SIMULATION OF SLACK LINE FLOW

A TUTORIAL

R. Edward Nicholas
 Simulation of Slack Line Flow
A Tutorial

R. Edward Nicholas
Scientific Software - Intercomp

Slack line flow behavior is described using steady state analysis. Methods of estimating slack line flow velocities
andjlow cross sectional areas arepresented. l%en, a transient simulation approach that has been used to develop a
real time slack line flow model is described and the complexities of simulating transient slack line flow are
discussed.

INTRODUCTION ... ...... .... ...... .... ...... ..... ...... ..... ...... ..... ...... ..... ....... .... ....... .... ...... .... ...... .... ...... .... ............*............ 2

THE SLACK LINE FLOW PHENOMENA ... ...... ..... ...... ..... ....... .... ....... .... ...... ..... ..... ..... ..... .... ..... .... ..... .... ...... .2

DISCUSSION OF STEADY STATE SLACK LINE FLOW ............................................................................ 3

DISCUSSION OF HEAD IN SLACK LINE FLOW REGIONS.. ...................................................................................... .
VELOCITY AND FLOW CROSS SECTIONAL AREA IN SLACK REGIONS .................................................................. .
Hydraulic Radius ............................................................................................................................................ 6
Discussion of the Manning Formula for Open Channel Flow ................... .;. ................................................. .6
Modification of the Fanning Equation for Slack Line Flow.. ......................................................................... 7
Area and Hydraulic Radius Equations for Partially Full Pipes.. .................................................................. .7
Computation of Cross Sectional Area in Slack Regions.. .............................................................................. .8
Discussion of Cross Sectional Flow Area and Velocity in Slack Line Flow .................................................. .9
Velocity in Slack Regions As A Function of Tight Line Velocity .................................................................. I1
SIMULATION OF TRANSIENT SLACK LINE FLOW ............................................................................... 12
TIGHTLINESIMULATION .................................................................................................................................. 13
Tight Line Flow Equations ........................................................................................................................... 13
Solution of the Tight Line Flow Equations By Implicit Finite Differences.. ................................................ .I4
SLACKLINESIMULATION .................................................................................................................................. 14
The Simplicity of Simulating Slack Line Flow .............................................................................................. 14
The Complexity of Simulating Slack Line Flow ............................................................................................ 15
OPEN ISSUES FOR RESEARCH RELATED TO SLACK LINE FLOW SIMULATION.. ..................... .17
THEPROBLEMOFBUBBLECOLLAPSE ............................................................................................................... 17
THEPROBLEMOFDRA ..................................................................................................................................... 17
THEPROBLEMOFFRICTIONFACTOR ................................................................................................................ 17

CONCLUSION ................................................................................................................................................... 17

TABLE OF SYMBOLS . ...... ... ...... .... ....... .... ....... .... ....... ..... ...... ..... ....... .... ...... ..... ...... ... ....... ... ...... .... .................. 18

Slack Line Flow - A Tutorial Page 1 PSIG - October 19-20,1995

Introduction
Slack line flow is a phenomena in which a pipeline transporting a liquid product develops vapor bubbles at
points at which the pipeline pressure falls below the vapor pressure of that liquid.
This flow condition is very common in liquid pipelines. Most liquid pipelines fall into one of the following
categories:
1. Slack line flow occurs all of the time in some section of the pipeline.
2. Slack line flow occurs occasionally during normal pipeline operations.
3. Slack line flow occurs only during abnormal pipeline operations.
Experiences installing real time simulation software on a number of liquid pipelines over the last 15 years have
included the following:
1. On one, slack line flow existed almost all of the time. However, until the real time simulation software
was installed, the pipeline operators were unaware of its occurrence. In this case, the total slack
volume (volume of vapor bubble) was only 50 to a few hundred barrels in size.
2. On another, slack line flow occurred only during some upsets. Again, the pipeline operators were
unaware of this condition until real time simulation so&are was installed. Since the simulation
software was not designed to handle the slack line flow, simulations during those periods were
inaccurate.
3. On others, slack line flow was expected as a normal part of the operation. On one large pipeline, slack
volumes are in the tens of thousands of barrels.
Slack line flow behavior is of interest to pipeline operators for a number of reasons:
1. Frictional head loss in slack regions is much greater than in tight line regions.
2. Slack line regions significantly change the hydraulics of the pipeline operations, particularly the
behavior of transients.
3. The collapse of slack line regions may cause pressure surges that must be considered when developing
operational procedures to avoid damaging the pipeline.
Slack line flow behavior poses a significant challenge to developers and users of liquid pipeline simulators. On
one hand, it greatly affects the transient behavior of the pipeline; on the other hand, simulation of this behavior
is much more difficult than the simulation of tight line flow and is subject to many more uncertainties than is
tight line simulation.

