as demands at nodes e.g. peak hour, variation of demand during a peak day,
and fire flows
• water surface elevations or heads at reservoirs in the system
• elevations of all nodes
• pump information
• fitting or minor loss information including bends and valves
• design criteria — for example, minimum allowable nodal pressure, maximum allowable
nodal pressures
A detailed description of considerations for data requirements is given below.

17.2 Overview of data requirements

A number of aspects need to be considered in determining the data necessary to perform a thorough
analysis of likely requirements and for the design of a water distribution system.

17.2.1 Sources of supply

The sources of water supply need to be identified including headworks reservoirs, water treatment plants
and pumping stations on rivers. The delivery systems of pipes should be capable of supplying adequate
water either directly into the water distribution system or to service reservoirs or tanks within the net-
work.

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Ade-
laide, Australia. Part of a book entitled Water Distribution Systems Engineering.

398
 Chapter 17 - File: Chap17-Data-needs.fm July 6, 2022 Version 6

17.2.2 The existing system

In an existing system that is to be either expanded or rehabilitated it is necessary to collect together data
to define the connectivity and sizes of components in the systems. In defining the connectivity of the
network, node numbers or other designations are usually assigned. In addition, pipe numbers or names
are also assigned. If the demand of a number of houses is aggregated at a node, the elevation associated
with a node is usually taken as the elevation of the delivery to the house with the highest elevation in the
nodal supply area. Nodal information that is required includes

• the elevation of the node with respect to a selected horizontal datum

• demands at each node
• a node number or ID label
Data required about pipes includes

• location of pipe
• labels or node numbers of the end of the pipe
• diameter (including actual internal diameter)
• length
• material
• age and condition
• roughness coefficient information (e.g. the roughness height or for the Darcy – Weisbach
friction factor and head loss equation or Hazen – Williams C value – both preferably from
a field test where pressure drop and flow rate are measured to enable calculation of pipe
roughness coefficient)
• wall thickness (important for water hammer analysis or selection of class of pipe)
• modulus of elasticity of the pipe wall material E (for wave speed calculations for water
hammer analysis)
• depth of cover above the pipe
• soil type information is also useful to identify possible pipe corrosion and possibility of
ground movement
Properties of water required include

• the typical temperature range expected

• the kinematic viscosity of water  for a selected temperature (for computation of Reynolds
number and the Darcy – Weisbach friction factor)
• the specific weight of water  (at a selected temperature) for computation of pressure
• the density of water  for a selected temperature (for computation of water hammer)
• the bulk modulus of elasticity of water K (for water hammer computations)
Pump information includes

• location of pumping station and the number of pumps

• labels or node numbers at the immediate upstream and downstream sides of each pump in
the pumping station
• head – flow curve
• rotational speed
• efficiency – flow curve
• rated operating conditions including rated head, rated flow, rated power (from which rated
efficiency may be computed)
• condition of pumps
• details of pumping station configuration including parallel or series combination of pumps
• pump operation schedules
• electricity tariff structures
• rotating moment of inertia of the pump (for water hammer analysis)
• discount rate for present value of annual pumping cost computations

399
 Chapter 17 - File: Chap17-Data-needs.fm July 6, 2022 Version 6

• design life of pumps

Pressure and flow control valve information includes

• location of pressure reducing valve, pressure sustaining valve, flow control valve
• labels or node numbers at the immediate upstream and downstream sides of the valve
• pressure setting to which the downstream pressure or upstream pressure is restricted or flow
control value setting
Storage and tank (or reservoir) information includes

• location of storage
• label or node number
• nature of reservoir or tank (is it covered?)
• normal water surface elevation of the tank
• volume of tank
• normal operating range of water surface levels (maximum and minimum)
• elevation – volume relationship for the tank (to enable volume – elevation changes to be
computed during extended period simulations), for example, the diameter of the tank
• assumed reservoir operating levels for the peak hour analysis condition or for fire loading
analysis
• sources of supply to tanks and reservoirs in the system
System layout information includes

• geographic location
• topography of the supply area
• identification of pressure zones and their boundaries (stop valves)
• cross – connections to other distribution zones is needed
• positions of all components in the system
Water quality information includes

• chlorine injection points

• method of disinfection — quantity and amount
Finally, an understanding of the current operation of the existing water distribution system is required.
Information on quality of the water being supplied from the various sources in the system should also be
gathered. If the water is treated, then treatment cost per volume may have an influence on controlling
preferable sources of supply.

17.2.3 Demands
Two aspects must be considered here. In the first instance, the types of demands expected in the area of
supply need to be estimated. These include domestic demands, however, the density of housing in terms
of number of lots must also be considered. Industrial demands, commercial demands, school and hospi-
tal demands and public park usage also need to be estimated. Analysis of historic demands may be use-
ful. Current and projected demands should also be determined.

Consideration of the peak day demands usually have a significant influence on the design of a water dis-
tribution system or network. The peak instantaneous demand during the peak hour of the peak day is
usually considered. An event with a particular annual recurrence interval (ARI) may be selected (e.g. a 1
in 7 year event may be used) or sometimes the peak day on record may form the basis of the design. His-
torical data is important in knowing the demand characteristics of the present system. Demographic
trends, as well as, changes in consumption with time, educational programs to reduce the demand all
need to be considered.

Projected demands into the future must be made assuming full development of the supply area. Popula-
tion projections and projected occupancy of households is also required. Domestic water usage patterns
should be identified including the proportion used for garden watering. In addition, climate historic
records provides information on likely water usage.

400
 Chapter 17 - File: Chap17-Data-needs.fm July 6, 2022 Version 6

Implications on the need to augment headworks storages or pumping from rivers, in addition to any
additional water treatment plants must also be considered.

The demands also need to be quantified over time so variation of demand within the day, average daily,
weekly and annual demands need to be assembled. An extended period simulation is needed to ensure
tanks refill after a critical demand period (for example, during a peak day or sometimes as long as a
week) need to be studied. In addition, average daily demands may provide an indication of pumping
costs on an annual basis.

17.2.4 Design criteria

Various design criteria may be applied to the design of the water distribution network including

• peak instantaneous demand

• fire demands (at particular nodes or perhaps individually at all nodes)
For both of these situations, minimum allowable nodal pressures are usually specified. For example,
under peak instantaneous demand conditions the minimum allowable nodal pressure may be 20.0 m of
pressure head. A static or steady state computer model analysis is usually conducted to analyze these
two load cases as specified above. Maximum allowable pressures at nodes are usually specified to pre-
vent the occurrence of damage to appliances such as dishwashers and washing machines.

During the peak day or critical peak period, the tanks in the system must refill after the end of the peak –
demand, so the volumes of these tanks must be of adequate size to ensure these tanks do not run dry.

Minimum pipe sizes and pipe classes for various materials acceptable in the system (e.g. 100 mm diam-
eter) need to be set. Other design criteria may include, maximum allowable velocities of flow in pipes,
and water quality design criteria. Pipe breakage or system reliability may be another consideration.
Necessity for loops also needs to be considered. Dead end mains should be avoided.

A water hammer analysis of the system is necessary. Maximum allowable pressures are required for var-
ious classes of pipes that in turn determines of the wall thickness of the pipe.

17.2.5 Cost of options

The options available must be identified. For example, the pipe layout options and pump and storage
location options must be identified. If necessary, additional sources of water may need to be identified.
Pipe cleaning may also be an option. Pipe cleaning methods may sufficiently increase the internal diam-
eter of the pipe by removing built up scale and also by reducing the pipe roughness. These costs must be
incorporated in the overall estimate of cost. In addition, aspects including the need for loops to ensure
reliability of supply if there is a pipe breakage, need to be assessed.

Costs of choices for components in the new or expanded part of the water distribution system need to be
assembled, including pipe purchase costs for a range of sizes and materials, pump purchase and operat-
ing costs, storage costs, and pressure reducing valve costs (if necessary). Pipe laying costs may differ if
an existing road is being torn up and restored versus laying the pipe through farmland being newly sub-
divided. Road resurfacing costs and long term pipe maintenance costs should also be included. The
resulting inconvenience to traffic, local businesses, and residents in the surrounding area should also be
taken into account. Land acquisition costs may also be needed.

Shutoff valves are required in systems to be able to isolate portions of the main in case of breakage.
These costs must be incorporated in the overall estimate of cost.

Annual operating costs for pumps based on a projected daily schedule (e.g. the pumping during the aver-
age day) may be extended to the computation of annual pump costs. A present value of the operating
costs may be computed based on an expected life of a pump and also for a selected interest rate.

Loops are essential in water distribution systems to ensure adequate flow circulation to maintain ade-
quate water quality especially chlorine residual for bacterial control. In addition, aspects including the
need for loops to ensure reliability of supply if there is a pipe breakage, need to be assessed.

401
 Chapter 17 - File: Chap17-Data-needs.fm July 6, 2022 Version 6

In summary, the cost information (and associated information needed for pipes includes)

• diameter (nominal and internal diameter or ID)

• material (e.g. uPVC, Hobas, DICL (ductile iron cement mortar lined), RC (reinforced
concrete) — often used in irrigation systems
• roughness coefficient (roughness height  for Darcy – Weisbach or Hazen – Williams coeffi-
cient C)
• class (gives maximum allowable pressure)
• length of segments
• modulus of elasticity of material of pipe wall (for water hammer analysis)
• expected design life of the pipe (to enable full life cycle costing to be estimated)
• wall thickness (for water hammer analysis)
Maintenance costs for the proposed network also need to be estimated. External protection may be
needed for buried steel pipes. Stephenson (1984) gave information on the components of pipe costs for
water pipelines in open country and typical conditions (with the proviso that the cost of a pipeline may
vary widely from situation to situation). These estimates are shown in Table 17.1.

Table 17.1 Pipe construction costs

Item Cost Notes
proportion
Supply of pipe 55% May reduce as new materials are developed
Excavation 20% Depends on terrain, may reduce as mechanical
excavation techniques improve
Laying and jointing 5% May increase with labor costs
Fittings and specials 5%
Coating and wrapping 2%
Structures (valves, chambers, 2%
anchors)

17.3 Static versus extended period simulation

Two types of steady state computer analysis of water distribution networks are commonly performed.
The first is 'static – analysis' for a particular point in time. The demand pattern selected is commonly the
peak hourly demand for an extremely high demand peak – day (e.g. 1 in 7 annual recurrence interval
event). Neither the variation in nodal demands nor the reservoir water levels remain constant over a
period of time. The second form of analysis is extended period simulation or dynamic analysis of water
distribution systems. The performance of storages in balancing supply and distribution may be checked.
Extended period simulation ensures an adequate level of service to consumers under varying conditions
of demands and reservoir water levels (Bhave 1988). Adequate flow rates and pressures (residual heads)
must be maintained at all times. Analysis of the network is usually carried out over a period of 24 to 48
hours under varying nodal demands and reservoir water levels (Bhave 1988).

17.4 Demands
Two aspects must be considered here. In the first instance, the types of demands expected in the area of
supply need to be estimated. These include domestic demands, however, the density of housing in terms
of number of lots must also be considered. Domestic uses for water are numerous including drinking
water, toilet flushing, washing, bathing, kitchen use, washing of clothes, household cleaning, car wash-
ing, swimming pools and air conditioning. Garden watering is a large component of domestic use in a
number of places where the climate is hot and dry. Industrial demands (breweries, canneries, laundries,
paper mills, steel mills), commercial demands, schools, hospitals, educational establishments, hotels and
public park usage also need to be estimated.

402
 Chapter 17 - File: Chap17-Data-needs.fm July 6, 2022 Version 6

Peak day consideration usually have a significant influence on the design of a network. The peak instan-
taneous demand during the peak hour of the peak day is usually considered. An event with a particular
annual recurrence interval (ARI) may be selected (e.g. a 1 in 7 year event may be used) or sometimes the
peak day on record may form the basis of the design. Historical data is important in knowing the
demand characteristics of the present system. Demographic trends, as well as, changes in consumption
with time, educational programs to reduce the demand all need to be considered.

Projected demands into the future must be made assuming full development of the supply area. Popula-
tion projections and projected occupancy of households are also required. Domestic water usage pat-
terns should be identified including the proportion used for garden watering. In addition, historic records
of climate provide information on likely water usage. Implications on the need to augment headworks
storages or pumping from rivers, in addition to any additional water treatment plants must also be con-
sidered. [[[Some of this seems to be duplicated earlier in the book – check]]]

To determine the water consumption for a city a good knowledge of consumption of volumes of water
by consumers is needed. These volumes need to be related to both population and time. The important
volumes or demands include

• total annual consumption in liters (L) or megaliters (ML) or gallons (g) or millions of gal-
lons (mg). The total number of people to be served by the water supply system is a signifi-
cant determining factor for the total annual consumption
• average daily consumption (L/d) or millions of liters per day (ML/d) or gallons per day
(gall/day) or millions of gallons per day (mgd)
• average daily consumption per capita in liters (L/d/capita) or gallons (gall/day/capita). In
addition, industrial, commercial, parkland and any other specialized water use estimated
flow rates are required
• average daily consumption per service in liters per service (L/service) or gallons per ser-
vice (gall/service)
• peak daily consumption or maximum day in L/s or gpm
• peak hourly consumption in L/s or gpm

The water demand fluctuates within the year (seasonal changes in climate), within a month, within a
week, within a day (changes in household and industrial activity) and within an hour. These fluctuations
in flow may be expressed as ratios of the value at a particular time within the period under consideration
to the average value for that period. For example, during a peak demand day in the summer when the
watering of gardens is high the peak hour ratio may be 1.45. This means that the demand during the peak
hour is 45% higher than the average of the peak day flows taken over 24 hours. Typical ranges of ratios
are shown in Table 17.2.

Table 17.2 Ratios for water demands

Ratio of Range Average Value
Maximum day to 1.2 to 1 THROUGH 1.5 to 1
average day 2.0 to 1
Peak hour to aver- 2.0 to 1 THROUGH 2.5 to 1
age hour 5.0 to 1

Peak hour to aver- 1.4 to 1 THROUGH 1.45 to 1

age of peak day 2.5 to 1

Average daily demand varies depending on the day of the week. Industrial demands often are lower on
weekends. Accurate estimates of the variations in water demand must be known in order to properly
dimension transmission or supply pipes, service reservoirs and tanks, and distributing pipes that make
up the water distribution system network. Quantities of water required for fire fighting are unpredict-
able. These demands must be taken into account. During a fire the flow rate is high but the volume with-
drawn from the water distribution system is relatively small. The smaller the size of the community the
more unpredictable and variable the demand.

403
 Chapter 17 - File: Chap17-Data-needs.fm July 6, 2022 Version 6

Estimation of design demands or consumptions require analysis of historic demand patterns. Periods or
particular days of extremely high demand are useful to enable estimation of peak day demands. The pro-
jection of estimates of future demands requires an analysis of social, demographic and economic trends.
Considerable research has been carried out related to the prediction of population growth. The areas of a
city that have been designated for growth, in other words, where new subdivisions will be built needs to
be known. Details of maggots for predicting demand growth are not given in this book. There are many
other books on water supply that provide adequate coverage of this topic. In addition, local factors
including climate, demographics and attitudes to water conservation all affect current and future demand
patterns that need to be considered in the design of the water distribution network.

Fair, Geyer and Okun (1971) give a comprehensive list of factors that affect water consumption. These
include climate, standards of living, types of industry, commercial activity, cost of water, availability of
private supplies, quality of untreated water, pressures maintained, the use of water meters and systems
management.

17.5 Design Criteria

17.5.1 Design periods
A design period for a major structure is the number of years for which the proposed system, component
structures and equipment are deemed to have a useful and serviceable life. Design periods vary for struc-
tures within different segments of the water supply system. Typically, dams and reservoirs have a useful
life of in excess of one hundred years. Water distribution pipes last for between 70 and 100 years.

A number of factors govern the appropriate selection of a design period for a particular element of a
water supply system. Fair, Geyer and Okun (1971) suggested the following factors should be considered

• the useful life of components, structures and equipment needs to be considered. Wear and
tear and obsolescence all need to be taken into account
• the ease of extension of a water distribution network whether existing or planned also needs
to be considered
• the level of interest rate to be used in the economic analyses of alternatives is important.
Project alternatives are often compared on a present worth or present value analysis. For
example, if pumping costs are considered in the analysis of a particular scheme then a
stream of annual costs over the assumed life of the component needs to be converted into a
present value.
• in the early years, the amount of growth in a particular area may be significantly less than
that expected at full build out. As a result, consideration must be given to the performance
of the works when they are not fully loaded.
A long design period should be selected if the element of the water supply system has

• a long useful life

• if there is great difficulty in implementing extensions
• if the growth rate is small
• for low rates of interest
• if the early years performance of the system is better
17.5.2 Levels of service
Many water utilities are beginning to specifically define the level of service to customers that they may
expect from the utility. In fact, in some cases there are penalties if these levels of service are not main-
tained. There are different ways of defining a level of service. In the past minimum allowable pressures
at the entrance to a property was defined. Different utilities specify different minimum allowable pres-
sure levels. For example, Adelaide, Australia currently uses a minimum allowable pressure of 20 m of
head as a design criteria. Melbourne, Australia uses a value of 15 m of head. England uses a value of
10 m of head.

404
 Chapter 17 - File: Chap17-Data-needs.fm July 6, 2022 Version 6

Another way of defining the level of service is to specify the minimum flow rate that a customer (usu-
ally a domestic dwelling) may expect.

17.6 Controlling demand growth

During the last 70 years the rate of growth of average water consumption per capita per day has been
around 1% per annum (Crockett and Carroll 1997). Water demand in a typical Western City in the 1920s
was on the order of to 250 liters per capita per day. In recent times, in Australia the demand has grown to
the order of 500 liters per capita per day.

In the past water utilities have maintained higher pressures (40 m or 60 psi in gage units) in the system
to cope with fire fighting occurrences. The introduction of motor pumpers in fire fighting appliances has
eliminated the need for elevated pressures. Improved construction of joints in pipelines has reduced
leakage. Seasonal variations in temperature may cause pipe joints to expand and contract. Colder tem-
peratures may result in increased leakage due to contraction of pipes at their joints. Pressure control or a
lowering of pressure also provides a reduction in leakage.

Water utilities should be able to account for a large percentage of water delivered. Water meters should
be able to record water deliveries to domestic dwellings, industry and commercial users. It has been
shown that metering reduces demand. Unaccounted for water is used for purposes such as fire fighting,
washing filters, flushing streets, mains and sewers and some watering of parklands and public gardens.

17.7 References
Bhave, P. R. (1988). “Extended period simulation of water systems - direct solution.” Journal of Envi-
ronmental Engineering, 114(5), October, 1146-1159.

Fair, G. M., Geyer, J. C., and Okun, D. A. (1971). Elements of water supply and wastewater disposal,
2nd Ed., Wiley, New York.

Stephenson, D. (1984). Pipeflow analysis. Developments in Water Science, No., Elsevier, Amsterdam,
84.

405
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

CHAPTER 18 PUMPS1

18.1 Introduction
Pumps are a member of the fluid machinery family that also includes turbines (see 58022). Pumps are
the focus of this Chapter. The aim of this Chapter is to enable informed judgment to be made in the
selection of pumps in Civil Engineering applications. In many regions of the world the use of pumps in
a water supply is extremely important. This book covers very little on the actual design of pumps. Many
other books provide detailed information on the design of pumps. The discipline area of pump design is
a specialist area and is usually in the domain of Mechanical Engineers and the pump manufacturers.

There are a number of important definitions of pump terminology that an engineer needs to become
familiar with. This Chapter starts out by giving the most important of these definitions. Different types
of pumps are then defined but only rotodynamic or impeller type pumps are considered in detail here.
Pump performance characteristics are defined. These are generally provided by the manufacturer. The
steady state operation of a pump is then considered by the introduction of the pump curve, the system
curve and the pump operating point. Specific speed for a pump is defined and ranges are given for cen-
trifugal, mixed flow and axial flow pumps. The pump affinity laws and dimensionless numbers associ-
ated with pumps are presented. Dimensionless pump curves are then developed. The use of the affinity
laws for similar pumps results in the development of two affinity laws. Pumps in series or pumps in par-
allel are sometimes used in pumping stations depending on the economic evaluation of alternatives. Two
sections are devoted to the consideration of pumps in series and pumps in parallel. Prevention of cavita-
tion of pumps is also considered. The concept of net positive suction head is introduced that enables a
designer to select an appropriate vertical elevation for the pump relative to the water surface elevation of
the suction side reservoir to avoid cavitation of the impeller of the pump. The rotating moment of inertia
of a pump is introduced. This is important when water hammer is considered.

18.2 Useful Definitions for Pumps (alphabetized)

TERM DEFINITION
BEP Best efficiency point on the pump efficiency  versus flow Q curve
Net positive suction head The amount of head available at the inlet of the pump in relation to
available (NPSHA) prevention of cavitation in the pump impeller. Depends on the suction
pipe work configuration.
Net positive suction head The amount of head required at the inlet of the pump to prevent cavi-
required (NPSHR) tation in the pump impeller. Provided by the pump manufacturer.
Power curve for pump Plot of pump shaft power P versus flow Q. Can be computed from the
(P – Q) pump head curve (HP versus Q) and the pump efficiency curve (
versus Q).
Pump curve (H – Q) Plot of pumping head HP versus flow Q for a pump at a particular
rotational speed. Usually obtained from the pump manufacturer.
Pump flow Q Flow through the pump
Pump efficiency  Ratio of water or hydraulic power (IEC 1999) QHP to the pump shaft
power P

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Ade-
laide, Australia. Part of a book entitled Water Distribution Systems Engineering.

406
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Pump efficiency curve Plot of efficiency  versus flow Q for a pump at a particular rotational
( – Q) speed. Usually obtained from the pump manufacturer.
Pump head HP The increase in head across the pump from the suction flange to the
discharge flange
Pump shaft power P Product of specific weight of fluid flowing , discharge Q and pump
head HP divided by the pump efficiency or P = QHP/
Pump rotational speed N Rotational speed of the pump impeller
Shut – off head Value of pump head from pump curve corresponding to zero flow
Q=0
Static head Hs Difference in elevation between the water surface at a lower reservoir
and the water surface elevation at the upper reservoir where water is
being pumped from the lower to the upper reservoir
Static suction head hs Elevation difference between the centerline of the intake to the pump
(2 and thus z 2 ) and the suction reservoir water surface (1 and thus z 1 )
such that h s = z 2 – z 1 where
• hs is positive when the pump is above the reservoir (m or
ft)
• hs is negative when the pump is below the reservoir water
surface elevation (m or ft)
System curve Sum of static head Hs plus pipe friction losses and minor losses in suc-
tion pipework and delivery pipework of the pumping system
Water or hydraulic power Product of specific weight of fluid flowing , flow Q and pump head
(IEC 1999) HP or QHP.

18.3 Main Features of a Pump

A pump involves a rotating system. Other types of rotating systems include compressors, turbines, pro-
pellers, fluid couplings and torque converters. A pump is a hydraulic machine that adds energy to a fluid
to move fluid from one point to another — usually the liquid is moved from a lower elevation to a higher
elevation.

18.3.1 Types of pumps

There are a number of different types of pumps including

• rotodynamic or impeller pumps. These have a continuous flow.

• reciprocating positive displacement pumps. These are a piston type pump with pulsating
flow.
• rotary positive displacement pumps (e.g. gear, vane and screw pumps). These have a con-
tinuous flow rather than a pulsating flow.
The impeller pump is the most commonly used type in water supply situations. As a result only impeller
type pumps is considered further in this Chapter.

18.4 Energy Equation Applied to a Pumping System

Applying the energy equation (Equation 3.67) across a pump in a system such as that is shown in
Figure 18.1 the head added to the fluid by the pump HP must be calculated from

2 2
p V1 p V2
z 1 + -----1- + --------
- + H P = z 2 + -----2- + --------
2g  f  L
-+ h + h (18.1)
 2g 

where

• z = elevation head (m or ft)

407
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

• p   = pressure head (m or ft)

• V = velocity (m/s or ft/s)
• g = gravitational acceleration (m/s2 or ft/s2)
• HP = pump head added to flow (m or ft)
• h f = the head loss along the pipe (m or ft)
• h L = the head loss due to minor losses (m or ft)

Figure 18.1 Rising main system supplied by a pump; note hf2 has been exaggerated to highlight its
existence - normally the loss per unit length on the suction side would be significantly smaller than
in the discharge pipe

Question – How would the pressure difference between the suction and discharge sides of a pump be obtained?

Answer – By using two Bourdon pressure gages that each has a dial reading the pressure on both sides of the pump
(on the suction intake pipe and on the discharge pipe) to determine the pressures. The head across the pump is the dif-
ference in HGL (sum of elevation head of the location of the centerline of the dial of the gage and the pressure head)
between the downstream location and the upstream location for the Bourdon gages.

Conventionally minor losses inside the pump between the suction flange and discharge flange of the
pump are not calculated separately. The manufacturer includes all losses as pump losses.

A typical cylindrical concrete tank in a pumping system is shown in Figure 18.2.

Figure 18.2 Photograph of a cylindrical water tank (South Australia)

408
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

18.5 Rotodynamic or Impeller Pumps

In rotodynamic or impeller type pumps there is only one moving part — the impeller — that rotates on a
shaft within a housing or the pump casing.

18.5.1 Classification of impeller pumps

Three basic classes of impeller type pumps are

• a centrifugal pump or radial flow — small flow Q, large pump head HP

• a mixed flow type pump — moderate Q, moderate HP
• an axial flow pump — large Q, small HP
The head produced by the pump depends on rotational speed or the periphery velocity of the tips of the
vanes of the impeller.

18.5.2 Centrifugal pumps

This type of pump is also referred to as a radial pump or volute or diffuser. In a centrifugal pump water
enters through the eye of the impeller and discharges radially at right angles to the axis of rotation of this
impeller due to the centrifugal force of the rotary motion. They provide a smooth and even flow. Types
of centrifugal impellers include:

• closed or fully – shrouded, single and double entry

• semi – open impeller
• fully open impeller
A single entry impeller is not hydraulically balanced and thus an axial thrust results that has to be sup-
ported by a balancing chamber. A double suction pump has virtually no axial hydraulic thrust. A single
stage pump has one impeller. A multi – stage pump has more than one impeller. Centrifugal pumps may
have a horizontal or vertical axis.

The load on the motor is very easy to cope with for a centrifugal pump as only a relatively small torque
is required to start the pump against a closed valve. The effect of closing the valve on the pump flow is
to increase the pump head by approximately 15 to 30% (to the shut – off head value corresponding to a
discharge of zero) and to reduce the power by 50 to 60% from those values at the best efficiency point
(ASCE 1975).

A typical gate valve that is often used on either side of a pump is shown in Figure 18.3 and Figure 18.4.

409
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.3 Photograph of a gate valve

Figure 18.4 Photograph of a gate valve showing the gate with a semi – circular bottom

410
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

18.5.3 Mixed flow pumps

The impeller vanes on a mixed flow type pump have some degree of twist and the water flow path is
somewhere between a radial or centrifugal pump and an axial flow type pump (discussed in the next
Section). The impeller for a mixed flow pump develops head both by centrifugal force and by lift of the
vanes on the water in the diffuser casing or bowl. Often a mixed flow type pump is of the vertical type
and located in a wet well.

18.5.4 Axial flow pumps

Often an axial pump is called a propeller pump because the impeller looks like a ship’s propeller. The
impeller in an axial flow type pump propels water along parallel to the axis of rotation. The increase in
head across the impeller of an axial flow type pump is achieved by the lifting action of the vanes on the
water. Axial flow pumps are often used in drainage applications.

18.5.5 Pump size specification

Usually a pump is specified or expressed in terms of the suction and the delivery flange nominal diame-
ters with delivery being specified first. For example if a pump is referred to as a 200/250 mm pump then
the 200 mm dimension refers to the diameter of the delivery flange and the 250 mm diameter refers to
the diameter of the suction flange.

Often the suction line pipe diameter leading to intake side of the pump is larger than the discharge pipe
downstream of the pump in order to minimize the possibility of cavitation. The concept of net positive
suction head (NPSH) in relation to cavitation potential for the pump is introduced later on in the Chap-
ter.

18.6 Pump Performance Characteristics

Basic pump performance factors for an impeller type pump include

• head HP usually in meters (m) in SI units (or feet — ft — in US customary units) of water
• flow Q in L/s (or US gpm or cfs)
• efficiency  taken as a decimal fraction between 0 and 1 (sometimes referred to as Ep)
• shaft power input to the pump P in kW (or HP)
• rotational speed N in rpm
18.6.1 Pump curves
Three curves are important for consideration of pumps in water distribution systems

• HP versus Q — called the pump curve

•  versus Q — called the efficiency curve
• P versus Q — called the power curve
These characteristics apply to just one value of the impeller rotational speed N — and this value should
always be specified on the diagrams.

Different classes of impeller or rotodynamic pumps (centrifugal, mixed flow and axial flow) have differ-
ent shaped curves.

If the pump head is increased by throttling the pump delivery valve then the flow must decrease as the
pump operating point moves to the left along the H – Q curve. When the flow is zero (Q = 0) this is
referred to as the shut – off head condition.

The point of maximum pump efficiency is called the best efficiency point (BEP). The most common
operating point for the pump should be close to the BEP to minimize the power costs (although it does
depend on initial capital cost as well).

411
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

18.6.2 Definition of water or hydraulic power

Water or hydraulic power is defined as the amount of power the water needs to receive to pump a fluid
of specific weight  at a flow Q at a pumping head of HP. It is defined as the following product

QH P (18.2)

PUMP WATER OR HYDRAULIC POWER

where

•  = specific weight of fluid (N/m3or lb/ft3)

• Q = pump flow (m3/s or ft3/s)
• HP = pump head (m or ft)

18.6.3 Pump efficiency

The pump efficiency  is defined as

water or hydraulic power - = --------------- QH P

 = ----------------------------------------------------------------------- (18.3)
shaft power from pump motor P

PUMP EFFICIENCY

where

•  = pump efficiency (dimensionless or expressed as a percentage)

• P = shaft power delivered by pump motor (W or HP - watts are usually converted to kW by
dividing by 1000)
Note that for a pump, in contrast to a turbine (see Section 22.9), the shaft input power from the pump
motor P . to the impeller of the pump is greater than the water or hydraulic power QH P .

18.6.4 Pump shaft power

The pump shaft power in terms of fluid specific weight, pump head, flow and efficiency is defined as

QH
P = ---------------P (18.4)


PUMP SHAFT POWER

where

• P = shaft power delivered by pump motor (W or HP - watts are usually converted to kW by

dividing by 1000)
•  = specific weight of fluid (N/m3or lb/ft3 e.g. 9789 N/m3 for water at 20C)
• Q = flow (m3/s or ft3/s – e.g. 1.2 m3/s – a typical large pump in a filtered water pumping
station at a water filtration plant)
• HP = pump head (m or ft – e.g. 25 m for filtered water pumps at a reservoir)
•  = efficiency of the pump (usually specified as a decimal, e.g. 0.82 but sometimes referred
to as a percentage). Efficiency gives a measure of the hydraulic and mechanical losses.
(dimensionless)

412
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

.
Example 18.1 Consider two pumps with pump curves represented by algebraic equations. The general form of the
pump curve for the pump head H P in terms of Q represented as an equation may be expressed as

b
H P = H SHUTOFF – aQ

For pump A, take H SHUTOFF = 50 meters , a = 0.0002 and b = 2.5 . Thus the pump head – flow relationship is
given by

2.5
H P = 50 –  0.0002 Q (18.5)

For pump A, efficiency– flow relationship is given by

2
 = – 0.0164Q + 2.3727Q + 0.0734 (18.6)

For pump B, take H SHUTOFF = 60 meters , a = 0.0002 and b = 2.68 . Thus the pump head – flow relationship
is given by

2.68
H P = 60 –  0.0002 Q (18.7)

(i) Compute the pump head for a flow of 40 L/s for both pump A and pump B. Compute the corresponding efficiency
for pump A.
(ii) Determine the head – flow curves for both pumps A and B over a range of 100 L/s at increments of 10 L/s.
(iii) Plot the pump curves (HP versus Q) for pump A and pump B on the same plot.

(iv) Plot the efficiency versus flow curve for pump A

(v) Plot the power versus flow curve for pump A

Answer – (i) Computation of the pump head at a flow of 40 L/s for both pump A and pump B.

The pumping heads may be obtained from the equations above. For pump A at a flow of 40 L/s the pumping head from
Equation 18.5 is:

2.5 2.5
H P = 50 –  0.0002 Q = 50 –  0.0002 40 = 48.0 m

For pump B at a flow of 40 L/s the pumping head from Equation 18.7 is:

2.68 2.68
H P = 60 –  0.0002 Q = 60 –  0.0002 40 = 56.1 m

For pump A for a flow of 40 L/s the efficiency fromEquation 18.6 is:

2 2
 = – 0.0164Q + 2.3727Q + 0.0734 = – 0.0164  40  + 2.3727  40  + 0.0734 = 68.74%

(ii) Determination of the head – flow curves for both pumps A and B.

The pump head, efficiency and power curves over the range of flows of 100 L/s are given in Table 18.1

Table 18.1 Pump operating information

Flow Head Efficiency Power Head Head

Q Pump A (m) Pump A (%) Pump A (kW) Pump B (m) Pumps in Series,
(L/s) Equation 18.5 Equation 18.6 Equation 18.4 Equation 18.7 A+B (m)
0 50.0 0 60 110
10 49.9 22.2 22 59.9 109.8
20 49.6 41 23.7 59.4 109
30 49.0 56.5 25.5 58.2 107.2
40 48.0 68.7 27.4 56.1 104.1
50 46.5 77.7 29.3 52.9 99.4
60 44.4 83.4 31.3 48.3 92.7
70 41.8 85.8 33.4 42.4 84.2
80 38.6 84.9 35.6 34.8 73.4

413
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Table 18.1 Pump operating information

90 34.6 80.8 37.7 25.5 60.1

100 30.0 73.3 40.1 14.2 44.2

(iii) Plots of the pump curves are given in Figure 18.5.

Equation 18.5 and Equation 18.7 have been used to generate the head values in Table 18.1.

Figure 18.5 Pump curves for Pump A and Pump B

(iv) Plot of the efficiency versus flow curve for pump A is given in Figure 18.6.

Equation 18.6 has been used to generate the efficiency values in Table 18.1.

Figure 18.6 Efficiency versus Q curve for Pump A

(v) Plot of the power versus flow curve for pump A is given in Figure 18.6.

Compute the power for each flow in Table 18.1. In particular, for Q = 40 L/s the power from Equation 18.4 is:

Q H 9789  0.040   48  27358

P = ---------------P = ------------------------------------------ = 27358 = --------------- = 27.36 kW (18.8)
 0.687 1000

414
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.7 Pump power curve versus Q for Pump A

Units of right – hand side of Equation 18.4 for the pump power delivered by the shaft to the pump impel-
ler are as follows

3
N- m
------ -------  m 
m
3 s N-m
------------------------------------ = ----------- = watts or W
1 s

The usual units of power for a pump are kilowatts or kW.

18.6.5 Pump energy

The energy provided to a pump over a period of time t is defined as

E = Pt (18.9)

PUMP ENERGY

where

• E = energy delivered to the pump shaft (Wh - usually converted to kWh by dividing by
1000)
• P = shaft power delivered by pump motor (W or HP)
• t = time over which pump operates (h or hours)

For Equation 18.9 to be valid the power must remain constant for the entire period t . In turn this
requires the head across the pump, the flow through the pump and the upstream and downstream reser-
voir levels to remain constant. In practice, this would be very unlikely. In order to correctly assess the
energy usage by the pump it would necessitate an EPANET extended period simulation hydraulic model
to be run at say a time step of 5 minutes over the t period. The reservoir levels, the pump flow and
hence the head across the pump would be updated every 5 minutes to update the power calculation to be
the correct value. The energy usage would be accumulated for all of the 5 minute periods in t .

415
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

18.6.6 Electrical energy input

The electrical energy provided to a pump motor depends on the efficiency of the motor and is given by:

Pt
EE = --------- (18.10)
Em

PUMP INPUT ELECTRICAL ENERGY

where

• EE = electrical energy delivered to the pump motor (Wh) - usually converted to kWh by
dividing by 1000
• Pt = E = energy delivered to the pump shaft (Wh)
• E m = efficiency of pump motor (usually around 95%)
• P = shaft power delivered by pump motor (W or HP)
• t = time over which pump operates (h or hours)
Input power to shaft of the motor is not readily measured directly so it is usually obtained from the elec-
trical power input Pe to the motor.

P = Pe Em

where

• P = shaft power delivered by pump motor (W or HP)

• P e = electrical power delivered to the pump motor (W or HP)
• Em = motor efficiency (usually 0.95 to 0.98) (dimensionless)

18.6.7 Cost of pump operation

The cost of operating a pump over a period of time t and assuming the flow conditions and upstream
and downstream reservoir heads do not vary (as described in Equation 18.6.5) is given by:

Cost = EE  c  (18.11)

PUMP COST OF OPERATION

where

• Cost = cost of pump operation ($)

• EE = electrical energy delivered to the pump motor (Wh - usually converted to kWh by
dividing by 1000)
• c = cost of electricity (cents per kWh - converted to $/kWh by dividing by 100)

Example 18.2 Compute the following (i) the shaft power for a pump head of H P = 41 m , a pump flow
Q = 28 L/s at the operating point and a pump efficiency of  = 0.755 . (ii) the electrical energy input to the pump
motor for 5 hours of continuous operation at constant pumping conditions (iii) the cost of electricity for pump opera-
tion assuming the pump is operating from 12 midnight to 5am on an off-peak electricity tariff of 6 cents/kWh.
Assume a motor efficiency of 96%.

Answer – (i) Computation of the shaft power for a pump.

The pump shaft power is given by Equation 18.4 as

Q H  9789   0.028 41.0 14880

P = ---------------P = ---------------------------------------------- = 14880 W = --------------- kW = 14.9 kW
 0.755 1000

416
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

(ii) Computation of energy.

The shaft energy over 5 hours is given from Equation 18.9 as:

E = Pt = 14880  5  = 74400 Wh = 74400

--------------- kWh = 74.4 kWh
1000
The electrical energy input to the pump motor is given from Equation 18.10 by:

Pt 74400 77500

EE = ---------- = --------------- = 77500 W = --------------- = 77.5 kWh
Em 0.96 1000

(iii) Computation of cost of electrical energy.

The cost of electrical energy input to the pump motor is given from Equation 18.11 by:

6
Cost = EE  c  = 77.5  --------- = $4.65
 100

___________________________________________________________________________________

18.6.8 Pump speed

Pumps may either be fixed speed or variable speed. Fixed speed pumps are the focus of the rest of this
Chapter. For direct coupled or close coupled electric units a synchronous speed must be used. Examples
of synchronous speeds and the number of poles are 3000 rpm (2 poles), 1500 rpm (4 poles), 1000 rpm (6
poles), 750 rpm (8 poles) etc. Actual speeds are slightly less due to slip losses e.g. 2900, 1480, 990,
745 rpm approximately. Usual nominal speeds for various classifications of pumps are

• centrifugal — higher speeds (3000 or 1500 rpm)

• mixed flow — medium speeds (1500, 1000 rpm)
• axial flow — lower speeds (1000, 750, etc. rpm)
As the speed of a pump is increased the head – flow curve translates upwards as shown in Figure 18.8. If
the pump speed is decreased the head – flow curve for the pump is translated downwards.

Figure 18.8 shows the change in the pump curve when the pump speed is altered from 1480 rpm to
either 2900 rpm or 750 rpm. The tabular values of the pump curves for different speeds are shown in
Table 18.2. Details of exactly how the curves at different speeds are obtained is given in Example 18.6.
In Figure 18.8, it is evident that doubling the speed increases the head by a factor of four while the flow
is doubled.

417
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Table 18.2 Computation of pump curves for other rotational speeds (D = 0.25 m)
Pump A Pump A N=750 N=750 N=2900 N=2900
N=1480 N=1480 rpm rpm rpm rpm
rpm rpm
Q HP Dimensionless Dimensionless Q HP Q HP
(L/s) 3 2 2 (L/s) (L/s)
(m) Q  N D  gH P  N D  (m) (m)

0 50 0 0.003582907 0 12.8 0 192

10 49.9 0.000432432 0.003575741 5.1 12.8 19.6 191.6
20 49.6 0.000864865 0.003554244 10.1 12.7 39.2 190.4
30 49 0.001297297 0.003511249 15.2 12.6 58.8 188.1
40 48 0.00172973 0.003439591 20.3 12.3 78.4 184.3
50 46.5 0.002162162 0.003332104 25.3 11.9 98 178.5
60 44.4 0.002594595 0.003181622 30.4 11.4 117.6 170.5
70 41.8 0.003027027 0.00299531 35.5 10.7 137.2 160.5
80 38.6 0.003459459 0.002766004 40.5 9.9 156.8 148.2
90 34.6 0.003891892 0.002479372 45.6 8.9 176.4 132.8
100 30 0.004324324 0.002149744 50.7 7.7 195.9 115.2

Figure 18.8 H – Q pump curves for three different speeds

18.6.9 Power – torque – speed relationship.

The relationship between pump shaft power and the applied torque is

P = T (18.12)

where

• P = shaft power delivered by pump motor (W or HP)

•  = angular speed (rad/s)
• T = torque (N – m or ft-lb)
The angular speed may be defined in terms of the rotational speed N in rpm as

418
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

2N
 = ----------- (18.13)
60

and thus

 2N - =  2
P = T
------------------- ------ TN (18.14)
60  60 

18.7 Pump System Curve

The pump system curve depends on components attached to the pump. These include the difference
between the elevations of the lower and upper reservoirs as well as the losses in the pipe work and fit-
tings attached to both sides of the pump. The static head for a pumping system is defined as the differ-
ence in elevation between the lower and the upper reservoirs as shown in Figure 18.1

H S = z2 – z1 (18.15)

where

• H S = static head for a pumping system (m or ft)

• z 2 = elevation of the water surface of the reservoir or tank on the suction side of the pump
(m or ft)
• z 1 = elevation of the water surface of the reservoir or tank on the discharge side of the
pump (m or ft)
The static head HS and friction losses (both in suction line and the discharge line and minor losses) must
be taken into account in developing the system curve for the pumping system. The static head or the ele-
vation difference between the upstream reservoir and the downstream reservoir is not a function of flow
Q through the system. The friction losses are losses due to pipe friction and minor losses vary approxi-
mately with the square of the flow. From the Darcy – Weisbach equation (Equation 4.3) the total friction
loss for the pipe work in series in the pumping system is given by

2 2
LV L Q
f =  f ---- ------ =
D 2g  f ---- -----------
D 2g A
(18.16)

Minor losses for multiple losses in the system from Equation 5.1 are given by

2 2
Q
 hL=  K V
------ =  K ------------2-
2g 2g A

The system curve for a pumping system is given as

2 2
fL- ------------
Q - + K ------------
Q -
system curve = H S +  -----
D 2g A 2  2g A 2
(18.17)

PUMPING SYSTEM CURVE

where

• H S = static head for a pumping system (m or ft)

419
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

2
fL Q
•  ------ ------------- = friction losses for the pipework attached to the pump both on the suction side
D 2g A 2
and the discharge side of the pump (m or ft)
2
Q
•  K ------------2- = minor losses for components in the pipework attached to the pump (m or ft)
2g A
18.8 Pump Operating Point
For the pump – pipeline system as depicted in Figure 18.1, the head – flow curve is shown in
Figure 18.9. The operating point is where the H – Q curve for the pump intersects with the system curve
of Hs+(sum of the pipe friction losses   h f  and minor losses   h L  in the suction and delivery
pipework) — see Equation 18.17.

Figure 18.9 H – Q pump curve, system curve and efficiency curve for a constant speed

The properties of the pipe network used to determine the system curve in Figure 18.9 is shown in
Table 18.3.

Table 18.3 Pipeline characteristics

Property Quantity Units
Pipe length 1000 meters
Pipe diameter 300 mm
Roughness height 0.25 mm
Kinematic viscosity of water –6 m2/s
1.140 10
Static head 35 meters

Example 18.3 Consider the following pumping system with Pump A (see Figure 18.5 and Table 18.1) connected to a
pipeline with a diameter D = 300 mm , pipe length L = 1000 m , pipe roughness  = 0.25 mm and water viscos-
–6 2
ity  = 1.140 10 m /s at 15 degrees Celsius. The pump – pipeline system connects two reservoirs with the dif-
ference in water surface elevations of the lower and upper reservoirs of 35 m (the static head H S = 35.0 meters ).

(i) Determine the system curve

(ii) Find the operating point for the pump.

Answer – (i) Determination of the system curve.

420
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

The system curve is based on the friction loss in the pipeline. Results are shown in Table 18.4. Consider a flow of 70 L/
s.

The area of the pipe is

2 2
D   0.3  2
A = ---------- = ------------------ = 0.07069 m
4 4

The velocity of flow is

 1000- = 0.99 m/s

V = Q
---- = 70
---------------------
A 0.07069

The Reynolds number is

 0.99   0.3 
Re = VD
-------- = ---------------------------
- = 260300
 1.141 10
–6

The Darcy – Weisbach friction factor from the Swamee and Jain equation (Equation 4.61) is

0.25 0.25
f = ---------------------------------------------------------
- = ----------------------------------------------------------------------------------
- = 0.0202
  5.74  2 0.25 5.74 2
log 10 ------------ + ------------ log 10 --------------------- + ----------------------------

 3.7 D 0.9  3.7  300  0.9
Re  260300 

The head loss based on the Darcy – Weisbach head loss formula (Equation 4.3) is

2 2
fL V 0.0202  1000   0.99 
h f = ------ ------ = --------------------------------- ------------------ = 3.36 m
D 2g 0.3 2  9.81 
The system curve head is

H S + h f = 35.0 + 3.4 = 38.4 m

The plot of the system curve head is shown in Figure 18.10.

_________________________________________________________________________________
Answer – (ii) The operating point.

The flow is 77 L/s and the pumping head is 39 m.

Figure 18.10 System curve, pump curve and operating point

421
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Table 18.4 System curve for Example 18.3

18.9 Pump Specific Speed

When an engineer is required to select a pump for a system, assistance is usually available from pump
manufacturers and is often quite valuable. However, it is always good practice for the engineer to be
fully informed on the basics of pumps and the operating characteristics so that pumps may be selected
correctly.

The specific speed for a pump is a very useful number to determine in the process of selecting a pump. It
is defined as

12
NQ -
N s = ---------------- (18.18)
34
HP

SPECIFIC SPEED OF A PUMP

where for SI units

• N s = specific speed (units depend on system of units being used - see Equation 18.19 and
Equation 18.20)
• N = rotational speed (rpm)
• Q = pump flow (m3/s or ft3/s – sometimes the flow may be given in L/s)
• HP = pump head (m or ft)
The units of the specific speed are not rpm but actually have the following dimensions for SI units

12 3 12
 N   Q   rpm   m /s 
--------------------------- = --------------------------------------- (18.19)
34 34
H P  m

For US customary units the units of the quantities used to define the specific speed of a pump are

• N in rpm
• Q in gpm (US gallons)
• HP in feet
In this case the units of Ns are

422
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

12
 N   Q   rpm   gpm  1  2
--------------------------- = -------------------------------------- (18.20)
34 34
H P   ft 

A convention is adopted here that whenever the specific speed for a pump is given then the units should
be specified in square brackets following the number. The reason that this practice should be followed is
so that it is very clear what the units of the quantities were that were used to calculate the value of spe-
cific speed for the particular pump. Examples of units are

• NS (SI units) = 20 [rpm, m3/s, m]

• NS (US customary units) = 70 [rpm, US gpm, ft] for a different pump than in the example
just above in previous bullet point
The relationship between specific speed in the two sets of units is

N S  in US Customary units 
N S  in SI units  = -----------------------------------------------------------------------
- (18.21)
51.7

Another way to think of specific speed of a pump is to consider it as the rotational speed for a unit flow
(i.e. 1 m3/s or 1 US gpm) and a unit pumping head (i.e. 1 m or 1 ft). For these conditions from
Equation 18.18 the specific speed equals the rotational speed as

NS = N (18.22)

where

• N s = specific speed
• N = actual rotational speed of the impeller of the pump (rpm)
18.10 Types of Pumps
Three types of pumps are considered in this Chapter that are relevant to water distribution systems.
These include centrifugal, mixed flow and axial flow pumps. A typical configuration for a pumping sys-
tem showing a pump, check valve and bypass line is shown in Figure 18.12. A reflux valve is shown in
A duo check reflux valve is shown in Figure 18.14 and Figure 18.15.

423
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.11 Photograph of a reflux valve and a butterfly valve (on the left) –
flow is from right to left

Figure 18.12 Pumping system with check valve and bypass line

424
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.13 Photograph of a duo check valve installed in a pumping system –

flow is from right to left

Figure 18.14 Photograph of a duo check valve – flow goes into the valve from this side

425
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.15 Photograph of a duo check valve – flow comes into the valve from the other side

18.10.1 Centrifugal pumps

This type of pump is used for a substantial increase in pressure (or head) for a relatively small flow rate.
Thus the magnitude of the specific speed is relatively small for a centrifugal pump.

12
NQ  small 
N s = ----------------- = N ----------------- (18.23)
HP
34  large 

The specific speed range for centrifugal pumps is as follows

• NS (SI units) = 10 to 60 [rpm, m3/s, m]

• NS (US customary units) = 500 to 3000 [rpm, US gpm, ft]
Flow through the impeller of a centrifugal pump occurs radially. The flow is perpendicular to the pump
shaft.

18.10.2 Mixed flow pumps

This type of pump is a combination of centrifugal and axial flow to provide a moderate pressure increase
for moderate flow rates. Thus the magnitude of the specific speed is moderate for mixed flow pump.

12
NQ  moderate 
N s = ----------------- = N --------------------------- (18.24)
HP
34  moderate 

The specific speed range for partly mixed flow pumps

• NS Sunits = 50 to 90 [rpm, m3/s, m]

• NS US customary units = 1500 to 4500 [rpm, US gpm, ft]
The specific speed range for fully mixed flow pumps is as follows

• NS Sunits = 90 to 155 [rpm, m3/s, m]

• NS US customary units = 4500 to 8000 [rpm, US gpm, ft]

426
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Mixed flow pumps are popular when the space available is restricted in places such as wells when multi
stages (pumps in series) are used.

18.10.3 Axial flow pumps

This type of pump is used for large flow rates for relatively small increases in head. The impeller resem-
bles a propeller. For an axial flow pump the fluid passes over the vanes parallel to the pump axis. Thus
the magnitude of the specific speeds is large for an axial flow pump.

12
 large 
N Q - = N -----------------
N s = ---------------- (18.25)
HP
34  small 

The specific speed range for axial flow pumps is

• NS Sunits = 155+ [rpm, m3/s, m]

• NS US customary units = 8000+ [rpm, US gpm, ft]
Each of the different types of pumps has a different shaped impeller as shown in Figure 18.16.

Figure 18.16 Impeller types for various specific speeds (Streeter and Wylie p.392)

Two pump selection examples are given below.

Example 18.4 A pump is required to handle a pumping head of 18 m and a flow of Q = 30 L/s. The available motor
speed is N = 1450 rpm. Compute the specific speed and select the appropriate type of pump.

Answer – Computation of the specific speed of a pump.

The specific speed is given by Equation 18.18 as

12 12
N Q - = ------------------------------------------
1450   0.030  - = 28.7 [rpm, m3/s, m]
N s = ----------------
34 34
HP  18 

As the specific speed of NS is much less than 60 (the approximate breakpoint between centrifugal and mixed flow
pumps) thus only a centrifugal pump is suitable for this application.

Example 18.5 A pump is required to handle a pumping head of 11.5 m and a flow of Q = 480 L/s. Again the avail-
able motor speed is N = 1450 rpm. Compute the specific speed and select the appropriate type of pump.

Answer – Computation of the specific speed of a pump.

The specific speed is given by Equation 18.18 as

427
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

12 12
N Q - = ------------------------------------------
1450   0.480  - = 160.9 [rpm, m3/s, m]
N s = ----------------
34 34
HP  11.5 

The specific speed for this application is on the border of a mixed flow pump and an axial flow pump. Thus either a
mixed flow or propeller type pump would be suitable.

___________________________________________________________________________________

18.11 Colt Industries Specific Speed Categorization for Pumps

Table 18.5 Colt Industries pump classification guidelines

Class Pump type Specific Specific Head D2/D1*
speed speed range
(NS in US (NS in
customary units) SI units) (m)
Centrifugal Radial 500 to 3000 10 to 60 HP > 45 m 2+
HP > 150'
Centrifugal Double Suction** 1000 to 3500 20 to 70 HP > 30 m 1.5
HP > 100'
Mixed Flow Francis 1500 to 4500 30 to 90 HP = 20 to 45 m 1.5
HP = 65 to 150'
Mixed Flow Mixed Flow 4500 to 8000 90 to 150 HP = 10 to 20 m 1.3 to 1.1
Francis HP = 35 to 65'
Axial Flow Propeller 8000 and up 150 and up HP = 0.3 to 12 m
HP = 1 to 40'

*where D = the distance from the centerline of the pump to the impeller exit and D
1 2 = distance from centerline to the
entrance of the impeller (m or ft).

** The specific speed for a double suction pump is obtained by using one – half the flow. For multi – stage pumps the
head per stage is used.

18.12 Pump Similarity

This section considers similarity of pumps. Dimensionless numbers are developed for pumps in
Section 18.13. There are dimensionless numbers associated with the following pump quantities — pump
head, flow, speed, torque and power. Pump similarity and the dimensionless numbers lead to pump
affinity laws as given in Section 18.15 that enable computation of pump operation of a geometrically
similar pump (perhaps of the same specific speed but of a different size diameter impeller).

18.12.1 Momentum and energy changes in a pump

In the impeller of a pump, each vane deflects the fluid jet and changes its momentum. Thus a force is
exerted on the jet by the pump vanes and thus work is done by the displacement of the vane. Pumps
work continuously on the fluid converting the torque of the motor shaft to energy of the fluid.

18.12.2 Requirements for similarity of pumps

Similarity relations may be developed for similar shaped pumps of different sizes. Similar pumps have
the same specific speed Ns. The following conditions are necessary for similarity

• geometric similarity
• kinematic similarity evidenced by similar velocity vector diagrams at the entrance to, and at
the exit from, the impeller

428
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

• dynamic similarity such that force ratios are the same in both pumps

Figure 18.17 The velocity vector diagram at the exit of a pump impeller

Often pump performance characteristics are determined in model study tests. These are then scaled up to
prototype size pumps. In considering similarity of pumps the viscous effects must be neglected because
it is not possible to maintain the same Reynolds number Re in the model and prototype while satisfying
the above conditions.

