Computational Mechanics

Enhanced Civil Engineering

Applications (WIND)

Prof. Dr.-Ing. Casimir Katz

Katz_13 / 2 Computational Mechanics
 Overview

Wind
Global Wind, local wind
Wind spectra, Aerodynamics
Design codes
Some Basic equations for more details (CFD)
Navier-Stokes Equation, Turbulence modelling
Boundary layers + Detachments
CFD Examples for Bluff Bodies
CFD Examples for Bridges
Dynamic Analysis for Bridges with wind

Katz_13 / 3 Computational Mechanics

 Wind Large Quantities of a light
medium with high velocities

Katz_13 / 4 Computational Mechanics

 Wind

Random, non deterministic:

Neither direction nor magnitude nor time
history are known in advance.
Dynamic
The natural wind is acting dynamically in
that way that wind speeds and thus forces
vary in time and space.
Meteorological observations + Statistics +
expected life time of structure
5 sec, 10 min, 1 year, 50 years, 100 years ?

Katz_13 / 5 Computational Mechanics

 The Davenport wind chain

Global Wind Terrain Local Wind

Aero Forces
Local Wind
dynamics

Forces Response Design

Katz_13 / 6 Computational Mechanics

 Dynamic Effects of wind

Gust (Ben) effects

(changing wind speed)
Periodic vortex shedding
(Karman effect, Rain-Wind induced oscillations)
Flutter + Interference
(Tacoma-Bridge)
Galloping
(Schlagschwingungen)
Theory VIII. Order
(Boundary layer effects)

Katz_13 / 7 Computational Mechanics

 Vortex Shedding

Katz_13 / 8 Computational Mechanics

 Global
Wind

Katz_13 / 9 Computational Mechanics

 Wind map of Germany

Katz_13 / 10 Computational Mechanics

 Wind map of Switzerland

Katz_13 / 11 Computational Mechanics

 Terrain Categories EN 1991-1-4

Z0=0.01 m

Z0=0.05 m

Katz_13 / 12 Computational Mechanics

 Terrain Categories EN 1991-1-4

Z0=0.30 m

Z0=1.05 m

Katz_13 / 13 Computational Mechanics

 The natural wind

U ( z ) u ( x, y, z, t )
1 zd
u v( x, y, z, t ) U ( z ) u* ln
z0
w( x, y, z , t )

Mean wind velocity U(z) as logarithmic profile

over the height
(exponential in some design codes)
Three turbulence components

Katz_13 / 14 Computational Mechanics

 Turbulence intensities
(Standard deviation / mean wind speed)

u ( z) v ( z) w ( z)
Iu ( z) Iv ( z) I w ( z)
U ( z) U ( z) U ( z)
Armitt / Hansen:

u A u* v 0.75 u w 0.50 u
A 2.5 for z0= 0.05 m und A 1.8 for z0= 0.30 m

Panofsky / Dutton:
u 2.39 u* v 0.80 u w 0.52 u

Katz_13 / 15 Computational Mechanics

 Turbulence intensities II

Simiu: u2 = bu*2

z0[m] 0.005 0.07 0.30 1.00 2.50

b 6.5 6.0 5.25 4.85 4.00

b 2.55 2.45 2.29 2.20 2.00

Katz_13 / 16 Computational Mechanics

 Turbulence intensities in the design code

v kI
Iv ( z) EN 1991-1-4 (4.7)
vm ( z ) c0 ( z ) ln( z / z0 ) kI = 1.0 A = 2.5

Terrain category I II III IV

z0[m] 0.01 0.05 0.30 1.05

u/vb 0.165 0.19 0.22 0.24

e 0.13 0.26 0.37 0.46

Windprofiles DIN EN 1991-1-4 NA.B/C u = const

Katz_13 / 17 Computational Mechanics

 Integral Length and Time Measures

T ( z ) ( z, ) d
T
u
L u ( z, rx ) drx
x
u
0 0

( z, ) exp
T

( )
u
T z

9 different Length measures!