The Slack Line Flow Phenomena

It is useful to consider what is occurring during slack line flow. From a thermodynamic perspective, when the
pressure falls below the vapor pressure, a vapor bubble is formed and the fluid pressure does not drop below the
vapor pressure.
From a mechanical point of view, slack line flow approximates open channel flow. An open channel is one in
which the fluid is not completely enclosed by solid boundaries and therefore has a free surface whose pressure
remains at atmospheric pressure (or vapor pressure in this case). The fact that the pressure on the free surface
remains constant requires, at steady flow, that the rate of change of potential energy of the fluid is exactly offset
by the frictional work of the “falling” fluid against the fluid surroundings.

Slack Line Flow - A Tutorial Page 2 PSIG - October 19-20, 1995

One can visualize this by considering what happens in open channel flow if one were to tilt that channel to a
greater angle. Clearly, the velocity of the fluid in the channel will increase. If a steady flow rate is maintained,
the cross sectional area of the fluid in the channel decreasesas the velocity increases. The velocity will reach
the velocity at which the energy dissipated by friction offsets the rate of change of potential energy of the fluid
flowing down the channel.

Discussion of Steady State Slack Line Flow

Discussion of Heud in Slack Line Flow Regions
The frictional head loss is given by the Darcy-Weisbach equation as:

Eqn. 1, Darcy- Weisbach Equation

h,=f--- +I h
w
The following equation relates the pressure head gradient, the frictional head loss per unit length, and the
elevation gradient for a liquid pipeline in steady flow at any point in the line.

Eqn. 2

filvl +dz = 0
L,+- 2gD dx
It is customary to define the pipeline head as the sum of the pressure head and the elevation head or:

Eqn. 3

However, for the purposes of this discussion, it is useful to define the head so that it is equal to the elevation
when the pressure is equal to the vapor pressure or:

Eqn. 4, Definition of Head

H = p - Pvapor+ z
Y
Eqn. 2 can be rewritten as:

Eqn. 5

dH
-=-- .f+j
dx 2gD

In portions of the pipeline containing slack line flow, the pressure is equal to the vapor pressure so the pressure
gradient is zero. Therefore, for slack segments of the line, Eqn. 2 is simplified to

Slack Line Flow - A Tutorial Page 3 PSIG - October 19-20, 1995

 Eqn. 6, Friction Loss - Elevation Gradient Relationship in Slack Line Flow

filvl ; dz - 0
2gD dx
This states that the head loss due to friction exactly equals the elevation head loss. Combining Eqn. 5 and Eqn.
6 gives
Eqn. 7, Head Gradient in Slack Regions

dH dz
m=z
Eqn. 5 and Eqn. 7 allow one to plot head versus distance for any slack line condition. The slope of this plot
will be equal to the head loss per unit length due to friction. In all tight line regions, this slope is constant
(assuming velocity and friction factor are constant). However, in slack line regions, the head is equal to the
pipeline elevation as specified by Eqn. 4 and the head curve follows the elevation profile until the point at
which slack line conditions no longer exist.
This is illustrated in Figure 1:

Figure I, Example Head Gradients in Slack Line Flow

Example Slack Line Hydraulic and Elevation Gradient

Distance (miles)

In this figure, segmentsAB, CD, and EF are in tight line flow. SegmentsBC and DE are in slack line flow and
the pressure in these regions is equal to the vapor pressure. This figure illustrates some basic facts of slack line
flow:
First, this figure illustrates that slack line flow is often unavoidable. Points G and H on the maximum allowable
operating head curve prevent the pressure from being raised high enough at the upstream or downstream ends
of the line to avoid the slack conditions.

Slack Line Flow - A Tutorial Page 4 PSIG - October 19-20, 1995

Next, it is important to consider what factors control the throughput during slack line flow. As long as the line
is in slack line flow, the head at B will be equal to the elevation. So, the flow rate in the line will be determined
by the head at A and the location and elevation of point B. This leads us to a basic fact about slack line flow:

Basic Fact 1

The throughput of a line flowing slack is governed by the head at the’upstream end and the
elevation and location of thefirst point at which theJlow becomesslack.

A consequence of this is that the head (or pressure) at the downstream end has absolutely no affect on the
throughput of the line. Therefore, the downstream end of the line may be operated at whatever pressure is
convenient. In some operations, a lower pressure is desired for safety. However, when a pump station is at the
downstream end, a higher operating pressure may be desired in order to reduce pumping costs. Furthermore,
the following is true.