18.12.3 Similarity of velocity vector diagrams for pumps

Two geometrically similar pumps having similar velocity vector diagrams are termed as homologous or
exhibit similarity. The velocity vector diagrams may be constructed both for the flow entering the pump
impeller and leaving the pump impeller. In addition homologous units have similar streamlines. The
vectors in the velocity diagram are defined as

2N
u = r = r -----------
60

where
• u = vector of peripheral velocity of the impeller (m/s or ft/s)
• r = radius of impeller (m or ft)
•  = angular speed (rad/s)

V = vector of absolute velocity of flow leaving the impeller (someone standing outside the impeller
would see the absolute velocity). This is the vector sum of u and v as follows:

V = u+v

The radial velocity is:

Vr  Q

where

• Vr = radial component of the absolute velocity V that is proportional to the flow

18.12.4 Similar velocity diagrams

Two different sized impellers with exactly the same shape have similar velocity diagrams. These pumps
are said to be homologous units and have the same specific speed NS.

429
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.18 Two similar velocity diagrams for different sized impellers of identical shape

The following quantity must be constant to have two pumps to be considered homologous units or to
exhibit similarity

Vr V sin 
------ = ---------------- = constant (18.26)
u u

where

•  = angle that the absolute velocity makes with the tangential direction u
•  = blade angle of the vane in the impeller

For homologous or similar units the blade and  or the angle  in the velocity diagrams must be the
same.

Equation 18.26 is a very important relation for the similarity of pumps.

18.12.5 Pump flow

The flow is proportional to the radial component of the absolute velocity normal to the impeller Vr. The
flow is also proportional to the area of flow D2 where D is the diameter of the impeller. Thus

2
Q  V rD (18.27)

or

2
Q  V sin D (18.28)

Thus the flow through the pump is directly proportional to the absolute velocity of flow leaving the
impeller, the sine of the angle that the absolute velocity makes with the tangential direction and the
diameter of the impeller squared.

18.12.6 Pump speed

The angular velocity u is related to the rotational speed N as

2N DN
u = r = r ----------- = ------------- (18.29)
60 60

or rearranging Equation 18.29 in terms of the rotational speed N gives

430
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

60u
N = ---------
D

It follows that

u
N  ----
D

or

u  ND (18.30)

Thus the magnitude of the peripheral velocity of the impeller is proportional to both the rotational speed
and the impeller diameter.

18.13 Dimensionless Numbers for Pumps

Five dimensionless numbers for pumps are developed in this Section.

18.13.1 Dimensionless pump flow

The first dimensionless number is related to the pump flow. From Equation 18.26 it follows that

V sin 
---------------- = C 1 (18.31)
u

and from Equation 18.28 (Q  V sin D2) it follows that

Q
V  -------------------- (18.32)
2
D sin 

Combining Equation 18.31 and Equation 18.32 gives

Q sin 
---------------------- = C 2
2
uD sin 

resulting in

Q- = C
--------- 2 (18.33)
2
uD

Now substituting Equation 18.30 (u  ND) into Equation 18.33 gives

Q
--------------------- = C 2 (18.34)
2
 ND D

resulting in

Q
C Q = ----------- = constant (18.35)
3
ND

DIMENSIONLESS PUMP FLOW

This is the first important dimensionless number for similar pumps.

431
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

The units of CQ are as follows

3
L
------
Q T
 C Q  = --------------------- = -------------------- =  1 
3 1 3
 N  D  ---  L 
T

The time unit may be different for the flow Q and the rotational speed N but this difference just means
that CQ is multiplied by a constant.

Equation 18.35 is only true for similar or homologous pumps of the same specific speed NS.

18.13.2 Dimensionless pump speed

The pump head HP is now introduced to obtain an expression related to the specific speed for the pump.
From the equation for the flow through an orifice that may be assumed to apply to flow through a pump

Q = C d A 2g H P (18.36)

where

• Q = flow through the pump (m3/s or ft3/s)

• Cd = discharge coefficient (dimensionless)
• A = area of flow (m2 or ft2)
• g = gravitational acceleration (m/s2 or ft/s2)
• HP = pump head (m or ft)
Substituting for the area in terms of the diameter of the impeller (D):

2
D
Q = C d  ---------- 2g H P (18.37)
 4 

Rearranging Equation 18.37 gives

Q
----------------------- = constant (18.38)
2
D gH P

Taking the square root of this expression and then cubing gives

32
1 Q
------- ------------------------- = constant (18.39)
3 34
D g H P

Combining with Equation 18.35 (Q/ND3 = constant) and inverting gives

32
N- ------------------------
Q
--- - = constant (18.40)
Q  g H 3  4
P

432
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

resulting in

12
1 - ----------------
NQ - = constant
C N = ---------- (18.41)
34 34
g HP

DIMENSIONLESS PUMP SPEED

This is the second important dimensionless number for similar pumps.

The specific speed as defined earlier is the second quotient in Equation 18.41 and it follows that

1
----------- N S = constant (18.42)
34
g

The omission of the gravity term of Equation 18.42 in the normal definition of the specific speed (as per
Equation 18.18) is the reason for the difference in specific speeds in the SI units system and for US cus-
tomary units.

The units of CN are dimensionless as given by

3 12
12 --1- L ------
1  N Q  1 T T
 C N  = ---------------- --------------------------- = -------------------- ----------------------------- =  1 
34 34 L- 3  4  L  3  4
g  H P  -----
2
T

18.13.3 Dimensionless pump head

A further dimensionless number may be determined by eliminating the flow Q in Equation 18.35 (Q/
2
ND3= constant) and Equation 18.38 (Q   D g H P  = constant ) and squaring to give

gH P
C H = -------------- = constant (18.43)
2 2
N D

DIMENSIONLESS PUMP HEAD

This is the third important dimensionless number for similar pumps.

The units of CH are dimensionless as shown below

L-  L 
-----
 g  H P  T
2
 C H  = -----------------------
- = ----------------------- =  1 
2 2 1 2
 N  D  ---  L 
2
T

Example 18.6 Consider a mixed flow pump with an impeller of diameter 150 mm, a flow of Q = 50 L/s, a rotational
speed of N = 1800 rpm and a pumping head of H = 4 m operating at BEP (best efficiency point).

(i) What speed should a pump of impeller diameter of 450 mm (3 times larger) be operated at to produce a flow of
Q = 0.9 m3/s (18 times larger) at the best efficiency point (BEP)?
(ii) What pumping head is developed by the 450 mm diameter pump operating at the speed computed in Part (i)?

Answer – (i) Computation of the pump speed.

433
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

For similar machines dimensionless pump flow number given by Equation 18.35

Q
C Q = ----------- = constant
3
ND

and

Q1 Q2
----------------
- = ----------------
-
3 3
N 1 D1 N 2 D2

Thus

Q1 50 900
----------------
- = --------------------------------- = ------------------------
3 3 3
N 1 D1  1800   150  N 2  450 

Solving for the rotational speed N2 gives

N2 = 1200 rpm

_________________________________________________________________________________
Answer – (ii) Computation of the head developed by the pump.

The head the pump developed is given from Equation 18.43 by

gHP
C H = -------------
- = constant
2 2
N D

or

gHP gHP
1 2
------------------
- = ------------------
-
2 2 2 2
N 1 D1 N 2 D2

Thus

 9.81   4.0   9.81  H P

2
------------------------------------ = -----------------------------------
-
2 2 2 2
 1800   150   1200   450 

Thus solving for the pumping head for the 450 mm diameter pump at the new speed of 1200 rpm gives

H P = 16 m
2

_________________________________________________________________________________

18.13.4 Dimensionless pump power

The pump shaft power is defined from both Equation 18.4and Equation 18.12 as

QH
P = T = ---------------P (18.44)


The rotational speed is defined as

2N
 = ----------- (18.45)
60

It then follows from Equation 18.44 that the shaft power is proportional to

P  QH P

for a given operating point on the pump curve. Thus it follows that

P
Q  -----------
H P

434
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

or

P -
Q  -------------- (18.46)
g H P

Eliminating the flow in Equation 18.46 using Equation 18.35 (Q/ND3 = constant) gives

P
---------------------------- = constant (18.47)
3
g H P N D

Solving for the head in Equation 18.43 gives

2 2
N D
H P  -------------- (18.48)
g

Eliminating the head HP between Equation 18.47 and Equation 18.48 gives

gP
---------------------- = constant (18.49)
3 5
g N D

resulting in

P - = constant
C P = ----------------- (18.50)
3 5
N D

DIMENSIONLESS PUMP POWER

This is the fourth important dimensionless number for similar pumps.

The units of CP are dimensionless as follows

2
ML
------------
3
P T
 C P  = --------------------------------- = ----------------------------------------- =  1 
3 5 M-  --- 1  3  L 5
 N   D -----
3  T
L

18.13.5 Dimensionless pump torque

The dimensionless torque may be derived from the power – torque relationship. From Equation 18.44 it
follows that

 2N  QH
T ---------------- = ---------------P (18.51)
60 

Rearranging Equation 18.51 gives

T  2N  2 TN
Q = ------------------------ =  ---------- ------------ (18.52)
60g H P  60g  H
P

Thus the flow at a given operating point along the pump curve is proportional to

435
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

TN
Q  ----------- (18.53)
H P

Combining Equation 18.53 with Equation 18.35 (Q/ND3 = constant) to eliminate Q gives

TN
----------------------------- = constant
3
H P  N D 

resulting in

T
------------------- = constant (18.54)
3
H P D

From Equation 18.43 (gHP/(N2D2) = constant) and taking gravity to be combined with the constant

HP
-------------- = constant (18.55)
2 2
N D

Combining Equation 18.54 with Equation 18.55 and eliminating HP gives

T
C T = ------------------ = constant (18.56)
2 5
N D

DIMENSIONLESS PUMP TORQUE

This is the fifth important dimensionless number for similar pumps.

The units of CT are dimensionless as shown below

2
ML
-----------
2
T  T
 C T  = --------------------------------- = ----------------------------------- =  1 
2 5 M 1 2
 N  D  ------ ---  L 
5
3 T
L

436
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

18.13.6 A summary of the dimensionless pump parameters

Thus a summary of the dimensionless parameters for pumps is as follows

Flow (Equation 18.35)

Q - = constant
C Q = ----------
3
ND

Speed (Equation 18.42)

12
1 NQ
C N = ----------- ----------------- = constant
34 34
g HP

Pump head (Equation 18.43)

gH P
C H = -------------- = constant
2 2
N D

Pump shaft power (Equation 18.50)

P
C P = ------------------ = constant
3 5
N D

Pump torque (Equation 18.56)

T
C T = ------------------ = constant
2 5
N D

18.14 Dimensionless Pump Curves

18.14.1 Dimensionless Head – Flow Curve
The H – Q curve and efficiency-Q curve for pump A (as seen earlier in the Chapter) are both shown in
Figure 18.19.

Figure 18.19 A head flow curve for pump A

The pump characteristic curve head – flow in dimensionless form is

437
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

gH P Q-
-------------- versus ---------- (18.57)
2 2 3
N D ND

or

C H versus C Q

Tabular values of the dimensionless head curve for pump A may be computed as shown in Table 18.6
for a rotational speed of N = 1480 rpm and pump impeller diameter D = 0.3 m. A plot of the dimension-
less head curve for pump A is shown in Figure 18.20.

Table 18.6 Pump A dimensionless H versus Q curve

Flow Pump Dimensionless Q Dimensionless H Efficiency Dimensionless P
Q Head A Equation 18.35 Equation 18.43 % Equation 18.50
(L/s) (m) Polynomial
Table 18.1 Equation 18.6
0 50.0 0.0000 0.002488 0
10 49.9 0.2503 0.002483 22.2 2.8E-09
20 49.6 0.5005 0.002468 41 3.01E-09
30 49.0 0.7508 0.002438 56.5 3.24E-09
40 48.0 1.0010 0.002389 68.7 3.48E-09
50 46.5 1.2513 0.002314 77.7 3.73E-09
60 44.4 1.5015 0.002209 83.4 3.98E-09
70 41.8 1.7518 0.002080 85.8 4.25E-09
80 38.6 2.0020 0.001921 84.9 4.53E-09
90 34.6 2.2523 0.001722 80.8 4.79E-09
100 30.0 2.5025 0.001493 73.3 5.1E-09

Example 18.7 For Pump A at a flow of 40 L/s (i) compute the dimensionless head and (ii) compute the dimension-
less flow (iii) compute the dimensionless power and (iv) compute the specific speed of the pump. The rotational
speed of the pump is N=1480 rpm and the impeller diameter is 300 mm.

Answer – (i) Computation of the dimensionless head.

Using Equation 18.57 give:

gHP  9.81   48  - = 0.002389

C H = -------------
- = ---------------------------------
2 2 2 2
N D  1480   0.3 

This is the value given in Table 18.6.

(ii) Computation of the dimensionless flow.

Using Equation 18.57 give:

Q 40
C Q = ------------ = -------------------------------- = 1.0010
3 3
ND  1480   0.3 

These are also the values given in Table 18.6.

(iii) Computation of the dimensionless power for pump A at 40 L/s.

438
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Using Equation 18.50 gives:

P - = -----------------------------------------------
27.4 -8
C P = ----------------- - = 0.3484 10
3 5 3 5
N D 998.2  1480   0.3 

This is the value given in Table 18.6.

(iv) Computation of the specific speed.

Switch to a discharge of 70 L/s and pump head of 41.8 m at the best efficiency point (85.8%). Thus the specific speed
is given by Equation 18.18 as

70 1  2
 1480   ------------
12  1000 
N Q - = -----------------------------------------
N s = ---------------- - = 24 [rpm, m3/s, m]
34 34
HP  41 8 

___________________________________________________________________________________

Figure 18.20 Dimensionless H – Q curve for an homologous series of pumps

For a particular pump specific speed the plot shown in Figure 18.20 is a single curve irrespective of the
size of the pump (assuming scale effects due to viscosity are ignored).

18.14.2 Dimensionless efficiency curve  versus CQ

The characteristic pump efficiency – flow curve in dimensionless form is

Q-
 versus ---------- (18.58)
3
ND

or

 versus C Q

A dimensionless efficiency for Pump A (from Example 18.1, Table 18.1 and Figure 18.6) is shown in
Figure 18.21. It is a plot of  versus CQ. Values are taken from Table 18.6.

439
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.21 Dimensionless  – Q curve for an homologous series of pumps

18.14.3 Power – flow curve in dimensionless form

The characteristic power – flow curve in dimensionless form from Equation 18.50 and Equation 18.35 is

P Q
------------------ versus ----------- (18.59)
3 5 3
N D ND

or

C P versus C Q

A dimensionless power curve for Pump A (from Example 18.1, Table 18.1 and Figure 18.7) is shown in
Figure 18.22. It is a plot of CP versus CQ. Values are taken from Table 18.6.

Figure 18.22 Dimensionless P – Q curve for an homologous series of pumps

18.15 Pump Affinity Laws

For homologous or similar machines the four pump dimensionless characteristics of head, torque, power
and efficiency (CH versus CQ, CT versus CQ, CP versus CQ and versus CQ), mentioned previously are
identical. Advantages of this type of representation include

• the performance of a whole range of geometrically similar machines are represented by a

few curves
• the above dimensionless curves may be obtained from one machine operating at one speed
over a range of flows

440
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

18.15.1 Pump affinity law number 1 — a change in pump speed

Law 1 — The pump Q, HP and P are functions of pump speed N for the same size impeller
(D = constant). This is usually achieved by the replacement or alteration of the existing motor of the
pump.

Flow – speed relation for pump affinity law number 1

From the expression for the dimensionless flow as given by Equation 18.35

Q - = constant
C Q = ----------
3
ND

it follows that

Q1 Q2
----------------
- = ----------------
- (18.60)
3 3
N 1 D1 N 2 D2

For a constant pump impeller diameter it follows that

D1 = D2 (18.61)

Thus from Equation 18.60 it follows that if the speed changes from N1 to N2 then the flow changes
according to the following law

Q Q
------1- = ------2-
N1 N2

or

Q1 N
------ = ------1- (18.62)
Q2 N2

Thus the pump flow varies linearly with an increase in rotational speed of the pump impeller.

Pump head – speed relation for pump affinity law number 1

From dimensionless head expression of Equation 18.43

gH P
C H = -------------- = constant
2 2
N D

Thus for two different pumping conditions

gH P gH P
1 2
-------------------
- = -------------------
-
2 2 2 2
N 1 D1 N 2 D2 (18.63)

Using the constant diameter condition of Equation 18.61 gives

HP HP
----------1 = ----------2
2 2
N1 N2

441
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

or

HP N 2
----------1 =  ------1- (18.64)
HP N 
2 2

Thus the pump head is related to the square of the rotational speed of the pump.

Power – speed relation for pump affinity law number 1

From dimensionless power expression of Equation 18.50

P
C P = ------------------ = constant
3 5
N D

For two different pumping situations

P1 P2
----------------------
- = ----------------------
- (18.65)
3 5 3 5
N 1 D 1 N 2 D 2

Using the constant diameter condition of Equation 18.61 gives

P1 P2
---------
- = ---------
-
3 3
N1 N2

or

P N 3
-----1- =  ------1- (18.66)
P2 N 
2

The pump shaft power is related to the cube of the rotational speed of the pump impeller.

A summary of the Law 1 pump affinity relationships for a constant diameter impeller follows

Q N
------1 = ------1- from Equation 18.62
Q2 N2

HP N 2
----------1 =  ------1-  from Equation 18.64
HP N 
2 2

P N 3
-----1- =  ------1-  from Equation 18.66
P2  N2 

PUMP AFFINITY LAW Number 1 – CONSTANT IMPELLER DIAMETER

18.15.2 Pump affinity law number 2 – a change in pump impeller diameter

Law 2 — Q, HP and P are functions of diameter of impeller. This is usually achieved by trimming the
impeller of the existing pump. Note that this is only applicable for a maximum of 10% change in diame-
ter.

The rotational speed in this case N = constant and thus

442
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

N1 = N2 (18.67)

Flow – diameter relation for pump affinity law number 2

From Equation 18.60

Q1 Q2
----------------
- = ----------------
-
3 3
N 1 D1 N 2 D2

For constant rotational speed given by Equation 18.67 it follows that

Q1 Q2
--------
- = --------
-
3 3
D1 D2

or

Q D 3
------1 =  ------1-  (18.68)
Q2  D2 

Thus the flow is a function of the cube of the diameter.

Pump head – diameter of impeller relation for pump affinity law number 2

From Equation 18.63

gH P gH P
1 2
-------------------
- = -------------------
-
2 2 2 2
N 1 D1 N 2 D2

For constant rotational speed given by Equation 18.67 it follows that

HP HP
----------1 = ----------2
2 2
D1 D2

or

HP D 2
----------1 =  ------1- (18.69)
HP  D 2
2

Thus the pump head is related to the square of the diameter ratio at a constant rotational impeller speed.

Power – diameter of impeller relation for pump affinity law number 2

From Equation 18.65

P1 P2
----------------------
- = ----------------------
-
3 5 3 5
N 1 D 1 N 2 D 2

For constant rotational speed given by Equation 18.67 it follows that

443
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

P1 P2
--------
- = --------
-
5 5
D1 D2

or

P D 5
-----1- =  ------1-  (18.70)
P2  D2 

Thus the pump power is related to the impeller of the diameter to the fifth power.

As mentioned earlier, a smaller impeller is usually achieved by trimming and using the same sized cas-
ing. If the assumption in Law No. 2 regarding the maximum allowable diameter change is violated it is
best to go back to manufacturer's data.

A summary of the Law 2 pump affinity relationships for a constant rotational speed follows

Q1 D 3
------ =  ------1-  from Equation 18.68
Q2 D 
2

HP D 2
----------1 =  ------1-  from Equation 18.69
HP D 
2 2

P D 5
-----1- =  ------1-  from Equation 18.70
P2 D 
2

PUMP AFFINITY LAW number 2 – CONSTANT ROTATIONAL SPEED

18.16 Multi – Impeller Machines

18.16.1 A multi – stage pump
A multi – stage pump has more than one impeller in series. If the impellers are the same then the charac-
teristics of a multistage pump has the total pump head HP (for all stages) equally split amongst each
impeller. From Equation 18.43 it follows that

H
g  -------P-
 n 
CH = ----------------- = constant (18.71)
2 2
N D

where

• n = number of stages in the multi stage pump

From Equation 18.35

Q
C Q = ----------- = constant (18.72)
3
ND

that is identical to the condition for the single stage impeller.

444
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

18.16.2 A multi – flow pump

A multi – flow pump (usually only 2 back to back) has flow equally split between the impellers. From
Equation 18.43

gH P
C H = -------------- = constant (18.73)
2 2
N D

that is identical to the condition for a single stage impeller.

From Equation 18.35 it follows that

Q ---- 
 n
CQ = ----------- = constant (18.74)
3
ND

where

• n = number of pumps in parallel

18.17 Effect of Reynolds Number on Pump Behavior
A characteristic length for the pump impeller is its diameter D. A characteristic velocity is the velocity
of tip of the runner as given by

u = ND (18.75)

The form of a pump Reynolds number (in contrast to Equation 2.15) is as follows

2
uD  ND D N D
Re = ------- = ---------------------- = --------------- (18.76)
  

One condition for dynamic similarity is that the Reynolds number be the same for each pumping head
along the pump curve.

18.17.1 Low Reynolds number flows

Oil flow often has a low Reynolds Number Re. Viscous effects dominate when the Reynolds number is
low. The dimensionless relations hold with limited accuracy even if Re is the same. If Re is ignored then
significant errors may result. Modified experimental data must be used.

18.17.2 High Reynolds number flows

Air and water are examples where it is usual that the flow occurs at very high Reynolds numbers. Simi-
larity between flow situations in different sized pumps is good even if Re is not the same for the two
machines, provided Re are of the same order or very large. For this situation, flow inside the pump is
very turbulent and viscous effects tend to be small. Sometimes air is used to test models of pumps and
turbines.

18.18 Pumps in Series

Series pumps are useful when different pumping heads are required for the same flow. The head charac-
teristic curve for pumps in series is obtained by adding the ordinates (head) of the head – flow character-
istic curves of the individual pumps for the same value of flow (ASCE 1975). By continuity, pumps in
series must have the same flow passing through them. Two centrifugal pumps in series are shown in
Figure 18.23

445
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.23 Photograph of two pumps in series – flow is from right to left

Consider n pumps in series, two of which are shown in Figure 18.24. For pumps in series the flow QSR is
the same through all n pumps

Q SR = Q 1 = Q 2 =  = Q n (18.77)

The subscript SR refers to series pump operation.

The total resultant pumping head H P from the suction side of the first pump to the discharge side of
SR

the nth pump is the sum of the heads for the individual pumps at the same flow.

HP = HP + HP +  + HP (18.78)
SR 1 2 n

Now consider two pumps A and B in series as shown in Figure 18.24.

Figure 18.24 Pumps A and B in series

The total power is the sum of the power for the individual pumps as follows

P SR = P A + P B (18.79)

The power of each of the two pumps A and B is given from Equation 18.4 as

Q A H P
P A = ---------------------A (18.80)
A

446
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

and

Q B H P
P B = --------------------B- (18.81)
B

Solving for the efficiencies of the two individual pumps in Equation 18.80 and Equation 18.81 gives

Q A H A
 A = ------------------
- (18.82)
PA

and

Q B H B
 B = ------------------
- (18.83)
PB

For the two pumps A and B in series the overall efficiency is given as

Q A + B H A + B
 SR =  A + B = ----------------------------------
- (18.84)
PA + B

But from Equation 18.77 it follows that

Q SR = Q A + B (18.85)

and from Equation 18.78

HP = HP =H + HP (18.86)
SR A+B PA B

The total power PA+B is the sum of the power for each of the individual pumps A and B. Thus from
Equation 18.80 and Equation 18.81 it follows that

Q A H P Q B H P
P SR = P A + B = P A + P B = ---------------------A + --------------------B- (18.87)
A B

Now substituting Equation 18.85, Equation 18.86 and Equation 18.87 into Equation 18.84 gives

Q A + B H A + B Q SR  H P + H P  Q SR  H P + H P 
 A + B = ---------------------------------- A B
- = --------------------------------------------- A B
- = ------------------------------------------------- (18.88)
PA + B PA + PB Q A H P Q B H P
---------------------A + --------------------B-
A B

It follows then that

Q SR  H P + H P 
 A + B = -------------------------------------------------
A B
(18.89)
Q SR H P Q SR H P
--------------------A- + --------------------B-
A B

and

447
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

HP + HP
 A + B = ---------------------------
A B
- (18.90)
HP HP
A B
---------- + ----------
A B

EFFICIENCY FOR PUMPS IN SERIES

A summary of the head, flow, power and efficiency relationships for two pumps in series are given
below

Q SR = Q A = Q B from Equation 18.85

HP = HP =H + HP from Equation 18.86

SR A+B PA B

Q A H P Q B H P
P SR = P A + B = P A + P B = ---------------------A + --------------------B- from Equation 18.87
A B

HP + HP
 SR =  A + B = ---------------------------
A B
- from Equation 18.90
HP HP
A B
---------- + ----------
A B

EFFICIENCY FOR PUMPS IN SERIES

18.19 Pumps in Parallel

Pumps in parallel are useful when the head is fairly constant while the flow rate is required to vary over
a significant range. The head characteristic curve for pumps in parallel is obtained by adding the
abscissa (flows) of the head – flow characteristic curves of the individual pumps for the same value of
head (ASCE 1975). Pumps in parallel must deliver the same pump head across the pump from suction
side to discharge side of the pump.

Consider n pumps in parallel, two of which are shown in Figure 18.25. For pumps in parallel the flow
QPL (where the subscript PL refers to the term “parallel”) is the sum of the flows through all n pumps

Q PL = Q 1 + Q 2 +  + Q n (18.91)

The pumping head H P from the suction side to the discharge side of each pump is the same for all n
PL

pumps.

HP = HP = HP =  = HP (18.92)
PL 1 2 n

Now consider two pumps A and B in parallel as shown in Figure 18.25.

448
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

Figure 18.25 Pumps A and B in parallel

The total power of the two pumps A and B in parallel is the sum of the power for the individual pumps is

P PL = P A + P B (18.93)

The power of each of the two pumps A and B is given from Equation 18.4 as

Q A H P
P A = ---------------------A (18.94)
A

and

Q B H P
P B = --------------------B- (18.95)
B

For the two pumps A and B in parallel the overall efficiency is given as

Q A + B H A + B
 PL =  A + B = ----------------------------------
- (18.96)
PA + B

But from Equation 18.91 it follows that

Q PL = Q A + Q B (18.97)

and Equation 18.92 gives

HP = HP = HP (18.98)
PL A B

The total power PA+B for the two pumps in parallel is the sum of the power for each of the individual
pumps A and B. Thus from Equation 18.93, Equation 18.94 and Equation 18.95 it follows that

Q A H P Q B H P
P PL = P A + B = P A + P B = ---------------------A + --------------------B- (18.99)
A B

Now substituting Equation 18.97, Equation 18.98 and Equation 18.99 into Equation 18.96 gives

Q A + B H A + B   Q A + Q B H P   Q A + Q B H P
 A + B = ---------------------------------- PL
- = ------------------------------------------ PL
= -------------------------------------------------
- (18.100)
PA + B P A + PB Q A H P Q B H P
---------------------A + --------------------B-
A B

It follows then that

449
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

 Q A + Q B H P
 A + B = ------------------------------------------------
PL
- (18.101)
QAH P QB H P
PL PL
--------------------- + ---------------------
A B

and

Q A + QB
 A + B = -------------------------
- (18.102)
Q A QB
------- + -------
 A B

EFFICIENCY FOR PUMPS IN PARALLEL

A summary of the head, flow, power and efficiency relationships for two pumps in parallel are given
below

Q PL = Q A + Q B from Equation 18.97

HP = HP = HP from Equation 18.98

PL A B

Q A H P Q B H P
P PL = P A + B = P A + P B = ---------------------A + --------------------B- from Equation 18.99
A B

Q A + QB
 PL =  A + B = -------------------------
- from Equation 18.102
Q A QB
------- + -------
 A B

18.20 Pump Cavitation

To maintain a high level of pump efficiency it is important that there is no cavitation occurring in the
inlet to the pump impeller. When a pump is cavitating it sounds as if gravel is going through it. This may
occur if the pressure drops too low (i.e. to around the vapor pressure of the liquid flowing) at the region
of the inlet to the pump. A low pressure in the inlet region may be due to too great an elevation differ-
ence between the water surface elevation of the suction reservoir (or intake) and the centerline of the
pump and/or a long suction line that causes a large pressure drop along its length. To analyze the poten-
tial for cavitation to occur a term is introduced referred to as the net positive suction head. There are two
forms of this term including the “Net Positive Suction Head Required (NPSHR)” and “Net Positive Suc-
tion Head Available (NPSHA).”

The concept of net positive suction head is very critical in the selection of pumps to prevent cavitation.
NPSH is defined as the pressure required to cause liquid to flow through the suction side piping and fit-
tings and into the pump. The minimum pressure point in a pump normally occurs at the entrance to the
impeller vanes.

18.20.1 Net positive suction head required (NPSHR)

NPSHR is an important pump characteristic curve that is normally obtained from the manufacturer of
the pump. It is plotted against pump flow Q and is provided by the manufacturer. NPSHR is equivalent
to the drop in pressure between the suction flange of the pump and the entrance to the impeller vanes
plus the velocity head at the suction flange. The NPSHR is the minimum allowable pressure head at the
pump entrance that is above the vapor pressure head of the fluid flowing in order to prevent cavitation in
the impeller of the pump.

450
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

18.20.2 Net positive suction head available (NPSHA)

NPSHA is the amount of head available to force water into the inlet of the pump in relation to preven-
tion of cavitation in the pump impeller. It is a function of the system in which the pump operates. It
depends on the elevation of the centerline of the pump relative to the suction side reservoir level and
depends on head losses in the suction pipework and fittings. It represents the pressure head available to
force the liquid into the pump and should be calculated for all systems.

To determine the Net Positive Suction Head Available (NPSHA) for a pumping system consider the suc-
tion side details in Figure 18.26.

Figure 18.26 Definition diagram for net positive suction head available (NPSHA) for a pump

Write the energy equation from Equation 3.67 between the suction side reservoir (1) and a fluid particle
at the centerline of the pipe just as it enters the inlet to the pump (2)

* 2 * 2
p V1 p V2
z 1 + -----1- + --------
- = z 2 + -----2- + --------
2g  f  L
-+ h + h (18.103)
 2g 

where

• the asterisk on each of the pressure terms refers to absolute units (rather than gage units that
is the convention)
Let the difference between the elevation of the centerline of the inlet pipe to the pump and the water sur-
face elevation of the suction reservoir be defined as the static suction lift hs or

hs = z2 – z1 (18.104)

Thus if the suction side reservoir (section 1) is below the centerline of the pump inlet (section 2) the
value of hs is positive (see Figure 18.26) while if the suction side reservoir (section 1) is above the cen-
terline of the pump inlet (section 2) the value of hs is negative. The term suction lift is rather misleading.
Water cannot be sucked through the inlet line to the pump. The water must be forced into the suction
intake and along to the pump by atmospheric pressure acting on the surface of the suction side reservoir.

In Equation 18.103 the pressure at the surface of the reservoir is zero (gage units) or equal to the baro-
metric pressure in the atmosphere (absolute units of pressure). It is usual to work in absolute units of
pressure when dealing with pump cavitation computations. Thus the pressure on the surface of the suc-
tion side reservoir is barometric pressure

* *
p p
-----1- = -----b- (18.105)
 

451
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

The velocity of a fluid particle on the surface of the reservoir may usually be assumed to be zero. Thus
in Equation 18.103

2
V1
------ = 0 (18.106)
2g

Substituting into Equation 18.103 and rearranging

* * 2
p p V2
-----b- + 0 =  z 2 – z 1  + -----2- + --------
2g  f  L
-+ h + h (18.107)
 

or

* 2 *
p V2 p
h s + -----2- + --------
- +  h f +  h L = -----b- (18.108)
 2g 

Rearranging and solving for the pressure head at the pump inlet gives

* * 2
p p V 2 
-----2- = -----b- – h s –  --------
- +  h f +  h L (18.109)
   2g 

The net positive suction head available is the amount of head available above the vapor pressure head of
the fluid flowing (in absolute units) or

* *
p p
NPSHA = -----2- – -----v- (18.110)
 

Thus an expression for the NPSHA for a particular pumping system results from combining
Equation 18.109 and Equation 18.110 as

* 2 *
p V 2  p
NPSHA = -----b- – h s –  --------
- +  h f +  h L – -----v- (18.111)
  2g  

Typical values of barometric pressure head and vapor pressure head of water (both in meters of water
absolute) at 15°C are

*
p
-----b- = 10.34 m


and

*
p
-----v- = 0.17 m


18.20.3 Preventing cavitation in a pump

The following relationship must hold for the pumping system to ensure that there is no cavitation in the
pump during normal operating conditions

NPSHA > NPSHR (18.112)

452
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

In other words the net positive suction head available (NPSHA) at the inlet of the pump (as computed
for the suction side pipework from Equation 18.111) must be greater than the minimum allowable or net
positive suction head required (NPSHR) as specified by the pump manufacturer.

The maximum allowable suction lift hs may be found from substituting Equation 18.111 into
Equation 18.112

* 2 *
p V 2  p
-----b- – h s –  --------
- +  h f +  h L – -----v-  NPSHR (18.113)
  2g  

Rearranging in terms of the static suction lift hs gives

* 2 *
p V 2  p
-----b- –  --------
- +  h f +  h L – -----v-  NPSHR + h s (18.114)
  2g  

Thus the static suction lift hs must be less than the quantity of the right – hand side of the following
equation

* 2 *
p V 2  p
h s  -----b- –  --------
- +  h f +  h L – -----v- – NPSHR (18.115)
  2g  

If in computing the suction lift hs it turns out to be negative (recall that it is defined as positive when
pump is located above the level of the suction reservoir), then the pump must be set below the water sur-
face of the suction reservoir
.
Example 18.8 Consider a design problem where it is required to determine the setting of a pump to prevent cavita-
tion. Water at a temperature of 20°C is to be pumped. The minimum expected barometric pressure (in absolute units
specified as a height of water) during the life time of the pump is expected to be

*
p
-----b- = 9.7 m


The vapor pressure of water for the warmest expected operating conditions during the life time of the pump is
*
p
-----v- = 0.24 m


The pipe friction losses and minor losses for a flow of 500 L/s in the inlet side suction pipework for the pump are

h f = 1.0 m

 hL = 0.1 m

From the pump manufacturer, the net positive suction head required for the pump is

NPSHR = 6.0 m

The velocity head of water entering the inlet of the pump for a velocity of 2.5 m/s.

Answer – Computation of the setting of a pump to prevent cavitation.

The velocity head at the inlet to the pump is

2 2
V2  2.5 
--------
- = ------------------ = 0.32 m
2g 2  9.81 

Now the minimum allowable static suction head from Equation 18.115 is

453
 Chapter 18 - File: Chap18-Pumps.fm July 6, 2022 Version 6

* 2 *
p V 2  p
h s  -----b- –  --------
- + h + h  – -----v- – NPSHR
  2 g  f  L 

or

h s  9.7 –  0.32 + 1.0 + 0.1  – 0.24 – 6.0 or h s  2.04 m

Because the value of the static suction lift is positive it means that the reservoir water surface elevation may be up to
but no more than 2.04 m below the level of the centerline of the intake to the pump.

Example 18.9 Consider a design problem in Example 18.8 except for different pipe friction and minor losses.
Determine the setting of a pump to prevent cavitation. The pipe friction losses and minor losses for a flow of 500 L/
s in the inlet side suction pipework for the pump are

h f = 5.0 m

 hL = 0.5 m

Answer – Computation of the setting of a pump to prevent cavitation.

Now the minimum allowable static suction head from Equation 18.115 is

* 2 *
p V 2  p
h s  -----b- –  --------
- + h + h  – -----v- – NPSHR
  2 g  f  L 

That is

h s  9.7 –  0.32 + 5.0 + 0.5  – 0.24 – 6.0

or

h s  – 2.36 m

Because the value of static suction lift is negative it means that the suction side reservoir water surface elevation must
at least be 2.36 m above the level of the centerline of the intake to the pump.

___________________________________________________________________________________

18.21 References
ASCE (1975). Pipeline design for hydrocarbon gases and liquids. Report of the Task Committee on
Engineering Practice in the Design of Pipelines, Committee on Pipeline Planning, Pipeline Division,
American Society of Civil Engineers, New York, 1-79.

454
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

CHAPTER 19 VALVES1

Three types of valves are considered in this Chapter including isolating valves (or control valves), regu-
lating valves and control valves. Isolating valves are extremely important in water distribution systems,
especially, when repairs to leaks need to be carried out. In terms of modeling water distribution systems,
isolating valves are not usually important. They may be taken into account as a localized or minor loss
via a head loss coefficient K. Regulating valves include pressure reducing valves (PRVs), pressure sus-
taining valves (PSVs) and flow control valves (FCVs). These are extremely important in hydraulic mod-
els and should be modeled to accurately reflect flow and pressure conditions in the water distribution
system.

This Chapter commences with a discussion of isolating valves. Regulating valves are then considered.
In the section on pressure reducing valves (PRVs) the various operating modes are described in detail.
Control valves that allow throttling of the flow are covered in the last section of the Chapter.

19.1 Isolating Valves

Isolating valves are important in water distribution systems and at pumping stations and storages. The
purpose of an isolating valve is to close off flow through a pipe (Stephenson 1984). Isolating valves
should only be operated as fully opened or fully closed.

Examples of isolating valves include

• gate valves. Often a gate valve is rectangular in shape. In the fully opened position the gate
is completely withdrawn from the flow. The inside of the valve is almost continuous with
the pipework upstream and downstream of the valve except for the recesses where the gate
seals shut. As a result, the losses through the valve in the fully opened position are small.
Gate valves are operated by a spindle with a screw thread.
• butterfly valves. These may be operated by a 90° turn handle. The leaf of the valve remains
in the flow in the fully opened valve position and thus the losses are higher than for gate
valves.
• rotary or spherical valves
Generally, isolating valves are designed to pass full flow with minimal head loss when in the fully
opened position. If an isolating valve is operated in a partly closed position it will have poor operating
characteristics. As the valve is moved from a fully opened position toward the fully closed position the
pressure reduction (or flow control) that occurs is small. A reasonable level of pressure control only
occurs when the isolating valve is almost closed. Due to the small opening when the valve is almost
closed, the velocities through the valve may be extremely high. As a result, cavitation and consequent
valve damage may occur. Cavitation occurs in association with high velocity flows. Fast rotating vorti-
ces in the flow may result in pressures within the vortex that drop to the vapor pressure of the fluid caus-
ing vapor bubbles to form. These bubbles are swept downstream into regions of higher pressure where
the bubbles implode. The implosion pressures may be on the order of hundreds of thousands of kPa.
Many implosions near a metal or concrete surface may erode the surface and cause significant damage.

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, The University of Adelaide, Ade-
laide, Australia. Part of a book entitled Water Distribution Systems Engineering.

455
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

A further potential problem with isolating valves is the possible occurrence of water hammer. Generally
the shutting of a gate valve is designed to close as rapidly as possible. As mentioned above, the flow
reduction effect does not actually occur until the valve is almost closed. In addition, the relationship
between the open bore area and degree of turning of the closure wheel is generally nonlinear. A conse-
quence of the significant flow reduction that occurs during the last portion of the closure is that severe
water hammer pressures may occur. The magnitude of the water hammer clearly depends on the length
of pipes attached to the valve. The warning here is that the total period of the valve closure may be mis-
leading. Consider a valve closure from fully opened to fully closed of say 3 minutes. The majority of the
flow reduction may actually only occur in the last 5 to 10% of the closure. As a result, the effective clo-
sure time that will produce water hammer may be 9 to 18 seconds. In some systems, this is fast enough
to result in the “fast” closure of the valve when Joukowsky water hammer pressure rise. The Joukowsky
pressure rise is given by

H = – --a- V (19.1)
g

where

• H = the pressure rise (m or ft)

• a = wave speed for the pipeline (the velocity at which a water hammer wave propagates
through the fluid in the pipeline) (m/s or ft/sec)_
• g = gravitational acceleration (m/s2 or ft/sec2)
• V = V f – V i = velocity change in the pipe due to the closure of the valve (assuming that a
reflection water hammer wave does not return to the valve before the valve is fully shut)
(m/s or ft/sec)
• V f = final velocity after the valve has stopped moving (m/s or ft/sec)
• V i = initial velocity before the valve begins to close (m/s or ft/sec)

Example 19.1 Consider an isolating valve where the closure causes a “fast” valve closure. Compute the maximum
head rise in the pipe upstream of the valve as it shuts completely. Assume the initial velocity in the pipe and through
the valve is 1.5 m/s and that the wave speed for the ductile iron cement mortar lined (DICL) pipe is 1150 m/s.

Answer – Computation of the maximum head rise in the pipe upstream of the valve.

The change in velocity of pipe flow due to the closure of the valve is given by

V = V f – V i = 0 – 1.5 = – 1.5 m/s

The Joukowsky pressure rise is determined from Equation 19.1 as

a 1150
H = – --- V = – ------------  – 1.5  = 176 m
g 9.81

___________________________________________________________________________________

19.2 Pressure Reducing Valves (PRV)

There are a number of different types of pressure regulating valves including

• pressure reducing valves (PRVs)

• pressure sustaining valves (PSVs)
• flow control valves (FCVs)

Each of these valves is considered in turn in the subsequent sections of this Chapter.
Pressure reducing valves are placed in pipelines to perform the following functions

• to keep the pressure at the downstream side of the PRV at a constant value whenever the
upstream pressure exceeds the preset value for the PRV (pressure setting)

456
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

• to avoid reverse flow when the pressure on the downstream side of the valve exceeds the
pressure on the upstream side of the valve
The PRV is a valuable and very useful device. They are cheaper than a Break Pressure Tank (BPT). A
PRV is represented within a computer hydraulic model of a water distribution system like a pipe, with an
upstream node and a downstream node.

19.2.1 Operating modes of pressure reducing valves (PRVs)

A pressure reducing valve is a hydromechanical device that acts as a control valve used in a water distri-
bution system. The PRV is usually operated by pressures immediately upstream and downstream of the
valve. The PRV is designed to maintain a constant pressure (i.e. a pressure setting or preset pressure) on
the downstream side of the valve irrespective of how large the upstream pressure is (Jeppson 1976).
PRVs do not have a defined head loss – flow relationship. There are three modes of operation for a PRV

• standard mode (or active position of PRV). Flow occurs from the upstream side of the
valve to the downstream side through the PRV. The pressure on the downstream side of the
valve is equal to the pressure setting for the PRV while the pressure on the upstream side of
the valve is greater than the PRV setting pressure. As the pressure increases or decreases on
the upstream side of the PRV the downstream pressure is held constant. The PRV valve
opens or closes to maintain the constant pressure on the downstream side of the valve. The
position of the valve element is adjusted until the head losses within the valve produce the
required outlet pressure. As a result, the head loss through the PRV is essentially variable.
The entire water distribution system section downstream of the PRV is protected from high
pressures.
• open mode (or inactive position). The pressure on the upstream side of the PRV is less than
the pressure setting for the PRV and as a result the PRV opens fully with the flow being
unrestricted. The pressure on the downstream side is equal to the pressure on the upstream
side (and both are less than PRV pressure setting). The PRV acts as a short pipe with a
larger diameter than the pipe and water flows unrestricted through the valve. Unlike a fully
opened check valve, a fully opened PRV in the inactive position has a low friction factor. A
small local head loss occurs.
• check valve mode (or closed position preventing back flow). The pressure on the down-
stream side of the PRV is greater than the pressure on the upstream side and as a result the
PRV shuts fully and acts as a check valve preventing reverse flow through the PRV. For this
occurrence the PRV could be eliminated from a hydraulic computer model of the water dis-
tribution system.
• a second situation may also lead to the PRV acting as a check valve. If the pressure on the
downstream side of the PRV increases to be above the set pressure then the PRV closes and
acts as a check valve. The pressure on the upstream side of the PRV may be greater than the
pressure on the downstream side of the valve. This situation is not possible if the water dis-
tribution system downstream of the PRV is completely isolated from the network upstream
of the PRV.
19.2.2 Elements of pressure reducing valves
There are a number of elements including

• a main valve
• a relay system
• fittings
The main valve is based on a simple globe valve, modified with a mobile central valve element, a relay
system allowing the specification of a set downstream pressure and fittings connecting the inlet, the out-
let and the chamber containing the mobile valve element.

457
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

19.2.3 Modeling of PRVs in computer hydraulic analysis software

The analysis of a water distribution system containing one or more PRVs must be capable of determin-
ing which of the three possible valve operating conditions (active, inactive or check valve mode) are
occurring at each of the PRVs in the system. PRVs are by far the most complex element to be included in
a general hydraulic computer analysis program due to different conditions of operations of PRVs. Com-
puter simulation of PRVs is difficult as their mode of operation cannot be predicted. The suggested
approach for the analysis to start with the PRV completely open. The mode of operation of a PRV is only
known after all the pressures at the nodes are determined. Usually an operational mode for each PRV is
assumed (i.e. guessed) the hydraulic model is solved, and then each PRV is checked based on the calcu-
lated nodal pressures as to whether the initial assumption for the mode of operation was correct.
Changes are made to the initially assumed PRV operating modes. For example, an active mode may
have been assumed for the PRV and then the hydraulic model may determine the PRV actually has a
check valve mode of operation. The hydraulic analysis must be recalculated as many times as necessary
until the assumed operating modes for each of the PRVs matches the computed nodal pressures
upstream and downstream of each PRV in the water distribution system. The presence of PRVs tend to
increase the computer computational time by changing a symmetric to a non – symmetric matrix and
sometimes resulting in convergence difficulties.

A PRV in active mode may be modeled by eliminating the connection between the upstream and down-
stream sides of the valve (results in a zero coefficient in the matrix). The downstream side of the valve is
considered a supply point or reservoir with known pressure. A pseudo – reservoir or dummy reservoir
with a fixed head or HGL replaces the PRV and is assumed to supply water to the water distribution sys-
tem downstream of the PRV. The PRV is disconnected from its upstream junction and the PRV is
replaced by a reservoir. After the withdrawal rate from the PRV (or pseudo – reservoir) is computed, the
outflow is applied as a demand on the upper system at the node located at the upstream side of the PRV.
The outflow from the pseudo – reservoir must be equal to flow out of the system on the upstream side of
the PRV. Flow continuity is imposed over the two halves of the water distribution system where the PRV
originally was located. Thus a demand is placed at the node just upstream of the PRV. In summary, the
water distribution system is disconnected at the PRV and replaced by a pseudo – reservoir and a demand.
The active operation of PRVs result in subnetworks that may be analyzed separately starting with the
lower elevation subnetworks. The solution from the lower network is used to determine demands at var-
ious junctions of the higher network. In a loop equation formulation of the governing equation for flow
in a water distribution system, an extra loop is needed to be added to the formulation for every PRV in
the system. Every PRV that is replaced by a reservoir results in an extra path or pseudo – loop equation.
The head loss in the path is equal to the difference between the “dummy” reservoir head and the head of
a real reservoir in the water distribution system.

The computer simulation of a PRV in inactive mode treats the model of the PRV as transparent (i.e. the
valve has a large diameter and its length is very short). Alternatively inputting a large pressure setting
for the PRV has the same effect. Water flows unrestricted through the valve.

For a PRV acting as a check valve the simulation model is allowed to find its own pressure level at the
downstream side of the PRV where the cutoff of flow has occurred (zero flow occurs in the pipes con-
nected to the PRV for a check valve operation mode). If the PRV acts as a check valve (having been
assumed to be in an active mode), the flow is calculated as negative through the valve.

Use of the loop method formulation in terms of unknown flows has difficulty in handling the operation
of PRVs. Heads must be known to be able to accurately assess whether a PRV is operating in the
assumed mode of operation.

458
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

Example 19.2 Consider a pressure reducing valve in the middle of a pipe line (600 m long, 500 mm in diameter, and
an internal roughness height of  = 0.25 mm ) between two reservoirs as shown in Figure 19.1 — the upstream res-
ervoir at 60.0 m and the downstream reservoir at 30.0 m. Set up the EPANET input file and run 4 cases. Assume the
–6
water is at 15°C — thus the density is 999.1 kg/m3 and the kinematic viscosity is 1.140 10 m2/s.

(i) Simulate the pipeline system without a PRV in the center of the pipe.
(ii) Simulate the PRV installed in the center of the pipe with a set pressure of 60 m.
(iii) Simulate the PRV with a set pressure of 40 m.
(iv) Simulate the PRV with a set pressure of 20 m.

Figure 19.1 Two pipe – PRV schematic and EPANET node and pipe numbers

Answer – (i) Simulation of the pipeline system when the PRV is removed.

First, the flow assuming that the PRV is removed from the system is calculated by the explicit (or non – computer meth-
ods) outlined in Chapter 4. Under this condition the HGL at the center of the 600 m pipeline would be 45 m (half way
between the upstream and downstream water levels).

Assume the flow is in the transitional flow region within the Moody diagram (Figure 4.2).

The flow for a pipe with flow in the transitional turbulent region from Equation 6.17 is

   
gh f D    5  –6 
5 –3
1.7748  9.81  30   0.5   0.25 10 1.7748  1.141 10 
Q = – 0.9646 ----------------- ln ------------ + --------------------- = – 0.9646 -------------------------------------ln ------------------------ + --------------------------------------------------
L  3.7 D 3 600  3.7  0.5  3 
 gh f D   9.81  30  0.5  - 
------------------------------------
 -----------------  600 
L

–3 –6
where g = 9.81 m2/s, h f = 30 m, D = 0.5 m, = 0.25 mm = 0.25 10 m and  = 1.141 10 m2/s.

Solving for the flow gives

3
Q = 1.057 m /s

The area of the pipe is

2 2
D   0.5  2
A = ---------- = ------------------ = 0.1963 m
4 4

The velocity of flow is

V = Q 1.057- = 5.385 m/s

---- = ---------------
A 0.1963

The Reynolds number is

VD 5.385  0.5  6
Re = -------- = ----------------------------- = 2.36 10
 1.141  10 
–6

The Darcy – Weisbach friction factor from the Swamee and Jain equation (Equation 4.61) is

0.25 0.25
f = ---------------------------------------------------------- = ---------------------------------------------------------------------------------------- = 0.0170
 - + -----------
5.74- 2  0.25  2
log 10  ----------- 5.74
log 10  --------------------- + -------------------------------- -
 3.7 D 0.9
3.7  500  0.9
Re   2.36 10  
6

459
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

A check on the head loss is

2 2
--------- --------------------
L- V = 5.385  = 30.15 m
h f = f --- ------  0.0170  600
D2g 0.5 2  9.81 

This is OK and the difference from 30 m may be attributed to using the Swamee and Jain formula (Equation 4.61)
instead of the Colebrook-White formula.

The shear stress at the boundary is

2 2
V  5.385  2
 o = f ------ =  0.0170   999.1  -------------------- = 61.6 N/m
8 8
The shear velocity is

o 61.6- = 0.248 m/s

u* = ----- = ------------
 999.1

The characteristic length for the viscous sublayer

–6
 - = 1.141  10  –6 –3
--------- = ---- ----------------------------- = 4.60 10 m = 4.60 10 mm
o u* 0.248
-----


The length ratio for the viscous sublayer is

 0.25
------------ = ------------------------ = 54.3
  u* 4.60 10
–3

The flow is in the transitional turbulent zone because


5  -----------
-  70
  u*

Thus the assumption that was made earlier was valid. Now a simulation of EPANET with the PRV removed follows.

The EPANET input file for the case where the PRV has been removed is given below.

[TITLE]
PRV Example - Example 19.2 (i)
PRV removed temporarily to determine maximum
possible flow

[JUNCTIONS]
;--------------------
; Elev Demand
; ID m L/s
;--------------------
2 0 0

[TANKS]
;----------
; Elev
; ID m
;----------
1 60
4 30

[PIPES]
;--------------------------------------------------
; Start End Length Diam Roughness
;ID Node Node m mm Height mm
; (epsilon)
;--------------------------------------------------
1 1 2 300 500 0.25
2 2 4 300 500 0.25
[REPORT]
FILE Ch19PRV1.out

[OPTIONS]

460
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

;----------------------------------------------
MAP Ch19PRV.map ; Map coordinates file
UNITS SI
HEADLOSS D-W
VISCOSITY 1.141E-6
ACCURACY 1E-06

[END]

The EPANET co-ordinates file for the case where the PRV has been removed is

[COORDINATES]
;Nodal coordinates for Chapter 19 Example 19.2 - PRV example
1 0 0
2 300 0
4 600 0

[LABELS]
0 20 "Upstream reservoir" 1
500 20 "Downstream reservoir" 4

The EPANET output file for the case where the PRV has been removed is

Page 1

******************************************************************
* E P A N E T *
* Hydraulic and Water Quality *
* Analysis for Pipe Networks *
* Version 2.0 *
******************************************************************

PRV Example - Example 19.2 (i)

PRV removed temporarily to determine maximum
possible flow

Input Data File ................... CH19PRV1.INP

Output Report File ................ Ch19PRV1.out
Verification File .................
Hydraulics File ...................
Map File .......................... Ch19PRV.map
Number of Pipes ................... 2
Number of Nodes ................... 3
Number of Tanks ................... 2
Number of Pumps ................... 0
Number of Valves .................. 0
Headloss Formula .................. Darcy-Weisbach
Hydraulic Timestep ................ 1.00 hrs
Hydraulic Accuracy ................ 0.000010
Maximum Trials .................... 40
Quality Analysis .................. None
Specific Gravity .................. 1.00
Kinematic Viscosity ............... 1.14e-06 sq m/sec
Chemical Diffusivity .............. 1.21e-09 sq m/sec
Total Duration .................... 0.00 hrs
Reporting Criteria:
All Nodes
All Links

Node Results:
-------------------------------------------------------------------
Elev. Demand Grade Pressure
Node m L/s m m
-------------------------------------------------------------------
2 0.00 0.00 45.00 45.00
1 60.00 -1054.78 60.00 0.00 Reservoir
4 30.00 1054.78 30.00 0.00 Reservoir

461
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

Page 2 PRV Example - Example 19.2

Link Results:
----------------------------------------------------------------
Start End Diameter Flow Velocity Headloss
Link Node Node mm L/s m/s /1000m
----------------------------------------------------------------
1 1 2 500.00 1054.78 5.38 50.00
2 2 4 500.00 1054.78 5.38 50.00

The HGL and flow for the situation above with the PRV removed are shown in Figure 19.2.

Figure 19.2 One pipe system with no PRV

___________________________________________________________________________________
Answer - (ii) Simulation of the pipeline when the set pressure for the PRV is 60 m.

This is the level of the upstream reservoir. As it is impossible to achieve this pressure, the PRV opens up completely
and the results are identical to the case above where there is no PRV present.

The EPANET input file for a PRV setting of 60 m is given below.