Counihan:
Lux C z m Luy 0.33 Lxu Lzu 0.50 Lxu
Simu: 0.3
z
L L10
x
u Luy 0.30 Lux Luz 0.20 Lux
z10
Katz_13 / 18 Computational Mechanics
 Integral Length Measure in the Design Code

EN 1991-1-4 Appendix B.1

z
L( z ) Lt Lt 300 m ; zt 200 m 0.67 0.05 ln( z0 )
zt

Windprofile DIN EN 1991-1-4 NA/C

e
z
L( z ) 300 e seeTable1
300

Katz_13 / 19 Computational Mechanics

 Local Wind depending on Terrain

Superposition of Boundary layers 0 20 40 60

V
m/s m

for every change of roughness:

Mean Wind speeds
3 sec Gust Peak values

ATMO = 60.00
beta = 0.000

Turbulence intensities
Effective Wave length

Bauwerk
(Integral measures) V-BOE

Rauhigkeit [m]
0.050

0.300

Katz_13 / 20 Computational Mechanics

 Wind profile
Hhenprofil Windgeschwindigkeit, Turbulenz und effektive Wellenlngen LF 91

V-BOE
m
LVER

LLAT

LLON

TVER

TLAT

TLON
V
40

30

20

10

0
0 10 20 30 40 50 60 m/s V, V-BOE

0 1 2 3 4 5 6 m/s TLON,TLAT,TVER

0 100 200 300 400 500 600 m LLON,LLAT,LVER

Katz_13 / 21 Computational Mechanics

 Wind profiles in Design Codes

Category 0 to IV according to Eurocode

Mixed profiles for rural and coastal regions in DIN
Provisions for
Wind pressure
Mean wind speed
Gust Speed = max. Design Pressure
Longitudinal Turbulence Intensity
Longitudinal Integral Length Scale
Simplified constant wind loadings on the save
side

Katz_13 / 22 Computational Mechanics

 Wind profiles

Continuous function of speed and pressure along

the height:
For rural regions the gust wind speed according
to chapter 10.3 is given :
For z 7m: q(z) = 1,5 qref
For 7m < z 50m: q(z) = 1,7 qref (z/10)0,37
For 50m < z 300m: q(z) = 2,1 qref(z/10)0,37

Katz_13 / 23 Computational Mechanics

 Wind-Spectra

Katz_13 / 24 Computational Mechanics

 Wind-Spectra
f S a1 X a2 X 2 a3 X 3

1 b X b
2 c
2
1 2 X
f f L
X
f0 V
a1 a2 a3 b1 b2 C

Karman 4.000 0.0 0.0 0.0 70.8 0.8333

longitudinal
Karman lateral 4.000 0.0 3021 0.0 283.0 1.8333

Davenport (long.) 0.0 0.6667 0.0 0.0 1.0 1.3333

Harris (long.) 0.337 0.0 0.0 1.5 0.0 0.8333

EC 6.800 0.0 0.0 10.2 0.0 1.6667

Katz_13 / 25 Computational Mechanics

 Spectral Analysis

Spectra are defined deterministic Energetic

quantities for selected frequencies.
Each frequency of a structure may be exited with a
value derived from the spectra

S y ( f ) a m S` z ( f )
2 2

a aerodynamic transfer function

m mechanical transfer function

Katz_13 / 26 Computational Mechanics

 Spectral Analysis

But for the superposition of the individual modes

we have a random effect introduced by the phase
of the response.
Coherencies (Kohrenzen) define how similar are
two wind loadings at neighboured locations.
Mathematics not easy, some key words to search
for:
Autocorrelation function
Normal distribution / Weibull-distribution

Katz_13 / 27 Computational Mechanics

 Time History Analysis

From the spectra and coherences we may generate

time histories for the wind speed at every point via
a Fourier transformation and a random generator.
This wind is acting on our structure.
All non linear effects may be accounted for.
You see what happens (Animated movements)
Tuned mass dampers (Tilger) are easy to include.
To assure the safety of the structure several
independent transient runs are needed.

Katz_13 / 28 Computational Mechanics

 Wind pressure coefficients

Katz_13 / 29 Computational Mechanics

 Wind pressure coefficients

Katz_13 / 30 Computational Mechanics

 Wind on walls

Katz_13 / 31 Computational Mechanics

 Wind on roof

Katz_13 / 32 Computational Mechanics

 Wind pressure coefficients
The missing link !
Current Steps for a wind analysis:
Profile of wind along the height and Spectra
Pressure coefficients from literature or wind tunnel
(transient in general)
Instability effects (Buffeting, Flutter etc.)
A CFD analysis could model the last two steps.
However this requires considerable effort and
experts argue, that they are not reliable at all.
Robust methods should yield acceptable
results even for coarse meshes!