Basic Fact 2

Pressure changes at the downstream end of a line flowing slack will never be observed at the
upstream end of the line. In other words, transients originating at the downstream end of the
line will be completely absorbed by the slack region (as long as the line remains slack).

One may also observe from Figure 1 that both the upstream and downstream pressures(head) influence the size
of the slack region. The length of the slack regions may be determined by drawing a straight line backwards
starting with the head at the downstream end and with a slope equal to the slope between the upstream end and
the first slack region. The point at which this line intercepts the elevation gradient is the end of the slack
region. If more than one slack region exists, this process may be duplicated for each additional tight line
section such as segment CD in Figure 1. We restate this as follows:

Basic Fact 3

The length of the slack regions for stea+lflow) depends upon both the head (pressure) at the
upstream end and the head (pressure) at the downstream end. The length of these regions
may be determined by graphical means on a headplot containing a plot of elevation versus
distance.

Finally, we state one additional fact that seemsobvious but is worth mentioning.

Basic Fact 4

Regions of slack line Jlow alwqs begin at an elevation peak or on the downstream side of a
gently sloping hill. They never begin on a uphill slope.

It is clear that the slack region cannot begin on an uphill slope. Both elevation changes and friction losses
decreasethe pressure as the fluid flows uphill, and it would be impossible for this to occur when the pressure is
already at vapor pressure. However, it may not be immediately obvious that slack regions may begin on the
downstream slope. This can occur if the elevation gradient is less steepthan the hydraulic gradient.
The only exception to this rule is when segments of a pipeline have emptied. In this case, as the pipeline is
refilling, there will be columns of fluid pushing their way up a hill with a vapor bubble ahead of them.

Slack Line Flow - A Tutorial Page 5 PSIG - October 19-20,1995

Velocity und Flow Cross Sectional Area in Sluck Regions
The equations in the preceding section are useful for determining the hydraulic gradient in regions of slack line
flow, particularly during steady state. However, additional equations are required to compute the velocity and
the flow cross-section in slack regions. These equations and their implications are discussed in this section.

Hydraulic Radius
In order to discussthe nature of the flow in the slack regions, it is necessary to use the concept of the hydraulic
radius. The hydraulic radius is defmed as the ratio of the flow cross-section to the wetted perimeter as follows:

Eqn. 8, Definition of Hydraulic Radius

For a full pipe, the wetted perimeter is 7cD and R,, = D 4 . (It is also interesting to note that the hydraulic
/
radius for a half full pipe is the same as the radius for a completely full pipe.) One may rewrite the Darcy-
Weisbach equation as

Eqn. 9, Fanning Equation

In this form, the equation is known as the Fanning Equation and is suitable for analyzing flow in non-circular
conduits, including partially full pipes such as occur in slack line flow. In terms of the hydraulic radius, Eqn. 3
becomes:

Eqn. 10

Note: The use of hydraulic radius is not a concept that is introduced in order to handle flow in non-circular
conduits. Instead, the ratio offlow cross-sectional area to the wetted perimeter is afundamental quantity in the
derivation of thefiiction equation. The Darcy- Weisbach equation is a specialization of the more general
Fanning equation for frow in pipes of circular cross section.

Discussion of the Manning Formula for Open Channel Flow

As stated earlier, flow in slack regions may be approximated as open channel flow. In 1890, Robert Manning
published one of the most widely used formulas for open channel flow. In English units, the formula is:

Eqn. 11, Manning Formula

where 1.49 is the cube root of 3.28, the number of feet in a meter. Velocity is in feet per second and the
hydraulic radius is in feet. In metric units the formula is

Slack Line Flow - A Tutorial Page 6 PSIG - October 19-20, 1995

 v2- $%dz
n2 hGk
From this formula, we seethat the frictional head loss may be written as:

Eqn. 12, Friction Loss - Head Relationship from Manning Formula

n2v2 dz
R;'3(1.49)2 = -z

It is useful to compare the friction factor with the parameter n. Combining Eqn. 10 and Eqn. 11 gives

Eqn. 13, Prediction of Friction Factor From Manning Formula

8gn2
f = (1.49)2R~ (English Units)

Values for the parameter n in the Manning formula are tabulated for different surfaces. Values of n for the
smoothest surfaces are 0.0 1. Substituting g and n=O.01 into the above equation gives:

0.0116
f=7 h
(for Rh in feet).