[TITLE]
PRV Example - Example 19.2 (ii)
PRV set point pressure set at HGL 60.0 m

[JUNCTIONS]
;--------------------
; Elev Demand
; ID m L/s
;--------------------
2 0 0
3 0 0

[TANKS]
;----------
; Elev
; ID m
;----------
1 60
4 30

[PIPES]

;------------------------------------------------
; Start End Length Diam Rough.
;ID Node Node m mm Height
; (epsilon)
;------------------------------------------------
1 1 2 300 500 0.25
2 3 4 300 500 0.25

[VALVES]
;id node1 node2 diam type setting (losscoeff)

462
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

3 2 3 500 PRV 60 0.0

[REPORT]
FILE CH19PRV2.out

[OPTIONS]
;----------------------------------------------
MAP Ch19PRV2.map ; Map coordinates file
UNITS SI
HEADLOSS D-W
VISCOSITY 1.141E-06

[END]

The EPANET co-ordinates file for a PRV setting of 60 m (also the same file is used for the other PRV settings below)
is

[COORDINATES]
;Nodal coordinates for Chapter 19 Example 19.2(ii) - PRV example
1 0 0
2 280 0
3 320 0
4 600 0

[LABELS]
0 20 "Upstream reservoir" 1
600 20 "Downstream reservoir" 4

The EPANET output file for a PRV setting of 60 m is

Page 1

******************************************************************
* E P A N E T *
* Hydraulic and Water Quality *
* Analysis for Pipe Networks *
* Version 2.0 *
******************************************************************

PRV Example - Example 19.2 (ii)

PRV set point pressure set at HGL 60.0 m

Input Data File ................... CH19PRV2.INP

Output Report File ................ CH19PRV2.out
Verification File .................
Hydraulics File ...................
Map File .......................... Ch19PRV2.map
Number of Pipes ................... 2
Number of Nodes ................... 4
Number of Tanks ................... 2
Number of Pumps ................... 0
Number of Valves .................. 1
Headloss Formula .................. Darcy-Weisbach
Hydraulic Timestep ................ 1.00 hrs
Hydraulic Accuracy ................ 0.001000
Maximum Trials .................... 40
Quality Analysis .................. None
Specific Gravity .................. 1.00
Kinematic Viscosity ............... 1.14e-06 sq m/sec
Chemical Diffusivity .............. 1.21e-09 sq m/sec
Total Duration .................... 0.00 hrs
Reporting Criteria:
All Nodes
All Links

463
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

Node Results:
-------------------------------------------------------------------
Elev. Demand Grade Pressure
Node m L/s m m
-------------------------------------------------------------------
2 0.00 0.00 45.00 45.00
3 0.00 0.00 45.00 45.00
1 60.00 -1054.78 60.00 0.00 Reservoir
4 30.00 1054.78 30.00 0.00 Reservoir

Page 2 PRV Example - Example 19.2 (ii)

Link Results:
----------------------------------------------------------------
Start End Diameter Flow Velocity Headloss
Link Node Node mm L/s m/s /1000m
----------------------------------------------------------------
1 1 2 500.00 1054.78 5.38 50.00
2 3 4 500.00 1054.78 5.38 50.00
3 2 3 500.00 1054.78 5.38 0.00 PRV

The HGL and flow for the situation above with a PRV setting of 60 m are shown in Figure 19.3.

Figure 19.3 Flow for PRV setting of 60 m

___________________________________________________________________________________
Answer – (iii) Simulation of the pipeline system when the set pressure for the PRV is 40 m.

As a result, the PRV closes down until the pressure on the downstream side of the valve is 40 m. The flow between the
reservoirs is thus determined by the difference in the pressure on the downstream side of the PRV and the elevation of
the downstream reservoir (EL 30.0 m).

The EPANET input file for a PRV setting of 40 m is given below.

[TITLE]
PRV Example - Example 19.2 (iii)
PRV set point pressure set at HGL 40.0 m
[JUNCTIONS]
;--------------------
; Elev Demand
; ID m L/s
;--------------------
2 0 0
3 0 0

[TANKS]
;----------
; Elev
; ID m
;----------
1 60
4 30

464
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

[PIPES]
;------------------------------------------------
; Start End Length Diam Rough.
;ID Node Node m mm Height
; (epsilon)
;------------------------------------------------
1 1 2 300 500 0.25
2 3 4 300 500 0.25

[VALVES]
;id node1 node2 diam type setting (losscoeff)
3 2 3 500 PRV 40 0.0

[REPORT]
FILE CH19PRV2.out

[OPTIONS]
;----------------------------------------------
MAP Ch19PRV2.map ; Map coordinates file
UNITS SI
HEADLOSS D-W
VISCOSITY 1.141E-06

[END]

The EPANET output file for a PRV setting of 40 m is

Page 1

******************************************************************
* E P A N E T *
* Hydraulic and Water Quality *
* Analysis for Pipe Networks *
* Version 2.0 *
******************************************************************

PRV Example - Example 19.2 (iii)

PRV set point pressure set at HGL 40.0 m

Input Data File ................... CH19PRV3.INP

Output Report File ................ CH19PRV3.out
Verification File .................
Hydraulics File ...................
Map File .......................... Ch19PRV2.map
Number of Pipes ................... 2
Number of Nodes ................... 4
Number of Tanks ................... 2
Number of Pumps ................... 0
Number of Valves .................. 1
Headloss Formula .................. Darcy-Weisbach
Hydraulic Timestep ................ 1.00 hrs
Hydraulic Accuracy ................ 0.001000
Maximum Trials .................... 40
Quality Analysis .................. None
Specific Gravity .................. 1.00
Kinematic Viscosity ............... 1.14e-06 sq m/sec
Chemical Diffusivity .............. 1.21e-09 sq m/sec
Total Duration .................... 0.00 hrs
Reporting Criteria:
All Nodes
All Links

Node Results:
-------------------------------------------------------------------
Elev. Demand Grade Pressure
Node m L/s m m
-------------------------------------------------------------------
2 0.00 0.00 50.00 50.00

465
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

3 0.00 0.00 40.00 40.00

1 60.00 -859.81 60.00 0.00 Reservoir
4 30.00 859.80 30.00 0.00 Reservoir
Page 2 PRV Example - Example 19.2 (ii)

Link Results:
----------------------------------------------------------------
Start End Diameter Flow Velocity Headloss
Link Node Node mm L/s m/s /1000m
----------------------------------------------------------------
1 1 2 500.00 859.81 4.38 33.33
2 3 4 500.00 859.80 4.38 33.33
3 2 3 500.00 859.81 4.38 10.00 PRV

The HGL and flow for a PRV setting of 40 m are shown in Figure 19.4.

Figure 19.4 Flow in a one pipe system containing a PRV with a pressure head setting of 40 m

___________________________________________________________________________________
Answer – (iv) Simulation of the pipeline system when the set pressure for the PRV is 20 m.

This causes the PRV to close completely. There is no flow between the reservoirs.

The EPANET input file for a PRV setting of 20 m is given below.

[TITLE]
PRV Example - Example 19.2 (iv)
PRV set point pressure set at HGL 20.0 m

[JUNCTIONS]
;--------------------
; Elev Demand
; ID m L/s
;--------------------
2 0 0
3 0 0

[TANKS]
;----------
; Elev
; ID m
;----------
1 60
4 30

[PIPES]
;------------------------------------------------
; Start End Length Diam Rough.
;ID Node Node m mm Height
; (epsilon)
;------------------------------------------------
1 1 2 300 500 0.25
2 3 4 300 500 0.25

466
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

[VALVES]
;id node1 node2 diam type setting (losscoeff)
3 2 3 500 PRV 20 0.0

[REPORT]
FILE CH19PRV4.out

[OPTIONS]
;----------------------------------------------
MAP Ch19PRV2.map ; Map coordinates file
UNITS SI
HEADLOSS D-W
VISCOSITY 1.141E-06

[END]

The EPANET output file for a PRV setting of 20 m is

Page 1

******************************************************************
* E P A N E T *
* Hydraulic and Water Quality *
* Analysis for Pipe Networks *
* Version 2.0 *
******************************************************************

PRV Example - Example 19.2 (iv)

PRV set point pressure set at HGL 20.0 m

Input Data File ................... CH19PRV4.INP

Output Report File ................ CH19PRV4.out
Verification File .................
Hydraulics File ...................
Map File .......................... Ch19PRV2.map
Number of Pipes ................... 2
Number of Nodes ................... 4
Number of Tanks ................... 2
Number of Pumps ................... 0
Number of Valves .................. 1
Headloss Formula .................. Darcy-Weisbach
Hydraulic Timestep ................ 1.00 hrs
Hydraulic Accuracy ................ 0.001000
Maximum Trials .................... 40
Quality Analysis .................. None
Specific Gravity .................. 1.00
Kinematic Viscosity ............... 1.14e-06 sq m/sec
Chemical Diffusivity .............. 1.21e-09 sq m/sec
Total Duration .................... 0.00 hrs
Reporting Criteria:
All Nodes
All Links

Node Results:
-------------------------------------------------------------------
Elev. Demand Grade Pressure
Node m L/s m m
-------------------------------------------------------------------
2 0.00 0.00 60.00 60.00
3 0.00 0.00 30.00 30.00
1 60.00 -0.00 60.00 0.00 Reservoir
4 30.00 0.00 30.00 0.00 Reservoir

Page 2 PRV Example - Example 19.2 (c)

Link Results:
----------------------------------------------------------------

467
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

Start End Diameter Flow Velocity Headloss

Link Node Node mm L/s m/s /1000m
----------------------------------------------------------------
1 1 2 500.00 0.00 0.00 0.00
2 3 4 500.00 0.00 0.00 0.00
3 2 3 500.00 0.00 Closed PRV

The HGL and flow for the situation of a PRV setting of 20 m are shown in Figure 19.5.

Figure 19.5 Flow for PRV setting of 20 m

___________________________________________________________________________________

PRVs protect sections of water distribution system being damaged by excessively high pressures. Trunk
mains often have a considerably higher pressure than distribution mains. PRVs are often used to reduce
the pressure for water flowing in distribution mains or in secondary or tertiary mains. Installation of
PRVs may be costly although in some instances there may be no alternative method for reducing the
pressure.

PRVs placed in a system cause the pressure in the portion of the water distribution system downstream
of the PRV to be independent of the remainder of the water distribution system upstream of the PRV.

19.2.4 Advantages of the use of PRVs

Some benefits of using PRVs include

• the pipe class (i.e. pipe wall thickness) may be reduced for pipes downstream of the PRV
thereby leading to a cost saving
• wear to plumbing and garden fixtures may be reduced by reducing the working pressure by
using a PRV
One possible disadvantage of a PRV is the tendency of blockages occurring especially for usage in
untreated water supply systems.

19.2.5 EPANET approach to modeling of PRVs

Initially all PRVs are assumed to be inactive (or fully open) as if they are not present in the water distri-
bution system. Once a hydraulic balance is completed the validity of the assumption of all PRVs being
open is checked and appropriate adjustments to status of all PRVs is made. Some example water distri-
bution systems that have PRVs present may take approximately 13 iterations to solve with the Salgado-
Castro hybrid iterative solution algorithm with EPANET.

468
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

19.3 Flow Control Valves (FCV)

A flow control valve regulates the flow in a pipeline to a set flow level. Usually the flow is measured
downstream of the valve and the measured value is compared to the set point. If the flow exceeds the set
point value the FCV closes to cause more head loss through the valve and therefore reduce the flow. If
the flow is below the set point the valve opens up until either the flow reaches the set point or the valve
is fully opened.

Example 19.3 Consider a flow control valve in the middle of a pipeline 600 m long, 500 mm in diameter and with a
roughness height of  = 0.25 mm between two reservoirs — the upstream reservoir at 60.0 m and the downstream
reservoir at 30.0 m (identical to the pipe system in Example 19.2). Set up the EPANET input file and run 3 cases.
–6
The water is at 15°C thus the density is 999.1 kg/m3 and the kinematic viscosity is 1.141 10 m2/s.

(i) Simulate the FCV set point such that the valve is fully opened — determine the maximum possible controllable
flow. Use a set flow for the FCV of 1500 L/s
(ii) Simulate the FCV set point at 600 L/s.

Figure 19.6 One pipe system containing an FCV (EPANET node and pipe numbers are also shown)

Answer – (i) The set flow for the FCV is 1500 L/s.

First, the flow assuming that the FCV is removed from the system is 1055 L/s — the same as for Example 19.2. Under
this condition the HGL at the center of the 600 m pipeline would be 45 m. It is therefore impossible to achieve a flow
of 1500 L/s, thus the FCV opens up completely and the results are identical to the case above where there is no FCV
present.

The EPANET input file for a FCV setting of 1500 L/s is given below.

[TITLE]
FCV Example - Example 19.3 (i)
FCV set point flow set at 1500 L/s

[JUNCTIONS]
;--------------------
; Elev Demand
; ID m L/s
;--------------------
2 0 0
3 0 0

[TANKS]
;----------
; Elev
; ID m
;----------
1 60
4 30

[PIPES]
;--------------------------------------------------

469
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

; Start End Length Diam Rough.

;ID Node Node m mm Height mm
; (epsilon)
;--------------------------------------------------
1 1 2 300 500 0.25
2 3 4 300 500 0.25

[VALVES]
;--------------------------------------------------
;id node1 node2 diam type setting (losscoeff)
; mm L/s K-open
;--------------------------------------------------
3 2 3 500 FCV 1500 0.0

[REPORT]
FILE CH19FCV1.out
STATUS YES

[OPTIONS]
;----------------------------------------------
MAP Ch19FCV2.map ; Map coordinates file
UNITS SI
HEADLOSS D-W
VISCOSITY 1.141E-06

[END]

The EPANET output file for a FCV setting of 1500 L/s is

Page 1

******************************************************************
* E P A N E T *
* Hydraulic and Water Quality *
* Analysis for Pipe Networks *
* Version 2.0 *
******************************************************************

FCV Example - Example 19.3 (i)

FCV set point flow set at 1500 L/s

Input Data File ................... CH19FCV1.INP

Output Report File ................ CH19FCV1.out
Verification File .................
Hydraulics File ...................
Map File .......................... Ch19PRV2.map
Number of Pipes ................... 2
Number of Nodes ................... 4
Number of Tanks ................... 2
Number of Pumps ................... 0
Number of Valves .................. 1
Headloss Formula .................. Darcy-Weisbach
Hydraulic Timestep ................ 1.00 hrs
Hydraulic Accuracy ................ 0.001000
Maximum Trials .................... 40
Quality Analysis .................. None
Specific Gravity .................. 1.00
Kinematic Viscosity ............... 1.14e-06 sq m/sec
Chemical Diffusivity .............. 1.21e-09 sq m/sec
Total Duration .................... 0.00 hrs
Reporting Criteria:
All Nodes
All Links

Node Results:
-------------------------------------------------------------------
Elev. Demand Grade Pressure
Node m L/s m m
-------------------------------------------------------------------

470
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

2 0.00 0.00 45.00 45.00

3 0.00 0.00 45.00 45.00
1 60.00 -1054.77 60.00 0.00 Reservoir
4 30.00 1054.77 30.00 0.00 Reservoir

Page 2 FCV Example - Example 19.3 (i)

Link Results:
----------------------------------------------------------------
Start End Diameter Flow Velocity Headloss
Link Node Node mm L/s m/s /1000m
----------------------------------------------------------------
1 1 2 500.00 1054.77 5.38 50.00
2 3 4 500.00 1054.77 5.38 50.00
3 2 3 500.00 1054.77 5.38 0.00 FCV

Warning error generated during run (in another Window when EPANET is run)

*** WARNING FCV 3 - Cannot deliver flow at 0:00 hrs.

The HGL and flow for a FCV setting of 1500 L/s are shown in Figure 19.7.

The maximum controllable flow by the pressure is 1055 L/s. If the valve setting is above this value, the valve is fully
opened and the condition as shown in Figure 19.7 results.

Figure 19.7 Flow for FCV setting of 1500 L/s

___________________________________________________________________________________
Answer – (ii) The set flow for the FCV is 600 L/s.

This causes the FCV to close and throttle the flow until it is equal to 600 L/s.

The EPANET input file for a FCV setting of 600 L/s is given below.

[TITLE]
FCV Example - Example 19.3 (ii)
FCV set point flow set at 600 L/s

[JUNCTIONS]
;--------------------
; Elev Demand
; ID m L/s
;--------------------
2 0 0
3 0 0

[TANKS]
;----------
; Elev
; ID m
;----------
1 60
4 30

471
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

[PIPES]
;--------------------------------------------------
; Start End Length Diam Rough.
;ID Node Node m mm Height mm
; (epsilon)
;--------------------------------------------------
1 1 2 300 500 0.25
2 3 4 300 500 0.25

[VALVES]
;--------------------------------------------------
;id node1 node2 diam type setting (losscoeff)
; mm L/s K-open
;--------------------------------------------------
3 2 3 500 FCV 600 0.0

[REPORT]
FILE CH19FCV2.out
STATUS YES

[OPTIONS]
;----------------------------------------------
MAP Ch19FCV2.map ; Map coordinates file
UNITS SI
HEADLOSS D-W
VISCOSITY 1.141E-06

[END]

The EPANET output file for a FCV setting of 600 L/s is

Page 1

******************************************************************
* E P A N E T *
* Hydraulic and Water Quality *
* Analysis for Pipe Networks *
* Version 2.0 *
******************************************************************

FCV Example - Example 19.3 (ii)

FCV set point flow set at 600 L/s

Input Data File ................... CH19FCV2.INP

Output Report File ................ CH19FCV2.out
Verification File .................
Hydraulics File ...................
Map File .......................... Ch19FCV2.map
Number of Pipes ................... 2
Number of Nodes ................... 4
Number of Tanks ................... 2
Number of Pumps ................... 0
Number of Valves .................. 1
Headloss Formula .................. Darcy-Weisbach
Hydraulic Timestep ................ 1.00 hrs
Hydraulic Accuracy ................ 0.001000
Maximum Trials .................... 40
Quality Analysis .................. None
Specific Gravity .................. 1.00
Kinematic Viscosity ............... 1.14e-06 sq m/sec
Chemical Diffusivity .............. 1.21e-09 sq m/sec
Total Duration .................... 0.00 hrs
Reporting Criteria:
All Nodes
All Links

472
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

Node Results:
-------------------------------------------------------------------
Elev. Demand Grade Pressure
Node m L/s m m
-------------------------------------------------------------------
2 0.00 0.00 55.09 55.09
3 0.00 0.00 34.91 34.91
1 60.00 -600.00 60.00 0.00 Reservoir
4 30.00 600.00 30.00 0.00 Reservoir

Page 2 FCV Example - Example 19.3 (ii)

Link Results:
----------------------------------------------------------------
Start End Diameter Flow Velocity Headloss
Link Node Node mm L/s m/s /1000m
----------------------------------------------------------------
1 1 2 500.00 600.00 3.06 16.35
2 3 4 500.00 600.00 3.06 16.35
3 2 3 500.00 600.00 3.06 20.19 FCV

The HGL and flow for a FCV setting of 600 L/s are shown in Figure 19.8.

Figure 19.8 Flow for a FCV setting of 600 L/s

___________________________________________________________________________________

19.4 Control Valves

Control valves, in contrast to isolating valves, are designed to operate well at a partial opening. In the
design of control valves, an attempt is made to linearly reduce the open area of the valve. The objective
is to contain the cavitation effect due to the reduction of pressure across the valve (Stephenson 1984).

Examples of control valves include

• needle valves. This type of valve is expensive. A needle valve is often used in water supply
headworks systems to release water from a pipe to a river.
• diaphragm type valves
• slotted radial flow sleeve valves
• plunger type control valves
• piston type valves
19.4.1 Cavitation in control valves
A measure of the potential for cavitation in valves is provided by the cavitation index. The cavitation
index is defined as;

473
 Chapter 19 - File: Chap19-PRVs.fm July 6, 2022 Version 6

* *
pd – pv
 = -----------------------
- (19.2)
* *
pu – pd

where

*
• p d = pressure in absolute units at the downstream side of the valve (N/m2 or psi)
*
• p u = pressure in absolute units at the upstream side of the valve (N/m2 or psi)
*
• p v = vapor pressure of the liquid in absolute units (N/m2 or psi)

* *
In Equation 19.2 the numerator of p d – p v represents how close the downstream pressure is to vapor
pressure. The closer this term is to zero the more potential there is for cavitation to occur. The denomi-
* *
nator of p u – p d represents the difference between the pressure on the upstream side of the valve and
the pressure on the downstream side of the valve. The larger the pressure difference (that in turn reflects
a higher head loss through the valve — often due to the valve being almost totally closed) the more likely
cavitation will occur.

In general terms, the value of sigma determined from Equation 19.2 and the possibility for cavitation are
given in Table 19.1.

Table 19.1 Values of cavitation index

Value of cavitation index  Cavitation potential

large number low
small number high

For streamlines valves, cavitation occurs when

  0.1

An interpretation is that the pressure is reduced by a factor of 10 across the valve.

For poorly designed valves, cavitation occurs when

  1.0

19.5 References
Jeppson, R. W. (1976). Analysis of flow in pipe networks, Ann Arbor Science, Ann Arbor, Michigan.

Stephenson, D. (1984). Pipeflow analysis. Developments in Water Science, No., Elsevier, Amsterdam.

474
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

CHAPTER 20 OPTIMIZATION OF WATER DISTRIBUTION

SYSTEMS1

20.1 Introduction
The cost of purchase, supply and installation of pipes, pumps, tanks and valves in water distribution sys-
tems to provide water to consumers is quite significant, often on the order of millions of dollars (Simp-
son et al. 1995). So far in this book, when designing a water distribution system, the designer has had to
choose the sizes of components or elements including pipe diameters, locations of and sizes and settings
of tanks, pumps and valves in the water distribution system. This is usually achieved by using computer
simulation models to test the performance of a trial water distribution system design to ensure it is satis-
factory. In turn, trial – and – error adjustment of the design is made based on the interpretation of the
results of computer simulation runs for the various water demand loading cases under consideration as
described in Chapter 17. Low pressures at nodes in the water distribution system are one indication that
a trial design is unsatisfactory. The designer may attempt to adjust the design to reduce the cost of the
water distribution system while at the same time achieving satisfactory pressures across the entire net-
work. This trial – and – error process is laborious and time – consuming with the outcome often being a
suboptimal design. What is needed is some way of automating the design process so that many designs
are considered both quickly and efficiently. Consideration of the optimization of the design and opera-
tion of water distribution systems is a logical next step, and is considered in detail in this Chapter.

Considerable research has been carried out in the past developing approaches for optimizing water dis-
tribution system design (Alperovits and Shamir 1977, Quindry et al. 1981, El – Bahwry and Smith 1985,
Goulter and Morgan 1985, Su et al. 1987, Simpson et al. 1994, Dandy et al. 1996). A large literature
exists. Lansey and Mays (1989) provided a comprehensive review of the published literature up to 1988.
Other reviews have been carried out since that time.

The main focus of this Chapter is the application of genetic algorithm (GA) optimization to the design
and operation of piped water distribution systems. A genetic algorithm is a member of a class of search
algorithms based on artificial evolution. GAs are a relatively new optimization technique pioneered by
Holland (1975) at the University of Michigan. Goldberg2 in 1989 published a book that was a turning
point in the development of the genetic algorithm (GA) technique and its application to engineering.
There is scant theoretical basis for genetic algorithms except from the schema theorem (Holland 1975),
however, there have been many studies where GA optimization has been shown to be very successful in
a wide range of optimization problems.

Genetic algorithm search methods have been applied to a number of other applications including

• gas pipeline operation optimization (Goldberg and Kuo 1987)

• structural optimization (Sved et al. 1991)
• control system optimization for aerospace applications (Krishnakumar and Goldberg 1990)
• musical composition (Horner and Goldberg 1991)
• sequencing of water resources projects (Dandy and Connarty 1995)

1. © Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,

Australia. Part of the book entitled Water Distribution Systems Engineering.
2. Goldberg (1989) was a graduate student who studied for a PhD under the supervision of Professor
John Holland in the early 1980s. Goldberg is a Civil Engineer by background and was co – supervised in
his Ph.D. by Professor Ben Wylie of Civil Engineering at the University of Michigan.

475
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Other evolutionary type techniques have included evolutionary strategy (Rechenberg 1973) and simu-
lated annealing (Bochachevsky et al. 1986).

The first application of evolutionary optimization techniques to water distribution system optimization
came from Cembrowicz and Krauter (1977). They used graph theory, linear programming and a search
procedure based on concepts from “biological evolution” (the evolutionary strategy) to optimize water
distribution systems. The evolutionary strategy approach was developed by Rechenberg (1973) in Ger-
many in parallel but independent of the development of genetic algorithms by Holland (1975). Cembro-
wicz and Krauter (1977) considered tree configurations of pipes (a branched network with no loops such
that every node is connected) that corresponded to a local cost minimum. The search procedure based on
biological evolution was composed of five major operators similar in nature to genetic algorithm opera-
tors. The evolutionary strategy technique was used to generate the tree networks while linear program-
ming was used to optimize the generated tree networks.

The research into the application of genetic algorithm optimization to the optimal design of water distri-
bution systems was pioneered in the early 1990s at the University of Adelaide, South Australia. First
results were presented in Murphy and Simpson (1992). The first application of GA optimization opti-
mized the sizing of the diameter of pipes in a water distribution system, as well as deciding on whether
pipes should be either cleaned or duplicated (with a new pipe parallel to the existing pipe — although not
necessarily the same diameter as the existing pipe). Since 1992, a sophisticated approach to the applica-
tion of genetic algorithm optimization has been developed including the ability to size and determine the
optimal location of pipes, pumps, tanks and pressure regulating valves, as well as for determining the
optimal operation of pumping. The GA technique finds low cost solutions (often near the global opti-
mum) in relatively few evaluations compared to the size of the search space. The efficiency of the GA
indicates a robustness of the search method that underlies the GA approach and the flexibility of the for-
mulation itself (Goldberg 1989). GAs are very well suited to complicated water distribution system opti-
mization (Simpson et al. 1995). Millions of dollars of savings of both capital and operating costs have
been achieved by the practical application of GA techniques to the optimal design and operation of
water distribution systems. The GA technique has the potential to be extremely useful in engineering
practice in the optimization of the design and operation of water distribution systems.

A primary objective of the design of water distribution systems is to determine the minimum cost sys-
tem that provides water for all the desired water demand conditions. The sum of the capital cost and the
operating cost (usually the pump power costs — expressed as an annual cost and taken back to a present
value cost) is a commonly selected objective that is minimized in the optimization procedure. Specified
design criteria (such as minimum allowable pressure at supply nodes, maximum allowable pressures and
maximum allowable velocities) usually need to be satisfied by the optimized design. An essential ele-
ment of the GA procedure is the running of a simulation model to determine whether a water distribu-
tion system meets the design constraints. As part of the computation of the cost of each water
distribution system in the genetic algorithm optimization process the unit costs of pipe (for different
diameters) are needed — including purchase and installation costs — as well as costs of all other new
components that could possibly be installed in the water distribution system.

Techniques for the optimization of water distribution systems are not new. In fact, many papers have
appeared on the optimization of water distribution systems in the research literature since the mid –
1960s. The main techniques that have been used in the past include linear programming, nonlinear pro-
gramming, pruned enumeration and dynamic programming. However, despite the availability of a num-
ber of papers, there has been virtually no transfer from the world of academia to professional practice of
the techniques developed in these technical publications.

This Chapter starts out by giving an overview of the genetic algorithm process. This is followed by a
review of the main elements of the common trial – and – error method of design traditionally used for the
design of water distribution systems. A discussion of combinatorially large problems is presented as an
introduction to optimization of water distribution systems. The genetic algorithm (GA) technique for the
optimization of water distribution system design is then described in detail. Some examples of applica-
tions of the GA technique to the optimization of water distribution systems are then given. The role of a
water distribution designer will change as a result of the use of genetic algorithm optimization, in com-

476
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

parison to the use of trial – and – error combined with experience and judgment design methods. A sec-
tion is devoted to discussing these issues. The Chapter ends by comparing the GA technique to other
traditional optimization techniques and exploring some of the reasons why traditional optimization tech-
niques have not been readily adopted by water distribution system design practitioners in the past.

20.2 An Overview of the Genetic Algorithm Process

The GA optimization technique has an analogy to natural genetics. A genetic algorithm operates on a
population of binary or integer strings using structured operators that include randomness that result in
an efficient search procedure (Dandy et al. 1996). GAs have been shown to produce near – global opti-
mum solutions for problems with complex search spaces. The outcome of the intensive computation
analysis on a high speed computer is a range of alternative low cost water distribution system designs
usually of different configurations with similar low costs.

In the case of water distribution system optimization, a population of solutions of water distribution sys-
tem designs is dealt with simultaneously in the GA technique. The individual strings are made up of
numbers or bits (zeroes and ones) that describe the problem to be optimized. These strings may be
thought of as genes within chromosomes when considering the analogy to natural population genetics.
The strings in a GA population represent complete design alternatives of possible combinations of ele-
ments in the water distribution system network that are being selected in the optimization process. The
population of designs is dealt with simultaneously in the GA process. The strings are manipulated by the
genetic algorithm to produce low cost water distribution system network designs. The aim of the genetic
algorithm optimization procedure is to identify a near – optimal combination (or combinations) of
design decision variable values such that cost is minimized and the design constraints are satisfied.

Particular decision variables are represented within the string by a series of substrings. For example, the
first substring may represent a pipe in the water distribution system that has to be sized. The second sub-
string may represent the pressure setting of a valve that needs to be determined, and so on along the
string. Choice or look – up tables for each decision variable represented by each substring need to be
provided. The corresponding cost data and design data must also be provided for each of the elements in
the choice tables. An example of a choice table would be a series of different pipe diameters, costs and
roughness height values that are possible selections for a particular pipe in the system.

The genetic algorithm process commences by randomly generating a population of strings (say 200)
where each string represents a different water distribution system design. These designs vary in their
effectiveness. A population of 200 usually represents only an infinitesimal fraction of the total search
space of all possible combinations. Thus it is very unlikely that even one of the randomly generated
solutions will be close to the global optimal solution. Genetic algorithm operators are applied to the ini-
tial population to produce the next generation of design solutions. To commence the process of applying
the genetic operators, it is necessary to compute a fitness of each of the strings in the initial population.
The cost of implementation of each design is computed. Each design may then be hydraulically evalu-
ated in a computer simulation model to assess whether it meets the specified design criteria. The overall
fitness of a coded string representing a water distribution system network design is determined by both
pipe costs and the hydraulic performance of the network design (Dandy et al. 1996). A hydraulic analy-
sis is required for all new strings during the GA run and this may be computationally intensive.

The beauty of the genetic algorithm optimization procedure is that it only requires the strings (the series
of numbers that represent a physical configuration of the water distribution system being optimized) and
a fitness measure for each string. No other information such as gradients is required. This situation is in
contrast to the requirements of other more traditional optimization techniques including linear program-
ming, nonlinear programming and dynamic programming.

Formulation of the fitness function for a string depends on the problem being optimized. For the optimi-
zation of water distribution systems the fitness often has two parts (Simpson et al. 1994). The first part is
an actual or physical cost (real or estimated) associated with the particular combination of design vari-
ables that the string represents including pipes, pumps, storage tanks and valves. Both capital costs and
operating costs may be included. The second part is a cost that is assigned if the particular combination
of elements that are represented by the string fail to meet all of the specified design criteria for the
design situation. A computer simulation model analysis of the water distribution system design for the

477
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

water demand loading case (or cases) enables the determination of whether the particular design meets
the design criteria (for example, the minimum allowable pressures). The failure to meet design con-
straints may be designated as a penalty cost (or costs). The total cost is then taken to be the sum of the
physical cost plus the penalty cost. The fitness of the string is usually taken as a function of the cost, for
example, often the inverse of the total cost is used. The genetic algorithm is often formulated to mini-
mize the cost of the system or to maximize the fitness. Once all the fitness values have been determined
for each of the strings in the population of water distribution system designs the next step is to apply
genetic operators to create, first, a mating pool and then the next generation of solutions. The most com-
mon genetic operators are selection, crossover and mutation.

The genetic algorithm optimization process is driven by two main features. In the first, a selection pro-
cess occurs (driven by the “survival of the fittest”) to create a mating pool for the next generation. The
new mating pool has a higher average fitness compared to the previous generation. The second feature
involves genetic operators including crossover and mutation. Crossover interchanges pieces of strings
between two randomly chosen strings from the mating pool formed by the selection process. The inter-
mixing of information between strings is critical to the process. Mutation is an operator that enables
diversity to be reinjected into the population by the alteration of one piece of the string from its existing
value to another value. The GA process proceeds by the repeated application of the genetic operators of
selection, crossover and mutation to produce many generations of populations of solutions. The best
solutions must be remembered along the way (for example — the 20 lowest cost solutions are stored cor-
responding to the strings with maximum fitness). It is necessary to remember the best solutions, as it is
possible that the crossover and mutation may eliminate top solutions during the process.

20.3 Simulation of Water Distribution Systems

As mentioned in Chapter 17, the simulation of piped water distribution systems has reached a mature
level of practice. The hydraulic conditions in the water distribution system are computed including the
pressures or HGLs at each of the nodes and flows in each pipe, pump and valve (including pressure
reducing valves, pressure sustaining valves and flow control valves). Water quality modeling (for exam-
ple, chlorine residual and water age) has been added to many of these computer simulation models over
the last five years. There are many models available for sale commercially and from government
sources. Some models, such as EPANET 2.0 (as discussed in Chapter 17) are available for no charge
from the website of the United States Environmental Protection Agency — Dr. Lewis Rossman devel-
oped the software at the Water Supply and Water Resources Division National Risk Management
Research Laboratory, Cincinnati, Ohio). Many of the models have sophisticated graphical interfaces
including the ability to retrieve data from Geographical Information Systems (GIS). The input and out-
put facilities and the methods for checking data consistency have also improved dramatically.

Hydraulic simulation models numerically solve a set of nonlinear equations governing steady – state
flow conditions in the water distribution system as described in Chapter 9. The fastest solver appears to
be EPANET. The source code for EPANET is available and has been incorporated into many commer-
cial solvers. Simulation is a very important part of the genetic algorithm procedure as is discussed in fol-
lowing sections.

20.4 Trial – and – Error Design

The objective in designing a water distribution system (whether it is a new system, or an expansion or
rehabilitation of an existing system) is to develop a water distribution system that delivers the required
flows under all of the water demand loading conditions. The designer aims to minimize the cost of the
water distribution system while at the same time satisfying all the design constraints. Thus it is neces-
sary to have available a database of costs of new components (both the capital cost of purchase and the
installation cost) including pipes, pumps, valves and tanks. For pipes, an assumed internal roughness of
the pipe is also required. If the cost of operation of pumps is being considered then the cost of electricity
(on – peak and off – peak rates) and pump efficiency curves are also required.

The standard method of design of water distribution systems currently used by the majority of engineers
in practice for the design is by a trial – and – error approach guided by engineering experience and judg-
ment. A computer hydraulic simulation model of the proposed water distribution system is developed.
Usually pipe routes and hence lengths, node elevations and water demands at nodes are given, although

478
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

there may be some alternative possible pipeline routes that need to be chosen (such as beneath tall build-
ings). All the locations and sizes of components in the computer model must be selected by the designer.
These components include sizes of pipe diameters and the sizes and locations of pumps, tanks and
valves. Valves also require settings to be specified such as pressure settings for pressure reducing valves
and pressure sustaining valves and the set flows for flow control valves. Tanks require a volume and a
normal operating level range to be specified. The designer usually selects sizes of pipes based on knowl-
edge of water demands in the system as well as engineering judgment. Once a trial water distribution
system has been developed, the computer hydraulic simulation model is run for all of the water demand
load cases of interest. These may include peak hour, peak day over an extended period simulation (EPS)
and fire loading cases. The output results for each of the computer simulation runs that are conducted
may then be analyzed. For example, in terms of specified design criteria including pressures at nodes —
that is, as to whether minimum allowable pressures are satisfied at all water demand supply nodes and
also whether maximum allowable pressures are not exceeded at water demand supply nodes. In addi-
tion, pipe velocities and head loss per km or 1000 ft are sometimes considered. Finally, an extended
period simulation (EPS) run is needed to ensure that tank or reservoir levels do not drop below mini-
mum allowable levels or even worse, run dry on a peak water demand day or a succession of peak water
demand days. The cost of the water distribution system is also computed from the unit cost information
from the database of costs.

An assessment is made of the acceptability of the results for a particular design of a water distribution
system relative to both the adopted design criteria and the objective of minimizing the cost of the design.
Once this evaluation is made, then changes to sizes of elements in the water distribution system are
made (for example, pipe sizes are increased or decreased) to either attempt to meet the design criteria
and/or reduce the cost of the design. Various combinations of pipe sizes, tanks, valves and pumps are
tried by the designer. The computer simulation of the revised water distribution system for all of the
water demand loading cases is repeated. The designer then continues the process of adjusting sizes of
components of elements in the water distribution system, computation of cost, hydraulic simulation for
all water demand loading cases, and assessment as to whether the design criteria are met over and over
in an attempt to improve the design of the water distribution system to achieve a low cost solution.
Clearly only a limited number of water distribution systems may be analyzed (Dandy et al. 1993).
Therefore the designer usually searches for a satisfactory rather than necessarily an optimum solution.
The inherent limitations using a computer simulation modeling approach for design of water distribution
systems is the total dependence on the designer to manually identify and input trial solutions to be tested
by the model. This places tremendous responsibility on the designer who is expected (often with limited
budget and time) to find hydraulically viable and cost effective solutions.

Where pressures at nodes in the water distribution system are lower than the minimum allowable pres-
sure, pipe sizes may be sometimes increased to improve these pressures. However, care must be taken
not in assuming that this is always the case. In looped systems it is sometimes possible to reduce the
sizes of certain pipes and as a consequence, increase the pressure to adequate levels at deficient nodes. A
redistribution of flow and hence pressure may occur when pipes are reduced in size in the looped system
depending on where the low pressure areas are. It is not always the case that reducing the size of pipes
makes the pressures lower.

20.5 Optimization of the Design of Water Distribution Systems

Goulter (1987) suggested the minimum cost design for a given layout and a single water demand load-
ing case is a branched water distribution system (that is, a network without any loops). In practice, loops
are an essential feature of actual water distribution systems as they provide an alternative flow path in
case there is a pipe failure or for maintenance purposes. One may achieve a degree of redundancy in
water distribution system optimization by ensuring that the layout has appropriate loops, and by specify-
ing for the optimization process that at least a minimum diameter pipe must be chosen for every new
pipe in the water distribution system.

479
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

20.5.1 Previous related genetic algorithm applications

In recent years a number of researchers have applied the genetic algorithm technique to certain aspects
of the design of pipeline systems. Goldberg and Kuo (1987) applied genetic algorithm optimization to
optimizing the steady state operation of a gas pipeline. They considered a steady – state serial gas pipe-
line consisting of 10 pipes and 10 compressor stations each containing four pumps in series. The objec-
tive was to minimize power used while supplying a specified flow and maintaining minimum allowable
pressures. A simple three – operator genetic algorithm found near – optimal pump operation alternatives
after only evaluating a small fraction of the total possible number of solutions (3500 from 1.10x1012
possible combinations). Davidson and Goulter (1994) used GAs to optimize the layout of a branched
rectilinear network, such as a rural natural gas or water distribution system. The optimal layout in that
case was assumed to be one of least length. The layout solutions were represented by blocks of binary
code, and new GA operators of recombination and perturbation were introduced to reduce the number of
infeasible solutions created by the traditional GA operators of crossover and mutation. A case of a water
distribution system with one demand loading case and no constraints on minimum pipe diameters was
considered by Walters and Lohbeck (1993). They showed the GA effectively converges to near –
optimal branched network layouts, as selected from a directed base graph that defines a set of possible
layouts.

20.5.2 The New York City water tunnels network

Consider an example of the optimization of the New York City tunnels network as shown in Figure 20.1
that was first studied in 1969 by Schaake and Lai at Massachusetts Institute of Technology (M.I.T.),
Boston, USA. A set of existing tunnels needed to have pipes added in parallel to the existing pipes,
referred to as duplication pipes (but not necessarily the same size), to meet increased water demand con-
ditions while maintaining adequate pressure at delivery nodes. Pipe and node data are given in
Table 20.2 and Table 20.3. US customary units are used to enable comparison with previously published
results. The primary tunnel system consisted of City Tunnels number 1 and number 2. City Tunnel num-
ber 1 extended from Hillview Reservoir to node 16 in Brooklyn by way of Manhattan. City Tunnel num-
ber 2 extended between Hillview reservoir and Richmond downtake by way of Queens. City Tunnel
number 1 was constructed around 1920 and City Tunnel number 2 was constructed around 1940 (de
Neufville et al. 1971).

The New York Tunnels network has been used as a case study by many researchers (Schaake and Lai
1969, Quindry et al. 1981, Gessler 1982, Bhave 1985, Morgan and Goulter 1985, Dandy et al. 1996,
Savic and Walters19 97 and Lippai et al. 1999).

The layout shown in Figure 20.1 indicates a series of tunnels that were in place in the late 1960s at the
time the optimization of the water distribution system was first considered. Future water demands had
been estimated for the water distribution system and minimum allowable HGL levels were specified at
all supply nodes. The objective in the design was to determine the least cost way of adding tunnels in
parallel to the existing tunnels to achieve the projected flows and meet the minimum allowable pressure
design criteria. The available pipe sizes and corresponding costs (in 1969 dollars) as listed in Table 20.2
have been discretized from the fitted cost function for the new paralleled pipes as given by

1.24
C f = 1.1D (20.1)

where

• C f = cost per foot length of pipe ($/m or $/ft)

• D = the diameter of pipe (in.)
There were really two questions in optimizing the design of the tunnels expansion — the first was that
for each of the existing tunnels — did a tunnel need to be built in parallel to the existing tunnel? If a tun-
nel was required then the second question was — what diameter or size should the new tunnel be?

480
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Table 20.1 shows the change in minimum cost configuration from 1969 to 1996 by researchers who
have used various optimization techniques to attempt to find the minimum cost expansion. The cost of
the water distribution system expansion that satisfies the minimum allowable pressure constraints has
been reduced to $38.8 million that is about one – half of the original cost.

Figure 20.1 The layout of the New York City Tunnels network

Table 20.1 Minimum cost solutions for the New York City Tunnels expansion determined by
various optimization techniques
Researchers Year Cost ($million – Optimization
1969 dollars) technique
Schaake and Lai 1969 78.1 Linear programming
(LP)
Quindry, Brill and Liebman 1981 63.6 Generalized reduced
gradient (GRG) —
nonlinear program-
ming (NLP)
Gessler 1982 41.8 Partial enumeration
Bhave 1985 40.2
Morgan and Goulter 1985 39.2 Iterative LP with heu-
ristics
Murphy, Simpson and 1996 38.8 Genetic algorithms
Dandy

481
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

The solution of Lai and Schaake (1969) at a cost of $78.1 million specified that 18 of 21 of the possible
pipes be duplicated. Proposed pipe diameters based on continuous rather than discrete pipe sizes, ranged
from 19.52 in. to 133.07 in. Quindry et al. (1981) used a nonlinear programming approach to identify a
solution with 10 pipes that lowered the cost to $63.6 million. Gessler (1982) presented a partial enumer-
ation technique (where the range of possible choices is severely pruned and pipes are grouped), all pos-
sible combinations were then enumerated and the lowest cost – feasible solution was selected. The
outcome was a solution with the duplication of just 7 pipes at a cost of $41.8 million. Morgan and
Goulter (1985) used linear programming combined with heuristics to develop a six pipe solution at
$39.2 million.

The genetic algorithm technique was used by Murphy et al. (1993) to identify a series of alternative
designs with costs less than earlier optimized solutions. Three designs were found with costs of $38.80,
$39.06 and $39.17 million that were less expensive than the Morgan and Goulter (1985) solution of
$39.20 million. Each solution calls for six pipes in parallel. Choices were made from 15 allowable sizes
(ranging from 36 in. to 204 in. in diameter in increments of 12 in.). Another interesting solution was
identified in the GA optimization analysis — a solution with a cost of $33.62 million that only required
five parallel pipes that was marginally infeasible with pressure head deficits of about 1.0 ft at three
nodes in the water distribution system. The sensitivity of the cost of upgrading the New Tunnels system
to the possibility of slightly relaxing the design criteria is a further outcome of the GA optimization pro-
cess.

Table 20.2 Pipe data for New York City tunnels network
Pipe Start End Length Diameter Pipe Start End Length Diameter
no. node node of existing no. node node of existing
Pipe pipe
(ft) (in.) (ft) (in.)
[1] 1 2 11,600 180 [12] [13] 12 12,200 204
[2] 2 3 19,800 180 [13 [14] 13 24,100 204
[3] 3 4 7,300 180 [14] [15] 14 21,100 204
[4] 4 5 8,300 180 [15] [1] 15 15,500 204
[5] 5 6 8,600 180 [16] [10] 17 26,400 72
[6] 6 7 19,100 180 [17] [12] 18 31,200 72
[7] 7 8 9,600 132 [18] [18] 19 24,000 60
[8] 8 9 12,500 132 [19] [11] 20 14,400 60
[9] 9 10 9,600 180 [20] [20] 16 38,400 60
[10] 11 9 11,200 204 [21] [9] 16 26,400 72
[11] 12 11 14,500 204

Table 20.3 Node data and minimum allowable HGLs

Node Water Minimum Node Water Minimum
no. demand allowable HGL no. demand allowable HGL
(cfs) (ft) (cfs) (ft)
1 reservoir 300.0 11 170.0 255.0
2 92.4 255.0 12 117.1 255.0
3 92.4 255.0 13 117.1 255.0
4 88.2 255.0 14 92.4 255.0
5 88.2 255.0 15 92.4 255.0
6 88.2 255.0 16 170.0 255.0

482
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Table 20.3 Node data and minimum allowable HGLs

7 88.2 255.0 17 57.5 272.8
8 88.2 255.0 18 117.1 255.0
9 170.0 255.0 19 117.1 255.0
10 1.0 255.0 20 170.0 255.0

Table 20.4 Available new pipe sizes and costs for New York City Tunnel expansion (as of 1969)
Diameter Pipe cost Diameter Pipe cost Diameter Pipe cost
(in.) ($/ft) (in.) ($/ft) (in.) ($/ft)
36 93.5 96 316.0 156 577.0
48 134.0 108 365.0 168 632.0
60 176.0 120 417.0 180 689.0
72 221.0 132 469.0 192 746.0
84 267.0 144 522.0 204 804.0

20.5.3 Combinatorially large design networks

The number of possible options in designing a water distribution system is usually unbelievably enor-
mous. Typical choices that are required to include

• diameters of pipes (new pipes or pipes to be laid parallel to existing pipes)

• pipes to be cleaned
• locations, sizes, and normal water surface operating levels of tanks
• locations of regulating valves and settings
• locations, sizes and operating schedules of pumps
For the New York City tunnels network (Schaake and Lai 1969) involving the decision as to which of
the 21 existing pipes to duplicate, with 16 pipe size choices for each pipe diameter (including zero diam-
eter) for the duplications the number of possible combinations is

21 84 25
16 = 2 = 1.934 10

Even if a computer was able to evaluate 1 million different design combinations per second, complete
enumeration of every possible alternative for the New York City tunnels network would take 613 billion
years. Thus complete enumeration of the entire search space to find the global optimum solution is usu-
ally infeasible even for small design networks. As a result, effective optimization techniques are
required to be able to quickly search for low – cost designs. Most water distribution systems networks
involve search space sizes that are much bigger than the example given for the New York City tunnels
network outlined above in Section 20.5.2.

20.6 The Genetic Algorithm Optimization Process

20.6.1 Introduction
During the 1990s “evolutionary” optimization techniques became popular. Genetic algorithm optimiza-
tion is a relatively new technique that has only become commonly considered as an optimization tech-
nique since the late 1980s and early 1990s. Genetic algorithms have their roots in computing science
(Holland 1975). It was really Goldberg (1989), with his widely used book entitled Genetic Algorithms in
Search, Optimization and Machine Learning, that first made inroads into the practical use of genetics
algorithms for application to engineering design problems.

483
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

A genetic algorithm (GA) is a powerful population orientated search algorithm based upon mechanisms
of natural selection and genetics (Holland 1975, Goldberg 1989). Application of genetic algorithms to
the design of water distribution systems was pioneered at the University of Adelaide, South Australia in
the early 1990s (Murphy and Simpson 1992). “Survival of fittest” relentlessly drives the genetic algo-
rithm towards improved designs of water distribution networks. The genetic algorithm selects, combines
and manipulates possible solutions in a search for the lowest cost water distribution system. The opera-
tors within a genetic algorithm search are analogous to survival in nature, reproduction and the combina-
tion of chromosomes in the search of the best adaptation in natural systems. Genetic algorithms exploit
coding similarities in a very general way (Goldberg 1989). GAs are not constrained by requirements
such as continuity, derivative existence, unimodality, linearity, etc. that are often necessary in other opti-
mization techniques. A genetic algorithm starts with a population of strings representing water distribu-
tion system designs and thereafter generates successive populations of strings. The GA relies on the
diversity of the initial population to move generation by generation in a parallel fashion towards
improved solutions.

One of the advantages of a genetic algorithm is the robustness of the technique. The GA exploits simi-
larities in codings and thus can deal with the optimization of a broader class of functions than can many
other optimization procedures.

In GA optimization of water distribution systems there is a very clean separation of the optimization
process and the simulation process for modeling of the behavior of various water distribution system
designs. All the GA sees is a population of strings (containing a series of numbers or bits) and their asso-
ciated fitnesses. These numbers in the strings represent values that are choices or options for decision
variables that are to be selected as outcome of the optimization procedure. In the case of optimization of
water distribution systems, the genetic algorithm passes to the simulator the sizes of components repre-
sented by the string of numbers (pipe diameters, pump sizes and locations, valve locations and settings,
tank volumes and normal operating levels). The GA is linked to the hydraulic simulation model of the
particular water distribution system it is trying to optimize. The hydraulic simulation model is then run
for each of the water demand loading cases (peak hour, peak day – extended period and fire demand
loadings). The performance of the water distribution system is then assessed by evaluating whether the
specified design criteria have been satisfied. The corresponding fitness of each particular string is deter-
mined for all members of the population and then passed back to the genetic algorithm. The genetic
operators are then applied to the population of strings driven by selection of the fittest strings and an
intermixing of the strings by crossover and mutation. Only the strings and fitnesses are used by the
genetic algorithm to create a new improved population of strings or design solutions. The GA does not
know whether it is optimizing a water distribution system design, a structure, a gas turbine design or a
financial system.

20.7 Genetic Algorithm Optimization

20.7.1 What is it?
The genetic algorithm (GA) technique is based on an analogy to natural genetics including natural selec-
tion and reproduction of species. The operators within a GA are analogous to survival in nature, repro-
duction and the combination of chromosomes in the search of the adaptation of natural systems. The
genetic algorithm technique is also referred to as an evolutionary technique or a stochastic search tech-
nique. In applying the technique, a GA deals with a population of solutions particular to the design prob-
lem being optimized. Decision variables that are to be selected as part of the design in a piped
distribution problem (e.g. diameters of pipes, whether a pipe should be cleaned or not, location of
pumps, sizes of pumps, locations of tanks, optimal normal water reservoir or tank operating elevations,
pump location and size and pump operation schedules) are coded as strings in the genetic algorithm
technique. Choices for each decision variable are provided. For example, a new pipe may be selected as
one of four pipe diameters of say, for example, 100, 200, 300 and 400 mm. Table 20.5 shows 8 pipe
diameter choices. The genetic algorithm also requires cost information (material cost plus installation)
as well as head loss characteristics of the pipe (e.g. relative roughness height  or Hazen – Williams C
value). The genetic algorithm selects, combines and mutates possible solutions of the strings that repre-
sent the water distribution network designs (Simpson 1996). The operators of selection (that allows

484
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

strings of higher fitness to go to the mating pool for the next generation), crossover (to allow mixing or
mating of strings) and mutation (to reinject diversity into the population) are used over and over on the
population of water designs to continuously improve the solutions. The GA optimization process is a
structured search technique that very efficiently and quickly narrows in on low cost designs.

Table 20.5 A decoding lookup table

Nominal Actual or Binary Integer Roughness height Unit cost
diameter internal coding coding  for
diameter Darcy – Weisbach
friction factor
(mm) (mm) (mm) ($/m)
150 151 000 0 0.2 49
200 199 001 1 0.2 63
250 252 010 2 0.2 95
300 305 011 3 0.2 133
375 384 100 4 0.2 171
450 464 101 5 0.2 220
500 516 110 6 0.2 270
600 615 111 7 0.2 330

20.7.2 Formulation of the optimization problem

The optimization problem may be stated as follows (Simpson et al. 1994, Dandy et al. 1996).

For a given possible layout of pipes and a set of specified water demand loadings patterns at nodes (for
example, peak hour, peak day and fire loadings)

• find the combination of pipe sizes (for new pipes, replacement pipes and pipes in parallel),
pump locations and sizes, tank locations and sizes and valve locations and settings (pres-
sure reducing valves, pressure sustaining valves and flow control valves)
• that minimizes the present value of material, construction and operating costs
• subject to specified constraints
Examples of constraints include

• minimum allowable and maximum allowable pressure constraints must be satisfied at cer-
tain nodes in the water distribution system
• maximum allowable velocity constraints must not be exceeded in pipes in the water distri-
bution system
• minimum diameter requirements may be specified for certain pipes in the water distribution
system
• only commercially available pipe diameter sizes, pump and valve sizes are to be used
• tanks of increments of certain volumes may be specified as choices for the water distribu-
tion system
Other design criteria that may be included in the optimization process include reliability of the system
for outage of certain pipelines. It may be acceptable to provide a proportion of the peak flow at a
reduced minimum allowable pressure when particular pipes are unable to be used. Looped water distri-
bution systems offer increased reliability for multiple loading cases especially in the case of off – farm
piped irrigation delivery systems (Dandy et al. 1996).

485
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

20.7.3 Steps in using genetic algorithms for water distribution system optimi-
zation
Goldberg (1989) referred to the application of the three operators of reproduction, crossover and muta-
tion as a standard genetic algorithm. The following steps summarize an implementation of a genetic
algorithm for optimizing the design of a water distribution network system (Simpson, Murphy and
Dandy 1993, Simpson, Dandy and Murphy 1994)

1. Develop a coding scheme to represent the decision variables to be optimized and the corre-
sponding lookup tables to represent the choices for the design variables (Section 20.8), e.g.
pipe diameters of new pipes, pipe diameters of duplicate pipes, locations of pumps, tanks
and pressure regulating valves, sizes of pumps, normal operating levels of reservoirs, pres-
sure regulating valve settings.
2. Choose the population size (N) — this may vary between 50 and 1000 (see Section 20.9).
3. Choose the form of the genetic algorithm operators (Section 20.12), e.g. population size,
selection scheme — tournament selection or biased Roulette wheel, crossover type — one –
point, two – point or uniform, and mutation type — bit – wise or creeping.
4. Choose values for the genetic algorithm parameters (e.g. crossover probability, mutation
probability).
5. Select a seed for the random number generator.
6. Randomly generate the initial population of water distribution network designs (see
Section 20.10). Each of the N strings represents a possible combination of components and
thus represents a different configuration of water distribution system network design.
7. Decode each string in the population by dividing into substrings and then determine the
corresponding decision variable choices (using the lookup tables).
8. Compute the cost of each of the water distribution system design in the population.
9. Analyze each water distribution system design with a hydraulic solver for each water
demand loading case to compute pipe flows, pressures and pressure deficits (if any).
10. Compute a penalty cost for each water distribution system where design constraints are vio-
lated.
11. Compute the fitness of each string based on the costs in steps 8 and 10.
12. Generate a new population of designs in the next generation using the genetic algorithm
operators — selection, crossover and mutation (see Section 20.13, Section 20.14 and
Section 20.15).
13. Record the lowest cost solutions from the generation.
14. Repeat steps 7 to 12 to produce successive generations of populations of designs — stop if
all members of the population are the same.
15. Select the lowest cost design and any similarly low cost design of different configuration.
16. Check if any of the decision variables have been selected at the upper bound of the possible
choices in the lookup table. If so, expand the range of choices and rerun the genetic algo-
rithm.
17. Repeat steps 4 to 15 for say, ten different starting random number seeds.
18. Repeat steps 4 to 16 for successively larger and larger population sizes.
20.8 Coding Schemes
A selected mapping between the coded substrings and the design variables associates the artificial
genetic code with the water distribution system network being optimized (Dandy et al. 1996). A genetic
algorithm coding scheme is required for each of the design variables to be selected as part of the design.
Examples include

1. Pipe diameter, material and class of new pipes.

2. Pipe diameter, material and class of duplicate pipes.
3. The possibility of eliminating or cleaning existing pipes.
4. Possible locations of source pumps and/or booster pumps.
5. Sizes of pumps (that is, rated head and rated flow)
6. Possible locations of storage tanks.