Katz_13 / 33 Computational Mechanics

 Motivation for CFD

From: CFD Round Table Madrid 2007

January jj / 34 Wind Analysis

 A Bad Example for a Wind Tunnel Test

Vela Hotel
M.O.Cornejo
F.M. Mato
IDEAM S.A.
IABSE 2010

No boundary layer
Blockage to high
CFD used instead

January jj / 35 Wind Analysis

 Firth of Tay
Insufficient Design and Workmanship

Katz_13 / 36
36 Computational Mechanics
 (28.12.1879, after train passing at 11-12 Bft)

Katz_13 / 37
37 Computational Mechanics
 Bridge over Firth of Forth

Consequence:
The bridge over the Firth of Forth
has been designed on the safe side
(i.e. 5-times) Wind loading.

Katz_13 / 38
38 Computational Mechanics
 Basic Equations for Fluid Dynamics

Fluid Dynamics = Many Formulas !?

Mass, Density
(also buoyancy for thermal problems)
Compressibility (may be neglected in many cases)
Conservation of mass, energy and momentum
e.g. law of Bernoulli derived from conservation of
energy:

p u 2 g z const
2

Katz_13 / 39 Computational Mechanics

 Scales of Forces of Fluids on Bodies

Inertial force
Fi 1/2AU lU
Viscous force
Fv mSdU/dH mlU
Gravitational force
Fg gV gl

Froude Number F = Fi / Fg = U/(gl)

Reynolds Number Re = Fi / Fv = Ul/m
Katz_13 / 40 Computational Mechanics
 Viscosity - Reynolds number

d u d u kinematicVis cos ity

Re
m m dynamicVis cos ity
2 2
m m
Air 15 10 6 ; Water 1.31 10 6
sec sec

Ground water:
d = 0.001 ; u = 0.01 Re = 10
Aerodynamics:
d = 10 ; u = 15 Re = 107

Katz_13 / 41 Computational Mechanics

 Navier-Stokes-Equation

ui u j ui ij 2 u j
p m gi
t x j x j xi 3 x j

u = Velocity
= Density
= Reynold Shear stress => turbulent viscosity
p = Pressure
m = dynamic Viscosity

Katz_13 / 42 Computational Mechanics

 Overview

RANS
Reynolds Averaged Navier Stokes
Complete Model of Turbulence, very effective
LES
Large Eddy Simulation
Model small scales, calculate large scales
DNS
Direct Solution of Navier Stokes-Equations
Calculate all effects,
limited to low Reynold numbers < 20000

Katz_13 / 43 Computational Mechanics

 Turbulence modelling
u(t ) ui ux' (t )

Kinetic Energy of turbulence k [m2/sec2]

1 ' ' 1
k u i ui U I x U I y U I z 0.9 U I x
2 2 2 2

2 2
Length Scale Lt

Dissipation rate e [m2/sec3] or Frequency [1/sec]

k 3/2 k 1/2
e 0.168
Lt Lt
Katz_13 / 44 Computational Mechanics
 Boundary Conditions

Inflow u (= supported Edge in statics)

Smooth Boundary Rand u/n (= free edge in statics)
Wall with friction

du Shearstress

dz Viscosity

laminar

turbulent
turbulent
laminar

Katz_13 / 45 Computational Mechanics

 Boundary Layer

Katz_13 / 46 Computational Mechanics

 Boundary Layer is not easy !

Katz_13 / 47 Computational Mechanics

 Modelling errors with a boundary layer

Turbulent energy for same input velocity but

different turbulent parameters:

Katz_13 / 48 Computational Mechanics

 Logarithmic Wall Law
u( z ) 1
u+
u *

k
ln Ey

u* /
y
y
y *
u
ln y+
1 20 300

Near Wall models y+<1 (k-)

Far Wall models y+>20 (k-e)
Mixed models
Katz_13 / 49 Computational Mechanics
 The full picture

Katz_13 / 50 Computational Mechanics

 Logarithmic wall law versus wind profiles

u( z ) 1 z 1
u *
ln
k z0 k
ln Ey

u* /

Katz_13 / 51 Computational Mechanics

 Boundary Layer
and change of roughness

Katz_13 / 52 Computational Mechanics

 Problem of the atmospheric boundary layer
There are two common velocity distributions
Exponential v = vref * (z/z0)
Logarithmic v = vref * ln (z/z0)
We need sound values for turbulent k and e
There is an analytic solution for k-e !