For a 48” pipeline flowing full, the hydraulic radius is one foot and the Manning formula would predict a
friction factor of 0.0116, a reasonable value.
Note: The Manning formula shows no dependence offriction on Reynolds number. The Moo&fiiction factor
curves indicate that this is only validfor large Reynolds numbers. However, large Reynolds numbers are
typical for most open channelJlow applications.

Modification of the Fanning Equation for Slack Line Flow

For our purposes, the primary usefulness of the Manning formula is the statement that, for open channel flow,
the frictional head losseshave a Ri413 dependence as opposed to the Ri’ dependence stated by the Fanning
Equation. We therefore modify Eqn. 10 as follows:

Eqn. 14, Friction Loss - Head Relationship for Slack Line Flow

fiIvl(s>“” + g = 0
8gRz3 dx

Eqn. 14 preserves the correct frictional losses for full pipe flow ( Rh = D / 4 ) yet provides the Rh4/3
dependence predicted by the Manning Formula. In this formula, f is the Moody friction factor computed for a
pipe of diameter D with velocity v.

Area and Hydraulic Radius Equations for Partially Full Pipes

Figure 2 shows the cross section of a partially full pipe. The cross sectional area and wetted perimeter may be
defined in terms of the angle 8 (in radians). From the figure, one can seethat 8 = 0 for an empty pipe and
8 = 71: for a full pipe. Now, equations for the wetted perimeter, cross sectional area, and hydraulic radius may
be written (with the help of a mathematical handbook for the area equation). They are:

Slack Line Flow - A Tutorial Page 7 PSIG - October 19-20, 1995

 Eqn. 15, Wetted Perimeter of a Partially Full Pipe

P=D8

Eqn. 16, Cross Sectional Area of Partially Full Pipe

A=$ e-- sin20

! 2 1

Eqn. 17, Hydraulic Radius of Partially Full Pipe

&A+(I--sin2820 1
Figure 2 - Cross Section of Partially Full Pipe

Computation of Cross Sectional Area in Slack Regions

It is often of great interest to be able to calculate the “empty volume” (vapor bubble size) of the slack region.
Of course, this volume affects the computation of pipe inventory and over/short and leak detection calculations.
In addition, this volume determines how long it will take to repack a slack region to be tight and often
determines how long it may take to restore normal pipeline operations after an upset.
In order to compute the volume of the “empty” region in slack line flow, one makes use of the fact that, for a
steady upstream flow rate, the flow rate in the slack region is equal to the tight line flow rate entering the slack
region.
Then, we have

Slack Line Flow - A Tutorial Page 8 RIG - October 19-20, 1995

 Eqn. 18

Aovo = Av
Where the subscript “o” refers to the tight line flow rate. We use Eqn. 16 and Eqn. 18 to eliminate v and A in
Eqn. 14. This gives

&!!d 7c2
. ,o,3+d”=*
w 02l-Fg!? di
L ( 1 I
or

Eqn. 19

f
x2-
f, -f
(32 l-%5$ 10/3 =
f,vh*I

( 1 20 /

Eqn. 19, particularly when combined with Eqn. 16, Cross Sectional Area of Partially Full Pipe, is very useful.
The numerator on the right hand side is the elevation gradient in the slack region which is also equal to the head
gradient in the slack region. The denominator of the right hand side is the tight line head gradient. The ratio
f / f, on the left hand side is the ratio of the friction factor computed for velocity V, (tight line flow) to the
friction factor computed in the slack region at velocity v . Since the friction factor does not change rapidly
with velocity for most pipeline flow rates, we can assumethat f/f, = 1, at least for the purposes of the
following discussion. Using the relationships expressed in the equations above, we can prepare Figure 3 and
Figure 4.
Note: It is d$kult to eliminate the parameter 0 from the equations above to directly derive relationships
between the various other variables. Instead, thefigures have been prepared by solving each of the equations
above for values of 8 ranging from 0 to 71and then plotting using 8 as a parameter.

Discussion of Cross Sectional Flow Area and Velocity in Slack Line Flow
Figure 3 shows the flow cross sectional area for slack regions as a function of the elevation or hydraulic
gradient. (The elevation gradient is equal to the hydraulic gradient in slack line flow as noted above.) The flow
cross sectional area is expressed as a fraction of the pipe cross sectional area. The elevation gradient is
expressed as a multiple of the frictional head loss per unit length upstream of the slack region.
Figure 4 shows the slack line velocity as a function of the elevation gradient. The velocity is expressed as a
multiple of the velocity upstream of the slack region.