486
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

7. Sizes and normal operating levels for the tanks.

8. Possible locations of pressure regulating valves (PRVs, FCVs and PSVs).
9. Pressure settings for the pressure regulating valves.
10. The operation schedule for pumps.
Each decision variable is usually coded as a finite – length string. As an example, consider 14 new pipes
that are to be sized using genetic algorithm optimization as shown in Figure 20.2 referred to as the 14 –
pipe water distribution system. A three bit substring is used to represent each decision variable for each
pipe whose diameter is to be optimally sized. Thus the lookup table as shown in Table 20.5 for each pipe
contains 8 different diameters. Note that in Figure 20.2 all the pipe sizes have been assigned the largest
pipe size and as a consequence this would lead to the largest possible cost of the water distribution sys-
tem. The decoded string is shown in Table 20.6. If all the pipes were minimum size pipes represented by
substring bit combinations of 000 then the least expensive water distribution system would result.

A string or chromosome for the design problem in Figure 20.2 is made up of the 8 substrings for the
three existing pipes that may either be cleaned or duplicated and the 5 new pipes.

Table 20.6 Genetic algorithm string coding scheme for the 14 – pipe network
Pipe Number [1] [4] [5] [6] [8] [11] [13] [14]
Substring 111 111 111 111 111 111 111 111
Diameter (mm) 615 615 615 615 615 615 615 615
Integer coding 7 7 7 7 7 7 7 7

Figure 20.2 A water distribution system showing genetic algorithm coding substrings

A range of pipe size diameter choices as shown in Table 20.5 is provided in the form a lookup table.

487
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

In Table 20.5, eight choices for pipe [14] are provided. For a binary coding system the number of
choices must be a multiple of 2n where n = 1, 2, 3, 4,.... etc. (that is 2, 4, 8, 16, 32 etc.). A different
decoding table may be specified for each of the decision variables if necessary. It may be seen that
binary coding is somewhat restrictive when 2n choices must be provided. An alternative to binary cod-
ing is integer coding that allows the number of choices to be any integer number (e.g. 3, 10, 15 etc.).
Integer codings are shown in Table 20.5.

20.9 Selecting the Population Size

Selection of the population size is an important parameter in applying the genetic algorithm procedure.
Tests have been carried out on a variant of the water distribution system shown in Figure 20.2 (Simpson
and Goldberg 1994). A population of 1000 members was needed to ensure that the known global mini-
mum was found by the GA search (the global minimum was known separately from the complete enu-
meration of approximately 16 million different designs). A population size of 100 did not always find
the optimal solution.

20.10 The Randomly Generated Initial Population

Once the coding representation has been selected an initial population is generated randomly. Lookup
tables of choices for each decision variable along with corresponding costs need to be developed. An
example of a lookup table was shown in Table 20.5. A binary coding scheme is used for the 24 – bit for
the optimal design of the water distribution system problem as shown in Figure 20.2. As an example,
consider a population of size N = 4 as shown in Table 20.7. Each bit position takes on either a zero (0) or
one (1). These four strings of length 24 – bits each could be randomly generated by the toss of a coin — 0
for heads and 1 for tails. In a computer code a uniform random number generator may be used with a
range of 0.0 to 1.0. For a random number less than 0.5 the bit is set to a value of zero while for random
numbers greater than or equal to 0.5 the bit is set to a value of one. Once the strings have been randomly
generated, they are decoded into the corresponding sizes using the lookup table in Table 20.5.
.

Table 20.7 Example of an initial population of size N = 4 based on binary coding

Pipe Number [1] [4] [5] [6] [8] [11] [13] [14]
String number 1 100 010 011 111 110 001 000 101
Diameter (mm) 384 252 305 615 516 199 151 464

String number 2 101 100 000 000 011 010 110 011
Diameter (mm) 464 384 151 151 305 252 516 305

String number 3 100 110 011 101 110 011 110 011
Diameter (mm) 384 516 305 464 516 305 516 305

String number 4 001 010 100 101 000 111 010 010
Diameter (mm) 199 252 384 464 151 615 252 252

20.11 The Fitness Function

A fitness function needs to be defined for the GA optimization of the water distribution system. A fit-
ness function that has been used successfully in the past has included two components (Murphy and
Simpson 1992). The first part is the cost of the physical elements required for the design represented by
the GA string of numbers. This usually includes costs of new pipes (purchase cost and installation
cost) — refer to this cost as C1. Examples include the costs of new pipes, costs of pumps and annual
pump operating costs taken back to a net present value. A penalty cost is added to the cost C1 to reflect

488
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

the degree by which the design does not meet the design constraints — refer to this cost as C2. The fitness
of each string may be computed from the total cost of the water distribution network design (the sum of
the real cost and the penalty cost). A number of different ways may be used to compute the fitness of a
string. A common way is to take the inverse of the total cost whereby the fitness is given as

1
Fitness = -------------------- (20.2)
C1 + C2

where

• C 1 = costs of new infrastructure (purchase cost and installation cost) ($)

• C 2 = penalty cost to reflect the degree by which the design does not meet the design
constraints ($).
An alternative method for computing the fitness is to take the negative of the total cost as

Fitness = –  C 1 + C 2  (20.3)

Different methods may be used to compute the penalty costs. Consider an example of a penalty cost that
is applied for pressures in the water distribution system being less than the minimum allowable pressure.
A penalty cost may be based on either

• Case 1: Only the node with the worst pressure deficit dmax
• Case 2: All nodes with a pressure deficit (d1, d2,..., dND) where ND = number of nodes in the
water distribution system with pressure deficits
A hydraulic simulation needs to be run for the design configuration to compute the flow distribution and
nodal pressures in the water distribution system. Consider the case where a pressure deficit occurs. A
cost penalty factor K needs to be assigned to enable computation of the penalty costs. The penalty multi-
plier factor K is a measure of worth per meter (or per foot length) attributed to having pressure heads
below the minimum allowable pressure heads as specified by the design criteria. This factor should be
selected such that near – optimal infeasible solutions are highly fit so that the optimum solution is
approached from both above and below (Simpson et al. 1994). Consider the penalty cost for Case 1 as

C 2 = K d max (20.4)

where

• K = penalty cost multiplier ($/m or $/ft)

• d max = the pressure deficit at the node with the worst pressure deficit (m or ft).

For Case 2, the penalty cost would be

ND
C2 =  Kdi (20.5)
i=1

where

• ND = number of nodes with pressure head deficits (m or ft)

• d i = the pressure deficit at the node i (m or ft).

For example, consider a K value of $5000/m of pressure deficit applied to the worst node in the water
distribution system where the pressure deficit is 7 m. The penalty cost from Equation 20.4 would be

C 2 = K d max = 5000  7  = $35,000

489
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Penalty costs may be cumulated over a number of water demand loading cases. In addition, if penalty
factors are applied for different design criteria then different penalty factor values may need to be used
for each of the different design criteria. For example, Table 20.8 shows different design criteria and cor-
responding penalty factor values.

Table 20.8 Penalty factor values

Design criteria Typical Penalty Units
requirement factor
symbol
Minimum allowable pressure > 20 m K1 $/m of pressure deficit
head
Maximum allowable pressure < 90 m K2 $/m of pressure excess
head
Maximum allowable velocity < 2.5 m/s K3 $/(m/s) of excess above the
minimum allowable velocity

The selection of appropriate values of the penalty factors K1, K2, and K3 depend on the magnitude of the
input costs of the water distribution system being optimized. A trial – and – error procedure is needed to
determine appropriate values. If the penalty factors are too large, then any design solution in the genetic
algorithm population is penalized so severely that it is very unlikely to be selected to go into the mating
pool for the next generation. Thus only feasible solutions would survive. On the other hand, if the pen-
alty factors are too small then after many generations of solutions from the GA process the population
would be dominated by infeasible solutions. Thus the lowest cost solutions do not satisfy the design
constraints. Reduction of the penalty factor makes the marginally infeasible solutions more prominent in
the search (Dandy et al. 1993). A balance between small and large values of penalty factors is required
so that the GA optimization process approaches the optimum solution from both the feasible side and
the infeasible side.

During the genetic algorithm process, it is desirable to keep solutions in the population that only just fail
to meet the design criteria. As a result, the selection of the magnitude of the penalty factors is important.
Consider an example, where the pressure at one node is in deficit by 0.1 m (19.9 m instead of the mini-
mum allowable pressure head of 20.0 m). This solution may be deemed as being acceptable. Clearly the
string that represents this solution for the design of a water distribution system contains useful genetic
material. If the penalty factor is set too high, this type of solution would be eliminated from the popula-
tion too readily (Simpson et al. 1995). It is best not to discard solutions that are only slightly infeasible
as a correctly placed crossover or mutation may produce an optimal string.

20.12 An Overview of Genetic Algorithm Operators

The operators within a genetic algorithm provide probabilistic transition rules to guide the search (Gold-
berg 1989). A simple genetic algorithm has three operators namely

• reproduction or selection to create a mating pool to base the next generation on

• crossover
• mutation
There are many variations of these operators.

20.13 The Selection Operator or the Reproduction Operator

The process of “survival of the fittest” relentlessly drives the genetic algorithm towards improved
designs of water distribution systems. Selection or reproduction selects a mating pool for the next gener-
ation from the current population of water distribution system designs based on fitness values of strings
in the previous generation. The selection process determines those members of the current population
that have copies in the next generation (usually more than one for the fitter of the strings in the popula-

490
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

tion). All selection methods rely on some form of the principle of “survival of the fittest” where more
copies of strings with larger fitness are preferred as candidates to go into the mating pool for the next
generation. Two commonly used operators are proportionate selection (also called Roulette wheel selec-
tion) and tournament selection.

20.13.1 Proportionate or weighted roulette wheel selection

Proportionate selection (Goldberg 1989) may be thought of as a weighted roulette wheel that has slots
sized according to the fitness of each member of the population. The selection operator assigns each
string in the population to a segment of the roulette wheel. The size of each segment is made propor-
tional to the fitness fi of the string. The probability of selection of a particular string is

fi
p i = ---------------------- (20.6)
N
 f j
j=1

ROULETTE WHEEL SELECTION PROBABILITY

where

• fi = fitness of string i
• pi = probability of selecting a string to be included in the mating pool for the next genera-
tion
• N = population size
To implement proportionate selection a cumulative distribution function is developed to enable the
weighted roulette wheel to be simulated. Consider as a trivial example a population size of N = 10. If the
fitness of 8 members is 0.08 and the fitness of the remaining two members (strings 9 and 10) is 0.18 then
the cumulative density function (CDF) has points 0.0, 0.08, 0.16, 0.24, 0.32, 0.40, 0.48, 0.56, 0.64, 0.82,
1.0. A uniformly distributed number between 0 and 1 is generated and depending on the segment of the
CDF that the random number “falls in” determines the member of the population that passes to the mat-
ing pool for the next generation. On average, strings 9 and 10 would receive more than twice the number
of copies than strings 1 to 8 because the size of the segment on the weighted roulette wheel is 0.18 com-
pared with 0.08. Proportionate selection was used in early studies for optimization of water distribution
networks (Murphy and Simpson 1992, Simpson et al. 1994). Proportionate selection may be improved
by using fitness scaling (Murphy, Simpson and Dandy 1993) as shown in Equation 20.7.

n
fi = fi (20.7)

where

• fi = fitness of string i
• n = scaling exponent
20.13.2 Tournament selection
There are various forms of tournament selection with the most common involving the comparison of
two randomly selected members of the population of water distribution system designs. Two – member
tournament selection are described first followed by the more general s – member tournament selection.

In two – member tournament selection (where s = 2), a contest is conducted between randomly selected
pairs of strings representing water distribution system designs from the population. If the current gener-
ation strings of population of size N is randomly divided into groups of two members selected without
replacement then a total of N/2 groups are formed. Each pair of strings is considered in turn. The string
that has the largest fitness out of the pair is transferred to the mating pool for the next generation. This
process is repeated for all N/2 groups. To fill the mating pool with a total of N strings it is necessary to
carry out the same process as described above another time. This second set of tournaments adds another

491
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

N/2 strings to the mating pool. Analysis of the N strings in the mating pool formed by tournament selec-
tion reveals that the fittest member of the previous generation obtains exactly 2 copies in the mating
pool (or more copies if there is more than one copy of the fittest string in the population). This occurs
because the fittest string always wins the tournament irrespective of which string it is paired up with.
The least – fit string in the population does not receive any copies in the mating pool as it cannot win a
tournament. All other strings in the population have between zero and two – members in the mating pool
depending on who their opponents were in the tournament.

An extension of the paired tournament selection process just described is s – member tournament selec-
tion. Now a contest is conducted between a specified number s of the strings from the population. The
current generation of population of size N is randomly divided into groups of s members selected with-
out replacement such that N/s groups are formed. It is assumed that N is exactly divisible by the tourna-
ment size s. The string in each subgroup of s that has the largest fitness is transferred to the mating pool
for the next generation. This process is repeated for all N/s groups. To fill the mating pool with a total of
N strings it is necessary to carry out the same process as described above another s – 1 times. The fittest
member of the previous generation obtains at least s – copies in the mating pool. The strings with the
( s – 1 )th fitness or smaller does not receive any copies in the mating pool as they cannot win a tourna-
ment. All other strings in the population have between zero and s – members in the mating pool depend-
ing on who their opponents were in the s – member tournament.

Tournament selection has been shown to be much more effective than proportionate selection (Simpson
and Goldberg 1994) and is the selection method of choice. The variable exponent in Equation 20.7 is
used to modify the fitness as part of the improved GA formulation (Dandy et al. 1996). The value of the
exponent n is allowed to increase in steps as the GA run develops. The initial population is randomly
generated and thus has a diversity of water distribution system designs. The average string fitness will
be relatively low and individual fitnesses will vary considerably. A low value of n should be used at the
start of the GA run, for example, n = 1. A low value of n preserves some population diversity in the early
generations. After a number of generations the value of the exponent n in the fitness function may be
increased in steps during the intermediate generations. As the GA run progresses further, the highly fit
string similarities that have evolved begin to dominate the population. A high value of n, say 3 or 4, is
needed to accentuate the small differences in string fitness as most of the strings will be constructed of
similar genetic code and their magnitudes of raw fitness will be similar.

20.14 Crossover Operators

Crossover interchanges bit information between two pairs of selected strings. A probability of crossover
pc needs to be specified in the range of 0.0 to 1.0 — with values of 0.6 to 1.0 being the most usual. The
strings in the mating pool that have resulted from the selection process as described in Section 20.13 are
randomly paired up. Each pair of strings is considered in turn. There are different forms of crossover
operators including

• one – point crossover

• two – point crossover
• uniform crossover
Crossover results in a pair of strings that contains characteristics of the two parent strings from the mat-
ing pool.

20.14.1 One – point and two – point crossover

In one – and two – point crossover, the crossover operator results in an exchange of blocks of bits
between strings. The crossover process for each pair of strings is controlled by two random decisions.
First, a decision needs to be made as to whether the strings will be subjected to crossover or not. Let us
assume the probability of crossover is 0.7. Then, on average 70% of the strings over all generations will
be crossed over. To make the decision on applying the crossover operator, a random number from a uni-
form distribution between 0.0 and 1.0 is generated. If the random number is less than 0.7 (say for exam-
ple, 0.46786) the crossover operation will be carried out. If the random number is greater than 0.7 (say
for example, 0.84567) then the crossover operator is not applied and the two strings pass to the next gen-
eration.

492
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

If crossover is to be performed, the next random operation is the selection of a crossover point (for the
case of one – point crossover). Consider a string of 8 bits as shown in Figure 20.3, a random number in
the range 1 to 7 with a uniform distribution is generated (say, for example 5). This random number then
becomes the crossover point. All bits from the first and second strings are interchanged to the right –
hand end of the strings past the crossover point. For the example, bits 6 to 8 are interchanged between
the strings as shown in Figure 20.3.

Figure 20.3 One – point crossover

For two – point crossover, two crossover points along the pair of strings are chosen randomly (for exam-
ple, points 3 and 6 in the 8 – bit string example as shown in Figure 20.4). The bits between the crossover
points are interchanged.

Figure 20.4 Two – point crossover

20.14.2 Uniform crossover

The strings in the mating pool that have resulted from the selection process are randomly paired up.
Each pair of strings is considered in turn. In uniform crossover, once the decision to perform crossover is
decided (as described above), then each pair of corresponding bits are considered in turn. The bits are
interchanged between the strings with a probability of 50% as shown in Figure 20.5. To carry out this
operation on a particular pair of bits for the two strings selected for crossover, a random number
between 0.0 and 1.0 from a uniform distribution is generated. If the random number (for example,
0.31189) is less than 0.5 the bits are not interchanged between the two strings. If the random number (for
example, 0.97321) is greater than 0.5 the bits are interchanged between the two strings. This process is
repeated for every pair of bits in both strings. Once all pairs of strings have been considered for cross-
over, the next operation is mutation.

Figure 20.5 Uniform crossover

493
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

20.15 Mutation Operators

20.15.1 Bit – wise mutation
Mutation involves the occasional alteration of a bit in a string (for a binary coding scheme). Mutation
ensures no genetic material is irretrievably lost from one particular bit position. The probability of muta-
tion is usually very small (e.g. 0.03). Consider a mutation rate of 0.03. On average over many genera-
tions 3% of the bits in the populations of strings would be mutated. For the mutation process each bit in
a string is considered in turn. For the possible mutation of each bit, a decision needs to made as to
whether the bit will be mutated or not. A random number between 0.0 and 1.0 from a uniform distribu-
tion is generated. If the random number is less than 0.03 (say 0.023763) then the bit is mutated. The bit
is flipped from a one to a zero or zero to a one, if a binary or Gray coding scheme has been used. If the
random number is greater than 0.03 (say 0.65923) the bit is not mutated and stays the same. The proba-
bility of mutation pm is usually in the range 0.01 to 0.05. Guidelines for computing pm are

1-
p m  --- (20.8)
N

where

• N = population size

1
p m  --- (20.9)
L

where

• L = number of bits in the string or the length of the string

As an example consider, a population size of N = 500 and a length of string of L = 24. Thus the total
number of bits in a particular generation or for the population of 500  24 = 12000 . The lower limit on
the probability of mutation is

1
p m  --------- = 0.002
500

For the lower limit of pm, then on average 12000  0.002 = 24 bits in each generation will be mutated.
The upper limit of the mutation rate is

1
p m  ------ = 0.0417
24

If the upper mutation rate is used, then on average 12000  0.0417 = 500 bits will be mutated out of
the 12,000 total bits in the population. If the mutation rate is too high, the population will be disrupted
and as a result the GA will not be able to converge on low cost solutions.

Thus the range of suggested values for the probability of mutation for a population size N = 500 and a
string length L = 24 is

0.002  p m  0.0417

Mutation may be regarded as of second – order importance in the GA process. Simpson and Goldberg
(1994) showed that there was little difference if mutation was or was not included in the GA process
when optimizing the design of a 14 – pipe network.

494
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

20.15.2 Adjacency mutation

The adjacency mutation operator or creeping mutation is applied to complete decision variable sub-
strings chosen randomly from the coded string (Murphy et al. 1993, Dandy et al. 1996). The adjacency
mutation operator mutates the selected complete decision variable substring to an adjacent decision vari-
able substring up or down the list of decision variable choices from the look – up table. This contrasts to
bitwise mutation that may or may not produce an adjacent substring as described above in
Section 20.15.1. The adjacency mutation operator may allow for adjustment of conditional probabilities
of upward and downward mutations (for example, a larger probability of moving to a smaller adjacent
pipe size rather than to the larger adjacent pipe size).

20.16 Successive Generations

For each new generation of solutions or designs of population size N that result by successive applica-
tion of the selection, crossover and mutation operators, costs are calculated and hydraulic analyzes are
performed. Once the fitness for all strings in the new generation are computed, the process of selection,
crossover and mutation is repeated over and over. The genetic algorithm process is usually continued for
many generations. The lowest cost designs must be remembered along the way. The final result is a
range of low cost water distribution system designs that are usually of significantly lower cost than
water distribution system designs developed by designers using traditional trial and error techniques
based on hydraulic computer simulation software. Typically 50,000 to 300,000 hydraulic evaluations of
water distribution system designs are carried out during a genetic algorithm run (100 to 1000 genera-
tions). The size of the water distribution system in terms of the number of decision variables determines
the total number of generations to be evaluated.

The maximum number of generations that the GA is to be run, is usually preset. If the mutation operator
is set to zero then the population will converge with the same solution being present for every member
of the generation. Thus an additional check for termination of the GA process is whether the standard
deviation of all of the fitnesses in the population has become so small that it is likely that all the mem-
bers of the population are the same. If this occurs, the GA process is terminated as there is no point con-
tinuing selection and crossover because no new strings can possibly arise from continuing the GA
process.

The GA technique is often referred to as a global search technique. The use of this terminology is mis-
leading as there is no guarantee that the GA will converge on the global optimum solution. If the popu-
lation size is too small it has been shown that the GA will converge on a sub – optimal solution (Simpson
and Goldberg 1994). However, once the population size has been selected such that it is large enough,
the GA has been shown to be able to produce lower cost solutions for the optimization of water distribu-
tion system designs than traditional optimization techniques (Dandy et al. 1993, Dandy et al. 1996).
These low cost optimized solutions may or may not be the global optimum solution — there is no way to
prove this — either way.

20.17 Application of Genetic Algorithm Optimization to Water Distribu-

tion System Design
The genetic algorithm (GA) technique may be used to optimize the layout and to determine the optimal
combination of the following

• optimal sizes of components (including pipes, pumps and tanks)

• optimal location of components (pipes, pumps, tanks, pressure regulating valves)
• optimal rehabilitation strategies (including replacement, duplication or cleaning of pipes)
• optimal operation of pumps
• optimize operating set points
• optimize hydrant flushing
• optimize chlorine booster locations

495
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

The techniques apply equally well to new water distribution systems, expansions to existing water distri-
bution systems or rehabilitation of existing water distribution networks. A computer simulation model
that is usually calibrated with field measurements is required to accurately simulate the existing water
distribution systems performance (Frey et al. 1996). A number of different applications of genetic algo-
rithms to generate least cost designs of all components of a water distribution network are now
described.

20.17.1 Layout of a new system

Aspects of the layout of a water distribution system that may be optimized using genetic algorithm opti-
mization includes

• deciding on which new pipes should be incorporated in the design. Often new pipes will
have service connections off them and thus they cannot be eliminated. These pipes may be
constrained to be at least a minimum diameter in the GA optimization process (Simpson et
al. 1995).
• choosing the optimal locations of new storage tanks
• choosing the optimal locations of source or booster pump stations
• choosing the optimal locations of valves in the water distribution system (pressure reducing
valves — PRVs, pressure sustaining valves — PSVs and flow control valves — FCVs)
20.17.2 Sizing of components in a new system
The genetic algorithm optimization technique for water distribution system design may be used to select
the best combination of

• diameters and materials of new pipes

• rated head and rated flow of new source or booster pump stations or an upgrade of an exist-
ing pump at a plant
• volumes and the normal optimal operating water surface level of new storage tanks
• settings of valves in a water distribution system (PRVs, PSVs and FCVs)
20.17.3 Expansion of an existing system
The genetic algorithm may be used to determine which pipes in the existing system need to be strength-
ened by duplicating with pipes in parallel to existing pipes. Thus the GA may be used to optimize the
selection of the following in the expansion area

• locations, diameters and materials of new pipes

• locations, volumes and normal operating levels of new storage tanks
• locations and sizes (rated head and rated flow) of source or booster pump stations
• locations and pressure settings of valves (for example, PRvs, PSVs and FCVs)
Within the existing pipe distribution network, the following additional decisions may be optimized by
the genetic algorithm

• duplication of existing pipes with new pipes in parallel (not necessarily of the same diame-
ter)
• removal of existing pipes
20.17.4 Rehabilitation
Rehabilitation actions that may be evaluated by the GA include

• cleaning (and lining) of pipes (in order to reduce the internal pipe roughness of a pipe)
• elimination of an existing pipe
• replacement of an existing pipe with a new pipe (not necessarily of the same diameter)
• refurbishment of pumps in existing pumping stations

496
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

20.17.5 Operation
A number of operational decisions may be optimized by the genetic algorithm including

• number and combination of pumps in operation at a particular time during a day of opera-
tion
• determining the schedule of operation of pumps to ensure tanks do not empty and then refill
on the peak water demand day
• determining the normal water surface operating elevations for reservoirs or storage tanks
• determining the settings for valves (PRVs, PSVs and FCVs)
The optimization by the GA evaluates the hydraulic feasibility of the trial designs by checking the pres-
sure heads at water demand nodes to ensure they are above the minimum allowable pressures and that
all other design constraints are satisfied.

20.17.6 Tender Evaluation

In the planning and design phases, firm costs for supply of components (e.g. pipes) are generally not
known. The GA optimization analysis in common with the normal trial – and – error design process, is
based on both pipe costs and construction costs. These costs are usually based on historical cost data
stored in a cost database.

The genetic algorithm technique may also be useful when tender bids are received especially for pipes
of different materials (e.g. reinforced concrete, Hobas and uPVC). Differences between materials of
pipes in actual internal pipe diameters and roughness values for the same nominal diameter may be
taken into account in the optimization process. The genetic algorithm enables the optimal configuration
of the design to be determined based on tender bid prices. This approach was adopted for the Cobdogla/
Loveday irrigation system optimization study in South Australia (Simpson et al. 1996).

20.18 Case Studies

A number of water distribution systems have been previously successfully optimized by using GAs.
Applications include both municipal water supplies and off – farm piped irrigation delivery systems
(Simpson et al. 1996).

In applying the genetic algorithm technique to the design of these water distribution systems, it has been
shown that lower cost designs have been identified consistently when compared with the designs of
experienced designers or compared to designs determined by other optimization techniques. In compar-
ing the designs, it is evident that a number of pipes have often actually increased in diameter in the GA
design (Simpson 1996). However, the GA was able to identify overall pipe combinations that reduced
the total cost of the water distribution system design components compared to the cost of the trial – and –
error developed network. The savings in using GA optimization have been shown to be in the range of
5% to 49%. In addition, the lowest cost feasible solution for the New York City Tunnels network has
been found using genetic algorithms (Murphy et al. 1993, Dandy et al. 1996). The savings are only 1%,
however, this problem has been previously optimized by many other researchers as described in
Section 20.5.2.

20.18.1 Optimization of a 14 – pipe network

In this example, the sizing decision associated with 8 pipes is to be optimized using the GA optimization
procedure (see Figure 20.2). The case study has some interesting features including

• selection of the diameters of five new pipes

• determining whether three existing pipes should be cleaned, duplicated with a pipe in paral-
lel (not necessarily of the same diameter) or left alone
• three water loading demand patterns are specified representing a peak hour demand loading
case, and two fire loading cases. Each loading case has an associated set of minimum
allowable pressure heads at nodes. The load cases are given in Table 20.9.
• two supply sources are available

497
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Table 20.9 Water demand loading patterns and associated minimum allowable pressures
Demand pattern 1 Demand pattern 2 Demand pattern 3
Demand Minimum Demand Minimum Demand Minimum
Node allowable allowable allowable
pressure pressure pressure
head head head
(L/s) (m) (L/s) (m) (L/s) (m)
2 12.62 28.18 12.62 14.09 12.62 14.09
3 12.62 17.61 12.62 14.09 12.62 14.09
4 0 17.61 0 14.09 0 14.09
6 18.93 35.22 18.93 14.09 18.93 14.09
7 18.93 35.22 82.03 10.57 18.93 14.09
8 18.93 35.22 18.93 14.09 18.93 14.09
9 12.62 35.22 12.62 14.09 12.62 14.09
10 18.93 35.22 18.93 14.09 18.93 14.09
11 18.93 35.22 18.93 14.09 18.93 14.09
12 12.62 35.22 12.62 14.09 50.48 10.57

In Figure 20.6, the solid lines represent the existing system while the dashed lines depict the new pipes.
Elevations, pipe lengths, diameters and Hazen – Williams C coefficients of the existing water distribu-
tion system were provided as known information.

To determine the effectiveness of the GA in identifying the global optimum solution, a complete enu-
meration of every possible combination for all three water demand loading cases was carried out. Com-
plete enumeration of each of the alternative designs was possible because the size of the search space
was not prohibitively large (Murphy and Simpson 1992). The lowest cost feasible water distribution sys-
tems were tracked during the complete enumeration of 16,777,216 possible combinations that involved
a total of 50,331,648 hydraulic simulations. The best feasible water distribution systems are shown in
Table 20.10. A number of different configurations have very similar costs. There are at least two water
distribution systems with the lowest possible cost that represent the global minimum with a cost of
$1.7503 million. All top 20 solutions have a duplication of pipe [4] with a 356 mm diameter pipe size
while pipe [1] is not duplicated or cleaned. Some of the solutions suggested cleaning of pipe [1] com-
bined with duplicating pipe [4] with a 305 mm diameter pipe. Some of the top solutions that duplicate
pipe [5] with either a 152 mm, 203 mm or 254 mm diameter pipe or clean pipe [5].

Table 20.10 Results of complete enumeration (Murphy and Simpson 1992, Simpson et al. 1994)
Rank Total Pipe Pipe Pipe Pipe Pipe Pipe Pipe Pipe
no. cost [1] [4] [5] [6] [8] [11] [13] [14]
($m)
1 1.7503 leave dup 356 leave 305 203 203 152 254
2 1.7503 leave dup 356 leave 305 203 254 152 203
3 1.7725 leave dup 356 leave 305 203 203 203 254
4 1.7725 leave dup 356 leave 305 203 254 203 203
5 1.7910 leave dup 356 dup 203 254 203 203 152 254
6 1.7999 leave dup 356 clean 254 203 203 203 254
7 1.8010 leave dup 356 leave 305 203 254 152 254
8 1.8010 leave dup 356 leave 305 254 203 152 254

498
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Table 20.10 Results of complete enumeration (Murphy and Simpson 1992, Simpson et al. 1994)
9 1.8010 leave dup 356 leave 305 254 254 152 203
10 1.8115 leave dup 356 leave 356 203 203 152 254
11 1.8115 leave dup 356 leave 356 203 254 152 203
12 1.8115 leave dup 356 leave 305 203 305 152 203
13 1.8115 leave dup 356 leave 305 203 203 152 305
14 1.8131 leave dup 356 dup 203 254 203 203 203 254
15 1.8232 leave dup 356 leave 305 203 203 254 254
16 1.8232 leave dup 356 leave 305 203 254 203 254
17 1.8232 leave dup 356 leave 305 254 203 203 254
18 1.8232 leave dup 356 leave 305 203 254 254 203
19 1.8232 leave dup 356 leave 305 254 254 203 203
20 1.8285 leave dup 356 clean 254 203 254 152 254

The actual pressure heads are shown for the known global optimal water distribution system (ranked as
number 1 in Table 20.10) with a cost of $1.7503 million.

Table 20.11 Actual pressure head for the global optimum solution (Murphy and Simpson 1992)
Demand pattern 1 Demand pattern 2 Demand pattern 3
Actual Min. Deficit Actual Min. Deficit Actual Min. Deficit
head allow. ( – ) or head allow. ( – ) or head allow. ( – ) or
Node head excess head excess head excess
(m) (m) (m) (m) (m) (m) (m) (m) (m)
2 36.2 28.18 +8.02 24.95 14.09 +10.89 30.48 14.09 +16.39
3 30.47 17.61 +12.86 19.32 14.09 +5.23 24.52 14.09 +10.43
4 26.84 17.61 +9.23 16.15 14.09 +2.06 20.46 14.09 +6.37
6 46.87 35.22 +11.65 18.68 14.09 +4.59 34.34 14.09 +20.25
7 50.06 35.22 +14.84 12.75 10.57 +2.18 37.54 14.09 +23.45
8 59.23 35.22 +24.01 41.29 14.09 +27.20 47.93 14.09 +33.84
9 51.88 35.22 +16.66 24.06 14.09 +9.97 34.61 14.09 +20.52
10 49.79 35.22 +14.57 22.38 14.09 +8.29 26.65 14.09 +12.56
11 47.53 35.22 +12.31 24.82 14.09 +10.73 18.27 14.09 +4.18
12 50.00 35.22 +14.78 27.29 14.09 +13.20 13.72 10.57 +3.15

The pipes for the global optimum solution are shown in Figure 20.6.

499
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Figure 20.6 The configuration of the global optimal solution ($1.7503 million)

The genetic algorithm procedure applied to the 14 – pipe network to search for low cost solutions
involved the following steps (Murphy and Simpson 1992, Simpson et al. 1994).

1. Representation of design problem with a binary string. There are 8 decision variables in this
design optimization problem (for pipes [1], [4], [5], [6], [8], [11], [13] and [14]). Each deci-
sion variable to be selected by the GA is assigned a choice or look – up table. For eight dif-
ferent choices of diameter, a substring of 3 binary bits is used to represent the choices. Thus
for eight pipes, a 24 – bit binary string, where each bit is either zero or one, comprises
eight – 3 bit substrings. Each substring represents each decision variable as shown in
Figure 20.2. The maximum possible number of combinations of pipes is 224 or 16,777,216
different alternatives as shown above in the discussion on complete enumeration of the
search space. An example of a string, corresponding pipe numbers for substrings and deci-
sion variables from the look up tables is shown in Figure 20.12.

Table 20.12 A string for the network

Substring 001 010 000 000 101 001 011 111
Pipe no. [1] [4] [5] [6] [8] [11] [13] [14]
Decoding duplicate clean leave as new new new new new
152 mm pipe is 152 mm 152 mm 152 mm 152 mm 152 mm

The last five pipes are new pipes for which diameters must be chosen. A 3 – bit binary rep-
resentation is used for each of the eight decision variables resulting in a binary string of
total length of 24 – bits. The 3 – bit binary codings and the choices, unit pipe costs and
Hazen – Williams C values are given in Table 20.13.

500
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Table 20.13 Options for each decision variable for new pipes
New pipe Binary Unit cost Hazen – Williams C
diameter coding roughness value for
(mm) ($/m) the new pipe
152 000 49.54 120
203 001 63.32 120
254 010 94.82 120
305 011 132.87 120
356 100 170.93 120
407 101 194.88 120
458 110 232.94 120
509 111 264.10 120

The first three pipes in the binary string are existing pipes that may either be cleaned or
duplicated with a parallel pipe (not necessarily of the same diameter). The lookup decod-
ing table for duplicate pipes and/or cleaning of the three existing pipes is shown in
Table 20.14. The options available for existing pipes have been arranged in Table 20.14 in
terms of increasing cost ($/m).
2. Selection of values for genetic algorithm operators and parameters. The values selected for
the genetic algorithm parameter values include
•population size N = 100
•probability of crossover pc = 0.9
•probability of mutation pm = 0.02
•penalty multiplier, K = $70,000/m pressure deficit at worst node
•maximum number of generations, G = 50

Table 20.14 Options for decision variables of existing pipes

(1) (2) (3) (4) (5) (6) (7)
Duplicate pipe Unit Binary Hazen – Diameter Unit cost Hazen –
diameter (mm) cost coding Williams C of pipe to for pipe Williams C
or cleaning roughness value be cleaned cleaning value after
($/m) for the new pipe (mm) ($/m) cleaning
leave as exists 0.0 000 not applicable 152 47.57 120
duplicate with 152 49.54 001 120 203 51.51 120
clean existing pipe col. (6) 010 col. (7) 254 55.12 120
duplicate with 203 63.32 011 120 305 58.07 120
duplicate with 254 94.82 100 120 356 60.70 120
duplicate with 305 132.87 101 120 407 63.00 120
duplicate with 356 170.93 110 120
duplicate with 407 194.88 111 120

3. Generation of the initial population. A random number generator based on a recurrence

called a linear congruential generator (Barnard and Skillicorn 1988) was used to generate
each of the 24 – bits in each string of a population size of N = 100 strings with a zero or one.
4. Computation of the water distribution system cost. The GA considered each string in turn in
the population of 100 strings. Each 24 – bit binary string was decoded into the correspond-
ing pipe sizes and options (cleaning for pipes [1], [4] and [5] when appropriate). The cost of
each design based on the unit pipe costs in Table 20.13 and Table 20.14 was computed.

501
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

5. Hydraulic analysis of each water distribution system for all three loading cases. Steady
state pipe flows, heads and nodal pressure heads were computed for each of the three load-
ing cases.
6. Computation of penalty costs. The GA assigns a penalty cost for each loading case if a
water distribution system does not satisfy the minimum pressure constraints. The pressure
violation at the worst node was used. The maximum pressure deficit was multiplied by a
penalty multiplier factor of K = $70,000/m of pressure head deficit.
7. Computation of the total water distribution system costs. The total cost of each network in
the population was taken as the sum of the network cost (Step 3 above) and any penalty
costs (Step 5 above).
8. Computation of string fitnesses. The fitness of each string was taken as the inverse of the
total water distribution system cost (Step 6 above). Thus from Equation 20.2 follows that

1
Fitness i = ------------------------------------------------- (20.10)
Total Network Cost i

9. Applying the genetic algorithm operators of selection, crossover and mutation. Proportion-
ate or biased roulette wheel selection, one – point crossover and bit – wise mutation were
used as the genetic algorithm operators to create a mating pool and then the next generation.
As the probability of crossover pc is 0.9 it would be expected on average that 90% of all
pairs of strings selected for crossover from the mating pool to be crossed over. Thus 10% of
the strings pass from the mating pool to the next generation uncrossed. In the same way, as
the probability of mutation pm is 0.02 then it would be expected that the average number of
bits to be mutated in a generation to be 2%.
10. Generation of successive generations. Once a new generation is created, Steps 4 to 9 are
repeated over and over until the specified maximum number of generations of G = 50 is
reached. Thus a maximum of 50,000 function evaluations of different water distribution
network designs were allowed for each run (0.298% of the total 16,777,216 total possible
combinations).
11. Different random number seeds. Ten GA runs were performed for the optimum design of
the 14 – pipe water distribution system network using different random number seeds. The
GA found one of the two global optimum solutions in 8 out of 10 runs and always identified
near – optimal solutions. The lowest cost solutions achieved in each run are presented in
Table 20.15.

Table 20.15 Genetic algorithm results for a 14 – pipe network with a variation in random number
seed for 50,000 evaluations (N = 100, pc = 0.9 and pm = 0.02) — Murphy and Simpson 1992
Run Random Best feasible Difference Evaluation Ranked
number seed total network from number solution
cost optimum achieved number from
($million) Table 20.10
1 1000 1.7725 1.27% 29,070 3
2 2000 1.7503 0 10,350 1
3 3000 1.7503 0 43,740 2
4 4000 1.8115 3.5% 40,860 10
5 5000 1.7503 0 17,190 2
6 6000 1.7503 0 11,070 2
7 7000 1.7503 0 10,080 1
8 8000 1.7503 0 41,490 2
9 9000 1.7503 0 12,150 1
10 10000 1.7503 0 19,890 1

502
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

12. Different sets of genetic algorithm parameters. Ten GA runs were performed for the case
study water distribution system network using different sets of GA parameters. The GA
found one of the two global optimum solutions in 9 out of 10 runs The genetic algorithm
parameters and the lowest cost solutions achieved in each run are presented in Table 20.16.

Table 20.16 Results for 10 different sets of genetic algorithm parameters

Run Population Probability of Probability of Lowest network cost Evaluation
no. size N crossover pc mutation pm found (% difference number
from optimum) achieved
1 20 0.7 0.02 1.7503 3,962
2 50 0.7 0.02 1.7503 20,775
3 100 0.7 0.02 1.7503 28,000
4 150 0.7 0.02 1.7503 9,765
5 100 0.1 0.02 1.7503 12,920
6 100 0.5 0.02 1.7999 (2.8%) 5,400
7 100 1.0 0.02 1.7503 13,800
8 100 0.7 0.0 1.7503 10,150
9 100 0.7 0.005 1.7503 10,150
10 100 0.7 0.03 1.7503 2,660

The genetic algorithm method using the standard three – operators was found to be particularly effective
in finding global or near optimal solutions for the case study water distribution system in relatively few
evaluations (< 0.3% of the total search space for each computer run). Computer times were compara-
tively long for 50,000 evaluations of each of the 10 genetic algorithm runs with different starting seeds.
However, the GA was still effective in finding the global solution for 15,000 evaluations for each com-
puter run. One difficulty with the GA technique is knowing how many evaluations will be sufficient.

20.18.2 Other case studies

A number of case studies have been previously carried out. These are summarized in Table 20.17

Table 20.17 Cost savings using genetic algorithm optimization

Water distribution Previous design Number Previous Genetic % Savings by
system technique using of lowest cost Algorithm savings using the
traditional pipes solution lowest cost GA
computer solution technique
simulation
Seaford Rise Devel- WATSYS 27 $1.29 million $1.14 million 12.0% $150,000
opment, hydraulic simu-
EL126 Zone Ade- lation model (1993)
laide, South Austra- design by
lia, Engineering and experienced
Water Supply water distribu-
Department of South tion network
Australia (E&WS) designers
(E&WS)
Seaford Rise Devel- WATSYS design 105 $2.59 million $2.43 million 6.2% $160,000
opment, EL80 by experienced
Zone Adelaide, water distribu- (1993)
tion network
South Australia
designers
(E&WS)
(E&WS)

503
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Table 20.17 Cost savings using genetic algorithm optimization

Fort Collins – EPANET design 313 $5.85 million $2.96 million 49.4% $2.88 mil-
Loveland Water by local consul- (1995) lion
System Colorado, tant in Fort Col-
lins, Colorado
USA (Fort Collins –
Loveland Water
District)
Cobdogla/Loveday WATSYS design 50 $4.26 million $3.79 million 11.0% $470,000
Irrigation Network by experienced (1995)
Riverland, South water distribu-
tion network
Australia (E&WS)
designers
(E&WS)
Corbie Hill WATSYS design 83 $2.91 million* $2.45 mil- 15.8% $458,000
Irrigation Network by experienced lion* (1995)
Leeton, NSW (Mur- engineering con-
sultant
rumbidgee Irriga-
tion and Department
of Water Resources,
NSW)

Bermuda System, KYPIPE3 322 $10.89 million Not available NA

2300 and 2420 design by expe-
Zones rienced water
distribution net-
Nevada, USA
work designers
(Las Vegas Valley
(Las Vegas Val-
Water District) ley Water Dis-
trict)
Cadell Irrigation WATSYS design 62 $2.824 million $2.575 mil- 8.8% $249,000
Area Pipe Network by experienced lion
Riverland, South water distribu-
tion network
Australia
designers (SA
(Central Irrigation
Water Corpora-
Trust) tion)

Willunga Reuse 40 Job completed

Scheme
Adelaide, South
Australia
Grand Prairie, Lynx model 3000 Job completed
Texas, USA design by expe-
rience by
CH2M – Hill,
Dallas, Texas
Western half of the WATSYS design
Northern Mallee by experience by
pipeline project AWT, Sydney
750Job in prog-
(Wimmera Mallee
ress
Water, Ouyen, Vic-
toria)

504
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Table 20.17 Cost savings using genetic algorithm optimization

Expansion of Reno Stoner SWS 2000 Job in progress

Sparks Water model design by
Transmission and experience
Distribution Sys-
tem, (Sierra Pacific
Power Co., Reno
Nevada)
Woorinen irriga- Genetic algo- 350 Job in progress
tion area rehabili- rithm design
tation, used from the
planning phase
Victoria (Goulburn
onwards
Murray Water)

* Includes present value of pumping cost

20.19 Altering the Current Role of the Designer

Use of GAs should provide a more challenging role for the experienced water supply designer. Instead
of requiring the trial – and – error process of adjusting sizes of components in a water distribution system
design using computer simulation tools, the designer may spend time on brainstorming an expanded set
of possible options (such as alternative water distribution system layouts) for the GA to consider and
also time for evaluating the ranges of alternative designs that come out of the GA optimization process.
The role of the designer shifts to a higher plane when GA optimization is available for use. Considerable
engineering judgment is still required to conceptualize possible alternative choices for the GA to con-
sider as well as to evaluate the GA generated designs. The GA optimization process must always be sup-
plemented by engineering experience. The GA technique does not replace the current simulation
modeling processes in a system planning, design, or operations study. It simply automates the critical
trial – and – error solution search step in the process. Rather than manually inputting trial solutions and
then evaluating them with a computer simulation program, the GA technique directs a very effective
search by generating and evaluating potential solutions in an automated structured search process that
quickly narrow in on low cost solutions. GA optimization thus supplements the expertise of water distri-
bution system specialists. The interpretation of the designs generated by the genetic algorithm optimiza-
tion process lends insight into the behavior and characteristics of the water distribution system being
studied.

The difficulty for a designer using trial – and – error improvement of the design of a water distribution
system is that once the number of pipes exceeds about 10, the size of the search space becomes incredi-
bly large. Even if the designer has a great deal of experience it is very difficult to be able to determine
the lowest cost solution or even a range of low cost solutions. The design engineer does not have time to
investigate all the potential cost savings using computer simulation models and some good designs may
not be obvious even to an experienced design engineer.

20.20 Advantages of Genetic Algorithm Optimization

There are a number of advantages of GA optimization as outlined in the following sections.

20.20.1 Solutions are based on commercially available pipe sizes

The choice or look – up tables for new, replacement or duplicate pipes are selected based on commer-
cially available pipe sizes. This contrasts to nonlinear programming where solutions are output as con-
tinuous diameters. As a result solutions have to be rounded up or down to select the best combination of
commercially available pipe sizes. For example, a solution from the nonlinear programming optimiza-
tion may suggest the optimal diameter for a particular pipe is 282.7 mm. Based on Table 20.5, a decision
now needs to be made as to whether to round up the pipe size to 305 mm or round down the pipe size to
252 mm. Rounding down may result in the modified water distribution system network design not satis-
fying the design constraints. Thus the issue of rounding up or rounding down many pipe diameter sizes

505
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

actually results in a further optimization problem. Other discrete options include the decision of cleaning
or not cleaning a pipe, locations of pumps, tanks and valves — that are either yes or no decisions. In
practice GAs are one of the few optimization techniques where direct use of commercially available
pipe sizes and other discrete choices are used directly in the optimization.

20.20.2 A range of solutions result

The GA process produces a range of different designs for the water distribution system optimization
problem. Many of these designs will have different configurations but will be of a similar low cost. As
an outcome, the designer may choose a preferred design based on some other non – quantifiable criteria.

As an example, consider the rehabilitation of the water distribution system in an existing area. Two dif-
ferent optimized solutions may be available that involve upgrading a pipeline on streets that are parallel.
The GA optimization process may not have taken into account the disruption and trading loss for shops
along one of the streets. Thus the designer may select the solution that involves putting the pipeline
along the less busy road that would have less impact on the shop owners.

20.20.3 Full utilization of simulation modeling capability

One of the difficulties in the past of using the traditional optimization methods was the clash between
introducing another program that appeared to be opposed to the simulation modeling approach. Design-
ers raised the question “How do we know the results of the optimization match the results of the simula-
tion models that we have developed?” The concern is valid and as an outcome reinforced resistance by
designers to acceptance of optimization. GA optimization does not share this problem. A simulation
model is a crucial part of the GA formulation. The GA evaluates the fitness of a solution by running the
proposed design through a simulation model evaluation. Thus simulation is incorporated directly into
the optimization process. The GA is an enhancement of the simulation process rather than being a differ-
ent model. As an outcome one of the benefits of GA optimization is that any improvements made in the
simulation model (for example, adding water quality modeling) are reflected directly in the GA optimi-
zation process.

When a simulation model is used as an integral part of the optimization procedure as in GAs then there
is no need to determine the underlying equations for a new feature and then to incorporate them within
the optimization formulation. This is necessary for many of the traditional optimization techniques and
is clearly one of the advantages that GA optimization has over traditional approaches.

20.20.4 GAs deal with a population of solutions

A GA process deals with a population of solutions at one time. These solutions are initially randomly
spread out throughout the solution space, so the chance of reaching low cost global or near global solu-
tions is increased significantly (Simpson et al. 1994).

20.20.5 GAs only require fitness function information

GAs compute the fitness of a water distribution system design solution based on real cost and penalty
costs to account for violation of design criteria. No other auxiliary information is required such as deriv-
atives. Existence and continuity of derivatives is not required in the genetic algorithm optimization pro-
cess in contrast to other traditional mathematical programming optimization techniques.

20.21 Other Varieties of Genetic Algorithm Optimization Formulations

There have been many varieties applied to alter the original standard genetic algorithm formulation of
Goldberg (1989). Dandy et al. (1996) developed an improved GA that included fitness scaling with pro-
portionate selection, Gray coding (rather than binary coding) and creeping mutation. Messy genetic
algorithms have also been developed to improve the effectiveness of the GA approach.

20.22 Other Optimization Methods

Many of the traditional optimization techniques have been applied previously to the optimization of
water distribution systems including

506
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

• complete enumeration
• pruned enumeration (Gessler 1982, Gessler 1985, Loubser and Gessler 1990)
• Linear programming (Calhoun 1971, Alperovits and Shamir 1977, Quindry et al. 1981,
Morgan and Goulter 1985)
• nonlinear programming (El – Bahrawy and Smith 1985, Su et al. 1985, Leibman et al. 1986,
Lansey and Mays 1989, Lansey et al. 1989, Vigus 1989, Waters 1989, Duan et al. 1990)
• dynamic programming
To implement the traditional optimization techniques the underlying equations governing the problem to
be solved need to be put into the form appropriate for the particular optimization technique.

20.22.1 Complete enumeration

Complete enumeration is one approach for the optimization of water distribution systems. The technique
simulates every possible combination of discrete pipe sizes and components in a simulation model. The
lowest cost water distribution system is selected that satisfies the pressure constraints. The main draw-
back as mentioned earlier is the possibility of prohibitive amounts of computer time that are involved to
completely enumerate the search space even for very small problems.

20.22.2 Partial or pruned enumeration of the search space

This technique was proposed by Gessler (1982, 1985) and Loubser and Gessler (1990). As mentioned
earlier complete enumeration evaluates every possible combination of specified pipe diameter sizes and
sizes of other components for the proposed water distribution system layout. For large water distribution
systems, complete enumeration is impossible due to the huge size of the search space and would require
billions of years even on the fastest of computers. Partial or pruned enumeration was proposed as a way
of cutting down the size of the search space to manageable size such that it became practical to enumer-
ate all possible solutions in the subset. The two techniques used to prune the search space included

• the grouping of a number of pipes together to be restricted to be the same pipe size
• the reduction of the number of choices that a group of pipes may take on
During the pruned enumeration process the cost of the water distribution system is firstly computed for
each combination of pipe groups. Each water distribution system design is then simulated to determine
the flows in pipes and the HGLs and pressures at the nodes. The pressures are compared with the mini-
mum allowable pressures specified in the design criteria. If the design criteria are not satisfied the water
distribution system is designated as infeasible. A list of the currently lowest cost networks is maintained
as well as the networks that just fail to satisfy the design constraints.

Two heuristics were introduced by Gessler (1982) to further reduce the number of networks to be evalu-
ated during the pruned enumeration process. These included

• a cost filter — a combination of pipe sizes does not need to be tested for hydraulic feasibil-
ity if the cost of the solution is greater than some cost cut off. The cut off cost is set to be the
lowest known cost solution that satisfies the design pressure constraints.
• a size filter — a combination of pipe sizes does not need to be tested for hydraulic feasibility
if all the selected sizes are equal or less than the pipe sizes of some other infeasible solution
A significant disadvantage of these heuristics as proposed by Gessler, is that they may lead to the elimi-
nation of networks that are in fact lower cost yet satisfy the design constraints (Murphy et al. 1993,
Dandy et al. 1993). Partial enumeration suffers from the disadvantage that an arbitrary grouping of pipes
into sets, each with the same diameter is likely to exclude solutions that may be of lower cost. GA opti-
mization (Dandy et al. 1993) has been shown to find lower cost solutions than partial enumeration of
Gessler and Loubser (1990).

507
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

In looped systems it is possible to increase the pressure head at a deficient node by decreasing the size of
a pipe downstream of the node (Murphy et al. 1993 — Water Journal). This actually improves the pres-
sure deficiency with a water distribution system of lower cost that is in direct contradiction to Gessler’s
(1982) heuristic above. Dandy et al. (1993) concluded that a clear disadvantage of the pruned enumera-
tion technique is that is likely that optimum and other low cost networks will be excluded from the
search space as an outcome of the search space pruning process.

As seen previously complete enumeration of all solutions even for a relatively small water distribution
system would take billions of years. When the total size of the search space exceeds about 230 (1.073 bil-
lion different alternative solutions) then the number of combinations becomes very large and complete
enumeration will take a very long time. It is when problems exceed this size that GA techniques become
particularly useful.

20.22.3 Linear programming

To apply linear programming to the optimization of water distribution system design — linear equations
need to be set up. Head is a particularly nonlinear function of the diameter (to the power five). In order
to apply linear programming to the selection of the set of optimal diameters in a water distribution net-
work the problem is set up to optimize the length of pipe diameters of a particular size such that head
loss is indeed directly or linearly proportional to the length. This “trick” allows the optimization to pro-
ceed. The outcome of the optimization process is for each pipe between nodes to be two adjacent diam-
eters of pipe with lengths that sum up to the total pipe length between two nodes. Thus the solution is in
fact a “continuous” solution rather than a discrete one as the two pipe diameters may be converted to a
pipeline of an equivalent diameter (that will usually not be a commercially available pipe size).