u*ABL z z0
u( z ) ln
0
z
u*ABL
2
k ( z)
Cm
u*ABL
3
e ( z)
z z0

Katz_13 / 53 Computational Mechanics

 Modelling with the analytic solution

Boundary condition at top required

There is a new boundary layer for small elements

Katz_13 / 54 Computational Mechanics

 Detachment

Stall point
Accelerated
Flow
Retarded
Flow
Pressure
increase
Detachment
Reattachment

Katz_13 / 55 Computational Mechanics

 Drag coefficients for Cylinder

Katz_13 / 56 Computational Mechanics

 Flow + Pressures

Katz_13 / 57 Computational Mechanics

 Mesh sensitivity (Reynolds Number = 100)

Level Volume Time Cd Cl S D-p CPU

elemen steps
t
2 552 500 2.28 -0.002 - 1.566 9

3 2208 500 2.96 -0.020 0.210 2.122 1205

4 8832 1000 3.12 0.877 0.261 2.610 20443

4 8832 500 2.97 0.558 0.263 2.331 32324

4 8832 200 2.96 -0.014 - 2.278 11982

Refer. 3.23 1.100 0.300 2.480

Katz_13 / 58 Computational Mechanics

 Cylinder Crisis

Low Reynolds Flow

Pressure before equals approximately pressure
behind: drag mainly caused by friction
With increasing Reynolds numbers, shear forces
become less
Detachment
we have a suction at the end of the cylinder
thus a resulting force from pressure
Reattachment
A positive pressure reduces the drag coefficient

Katz_13 / 59 Computational Mechanics

 Teamgeist

Analogue for spheres

A foot ball will be shot in the trans critical region
If it becomes slower the drag coefficient will raise
quite suddenly, the ball will drop down.
The optimisation of the roughness of the ball to
achieve the best curve for a goal keeper shot is a
really difficult task.
Katz_13 / 60 Computational Mechanics
 Turbulence models RANS

Variants
Standard k-e Model
ReNormalisation Group k-e Model (RNG)
k-e Model according Murakami, Mochida and Kondo
(MMK)
Standard k- model
SST model (Mixture from k-e model and k- model)
Remarks
For the k-e model the boundary layer is described by a
wall law y+ >30
For the k- model the boundary layer is resolved y+ < 1
(the finer the better)
Meshes are sensitive to the geometric shape, hybrid
meshes are strongly recommended

Katz_13 / 61 Computational Mechanics

 Comparison Turbulence model
Standard - RNG

Turb. Energy
Re = 186000

Turb. Energy
Re = 670000

Katz_13 / 62 Computational Mechanics

 Pressures
Re = 0.186 / 0.670 / 11.400 106

Katz_13 / 63 Computational Mechanics

Results Re = 46 500

Pressures for cylinder

4 % blockage 1 % blockage
Cd cp- cp- cp- Cd cp- cp- cp-
luv min lee luv min lee
Reference 1,200 -1,00 1,29 1,25

k-e standard 0,720 -1,03 1,66 0,60 0,673 -1,05 1,54 0,51

k-e RNG 0,916 -1,02 1,16 0,77 0,839 -1,02 1,25 0,63

k-e MMK 0,791 -1,01 1,50 0,66 0,735 -1,02 1,40 0,57

k- 0,424 -1,02 2,31 0,42

 Results Re = 186 000

Pressures for cylinder

4 % blockage 1 % blockage
Cd cp- cp- cp- Cd cp- cp- cp-
luv min lee luv min lee
Reference 1,200 -1,00 1,10 1,20

k-e standard 0,590 -1,08 1,96 0,46 0,563 -1,09 1,87 0,40

k-e RNG 0,895 -1,02 1,57 0,72 0,796 -1,03 1,25 0,60

k-e MMK 0,745 -1,02 1,61 0,64 0,701 -1,03 1,45 0,55

k- 0,429 -1,09 2,30 0,32

 Results Re = 670 000

Pressures for cylinder

4 % blockage 1 % blockage
Cd cp- cp- cp- Cd cp- cp- cp-
luv min lee luv min lee
Reference 0,42- -1,00 2,45 0,20
1,00