Slack Line Flow - A Tutorial Page9 PSIG - October 19-20, 1995

 Figure 3 - Cross Sectional Area in Slack Line Flow As a Function of Elevation Gradient

Cross Sectional Area versus Hydraulic Gradient

1 10 100 1000
Slack Line Hydraulic Gradient Relative to Tight Line Gradient
[ (-dzldx) I (fVo’W(2Dg)) ]

Figure 4 - Velocity in Slack Line Flow Versus Elevation Gradient

Velocity versus Hydraulic Gradient For Slack Line Flow

1 10 100 1000
Slack Line Hydraulic Gradient Relative to Tight Line Gradient
[ (-dzldx) I (fVo’W(2Dg)) ]

For crude oil pipelines, the head loss per mile of pipe is typically in the range of 10 to 30 feet per mile.
Elevation gradients in slack regions are typically 10 to 100 times this large. (There is a limit on elevation

Slack Line Flow - A Tutorial Page 10 PSIG - October 19-20, 1995

gradient as defined here since x is measured along the pipe and Idz / cixl cannot be greater than 1 (5280
Nmile).)
These figures may be used to estimate slack line velocities and cross sectional areasfor any pipeline. For
example, consider a pipeline that has a head gradient of 10 feet per mile and a velocity of 5 fVsec with slack
line flow in a steeply sloping section with a slope of 3000 feet per mile. Then, the slack line hydraulic gradient
is also 3000 feet per mile. The ratio of the slack line gradient to the tight line gradient is 300 (3000/10). From,
Figure 4, this corresponds to a normalized slack line velocity of about 9. The actual velocity in the slack region
is therefore 45 ft/sec (9*5 Msec). Figure 3 shows that the cross sectional flow area is approximately 0.11 of the
pipe cross section .
Figure 3 shows that the flow cross sectional area is less than 50% of the pipe cross sectional area for all but
very gentle gradients This brings us to our next basic conclusion about slack line flow.

Basic Fact 5

Thefiaction of liquid in slack regions varies slowly with elevation gradientfor all but very
gently sloping slack regions. Therefore, there is little needfor great detail in the elevation
gradient when simulating slack conditions in steep regions. However, for gently sloping
regions, the converse may be true.

One interesting feature of Figure 3 is the is that the hydraulic gradient (and therefore the frictional losses) can
be slightly less in slack line flow than in tight line flow for the samethroughput This is due to the fact that the
most efficient flow cross section is not full pipe flow. Rather, flow in a nearly full pipe is slightly more
efficient than full pipe flow. It would be nice to be able to operate pipelines in this state; however, this could
only be achieved by controlling the elevation gradient which is obviously not possible. In almost all practical
slack line flow conditions, the slack line flow frictional losses are far greater than the tight line losses. We
summarize this as follows:

Basic Fact 6

Thefrictional losses in slack line Jlow are almost always much greater than thefrictional
loses in tight line flow.

Velocity in Slack Regions As A Function of Tight Line Velocity

Finally, it is useful to plot the velocity in the slack region against the velocity in the tight line region for a given
elevation gradient. To do this, we will normalize each velocity by the “design” velocity, v,, . Solving for V,
using Eqn. 19 and normalizing by the design velocity gives the following equation:

Eqn. 20

The denominator in the radical is the frictional head loss per unit length (hydraulic gradient) of the line in tight
line flow at the design flow rate. Assuming that f,, / f is approximately 1, we now have an expression for
vo ’ vnm in terms of 8 and the ratio of the slack line gradient to the design pipeline hydraulic gradient. From

Slack Line Flow - A Tutorial Page 11 PSIG - October 19-20,1995

our previous equations we have relationships between v , V, and 8 so we can now plot the velocity in the
slack region versus the velocity in the tight line region normalizing both by the design velocity for specific

values of the relative elevation gradient, (-$/‘mm$I).ThisisdoneinFigure5,

Figure 5 - Slack Line Versus Tight Line Velocity

Slack Line Velocity versus Tight Line Velocity for Specified Elevation
Gradient
9
I

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Velocity of Tight Line Region as Fraction of Design Velocity

A notable fact is illustrated by this figure is that for a given elevation gradient, the slack line velocity is
generally several times greater than the tight line velocity. When a portion of the line goes slack, the velocity
swiftly changesto a much greater velocity.

Simulation of Transient Slack Line Flow

This section describes an approach to simulating transient slack line flow. The approach described has been
implemented in a real time pipeline model which is now installed on a major pipeline for the purpose of
providing real time transient tight and slack line pipeline simulation and leak detection capability. The model
described below has been rigorously tested over a wide range of slack line conditions including pipeline
shutdowns and startups and has handled all of these conditions robustly without failure.