20.22.4 Nonlinear programming

The decision variables in nonlinear programming take on a value that is assumed to be any value within
a continuous range of sizes. The objective function is based on a fitted cost function of pipe diameter
and length. Constraints include minimum diameters for particular pipelines, minimum allowable pres-
sures at supply nodes and possibly maximum allowable pressure heads. Nonlinear programming uses a
search technique to find optimum designs using a constrained generalized reduced gradient (GRG) tech-
nique to identify a local minimum. Commercial packages have been applied to the optimization includ-
ing GINO and LINGO (Brook at al. 1988). Constraints are explicitly included in the formulation of the
nonlinear optimization model. This optimization technique suffers from three disadvantages

1. The solution produced may only be a local optimum. Thus the problem needs to be solved a
number of times with different starting conditions.
2. The sizes of pipe diameters obtained are continuous values and not commercially available
pipe sizes. As mentioned earlier, pipe sizes in the solution must be rounded up or rounded
down to obtain a practical solution. The rounded solution needs to be checked to see if it
satisfies the design criteria. For continuous pipe size diameters as determined by the nonlin-
ear programming optimization process there are 2NP alternatives for rounding up or down to
the nearest commercially available pipe diameter. This rounding process requires consider-
able judgement, particularly for a large water distribution system network as it is not clear
that the solution after rounding will automatically satisfy the minimum allowable pressure
constraints at all nodes. Usually each rounded solution is determined by trial – and – error
and then each rounded solution is checked with a hydraulic solver. Many of the combina-
tions of round up and round down will not satisfy the minimum allowable pressure head
constraints. Thus, the issue of determining the best approach to round up and round down is
an optimization problem in its own right (Dandy et al. 1993). In any case there is no guaran-
tee that the rounded continuous solution is the optimum solution to the discrete pipe prob-
lem.
3. The order of the governing equations and constraints appears to influence the optimum
solution obtained.
4. The fitting of the cost function is difficult when pipe duplications are considered (Dandy et
al. 1993). An abrupt change occurs in the cost function for which it is difficult to obtain a
simple continuous function over the desired range of pipe diameters. The fitting of the cost

508
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

function based on discrete data for commercially available pipe sizes may have an impact
on the optimum solution obtained (Dandy et al. 93).
5. There is a limitation on the number of constraints that may be handled and hence this in
turn limits the size of the water distribution system and/or the number of loading cases that
may be handled (Simpson et al. 94). Adding an additional water demand loading case adds
a significant number of additional constraints for the optimization problem.
Consider a case of a new water distribution system water distribution system where all NP pipe diameter
sizes are to chosen. Thus for each pipe there are both NP pipe diameters and NP pipe flows to be com-
puted. In addition, there are NJ unknown heads at the non – reservoir nodes in the water distribution sys-
tem. Thus the total number of unknowns to be solved for in a nonlinear programming formulation of the
optimization problem is

2  NP + NJ

The requirements are that the flows and heads in the water distribution system must satisfy the hydraulic
constraints including continuity at all junctions and the head loss relationships for each pipe leads to the
number of constraints being equal to

NP + NJ

Walski (1988) suggested that the most successful of these traditional techniques involved “gradient
search” where the partial derivative of cost with respect to some or all of the variables in the problem are
calculated and used to determine a new solution. Simpson et al. (1994) highlighted some of the prob-
lems in using a nonlinear programming approach based on a generalized reduced gradient – technique
(GRG) as

1. Often the solution procedure gets stuck on a single local minimum solution. In fact, the
solution that the technique finds seems to depend on the order of the equations put into the
nonlinear optimization model.
2. Sometimes the method fails to converge on any solution.
3. The technique shows a dependence of the solution found on the order of equations and con-
straints in the optimization model. Different local minima are approached, as a result.
4. There is difficulty in being able to handle more than one steady state water demand loading
case. The inclusion of consideration of an extended period simulation is virtually impossi-
ble without making the size of the optimization problem unmanageable (in terms of a large
number of constraints).
5. Once the solution of the diameters is obtained there is a need to round – up or round – down
“continuous” solutions (e.g. a pipe diameter of 141.23 mm) to commercially available “dis-
crete” sizes. This ends up being a significant optimization problem in its own right.
El – Bahrawy and Smith (1985) applied the nonlinear programming package called MINOS to the
design of water collection and distribution systems. This technique was applied to a number of case
studies (El – Bahrawy and Smith 1985). Their model included a pre – processor to set up the data files
and a post – processor to round off the pipe sizes to commercial sizes. The model for water distribution
systems also included check valves and pressure regulating valves. They demonstrated its ability

• to handle pumps and valves

• to find the optimal location of booster pumps and their optimal lifts
• to address the optimal layout problem
One advantage of nonlinear programming over other methods is that computer run times are relatively
short (Dandy et al. 1993).

509
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

20.22.5 Heuristic methods for optimization

Montbaliu et al. (1990) have proposed a type of gradient search technique to achieve an efficient design.
Initially, all pipes were set at their minimum diameters and a simulation model was used to determine
the pressure at all nodes in the water distribution system. If the minimum pressures were not satisfied
the pipe with maximum head loss per unit length was increased to the next available diameter and a fur-
ther simulation was carried out. The process was repeated until all pressure constraints were satisfied.
Two near – optimal solutions were obtained for some example networks.

20.22.6 Why hasn’t optimization been used in the past?

Walski et al. (1988) suggested the following reasons for the lack of use of optimization in the water
industry

• optimization techniques usually are derived from the field of operations research and as a
result it is usually necessary to simplify the field problem to fit the solution technique
• there has been difficulty in incorporating reliability considerations into the design in an
appropriate manner
• there are generally shortcomings in design criteria
I believe the first point made by Walski is accurate. Traditional optimization techniques (e.g. linear pro-
gramming, nonlinear programming and dynamic programming) are difficult to apply unless an expert –
user is involved, as has been indicated in this Chapter.

I disagree with Walski’s second point above and believe that the issue of optimization may be treated
separately from reliability. The issue of reliability is an important one that must be addressed in relation
to design of all water distribution systems whether optimization is used or not. Reliability constraints or
design criteria may be incorporated into the formulation of the optimization problem but the way in that
reliability is provided or the design criteria associated with reliability in a water distribution system may
be determined separately from consideration of the optimization. Excellent guidance on pipe sizes and
layout, pump sizes and locations and pressure valve locations and pressure setting may be provided by
optimization. Reliability may be built into the system by determining appropriate design criteria (e.g.
providing a reduced flow at a lower than normal minimum pressure if a pipe breaks in the system) or by
providing adequate looping within the system.

Finally, Walski’s last point above is difficult to follow. Any shortcomings in design criteria will be pres-
ent if an engineer uses computer hydraulic simulation and a trial – and – error system improvement pro-
cess to develop lower cost designs or whether optimization is used. It is important to develop the best
possible design criteria. Once the design criteria are specified they are used to guide the trial – and –
error improvement of the design via evaluation of the computer simulation results. These same design
criteria may also be used in the optimization process. If there is uncertainty in data or shortcomings in
design criteria the issues associated with their use in the design process are equally as problematic if
optimization is or is not used. A sensitivity analysis may be needed to assess the effect on the optimiza-
tion results of changes in uncertain data or design criteria.

20.23 Summary
Genetic algorithm optimization for water distribution network design (both for irrigation and urban sys-
tems) is a very powerful technique. Low cost urban water distribution and irrigation pipe network
designs result from the use of genetic algorithms. This optimization technique has significant advan-
tages over the more traditional optimization techniques

• lower cost designs are generated

• designs are based on commercially available pipe sizes and discrete options (e.g. to clean or
not to clean a pipe)
• a range of similarly low cost but perhaps very different configurations are an outcome of
the genetic algorithm procedure
• the full capability of simulation of hydraulic water distribution system is possible including
pumps, pressure regulating valves, tanks and extended period simulation

510
 Chapter 20 - File: Chap20-Optimization.fm July 6, 2022 Version 6

Case studies have shown the effectiveness of the genetic algorithm optimization technique for identify-
ing significant capital cost savings in the design of water distribution systems. Wide spread adoption of
the use of these techniques could result in millions of dollars of savings in water supply and irrigation
infrastructure.

511
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

CHAPTER 21 WATER HAMMER ANALYSIS1

21.1 Concepts of Unsteady Flow

Unsteady flow or water hammer occurs within pipes in a water distribution system. It involves a change
in the flow velocity. A change in a device [such as a valve (opening or closing) or pump (starting up or
shutting down - potentially due to an electricity power failure)] or an accident is usually the cause. Other
examples of changes that may cause water hammer include closure of a reflux or non – return valve on
the discharge side of a pump or a change in demand usage in the network. Examples of accidental occur-
rences that may cause severe water hammer pressure are the rupture of a pipe or the accidental fast clo-
sure of a valve in the water distribution system or single pipe. The changes in fluid velocity in a pipeline
(usually by 1 or 2 m/s) are accompanied by a corresponding change (usually large) in head or pressure in
the fluid; and thus a change in hydraulic grade line elevation (HGL) or head. The transmission of pres-
sure waves through flowing water within pipes in a water distribution system (often at a wave speed des-
ignated as “a” that exceeds 1000 m/s - for a metallic pipe), and as a result of the fluid velocity changes,
is known as water hammer, a transient event or unsteady flow. As mentioned, these water hammer
waves may result in large pressures in the pipes. The elasticity of the fluid reflected by its bulk modulus
of elasticity K of the water and the degree of radial outward movement of the pipe wall (controlled by
Young’s modulus of elasticity E) both influence the speed of travel of water hammer waves through the
fluid along the pipe.

This Chapter does not attempt to present an exhaustive coverage of water hammer analysis. There are
many fine books that already cover water hammer analysis in detail. Examples include Wylie and
Streeter’s 1993 Fluid Transients in Systems and Chaudhry’s 2014 Applied Hydraulic Transients. The
basics of water hammer are presented here in order to provide the water distribution design engineer an
understanding of the underlying principles. Water hammer analysis is an important aspect of the design
of water distribution transmission and distribution systems. Water hammer analysis often governs the
class of pipe, and therefore the pipe wall thickness, that must be selected for the design of particular
pipes in a water distribution system network.

The topics presented in this Chapter are now described. The concept of a water hammer wave traveling
in an infinite fluid leads to the concept of wave speed and the important fluid properties that govern the
value of speed of propagation of pressure waves generated during water hammer events. The Joukowsky
head rise and the sequence of events following the fast closure of a valve in a pipeline system are then
presented. The time during which a valve is closed results in either a rapid or fast valve closure or a
slow valve closure - this delineation depends on the wave return time of 2L/a where L is the length of the
pipeline. These concepts are important in the understanding of water hammer and are described in this
Chapter. The governing equations of unsteady flow in pipelines are presented without a significant
amount of derivation. The unsteady equation of mass and momentum lead to a set of hyperbolic partial
differential equations. These equations are solved using a method of characteristics transformation that
results in four ordinary differential equations. Integration of the compatibility equations along the char-
acteristic lines results in a relatively straight forward set of algebraic equations that may be solved in a
pipeline system to determine the evolution of the pressure as a water hammer event proceeds. The
method of characteristic equations are the basis of a number of commercial computer programs includ-
ing Hytran (Hytran Solutions – Dr. Norman Lawgun, New Zealand), and HCP Software Pty Ltd (Chris
Vigus - Western Australia - originally written by Max Olde (HCP Software Pty Ltd (2019)) of Austra-

1.© Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,
Australia. Part of the book entitled Water Distribution Systems Engineering.

512
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

lia’s WATHAM. Boundary. Other programs are Hammer by Bentley Systems and Surge by
Innovyze.Boundary conditions including reservoirs, dead ends and valves are also described in this
Chapter. The Chapter concludes with a discussion of some typical devices used for the control and miti-
gation of water hammer in piping systems.

21.2 Terms for Water Hammer

A number of different terms are commonly used to refer to water hammer including

• fluid transients
• unsteady flow
• surge
Significant damage may occur as a result of water hammer as demonstrated in Figure 21.1.

Figure 21.1 Hydropower accident water hammer failure Oigawa, Japan (Bonin, 1960)

21.3 Transmission of Water Hammer Waves at a Finite Speed

Consider a reservoir – pipeline – valve system as shown in Figure 21.2. If the valve at the downstream of
the pipeline is at the fully opened position, a steady state flow situation will result. In this case, the HGL
is horizontal as it is assumed that there is no friction loss along the pipeline. Rapid closure of the valve
will result in an almost instantaneous reduction in the velocity of the fluid layers immediately adjacent
to the valve. Due to the fact that both the fluid and the pipe wall are elastic the effect of the valve closure
is not felt immediately at other points within the fluid in the pipeline (except immediately adjacent to the
valve). Successive fluid layers will be stopped as the effect of the rapid valve closure passes in an
upstream direction through the fluid within the pipeline (Lupton 1953). Thus the effect is transmitted
along the pipeline at a finite speed as indicated in Figure 21.4. The stoppage of the layers as it passes up
the pipeline is in effect the movement of the water hammer wave. The speed of travel of the stoppage of
fluid layers within the pipe following the initiation of a water hammer event is referred to as the wave
speed. In an infinite fluid the value of the wave speed is approximately 1485 m/s. This is also the value

513
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

in a perfectly rigid pipe such as a concrete lined tunnel through rock. If the pipe wall is elastic and
deforms as the water hammer wave passes along the pipeline, then the wave speed is reduced. For a steel
pipeline the wave speed drops from the theoretical maximum of 1485 m/s to around 1000 to 1200 m/s
while for a plastic pipe such as uPVC the wave speed may be as low as 300 m/s.

As the velocity of flow changes a corresponding head rise (or head drop — if a valve is suddenly opened)
occurs. Lupton (1953) suggested that the change in pressure or head resulting from the velocity change
persists. In fact, there will be little change in the pressure at a point in the pipeline until a reflected water
hammer wave arrives back at that point. Thus for an instantaneous valve closure as in Figure 21.2 the
head increase will persist at the valve for 2L/a seconds (the wave return time for the reservoir-pipe-valve
system). Water hammer reflections occur when a water hammer wave reaches a reservoir, a junction in a
water distribution network or a dead end.

Figure 21.2 A reservoir – pipeline – valve system for water hammer description

21.4 Fluid Compressibility

The compressibility of a fluid represented by the bulk modulus of elasticity K is very important in water
hammer or unsteady flow in pipelines for situations involving either sudden or great changes in pres-
sure. If the pressure on a surface increases (on say a square box of fluid) from p to p + p the relation-
ship between the change of pressure p and change of volume VM within the square box of fluid
depends on the bulk modulus of elasticity of the material.

The bulk modulus of elasticity is defined as the change in pressure p divided by the volumetric strain
V M  V M from Equation 2.29 as

p p
K = – -------------------------- = -------------- (21.1)
V M  V M   

where

• K = bulk modulus of elasticity (Pa or psi)

• p = change in pressure (Pa or psi)
• V M = change in volume due to the application of the change of pressure (m3 or ft3)
• V M = initial volume before the application of the change of pressure (m3 or ft3)
•  = change in density due to the application of the change of pressure (kg/m3 or slugs/ft3)
•  = initial density before the application of the change of pressure (kg/m3 or slugs/ft3)
• V M  V M = relative change of volume
•    = relative change of density
This definition applies to a gas, a liquid or a solid. The bulk modulus of elasticity K depends on the rela-
tionship between pressure and volume (or density). Since density is affected by temperature then K is
also temperature dependent.

514
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Typical values of bulk modulus of elasticity are

9
• water, K = 2.20 10 N/m2 at 20°C or 320,000 psi at 68°F
9
• oil, K = 1.62 10 N/m2 or 235,000 psi
11 6
• steel, E = 2.05 10 N/m2 or 29.7 10 psi
Thus water is about 93 times more compressible than steel.

21.5 Wave Speed of a Simple Elastic Wave in an Infinite Compress-

ible Fluid
Consider an elastic wave traveling through an infinite fluid at the acoustic wave speed of a with the
property of fluid density o. The initial steady state pressure, velocity and density in the undisturbed
region of the infinite fluid are assumed to be po, Vo and while the values in the zone behind the pres-
sure wave are po + dp, Vo + dV and  + d. For compressible fluids the density  is an additional vari-
able to be solved for.

This unsteady flow situation may be made steady by traveling with the wave at a velocity of a. The rela-
tive velocity in the undisturbed zone of the infinite fluid is now  V o – a  and in the disturbed zone
 V o + dV – a  behind the elastic wave.

Application of the steady state continuity equation across the pressure wave gives

 o A o  V o – a  =   o + d   A o  V o + dV – a   (21.2)

Cancelling Ao on both sides and expanding gives

 o V o –  o a =  o V o +  o dV –  o a + d  V o + dV – a 

Simplifying

0 =  o dV + d  V o + dV – a  (21.3)

Figure 21.3 A simple elastic wave traveling in an infinite fluid

515
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Usually the wave speed ( a  1000 m/s ) is much greater than the initial steady state fluid velocity
(Vo = 1 to 2 m/s) such that

a >>  V o + dV  (21.4)

This simplifies the continuity equation relating the change in velocity to the change in density as

dV = ad
---------- (21.5)
o

Recalling the bulk modulus of elasticity for a fluid in terms of the density change is

p -
K = ------------- (21.6)
  

Combining these two equations gives an expression for the wave speed in an infinite fluid

a = K dV
------- (21.7)
dp

It will be shown later that the relation (the Joukowsky head or pressure head rise) between the change in
pressure d p d and the change in velocity dV due to the passage of the water hammer wave is

d p d = adV (21.8)

or

dV 1
---------- = ------ (21.9)
d pd a

Now combining these two equations in Equation 21.7 and Equation 21.9 gives

K
a = ---- (21.10)


WAVE SPEED IN AN INFINITE FLUID

where

• a = wave speed (m/s or ft/s)

• K = bulk modulus of elasticity (Pa or psi)
•  = density of the infinite fluid (kg/m3 or slugs/ft3)
This value is the upper limit of the wave speed in a fluid in a pipeline (in which the walls must be
infinitely rigid). Usually the wave speed in a thin walled pipeline will be less that this value.

21.5.1 Example of wave speed computation in an infinite fluid

The following shows an example for the computation of the wave speed in an infinite body of water.

2.2 10 N/m2 and the fluid density of

9
Example 21.1 For water the bulk modulus of elasticity is K = 2.2 GPa =
water is 998.2 kg/m3 at 20°C. Compute the wave speed.

Answer – Computation of the wave speed in an infinite fluid.

516
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

The wave speed in an infinite fluid is

9
K- = 2.2 10 - = 1485 m/s
a = --- ------------------
 998.2

This would be the maximum wave speed that would be expected to be found in a pipeline filled with water.

Example 21.2 For water the bulk modulus of elasticity is K = 320,000 psi (lbf/in2) and the fluid density of water is 
1.938 slugs/ft3 at 68°F (from Table 2.3). Compute the wave speed.

Answer – Computation of the wave speed in an infinite fluid.

The wave speed in an infinite fluid is

slug.ft
1 --------------- -
2 2 2
K 320000 lbf/in s 144 in
a = ---- = ------------------------------------  ----------------------  ------------------ = 4876 ft/s
 1.938 slugs/ft
3 1 lbf 1 ft
2

where the relationship between slugs (mass in US Customary units) and lbf (pound-force) from Newton’s second law
of motion F = ma is

slug.ft
1 lbf = 1 ----------------
2
s

Recalling from Chapter 2 that a slug is a unit of mass in US customary units that accelerates by 1 ft/s2 when a force of
one pound (1 lbf) is exerted on it.

_______________________________________________________________________________

21.6 The Joukowsky Head Change — Basic Water Hammer Equation

Before introducing the Joukowsky head change equation, the concept of the wave return travel time for
a reservoir-pipeline-valve system will be considered.

21.6.1 Time of travel for a pipeline

The concept of the water hammer wave return travel time that was mentioned earlier in this Chapter is
now dealt with in detail. The time of travel of a water hammer wave from one end of a pipe at the valve
(assumed to be at the downstream end of the pipeline where the water hammer event is initiated) to the
reservoir (at the upstream end of the pipeline as shown in Figure 21.4) is

--L- seconds
a

where

• L = pipe length (m or ft)

• a = wave speed (m or ft) - usually this is taken as the wave speed in the pipeline being con-
sidered (thus Equation 21.48 or Equation 21.52) rather than the wave speed in an infinite
fluid (Equation 21.10).
Thus the time of travel for a water hammer wave from the valve to the reservoir and then back to the
valve for the first time is

2L
------ seconds (21.11)
a

WATER HAMMER WAVE RETURN TRAVEL TIME

517
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.4 Joukowsky's head rise upstream of a valve

Once the water hammer wave arrives at the valve for the first time - pressure relief occurs and the high
pressure (due to the Joukowsky head rise) drops down to a lower value. Thus for the Joukowsky head
rise equation it is assumed that the valve is closed in a time of less than 2L  a seconds.

21.6.2 Rapid valve closure

When the time of closure of the valve is less than the water hammer return travel time of 2L  a as given
by Equation 21.11 this is referred to as a rapid valve closure. At the valve, the full Joukowsky head rise
will be experienced after the valve is fully closed.

21.6.3 Slow valve closure

When the time of closure of the valve is greater than the water hammer return travel time of 2L  a as
given by Equation 21.11 this is referred to as a slow valve closure. For this case, the maximum head in
the pipeline due to the valve closure will be less than the full Joukowsky head rise.

21.6.4 The Joukowsky head change at the upstream side of a valve

A valve at the downstream end of a pipeline in Figure 21.4 is closed rapidly such that the flow velocity
is reduced from the initial steady state value of Vo to zero. Assume that the friction loss along the pipe-
line is zero (thus the HGL is horizontal) and also that minor losses are ignored.

Following the valve closure, the head change in the pipe just upstream of a valve is given by

H = – --a- V = – --a-  V f – V i  (21.12)

g g

JOUKOWSKY HEAD CHANGE ON UPSTREAM SIDE OF THE VALVE

where

• H = change of head associated with the velocity change V (m or ft)

• a = wave speed for the fluid in the pipeline (m/s or ft/s) - (thus Equation 21.48 or
Equation 21.52).
• g = gravitational acceleration (m2/s or ft2/s)
• V f = final velocity of flow (m/s or ft/s)
• V i = initial velocity of flow before initiation of water hammer event (m/s or ft/s)

The magnitude of the head change H does not depend on the type of fluid flowing, however, the actual
pressure change will obviously depend on the fluid as follows

p = H (21.13)

518
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

where

•  = g is the specific weight of the fluid flowing (N/m3 or lbf/ft3)

A more rigorous derivation of the Joukowsky equation will be presented later on.

An example of the application of the Joukowsky equation follows.

Example 21.3 Determine the Joukowsky water hammer head change for an initial steady state velocity of flow
Vo = 1.0 m/s reduced to zero almost instantaneously. Assume the pipe is steel.

Answer – Computation of the Joukowsky water hammer head change.

A typical wave speed for water in a steel pipe is a = 1000 m/s to 1300 m/s. Use 1300 m/s.

Thus the Joukowsky head change from Equation 21.12 is

 H = – --a- V = – --a-  V f – V i  = – 1300

------------  0 – 1.0  = 132.5 m
g g 9.81

___________________________________________________________________________________

In Example 21.3, the change in velocity is defined as the final velocity after the valve has been moved to
its final position minus the initial velocity that existed prior to the valve being moved

V =  V f – V i  (21.14)

Thus in Example 21.3, it follows that the change in velocity for the water hammer event of rapidly clos-
ing the valve at the downstream end of the pipe is minus 1 m/s and thus

V =  V f – V i  =  0.0 – 1.0 = – 1.0 m/s

21.6.5 The Joukowsky head change at the downstream side of a valve

Now consider the situation where the valve is moved from the downstream end of the pipe to the mid-
point of the pipeline as shown in Figure 21.5. A sudden valve closure for this case causes the head to
drop on the downstream side of the valve (a head rise still occurs on the upstream side of the valve).
Thus the Joukowsky head change in this case in the fluid downstream of the valve is

H = --a- V = --a-  V f – V i  (21.15)

g g

JOUKOWSKY HEAD CHANGE ON DOWNSTREAM SIDE OF THE VALVE

where

• H = change of head associated with the velocity change V (m or ft)

• a = wave speed for the fluid in the pipeline (m/s or ft/s) - (thus Equation 21.48 or
Equation 21.52).
• g = gravitational acceleration (m2/s or ft2/s)
• V f = final velocity of flow (m/s or ft/s)
• V i = initial velocity of flow before initiation of water hammer event (m/s or ft/s)

Note the minus sign has now been omitted from the right – hand side of Equation 21.15.

Returning to the example above it follows that:

519
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Example 21.4 Determine the Joukowsky water hammer head drop for an initial steady state velocity of flow
Vo = 1.0 m/s in Figure 21.5 that reduces to zero for the fluid just downstream of a valve. Assume the pipe is made of
steel.

Answer – Computation of the Joukowsky water hammer head drop.

As for Example 21.3, a typical wave speed for water in a steel pipe of 1300 m/s will be used.

Thus the Joukowsky head drop from Equation 21.15 is

1300
 H = --a- V = --a-  V f – V i  = ------------  0 – 1.0  = – 132.5 m
g g 9.81
This assumes the pressure of the fluid remains above the vapor pressure of the fluid as indicated in Figure 21.5.

Figure 21.5 Joukowsky's head drop downstream of a valve

___________________________________________________________________________________

21.7 Sequence of Events After Sudden Closure of a Valve

21.7.1 Steady state flow with an initial steady state velocity of Vo
Consider a reservoir – pipeline – valve frictionless system in which the minor losses are ignored. The
steady state HGL is horizontal and is as shown in Figure 21.6.

A head loss of Hss occurs across the valve such that

2
Vo
H ss = K --------
- = Ho (21.16)
2g

for a frictionless pipe where

• K = the head loss coefficient for the valve (rather than the bulk modulus of elasticity — also
referred to by K) (dimensionless)
• Vo = the initial steady state velocity (m/s or ft/s)
• g = gravitational acceleration (m/s2 or ft/s2)
• Ho = the reservoir HGL (m or ft)

21.7.2 Immediately following the valve closure

Immediately following the instantaneous closure of the valve, the velocity of the fluid layer immediately
adjacent to the valve is reduced from Vo to zero as shown in Figure 21.7.

520
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.6 Steady state HGL for a reservoir – pipeline – valve system

The fluid layer adjacent to the valve that has had its velocity reduced to zero results in the kinetic energy
of the flowing water being converted into the elastic work. The pressure in the stopped fluid layer
increases. The elastic work goes to the stretching the conduit wall and compressing the water in the fluid
layer (thus increasing the density of the fluid). The fluid is brought to rest by the impulse of the higher
pressure developed at the face of the valve. Once the first fluid layer is brought to rest, the same action is
applied to the next fluid layer. The successive stopping of these fluid layers is actually the propagation
of the water hammer in the upstream direction. Thus a water hammer pressure wave or a positive wave
travels upstream at a propagation wave speed of a as shown in Figure 21.7. The pressure rise at the
water hammer wave applies an impulse to the fluid that is just sufficient to bring the fluid to rest.

21.7.3 The first L/a seconds, 0 < t < L/a

A water hammer wave takes L/a seconds to travel from the valve to the reservoir as shown in
Figure 21.7. Following the instantaneous closure of the valve as shown in Figure 21.7(a), the water
hammer wave travels away from the valve. As the water hammer wave moves towards the reservoir,
fluid continues to enter the pipeline from the reservoir. It is this fluid that supplies the additional fluid
into the high pressure zone (behind the water hammer wave) that is stored in the expanded pipe section
and in the compressed fluid. Figure 21.7(b) shows the pipeline at a time of  -seconds after the valve
closure. Figure 21.7(c) shows the pipeline at a time of L  2a -seconds after the valve closure as the
water hammer wave arrives at the mid-point of the pipeline for the first time. Figure 21.7(d) shows the
pipeline at a time of L  a –  -seconds after the valve closure as the water hammer approaches the reser-
voir for the first time.

The quantity of additional water that enters the pipe during the first L/a seconds (in which the water
hammer wave travels for the valve to the reservoir a shown in Figure 21.7(e)) is

V o AL
-------------- (21.17)
a

21.7.4 At L/a seconds the positive water hammer wave reaches the
reservoir
When the wave reaches the reservoir after a time of t = L/a seconds as shown in Figure 21.7(e) then

• all the fluid is at rest V = 0 along the entire pipeline

• all the fluid in pipeline is subject to an additional head given by the Joukowsky formula as

H = --a- V o (21.18)
g

• all the momentum has been lost

• the fluid along the entire pipe is compressed to greater than the normal density
• all the kinetic energy has been converted into elastic energy of the expanded pipe wall and
the compressed fluid

521
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

There is an unbalanced head condition at the reservoir just as the positive pressure wave arrives at the
reservoir. The head or pressure in the reservoir remains unchanged with an HGL of Ho as the water ham-
mer reflects from the reservoir.

The HGL in the pipe as the water hammer wave arrives at the reservoir at L/a seconds is at an elevated
level of

H o + H = H o + --a- V o (21.19)
g

The unbalanced head condition between the inside of the pipe and the reservoir at L/a results in a flow
backwards from the reservoir into the pipe with a velocity of – Vo. The resulting backwards flow into the
reservoir returns the head or pressure in the pipeline to the level of the reservoir head of Ho in the fluid
layer adjacent to the reservoir. Thus the head drop is

H = --a-  – V o – 0  = – --a- V o (21.20)

g g

A water hammer wave now travels downstream from the reservoir towards the valve. s the water ham-
mer now moves away from the reservoir (after L/a seconds), the fluid layer just behind the water ham-
mer wave the pipe wall contracts to normal size and the fluid expands to normal density. A reflected
negative wave now travels downstream towards the valve. Water is discharging out of the pipe back into
the reservoir.

21.7.5 The second L/a seconds, L/a < t < 2L/a

The reduction in pressure and initiation of a velocity Vo in the negative direction travels progressively
downstream as a water hammer wave from the reservoir towards the valve as shown in Figure 21.8. The
water expelled in the upstream direction due to the return of the pipe to normal size and expansion of the
fluid to the normal density level serves to maintain the negative velocity on the upstream side of the neg-
ative wave as it proceeds downstream at acoustic speed (Rich 1951, p. 3). In other words, water empties
upstream from the shrinkage space (Rich 1951, p. 2).

Figure 21.8(f) shows the pipeline at a time of L  a +  -seconds after the valve closure with the water
hammer wave traveling from the reservoir back towards the valve. Figure 21.8(g) shows the pipeline at
a time of 3L  2a -seconds after the valve closure as the water hammer wave arrives at the mid-point of
the pipeline for the second time. Figure 21.8(h) shows the pipeline at a time of 2L  a –  -seconds after
the valve closure as the water hammer approaches the valve for the first time.

522
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.7 Water hammer waves following instantaneous valve closure [0 < t < L/a]

21.7.6 At 2 L/a seconds the reflected negative wave reaches the valve
When the wave reaches the valve after a time of t = 2L/a seconds as shown in Figure 21.8(i) then

• the fluid velocity in the pipeline is – Vo towards the reservoir along the entire pipe
• the head along the entire pipeline is normal at Ho
• the conduit is of normal size
• the fluid is at normal density

523
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Since the valve is closed when the reflected water hammer wave arrives at 2L/a seconds there then addi-
tional fluid is not available at the closed valve to maintain the negative velocity – Vo towards the reser-
voir. As a result the velocity instantaneously reduces from – Vo to zero in association with a pressure
drop. This occurs in order to stop the flow from detaching from the valve. It is assumed that the head at
the valve does not drop below the vapor pressure head of the fluid at the valve (usually about 10 m or 33
ft below the centerline elevation of the pipe).

For velocity to become zero in the fluid layer just adjacent to the valve as the water hammer wave
reflects from the valve, the head at the valve must drop below the reservoir level by an amount of aVo/g.
Thus an impulse is applied to the fluid due to the pressure drop that just reduces the velocity from – Vo
to zero. The kinetic energy of the reverse flow is converted into negative elastic work of sucking down
the size of the conduit below the initial diameter at the section immediately adjacent to the valve. A
reflected negative water hammer wave now begins to travel from the closed valve towards the reservoir.
The conduit shell behind the negative water hammer wave now contracts to less than the normal size.
The water behind the negative water hammer wave expands to less than the normal density.

21.7.7 The third L/a seconds, 2L/a < t < 3L/a

The additional water expelled from behind the negative water hammer wave due to the contraction of
the pipe below the normal size and the expansion of the fluid to below the normal density level serves to
maintain the negative velocity upstream and ahead of the negative wave front as shown in
Figure 21.9(j). The water hammer proceeds upstream in the pipe at the acoustic speed (Rich 1951, p. 3).

Figure 21.8(j) shows the pipeline at a time of 2L  a +  -seconds after the valve closure with the water
hammer wave traveling from the reservoir back towards the reservoir. Figure 21.8(k) shows the pipeline
at a time of 5L  2a -seconds after the valve closure as the water hammer wave arrives at the mid-point
of the pipeline for the third time. Figure 21.8(l) shows the pipeline at a time of 3L  a –  -seconds after
the valve closure as the water hammer approaches the reservoir for the second time.

21.7.8 At 3L/a seconds the negative wave reaches the reservoir

When the negative wave reaches the reservoir after a time of t = 3L/a seconds as shown in
Figure 21.9(m) then

• the fluid velocity is zero along the entire pipe

• the head along the entire pipeline is at a depressed level of

H o –  H = H o – --a- V o (21.21)
g

that is below the normal reservoir level

• the entire conduit is contracted below the normal level
• the fluid along the entire conduit is expanded to be at less than the normal density due to the
reduced pressure
• water has been entirely emptied from the shrinkage space back into the reservoir
Considering the upstream end of the pipeline just as the negative water hammer wave arrives at the res-
ervoir as shown in Figure 21.9(m), there is an unbalanced head condition between the reservoir and the
pipeline. The head in the reservoir remains unchanged with an HGL of Ho. The head in the pipeline is
lower than that in the reservoir. The higher head in the reservoir or unbalanced condition results in a
flow into the pipe from the reservoir with a velocity of Vo. This flow from the reservoir returns the pres-
sure in the pipeline to the reservoir pressure in the fluid layer adjacent to the reservoir.

a a a
H = --- V = ---  V o – 0  = --- V o (21.22)
g g g

A reflected positive wave travels back towards the valve.

524
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.8 Propagation of water hammer waves in the second L/a period [L/a < t < 2L/a]

21.7.9 The fourth L/a seconds, 3L/a < t < 4L/a

After the reflection of the water hammer wave from the reservoir at 3L  a -seconds, the increase in pres-
sure in the pipeline behind the wave (from the depressed level back to the reservoir level) and the initia-
tion of a velocity Vo (from a value of zero) in the pipe towards the valve travels progressively moves
downstream as a water hammer wave as shown in Figure 21.10(n). The fluid layer behind the passing
water hammer wave contracts to normal density. The pipe wall behind the passing water hammer wave
expands or stretches back to normal. The water entering the pipe from the reservoir fills the volume
resulting from the pipe expanding to normal size and contributes fluid to the increase the fluid density to
the normal level behind the positive reflected wave as it proceeds downstream with acoustic speed.

525
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.8(n) shows the pipeline at a time of 3L  a +  -seconds after the valve closure with the water
hammer wave traveling from the reservoir back towards the valve for the second time. Figure 21.8(o)
shows the pipeline at a time of  7L   2a -seconds after the valve closure as the water hammer wave
arrives at the mid-point of the pipeline for the fourth time. Figure 21.8(p) shows the pipeline at a time of
4L  a –  -seconds after the valve closure as the water hammer approaches the valve for the second
time.

Figure 21.9 Propagation of water hammer waves in the third L/a period [2L/a < t < 3L/a]

21.7.10 At 4L/a seconds the positive reflected wave reaches the valve
When the water hammer wave reaches the valve after a time of t = 4L/a seconds as shown in
Figure 21.10(q) then the following is true in the pipe

• the fluid velocity is Vo towards the valve along the entire pipe
• the head along the entire pipeline is the normal reservoir level of Ho
• the fluid density is normal
• the conduit size is normal

526
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

The conditions are now identical to those in the pipe at time of zero immediately following the instanta-
neous closure of the valve. The cycle has been completed. This cycle will repeat itself indefinitely every
4L/a seconds.

For a real pipeline, friction and both an inelastic fluid and pipe wall material dissipates energy as heat
and gradually damps down the oscillation to quiescence of zero velocity and normal reservoir head for
the entire pipe.

Figure 21.10 Propagation of water hammer in the fourth L/a period [3L/a < t < 4L/a]

21.8 The Joukowsky Head Rise Equation from the Unsteady

Momentum Equation
The resultant force acting on a control volume is equal to the time rate of change of increase of linear
momentum within the control volume plus the net efflux of linear momentum from the control volume
(Streeter et al. 1998).

d 
 F = d t  mv  =  t  v dV +  v  v d A (21.23)
cv cs

527
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

where

• F = the resultant force on the control volume (N or lbf)

• m = mass in control volume (kg or slugs)
• = the fluid density (kg/m3 or slugs/ft3)
• v = the velocity vector for the center of mass (m) of the system (m/s or ft/s)
• dV = an elemental volume (m3 or ft3)
• dA = a vector representing an area element on the control surface (m2 or ft2)
• cs is the control surface
• cv is the control volume
Now consider application of the force – momentum flux equation in Equation 21.23 to the unsteady
pipeline flow situation in Figure 21.11. The mass of fluid in the control volume having its momentum
changed is

m = A  a – V o t (21.24)

where

• V o = initial steady state velocity before the water hammer event

• a = water hammer wave speed
The dominant forces are the pressure forces. Application of the unsteady momentum equation for the x –
direction to the control volume in Figure 21.11 gives

A  a – V o t  V o + V – V o  2
p s A –  p s + p d  A = -------------------------------------------------------------------------
- + A  V o + V  – AV o (21.25)
t

where

• ps = steady state pressure on the upstream face of the control volume

• p d = the pressure rise due to the water hammer event
• A = area on the upstream face of the control volume
• t = time of travel of the water hammer wave in the control volume
• V = velocity change in the fluid
• a – Vo = absolute velocity of the water hammer wave
Simplifying Equation 21.25 gives

2 2 2
– p d A = A  a – V o   V  + A  V o + 2V o V +  V   – AV o (21.26)

2 2 2
– p d =   aV – V o V + V o + 2V o V +  V  – V o  (21.27)

2
– p d =   aV + V o V +  V   (21.28)

– p d = V  a + V o + V  (21.29)

528
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.11 Control volume for the unsteady momentum equation

Consider the case of a water hammer event in a pipeline where the velocity change is caused by instan-
taneous closure of the valve from an initial steady state velocity of V o to zero such that

V = V f – V i = – V o (21.30)

with variables as defined in Equation 21.12.

Substituting from Equation 21.30 into Equation 21.29 gives

2 2
– p d =   – a V o –  V o  +  – V o   (21.31)

Thus the dynamic pressure increase (or Joukowsky pressure increase) above the steady state pressure ps
is

p d = aV o (21.32)

The change or increase in head or hydraulic grade line is related to the change in pipeline pressure pd as

p p
H = ---------d- = ---------d- (21.33)
 g

Combining the Equation 21.32 and Equation 21.33 leads to the basic water hammer or Joukowsky head
change equation

aV
H = ---------o- (21.34)
g

or

H = – aV
----------- (21.35)
g

JOUKOWSKY HEAD CHANGE ON UPSTREAM SIDE OF THE VALVE

Recall that this equation applies for no friction and no reflections of water hammer waves that have
returned to the point in the pipeline under consideration.

21.8.1 For a partial valve closure where the velocity change is not V = – Vo
From Equation 21.29

529
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

– p d = V  a + V o + V  (21.36)

Usually in water hammer flow the wave speed a is much larger than the initial steady state fluid velocity
V o in the pipeline such that

a >>  V o + V 

Thus applying this approximation to Equation 21.36 gives

– p d = Va (21.37)

or

aV
H = – ----------- (21.38)
g

which again leads to the Joukowsky head change equation for a partial or full valve closure.

21.9 Wave Speed for the Fluid in a Pipeline

As a high pressure water hammer wave moves along the pipe, the pipe wall is stretched due to the pres-
sure change thereby resulting in a lateral or circumferential strain. The elasticity of the pipeline wall
reduces the wave speed compared to the wave speed in either an infinite fluid or a completely inelastic
rock tunnel (where a = 1485 m/s).

Application of the continuity equation to the fluid flow situation during the first L/a seconds following
the instantaneous closure of the valve and combining with the Joukowsky head rise equation gives

2 gH
a = --------------------- (21.39)
A 
------- + -------
A 

Recalling the bulk modulus of elasticity of Equation 21.6 as

p gH
K = -------------- = ---------------
     

or

K 
gH = ----  ------- (21.40)
  

Substituting Equation 21.40 into Equation 21.39 gives

K-  
--- -------
2   
a = --------------------- (21.41)
A 
------- + -------
A 

or

530
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

K
----
2 
a = ------------------------- (21.42)
A 
------- ------- + 1
A 

Also from the bulk modulus of elasticity of Equation 21.6 gives

- = ------
K-
------ (21.43)
 p

Substituting Equation 21.43 into Equation 21.42 and rearranging terms gives

K
----
2 
a = -------------------------------- (21.44)
1 A
1 + K  --- -------
 A p 

K-
---

a = -------------------------------- (21.45)
1 A
1 + K  --- -------
 A p 

where

• K = bulk modulus of elasticity (N/m2 or lbf/ft2 - usually converted from psi)

• = fluid density (kg/m3 or slugs/ft3)
• A = area of the pipe flow (m2 or ft2)
• A = change in area of the pipe flow area for an increase of pressure p (m2 or ft2)
• p = change in pressure (N/m2 or lbf/ft2)
Equation 21.45 may now be compared to the wave speed for an infinite fluid from Equation 21.10 or

K
a = ---- (21.46)


Thus the wave speed in a pipe conduit will be smaller than the wave speed in an infinite fluid as the term
1 A
K  --- ------- will be greater than zero.
 A p 

The term in the denominator of Equation 21.45 represents the amount of change of area of the pipe for a
change of internal pressure

A
-------
p

Thus the larger the value of the area A for a given change in pressure p due to a water hammer event
1 A
the bigger the term K  --- ------- will be, and consequently the smaller the magnitude of the wave speed a.
 A p 
Imagine a water hammer traveling in a rubber hose. The amount of area change will be relatively large
compared with say a steel pipe (for the same pressure increase due to a water hammer event) and thus
the wave speed in a rubber hose is considerably slower.

531
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

21.9.1 Wave speed for a thin – walled pipe

Consider a pipeline of diameter D and wall thickness e. A thin walled pipe is defined when the follow-
ing condition holds

D-  25
---
e

The change in area for a pressure change inside the pipe depends on the restraint condition along the
length of the pipeline. In Equation 21.44 the change in area to change in pressure ratio may be expressed
as

1 A 1D
--- ------- = --- ---- c 1
A p Ee

where

• c1 depends on the restraint condition as given in Table 21.1

Thus Equation 21.44 becomes

K K
---- ----
2  
a = ------------------------------------- = ------------------------ (21.47)
1D K- ---
D- c
1 + K  --- ---- c 1  1 + ---
Ee  Ee 1

K-
---

a = ------------------------ (21.48)
KD
1 + ---- ---- c 1
Ee

WATER HAMMER WAVE SPEED FOR A THIN WALLED PIPE

where

• K = bulk modulus of elasticity (N/m2 or lbf/ft2 - psi converted)

• = fluid density (kg/m3 or slugs/ft3)
• E = Young’s modulus of elasticity (N/m2 or lbf/ft2 - usually converted from psi)
• D = the internal diameter of the pipe (m or ft) (sometimes specified in mm or in.)
• e = the pipe wall thickness (m or ft) (sometimes specified in mm or in.)
• c1 = parameter reflecting how the pipeline is restrained along its length (dimensionless)
(see Table 21.1)
• = Poisson’s ratio for the pipe wall material (dimensionless)

Table 21.1 Variation of c1 with pipe restraint condition

Pipe restraint condition c1
Anchored at upstream end 
1 – ---
2
Anchored throughout entire length 2
1–
Expansion joints throughout length 1.0

532
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

In Table 21.1,  = Poisson’s ratio. Poisson’s ratio is the ratio between the contraction at right angles to a
stress and the direct extension. Poisson’s ratio for steel is in the order of 0.3.

Example 21.5 Consider the following pipeline with an internal diameter of 752.5 mm and a pipe wall thickness of
7.94 mm. Assume the pipe has expansion joints throughout. For water assume the bulk modulus of elasticity is
9 2 3
K = 2.2 GPa = 2.2 10 N/m , the fluid density of water is 998.2 kg/m at 20°C, while Young’s modulus of
steel E = 210 GPa = 210 10 N/m2. Compute the wave speed.
9

Answer – Computation of the wave speed in a thin walled steel pipeline.

First check that the pipe is thin – walled. Compute:

D 752.5
---- = ------------- = 94.8 > 25
e 7.94

Thus the pipe is thin – walled. Because the pipeline has expansion joints throughout the c 1 value is 1.0.

Thus from Equation 21.48 the wave speed in a thin – walled pipeline is

9
K- 2.2 10 -
--- ------------------
 998.2 2203967.1 1484.6
a = ------------------------- = --------------------------------------------------------- = ------------------------------------------------- = ------------------------------
KD 2.2 10 752.5
9 1 +  0.01048   94.8  1 + 0.9935
1 + ---- ---- c 1 1 + --------------------- -------------  1.0 
Ee
210 10 7.94
9

Note the numerator of the last term on the right is the wave speed in an infinite fluid. Thus the wave speed is

a = 1051.5 m/s

___________________________________________________________________________________

21.9.2 Wave speed for a cement mortar lined pipe

If a pipeline has a cement mortar lining on the inside of a metal or steel, ductile iron (DICL) or mild
steel (MSCL) pipe then an equivalent thickness of metal represented by the cement mortar lining needs
to be computed as:

E
t eq = t c  -----c- (21.49)
Es

where

• t eq = equivalent thickness of steel or metal (m or ft)

• tc = thickness of the cement mortar lining (m or ft)
• Ec = Young’s modulus of the cement mortar lining (typically 25-28 GPa or 4,000,000 psi)
• Es = Young’s modulus of the steel or metal pipe wall (typically 207-210 GPa or 30,000,000
psi)
Thus the combined steel  t s  (or metal) and cement mortar lined  t eq  equivalent thickness of steel (or
metal) is

t s – eq = t s + t eq (21.50)

where

• t Seq = composite equivalent thickness of steel or metal pipe wall plus the cement mortar
lining (m or ft)
• ts = thickness of the steel or metal wall of the pipeline (m or ft)

533
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

• t eq = equivalent thickness of steel or metal (m or ft)

Consider the following examples.

Example 21.6 Consider the following pipeline with an internal diameter of 752.5 mm and a pipe wall thickness of
7.94 mm (the identical internal diameter and steel pipe wall thickness as in Example 21.5) that has a cement mortar
lining thickness of 12.0 mm. Assume the pipe has expansion joints throughout. For water assume the bulk modulus
of elasticity is K = 2.2 GPa = 2.2 10 N/m2, the fluid density of water is 998.2 kg/m3 at 20°C, while Young’s
9

modulus of steel E = 210 GPa = 210 10 N/m2. Compute the wave speed.
9

Answer – Computation of the wave speed in a thin walled pipeline with a cement mortar lining.

First determine the equivalent thickness of steel as represented by the cement mortar lining thickness from
Equation 21.49 as:

E 25
t eq = t c  ------c = 12.0  --------- = 12  0.12  = 1.44 mm
Es 210

Thus the equivalent thickness of steel incorporating the cement mortar lining thickness from Equation 21.50 is

t Seq = 7.94 + 1.44 = 9.38 mm

Now check that the pipe is thin – walled. Compute:

D- = 752.5
--- ------------- = 80.2 > 25
e 9.38

Thus again as for the previous example the pipe is thin – walled. Because the pipeline has expansion joints throughout
the c 1 value is 1.0.

The wave speed in a thin – walled pipeline with a cement mortar lining from Equation 21.48 using t Seq instead of e
for the pipe wall thickness is

9
K 2.2 10
---- -------------------
 998.2 2203967.1 1484.6 -
a = ---------------------------------- = --------------------------------------------------------- = ------------------------------------------------
- = -----------------------------
K D -c 2.2 10 - 752.5
9 1 +  0.01048   80.2  1 + 0.8405
1 + ---- ------------- 1 + -------------------- -------------  1.0 
E  t Seq  1
210 10 9.38
9

Note the numerator of the last term on the right is the wave speed in an infinite fluid. Thus the wave speed for a cement
mortar lined steel pipe is

a = 1094.4 m/s

Thus compared to Example 21.5 where the steel pipe is unlined, the wave speed increases from a = 1051.5 m/s to
a = 1094.4 m/s by an amount of 4.1% due to the presence of the 12 mm thickness of the cement mortar lining.

Now let us assume the cement mortar lining is now placed on inside the pipe in Example 21.5 thereby
reducing its diameter.
_
Example 21.7 Consider the following pipeline with an internal diameter of 728.5 mm (this is 24 mm less that the
pipe in Example 21.5 due to the presence of the cement mortar lining thickness) and a pipe wall thickness of 7.94
mm that has a cement mortar lining thickness of 12.0 mm. Assume the pipe has expansion joints throughout. For
9 2
water assume the bulk modulus of elasticity is K = 2.2 GPa = 2.2 10 N/m , the fluid density of water is
998.2 kg/m3 at 20°C, while Young’s modulus of steel E = 210 GPa = 210 10 N/m2. Compute the wave speed.
9

Answer – Computation of the wave speed in a thin walled pipeline with a cement mortar lining.

First determine the equivalent thickness of steel as represented by the cement mortar lining thickness from
Equation 21.49 as:

E 25
t eq = t c  ------c = 12.0  --------- = 12  0.12  = 1.44 mm
Es 210

Thus the equivalent thickness of steel incorporating the cement mortar lining thickness from Equation 21.50 is

534
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

t Seq = 7.94 + 1.44 = 9.38 mm

Now check that the pipe is thin – walled. Compute:

D- = 728.5
--- ------------- = 77.66 > 25
e 9.38

Thus again as for the previous example the pipe is thin – walled. Because the pipeline has expansion joints throughout
the c 1 value is 1.0.

The wave speed in a thin – walled pipeline with a cement mortar lining from Equation 21.48 using t Seq instead of e is

9
K 2.2 10
---- -------------------
 998.2 2203967.1 1484.6
a = ------------------------------------ = --------------------------------------------------------- = ---------------------------------------------------- = ------------------------------
K  D  2.2 10 728.5
9 1 +  0.01048   77.66  1 + 0.8139
1 + ---- -------- - c 1 + --------------------- -------------  1.0 
E t  1
210 10 9.38
Seq 9

Note the numerator of the last term on the right is the wave speed in an infinite fluid. Thus the wave speed is

a = 1102.3 m/s

Thus compared to Example 21.5 where there is no cement mortar lining, the wave speed increases from
a = 1051.5 m/s to a = 1102.3 m/s by an amount of 4.9 % due to the presence of the 12 mm thickness of the cement
mortar lining and also because the internal diameter is reduced by 24 mm due to the thickness of the pipe wall lining.

This example is a more realistic comparison than Example 21.6 due to the internal diameter of the steel pipe alone
being kept at the same value.

___________________________________________________________________________________

21.9.3 Wave speed for a thick – walled pipe

A thick walled pipe is defined for a pipe that has D/e ratio that is approximately less than 25. The stress
is not uniformly distributed throughout the walls when they are thick (Wylie and Streeter 1993). For the
case of a pipeline with expansion joints throughout its length then to account for the pipe wall being
thick the value of c1 becomes:

2e D
c 1 = ------  1 +   + ------------- (21.51)
D D+e

Thus the wave speed for a thick walled pipe becomes from Equation 21.48:

K
----

a = --------------------------------------------------------------------- (21.52)
K- ---
D- 2e D -
1 + --- ------  1 +   + ------------
Ee D D+e

WATER HAMMER WAVE SPEED FOR A THICK WALLED PIPE

where

• a = wave speed in the fluid (m/s or ft/s)

• K = bulk modulus of elasticity (N/m2 or lbf/ft2 - converted from psi)
• = fluid density (kg/m3 or slugs/ft3)
• E = Young’s modulus of elasticity (N/m2or lbf/ft2 - converted from psi)
• D = internal diameter of the pipe (m or ft)
• e = pipe wall thickness (m or ft)
• = Poisson’s ratio for the pipe wall material (dimensionless)

535
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Example 21.8 Consider the following pipeline with an internal diameter of 752.5 mm and a pipe wall thickness of
40.0 mm. Assume the pipe has expansion joints throughout. For water assume the bulk modulus of elasticity is
9 2 3
K = 2.2 GPa = 2.2 10 N/m , the fluid density of water is 998.2 kg/m at 20°C, while Young’s modulus of
steel E = 210 GPa = 210 10 N/m2 and the Poisson’s ratio for steel is  = 0.3 . Compute the wave speed.
9

Answer – Computation of the wave speed in a thick walled pipeline.

Check whether the pipe is thin – walled or thick walled. Compute:

D 752.5
---- = ------------- = 18.8 < 25
e 40

Thus the pipe in this case is thick walled rather than thin walled.

Thus from Equation 21.52 the wave speed in a thick – walled pipeline fluid is

9
K- 2.2 10 -
--- ------------------
 998.2
a = ---------------------------------------------------------------------- = ------------------------------------------------------------------------------------------------------------------------
K- ---
D- 2e D - 9
1 + --- ------  1 +   + ------------ 2.2 10 752.5 2  40  752.5
1 + --------------------- ------------- --------------  1 + 0.3  + -------------------------
Ee D D+e
210 10 40 752.5 752.5 + 40
9

Thus

2203967.1 2203967.1 1484.6

a = ------------------------------------------------------------------------------------------------------ = --------------------------------------------------------------- = ------------------------------
1 +  0.01048 18.8  0.1063  1.3  + 0.9495  1 +  0.01048 18.8  1.087  1 + 0.2142

Note the numerator of the last term on the right is the wave speed in an infinite fluid. Thus

a = 1347.3 m/s

Thus compared to Example 21.5, the wave speed increases from a = 1051.5 m/s to a = 1347.3 m/s by an amount
of 28.1 % due to the increase of pipe wall thickness from 7.94 mm to 40.0 mm.

___________________________________________________________________________________

21.9.4 Wave speed for a completely rigid pipe

To confirm the limits of the equation developed in the section above for the wave speed in the pipeline
(Equation 21.44) consider a rigid pipeline and then a very flexible pipeline. For a rigid pipeline the
change in area as a water hammer travels along the pipeline will be very small as

A  0 (21.53)

Equation 21.45 becomes

K K-
---- ---
  K
a = ----------------------------------- = ---------------------- = ---- (21.54)
1 A 1 + K 0 
1 + K  --- ------- 
 A p 

Thus:

K
a = ---- (21.55)


WATER HAMMER WAVE SPEED FOR A COMPLETELY RIGID PIPE

where

• a = wave speed in the fluid (m/s or ft/s)

536
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

• K = bulk modulus of elasticity (N/m2 or lbf/ft2 - converted from psi)

• = fluid density (kg/m3 or slugs/ft3)

Example 21.9 Consider the following pipeline with an internal diameter of 752.5 mm that is completely rigid. For
water assume the bulk modulus of elasticity is K = 2.2 GPa = 2.2 10 N/m2 and the fluid density of water is
9

998.2 kg/m3 at 20°C. Compute the wave speed.