k-e standard 0,540 -1,08 2,08 0,39 0,517 -1,10 2,00 0,34

k-e RNG 0,878 -1,02 1,63 0,75 0,779 -1,03 1,39 0,60

k-e MMK 0,724 -1,02 1,69 0,62 0,675 -1,03 1,52 0,53

k- 0,391 -1,10 2,40 0,22

 Results Re = 11 400 000

Pressures for cylinder

4 % blockage 1 % blockage
Cd cp- cp- cp- Cd cp- cp- cp-
luv min lee luv min lee
Reference 0,63- -1,00 1,55 0,50
1,04

k-e standard 0,474 -1,08 2,24 0,27 0,458 -1,10 2,15 0,21

k-e RNG 0,933 -1,02 1,72 0,96 0,725 -1,03 1,55 0,55

k-e MMK 0,650 -1,02 1,82 0,58 0,622 -1,02 1,66 0,51

k- 0,336 -1,10 2,67 -0,18

 First Remarks

There are some problems for CFD

(It is nearly impossible to calculate the problem of
a cylinder for 200 000 < Re < 600 000 )
Time step and mesh size are critical
Where is the detachment / reattachment ?
Rotating Flow is also known to be very difficult
Wall law Roughness Formulations ?

Katz_13 / 68 Computational Mechanics

 Cylinder in boundary layer

Reynolds number 20000

Profile exponent 0.14 / 0.22
Blockage 1.3 %

Katz_13 / 69 Computational Mechanics

 Longitudinal section

Katz_13 / 70 Computational Mechanics

 Pressures

Pressures very well

Suction at sides to large by a factor of 1.7
Suction at tail to small by 30 %
Katz_13 / 71 Computational Mechanics
 Problematic Building

Katz_13 / 72 Computational Mechanics

 Parameter / mesh generation

Re = 4107, profile exponent 0.25

Dimension B/H=55/25 m
Mesh density with 0.25 cm (30 y+) => 12500
volumes

Katz_13 / 73 Computational Mechanics

 Pressures

RNG

MMK
STD

Katz_13 / 74 Computational Mechanics

 Turbulent Energy

STD

RNG

Katz_13 / 75 Computational Mechanics

 Reduction of viscosity
for inflow

RNG

Katz_13 / 76 Computational Mechanics

 Assessment

Maximum values are quite precise

The flow detaches in the experiment at the middle of the roof,
in the analysis only in the last end of the roof.
Effect is well known for steep air foils, detachment occurs in
the mid range, all RANS are within the last 10 %
=> LES / DES
Questions of Reynold scaling effects are still open

Katz_13 / 77 Computational Mechanics

 Rectangular towers

Pressures:

Front
Side
Rear

Katz_13 / 78 Computational Mechanics

 Pressures

Much better pressure values achievable

Katz_13 / 79 Computational Mechanics
 Wind on flat roof

Measurements from
Prof. Gerhardt
Suction peak
values at the edge
Height of roof 11.20
m
v = 22.5 m/sec
q = 316 N/m2
Mixed profile

Katz_13 / 80 Computational Mechanics

 Wind on Flat roof

Katz_13 / 81 Computational Mechanics

 Coarse mesh pressure distribution

Katz_13 / 82 Computational Mechanics

Pressure distribution

Katz_13 / 83 Computational Mechanics

 Velocity distribution

Katz_13 / 84 Computational Mechanics

 Comparison of Results
cp-d cp-d cp-s cp-s
peak value mean peak value middle/far
value end
Measurements - - -2.8 -0.6 ?
Coarse mesh +0.81 +0.56 -1.06 -0.45
without turbulence

Coarse mesh +1.06 +0.66 -1.25 -0.35

with turbulence

Intermediate mesh +1.35 +0.60 -1.06 -0.27

with turbulence

Fine mesh +1.63 +0.71 -1.55 -0.50

with turbulence
Fine mesh II +1.73 +0.69 -2.20 -0.35
with turbulence

Values according - +0.70 -1.80 -0.70/+0.20

DIN 1055

Katz_13 / 85 Computational Mechanics

 Measurements at the Silsoe Cube

z U Lux
[m] [m/sec] Iu Iv Iw
[m]

1 6.97 0.243 0.196 0.077 11

3 8.65 0.208 0.166 0.072 33

6 9.52 0.193 0.150 0.078 53

10 10.13 0.186 0.151 0.083 62

Katz_13 / 86 Computational Mechanics

 Turbulence Intensities at the Silsoe Cube

z0=0.008 u*=0.575 m/sec z0=0.01 u*=0.63 m/sec z0=0.05 u*=0.94 m/sec

z U K U K U K
[m] [m/sec] [m/sec] [m/sec] [m/sec] [m/sec] [m/sec]