In the preceding sections, the nature of steady state slack line flow has been described. However, much of the
interest in slack line flow behavior is associatedwith dynamic effects. Transient simulation software must be
able to simulate the onset, growth, and collapse of slack regions and the effects of these regions on the pipeline
hydraulics and packing rates. Furthermore, a slack line model should be able to survive the complete emptying
of pipeline sections (a significant difficulty).
A complete model of slack flow would include the thermodynamic effects of the liquid-vapor phase transition,
the mechanical effects of the friction between pipeline wall and the liquid phase, the interfacial friction between
the liquid and vapor phases, and the friction between the vapor phase and the pipe wall. One would probably

Slack Line Flow - A Tutorial Page 12 PSIG - October 19-20,1995

simplify the problem by assuming that the flow is stratified flow and ignore the possibility of bubbly flow and
other two phase flow regimes.
However, a purely mechanical model is sufficient to adequately simulate slack line flow in many pipelines and
to give a reasonable quantitative representation of the pressure profile and fluid volume in the slack region. A
reasonable mechanical model is to ignore the vapor phase altogether. This implies ignoring the volume of the
fluid in the vapor phase and the friction of the vapor phase against the pipe wall and against the liquid phase.
These are the same assumptions used in the steady state analysis of the preceding sections. The mechanical
model must represent the reduction in flow cross sectional area properly, the increase in flow velocity, and the
consequent increase in pressure drop in the slack region.
Under these assumptions, the slack line flow model usesmost of the same equations as the transient model of
tight line flow. Slack sections are modeled using the same continuity, momentum, and energy equations as used
in tight line flow. These equations are in five unknowns: pressure, temperature, velocity, density, and area.

Tight Line Simulation

The transient pipeline simulator must simulate both tight and slack regions and the transition between one flow
regime and the other.
The tight line pipeline flow hydraulics (transient and steady) are computed by solving the equations of
momentum, continuity, energy, the equation of state and an equation for the variation of pipe area with pressure
and temperature. These equations are presented below. (In addition, the equation for transient heat flow into the
pipeline surroundings is solved but not discussedhere.).

Tight Line Flow Equations

Eqn. 21, Momentum Equation

(VI+vvx)+Lpx +-f+l +$o

P 20 dx

Eqn. 22, Continuity Equation

Eqn. 23, Energy Equation

This formulation of the energy equation lumps the heat capacity of the pipe wall in with the heat capacity of the
fluid when considering the thermal massof the system. The temperature gradient term represents the heat flow
to the pipe surroundings.

Eqn. 24, Equation of State

P = J-w?

Slack Line Flow - A Tutorial Page 13 PSIG - October 19-20, 1995

 Eqn. 25, Flow Cross Sectional Area Equation for Tight Line Flow

where

For a length of pipe with ends free to move laterally:

For length of pipe with ends fixed (no lateral movement):

and
1 dA
--=2a,@,
4i aT
(In the above set of equations, the subscript “std” refers to conditions at atmospheric pressure and 60” F.)

Solution of the Tight Line Flow Equations By Implicit Finite Differences

This set of five equations is solved by finite differences, dividing each pipeline segment into a set of finite
intervals. The boundary of each interval is termed a knot and the five equations are written in terms of the
change (over a time step) in each of the five variables at each knot ( Api, Avi, AT, A/ii, Ap i where the
subscript i refers to the knot). If N is the number of knots, the number of unknowns to be solved for each time
step is 5N. One energy, area, and equation of state equation may be written for each knot for a total of 3N
equations. Also, N-Z continuity and N-Z momentum equations may be written (one for each interval between
knots. This results in 5N-2 equations. Two boundary conditions are required (e.g. A pressure at each end of
the segment) to give a total of 5N equations to solve for the 5N unknowns. (Actually, we reduce the number of
unknowns and equations to 4N by eliminating density as a variable using the equation of state.)
These equations are solved for the unknowns by matrix methods.

Slack Line Simulation

The Simplicity of Simulating Slack Line Flow
The following assumptions are used in the slack line simulation:
. The mechanics of the vapor phase may be ignored.