Answer – Computation of the wave speed in a completely rigid pipeline.

From Equation 21.55 the wave speed is:

9
K 2.2 10
a = ---- = ------------------- = 1484.6 m/s
 998.2

Note the wave speed here is the same for that in an infinite fluid. In addition, it is independent of the pipe material.

___________________________________________________________________________________

21.9.5 Wave speed for a highly flexible pipe

For a highly flexible pipeline the change in area term in the denominator of Equation 21.44 dominates
the other term as

1 A
K --- ------- » 1 (21.56)
A p

Equation 21.44 becomes

K- K-
--- ---
  Ap-
a= --------------------------------  ---------------------- = ---------- (21.57)
 1 A   1 A  A
1 + K --- ------- K --- -------
 A p   A p 

Thus Equation 21.44 for a highly flexible pipeline reduces to:

Ap
a = ----------- (21.58)
A

WATER HAMMER WAVE SPEED FOR A HIGHLY FLEXIBLE PIPE

where

• a = wave speed in the fluid (m/s or ft/s)

• A = area of the pipe flow (m2 or ft2)
• p = change in pressure ((N/m2 or lbf/ft2)
• = fluid density (kg/m3 or slugs/ft3)
• A = change in area of the pipe flow area for an increase of pressure p (m2 or ft2)
Thus the wave speed is not a function of the bulk modulus of elasticity K of the fluid

21.10 Hoop Stress and Strain in a Thin Walled Pipe

Consider the free body diagram of a semi-cylindrical portion of a pipe of unit length with internal diam-
eter D as shown in the Figure 21.12. The axial direction along the pipe (into the page in Figure 21.12) is
designated by the subscript x. The circumferential direction is designated by the subscript 

537
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.12 Forces and stresses in a pipe wall due to water hammer

Assume the pressure in a pipeline is increased by p due to the passage of a water hammer wave. The
increase in the resultant force F on the inside of a unit length of the pipe resulting from the pressure
change (p or pd) is

F = pA (21.59)

Now substituting for the area for a 1 m length of pipe and using p = H

F = pD  1  = HD (21.60)

Let the increase in circumferential or lateral tensile force per unit length in the pipe wall be T. Thus

F = 2T (21.61)

Thus

F pD
T = -------- = ----------- (21.62)
2 2

The increase in circumferential normal stress in the pipe wall is

T pD
  = ---------- = ----------- (21.63)
e1 2e

Thus:

pD
  = ----------- (21.64)
2e

CHANGE IN STRESS IN PIPE WALL

where

•   = change in normal stress due to a change of pressure in the pipe (N/m2 or lbf/ft2)
• p = change in pressure (N/m2 or lbf/ft2) - and also equivalent to a change in transient
pressure head H (see Figure 21.12)
• D = internal diameter of the pipe (m or ft)
• e = thickness of the pipe wall (m or ft)
21.11 Design Pressure Level
The design pressure level or DPL is the maximum permissible pressure (or head) for a pipeline that it is
designed to withstand. It is the sum of the pumping head, water hammer pressures and a factor of safety.
For a section of main that has both a constant pipe diameter and wall thickness, the DPL is often taken
as the maximum pressure at the worst point in the pipeline profile. This is generally where the maximum

538
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

elevation difference occurs between the ground surface and the maximum hydraulic grade line. The
DPL is often taken to be a horizontal – although in reality there may be points in the pipeline where the
ground surface rises up and the maximum possible pressure does not exceed the worst possible pressure
in the pipeline.

21.12 Unsteady Continuity Equation for Pipe Flow

21.12.1 Assumptions
The following assumptions are made in deriving the unsteady continuity and unsteady momentum equa-
tions

• one – dimensional flow is assumed. Thus it follows that

• a common pressure is assumed at a pipeline section (usually taken to be the center-
line of the pipe)
• the hydrostatic variation across a vertical section of the pipeline is ignored
• an average velocity is assumed for each pipe section
• the pipe is full of fluid and remains full during the transient
• there is no column separation during the transient (thus pressures stay above vapor pressure
everywhere along the pipeline)
• the pipe wall and fluid are assumed to behave linearly elastically (i.e. stress is proportional
to strain)
• the pipe may be non – prismatic (non – constant diameter)
• unsteady viscous losses may be approximated by steady state friction losses
21.12.2 The continuity equation
From Reynolds transport theorem the conservation of mass is

0 =    dV +  v  d A (21.65)
 t cv cs

where the variables in the equation are as for Equation 21.23.

The first term in Equation 21.65 is the change of mass in the control volume while the second term is the
net efflux of mass across the control surface.

Consider the control volume in Figure 21.13.

Figure 21.13 Control volume for the unsteady flow continuity equation

Application of Equation 21.65 to a control volume of length x gives

539
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

0 =   A x +   AV x (21.66)

t x

Expanding, introducing the wave speed and changing from pressure to HGL leads to

2
H H a V
+V – V sin  + ----- = 0 (21.67)
t x g x

UNSTEADY CONTINUITY EQUATION FOR PIPE FLOW

where

• H = head (a dependent variable) (m or ft)

• t = time (an independent variable) (s)
• V = velocity (a dependent variable) (m/s or ft/s)
• x = distance along the pipe (an independent variable) (m or ft)
•  = angle of the slope of the pipe to the horizontal (degrees)
• a = wave speed (m/s or ft/s)
• g = gravitational acceleration (m/s2) = 9.81 m/s2 or 32.2 ft/s2
Equation 21.67 is simplified by

• assuming the Vsin term is small and therefore dropping this term out
• assuming the relative size of the following terms is such that

H >> V H
t x

Thus the simplified form of Equation 21.67 is

2
H a V
+ ----- = 0 (21.68)
t g x

SIMPLIFIED UNSTEADY CONTINUITY EQUATION

21.13 Unsteady Momentum Equation for Pipe Flow

Consider the control volume in Figure 21.14.

540
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.14 Control volume for the unsteady momentum equation

In Equation 21.23 (from the Reynolds theorem) the main forces involved include pressure forces, grav-
ity and shear resistance at the boundary. The two types of pressure forces that are considered are the sur-
face contact normal pressures and the pressure component at the periphery or annular space due to the
diverging pipeline. It is also assumed that unsteady friction may be adequately modeled by steady state
friction.

The resulting governing partial differential equation is given by

V V H f V V
+V +g + ------------------ = 0 (21.69)
t x x 2D

UNSTEADY MOMENTUM EQUATION FOR PIPE FLOW

where

• V = velocity (a dependent variable) (m/s or ft/s)

• t = time (an independent variable) (s)
• H = head (a dependent variable) (m or ft)
• x = distance along the pipe (an independent variable) (m or ft)
• g = gravitational acceleration (m/s2) = 9.81 m/s2 or 32.2 ft/s2
• f = Darcy-Weisbach friction factor (dimensionless)
• D = internal diameter of the pipe (m or ft)
Equation 21.69 is simplified by

• Assuming the relative size of the following term (the convective acceleration) is such that

V V
>> V
t x

541
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Thus the simplified form of Equation 21.69 is

V + g H + -----------------
f V V- = 0 (21.70)
t x 2D

SIMPLIFIED UNSTEADY MOMENTUM EQUATION

The two partial differential equations (PDEs) of Equation 21.68 and Equation 21.70 form a set of
quasi – linear hyperbolic PDEs. The two dependent variables are velocity V (we will replace this by the
discharge Q later on in the Chapter) and hydraulic grade line H or V (x, t) and H(x, t). The independent
variables are distance along the pipeline x and time t.

A summary of the two unsteady flow or water hammer partial differential equations is as follows:

2
H a V
+ ----- =0
t g x
V H f V V
+g + ------------------ = 0
x x 2D

UNSTEADY CONTINUITY AND MOMENTUM EQUATIONS

where

• H = head (a dependent variable) (m or ft)

• V = velocity (a dependent variable) (m/s or ft/s)
• t = time (an independent variable) (s)
• x = distance along the pipe (an independent variable) (m or ft)
• a = wave speed (m/s or ft/s)
• g = gravitational acceleration (m/s2) = 9.81 m/s2 or 32.2 ft/s2
• f = Darcy-Weisbach friction factor (dimensionless)
• D = diameter of the pipe (m or ft)
21.14 The Method of Characteristics Transformation
The set of partial differential equations governing unsteady flow behavior in pipelines does not have a
general solution. A method of characteristics transformation may be performed to convert the two par-
tial differential equations to four ordinary differential equations. The four ordinary differential equations
are treated as two pairs of equations. Each pair of ordinary differential equations is linked and must
always be considered together.

An unknown multiplier method may be applied to the two partial differential equations of
Equation 21.68 and Equation 21.70. Let the simplified unsteady continuity equation be represented as

2
H a V
L2 = + ----- (21.71)
t g x

Let the simplified unsteady momentum equation be represented as

V H f V V
L1 = +g + ------------------ (21.72)
t x 2D

Consider a linear combination of L1 and L2 as

L = L1 +  L2 (21.73)

542
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Selection of the following value of  as

g
 =  ---- (21.74)
a-

leads to the following four ordinary differential equations in two pairs.

+
The C pair of ordinary differential equations include the compatibility equation

g dH + dV + --------------
+ ----
fV V
- = 0 (21.75)
a- d t dt 2D

+
that is valid along the C characteristic line of

dx
= a (21.76)
dt

_
The C pair of ordinary differential equations include the compatibility equation

fV V
g dH + dV + ----------------
– ---- - = 0 (21.77)
a- d t dt 2D

_
that is valid along the C characteristic line of

dx
= –a (21.78)
dt

where in Equation 21.75 to Equation 21.78

• g = gravitational acceleration (m/s2) = 9.81 m/s2 or 32.2 ft/s2

• a = wave speed (m/s or ft/s)
• H = head (a dependent variable) (m or ft)
• t = time (an independent variable) (s)
• V = velocity (a dependent variable) (m/s or ft/s)
• f = Darcy-Weisbach friction factor (dimensionless)
• D = diameter of the pipe (m or ft)
• x = distance along the pipe (an independent variable) (m or ft)
21.15 Integration of the MOC Compatibility Equations
Each of the compatibility ordinary differential equations is integrated along its corresponding character-
istic line. The two equations representing characteristics lines are depicted on an x – t plot as shown in
+
Figure 21.15. Line AP in Figure 21.15 is the C characteristic line given by Equation 21.79. Line BP is
_
the C characteristic line in Figure 21.15. The pipe in Figure 21.15 has been divided into 2 reaches
(Reach 1 is between the reservoir and the mid-point of the pipeline and Reach 2 between the mid-point
of the pipeline and the valve). that correspond to 3 computational sections at the reservoir, at the mid –
point of the pipeline and at the valve.

21.15.1 The characteristic grid on the x – t plane

The characteristic grid on the x – t plane (as shown in Figure 21.15) may be formed from the discretiza-
tion of the ordinary differential equation representing the characteristic

543
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

dx
= a (21.79)
dt

Now in discrete form:

x
------ = a (21.80)
t

DISCRETIZATION OF THE CHARACTERISTIC LINE

where

• x = reach length between two computational sections (m or ft)

• t = time step for transient computations (s)
• a = wave speed (m/s or ft/s)
There is always a defined relationship between the reach length (the distance between computational
steps in a pipeline) and the computational time step. Thus once a reach length is selected for the pipeline
in Equation 21.80, usually in such a way as to have an even number of reaches, the corresponding time
step is

x
t = ------ (21.81)
a

Figure 21.15 Characteristic lines on an x – t plot and computation of interior point at P

on the x-t plane at L/2a seconds

Note the use of the cross-over grid in Figure 21.15 where every section (even numbered and odd num-
bered sections) are computed at each time step. This is in contrast to the use of a diamond grid where
even and odd reaches are computed sequentially at each alternative time step.

21.15.2 Integrating the C+ compatibility equation

+
Consider the C compatibility equation. In Figure 21.15, integrate the compatibility equation between
point A and point P both on the x-t plane. The velocity and HGL are assumed to be known at point A and
unknown at point P.

544
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

H t
g P dH + V P dV + P --------------
fV V
+ ----
-
a H    - dt = 0 (21.82)
A V t A 2D
A

f t P V V dt = 0
g  H – H  +  V – V  + -------
+ ---- (21.83)
2D t
a - P A P A
A

where

• g = gravitational acceleration (m/s2) = 9.81 m/s2 or 32.2 ft/s2

• a = wave speed (m/s or ft/s)
• H P = unknown transient head or HGL at point P in Figure 21.15 (m or ft)
• H A = known transient head or HGL at point A in Figure 21.15 (m or ft) at a time t earlier
than where H P is being computed
• V P = unknown velocity at point P in Figure 21.15 (m or ft)
• V A = known velocity at point A in Figure 21.15 (m or ft) at a time t earlier than where
V P is being computed
• f = Darcy-Weisbach friction factor (dimensionless)
• D = internal diameter of the pipe (m or ft)
• t P = time associated with Point P (s)
• t A = time associated with Point A (s)
• V = velocity along the characteristic line between Point A and Point P (m/s or ft/s)
• t = time (s)
Converting to flow (from velocity) and using an integrating by parts approach for the last term gives

a fx
 H P – H A  + -------  Q P – Q A  + ------------------ Q P Q A = 0 (21.84)
gA 2gD A
2

where

• H P = unknown transient head or HGL at point P in Figure 21.15 (m or ft)

• H A = known transient head or HGL at point A in Figure 21.15 (m or ft) a time t earlier
than where H P is being computed
• a = wave speed (m/s or ft/s)
• g = gravitational acceleration (m/s2) = 9.81 m/s2 or 32.2 ft/s2
• A = area of the pipe (m2 or ft2)

• Q P = unknown discharge or flow at point P in Figure 21.15 (m3/s or ft3/s)

• Q A = known discharge at point A in Figure 21.15 (m or ft) a time t earlier than where
Q P is being computed
• f = Darcy-Weisbach friction factor (dimensionless)
• x = reach length (m or ft)
• D = internal diameter of the pipe (m or ft)

545
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

The characteristic impedance (B) for a pipeline is defined as

a
B = ------- (21.85)
gA

CHARACTERISTIC IMPEDANCE

where

• B = characteristic impedance (s/m2 or s/ft2)

• a = wave speed (m/s or ft/s)
• g = gravitational acceleration (m/s2) = 9.81 m/s2 or 32.2 ft/s2
• A = area of the pipe (m2 or ft2)
The units of B are

m
----
s s
 B  = ---------------------- = ------
m 2 2
-----  m  m
2
s

The reach resistance term (R) is defined as

fx
R = ------------------ (21.86)
2
2gD A

PIPE REACH RESISTANCE

• R = reach resistance (units are given in Equation 21.87)

• f = Darcy-Weisbach friction factor (dimensionless)
• x = reach length (m or ft)
• g = gravitational acceleration (m/s2) = 9.81 m/s2 or 32.2 ft/s2
• D = internal diameter of the pipe (m or ft)
• A = area of the pipe (m2 or ft2)
The units of R are

2
 -   m  - = ------ s
R = ------------------------------- (21.87)
m 2 3
-----  m   m  m
2
s

Substituting Equation 21.85 and Equation 21.86 into Equation 21.84 gives

H P = H A – B  Q P – Q A  – RQ P Q A (21.88)

H P =  H A + BQ A  –  B + R Q A Q P (21.89)

Recall that point P on the x-t plane is the current time step of L/2a seconds with the values of QP and HP
that are unknown and point A is the previous time step of zero seconds at which values of QA and HA are
known. Assign values in Equation 21.89 that are at the previous time step of point A where conditions
are assumed to be known as

546
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

C pc = H A + BQ A (21.90)

B pc = B + R Q A (21.91)

where

• the subscript pc refers to “positive characteristic”

Thus the outcome of the integration of the compatibility equation along the characteristic equation con-
verts an ordinary differential equation into a simple linear algebraic expression

H P = C pc – B pc Q P (21.92)

POSITIVE COMPATIBILITY EQUATION

where in Equation 21.90 to Equation 21.93

• C pc = “C” constant for positive characteristic line between Point A and Point P
• H A = known transient head or HGL at point A in Figure 21.15 (m or ft) at a time t earlier
than where H P is being computed
• B = characteristic impedance (s/m2 or s/ft2)
• Q A = known discharge at point A in Figure 21.15 (m or ft) at a time t earlier than where
Q P is being computed
• B pc = “B” constant for positive characteristic line between Point A and Point P
• R = reach resistance (units are given in Equation 21.87)
• H P = unknown transient head or HGL at point P in Figure 21.15 (m or ft)
• Q P = unknown discharge or flow at point P in Figure 21.15 (m3/s or ft3/s)

21.15.3 Integrating the C – compatibility equation

_ _ _
The C pair of equations is now considered. The C compatibility equation is integrated along the C
characteristic line that is represented in discretized form as

x
------ = – a (21.93)
t

_ _
Integrating the C compatibility equation along the C characteristic line between points B and P both
on the x-t plane in Figure 21.15 results in

t f
g H P dH + V P dV + P ----------------
– ----
V V
a H
-   2D - dt = 0 (21.94)
B V B t B

Integrating and multiplying by minus one gives

g f tP
+ -----  H P – H B  –  V P – V B  – -------  V V dt = 0 (21.95)
a 2D t B

Converting to flow (from velocity) and using an integrating by parts approach for the last term gives

547
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

a fx
 H P – H B  – -------  Q P – Q B  – ------------------ Q P Q B = 0 (21.96)
gA 2gD A
2

Substituting Equation 21.85 and Equation 21.86 for the characteristic impedance B and the pipe reach
resistance R into Equation 21.96

H P = H B + B  Q P – Q B  + RQ P Q B (21.97)

H P =  H B – B Q B  +  B + R Q B Q P (21.98)

Recall that point P on the x-t plane is the current time step of L/2a seconds with the values of QP and HP
that are unknown and point B is the previous time step of zero seconds at which values of QB and HB are
known. Assign values in Equation 21.98 that are at the previous time step of point B again where condi-
tions are assumed to be known as

C mc = H B – BQ B (21.99)

B mc = B + R Q B (21.100)

where

• the subscript mc refers to “minus characteristic”

Thus the outcome of the integration of the compatibility equation along the characteristic equation con-
verts an ordinary differential equation into a simple linear algebraic expression

H P = C mc + B mc Q P (21.101)

NEGATIVE COMPATIBILITY EQUATION

where in Equation 21.99 to Equation 21.101

• C mc = “C” constant for negative characteristic line between Point B and Point P
• H B = known transient head or HGL at point B in Figure 21.15 (m or ft) a time t earlier
than where H P is being computed
• B = characteristic impedance (s/m2 or s/ft2)
• Q B = known discharge at point B in Figure 21.15 (m or ft) at a time t earlier than where
Q P is being computed
• B mc = “B” constant for negative characteristic line between Point B and Point P
• R = reach resistance (units are given in Equation 21.87)
• H P = unknown transient head or HGL at point P in Figure 21.15 (m or ft)
• Q P = unknown discharge or flow at point P in Figure 21.15 (m3/s or ft3/s)

548
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Thus an interior computational section is governed by the two integrated compatibility equations of
Equation 21.92 and Equation 21.101. Solving for the unknown head HP between these two equations
give

C pc B mc + C mc B pc
H P = ---------------------------------------------- (21.102)
B pc + B mc

HEAD AT INTERIOR PIPELINE SECTION

and solving for QP between these two equations of Equation 21.92 and Equation 21.101 gives

C pc – C mc
Q P = -------------------------
- (21.103)
B pc + B mc

DISCHARGE AT INTERIOR PIPELINE SECTION

where in Equation 21.102 to Equation 21.103

• H P = unknown transient head or HGL at point P in Figure 21.15 (m or ft)

• C pc = “C” constant for positive characteristic line between Point A and Point P where
C pc = H A + BQ A
• B mc = “B” constant for negative characteristic line between Point B and Point P where
B mc = B + R Q B
• C mc = “C” constant for negative characteristic line between Point B and Point P where
C mc = H B – BQ B
• B pc = “B” constant for positive characteristic line between Point A and Point P where
B pc = B + R Q A
• Q P = unknown discharge or flow at point P in Figure 21.15 (m3/s or ft3/s)

An example of the calculation of the head at the mid – point of a pipe due to the arrival of a water ham-
mer generated by rapid closure of a valve at the downstream end of the pipe is given in Example 21.10.
See Figure 21.7 (c) for details of arrival of the water hammer wave at the mid – point of the pipeline.

21.15.4 Computational examples

A number of examples of MOC computations will be given in the remainder of the Chapter.

These include (and are also shown in Figure 21.16):

1. Computation at an interior node at the mid-point of the pipeline at Point P2 on the x-t plane
at time L/2a seconds (Section 21.15.5 and Example 21.10). The interior nodes are usually
computed prior to the boundary conditions. This is why Point P2 is presented first before
Point P1.
2. Computation in the pipeline just adjacent to the downstream boundary condition of a closed
valve at Point P1 on the x-t plane at time zero seconds (Section 21.16.2 and
Example 21.11). A significant increase in head occurs at the valve due to the instantaneous
closure of the valve.
3. Computation in the pipeline just adjacent to the upstream boundary condition of a reservoir
at Point P3 on the x-t plane at time L/2a seconds (Section 21.7.2 and Example 21.12). This
point is still at the initial steady state conditions so no change in head or discharge occurs.
The method of characteristics computations verify that this is the case.

549
Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

4. Computation in the pipeline just adjacent to the downstream boundary condition of a closed
valve at Point P4 on the x-t plane at time L/2a seconds (Section 21.18.1 and
Example 21.13).
5. Computation in the pipeline just adjacent to the upstream boundary condition of a reservoir
at Point P5 on the x-t plane at time L/a seconds (Section 21.18.2 and Example 21.14). Flow
reversal occurs at the reservoir as the water hammer wave reflects.
6. Computation in the pipeline just adjacent to the downstream boundary condition of a closed
valve at Point P6 on the x-t plane at time L/a seconds (Section 21.18.2 and Example 21.15).
7. Computation at an interior node at the mid-point of the pipeline at Point P7 on the x-t plane
at time 3L/2a seconds (Section 21.19.1 and Example 21.16).
8. Computation in the pipeline just adjacent to the downstream boundary condition of a closed
valve at Point P8 on the x-t plane at time L/a seconds (Section 21.19.2 and Example 21.17).
The head at the valve drops significantly when the water hammer wave reflects from the
closed valve for the first time.

Figure 21.16 Sequence of method of characteristics calculations in Examples below

550
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

21.15.5 An MOC example calculating conditions at an interior computational

section at L/2a seconds (Point P2 on the x-t plane)

Example 21.10 Consider a reservoir – pipe – valve system in Figure 21.18 with the following characteristics

• reservoir elevation HR = 200 m,

• pipe length L = 980 m,
• pipe diameter D = 0.5 m,
• Darcy – Weisbach friction factor f = 0.02,
• initial velocity V = 1.0 m/s,
• wave speed a = 980 m/s,
• gravitational acceleration g = 9.81 m/s2.
The valve at the downstream end is closed instantaneously at a time of zero seconds and generates a head rise to an
HGL of 298.90 m (see Example 21.11 for the computation of HGL at the valve [Point P2 on the x-t plane] immedi-
ately following valve closure).

Compute the head and flow at the mid-point of the pipeline (Point P2 on the x-t plane) at L/2a seconds.

Answer – Computation of the head and flow at the mid-point of the pipeline at L/2a seconds

Consider the x – t plane in Figure 21.18.

Computing transient conditions at section 2 at the mid – point of pipe (Point P2 on the x-t plane) at L/2a seconds

+
The conditions at the mid – point of the pipeline P2 at time t = L/2a must be determined from the C characteristic line
_
between Points A2 and P2 and the C characteristic line between Points B2 and P2 as shown in Figure 21.18, all these
points being on the x-t plane. Conditions for head and discharge are needed at Points A and B both on the x-t plane in
Figure 21.18 at the previous time step of zero seconds. The usage of the number 2 in A2, B2 and P2 refers to the second
method of characteristics point computation that is carried out and does not refer to the section numbers at ends of
reaches along the pipeline. The first calculation is actually for Point A1 on the x-t plane as shown in Example 21.11 (as
Point P1 on the x-t plane).

Two reaches are selected for the method of characteristics calculations. Thus the reach length is one – half the length of
the pipeline

x = --L- = 980
--------- = 490 m (21.104)
2 2

Corresponding to the two reaches in the pipeline there are three computational sections — one at the entrance just inside
the pipeline (section 1), one at the mid – point of the pipeline (section 2) and one just upstream of the valve at the
downstream end of the pipeline (section 3).

The time step corresponding to the selected reach length of 490 m must now come from the equation for the character-
istic line (Equation 21.81) in discrete form or

x = 490
t = ------ --------- = 0.5 s (21.105)
a 980

Steady state conditions in the pipeline prior to the closure of the valve at zero seconds

The area of the pipeline is

2 2
A =  D- =   0.5  - = 0.1964 m 2
--------- ----------------- (21.106)
4 4

The flow along the entire pipeline, before the valve closure, prior to a time of zero seconds is constant and at steady
state conditions. Thus the flow at the mid – point of the pipeline (section 2) is

3
Q 2 = VA = 1.0  0.1964  = 0.1964 m /s

This is the flow at the mid – point of the pipeline until the arrival of the water hammer wave at L/2a seconds.

The HGL at the reservoir at section 1 at time t = 0 seconds is

H 1 = 200.0 m

551
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.17 Steady state hydraulic grade line just before the rapid valve closure

The initial steady state hydraulic grade line is given by the head loss from the Darcy – Weisbach equation
(Equation 4.3)

2 2
fL- V = 0.02  980   1.0  =
h f = ----- ------ ------------------------ ------------------ 1.998 m (21.107)
D 2g  0.5  2  9.81 

The initial steady state HGL at section 2 (the mid – point of the pipeline – see Figure 21.17) before the arrival of the
water hammer wave at L/2a seconds is

1.998
H 2 = 200.0 – ------------- = 199.001 m (21.108)
2

The HGL at section 3 (just upstream of the valve at the downstream end – see Figure 21.17) prior to the closure of the
valve at zero seconds is

H 3 = 200.0 – 1.998 = 198.002 m (21.109)

A summary of conditions at the three sections along the pipeline at just prior to zero seconds is given in Table 21.2.

Conditions at section 1 – at the reservoir (Point A2 on the x-t plane) – steady state conditions at zero seconds

The HGL at the reservoir at section 1 or point A2 on the x-t plane at time t = 0 seconds is

H A2 = 200.0 m

The flow in the pipeline at section 1 or point A2 on the x-t plane at time t = 0 seconds is

3
Q A2 = 0.1964 m /s

Conditions at section 3 – at the closed valve (Point B2 on the x-t plane) at zero seconds

+
Conditions at point B2 on the x-t plane at time t = 0 seconds immediately after the closure of the valve are taken from
Example 21.11 (Point P1 on the x-t plane) as

3
H B2 = 298.90 m and Q B2 = 0.0 m /s

Table 21.2 Initial steady state conditions in pipeline just prior to valve closure

Location At reservoir At mid – point At valve

Section 1 2 3
Time (s) 0 0 0
Time (L/a units) 0 L/a 0 L/a 0 L/a
HGL (m) 200.0 199.001 198.002
Condition Steady state conditions. Steady state conditions. Steady state conditions.

Flow Q (m3/s) 0.1964 0.1964 0.1964

Velocity V (m/s) 1.0 1.0 1.0

552
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.18 Calculation of head and flow at the mid-point of the pipe at P2
on the x-t plane at L/2a seconds

Method of characteristics calculations at Point P2 on the x-t plane (or at section 2) at a time of L/2a seconds

The transient was generated by the closure of the valve at the downstream end of the pipeline at zero seconds. In the
first L/2a seconds the water hammer wave has traveled from the valve to the mid – point of the pipeline (Point P2 on
the x-t plane) where it generates a significant head rise and a corresponding velocity reduction.

The water hammer HGL and the discharge at the mid – point are given from the method of characteristics equations of
Equation 21.102 and Equation 21.103 respectively as

C pc B mc + C mc B pc C pc – C mc
H P2 = --------------------------------------------
- and Q P2 = ------------------------
-
B pc + B mc B pc + B mc

The characteristic impedance for the pipeline is

a- = -------------------------------
980
B = ------ - = 508.65 (21.110)
gA 9.81  0.1964 

The reach resistance is

0.02  490 
fx - = -------------------------------------------------------
R = ----------------- = 25.90 (21.111)
2 2
2 gD A 2  9.81   0.5   0.1964 

Note that R depends on the reach length  x. Thus if the number of computational reaches is increased the R value will
decrease.

Compute the values of the “constants” of C pc and B pc for the previous time step from Equation 21.90 and
Equation 21.91 as

C pc = H A2 + BQ A2 = 200.0 + 508.65  0.1964  = 299.90

B pc = B + R Q A2 = 508.65 + 25.90 0.1964 = 513.74

Now compute the values of the “constants” of C mc and B mc for the previous time step from Equation 21.99 and
Equation 21.100 as

C mc = H B2 – B Q B2 = 298.90 – 508.65  0.0  = 298.90

B mc = B + R Q B2 = 508.65 + 25.90 0.0 = 508.65

Thus the water hammer HGL at the mid-point of the pipeline at L/2a seconds is

553
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

C pc B mc + C mc B pc 299.90  508.65  + 298.90  513.74 

H P2 = --------------------------------------------
- = ----------------------------------------------------------------------------------- = 299.40 m
B pc + B mc 513.74 + 508.65

The corresponding transient flow at the mid – point at L/2a seconds corresponding to the arrival of the water hammer
wave from Equation 21.103 is

C pc – C mc 299.90 – 298.90 3
Q P2 = ------------------------
- = --------------------------------------- = 0.00098 m /s
B pc + B mc 513.74 + 508.65

Table 21.3 Transient conditions at mid-point of the pipeline at L/2a seconds

after the valve closure (Point P2 on the x-t plane)

Location At reservoir At mid – point (Point P2) At valve

Section 1 2 3
Previous HGL (m) at zero sec 199.001
Time (s) 0.5 0.5 0.5
Time (L/a units) L/2a L/2a L/2a
HGL (m) 200.0 299.40 298.90
Head rise H (m) Has not 100.40 None since last calculation
changed. +
time step of 0 seconds
Steady state
[see Example 21.11 (Point
conditions.
P1 on the x-t plane)]

Flow Q (m3/s) 0.1964 0.00098 0.0

Previous V (m/s) at zero sec 1.0

Velocity V (m/s) 1.0 0.005 0.0
Velocity change V (m/s) Has not – 0.995 None since last calculation
changed. +
time step of 0 seconds
Steady state
[Example 21.11 (Point
conditions.
P1on the x-t plane)]

Note this value is not zero, as one would expect.

A summary of conditions at the three computational sections along the pipeline at L/2a seconds is given in Table 21.3.
Values computed for this time step are shown in bold. The final results are also shown in Figure 21.19.

Figure 21.19 Summary of method of characteristics results for calculation at the mid – point of the pipeline
at Point P2 on the x-t plane at L/2a seconds

___________________________________________________________________________________

554
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

21.16 The Closed Valve or Dead End Boundary Condition

21.16.1 The governing equations
Consider the reservoir – pipe – valve system in Figure 21.20. In the method of characteristics solution
method, conditions at the valve (section 3 or Point P on the x-t plane) at a time of zero seconds are influ-
enced by the conditions at the computational section immediately upstream (section 2 or Point A on the
x-t plane) at a time of minus L/2a seconds at the mid – point of the pipeline.

Figure 21.20 Closed valve boundary condition at Point P on the x – t plane

+
The C integrated compatibility equation between point A and P both on the x-t plane in Figure 21.20
+
from Equation 21.92 is used. It is valid along the C characteristic line.

H P = C pc – B pc Q P (21.112)

where the “constants” for the previous time step at point A at L/2a seconds are

C pc = H A + BQ A and B pc = B + R Q A

with the characteristic impedance and the reach resistance as

a fx
B = ------- and R = ------------------
gA 2gD A
2

_
There are two unknowns of HP and QP in Equation 21.112. There is no C equation available on the
downstream side of the closed valve. In actual fact, one of the two unknowns is known with the valve
being fully closed (as long as the pressure in the pipe is above the vapor pressure head) and hence

Q P = 0.0 (21.113)

555
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Substituting Equation 21.113 into Equation 21.112 gives an expression to solve for the HGL at the valve
at section 3 based on conditions at the previous time step at the section immediately upstream

H P = C pc (21.114)

HEAD FOR A CLOSED VALVE BOUNDARY CONDITION

where

• H P = unknown transient head or HGL at point P in Figure 21.20 (m or ft)

• C pc = “C” constant for positive characteristic line between Point A and Point P where
C pc = H A + BQ A

21.16.2 An MOC example for the head rise due to instantaneous valve clo-
sure at zero seconds (Point P1 on the x-t plane)
Consider an example to illustrate the head rise at the closed valve immediately after the valve is closed.
See Figure 21.7 (b) for details of generation of the water hammer wave at the valve following instanta-
neous closure.

Example 21.11 Consider a reservoir – pipe – valve system in Figure 21.21 with the same characteristics as the sys-
tem described in Example 21.10: reservoir elevation HR = 200 m, pipe length L = 980 m, pipe diameter D = 0.5 m,
Darcy – Weisbach friction factor f = 0.02, initial velocity V = 1.0 m/s, wave speed a = 980 m/s and gravitational
acceleration g = 9.81 m/s2.
Compute the hydraulic grade line at the valve (Point P1 on the x-t plane) due to instantaneous closure that occurs at
+
zero seconds (t = 0 ). Compare the methods of characteristics head rise with the Joukowsky head rise.

+
Answer – Computation of the HGL at the valve immediately following instantaneous closure at 0 seconds.

Consider the x – t plane in Figure 21.21.

Computing transient conditions at section 1 at the valve (Point P1 on the x-t plane)

The conditions at point P1 on the x-t plane immediately after the valve closure must be determined from the character-
istic line between A1 and P1 in Figure 21.21. The usage of the number 1 in A1 and P1 refers to the first method of
characteristics point computation that is carried out and does not refer to the section numbers at ends of reaches along
the pipeline. The characteristic line between A1 and P1 (both on the x-t plane) actually reaches backwards to section 2
or point A1 at time – L/2a seconds where conditions in the pipeline are at steady state — in fact, identical to conditions
along the entire pipeline at a time of t = 0. Steady state conditions for the pipeline prior to the closure of the valve are
given in Table 21.2.

The transient is generated by the closure of the valve at the downstream end of the pipeline at zero seconds. As for
Example 21.10 the reach length x is 490 m or one – half the length of the pipeline as computed in Equation 21.104.
The corresponding time step from the characteristic equation t is 0.5 seconds from Equation 21.105.

556
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.21 x – t plane for instantaneous valve closure

2
From Example 21.10 the area ( A = 0.1964 m ), characteristic impedance ( B = 508.65 ), reach resistance
( R = 25.90 ) for the pipe are given by Equation 21.106, Equation 21.110 and Equation 21.111 respectively.

Conditions at section 2 at the mid – point (Point A1 on the x-t plane) with steady state conditions at – L/2a seconds

The flow along the entire pipeline is constant for steady state conditions prior to zero seconds, thus the flow at the
mid – point of the pipeline is

3
Q A1 = VA = 1.0  0.1964  = 0.1964 m /s

The initial steady state hydraulic grade line is given by the head loss of h f = 1.998 m from the Darcy – Weisbach
equation from Equation 21.107 in Example 21.10. The initial steady state HGL at section 2 or point A1 on the x-t plane
(the mid – point of the pipeline) before the arrival of the water hammer wave at L/2a seconds from Equation 21.108 is

H A1 = 199.001 m

+
Method of characteristics calculations at point P1 on the x-t plane (at section 3) at t = 0 seconds

+
The C compatibility equation from Equation 21.92 is

H P1 = C pc – B pc Q P1

Compute the values of the “constants” of C pc and B pc for the previous time step from Equation 21.90 and
Equation 21.91 as

C pc = H A1 + BQ A1 = 199.001 + 508.65  0.1964  = 298.90

Note the term BQ A1 has been added to the HGL at section 2 at the mid-point of the pipeline rather than to the HGL at
section 3 at the valve. This leads to an error in the MOC calculation due to the fact that only two reaches are used. For
increased accuracy more computational reaches must be used.

B pc = B + R Q A1 = 508.65 + 25.90 0.1964 = 513.74

+
The flow at the valve following the instantaneous closure of the valve at t = 0 is

3
Q P1 = 0.0 m /s

This is a closed valve boundary condition.

The water hammer HGL at the valve following the instantaneous valve closure is

H P1 = C pc – B pc Q P1 = 298.90 – 513.74  0.0  = 298.90 m

557
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

+
Conditions at the valve just after the valve closure at t = 0 are shown in Table 21.4. Values computed for this time
step are shown in bold. The final results are also shown in Figure 21.22.

Table 21.4 Transient conditions in pipeline just after the valve closure at Point P1 on the x-t plane
at zero seconds

Location At reservoir At mid – point At valve (Point P1 on

the x-t plane)
Section 1 2 3
Previous HGL (m) at 198.002
minus L/a seconds
Time (s) + + +
0 0 0
Time (L/a units) + + +
0 L/a 0 L/a 0 L/a
HGL (m) 200.0 199.001 298.90
Head rise H (m) Has not changed. Steady Has not changed. Steady 100.90
state conditions. state conditions

Flow Q (m3/s) 0.1964 0.1964 0.0

Previous V (m/s) at minus 1.0

L/a seconds
Velocity V (m/s) 1.0 1.0 0.0
Velocity change V (m/s) Has not changed. Steady Has not changed. Steady – 1.0
state conditions state conditions

Comparison with the Joukowsky head rise

The head rise above the initial HGL just upstream of the valve at section 3 or point P1 on the x-t plane as computed
from the method of characteristics method is

 H P1 = 298.90 – 198.002 = 100.898 m

This value is in fact incorrect as will be demonstrated below.

Compare the head rise with that from the Joukowsky formula of Equation 21.35 as

aV 980  0.0 – 1.0 

 H P1 = – ----------- = – ----------------------------------- = 99.90 m (21.115)
g 9.81

Figure 21.22 Summary of method of characteristics results for calculation at the closed valve
at Point P1 on the x-t plane at zero seconds

This value is in fact correct. The HGL for P1 on the x-t plane should be 297.90 m and not 298.90 m.
The discrepancy of 1.0 m here actually highlights a problem with the MOC method. Using many more reaches would
reduce the error.

If the positive characteristic was applied over a very short distance so that Point A1 on the x-t plane was very close to
P1 such that  x  0 then R  0 . For this computation, H A1 becomes 198.002 m instead of 199.001 m, a significant

558
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

change.

Now recompute the values of the “constants” of C pc and B pc for the previous time step from Equation 21.90 and
Equation 21.91 as

C pc = H A1 + BQ A1 = 198.002 + 508.65  0.1964  = 297.90

B pc = B + R Q A1 = 508.65 + 0.0  25.90  0.1964 = 508.65

The recomputed water hammer HGL at the valve following the instantaneous valve closure is now

H P1 = C pc – B pc Q P1 = 297.90 – 513.74  0.0  = 297.90 m

Thus the Joukowsy head rise is recomputed water hammer HGL at the valve following the instantaneous valve closure
is now

 H P1 = 297.90 – 198.002 = 99.90 m

This now matches the actual Joukowsy head rise computed as Equation 21.115.

___________________________________________________________________________________

21.17 The Reservoir Boundary Condition

21.17.1 The governing equations
Consider the reservoir – pipe – valve system in Figure 21.23. In the method of characteristics solution
method, the conditions at the reservoir (section 1 or point P on the x-t plane) at time L/2a seconds are
influenced by the conditions at the computational section immediately downstream (section 2 or point B
on the x-t plane) at the mid – point of the pipeline at time of zero seconds in Figure 21.23.

Figure 21.23 Reservoir boundary condition in the x – t plane

_
The C integrated compatibility equation between point B and P (both on the x-t plane) in Figure 21.23
_
from Equation 21.101 is used. It is valid along the C characteristic line. Information may be thought of
as propagating along the characteristic line from B to P from a time zero to time L/2a seconds.

H P = C mc + B mc Q P (21.116)

where

559
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

• C mc and B mc depend on the “known” conditions of H B and Q B at point B on the x-t

plane (the adjacent downstream reach) at the previous time step of zero seconds

C mc = H B – B Q B and B mc = B + R Q B

with the characteristic impedance and reach resistance of

f x
a- and R = -----------------
B = ------ -
gA 2gD A
2

+
There are two unknowns of HP and QP in Equation 21.116. There is no C equation available on the
reservoir side of section 1 or point P (on the x-t plane) at the reservoir. In actual fact, one of the two
unknowns is known and for a stationary reservoir water surface is always specified and constant

HP = HR (21.117)

where

• H P = unknown transient head or HGL at point P in Figure 21.23 (m or ft)

• H R = upstream reservoir head or HGL at point P in Figure 21.23 (m or ft)

Substituting Equation 21.117 into Equation 21.116 gives an expression to solve for the flow at the reser-
voir at section 1 or point P (on the x-t plane) based on conditions at the previous time step at the section
immediately upstream

H R – C mc
Q P = -----------------------
- (21.118)
B mc

DISCHARGE FOR A RESERVOIR BOUNDARY CONDITION

where

• Q P = unknown discharge or flow at point P in Figure 21.23 (m3/s or ft3/s)

• H R = upstream reservoir head or HGL at point P in Figure 21.23 (m or ft)
• C mc = “C” constant for negative characteristic line between Point B and Point P
• B mc = “B” constant for negative characteristic line between Point B and Point P

Thus to alter the flow at the reservoir then either the head HB or the flow QB must be altered by changing
conditions at the valve at the downstream end of the pipeline.

21.17.2 An MOC example for steady state conditions at the reservoir at L/2a
seconds (Point P3 on the x-t plane)
The calculations in the following Example are not really necessary given an understanding of the wave
travel characteristics of the valve closure. The water hammer wave will take L/a seconds to travel from
the valve to the reservoir. Thus the water hammer wave at L/2a seconds will have only reached the mid-
point of the pipeline and as a result the conditions at the reservoir will be still at steady state flow condi-
tions. However, if the method of characteristics is implemented within a computer code then the follow-
ing calculations will actually be carried out as it will not know that the conditions at the reservoir are at
steady state. The same is true for the reservoir at a time of zero seconds - a computer method of charac-
teristics calculation will also be carried out for this point.

560
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Example 21.12 Consider a reservoir – pipe – valve system in Figure 21.24 with the following characteristics (same
as for Example 21.10): reservoir elevation HR = 200 m, pipe length L = 980 m, pipe diameter D = 0.5 m, Darcy –
Weisbach friction factor f = 0.02, initial velocity V = 1.0 m/s, wave speed a = 980 m/s and gravitational acceleration
g = 9.81 m/s2.
The valve at the downstream end is closed instantaneously at a time of zero seconds. Compute the flow and velocity
at the reservoir (Point P3 on the x-t plane) at L/2a seconds.

Answer – Computation of the flow and velocity at the reservoir at L/2a seconds.

Consider the x – t plane in Figure 21.28.

Computing transient conditions at section 1 – at the reservoir (Point P3 on the x-t plane)

_
The conditions at the reservoir at P3 at time t = L/2a must be determined from the C characteristic line from the con-
ditions at the mid – point between B3 on the x-t plane at time zero seconds and P3 or section 1 as shown in
Figure 21.24. The usage of the number 3 in B3 and P3 refers to the third method of characteristics point computation
that is carried out and does not refer to the section numbers at ends of reaches along the pipeline.

The transient was generated by the closure of the valve at the downstream end of the pipeline at zero seconds. In the
first L/2a seconds the water hammer wave has traveled from the valve to the mid-point of the pipeline.

As for Example 21.10, the reach length  x is 490 m or one – half the length of the pipeline as computed in
Equation 21.104 of Example 21.10. The corresponding time step x from the characteristic equation is 0.5 seconds
from Equation 21.105.

2
The area ( A = 0.1964 m ), characteristic impedance ( B = 508.65 ), reach resistance ( R = 25.90 ) for the pipe are
given by Equation 21.106, Equation 21.110 and Equation 21.111 of Example 21.10 respectively.

Conditions at section 2 – at the mid – point of the pipeline (point B3 on the x-t plane) at a time of zero seconds

Conditions are taken from Example 21.10 as

3
H B3 = 199.001 m and Q B3 = 0.1964 m /s

Method of characteristics calculations at point P3 on the x-t plane (or at section 1) at L/2a seconds

From Equation 21.118 the flow at the reservoir is

H R – C mc
Q P3 = -----------------------
-
B mc

Figure 21.24 Calculation of the flow and velocity at the reservoir (Point P3 on the x-t plane) at L/2a seconds

Compute the values of the “constants” of C mc and B mc for the previous time step from Equation 21.99 and

561
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Equation 21.100 as

C mc = H B3 – B Q B3 = 199.001 – 508.65  0.1964  = 99.10

B mc = B + R Q B3 = 508.65 + 25.90 0.1964 = 513.74

Thus the flow at the reservoir from Equation 21.118 is

H R – C mc 200.0 – 99.10 3
Q P3 = -----------------------
- = --------------------------------- = 0.1964 m /s
B mc 513.74

The corresponding velocity is

Q P3
- = 0.1964
V P3 = -------- ---------------- = 1.0 m/s
A 0.1964

Thus the exact initial steady state condition for discharge and velocity have been calculated, which is what one would
expect, as the water hammer wave from the valve closure has only reached the mid-point of the pipeline.

Conditions at the reservoir just after the water hammer wave arrives at the mid-point of the pipeline for the first time
are shown in Table 21.7. Values computed for this time step are shown in bold. The final results are also shown in
Figure 21.25.

Table 21.5 Transient conditions in pipeline just after the valve closure at Point P1 on the x-t plane
at zero seconds

Location At reservoir At mid – point At valve

(Point P3 on the x-t
plane)
Section 1 2 3
Time (s) 0.5 0.5 0.5
Time (L/a units) L/2a L/2a L/2a
HGL (m) 200.0 299.40 298.90
Does not change. [Example 21.10 [Example 21.11 |(Point
Steady state. (Point P2 on the x-t plane)] P1 on the x-t plane)]
Head rise H (m) Does not change.
Steady state.

Flow Q (m3/s) 0.1964 0.00098 0.0

Previous V (m/s) at zero 1.0 1.0 0.0

L/a seconds
Velocity V (m/s) 1.0 0.005 0.0
Does not change.
Steady state.
Velocity change V (m/s) 0.0

Figure 21.25 Summary of method of characteristics results for calculation at the reservoir
at Point P3 on the x-t plane at L/2a seconds

562
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

___________________________________________________________________________________

21.18 Further MOC Computations

21.18.1 An MOC example for conditions at the closed valve at L/2a seconds
(Point P4 on the x-t plane)
Consider the reservoir – pipe – valve system in Figure 21.26. As was seen in Example 21.11 the head at
the valve (section 1 or Point P4 on the x-t plane) is computed from conditions at the computational sec-
tion immediately upstream (section 2 or Point A4 on the x-t plane) at the mid – point of the pipeline.

Example 21.13 Consider a reservoir – pipe – valve system as shown in Figure 21.26 with the same characteristics as
the system described in Example 21.10: reservoir elevation HR = 200 m, pipe length L = 980 m, pipe diameter
D = 0.5 m, Darcy – Weisbach friction factor f = 0.02, initial velocity V = 1.0 m/s, wave speed a = 980 m/s and gravi-
tational acceleration g = 9.81 m/s2.
The valve at the downstream end was closed instantaneously at a time of zero seconds. Compute the hydraulic grade
line at the valve (Point P4 on the x-t plane) at L/2a seconds.

Answer – Computation of the HGL at the valve at L/2a seconds.

Consider the x – t plane in Figure 21.26.

Computing transient conditions at section 1 at the valve (Point P4 on the x-t plane)

+
The head at P4 needs to be calculated from the C characteristic line from the conditions at the mid – point of the pipe-
line at a time of zero seconds (designated as point A4 on the x-t plane in Figure 21.26).

The transient was generated by the closure of the valve at the downstream end of the pipeline at zero seconds. In the
first L/2a seconds the water hammer wave has traveled from the valve to the mid-point of the pipeline. Conditions at
the valve are recalculated at this time. At the valve at L/2a seconds the discharge remains at zero.

Two reaches are selected as for the previous examples. Thus the reach length  x is 490 m or one – half the length of the
pipeline as computed in Equation 21.104 of Example 21.10. The corresponding time step  t from the characteristic
equation is 0.5 seconds from Equation 21.105.

Figure 21.26 x – t plane for valve boundary condition at L/a seconds

2
The area ( A = 0.1964 m ), characteristic impedance ( B = 508.65 ), reach resistance ( R = 25.90 ) for the pipe are
given from Example 21.10 by Equation 21.106, Equation 21.110 and Equation 21.111 respectively.

Conditions at section 2 – at the mid – point (Point A4 on the x-t plane) at zero seconds

Conditions at point A4 on the x-t plane at the mid – point of the pipeline are taken from Example 21.10 as the initial
steady state conditions of

563
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

3
H A4 = 199.001 m and Q A4 = 0.1964 m /s

Method of characteristics calculations at point P4 on the x-t plane (or at section 3) at L/2a seconds

+
The C compatibility equation from Equation 21.92 is

H P4 = C pc – B pc Q P4

Compute the values of the “constants” of C pc and B pc for the previous time step from Equation 21.90 and
Equation 21.91 as

C pc = H A4 + BQ A4 = 199.0 + 508.65  0.1964  = 298.90

B pc = B + R Q A4 = 508.65 + 25.90 0.1964 = 513.74

The flow at the valve (assuming the pressure stays above the vapor pressure of the fluid) at t = L/2a is

3
Q P4 = 0.0 m /s

The water hammer HGL at the valve at t = L/2a is

H P4 = C pc – B pc Q P4 = 298.90 – 513.74  0.0  = 298.90 m

Thus the conditions at the valve are identical to those immediately following valve closure as shown for Point P1 on
the x-t plane in Example 21.11 and as shown in Table 21.4.

Conditions at the valve just after the valve closure at a time of L/2a seconds are shown in Table 21.6. Values computed
for this time step are shown in bold. The final results are also shown in Figure 21.27.

Table 21.6 Transient conditions in pipeline at the closed valve at Point P4 on the x-t plane at L/2a seconds
after the valve closure

Location At reservoir At mid – point At valve

(Point P4 on the x-t plane)
Section 1 2 3
Previous HGL (m) at zero sec 298.90
Time (s) 1.0 1.0 1.0
Time (L/a units) L/2a L/2a L/2a
HGL (m) at L/2a sec 200.0 299.40 298.90
Head rise H (m) None since last calculation time
+
step of 0 seconds
(Example 21.11
[Point P1 on the x-t plane])

Flow Q (m3/s) at L/2a sec 0.1964 0.00098 0.0

Previous V (m/s) at zero sec 0.0

Velocity V (m/s) 1.0 0.005 0.0
Velocity change V (m/s) None since the calculation time
+
step of 0 seconds

564
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.27 Summary of method of characteristics results for calculation at the closed valve
at Point P4 on the x-t plane at L/2a seconds

___________________________________________________________________________________

21.18.2 A MOC example of flow reversal at the reservoir at L/a seconds

(Point P5 on the x-t plane)
Consider now an example to illustrate the reversal of velocity and flow at the reservoir when the sharp
water hammer reflects from the reservoir for the first time at L/a seconds. See Figure 21.7(e) and
Figure 21.8(f) for details of arrival and reflection of the water hammer wave at the reservoir for the first
time.

Example 21.14 Consider a reservoir – pipe – valve system in Figure 21.28 with the following characteristics (same
as for Example 21.10): reservoir elevation HR = 200 m, pipe length L = 980 m, pipe diameter D = 0.5 m, Darcy –
Weisbach friction factor f = 0.02, initial velocity V = 1.0 m/s, wave speed a = 980 m/s and gravitational acceleration
g = 9.81 m/s2.

The valve at the downstream end is closed instantaneously at a time of zero seconds. Compute the reversal of both
the flow and velocity at the reservoir (Point P5 on the x-t plane) following the reflection of the water hammer wave
from the reservoir at L/a seconds.

Answer – Computation of the reversal of the flow and velocity at the reservoir at L/a seconds.

Consider the x – t plane in Figure 21.28.

Computing transient conditions at section 1 – at the reservoir (Point P5 on the x-t plane)

_
The conditions at the reservoir at P5 on the x-t plane at time t = L/a must be determined from the C characteristic line
from the conditions at the mid – point between B5 on the x-t plane at time L/2a seconds and P5 or section 1 as shown in
Figure 21.28.

The transient was generated by the closure of the valve at the downstream end of the pipeline at zero seconds. In the
first L/a seconds the water hammer wave has traveled from the valve to the reservoir where it reflects. The reflection of
the water hammer wave from the reservoir is being calculated here.

As for Example 21.10, the reach length  x is 490 m or one – half the length of the pipeline as computed in
Equation 21.104. The corresponding time step t from the characteristic equation is 0.5 seconds from
Equation 21.105.

565
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.28 Calculation of reversal of the flow and velocity at the reservoir (Point P5 on the x-t plane) at L/a seconds

2
The area ( A = 0.1964 m ), characteristic impedance ( B = 508.65 ), reach resistance ( R = 25.90 ) for the pipe are
given by Equation 21.106, Equation 21.110 and Equation 21.111 of Example 21.10 respectively.

Conditions at section 2 – at the mid – point of the pipeline (point B5) at a time of L/2a seconds

Conditions are taken from Example 21.10 as

3
H B5 = 299.40 m and Q B5 = 0.00098 m /s

Method of characteristics calculations at point P5 on the x-t plane (or at section 1) at L/a seconds

From Equation 21.118 the flow at the reservoir is

H R – C mc
Q P5 = -----------------------
-
B mc

Compute the values of the “constants” of C mc and B mc for the previous time step from Equation 21.99 and
Equation 21.100 as

C mc = H B5 – B Q B5 = 299.40 – 508.65  0.00098  = 298.90

B mc = B + R Q B5 = 508.65 + 25.90 0.00098 = 508.68

Thus the flow at the reservoir from Equation 21.118 is

H R – C mc 200.0 – 298.90 3
Q P5 = -----------------------
- = ------------------------------------ = – 0.1944 m /s
B mc 508.68

The corresponding velocity is

Q P5
- = –------------------
0.1944- = – 0.990 m/s
V P5 = --------
A 0.1964

Thus after the reflection of the water hammer wave at the reservoir, the flow is out of the pipe into the reservoir at a
velocity of negative 0.990 m/s. Just before the arrival of the water hammer wave at the reservoir the velocity was into
the pipe at a velocity of 1.0 m/s. Thus there has been a virtually total reversal of the velocity at the reservoir. Conditions
at the reservoir just after the water hammer wave arrives at the reservoir for the first time are shown in Table 21.7. Val-
ues computed for this time step are shown in bold.
The final results are also shown in Figure 21.29.