1 6,94 1,10 7,25 1,32 7,04 2,95

3 8,52 1,10 8,98 1,32 9,62 2,95

6 9,52 1,10 10,08 1,32 11,25 2,95

10 10,25 1,10 10,88 1,32 12,45 2,95

Katz_13 / 87 Computational Mechanics

 Integral Length Scales at the Silsoe Cube

z L[m] L[m] L[m] L[m]

[m] measured Eurocode DIN/NAD CFD

1 11 31 170 0.41

3 33 50 189 1.23

6 53 67 203 2.46

10 62 83 214 4.09

Katz_13 / 88 Computational Mechanics

 Measurements and LES Analysis

Katz_13 / 89 Computational Mechanics

 Fluid velocities with RANS

Katz_13 / 90 Computational Mechanics

 Pressures
of RANS
Analysis

Significant influence
of the boundary
layer
Low influence of the
turbulence intensity
Which pressure
distribution on the
roof is correct ?

Katz_13 / 91 Computational Mechanics

 A larger example

Katz_13 / 92 Computational Mechanics

 Highlight Towers Munich

H = 129 m
B = 48 m
D = 78 m

Katz_13 / 93 Computational Mechanics

 Structured Mesh

150 000 to
500 000
elements

Katz_13 / 94 Computational Mechanics

Which Wind Profile ?

Wind zone DIN 1055 Expertise

WZ 2
vref = 25
Land
m/sec III IV III IV
Side

vtop 1.50 vref 1.35 vref 1.20 vref 1.33 vref 1.16 vref

qref,top
kN/m2 0.880 0.708 0.562 0.691 0.572
 Pressures

Katz_13 / 96 Computational Mechanics

 Pressures are to high !

Cp >> 1.0 ?
What may be the reason ?
Law of Bernoulli is not valid for viscous flow with shear
stresses, but the viscosity should reduce the energy
available to be converted in pressures.
The turbulent kinetic energy increases the pressures,
but the effect is not as large as observed here. A
turbulence intensity of 10 % will increase the total
energy by only 1 %, but an increase of the velocity by 10
% will increase the pressure by a factor of 1.21.
If the flow is accelerated locally and hits a body behind,
the pressure will be larger of coarse.

Katz_13 / 97 Computational Mechanics

 What to do ?

Select a finer mesh ?

Change some boundary conditions ?
Larger Mesh ?
Structured Mesh ?
Change Integration parameters (app. 50 values)
e.g. Relaxation, Solver-Type
Transient Solution ?
Quasi-steady state solution with a rather large
time step of 10 to 100 times Dh/v

Katz_13 / 98 Computational Mechanics

 Try effects with a 2D Mesh !

Katz_13 / 99 Computational Mechanics

 Squeezed Flow

Katz_13 / 100 Computational Mechanics

 Free Flow

Katz_13 / 101 Computational Mechanics

 Free Flow - Detail

Katz_13 / 102 Computational Mechanics

Effects of mesh size and
total meshed region
Distance No. Nodes Pressure Suction
40 m 1154 2078 7683
13365 2020 7770
60 m 1749 1136 1903
6620 1153 1910
160 m 2519 1030 1200
8664 1162 1130
20257 1212 1087
Maximum value vref = 37.24 867
 Skewed mesh

Katz_13 / 104 Computational Mechanics

 Karman !

Katz_13 / 105 Computational Mechanics

 Pressures from Experiment

Katz_13 / 106 Computational Mechanics

 Pressures 2D Analysis Boundary
Conditions Type I

Katz_13 / 107 Computational Mechanics

 Pressures 2D Analysis Boundary
Conditions Type II

Katz_13 / 108 Computational Mechanics

 Pressures 2D Analysis Boundary
Conditions Type IIw

Katz_13 / 109 Computational Mechanics

 Pressures 3D ; a = 80 m

Katz_13 / 110 Computational Mechanics

 Pressures 3D ; a = 160 m

Katz_13 / 111 Computational Mechanics

 Pressures 3D ; a = 320 m

Katz_13 / 112 Computational Mechanics

 Pressures 3D ; a = 480 m

Katz_13 / 113 Computational Mechanics

 Comparison of Pressures

Distance / h Nodes Pressure B2 Pressure B1 Suction B1

80.0 / 2.5 m 80717 2261 2057 -1077 576

160.0 / 2.5 91875 839 874 - 194 517

160.0 / 1.0 783738 1068 1055 - 238 562

320.0 / 2.5 111325 1034 1006 - 272 550

320.0 / 1.0 1079473 852 793 - 215 409

480.0 / 1.0 1079189 847 787 - 199 420

Measureme 502 687 - 284 250
nts
DIN-Values GKL II 876 ? 438
 Velocities along the wind