J;lVl($)” as in Eqn. 14.

l The frictional head loss in slack line regions is given by
8 gR;13
l The flow may be treated as one dimensional flow. This means that surface waves in the slack region
are not simulated.
l The flow in the slack region may be approximated as open channel flow (as described in the steady
state analysis earlier).
Under these assumptions, the continuity, energy, and equation of state for slack line flow are the same as for the
tight line regions. The momentum equation becomes:

Slack Line Flow - A Tutorial Page 14 PSIG - October 19-20, 1995

 Eqn. 26, Momentum Equation (Slack Line Flow)

(VI+vvJ+l-p, + filVl(S>“”+&+=o
P 8R,4/3

This is equivalent to the tight line momentum equation when the pipe is full ( R,, = D / 4 ) and therefore, this
form of the momentum equation may be used for both tight and slack line conditions.
Therefore, we have four of the five equations required to simulate slack line flow. All that is needed is to
obtain a replacement for the area equation.
In slack line flow, Eqn. 25, Flow Cross Sectional Area Equation, is no longer valid. Instead, the cross sectional
flow area reduces until the velocity is sufficient to give the appropriate frictional head loss gradient required to
maintain the pressure at the vapor pressure.
The approach to replacing the area equation seemsobvious: replace the area equation for knots in slack regions
with the following equation:

Eqn. 27, Slack Line Pressure Requirement

Then, after all 5N equations are solved for the 5N unknowns, the areasat the slack knots will be whatever is
required to maintain the pressures at those knots at vapor pressure.

The Complexity of Simulating Slack Line Flow

The approach outlined above can be implemented as a simple modification to a transient tight line pipeline flow
simulator, and is a good description of the basic approach that one may take to simulating slack line flow.
Unfortunately, as outlined above, the simplified approach is prone to instabilities and does not handle some of
the complexities of slack line flow. The following additional items must be dealt with for a robust simulation
of slack line flow:
1. If the flow goes to zero, some portions of the line will empty. In an empty line, none of flow
equations have any validity.
2. When slack line flow occurs, the behavior of the pipeline fundamentally changes. In a transient
pipeline flow simulation, equations are written at discrete points in the line and the fundamental flow
behavior may change many times as the different knots enter or leave slack line flow. Numerical
solutions of partial differential equations are prone to instabilities when the discontinuities like these
are imposed on the solution.
3. The collapse of slack regions may be very violent (this is discussed further in a later section) and may
cause instability in the simulation, particularly since the set of equations being solved are being
switched at the same time.

Handling Zero Flow

When the flow goes to zero in the pipeline, slack regions will empty. In this case, the flow equations have no
meaning. There are basically two ways of dealing with this problem:
1. Handle empty sections with a special set of (trivial) equations.
2. Simulate the pipeline as always having some fluid in it.

Slack Line Flow - A Tutorial Page 15 PSIG - October 19-20, 1995

Both approaches were tried and the latter was adopted. The first approach requires making a set of equation
substitutions and poses some difficult problems. The latter approach requires making some small modifications
to the physics of the problem but results in a stable simulation.
One way to simulate the pipeline as always having some fluid in it is to replace Eqn. 27 with the following
relationship:

A = (4 - Aminimum )f( Pvapor - P ) + Aminimum

The functionfmust have the following characteristics:

1. f(o) = 1. Th’1ssets the area equal to the pipe cross sectional area when the pressure is equal to the
vapor pressure.

2. f(Pwpor - P) must drop very quickly (within a couple of psi) to about 0.1. These meansthat the
cross sectional area will maintain the pressure in the slack region within a psi or two of the vapor
pressure for most slack line conditions.

3. The function approaches 0 asymptotically as pvaPOr- p + co. The area is never allowed to drop
be1ow Aminimum which is set as a very small fraction of the pipe cross sectional area.

This approach has proven to be very effective in obtaining a robust solution while at the same time maintaining
good accuracy in the simulation results.

Maintaining Stability During Flow Regime Changes at Knots

As any knot switches from slack line to tight line conditions (or vice versa), an instability in the solution may
develop in the solution. Handling this required adjusting some of the parameters in the finite difference
representation of the differential equations and adjusting computational time steps so that the transitions were
made as smoothly as possible.

Maintaining Stability During Vapor Bubble Collapse

The model of slack line flow presented here presumes that the vapor bubble will collapse as soon as the
pressure rises above the vapor pressure. This collapse may therefore be very violent. This phenomena was the
subject of a recent article.’ The violent collapse arises from the fact that the flow rates upstream and
downstream of the slack region are nearly independent. For instance, if the line is being repacked after a
pipeline shutdown, the flow downstream of the slack region may be nearly zero. However, pumps upstream of
the slack region will be attempting to refill the line and the flow entering the slack region may be at the fir11
design flow rate. This can persist until the slack region finally fills. The collapse of the vapor bubble can
appear like an nearly instantaneous valve closure.
The result of this collapse is a pressure transient whose rise time is as short as the time step over which the
bubble collapse actually occurred. This poses a dilemma for pipeline simulators. Generally, smaller time steps
help preserve stability during violent transients. However, in this case, smaller time steps may contribute to
more violent behavior. A particularly troublesome situation occurs when the multiple slack regions exist in
close proximity to each other. The violent collapse of one slack region may trigger violent behavior in the
neighboring slack regions.
Resolving difficulties with this type of behavior was perhaps the most difficult part of developing a robust slack
line flow simulator and required a significant amount of effort.