566
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Table 21.7 Transient conditions in pipeline at the reservoir at Point P5 on the x-t plane at L/a seconds
after the valve closure

Location At reservoir (Point P5 At mid – point At valve

on the x-t plane)
Section 1 2 3
Time (s) 1.0 1.0 1.0
Time (L/a units) L/a L/a L/a
HGL (m) 200.0 299.40 298.90
Head rise H (m) Does not change

Flow Q (m3/s) – 0.1944 0.00098 0.0

Previous V (m/s) at 1.0

L/2a seconds
Velocity V (m/s) – 0.990 0.005 0.0
Velocity change V – 1.990
(m/s)

Figure 21.29 Summary of method of characteristics results for calculation at the reservoir at Point P5 on the x-t plane
at L/a seconds - the flow and velocity both reverse

___________________________________________________________________________________

21.18.3 An MOC example for conditions at the closed valve at L/a seconds
(Point P6 on the x-t plane)
As was seen in Example 21.11, the head at the valve (section 1 or point P6 on the x-t plane) is computed
from conditions at the computational section immediately upstream (section 2 or point A6 on the x-t
plane) at the mid – point of the pipeline.

Example 21.15 Consider a reservoir – pipe – valve system as shown in Figure 21.30 with the same characteristics as
the system described in Example 21.10: reservoir elevation HR = 200 m, pipe length L = 980 m, pipe diameter
D = 0.5 m, Darcy – Weisbach friction factor f = 0.02, initial velocity V = 1.0 m/s, wave speed a = 980 m/s and gravi-
tational acceleration g = 9.81 m/s2.

The valve at the downstream end was closed instantaneously at a time of zero seconds. Compute the hydraulic grade
line at the valve (Point P6 on the x-t plane) at L/a seconds.

Answer – Computation of the HGL at the valve at L/a seconds.

Consider the x – t plane in Figure 21.30.

Computing transient conditions at section 1 at the valve (Point P6 on the x-t plane)

567
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

+
The head at P6 on the x-t plane needs to be calculated from the C characteristic line from the conditions at the mid –
point of the pipeline at a time of L/2a seconds (designated as point A6 on the x-t plane in Figure 21.30).

The transient was generated by the closure of the valve at the downstream end of the pipeline at zero seconds. In the
first L/a seconds the water hammer wave has traveled from the valve to the reservoir. Conditions at the valve are recal-
culated at this time. At the valve at L/a seconds the discharge remains at zero.

Two reaches are selected as for the previous examples. Thus the reach length  x is 490 m or one – half the length of the
pipeline as computed in Equation 21.104 of Example 21.10. The corresponding time step  t from the characteristic
equation is 0.5 seconds from Equation 21.105.

Figure 21.30 x – t plane for valve boundary condition at L/a seconds

2
The area ( A = 0.1964 m ), characteristic impedance ( B = 508.65 ), reach resistance ( R = 25.90 ) for the pipe are
given from Example 21.10 by Equation 21.106, Equation 21.110 and Equation 21.111 respectively.

Conditions at section 2 – at the mid – point (Point A6 on the x-t plane) at L/2a seconds

Conditions at point A6 on the x-t plane at the mid – point of the pipeline are taken from Example 21.10 as the transient
conditions of

3
H A6 = 299.40 m and Q A6 = 0.00098 m /s

Method of characteristics calculations at point P6 on the x-t plane (or at section 3) at L/a seconds

+
The C compatibility equation from Equation 21.92 is

H P6 = C pc – B pc Q P6

Compute the values of the “constants” of C pc and B pc for the previous time step from Equation 21.90 and
Equation 21.91 as

C pc = H A6 + BQ A6 = 299.4 + 508.65  0.00098  = 299.90

B pc = B + R Q A6 = 508.65 + 25.90 0.00098 = 508.68

The flow at the valve (assuming the pressure stays above the vapor pressure of the fluid) at t = L/a is

3
Q P6 = 0.0 m /s

The water hammer HGL at the valve at t = L/a seconds is

H P6 = C pc – B pc Q P6 = 299.90 – 513.74  0.0  = 299.90 m

Note that in comparison to the valve head computed at L/2a seconds the head here has increased from 298.90 m to
299.90 m, a one meter increase. This is a result of the friction or slope of the HGL in the system.

568
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Conditions at the valve just after the valve closure at a time of L/a seconds are shown in Table 21.8. Values computed
for this time step are shown in bold. The final results are also shown in Figure 21.31.

Table 21.8 Transient conditions in pipeline at the closed valve at Point P6 on the x-t plane at L/a seconds
after the valve closure

Location At reservoir At mid – point At valve

(Point P6 on the x-t plane)
Section 1 2 3
Previous HGL (m) at L/2a sec 298.90
Time (s) 1.0 1.0 1.0
Time (L/a units) L/a L/a L/a
HGL (m) at L/a sec 200.0 Not computed 299.90
Head rise H (m) +1.0
since last calculation time
step of L/2a seconds (Point
P4 on the x-t plane
[Example 21.13])

Flow Q (m3/s) at L/a sec -0.1944 Not computed 0.0

(Point P3 on the x-t
plane [Example 21.12])
Previous V (m/s) at L/2a sec 1.0 0.0
Velocity V (m/s) -0.990 0.0
Velocity change (m/s) -1.990 None since the calculation
time step of L/2a seconds

Figure 21.31 Summary of method of characteristics results for calculation at the closed valve
at Point P6 on the x-t plane at L/a seconds

21.19 Further Examples of Applying the Method of Characteristics

The sequence of events during a water hammer event in a reservoir – pipeline – valve system have been
considered in detail earlier in the Chapter in Figure 21.7 to Figure 21.10. A period of 4L/a seconds has
been considered. Consider now the following additional cases during the first 2L/a seconds of the tran-
sient following instantaneous closure of the valve

• At 3L/2a when the water hammer wave reaches the mid-point of the pipeline for the second
time (presented in Example 21.16 at Point P7 on the x-t plane)
• At 2L/a when the water hammer wave reflects from the closed valve for the first time (pre-
sented in Example 21.17 at Point P8 on the x-t plane)

569
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

21.19.1 An MOC example for conditions at the mid – point of the pipeline at
3L/2a seconds (Point P7 on the x-t plane)
Consider the x – t plane and the reservoir – pipeline – valve system in Example 21.16 and Figure 21.32.
See Figure 21.8 (g) for details of arrival of the water hammer wave at the mid – point after it has
reflected from the reservoir for the first time. This corresponds to Figure 21.8 (g).

Example 21.16 Consider a reservoir – pipe – valve system in Figure 21.32 with the same characteristics as the sys-
tem described in Example 21.10: reservoir elevation HR = 200 m, pipe length L = 980 m, pipe diameter D = 0.5 m,
Darcy – Weisbach friction factor f = 0.02, initial velocity V = 1.0 m/s, wave speed a = 980 m/s and gravitational
acceleration g = 9.81 m/s2.
The valve at the downstream end is closed instantaneously at a time of zero seconds. Compute the hydraulic grade
line at the mid – point of the pipe (Point P7 on the x-t plane) at 3L/2a seconds.

Answer – Computation of the head and flow at the mid-point of the pipeline at 3L/2a seconds

Consider the x – t plane in Figure 21.32.

Computing transient conditions at section 2 at the valve (Point P7on the x-t plane)

+
The head at P7 on the x-t plane at time 3L/2a needs to be calculated from the C characteristic line from the reservoir
-
at a time of L/a seconds (designated as Point A7 on the x-t plane in Figure 21.32) and from the C characteristic line
from the valve at a time of L/a seconds (designated as Point B7 on the x-t plane in Figure 21.32).

The transient was generated by the closure of the valve at the downstream end of the pipeline at zero seconds. In the
first 3L/2a seconds the water hammer wave has traveled from the valve to the reservoir where it reflects and has trav-
eled back to the mid – point of the pipeline where it generates a significant head decrease and a corresponding velocity
reversal.

Two reaches are selected as for the previous examples. Thus the reach length  x is 490 m or one – half the length of the
pipeline as computed in Equation 21.104 of Example 21.10. The corresponding time step  t from the characteristic
equation is 0.5 seconds from Equation 21.105.

Figure 21.32 Calculation of head and flow at the mid – point of the pipe at Point P7 on the x-t plane at 3L/2a seconds

2
The area ( A = 0.1964 m ), characteristic impedance ( B = 508.65 ), reach resistance ( R = 25.90 ) for the pipe are
given in Example 21.10 by Equation 21.106, Equation 21.110 and Equation 21.111 respectively.

Conditions at section 1 – at the reservoir (point A7 on the x-t plane) at L/a seconds

Conditions at point A7 on the x-t plane at the reservoir are taken from Example 21.14 as

570
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

3
H A7 = 200.00 m and Q A7 = – 0.1944 m /s

Conditions at section 3 – at the closed valve (point B7 on the x-t plane) at L/a seconds

Conditions at point B7 on the x-t plane at the closed valve are taken from Example 21.13 as

3
H B7 = 299.90 m and Q B7 = 0.0 m /s

Method of characteristics calculations at point P7 on the x-t plane (or at section 2) at 3L/2a seconds

The water hammer HGL and the discharge at the mid – point at Point B7on the x-t plane are given from the method of
characteristics of Equation 21.102 and Equation 21.103 respectively as

C pc B mc + C mc B pc C pc – C mc
- and Q P7 = ------------------------
H P7 = -------------------------------------------- -
B pc + B mc B pc + B mc

Compute the values of the “constants” of C pc and B pc for the previous time step at L/a seconds from Equation 21.90
and Equation 21.91 as

C pc = H A7 + BQ A7 = 200.0 + 508.65  – 0.1944  = 101.12

B pc = B + R Q A7 = 508.65 + 25.90 – 0.1944 = 513.68

Now compute the values of the “constants” of C mc and B mc for the previous time step at L/a seconds from
Equation 21.99 and Equation 21.100 as

C mc = H B7 – BQ B7 = 299.90 – 508.65  0.0  = 299.90

B mc = B + R Q B7 = 508.65 + 25.90 0.0 = 508.65

Thus the water hammer HGL at the mid-point of the pipeline at L/2a seconds is

C pc B mc + C mc B pc  508.65  + 299.90  513.68 - = 201.0 m

- = 101.12
H P7 = -------------------------------------------- ----------------------------------------------------------------------------------
B pc + B mc 513.68 + 508.65

The corresponding transient flow at the mid – point at 3L/2a seconds corresponding to the arrival of the water hammer
wave from Equation 21.103 is

C pc – C mc 101.12 – 299.90 = – 3
Q P7 = ------------------------
- = --------------------------------------
- 0.1944 m /s
B pc + B mc 513.68 + 508.65

The corresponding velocity is

Q P5 – 0.1944
V P7 = --------
- = ------------------- = – 0.990 m/s
A 0.1964

A summary of conditions at the three sections along the pipeline at 3L/2a seconds is given in Table 21.9. Values com-
puted for this time step are shown in bold. The final results are also shown in Figure 21.33.

571
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Table 21.9 Transient conditions in pipeline at the mid-point at Point P7 on the x-t plane at 3L/2a seconds
after the valve closure

Location At reservoir At mid – point At valve

(Point P7 on the x-t
plane)
Section 1 2 3
Previous HGL (m) at L/2a sec 299.40
Time (s) 1.5 1.5 1.5
Time (L/a units) 3L/2a 3L/2a 3L/2a
HGL (m) Not computed 201.0 Not computed
Head rise H (m) – 98.40

Flow Q (m3/s) Not computed – 0.1944 Not computed

Previous V (m/s) at L/2a sec 0.005

Velocity V (m/s) – 0.990
Velocity change V (m/s) – 0.995

Figure 21.33 Summary of method of characteristics results for calculation at the mid – point of the pipeline
at Point P7 on the x-t plane at 3L/2a seconds

___________________________________________________________________________________

572
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

21.19.2 An MOC example of inversion of head at the closed valve at 2L/a

seconds (Point P8 on the x-t plane)
Consider the x – t plane and the reservoir – pipeline – valve system in Example 21.17 and Figure 21.34.
See Figure 21.8 (i) for details of arrival of the water hammer wave at the valve for the first time after it
has reflected from the reservoir and traveled back towards the valve.

Example 21.17 Consider a reservoir – pipe – valve system in Figure 21.34 with the same characteristics as the sys-
tem described in Example 21.10: reservoir elevation HR = 200 m, pipe length L = 980 m, pipe diameter D = 0.5 m,
Darcy – Weisbach friction factor f = 0.02, initial velocity V = 1.0 m/s, wave speed a = 980 m/s and gravitational
acceleration g = 9.81 m/s2.
The valve at the downstream end is closed instantaneously at a time of zero seconds. Compute the hydraulic grade
line at the valve (Point P8 on the x-t plane) at 2L/a seconds.

Answer – Computation of the inversion of the HGL at the valve at 2L/a seconds

Consider the x – t plane in Figure 21.34.

Computing transient conditions at section 3 at the valve (Point P8 on the x-t plane)

+
The head at P8 on the x-t plane needs to be calculated from the C characteristic line from the conditions at the mid –
point of the pipeline at a time of 3L/2a seconds (designated as Point A8 on the x-t plane in Figure 21.34).

The transient was generated by the closure of the valve at the downstream end of the pipeline at zero seconds. In the
first 2L/a seconds the water hammer wave has traveled from the valve to the reservoir and back to the valve for the first
time where it generates a significant head reduction.

2
The area ( A = 0.1964 m ), characteristic impedance ( B = 508.65 ), reach resistance ( R = 25.90 ) for the pipe are
given by Example 21.10 from Equation 21.106, Equation 21.110 and Equation 21.111 respectively.

Conditions at section 2 – at the mid – point (Point A8 on the x-t plane) at 3L/2a seconds

Conditions at point A8 on the x-t plane at the mid – point of the pipeline are taken from Example 21.16 as

3
H A8 = 201.0 m and Q A8 = – 0.1944 m /s

Method of characteristics calculation at point P8 on the x-t plane (or at section 3) at 3L/2a seconds

+
The C compatibility equation from Equation 21.92 is

H P8 = C pc – B pc Q P8

573
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Figure 21.34 Transient conditions in pipeline at the closed valve at Point P8 on the x-t plane at 2L/a seconds
- the water hammer arrives back at the valve for the first time

Compute the values of the “constants” of C pc and B pc for the previous time step from Equation 21.90 and
Equation 21.91 as

C pc = H A8 + BQ A8 = 201.0 + 508.65  – 0.1944  = 102.12

B pc = B + R Q A8 = 508.65 + 25.90 – 0.1944 = 513.68

The flow at the valve (assuming the pressure stays above the vapor pressure of the fluid) at t = 2L/a is

3
Q P8 = 0.0 m /s

The water hammer HGL at the valve at t = 2L/a is

H P8 = C pc – B pc Q P8 = 102.12 – 513.68  0.0  = 102.12 m

Just before the arrival of the water hammer wave at the valve for the first time at 2L/a seconds, the head at the valve
was 299.90 m. After the reflection of the water hammer wave at the valve at 2L/a seconds, the head drops to 102.12 m.
Thus the head has dropped almost 200 m from a very high HGL to a low HGL. It has been effectively inverted from a
high head to a low head. Conditions at the valve just after the water hammer wave arrives at the valve for the first time
at 2L/a seconds are shown in Table 21.10. Values computed for this time step are shown in bold. The final results are
also shown in Figure 21.35.

574
Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

Table 21.10 Transient conditions in pipeline at the valve at Point P8 on the x-t plane at L/a seconds
after the valve closure

Location At reservoir At mid – point At valve

(Point P8 on the x-t
plane)
Section 1 2 3
Previous HGL (m) at L/a sec 299.90
Time (s) 2.0 2.0 2.0
Time (L/a units) 2L/a 2L/a 2L/a
HGL (m) Not computed Not computed 102.12
(this is above vapor
pressure head)
Head rise H (m) – 197.78

Flow Q (m3/s) Not computed Not computed 0.0

Previous V (m/s) at L/a sec 0.0

Velocity V (m/s) 0.0
Velocity change V (m/s) None since the calcula-
+
tion time step of 0
seconds

Figure 21.35 Summary of method of characteristics results for calculation at the closed valve
at Point P8 on the x-t plane at 2L/a seconds

575
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

21.19.3 The MOC calculation sequence in a computer program

In the examples in the previous sections, only selected points were considered. In a computer program
all the points as designated by circles as shown in Figure 21.36 would actually be computed starting
with the three sections at time zero (the mid-point, the reservoir boundary condition and the closed valve
boundary condition). Computations would proceed every L/2a seconds.

Figure 21.36 All the points for computation in a computer implementation

of the method of characteristics

___________________________________________________________________________________

21.20 Water hammer control methods

Large changes in velocity are usually responsible for large pressure variations in piping systems. A pipe-
line may usually be designed with a liberal factor of safety by selecting an adequate class of pipe (i.e
pipe wall thickness) that will enable the pipe to withstand both minimum and maximum pressures.
Excessive pressures and column separation (where the pressure drops to the vapor pressure head of the
fluid) are usually undesirable. Water hammer control devices are used to reduce the large changes in
velocity. Thus a smaller class of pipe may be selected and hence the cost of the pipe will be reduced. In
some instances, it may be more cost effective to use a higher class of pipe and eliminate the need for a
water hammer control device, but in long pipelines this is unlikely to be the case. An economic analysis
should be carried out to determine the least cost alternative. Some water hammer devices are expensive.

21.20.1 Common water hammer control measures

These include

576
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

• surge tanks
• air tanks or air chambers
• pressure relief valves
• additional pump inertia in the form of a flywheel
• a by – pass line around a pump
• wave speed reduction by selection of an alternative pipe material
• slower valve closure
• alteration of the pipeline profile
• increasing the diameter of the pipeline to reduce the velocity
An example of a surge tank is given in Figure 21.37.

Figure 21.37 A surge tank – 40 m (130 ft) high and 6 m (20 ft) diameter on the Mannum –
Adelaide pipeline (South Australia) for an above ground pipeline

An example of a Neyrtec pressure relief valve is shown in Figure 21.38.

Figure 21.38 Neyrtec pressure relief valves for water hammer control to prevent high pressures

577
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

21.20.2 Overview of surge tanks

A surge tank is an open tank connected to the piping system. Other names for a surge tank are surge tow-
ers or stand pipes. Inclusion of a surge tank can enable the very high rejoinder pressures due to column
separation cavity collapse to be eliminated. In addition to eliminating the high pressures – column sep-
aration can also be eliminated by use of a surge tank. Sometimes surge tanks are used on the suction
sides of pumping stations to eliminate upsurge resulting from pump power failure.

When the pressure in the pipeline suddenly increases, water flows up into the surge tank thus storing
water. For low pressures in the pipeline, the surge tank supplies water for the pipeline. Pressure waves
reflect from the surge tank. The result is that the transient occurs much more slowly with a large reduc-
tion in the maximum pressure in the system.

The surge tank is extremely simple and effective, with no dependence on moving or mechanical parts. A
surge tank should be regarded as an integral part of the pipeline and should be on – line at all times. For
a surge tank on a pumping line a long periodic oscillation will occur in the system following pump fail-
ure. It may take a long time for the oscillations to damp out. Thus a reasonable time interval should be
allowed prior to re – starting the pumps in order to avoid the possibility of overflow of the surge tank.

Some categories of surge tank include (see Figure 21.39)

• simple surge tank.

• orifice surge tank. Water flows out of the tank more easily during downsurge in order to
prevent column separation while water flows less easily into the tank during upsurge.
• differential surge tank. Has a riser inside the surge tank.
• a one – way surge tank. Assists in preventing column separation during downsurge. Works
when the HGL in the pipe drops below the level in the one – way surge tank. A non – return
valve prevents water from flowing into the surge tank.
• a closed surge tank. These are air chambers and will be discussed in a later section.

Figure 21.39 Various type of surge tanks

21.20.3 Open surge tanks

In order to avoid a surge tank tower that is too tall, open surge tanks are only practicable when there is a
location such that the normal operating level is less than say 30 to 40 m (100 to 130 ft) above the ground
surface. Thus the hydraulic grade line level must be close to the pipeline to justify the use of an open
surge tank. If an open surge tank is going to be too tall, a logical alternative is a closed surge tank or air
vessel (sometimes referred to as an air accumulator). A surge tank needs to be large enough to prevent

578
 Chapter 21 - File: Chap21-Waterhammer.fm July 6, 2022 Version 7

air from being drawn from the tank into the pipeline during the downsurge portion of a transient event.
An example of a surge tank is on the Mannum – Adelaide pipeline and is 40 m (130 ft) tall and has a
diameter of 6.1 m (20 ft). It is constructed of reinforced concrete and connects to the 1450 mm (57 in.)
diameter pipeline.

Sometimes a surge tank will be designed to accommodate for overflow during the transient event in
order to restrict the height of the surge tank. If a surge tank is not specifically designed to overflow then
an overflow condition usually indicates that there is something wrong. This may be caused by a block-
age or a closed (or partially closed valve) in the pipeline on the downstream side of the surge tank.

Often the inlet to the open surge tank often serves as the outlet. In some cases, both the inlet and outlet
may be separate pipe connecting to the main pipeline. If there is a shutoff valve in the connecting pipe
between the main pipeline and the surge tank – it is extremely important that this valve remains open
when the pumping system is operational.

Extensive coverage of water hammer control devices is given by Wylie and Streeter (1993) and
Chaudhry (2014).

21.21 References
Bonin, C. C., 1960, “Water-Hammer Damage to Oigawa Power Station,” Jour. Engineering for Power,
American Society of Mechanical Engineers., April, 111-119.

Chaudhry, M.H. (2014). Applied hydraulic transients. 3rd ed., Springer, New York.

HCP Software Ptd Ltd. http://www.hcp.net.au/ Accessed 12 November 2019.

Lupton, H.R. (1953). “Graphical Analysis of Pressure Surges in Pumping Systems.” Journal of the Insti-
tution of Water Engineers, Vol. 7, 87-156.

Rich, G. R. (1951). Hydraulic transients, McGraw-Hill Book Co., New York.

Streeter, V. L., Wylie, E. B., and Bedford, K. W. (1998). Fluid mechanics. WCB McGraw–Hill., Boston.

Wylie, E.B. and Streeter, V.L. (1993). Fluid transients in systems, Prentice Hall, Englewood Cliffs, New
Jersey.

579
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

CHAPTER 22. HYDROPOWER AND TURBINES1

22.1 Introduction
Turbines are used in hydropower generation of electricity. Hydropower plants have been in existence for
almost 140 years. The world's first hydroelectric power plant began operating in the United States along
the Fox River in Appleton, Wisconsin in 1882 (National Geographic 2019). Rotational energy is pro-
duced when flowing fluid strikes the blades of the turbine such that the turbine blades are displaced. The
rotational mechanical energy is transferred by the turbine generator shaft to the electric generator to pro-
duce electrical energy. The USGS (USGS 2019) quotes the Corps of Engineers (USA) in describing the
conversion of energy that occurs in hydropower as “A hydraulic turbine converts the energy of flowing
water into mechanical energy. A hydroelectric generator converts this mechanical energy into electric-
ity.”

Hydroelectric power “supplies 17.5% of the world's electricity (99% in Norway, 57% in Canada, 55% in
Switzerland, 40% in Sweden, 7% in USA)” (Electropaedia 2019). Two types of hydropower plants are
common 1) peaking plants and 2) run-of-river plants. The normal time of day usage of hydropower (for
peaking plants) is at peak-load demand because it can so easily started and stopped (Electropaedia
2019). However, accidents may occur due to transient water hammer pressures induced by fast actions -
so as a result often operators may be reluctant to start and stop turbine generation too frequently. In con-
trast to peaking plants, run-of-river hydropower generating plants provide a continuous supply that con-
tributes to base load generation.

Hydropower has some significant advantages - “Hydroelectric power generation is by far the most effi-
cient method of large scale electric power generation. Energy flows are concentrated and can be con-
trolled. The conversion process captures kinetic energy and converts it directly into electric energy.
There are no inefficient intermediate thermodynamic or chemical processes and no heat losses. The
overall efficiency can never be 100% however since extracting 100% of the flowing water's kinetic
energy means the flow would have to stop” (Electropaedia 2019). Other advantages of hydroelectric
power have been compiled by the USGS(2019) as:

• “Fuel is not burned so there is minimal pollution

• Water to run the power plant is provided free by nature
• Hydropower plays a major role in reducing greenhouse gas emissions
• Relatively low operations and maintenance costs
• The technology is reliable and proven over time
• It's renewable - rainfall renews the water in the reservoir, so the fuel is almost always
there”
The USGS (2019) also list some disadvantages of hydroelectric power:

• “High investment costs

• Hydrology dependent (precipitation)
• In some cases, inundation of land and wildlife habitat
• In some cases, loss or modification of fish habitat
• Fish entrainment or passage restriction
• In some cases, changes in reservoir and stream water quality

1.© Copyright 2022Jul. Prof. Emeritus Angus R. Simpson, University of Adelaide, Adelaide,
Australia. Part of the book entitled Water Distribution Systems Engineering.

580
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

• In some cases, displacement of local populations”

Three Gorges Hydroelectricity plant on the Yangtze River in China is the largest in the World as is
shown in Figure 22.1 It has 22,500 MW of installed capacity (USGS 2019b).

Figure 22.1 Three Gorges Hydropower Plant, China (Public Domain)

As coal fired power plants and gas fired power plants are phased out of use over the next 30 to 50 years
due to the impact of climate change and the need to de-carbonize the economy, the importance of hydro-
power and in particular pumped hydropower (pumped hydro) will increase to provide “battery storage”
as a complement to renewable intermittent electrical energy sources of wind and solar photovoltaic.

22.2 The Snowy Hydro Scheme

From 1949 to 1974 a number of dams and hydroelectric stations were constructed (4100 MW capacity).
The scheme has seven power stations, 16 major dams, 145 km of inter-connected tunnels and 80 km of
aqueducts (snowyhydro 2019). Tumut 3 Power Station which was part of the Snowy Scheme was the
first major pump-storage scheme in Australia (snowyhydro 2019). In 1967 the American Society of
Civil Engineers rated the Snowy Mountains Scheme as one of the civil engineering wonders of the
world.

22.3 Useful Definitions for Turbines (alphabetized)

TERM DEFINITION
BEP Best efficiency point on the turbine efficiency  versus flow Q curve
for a particular wicket gate opening
Power curve for turbine Plot of turbine generator shaft power PT versus flow Q. Can be com-
(P – Q) puted from the turbine head curve (HT versus Q) and the turbine effi-
ciency curve ( versus Q).
Turbine head curve Plot of turbine head HT versus flow Q for a turbine at a particular rota-
(HT – Q) tional speed. Usually obtained from the turbine manufacturer.
Turbine flow Q Flow through the turbine
Turbine efficiency  Ratio of the turbine generator shaft power PT to the water or hydraulic
power QHT associated with the water passing through the turbine
Turbine efficiency curve Plot of efficiency  versus flow Q for a turbine at a particular rota-
( – Q) tional speed and at a particular wicket gate opening. Usually obtained
from the turbine manufacturer.

581
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

Turbine head HT The drop in head across the turbine is taken from the inlet of the spiral
case of the turbine to the outlet of the draft tube on the discharge side
of the turbine (for a Francis or Kaplan turbine). This head is extracted
from the water and transferred to the rotational energy of the turbine
and then into electricity via the generator.
Turbine generator shaft Product of the turbine efficiency, the specific weight of fluid flowing
power PT , the turbine discharge Q and turbine head HT or PT = QHT
Turbine rotational speed N Rotational speed of the turbine
Static head Hs Difference in elevation between the water surface at the upper reser-
voir and the water surface elevation at the lower reservoir where water
is being delivered after passing through the turbine
System curve for penstock The static head Hs minus pipe friction losses and minor losses in pen-
pipework stock/tunnel pipework of the hydropower system
Water or hydraulic power Product of specific weight of fluid flowing , flow Q and turbine head
(IEC 1999). HT or QHT.

22.4 Components of Hydropower Generating Plants

A hydropower or hydroelectric power generating plant (see Figure 22.2) is made up of a number of
components (IEC 1999) including:

• Upstream reservoir
• Penstock or tunnel hydraulic passages (metallic or concrete) - pipelines or conduits that
transport the water to the turbine
• Control valves
• Needle valve (Pelton turbine)
• Spiral case or volute (Francis turbine)
• Stay or diffuser vanes (Francis turbine)
• Wicket gates or guide vanes (Francis and Kaplan turbine -see Figure 22.3)
• Turbine impeller or runner (Pelton (buckets), Francis, Kaplan (blades), bulb, Turgo and oth-
ers - see Section 22.7)
• Turbine generator shaft connecting the runner to the generator -see Figure 22.3
• Generator (with a rotor and a stator) -see Figure 22.3
• Transformer
• Switch yard
• Transmission cables or power lines
• Draft tube (Francis turbine and Kaplan turbine)
• Tailrace

Figure 22.2 A hydroelectric generating station showing a Francis turbine

582
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

(Teach Nuclear, Canada 2019

Figure 22.3 A Kaplan turbine and generator (Wikimedia: Creative Common License)

22.5 Classification of Hydropower Generating Plants by Size

Table 22.1 gives a classification of hydropower plants by size.

Table 22.1 Classification of hydropower (Hoes et al. 2017)

Category Size
Large > 10MW
Small < 10MW
Mini < 1MW
Micro < 0.1 MW
Pico < 0.005 MW

22.6 Turbine Specific Speed

The specific speed for a turbine is a very useful number to determine in the process of selecting a tur-
bine. It is a term to describe the geometry or shape on the turbine impeller (IEC 2019). The specific
speed for a turbine is defined as

12
N PT
N s = -------------------
- (22.1)
54
HT

SPECIFIC SPEED OF A TURBINE

where for both SI and US customary units

• N s = specific speed of turbine (actual units depend on system of units being used - see
Equation 22.2 and Equation 22.3)

583
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

• N = rotational speed (rpm)

• P T = turbine power delivered to the turbine generator shaft (kW or HP)
• HT = turbine head (m or ft)
The units of the turbine specific speed are not rpm but actually have the following dimensions for, first,
SI units”:

12
 N  PT   rpm   kW  1  2
------------------------------ = ------------------------------------ (22.2)
54 54
HT  m

In the case of US customary units the units of Ns are

12
 N  PT   rpm   HP  1  2
-----------------------------
- = ----------------------------------- (22.3)
54 54
HT   ft 

Due to the fact that the definition of specific speed for a turbine depends on the units system then the
numeric value of a specific speed for the same turbine expressed in both SI units and US customary
units are in fact different. Note that the turbine specific speed definition differs from that for a pump.

The relationship between specific speed for a turbine in the two sets of units is

N S  in SI units  = 3.81  N S  in US Customary units  (22.4)

A convention is adopted here that whenever the specific speed for a turbine is given then the units
should be specified in square brackets following the number. The reason that this practice should be fol-
lowed is so that it is very clear what the units of the quantities were that were used to calculate the value
of specific speed for the particular turbine. Examples of units are:

• NS (SI units) = 20 [rpm, kW, m]

• NS (US customary units) = 70 [rpm, HP, ft] for a different turbine than in the example just
above in previous bullet point
22.7 Turbine Types
A turbine is classified as either an impulse turbine or a reaction turbine. For an impulse turbine the
change of head is brought about primarily by a change in velocity. In contrast, for a reaction turbine the
change is head is brought about primarily by a change in pressure (University of Manchester 2019).

Four turbine types are described in this Chapter including:

• Pelton or impulse turbine (high head, low specific speed - see Section 22.6 for definition).
• Turgo turbine (medium head, low specific speed)
• Francis turbine (moderate head, moderate specific speed).
• Kaplan turbine or propeller turbine (low head, high specific speed).
Both Pelton wheel and Turgo turbines are impulse turbines. Both the Francis and Kaplan turbines are
reaction turbines. More details of these three turbines are given in following sections.

Wikipedia (2019b) lists a number of different types of water turbines as shown in Table 22.2.

584
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

Table 22.2 Types of water turbines (Wikipedia 2019b)

Reaction turbines Impulse turbines
VLH turbine Water wheel
Francis turbine Pelton wheel
Kaplan turbine Turgo turbine
Tyson turbine Cross-flow turbine
Deriaz turbine Jonval turbine
Gorlov helical turbine Reverse over-shot water
wheel
Bulb turbine Screw turbine
Barkh turbine

22.7.1 Pelton wheel turbines

Lester Allan Pelton invented the Pelton turbine. “The first operational Pelton wheel was installed at the
Mayflower Mine in Nevada City, USA in 1878.” (Crawford 2012). By 1900, over 11,000 turbines were
in use. (Wikipedia 2019c). Today the largest Pelton units exceed 400 MW. In a Pelton turbine, very high
velocity streams of water are directed by nozzles at the buckets on a wheel. The Pelton runner has buck-
ets and nozzle valves that direct jets of water onto a rotary series of spoon-shaped buckets from the pen-
stock/spiral case. The wheel spins and the momentum of the water jets is transferred to the wheel. Water
flow is split and deflected left and right into the dual buckets (see Figure 22.4). “This simple idea easily
doubled the efficiency of the standard water wheel, raising it from 30-40% to as much as 90%....... This
unique design revolutionized the use of turbines in high-energy sites, such as small, fast-running moun-
tain streams.” (Crawford 2012). Virtually all of the kinetic energy of the water jets is transferred to the
wheel. Spear-jets via needle valves are used to regulate the flow rate through both Pelton wheel turbines
and Turgo turbines (Renewables First 2019).

Figure 22.4 Photograph of a Pelton wheel turbine and generator (General Electric 2019a)

22.7.2 Turgo turbine

“The Turgo turbine is an impulse water turbine designed for medium head applications.” (Wikipedia
2019e). It was invented in 1919 by Eric Crewdson (Crewdson 1922) (see Figure 22.5) and has certain

585
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

advantages over Francis and Pelton wheel turbines for certain applications. It “provides a unique and
novel solution to increasing the capacity of a hydraulic impulse turbine while maintaining the nozzle
and spear injector system (as used in Pelton turbines) for flow regulation” (Benzon et al. 2016). The
Turgo turbine is a side entry impulse turbine (Gilkes 2019a). “The jet on the Turgo enters the runner at
an acute angle (fixed at 20 degrees), passes through the wheel, and discharges from the other side.
(Gilkes 2019b). “It converts the kinetic energy from the water jet into mechanical energy” (Turbines
Info 2019). The turbine is protected from overspeed conditions using jet deflectors, which also protects
the penstock from surge pressures in a load loss event (Benzon et al. 2016). Typically it runs at twice the
speed of a Pelton turbine when operating at the same head (Gilkes 2019a). It is less expensive than a
Pelton wheel turbine and does not need to be airtight like a Francis turbine. Water leaves the turbine at
atmospheric pressure. Interference of the outflow with the runner and jets is minimized, which is a
known problem for Pelton turbines (Benzon et al. 2016). The specific speed is higher than a Pelton
wheel and thus can handle a greater flow with the same diameter (Wikipedia 2019e). A family Gilkes
company still produces Turgo turbines 100 years after they were invented (Gilkes 2019a). Other
companies also produce Turgo turbines. The Turgo turbine has been installed in thousands of locations
across the globe (Benzon et al. 2016). However, “the current deployment of Turgo turbines is rather
restricted which could be attributed to the more difficult fabrication of its complex runner and the
somewhat lower hydraulic efficiency compared to Pelton turbines” (Benzon et al. 2016).

Figure 22.5 Photograph of a Turgo turbine runner (Gilkes 2012b)

22.7.3 Francis turbine

“The Francis turbine is a type of water turbine that was developed by James B. Francis in Lowell, Mas-
sachusetts” (Wikipedia 2019d). Francis developed the turbine in the 1850s. A Francis turbine has the
following elements: 1) wicket gates which adjust their position to control the amount of flow released
into the turbine from the penstock 2) a spiral case (or volute case) to distribute the flow evenly around
the circular runner 3) a stay ring of fixed blades 4) the Francis runner with a series of blades that rotate
to produce a torque and 5) a draft tube to reduce the velocity of the discharged water to minimize the
loss of kinetic energy. A Francis runner has more blades (16 to 24) than a Kaplan turbine therefore has
more friction loss. “Francis turbines are the most common water turbine in use today” Wikipedia
(2019d). Unlike a Pelton turbine, the Francis turbine is best completely filled with water Wikipedia
(2019d).

586
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

Figure 22.6 Photograph of a Francis turbine runner (General Electric 2019b)

Figure 22.7 Photograph of a Francis turbine runner and generator (General Electric 2019c)

22.7.4 Kaplan turbine

The Kaplan turbine with adjustable blades was developed in 1913 by Viktor Kaplan in Austria. (Wikipe-
dia 2019). A Kaplan turbine has adjustable blades that can be positioned to maximize the efficiency of
the turbine. These are expensive. “The Kaplan turbine was an evolution of the Francis turbine. ...Viktor
Kaplan, who combined automatically adjusted propeller blades with automatically adjusted wicket

587
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

gates to achieve efficiency over a wide range of flow and water level.” (Wikipedia 2019). The efficiency
of a Kaplan turbine is better than a Francis turbine. However, difficulties with cavitation delayed the
first use of the turbine. To avoid cavitation, the Kaplan runner needs to be submerged which can increase
the Civil Works costs for the project.

Figure 22.8 Photograph of a Kaplan turbine runner (Hydro Quebec 2019)

Figure 22.9 Photograph of a Kaplan turbine runner, wicket gates and turbine generator shaft and
generator (General Electric 2019d)

588
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

22.7.5 Ranges of parameters for turbines

Head, discharge, rotational speed, power and specific speed ranges for various types of turbines for both
SI Units and US customary units are given in Table 22.3 and Table 22.4. The definition of specific speed
for a turbine (or power specific speed - Encyclopedia Britannica 2019) is given in Section 22.6 as
12 54
N s =  N PT   HT . For selection of the appropriate rotational speed, four poles (Krueger 1959)or
a larger number is preferred for the generator. Turbine efficiencies are given in Table 22.5.

Table 22.3 Turbine head and specific speed ranges (SI units) (Wikipedia 2019)
Type of Turbine head Turbine power Rotational speed Specific
turbine range range range *** speed
HT PT N (NS in SI units)
(m) (kW) (rpm)
Kaplan 2 m < HT < 70 m** 75 kW to 330000 70 to 430 rpm 110 < NS < 5500
kW
Francis 40 m < HT < 600 m+ Few kW up to 75 to 1000 rpm+ 17 < NS < 2900
800,000 kW
Turgo 15 m < HT < 300 m Up to 10000 kW
Pelton 80 m < HT < 1600 m Up to 400,000 kW 65 to 800 rpm 8 < NS < 730
wheel

* Encyclopedia Britannica (2019) ** These ranges may vary considerably depending on the source. ***

Table 22.4 Turbine head and specific speed ranges

(US customary units converted from Table 22.3) (Wikipedia 2019)
Type of Turbine head Turbine power range (HP) Specific
turbine range speed
HT (NS in US
(ft) customary units-rpm,
HP, ft)
Kaplan ft < HT < ft** 13,400 HP < PT < 1,670,000 HP 28 < NS < 1430
Francis 130 ft < HT < 2000 ft 16,000 HP < PT < 390,000 HP 4.5 < NS < 755

Turgo 50 ft < HT < 1000 ft Up to 12800 HP

Pelton 260 ft < HT < 4600 ft 16,000 HP < PT < 390,000 HP 2 < NS < 190
wheel

Mane, N. SlideShare (12 Feb. 2017) +Wikipedia (2019d).

589
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

Table 22.5 Turbine efficiencies (First Renewables 2019)

Type of turbine Efficiency (%)
Kaplan > 90%
Francis 85-90% >90%
Turgo 87%
Pelton wheel 90%

* Expensive at this smaller size.

Example 22.1 Consider the following hydropower generating systems. Compute the specific speeds and select an
appropriate type of turbine for each. (i) H T = 1000 meters , Q = 2000 L/s , Efficiency = 0.9, N = 600 rpm
(ii) H T = 3280 ft , Q = 71cfs , Efficiency = 0.9, N = 600 rpm (iii) H T = 200 meters , Q = 10000 L/s ,
Efficiency = 0.85, N = 500 rpm (iv) H T = 660 ft , Q = 350cfs , Efficiency = 0.85, N = 500 rpm
(v) H T = 30 meters , Q = 100000 L/s , Efficiency = 0.90, N = 200 rpm (vi) H T = 100 ft , Q = 3500cfs,
Efficiency = 0.90, N = 200 rpm

Answer – (i) Determination of the turbine specific speed. H T = 1000 meters , Q = 2000 L/s , Efficiency = 0.9,
N = 600 rpm

Compute the turbine power

17620200
P T = Q H T = 0.9  9789 2.0  1000  = 17620200 = ------------------------ = 17620.2 kW
1000

The specific speed is:

12 0.5
N PT  17620  - = 14
N s = ------------------- = 600
----------------------------------
54 1.25
HT 1000

From Table 22.3 the selected turbine is a Pelton wheel. Very high head. Relatively low discharge.

(ii) Determination of the turbine specific speed. H T = 3280 ft , Q = 71cfs , Efficiency = 0.9, N = 600 rpm

Compute the turbine power

P T = Q H T = 0.9  62.4 71  3280  = 13078541 = 13078541

------------------------ = 23779 HP
550

The specific speed is:

12 0.5
N PT  23779  - = 3.6
N s = ------------------- = 600
----------------------------------
54 1.25
HT 3280

From Table 22.4 the selected turbine is a Pelton wheel. Very high head. Relatively low discharge.

(iii) Determination of the turbine specific speed. H T = 200 meters , Q = 10000 L/s , Efficiency = 0.85,
N = 500 rpm

Compute the turbine power

16641300
P T = Q H T = 0.85  9789 10.0  200  = 16641300 = ------------------------ = 16641 kW
1000

The specific speed is:

12 0.5
N PT 500  16641 
N s = ------------------- = ----------------------------------- = 86
54 1.25
HT 200

From Table 22.3 the selected turbine is a Francis turbine. Moderately high head. Moderate discharge.

590
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

(iv) Determination of the turbine specific speed. H T = 660 ft , Q = 350cfs , Efficiency = 0.85, N = 500 rpm

Compute the turbine power

P T = Q H T =  0.85   62.4 350  660  = 12252240 = 12252240

------------------------ = 22277 HP
550

The specific speed is:

12 0.5
N PT  22277  - = 22.3
N s = ------------------- = 500
----------------------------------
54 1.25
HT 660

From Table 22.4 the selected turbine is a Francis turbine. Moderately high head. Moderate discharge.

(v) Determination of the turbine specific speed. H T = 30 meters , Q = 100000 L/s , Efficiency = 0.90,
N = 200 rpm

Compute the turbine power

26430300
P T = Q H T = 0.9  9789 100.0  30  = 26430300 = ------------------------ = 26430 kW
1000

The specific speed is:

12 0.5
N PT 200  26430 
N s = ------------------- = ----------------------------------- = 463
54 1.25
HT  30 

From Table 22.3 the selected turbine is a Kaplan turbine. Low head. Large discharge.

(vi) Determination of the turbine specific speed. H T = 100 ft , Q = 3500cfs, Efficiency = 0.90, N = 200 rpm

Compute the turbine power

P T = Q H T = 0.9  62.4 3500  100  = 19656000 = 19656000

------------------------ = 35738 HP
550

The specific speed is:

12 0.5
N PT 200  35738 
N s = ------------------- = ----------------------------------- = 119.6
54 1.25
HT 100

From Table 22.4 the selected turbine is a Kaplan turbine. Low head. Large discharge.

_________________________________________________________________________________

22.7.6 Turbine rotational speeds

A turbine should rotate at an exact rotational speed to ensure that all generating units in a power system
are synchronized (Hydro Quebec 2019). The exact rotational speed for a turbine depends on the number
of poles in the generator which can be as little as 4 and as large as approximately 100. In Australia, Brit-
ain and Europe the frequency of electricity generated is 50 Hz. In contrast, in North America the fre-
quency is 60 Hz. The rotational speed of a turbine may for calculated from (for Australia) as:

3000
N = ---------------- (22.5)
N Pairs

where

• N = rotational speed (rpm)

• N Pairs = number of pairs of electromagnetic poles in the generator

591
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

For the USA or Canada the 3000 value would need to be replaced by 3600 on the right hand of
Equation 22.5. An example (Hydro Quebec 2019) of the computation of the rotational speed is given
below.

Example 22.2 Compute the rotational speed of a turbine that has 32 pairs of poles (or electromagnets) where the tur-
bine is located in Canada (at La Grande-3 generating station.

Answer – (i) Determination of the turbine rotational speed

Using Equation 22.5 adjusted for North American conditions gives:

3600 3600
N = ---------------- = ------------ = 112.5 rpm
N Pairs 32

at a frequency of 60 Hz alternating current.

For the various numbers of pairs of poles see the Table below for the corresponding rotational
speed of a turbine located in Australia.

Table 22.6 Turbine rotational speeds (50 Hz frequency)

Number of pairs of Rotational speed
poles (rpm)
3 1000
6 500
8 375

22.8 Energy Equation Applied to a Hydropower Turbine System

Applying the energy equation (Equation 3.67) across a turbine in a system such as that is shown in
Figure 22.10 the turbine head HT r removed from the fluid by the turbine must be calculated from

2 2
p V1 p V2
z 1 + -----1- + --------
- = z 2 + -----2- + --------
- + H T +  h f +  hL (22.6)
 2g  2g

The head across the turbine is the difference in EGL between the inlet to the spiral case (IN) and the out-
let of the draft tube (OUT) or:

2 2
 p IN V IN   p OUT V OUT 
HT =  z IN + --------
- + ------------ –  z OUT + -------------
- + ----------------- (22.7)
  2g    2g 

592
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

Figure 22.10 Hydropower penstock-turbine-draft tube showing turbine head H T

and static head H S = z 1 – z 2

The penstock/tunnel pipework system curve for a hydropower system depends on the difference
between the elevations of the upstream and downstream reservoirs as well as the losses in the penstock/
tunnel and fittings attached to both sides of the turbine. The static head for a hydropower system is
defined as the difference in elevation between the upper and the lower reservoirs as shown in
Figure 22.10 (or from the upper reservoir to the centerline of the Pelton wheel that discharges to the
atmosphere):

H S = z1 – z2 (22.8)

The static head HS and friction losses in the penstock/tunnel/draft tube pipe work must both be taken
into account in developing the system curve for the penstock/tunnel system. The static head or the eleva-
tion difference between the upstream reservoir and the downstream reservoir is not a function of flow Q
through the system. The friction losses are losses due to pipe friction and minor losses vary approxi-
mately with the square of the flow. These friction losses (and also any minor losses) must be subtracted
from the static head as they will not be available to generate electricity at the turbine. From the Darcy –
Weisbach equation (Equation 4.3) the total friction loss for the pipe work in series in the penstock/tun-
nel of the hydropower generating system is given by

2 2
LV L Q
f =  f ---
- ------ =  f ---- -----------
D 2g D 2g A
(22.9)

The hydraulic losses in the draft tube may be counted in determining the head across the turbine H T ,
thus they may need to be excluded when computing the system curve.

Minor losses for multiple losses in the system from Equation 5.1 are given by

2
 hL=  K V
------
2g

The system curve for the pipework associated with hydropower generating system is given as
2 2
fL Q V
penstock/tunnel system curve = H S –  ------ ------------- –  K ------ (22.10)
D 2g A 2 2g

PENSTOCK/TUNNEL PIPEWORK SYSTEM CURVE

593
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

Example 22.3 Consider the following hydropower generating system with a penstock passing a flow of
Q = 2500 L/s connecting an upper reservoir to a turbine with a penstock diameter D = 1290 mm , penstock
–6 2
length L = 4000 m , pipe roughness  = 0.25 mm and water viscosity  = 1.141 10 m /s . The penstock –
turbine system connects two reservoirs with the difference in water surface elevations of the lower and upper reser-
voirs of 105 m (thus the static head is H S = 105.0 meters ).

(i) Determine the turbine head H T

Answer – (i) Determination of the turbine head.

The turbine head is based on both the static head and the friction loss in the pipeline. The flow is 2500 L/s.

The area of the pipe is

2 2
D   1.290  2
A = ---------- = ------------------------ = 1.3070 m
4 4

The velocity of flow is

 1000- = 1.913 m/s

V = Q
---- = 2500
---------------------------
A 1.3070

The Reynolds number is

VD  1.913   1.29 
Re = -------- = ---------------------------------- = 2162813
 1.141 10
–6

The Darcy – Weisbach friction factor from the Swamee and Jain equation (Equation 4.61) is

0.25 0.25
f = ---------------------------------------------------------
- = ----------------------------------------------------------------------------------------
- = 0.0142
 5.74 2 0.25 5.74 2
log 10  ------------ + ------------ log 10  ------------------------ + -------------------------------
 3.7 D 0.9  3.7  1290  0.9
Re  2162813 

The head loss based on the Darcy – Weisbach head loss formula (Equation 4.3) is

2 2
fL V 0.0142  4000   1.913 
h f = ------ ------ = --------------------------------- -------------------- = 8.213 m
D 2g 1.29 2  9.81 
The turbine head is

H T = H S – h f = 105.0 – 8.213 = 96.8 m

_________________________________________________________________________________

22.9 Turbine Performance Characteristics

Basic performance factors for a turbine include

• turbine head HT usually in meters (m) for SI units (or feet — ft — in US customary units) of
water. This is the head extracted from the water by the turbine to generate electricity.
• turbine flow Q in L/s (or US gpm or cfs)
• efficiency  taken as a decimal fraction between 0 and 1 (sometimes referred to as Ep)
• shaft power output from the turbine P in kW (or HP)
• rotational speed N in rpm
22.9.1 Turbine curves
Three curves are important for consideration of turbines:

• HP versus Q — called the turbine curve (often a series of curves each at different wicket gate
openings)
•  versus Q — called the efficiency curve
• P versus Q — called the power curve

594
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

These characteristics apply to just one value of the impeller rotational speed N — and this value should
always be specified on the drawings. Generally manufacturers do not provide these curves, however,
characteristic curves are provided in another form which can be transformed into the curves above.

Different classes of turbines (Francis, Pelton or impulse and mixed flow and Kaplan flow) have different
shaped curves.

The point of maximum turbine efficiency is called the best efficiency point (BEP) - similar to the defini-
tion for pumps in Chapter 18. The most common operating point for the turbine should be close to the
BEP to maximize the power output from the generator.

22.9.2 Definition of water or hydraulic power for a turbine

Water or hydraulic power is defined as the amount of power the water a turbine runner receives that will
be transferred to shaft power and consequentially electricity from the rotation of the generator. It is
defined as the following product

QH T (22.11)

TURBINE WATER OR HYDRAULIC POWER

where

•  = specific weight of fluid flowing (N/m3or lb/ft3)

• Q = turbine flow (m3/s or ft3/s)
• HT = turbine head (m or ft)
This is similar to the definition of water or hydraulic power for a pump (see Equation 18.2).

22.9.3 Turbine efficiency

The turbine efficiency  is defined as

PT
 = shaft power output from turbine- = ---------------
--------------------------------------------------------------------------- (22.12)
water power QH T

TURBINE EFFICIENCY

where

•  = turbine efficiency (dimensionless or expressed as a percentage)

• P T = shaft power output delivered by turbine (W or HP - watts are usually converted to kW
by dividing by 1000)
• QH T = turbine water or hydraulic power (W or HP)

Note that unlike for a pump, for a turbine, the water power QH P is greater than the turbine shaft output
power P T .

22.9.4 Turbine shaft power output

Thus the turbine shaft power output is defined as

P T = QH T (22.13)

TURBINE SHAFT POWER OUTPUT

595
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

where

• P T = shaft power delivered by turbine (W or HP - watts are usually converted to kW by

dividing by 1000)
•  = efficiency of the turbine (usually specified as a decimal, e.g. 0.88 but sometimes
referred to as a percentage). Efficiency gives a measure of the turbine hydraulic and
mechanical losses in transferring water power from the runner to the shaft. (dimensionless)
•  = specific weight of fluid flowing (N/m3or lb/ft3 e.g. 9789 N/m3 for water at 20C)
• Q = turbine flow (m3/s or ft3/s – e.g. 1.2 m3/s – a typical medium sized Pelton wheel tur-
bine)
• HT = turbine head (m or ft – e.g. 170 m for a typical medium sized Pelton wheel turbine
with a high head)
Units of right – hand side of Equation 22.13 for the turbine output power to the shaft are as follows:

3
N m N-m
 1  ------- -------  m  = ----------- = watts or W
m
3 s s

The usual units of power from a turbine are kilowatts or kW.

Output power from the generator is given by:

P
P T = ------G- (22.14)
EG

where

• P T = power delivered from the turbine shaft to the generator (W or HP); note this is larger
than the electrical power delivered to the electrical grid by the generator P e
• P G = electrical power delivered by the generator (W or HP)
• EG = generator efficiency (usually 0.95 to 0.98) (dimensionless)

22.9.5 Turbine energy output

The energy provided the turbine runner to the shaft of the generator over a period of time t is defined
as

E T = P T t (22.15)

TURBINE ENERGY OUTPUT

where

• E T = energy delivered by the shaft of the turbine (Wh - usually converted to kWh by divid-
ing by 1000)
• P T = shaft power delivered by turbine to the generator (W or HP)
• t = time over which turbine operates (h or hours)

For Equation 22.15 to be valid the turbine power output must remain constant for the entire period t .
In turn this requires the head across the turbine, the flow through the turbine and the upstream and
downstream reservoir levels to remain constant. In many cases, the guide vanes are adjusted by a PID
governor to ensure the power output is kept constant. In order to correctly assess the energy output from
a turbine it would necessitate that an extended period simulation hydraulic model be run at say a time

596
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

step of 5 minutes over the t period. The reservoir levels, the turbine flow and hence the head across
the turbine would be updated every 5 minutes to update the power calculation to be the correct value.
The energy usage would be accumulated for all of the 5 minute periods in t . The process here would
be very similar to that described for pumps in Section 18.6.5.

22.9.6 Electrical energy output

The electrical energy provided by the generator connected to the turbine depends on the efficiency of the
generator and is given by:

P T t
EE T = ------------
- (22.16)
EG

GENERATOR ELECTRICAL ENERGY OUTPUT

where

• EE T = electrical energy delivered by the generator (Wh) - usually converted to kWh by

dividing by 1000
• P T t = E T = energy delivered to the turbine shaft (Wh)
• E G = efficiency of generator (usually around 95%)
• P T = shaft power delivered by the turbine runner to the shaft (W or HP)
• t = time over which turbine operates (h or hours)

22.9.7 Monetary benefits from turbine operation

The benefits from selling the electricity output from a hydropower plant by operating a turbine over a
period of time t and assuming the turbine flow conditions, turbine settings as well as the upstream and
downstream reservoir heads do not vary (as described in Section 22.9.5) is given by:

Benefit = EE T  p  (22.17)

MONETARY BENEFIT OF TURBINE OPERATION

where

• Benefit = monetary benefit or income of selling the electrical energy from the turbine oper-
ation ($)
• EE T = electrical energy delivered by the generator (Wh - usually converted to kWh by
dividing by 1000)
• p = price of sale of electricity (cents per kWh - converted to $/kWh by dividing by 100)

Example 22.4 Compute the following (i) the shaft power for a turbine head of H T = 170 m , a turbine flow of
Q = 1200 L/s at the operating point and a turbine efficiency of  = 0.88 (ii) the electrical energy output from the
generator for 5 hours of continuous operation at constant generating conditions (iii) the monetary benefit of the sale
of electricity for turbine operation assuming the turbine is operating from 4pm to 9pm on an peak electricity tariff of
42 cents/kWh. Assume a generator efficiency of 97%.