’ Burnett, R. R., “Predicting UnbalancedForces In Liquid Lines Can Avert Damage”, Oil & Gas Journal, Vol. 93, No. 13,
March 27, 1995.

Slack Line Flow - A Tutorial Page 16 PSIG - October 19-20, 1995

Open Issues for Research Related to Slack Line Flow
Simulation
The Problem of Bubble Collapse
The simulation approach described above assumesthat vapor bubbles are reabsorbed into the pipeline fluid as
soon as the line pressure rises above the vapor pressure. There is some cause to question this assumption. For
example, some pipelines operate in slack conditions for months on end. As the pipeline fluid passesthrough
the slack line region, the gasesthat are least soluble are most likely to vaporize and collect in the vapor bubble.
On some pipelines, this vapor bubble may be 10’s of thousands of barrels in size. When the slack region begins
to collapse, it is not clear that the vapor bubble will be completely redissolved in the fluid immediately.

The Problem of DRA

The simulator described in this paper was first developed for a pipeline in which DRA (drag reducing additive)
is used regularly to reduce pumping costs. Unfortunately, DRA degradeswhen subjected to large shear stresses
such as occur in the high velocities and turbulence of slack regions. The completeness of this degradation is not
well known and probably varies as a function of the slope of the slack regions.
Unfortunately, the computation of the size of the slack region depends directly upon the accurate computation
of the hydraulic gradient downstream of the slack region as is illustrated in Figure 1 . Therefore, uncertainties
in the DRA degradation in slack regions will result in uncertainties in the slack line model.

The Problem of Friction Factor

We have used a modification of the Fanning Equation, utilizing the Ri” dependency of the Manning Formula,
for frictional head loss in slack regions.

When using this approach, we have used the Moody friction factor using the relative roughness for the a pipe of
diameter D. Alternatively, one could ignore the Manning Formula and compute a Moody friction factor for a
pipe of diameter 4R, . In other words, the relative roughness would change as the height of the fluid in the
pipe changes. For fully developed turbulent flow, one may determine from the Moody friction factor curves
that the friction factor varies approximately as Ri1’5whereas the Manning formula predicts a
Ri”3 dependence (Eqn. 13).

Conclusion
Slack line flow is a very common phenomena in liquid pipelines and the simulation of slack line flow is often
necessary to provide a reliable representation of actual pipeline operations. Some aspectsof slack line flow
phenomena have direct impact on the safety of the pipeline operations, particularly the potential for water
hammer type effects upon the collapse of vapor bubbles.
Several figures have been presented in this paper that should be useful in estimating flow rates, velocities, and
slack line volumes. In addition, the basics of transient slack line flow simulation have been reviewed and the
development of a transient slack line flow simulator has been reported.

Slack Line Flow - A Tutorial Page 17 PSIG - October 19-20, 1995

Table of Symbols
Term Definition
A cross sectional flow area (ft2)
cpipe heat capacity of the pipe wall

C” specific heat at constant volume

D diameter of the pipe
dz elevation gradient (slope of the ground profile).
25
Young’s modulus
; friction factor
g acceleration of gravity (32.2 ft/s’)
H pipeline head.
4 frictional head loss per unit length
n parameter in Manning formula
P wetted perimeter
P pressure
P vapor vapor pressure of the fluid.
hydraulic radius (fi)
Rtl
T temperature
V velocity of the flowing liquid
W wall thickness of the pipe
X distance along the pipe
Z local elevation of the pipe
a pipe
coefficient for thermal expansion of the pipe material.
Ax length of a section of pipeline
Y specific weight (lbs/ft3)
K effective heat loss coefficient determined by the thermal shell model
CL Poisson’s ratio
P density of the fluid
0 angle used to specify height of fluid in partially empty pipe
density of the pipe material
Ppipe

Subscripts
x Refers to partial derivatives with respect to distance.
t Refers to partial derivatives with respect to time.
0 Refers to conditions in tight line flow, usually immediately upstream of the slack region.
max Refers to conditions at design flow rates.
std Conditions at atmospheric pressure and 60” F.

Slack Line Flow - A Tutorial Page 18 PSIG - October 19-20, 1995