Answer – (i) Computation of the shaft power for a turbine.

The output shaft power from a turbine is given by Equation 22.13 as

P T = Q H T = 0.88  9789   1.2   170  = 1757321 W = 1757321

--------------------- kW = 1757 kW
1000

597
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

(ii) Computation of energy.

The turbine shaft energy over 5 hours is given from Equation 22.15 as:

E T = Pt = 1757  5  = 8785 kWh

The electrical energy output from the generator is given from Equation 22.16 by:

EE T = E G  Pt  = 0.97  8785  = 8521 kWh

(iii) Computation of monetary benefit from the sale of electricity from the generator.

The monetary benefit of the sale of electrical energy from the generator is given from Equation 22.17 by:

42
Benefit = EE T  p  = 8521  --------- = $3579
 100

___________________________________________________________________________________

22.10 Pumped hydroelectricity (or pumped hydro)

Pumped hydropower storage (PHS) technology was first applied at the beginning of the 20th century.
Significant construction of PHS occurred after the second world war (IHA 2018). PHS generates electri-
cal energy by using water that has been previously pumped from a lower reservoir to an upper reservoir.
“This is one of a few methods that allow temporary excess electrical capacity to be stored for later utili-
zation.” (Wikipedia 2019d),

There are two considerations in relation for PHS including (IHA 2018):

• the power generated in MW (e.g. 2000 MW)

• the energy storage capacity in MWh (e.g 10,000 MWh) - this depends on size of the upper
storage
Solar and wind renewable energy sources can be used to pump water uphill from the lower reservoir to
the upper reservoir in PHS (IHA 2018). When electrical energy demand increases (say during peak hour
demand in the evening), water is passed from the upper reservoir downhill through the turbines to gener-
ate electricity. The combination of variable renewable energy with pumped hydro can provide electricity
that is reliable, dispatchable and low carbon.

22.10.1 The Snowy 2.0 pumped hydro project

The Snowy 2.0 pumped hydro project is proposed to connect 2 existing dams with tunnels to utilize the
678 meters head difference and thereby will be able to produce 2000 MW of electrical power for a
period of up to 5 hours. The objective is to provide storage that will complement the intermittency of
wind and solar generating capacity (snowyhydro 2017). The significant increasing supply of electricity
coming from renewables means that there can be electricity shortages when the sun does shine and the
wind does not blow. Pumped hydro absorbs supply-side volatility and resolves the intermittency of
renewables (snowyhydro 2017). Adequate and deep transmission augmentation will be needed to
increase the capacity of the transmission lines between the Scheme and the load centers in NSW and
Victoria (snowyhydro 2017).

22.11 Water hammer control for hydropower systems

Water hammer due to changes in power output as a result of changes in flow, load rejection and runaway
conditions all must be dealt with in hydropower systems. Surge tanks and air vessels are commonly used
to absorb water hammer pressure changes. Deflectors in Pelton wheel turbine and Turgo turbine systems
allow water jets to be deflected away from the runner so that valves can be operated slowly to minimize
the pressure rises in the penstock/tunnel pipework that deliver the water to the hydropower plant.

22.12 References
Benzon, D.S., Aggidis, G.A. and Anagnostopoulos, J.S. (2016). “Development of the Turgo Impulse
Turbine: Past and Present.” Applied Energy, Elsevier, vol. 166(C), pages 1-18.

598
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

Crawford, M. (2019). “Lester Allan Pelton” The American Society of Mechanical Engineers, https://
www.asme.org/topics-resources/content/lester-allan-pelton. Accessed 13 November 2019.

Crewdson E. “Design and performance of a new impulse water-turbine.” In: Minutes of Proceedings of
the Institution of Civil Engineers; 1922. p. 396–407.

Electropaedia (2019). “Hydroelectric Power.” https://www.mpoweruk.com/hydro_power.htm Accessed

15 November 2019

Encyclopedia Britannica (2019) https://www.britannica.com/technology/turbine/Turbine-selection-on-

the-basis-of-specific-speed Accessed 13 November 2019

General Electric (2019a). https://www.ge.com/content/dam/gepower-renewables/global/en_US/down-

loads/3d-still-images/hydro-cgi-pelton-generating-unit-3000px.jpg Accessed 20 November 2019

General Electric (2019b). https://www.ge.com/content/dam/gepower-renewables/global/en_US/down-

loads/photographs/hydro-canada-sorel-tracy-francis-turbine-lifted-3000px.jpg Accessed 20 November
2019

General Electric (2019c). https://www.ge.com/content/dam/gepower-renewables/global/en_US/down-

loads/3d-still-images/hydro-cgi-francis-turbine-generator-3000px.jpg Accessed 20 November 2019

General Electric (2019d). https://www.ge.com/content/dam/gepower-renewables/global/en_US/down-

loads/3d-still-images/hydro-cgi-kaplan-generating-unit-3000px.jpg Accessed 20 November 2019

Gilkes (2019a). https://www.gilkes.com/hydropower-systems-manufacturers/gilkes-turbines/turgo-tur-

bines Accessed 19 November 2019

Gilkes (2019b). “Gilkes Turgo Impulse Hydro Turbine” pdf downloaded from https://www.gilkes.com/
hydropower-systems-manufacturers/gilkes-turbines/turgo-turbines <http://www.gilkes.com/user_up-
loads/turgo%20paper2.pdf> Accessed 19 November 2019.

Hoes, O.A.C., Meijer, L.J.J, van der Ent, R.J., and van de Giesen, N.C. (2017). “Systematic high-resolu-
tion assessment of global hydropower potential.” PLoS One, 2017, Vol. 12(2), pp.urn:issn:1932-6203

Hydro Quebec (2019). “Turbine-generator unit” http://www.hydroquebec.com/learning/hydroelectric-

ite/turbine-alternateur.html Accessed on 19 December 2019.

IEC (1999). “Hydraulic Turbines, Storage Pumps and Pump-Turbines—Model Acceptance Tests,” 2nd
ed., International Electrotechnical Commission, Geneva, Switzerland,, Standard No. IEC 60193:1999.

International Hydropower Association -IHA (2018). “The world’s water battery: Pumped hydropower
storage and the clean energy transition.” IHA Working Paper, Dec. 2018.

Krueger, R. E. (1959). Selecting Hydraulic Reaction Turbines, Technical Information Branch, USBR.

National Geographic 2019. https://www.nationalgeographic.com/environment/global-warming/hydro-

power. Accessed 12 November 2019.

Rehman, S. Al-Hadhrami, L.M and Alam, M.M. (2015). “Pumped hydro energy storage system: a tech-
nological review” Renewable Sustainable Energy Review, 44 (2015), 586-598

Renewables First (2019). https://www.renewablesfirst.co.uk/hydropower/hydropower-learning-centre/

pelton-and-turgo-turbines/ Accessed 19 November 2019.

snowyhydro (2109). https://www.snowyhydro.com.au/our-energy/hydro/the-history/ Accessed 21

November 2019.

snowyhydro (2017). “Snowy 2.0 Feasibility Study Report Volume 1.” https://drive.google.com/
uc?id=1pYUvSCuktSxDNDEhxma4j_-mtGtIttZD&export=download

Teach Nuclear, Canada (2019). https://teachnuclear.ca/all-things-nuclear/energy-demand-and-sources/

present-energy-sources/hydroelectricity/ Accessed 15 November 2019.

599
 Chapter 22 - File: Chap22-Turbines.fm July 6, 2022 Version 6

Turbines Info (2019). https://www.turbinesinfo.com/turgo-turbines/ Accessed 19 November 2019.

University of Manchester (2019). “TOPIC T4: PUMPS AND TURBINES.” https://www.google.com/

url?sa=t&rct=j&q=&esrc=s&source=web&cd=11&ved=2ahUKEwjGhpik3ejlAhVFILcAHWm-
DqsQFjAKegQIARAC&url=https%3A%2F%2Fsiteproxy.ruqli.workers.dev%3A443%2Fhttps%2Fpersonalpages.manchester.ac.uk%2Fstaff%2Fdavid.d.
apsley%2Flectures%2Fhydraulics2%2Ft4.pdf&usg=AOvVaw0KItYwj9z9iuuqCQFzcgk7 Accessed 14
November 2019.

USGS (2019). https://www.usgs.gov/media/images/a-turbine-connected-a-generator-produces-power-

inside-a-dam Accessed 19 December 2019.

Wikipedia (2019a). “Kaplan turbine.” https://en.wikipedia.org/wiki/Pelton_wheel Accessed 15 Novem-

ber 2019.

Wikipedia (2019b). “Water turbine.” https://en.wikipedia.org/wiki/Water_turbine Accessed 13 Novem-

ber 2019.

Wikipedia (2019c). “Pelton wheel.” https://en.wikipedia.org/wiki/Kaplan_turbine Accessed 13 Novem-

ber 2019.

Wikipedia (2019d). “Francis turbine.” https://en.wikipedia.org/wiki/Francis_turbine#cite_note-1

Accessed 15 November 2019.

Wikipedia (2019e). “Turgo turbine.” https://en.wikipedia.org/wiki/Turgo_turbine Accessed 19 Novem-

ber 2019.

600
 Appendix A - File: AppendixA - New.fm July 6, 2022 Version 6

APPENDIX A PROPERTIES OF WATER

A.1 Density, Specific Weight, Viscosity, Bulk Modulus of Elasticity,

Vapor Pressure and Surface Tension

601
 Appendix A - File: AppendixA - New.fm July 6, 2022 Version 6

Table A.1 Properties of Water (SI Units)

T Density Specific Bulk Absolute Kinematic Vapor Vapor Surface T
Weight Modulus Viscosity Viscosity* pressure Pressure Tension
deg rho gamma K mu nu pv (abs.) Head sigma deg
(abs.)
°C (kg/m^3) (N/m^3) (Pa) (Pa.s) (m^2/s) (kPa) (m) (mN/m) °C
0 999.9 9809 1.968E+09 1.792E-03 1.791E-06 0.612 0.06 75.6 0
1 999.9 9809 1.982E+09 1.731E-03 1.731E-06 0.658 0.07 75.5 1
2 1000.0 9810 1.996E+09 1.674E-03 1.674E-06 0.706 0.07 75.4 2
3 1000.0 9810 2.010E+09 1.620E-03 1.620E-06 0.759 0.08 75.2 3
4 1000.0 9810 2.023E+09 1.568E-03 1.568E-06 0.814 0.08 75.1 4
5 1000.0 9810 2.036E+09 1.519E-03 1.520E-06 0.873 0.09 74.9 5
6 1000.0 9810 2.049E+09 1.473E-03 1.473E-06 0.936 0.10 74.8 6
7 999.9 9809 2.061E+09 1.428E-03 1.429E-06 1.003 0.10 74.7 7
8 999.9 9809 2.073E+09 1.386E-03 1.387E-06 1.074 0.11 74.5 8
9 999.8 9808 2.085E+09 1.346E-03 1.346E-06 1.149 0.12 74.4 9
10 999.7 9807 2.096E+09 1.307E-03 1.308E-06 1.229 0.13 74.2 10
11 999.6 9806 2.107E+09 1.270E-03 1.271E-06 1.314 0.13 74.1 11
12 999.5 9805 2.118E+09 1.235E-03 1.236E-06 1.404 0.14 73.9 12
13 999.4 9804 2.129E+09 1.201E-03 1.202E-06 1.499 0.15 73.8 13
14 999.3 9803 2.139E+09 1.169E-03 1.170E-06 1.600 0.16 73.6 14
15 999.1 9801 2.149E+09 1.138E-03 1.140E-06 1.707 0.17 73.5 15
16 999.0 9800 2.159E+09 1.109E-03 1.110E-06 1.820 0.19 73.3 16
17 998.8 9798 2.168E+09 1.080E-03 1.082E-06 1.939 0.20 73.2 17
18 998.6 9796 2.178E+09 1.053E-03 1.055E-06 2.066 0.21 73.0 18
19 998.4 9794 2.187E+09 1.027E-03 1.029E-06 2.200 0.22 72.9 19
20 998.2 9792 2.195E+09 1.002E-03 1.007E-06 2.341 0.24 72.7 20
21 998.0 9790 2.204E+09 9.780E-04 9.800E-07 2.490 0.25 72.6 21
22 997.8 9788 2.212E+09 9.550E-04 9.570E-07 2.647 0.27 72.4 22
23 997.6 9786 2.220E+09 9.320E-04 9.340E-07 2.813 0.29 72.3 23
24 997.3 9784 2.228E+09 9.110E-04 9.130E-07 2.987 0.31 72.1 24
25 997.1 9782 2.236E+09 8.900E-04 8.920E-07 3.172 0.32 72.0 25
26 996.8 9779 2.243E+09 8.700E-04 8.730E-07 3.366 0.34 71.8 26
27 996.5 9776 2.250E+09 8.510E-04 8.540E-07 3.570 0.37 71.7 27
28 996.3 9774 2.257E+09 8.320E-04 8.350E-07 3.785 0.39 71.5 28
29 996.0 9771 2.263E+09 8.150E-04 8.170E-07 4.011 0.41 71.3 29
30 995.7 9768 2.270E+09 7.970E-04 8.000E-07 4.249 0.43 71.2 30
31 995.4 9765 2.276E+09 7.810E-04 7.840E-07 4.499 0.46 71.0 31
32 995.1 9762 2.282E+09 7.640E-04 7.680E-07 4.762 0.49 70.9 32
33 994.7 9758 2.287E+09 7.490E-04 7.530E-07 5.038 0.52 70.7 33
34 994.4 9755 2.293E+09 7.340E-04 7.380E-07 5.328 0.55 70.6 34
35 994.1 9752 2.298E+09 7.190E-04 7.230E-07 5.632 0.58 70.4 35

602
 Appendix A - File: AppendixA - New.fm July 6, 2022 Version 6

Table A.1 Properties of Water (SI Units)

T Density Specific Bulk Absolute Kinematic Vapor Vapor Surface T
Weight Modulus Viscosity Viscosity pressure Pressure Tension
deg rho gamma K mu nu pv(abs.) Head sigma deg
(abs.)
°C (kg/m^3) (N/m^3) (Pa) (Pa.s) (m^2/s) (kPa) (m) (mN/m) °C
36 993.7 9748 2.303E+09 7.050E-04 7.090E-07 5.951 0.61 70.2 36
37 993.4 9745 2.308E+09 6.910E-04 6.960E-07 6.286 0.65 70.1 37
38 993.0 9741 2.313E+09 6.780E-04 6.830E-07 6.637 0.68 69.9 38
39 992.6 9737 2.317E+09 6.650E-04 6.700E-07 7.004 0.72 69.8 39
40 992.2 9733 2.321E+09 6.530E-04 6.580E-07 7.389 0.76 69.6 40
41 991.9 9731 2.326E+09 6.410E-04 6.460E-07 7.792 0.80 69.4 41
42 991.5 9727 2.329E+09 6.290E-04 6.350E-07 8.214 0.84 69.3 42
43 991.1 9723 2.333E+09 6.180E-04 6.240E-07 8.656 0.89 69.1 43
44 990.7 9719 2.337E+09 6.070E-04 6.130E-07 9.117 0.94 68.9 44
45 990.2 9714 2.340E+09 5.960E-04 6.030E-07 9.600 0.99 68.8 45
46 989.8 9710 2.343E+09 5.860E-04 5.930E-07 10.105 1.04 68.6 46
47 989.4 9706 2.346E+09 5.760E-04 5.830E-07 10.632 1.10 68.4 47
48 989.0 9702 2.348E+09 5.660E-04 5.730E-07 11.183 1.15 68.3 48
49 988.5 9697 2.351E+09 5.560E-04 5.640E-07 11.758 1.21 68.1 49
50 988.1 9693 2.353E+09 5.470E-04 5.550E-07 12.359 1.28 67.9 50
51 987.6 9688 2.355E+09 5.380E-04 5.460E-07 12.985 1.34 67.8 51
52 987.1 9683 2.357E+09 5.290E-04 5.380E-07 13.638 1.41 67.6 52
53 986.7 9680 2.359E+09 5.210E-04 5.290E-07 14.320 1.48 67.4 53
54 986.2 9675 2.361E+09 5.120E-04 5.210E-07 15.030 1.55 67.3 54
55 985.7 9670 2.362E+09 5.040E-04 5.140E-07 15.770 1.63 67.1 55
56 985.2 9665 2.364E+09 4.970E-04 5.060E-07 16.542 1.71 66.9 56
57 984.7 9660 2.365E+09 4.890E-04 4.990E-07 17.345 1.80 66.8 57
58 984.2 9655 2.366E+09 4.810E-04 4.910E-07 18.181 1.88 66.6 58
59 983.7 9650 2.367E+09 4.740E-04 4.840E-07 19.051 1.97 66.4 59
60 983.2 9645 2.367E+09 4.670E-04 4.770E-07 19.957 2.07 66.2 60
61 982.7 9640 2.368E+09 4.600E-04 4.710E-07 20.899 2.17 66.1 61
62 982.2 9635 2.368E+09 4.530E-04 4.640E-07 21.878 2.27 65.9 62
63 981.6 9629 2.368E+09 4.470E-04 4.580E-07 22.896 2.38 65.7 63
64 981.1 9625 2.369E+09 4.400E-04 4.520E-07 23.954 2.49 65.5 64
65 980.6 9620 2.369E+09 4.340E-04 4.460E-07 25.054 2.60 65.4 65
66 980.0 9614 2.368E+09 4.280E-04 4.400E-07 26.196 2.72 65.2 66
67 979.4 9608 2.368E+09 4.220E-04 4.340E-07 27.382 2.85 65.0 67
68 978.9 9603 2.367E+09 4.160E-04 4.290E-07 28.613 2.98 64.8 68
69 978.3 9597 2.367E+09 4.100E-04 4.230E-07 29.890 3.11 64.7 69
70 977.7 9591 2.366E+09 4.050E-04 4.180E-07 31.216 3.25 64.5 70
71 977.2 9586 2.365E+09 3.990E-04 4.130E-07 32.591 3.40 64.3 71
72 976.6 9580 2.364E+09 3.940E-04 4.080E-07 34.016 3.55 64.1 72
73 976.0 9575 2.363E+09 3.890E-04 4.030E-07 35.494 3.71 63.9 73

603
 Appendix A - File: AppendixA - New.fm July 6, 2022 Version 6

Table A.1 Properties of Water (SI Units)

T Density Specific Bulk Absolute Vapor Vapor Surface T
Weight Modulus Viscosity pressure Pressure Tension
deg rho gamma K mu pv (abs.) Head sigma deg
(abs.)
°C (kg/m^3) (N/m^3) (Pa) (Pa.s) (kPa) (m) (mN/m) °C
74 975.4 9569 2.361E+09 3.840E-04 3.980E-07 37.026 3.87 63.8 74
75 974.8 9563 2.360E+09 3.790E-04 3.930E-07 38.613 4.04 63.6 75
76 974.2 9557 2.358E+09 3.740E-04 3.880E-07 40.257 4.21 63.4 76
77 973.6 9551 2.356E+09 3.690E-04 3.840E-07 41.959 4.39 63.2 77
78 973.0 9545 2.355E+09 3.640E-04 3.790E-07 43.722 4.58 63.0 78
79 972.3 9538 2.352E+09 3.600E-04 3.750E-07 45.547 4.78 62.9 79
80 971.7 9532 2.350E+09 3.550E-04 3.710E-07 47.435 4.98 62.7 80
81 971.1 9526 2.348E+09 3.510E-04 3.670E-07 49.388 5.18 62.5 81
82 970.5 9521 2.346E+09 3.460E-04 3.630E-07 51.409 5.40 62.3 82
83 969.8 9514 2.343E+09 3.420E-04 3.590E-07 53.498 5.62 62.1 83
84 969.2 9508 2.341E+09 3.380E-04 3.550E-07 55.658 5.85 61.9 84
85 968.5 9501 2.338E+09 3.340E-04 3.510E-07 57.891 6.09 61.7 85
86 967.8 9494 2.335E+09 3.300E-04 3.470E-07 60.198 6.34 61.6 86
87 967.2 9488 2.332E+09 3.260E-04 3.440E-07 62.581 6.60 61.4 87
88 966.5 9481 2.329E+09 3.220E-04 3.400E-07 65.043 6.86 61.2 88
89 965.8 9474 2.326E+09 3.190E-04 3.370E-07 67.585 7.13 61.0 89
90 965.2 9469 2.322E+09 3.150E-04 3.330E-07 70.209 7.41 60.8 90
91 964.5 9462 2.319E+09 3.110E-04 3.300E-07 72.918 7.71 60.6 91
92 963.8 9455 2.315E+09 3.080E-04 3.270E-07 75.713 8.01 60.4 92
93 963.1 9448 2.312E+09 3.040E-04 3.230E-07 78.597 8.32 60.2 93
94 962.4 9441 2.308E+09 3.010E-04 3.200E-07 81.572 8.64 60.1 94
95 961.7 9434 2.304E+09 2.980E-04 3.170E-07 84.640 8.97 59.9 95
96 961.0 9427 2.300E+09 2.950E-04 3.140E-07 87.803 9.31 59.7 96
97 960.3 9421 2.296E+09 2.910E-04 3.110E-07 91.064 9.67 59.5 97
98 959.5 9413 2.292E+09 2.880E-04 3.080E-07 94.424 10.03 59.3 98
99 958.8 9406 2.288E+09 2.850E-04 3.050E-07 97.887 10.41 59.1 99
100 958.1 9399 2.283E+09 2.820E-04 3.030E-07 101.454 10.79 58.9 100

 – 16.921 + 396.13   107.41 + T  

* from IEC (1999) page 175; kinematic viscosity  = e

A.2 References
IEC (1999). “Hydraulic Turbines, Storage Pumps and Pump-Turbines—Model Acceptance Tests,” 2nd
ed., International Electrotechnical Commission, Geneva, Switzerland, Standard No. IEC 60193:1999.

604
 Chapter B - File: AppendixB - New.fm July 6, 2022 Version 6

APPENDIX B Units and Units Conversions

B.1 Units of derived quantities and conversions

Table B.1 Units of derived quantities and conversions

Quantity Symbol SI units US Conversion
and generic customary units
units or
non – SI units
Length L m (meter) ft (feet) 1 ft = 0.304800 m
[L] mm in. (inch) 1 in. = 25.40 mm
km mile 1 mile = 1.609 km
Area A m2, ha ft2, acre 1 acre = 660 x 660 ft2
[L2] (hectare) 1 acre = 4047 m2
1 ha = 10,000 m2
Volume V m3, L (liter), ft3, 1 gal = 0.003875 m3
[L3] ML gal (gallon – US), 1 US gallon = 3.875 L
(megaliters) acre – ft, 1 UK gallon = 4.54609 L
MG (million 1 acre – ft = 1233.48 m3
gallons) 1 MG (US) = 3.875 ML
Velocity V m/s fps (ft/sec) 1 ft/sec = 0.304800 m/s
[LT – 1]
Accelera- a m/s2 ft/sec2 1 ft/sec2 = 0.304800 m/s2
tion [LT – 2]
Flow or Q m3/s, L/s, cfs (ft3/s), gpm or 1 ft3/s = 28.3168 10– 3 m3/s
Discharge [L T – 1]
3
ML/day gal/min (gallons
1 ft3/s = 28.3168 L/s
per minute – US),
1 gal/min (US)
MGD (million 3
gallons per day – = 0.003785 m /min 1 MGD (US)
US) = 1.5473 ft3/s
1 ML/day = 11.574 L/s
Mass m kg slugs, tons 1 slug = 14.5939 kg
[M] 1 ton = 1016.05 kg (short ton)
Density  kg/m3 slugs/ft3 1 slug/ft3 = 515.379 kg/m3
[ML – 3]
Specific  N/m3 lb/ft3 1 lb/ft3 = 157.09 N/m3
weight [ML – 2T – 2]
Specific S dimensionless dimensionless same in all unit types
gravity [–]
Force F N lb or lbf, 1 lb = 4.44822 N
[MLT – 2] (newton) (pounds force), kgf 1 kgf = 9.80665 N
(kilograms force)

605
 Chapter B - File: AppendixB - New.fm July 6, 2022 Version 6

Table B.1 Units of derived quantities and conversions

Pressure p Pa (or N/m2) psi (lb/in2), 1 Pa = 1 N/m2
 (pascal) atm (atmospheres), 3
Normal 1 lb/in2 = 6.89476 10 Pa
 kPa bars
stress MPa 1 ft water = 2.988 kPa
[ML – 1T – 2] 1 atm = 101,325 Pa
Shear 1 atm = 14.6960 psi
stress 1 bar = 0.98692 atm
Vapor pv* Pa (or N/m2) psi (lb/in2) absolute 1 lb/in2 = 6.89476 103 Pa
pressure
[ML – 1T – 2] absolute1
bulk K Pa (N/m2) psi (lb/in2) 3
1 lb/in2 = 6.89476 10 Pa
modulus [ML – 1T – 2]
of
elasticity
Viscosity  N – s/m2 lb – s/ft2, 1 Poise = 0.1 N – s/m2
(absolute [ML T ]
–1or Pa – s, kg/m –
–1 Poise, 1 lb – s/ft2 = 47.8803 N – s/m2
or s slugs/ft – s 1 slug/ft – s = 47.8803 kg/m – s
dynamic)
Kinematic  m2/s ft2/s, –3
1 ft2/s = 92.9030 10 m2/s
viscosity [L2T – 1] centistoke
1 ft2/s = 92903 centistokes
Rota- N rpm rpm Same
tional [T – 1] (revolutions per
speed minute),
Hz (hertz) or
cycle/s
Angular  rad/s rad/s 1 Hz = 6.28319 rad/s
speed [T – 1] (radians per (radians per 1 radian = 180°/
second) second)
Energy or E, W, T N – m or kWh ft – lb 1 ft – lb = 1.35582 J
work or [ML2T – 2] or joule (J)
torque
Power P N – m/s ft – lb/s 1 N – m/s = 1 W
(watt or W), HP (horsepower) 1 ft – lb/s = 1.35582 W
[ML2T – 3] kW, MW 1 HP = 550 ft – lb/s
1 HP = 745.7 W
Dynamic I kg – m2 slugs – ft2 1 slugs – ft2 = 1.35582 kg – m2
moment or (W/g)
of inertia (useful for pump water hammer
studies)
Conver- = C + 273 K = F + 460 R
sion to
absolute SI units: US customary
units of degrees units: degrees
tempera- Celsius to Fahrenheit to
ture Kelvin degrees Rankine
1. The asterisk on the pressure symbol pv* refers to absolute pressure units as discussed in Section 2.11.

606
 Chapter B - File: AppendixB - New.fm July 6, 2022 Version 6

Table B.2 Common abbreviations

Term Abbreviation SI units Term Abbreviation SI units
or US or US
custom- custom-
ary unit ary unit
absolute abs both megawatt MW SI
atmosphere atm both mercury Hg both
cubic feet cfs US megaliters per day ML/day SI
per second
degrees Celsius C SI meter m SI
degrees Fahrenheit F US miles per hour mph US
Kelvin K SI million gallons MGD US
per day (US)
degrees Rankine R US millimeter mm SI
foot, feet ft US minute min both
feet per second ft/sec or fps US milliliter mL SI
foot – pound ft – lb US newton N SI
gallon (US) gal US outside diameter OD both
(gallon per minute gpm US pascal Pa (N/m 2) SI
(US)
hectare ha SI pound (pound lb (lbf) US
force)
hertz Hz both pounds per psi US
square inch (gage or lb/in2
or absolute units)
horsepower HP or hp US pounds per psf US
square foot or lb/ft2
hour hr both radian rad both
inch in. US radians per second rad/s SI
internal diameter ID both revolutions per rpm
minute
joule J SI second s both
kilogram kg SI watt W SI
kilometer km SI
kilowatt kW SI
liter L SI

607
 Appendix C - File: AppendixC-epsilonvalues - New.fm July 6, 2022 Version 6

APPENDIX C PIPE ROUGHNESS VALUES

608
 Appendix C - File: AppendixC-epsilonvalues - New.fm July 6, 2022 Version 6

C.1 Roughness Height Values for Darcy – Weisbach Calculations

Table C.1 Collected information on pipe interior surface roughness: values of  (based on Skeat —
Manual of British Water Engineering Practice 1969, p. 158)

Equivalent surface roughness

heights  specified in mm ( in.)
Type of Pipe
Mean Recommended
experimental design
value value
ROUGH PIPES:
New uncoated cast iron 0.23 (0.0089) 0.25 (0.010)
New coated cast iron 0.10 (0.004) 0.13 (0.005)
New coated spun iron 0.056 (0.0022) 0.051 (0.002)
New galvanized iron 0.10 (0.004) 0.13 (0.005)
New wrought iron 0.051 (0.002) 0.051 (0.002)
New coated steel 0.045 (0.0018) 0.051 (0.002)
New uncoated steel 0.028 (0.0011) 0.038 (0.0015)
Cast iron mains immediately after scraping 0.47 (0.0185) 0.508 (0.020)
Cast iron mains immediately after brushing with whalebone 1.02 (0.040) 1.02 (0.040)
brush
Tate relined pipes 0.33 (0.013) 0.38 (0.015)
Prestressed concrete (Freyssinet) 0.030 (0.0012) 0.038 (0.0015)
Spun concrete pipe (Bonna, Socoman, etc.) 0.198 (0.0078) 0.25 (0.010)

Concrete pipes: Scobey’s Class 1 5.08 (0.20) 5.08 (0.20)

(old concrete pipes, mortar not wiped from joints)
Concrete pipes: Scobey’s Class 2 1.27 (0.05) 1.27 (0.05)
(modern dry – mix pipe, monolithic pipe, or tunnel linings made
over rough forms)
Concrete pipes: Scobey’s Class 3 0.41 (0.016) 0.508 (0.020)
(small wet – mix pipes in short lengths, average monolithic pipes
made on steel forms, pressure – made pipes with interior float of
neat cement by mechanical trowel)
Concrete pipes: Scobey’s Class 4 0.178 (0.007) 0.25 (0.010)
(monolithic pipes with joint scars and all irregularities removed,
first – class concrete against oiled steel forms)
New steel pipes: Scobey’s Class 1 0.305 (0.012) 0.38 (0.015)
(full riveted pipes up to 3/16 in metal, girth – riveted pipes)
New steel pipes: Scobey’s Class 2 1.52 (0.060) 1.52 (0.060)
(full riveted pipes 4.8 mm (3/16 in.) to 12.5 mm (½ in.) in metal)
New steel pipes: Scobey’s Class 3 0.051 (0.002) 0.051 (0.002)
(continuous interior pipes unmarred by plate offsets, rivet heads, to See values for
etc.) 0.127 (0.005) coated and
uncoated steel
above)
New uncoated asbestos cement 0.028 (0.0011) 0.025 (0.001)

609
 Appendix C - File: AppendixC-epsilonvalues - New.fm July 6, 2022 Version 6

Table C.1 Collected information on pipe interior surface roughness: values of  (based on Skeat —
Manual of British Water Engineering Practice 1969, p. 158)

Equivalent surface roughness

heights  specified in mm ( in.)
Type of Pipe
Mean Recommended
experimental design
value value
HYDRAULICALLY SMOOTH PIPES: Smooth pipe Smooth pipe
law law
Smooth drawn non – ferrous pipes — aluminium, brass, copper, Smooth pipe Smooth pipe
lead, etc. law law
Plastic pipes — glass, Polyethylene, PVC,* etc. Smooth pipe Smooth pipe
law law
Coated asbestos – cement pipes Smooth pipe Smooth pipe
law law
Spun lined pipes — spun concrete or bitumen – lines pipes free Smooth pipe Smooth pipe
from slime law law

* Average roughness height values of up to 0.064 mm (0.0025 in.) have been recorded in experiments on
commercial PVC pipes as the result of pronounced internal waviness in the extrusion process. Visual
inspection of recently manufactured British PVC pipes (as of 1969 book by Skeat) has confirmed that wav-
iness has now been substantially eliminated by improved extrusion techniques.

610
 Appendix C - File: AppendixC-epsilonvalues - New.fm July 6, 2022 Version 6

C.2 Hazen – Williams C Values

Table C.2 Values of C in Hazen – Williams formula (based on Skeat — Manual of British Water
Engineering Practice 1969, p.163 and 164)

Type of pipe Values of C for pipes of diameter

25 mm 75 mm 150 mm 300 mm 600 mm 1000 1200

(1 in.) (3 in.) (6 in.) (12 in.) (24 in.) mm mm
(40 in.) (48 in.)
Uncoated cast iron: smooth and new 121 125 130 132 134
Coated cast iron:
Smooth and new 129 133 138 140 141
30 years old:
Trend 1: slight attack 100 106 112 117 120
Trend 2: moderate attack 83 90 97 102 107
Trend 3: appreciable attack 59 70 78 83 89
Trend 4: severe attack 41 50 58 66 73
60 years old:
Trend 1: slight attack 90 97 102 107 112
Trend 2: moderate attack 69 79 85 92 96
Trend 3: appreciable attack 49 58 66 72 78
Trend 4: severe attack 30 39 48 56 62
100 years old:
Trend 1: slight attack 81 89 95 100 104
Trend 2: moderate attack 61 70 78 83 84
Trend 3: appreciable attack 40 49 57 64 71
Trend 4: severe attack 21 30 39 46 54
Miscellaneous:
Newly scraped mains 109 116 121 125 127
Newly brushed mains 97 104 108 112 115
Coated spun iron:
Smooth and new 137 142 145 148 148
Old Take as coated cast iron of same age
Galvanized iron: smooth and new 120 129 133
Galvanized steel:
New 150
25 years old 130
50 years old 100
Badly corroded 60
Wrought iron: smooth and new 129 137 142
Coated steel: smooth and new 129 137 142 145 148 148
Uncoated steel: smooth and new 134 142 145 147 150 150

611
 Appendix C - File: AppendixC-epsilonvalues - New.fm July 6, 2022 Version 6

Table C.2 Values of C in Hazen – Williams formula (based on Skeat — Manual of British Water
Engineering Practice 1969, p.163 and 164)

Type of pipe Values of C for pipes of diameter

25 mm 75 mm 150 mm 300 mm 600 mm 1000 1200

(1 in.) (3 in.) (6 in.) (12 in.) (24 in.) mm mm
(40 in.) (48 in.)
Riveted steel:
New 120
25 years old 100
50 years old 80
Badly corroded 45
Coated asbestos cement: clean 147 149 150 152
Uncoated asbestos cement: clean 142 145 147 150
New 150
25 years old 130
50 years old 120
Badly corroded 100
Spun cement lined and spun bitumen 147 149 150 152 153 153
lined: clean
25 years old 130
50 years old 100
Badly corroded 60
Smooth pipe (including lead, brass, 140 147 149 150 152 153
copper, polyethylene, smooth PVC):
clean
PVC (wavy): clean 134 142 145 147 150 150
PVC:
New 150
25 years old 140
50 years old 140
Badly corroded 130
Concrete: (Scoby)
Class 1 (Cs = 0.27) Clean 69 79 84 90 95
Class 2 (Cs = 0.31) Clean 95 102 106 110 113
Class 3 (Cs = 0.345) Clean 109 116 121 125 127
Class 4 (Cs = 0.37) Clean 121 125 130 132 134
Best (Cs = 0.40) Clean 129 133 138 140 141
Concrete — smooth
New 150
25 years old 130
50 years old 120
Badly corroded 100

612
 Appendix C - File: AppendixC-epsilonvalues - New.fm July 6, 2022 Version 6

Table C.2 Values of C in Hazen – Williams formula (based on Skeat — Manual of British Water
Engineering Practice 1969, p.163 and 164)

Type of pipe Values of C for pipes of diameter

25 mm 75 mm 150 mm 300 mm 600 mm 1000 1200

(1 in.) (3 in.) (6 in.) (12 in.) (24 in.) mm mm
(40 in.) (48 in.)
Tate Relined Pipes: Clean 109 116 121 125 127
Prestressed Concrete Pipes: Clean 147 150 150
Woodstave:
New 120
25 years old 100
50 years old 80
Badly corroded 45

NOTES: (1) The above table was compiled from an examination of 372 records (except for 1000 mm diameter pipes
listed), and emphasizes that the Hazen – Williams formula is not suitable for values of the efficient C appreciably below
100. However, the values in the above table are approximately correct at a velocity of 0.9 m/s (3 ft/sec). For other
velocities the approximate corrections shown in Table C.1 should be applied to values of C. (2) For diameters less than

1000 mm subtract 0.1 C (Skeat 1969).

613
 Chapter D - File: AppendixD - New.fm July 6, 2022 Version 6

APPENDIX D MOODY DIAGRAM FOR COMPUTER

SIMULATION OF WATER DISTRIBUTION SYSTEMS

A schematic of the Moody diagram is shown in Figure D.1. Dunlop(2001) also came up with a form of
the Moody diagram where he developed curves to smoothly transition from the laminar region to the
transition turbulent region. This form is useful in the computer simulation of water distribution networks
to enable an appropriate transition occur between laminar flow and turbulent flow. Rossman (2000) used
these curves in the EPANET computer code.

614
 Chapter D - File: AppendixD - New.fm July 6, 2022 Version 6

Figure D.1 The Moody diagram (after Dunlop 1998)

D.1 References
Dunlop, E. (2001). “Computation of Friction Factor,” CCWI Conference - Water Software Systems: the-
ory and applications, Vol. 2, Ed. B Ulanicki, B Coulbeck, J Rance; Research Studies Press Ltd, 203.

615
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

A
abbreviations 12, 15, 17, 605
abrupt contraction 50, 131, 132, 138, 138, 139, 140, 142, 146
head loss coefficient 140, 141
abrupt expansion 131, 132, 134, 135, 136, 137, 140, 146, 147
head loss coefficient 137
absolute (abs) 605
absolute pressure 9, 33, 34, 604
vapor pressure 604
absolute temperature 22
units 604
absolute viscosity 13, 16, 41, 42, 45, 46, 83, 132, 604
acceleration 12, 13, 46, 51, 131, 143, 539, 603
gravity 12
accumulator 576
acre, acre-ft 603
affinity laws
pumps 404
age of pipe 397
aging of pipes 104
air
density 13, 23, 25
gas constant 22
specific gravity 25
specific weight 25
air vessel 576
angular speed 604
area 603
assumptions
continuity equation 47
energy equation 55
Euler equation 60
pipe flow analysis 47
unsteady continuity equation 537
atm 605
units 604
atmosphere 16
absolute pressure 449
barometric pressure 449
local atmospheric 33
pressure 16, 33, 604, 605
atmospheric pressure 11
local 34
axial flow pump 404, 407, 409, 409, 415, 421, 424, 425, 425, 426
B
barometer 35
aneroid 35
barometric pressure 26, 34, 449, 450, 451
bellmouth

614
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

entrance 145
bend 54, 63, 65, 75, 131, 143, 143, 145, 396
forces on 53
Bernoulli equation 39, 56, 60, 61, 62
Bernoulli flow 60
bifurcation 131
body force 52
boundary layer 42, 76, 92, 92
Bourdon gage 35, 406
branch 131, 146
network 474, 477, 478
Buckingham pi theorem 45, 46, 82, 132
bulk modulus of elasticity 14, 27, 397, 510, 512, 514, 518, 528, 535, 599
units 604
butterfly valve loss 147
C
cast iron
fittings 75, 76
cast iron pipe 74, 75, 105, 112
fittings 75, 76
Hazen Williams C 80
roughness height 79
cavitation
pumps 404
cavitation damage
photograph 27
Celsius 605
units 604
centistoke 604
centrifugal pump 404
coefficient
contraction 138
Hazen-Williams C 106, 112
Manning n 128
minor loss 131
Colebrook 80, 98, 104
Colebrook-White formula 98, 99, 158
complete turbulence 77
computer analysis of water distribution systems 3
conservation
energy 64, 218
mass 48, 227, 537
water 402
constant
gas 22, 23
continuity 218, 224, 227
assumptions 47
equation 39
non-uniform velocity distribution 70

615
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

continuity equation 228, 231, 246, 261, 264, 287, 311, 336, 354
contraction 63
abrupt 50, 131, 132, 138, 138, 139, 140, 142, 146
coefficient of 141
gradual 131
contraction coefficient Cc 138
contractions 131
conversion of units 14
conversions 603
critical velocity 91
D
Darcy-Weisbach
comparison to Hazen-Williams coefficient
113
head loss equation 76
Darcy-Weisbach fluid friction factor
laminar flow 87
Darcy-Weisbach friction factor
Prandtl-Nikuradse formula 101
Swamee and Jain equation 93
Darcy-Weisbach head loss equation 86
definition
absolute viscosity 16
delta(QL)-equations 218
demand 228
peak hour 2
demand estimation 1
demands 223
water 1
density 20, 397
air 13, 14, 23, 25
units 603
water 14, 21
design life
water distribution systems 2
determinate networks 223
diameter 397
equivalent 112, 113
unknown for a smooth pipe 171, 173
unknown pipe 149
diameter calculation
fully rough turbulent flow 173
diameter unknown
smooth pipe 171, 172, 173
DICL (ductile iron cement lined) 75
DICL (ductile iron cement mortar lined) 75
dimensional analysis 82, 132
dimensionless parameters 132
discharge

616
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

unknown 154, 162

unknown pipe 149, 152
unknown pipe for transitional turbulent 158
discharge unknown
fully rough turbulent 162
smooth turbulent flow 153
transitional turbulent flow 158
disconnected networks 222, 228, 241, 249
discount rate 397
distribution
water 1
draft tube 580
ductile iron 75
dynamic moment of inertia 604
dynamic programming 7
dynamic viscosity 604
E
efficiency
pump 404
efficiency curve
pump 397
efficiency of generator 595
EGL 67, 86, 106, 107, 181, 219
elbow 145
electricity tariff 397
elevation head 56, 57, 61, 64, 66, 67, 406
elevation-volume relationship
tank 398
energy 55
conservation 64
equation 39
heat 59
kinetic 57
mechanical 59
potential 56
units 604
energy conservation 218
Energy equation 244
energy equation 39, 55, 65, 65, 66, 68, 82, 90, 135, 149, 150, 151, 155, 159, 163, 165,
167, 169, 170, 173, 176, 195, 197, 231, 232, 232, 233, 236, 241, 295, 341, 405, 449
assumption 55
energy grade line 66
energy, pressure 58
entrance
bellmouth 145
re-entrant 144
sharp-edged 144
entrance loss 141, 144
entrances 131

617
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

enumeration 7
EPANET 248
EPS 2
equivalent pipe diameter 112, 113
equivalent pipe length 147
estimating water demands 1
Euler equations
assumptions 60
Euler’s equation 59
exit loss 147
exits 131
expansion
abrupt 132, 134, 135, 136, 140, 146
abrupt head loss coefficient 137
expansions 131
extended period simulation 2, 4
F
Fahrenheit 604, 605
FCV 453, 454, 467
feet
units 603
fire demands 1
fire loading 398
fittings
cast iron 75, 76
cast iron pipe 75
fixed grade nodes 222
flow
non-uniform 40
flow control valve 398
flow equations 218
flow meters 131
flows
solving as unknowns 245
fluid
definition 10
ideal 42
real 42
force
body 52
definition of 12
surface 52
units 603
Formulation of equations
Q-H 248
formulation of equations
Q-H 250
fps (feet per second) 603
Francis turbine 582, 584

618
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

friction factor
rough turbulent flow 101
friction loss 63
ft (feet) 603
fully rough 87, 166, 169, 173
fully rough turbulent flow 162
G
gage
Bourdon 35, 406
gage pressure 25, 26, 30, 33, 34, 449
pressure
gage 9
gal (gallon) 605
gallon 603, 605
UK 603
US 603
gallon (gal) 605
gas constant 22, 22, 23
air 22
water vapor 22
gas law
perfect 21, 23, 33
gate valve loss 146
genetic algorithm optimization 7
genetic algorithms 2
geographical information systems 1
gigaliter 14
GIS 1
globe valve loss 147
governing equations 218
water distribution systems 2
gpm (gallons per minute) 605
gradual contraction 131, 142
gradual expansion 142
graph 222
graph theory 222
gravitational acceleration 12
H
ha (hectare) 605
Hardy Cross 2, 3
Hazen-Williams coefficient
adjustments to 113
comparison to Darcy-Weisbach 113
typical values 112
Hazen-Williams equation 105
accuracy of 115
head 56
piezometric 68
head equations 218

619
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

head loss 229

abrupt contraction 139, 140
Darcy-Weisbach 86
unknown 149
unknown pipe 150, 151
head loss coefficients
typical 144
heat 64
heat energy 59
hectare 603
hectare (ha) 605
H-equations 218
hertz (Hz) 605
HGL 67, 219
horsepower (HP) 604, 605
HP (horsepower) 604, 605
hydraulic grade line 66, 67
hydraulic radius 106, 107, 128, 129
hydropower
classification 581
hydropower generation 578
Hz (hertz) 605
hertz (Hz) 604
I
ID (internal diameter) 605
ideal fluid 42
Imperial units (see also US Customary units) 9
impulse (as in momentum) 51
in. (inch) 605
inch (in.)
in. (inch) 605
inertia
moment of 604
internal diameter 72, 74
internal energy 64
J
J (joule) 605
Jacobian matrix 248
joule (J) 604
junctions
number of 249
K
Kaplan turbine 582, 585
Kelvin 22, 605
units (temperature) 604
kg (kilogram) 603, 605
kgf (kilograms force) 603
kilogram (kg) 603, 605
kilograms force (kgf) 603

620
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

kiloliter 14
kilopascal (kPa) 604
kilowatt 605
kinematic viscosity 19, 397
units 604
kinetic energy 56, 57, 64
kinetic energy correction factor 56, 68, 129, 602, 613
laminar flow 70
turbulent flow 70
km (kilometer) 603
kPa
kilopascal 604
kW 605
kW (kilowatt) 604
kWh (kilowatt-hour) 604
L
L (liter) 603, 605
laminar flow 41, 77, 87
laminar sublayer 78, 92
length 603
linear programming 7
links 222
liter 14
liter (L) 603
loop 218
loop flow correction equations 218
loops
path 222, 234, 249
simple 233, 241, 249
M
m (meter) 603
Manning n coefficient
typical values 129
Manning’s equation 128
Mannum – Adelaide pipeline 575
manometers 36
mass
units 603
weight relationship 12
mass conservation 48, 227, 537
Matrices
symmetric 248
matrices
Jacobian. See Jacobian matrix 248
reservoir node incidence matrix 224, 226, 227
unknown heads node incidence matrix 224, 226
mechanical energy 59
megaliter 14
megaliters (ML) 603

621
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

megawatt 605
megawatt (MW) 604
meter 603
MG (million gallons) 603
MGD (million gallons per day) 603
mile 603
million gallons (MG) 603
minor loss 63
abrupt expansion 134, 146, 147
branch 146
butterfly valve 147
contraction 146
elbow 145
entrance 141, 144
exit 147
exit to a reservoir 137
gate valve 146
globe valve 147
gradual contraction 142
gradual expansion 142
re-entrant entrance 141
reflux valve 147
smooth entrance 141
sudden contraction 138, 139, 142
sudden expansion 135
taper 146
tee 145
valve 146
valves 144
minor loss coefficient 131
abrupt contraction 140
abrupt expansion 137
minor losses 131, 218
bend 143
ignoring 147
mixed flow pump 404
ML (megaliters) 603
mm (millimeter) 603
molecular weight 22
moment of inertia 604
momentum equation 39, 134
momentum flux 50
momentum, equation of 50, 51
Moody 80
Moody diagram 79
MW 605
MW (megawatt) 604
N
net positive suction head (NPSHA, NPSHR) 404

622
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

newton
units 603
Newton’s law of viscosity 18, 41
Newton’s second law 12
Nikuradse 80
node
definition 218
non-linear programming 7
non-uniform flow 40, 131
norm
1-norm 263
2-norm 263
infinity-norm 263, 264
pth-norm 263
normal stress
units 604
numerical solutions
water distribution system equations 2
O
objectives
water distribution system design 2
OD 605
one dimensional flow 56
one-dimensional flow 42
optimization methods 7
optimization of water distribution systems 2
orifice plates 131
outside diameter 605
P
Pa (pascal) 604, 605
pascal (Pa) 604
path 222, 234, 249
definition 223
path. See also loops.
peak hour 1
peak hour analysis 398
peak hour demand 2
Pelton wheel 583
Pelton wheel turbine 582
penstock 580
perfect gas law 21
piezometers
open 36
piezometric head 68
Pilati. See Todini and Pilati
pipe
cast iron 74, 75, 79, 105, 112
ductile iron 75
steel 75

623
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

pipe flow analysis

assumptions 47
pipe numbers or names 397
pipe systems
simple 149
pipes 1
aging 104
number of 249
Poiseuille flow 90
population projections 3
potential energy 56, 64
potentiated network 222
pounds force (lbf or lb) 603
power
pump 66
turbine 66
units 604
Prandtl 80, 92, 101, 162, 164, 166, 199
Prandtl-Nikuradse formula
Darcy-Weisbach friction factor 101
present value 397
pressure
absolute 9, 33, 34, 604
barometric 26, 34, 449, 450
definition 11
gage 25, 26, 30, 33, 34, 449
measurement 35
minimum allowable 2
units 604
vacuum 35
vapor 25, 604
variation in a fluid 29
variation with depth 31
pressure energy 56, 58
pressure head 58
definition 31
pressure reducing valve 398
pressure relief valves 575
pressure setting 398
pressure sustaining valve 398
properties of water 13
PRV 453, 454, 455, 456, 466
psi (pounds per square inch) 605
pump
affinity laws 404
axial flow 404, 407, 409, 409, 415, 421, 424, 425, 425, 426
best efficiency point (BEP) 404, 579
cavitation 404
centrifugal 404, 407

624
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

efficiency curve 397

energy equation 405
head 405
impeller 407
impeller types 407
mixed flow 404
net positive suction head 404
operating point 404
parallel 404
rotating moment of inertia 404
shutoff head 405
specific speed 404
speed 405
static head 405, 580
static suction head 405
system curve 405, 580
water power 405, 580
pump curve 397, 404
pump operation 397
pumped hydro 596
pumps 1
impeller type 404
rotodynamic 404
series 404
types 405
Q
Q-equations 218
Q-H equations 218, 257, 351
R
rad (radian, radians) 604, 605
Rankine 22, 605
units (temperature) 604
rated flow
pump 397
rated head
pump 397
real fluid 42
re-entrant
entrance 144
re-entrant entrance 141
reflux valve loss 147
relative roughness 74
repeating variables 83, 133
reservoir 398
reservoir loss 137
reservoirs 1
revolutions per minute (rpm) 604
Reynolds 80
Osborne 91

625
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

Reynolds number 77, 91

rotating moment of inertia 397, 404
rotational speed
units 604
rough turbulent flow 72, 77, 78, 87, 102, 164, 166, 169, 173
requirement for 102
roughness
relative 74
rpm (revolutions per minute) 605
rpm (revolutions per minutes) 604
S
secondary flow 143
shaft work 64
shear stress 10, 81, 83, 85
pipe wall 82, 86
units 604
shear velocity 94, 102
SI units 9, 604
sign convention 223, 228
similitude 132
simple pipe systems 149
simulation of the water distribution systems 1
slugs 603
smooth entrance 141
Smooth pipe 610
smooth pipe 93, 112, 116, 117, 118, 169, 273, 274, 608
smooth pipe turbulent flow 77, 78, 93, 94, 153, 169
requirement for 95, 99
Snowy 2.0 pumped hydro project 596
Snowy Hydro Scheme 579
software
simulation of hydraulics 1
specific gravity 24
air 25
units 603
water 24
specific speed 404
turbine 581
specific volume 21
specific weight 23, 397
air 25
units 603
specific weight of water 14
speed
angular 604
rotational
units 604
standard conditions 14, 21
static head

626
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

turbine 590
steady flow 39, 55
steady state analysis 2, 4
steel pipe 75
stopping test 262
storage 398
streamline 56, 59
stress
normal 9, 604
shear 10, 604
sublayer
laminar 78, 92
viscous 72, 92, 92, 93
surface force 52
surge tank 575
surge tanks 576
Swamee and Jain equation 93
switch yard 580
system curve 404
pump 405, 580
turbine 591
T
T junction 131
tailrace 580
tank 398
water tank (photograph) 406
tanks 1
taper 146
Taylor series 258, 261, 286, 310, 335, 353, 373
tee 145
temperature
absolute
units 604
thick walled pipe 73
thin walled pipe 73
Todini and Pilati method 245, 248
tons 603
torque 604
transformer 580
transitional 87, 169
transitional turbulent flow 77, 78, 87, 98, 169
transitions 131
transmission 1
treed 223
trial and error design 3
trial and error design of water distribution systems 6
tuberculation 104
tunnel 580
Turbine

627
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

energy output 594

turbine 578
curves 592
efficiencies 588, 590
energy equation 590
impulse 582
reaction 582
shaft power 593
system curve 591
water power 592
turbine Types 582
turbulent flow 41, 87, 166, 169, 173
fully rough 72, 77, 78, 101, 102, 162, 164
rough 102
smooth pipe 77, 78, 93, 94, 153
transitional 77, 78, 98, 158
Turgo turbine 582, 583
U
uniform 40
uniform flow 131
units 603
units of measurement 11
unknown
head loss in pipe 151
pipe diameter 149, 171, 172, 173
pipe discharge 149, 152, 153, 158, 162
pipe head loss 149, 150
unknown discharge
fully rough turbulent 162
smooth pipe 154
unknown heads node incidence matrix 224, 226
unsteady continuity equation
assumptions 537
unsteady flow 39
US customary 604
US customary units 9
V
vacuum pressure 35
valve
angle 144
butterfly 144
check 144
gate 144
globe 144
sleeve 27
valve loss 146
valves 1, 131, 144
vapor pressure 14, 25
units 604

628
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

vapor pressure of water 25

velocity
shear 94
units 603
velocity distribution
laminar flow 90
Prandtl 70
turbulent flow 70
velocity head 56, 57, 62, 219
Vena contracta 138
viscosity
absolute 16, 604
dynamic 604
eddy 42
kinematic 19, 604
Newton’s law 18, 41
units 604
viscosity of water
absolute 13
kinematic 13
viscous sublayer 72, 92, 92, 93
characteristic length or height 95
volume 603
tank 398
von Karman 80, 92, 96, 101, 117, 153, 156
von Karman fluid resistance formula 94
W
W (watt) 604, 605
water
density 14, 21
fire demands 1
properties 13
water conservation 402
water demands 1
commercial 1
industrial 1
water distribution systems
governing equations 2
water hammer 2
water power
turbine 592
water surface elevation
elevation
water surface 398
water treatment plants 1
water vapor
gas constant 22
watt (W) 604, 605
weight

629
 Chapter E - File: WDSv4IX.fm July 6, 2022 Version 7

mass relationship 12
wetted perimeter 82
White 80, 98, 104
wicket gates 580
work 58
units 604
Y
Young’s modulus of elasticity 28

630