Katz_13 / 115 Computational Mechanics

 Pressures along the wind

Katz_13 / 116 Computational Mechanics

 Benchmark for a 2:1:1 building

Yoshie, Mochida, Tominaga, Kataoka, Harimoto et. al.

Comparative project for CFD Prediction of wind env.
J. Wind Engineering & Industrial Aerodynamics, 2007

January jj / 117 Wind Analysis

 Practical Example

Two connected high

rise buildings with a
height of 146 m at
Krasnobogatirsky
Place in Moscow
Use of CFD for the
analysis of wind loads
instead of a wind
tunnel test is not
excepted very often,
because there are
known defects of the
numerical procedures

Katz_13 / 118 Computational Mechanics

 Wind map of Russia in SNIP

Katz_13 / 119 Computational Mechanics

 Environment

located in the vicinity of buildings with up to 4 storeys

and a forest, which can not be modelled with
reasonable effort, thus we had to select a boundary
layer wind profile based on the SNIP 2.01.07-85 wind
design code.
Moscow is in Wind zone I yielding a specified wind
pressure of 230 Pa.
The SNIP provides in Table 6 and 7 pressure factors k
and pressure pulses -gust along the height for a
given roughness class B. From those two values we
may get the mean wind velocity and the turbulence
intensity:
kg vgust vmean
vmean 2 p / 1.25 3 vgust vmean 1 I
m 2.65 vmean
Katz_13 / 120 Computational Mechanics
 Complete Wind profile
Distribution of speed, turbulences and effective wave lengths LC 100

v-gust ATMO = 41.24 AT 2106

TLON
TVER

LLON
LVER

TLAT
LLAT
m

V
500

450

400

350

300

250

200

150

100

50

0
0 10 20 30 40 50 60 70 80 90m/s V, V-gust

0 1 2 3 4 5 6 7 8 9 m/s TLON,TLAT,TVER

0 100 200 300 400 500 600 700 800 900m LLON,LLAT,LVER

Katz_13 / 121 Computational Mechanics

 Turbulence Model

Reynolds number of about 5 107

Turbulence model is required
standard k-e model and the hybrid difference
scheme, developed by Spalding is very robust.
standard k-e model over estimates the turbulence
production in the stall point, and we might use
other models like the MMK variant:
Reduction of pressure in stall point,
theoretical maximum value is 1.0 (!)
Increase of suction values at the rear side
other difference schemes may be used, but
difficult to decide. We have not found significant
improvements

Katz_13 / 122 Computational Mechanics

 Inflow Boundary Conditions

We have to specify the wind speed and the

turbulence parameters.
The turbulence energy can be taken from the
SNIP specifications directly as stated above.
We have also provisions for the turbulence length
scale either by the integral wave length specified
in the Eurocode between 120 and 220 m or the
horizontal dimensions of the building with about
35 m. We then may use this length to calculate
the dissipation constant e by the formula: k 3/ 2
e
Lt
Katz_13 / 123 Computational Mechanics
 Boundary Conditions
1
1. Inflow Boundary 1 3a
Velocities
Turbulence parameters
2
2. Outflow boundary
Velocities 2
3. Pressures 1
Arbitrary
reference point
Outflow Area
4. Walls + Surface
Logarithmic wall law

3
Katz_13 / 124 Computational Mechanics b
 Wall Boundary Condition
Logarithmic Wall Law
Boundary Layer with structured mesh

Katz_13 / 125 Computational Mechanics

 Mesh generation

Model a sufficiently large air volume around the building

A blocking of the stream surface with more than 3 percent
increases the obtained pressures considerably, it is
recommended not to exceed 1 percent
Model a proper boundary layer for the wall boundary condition we
have to take special care on the mesh generation.
An unstructured mesh behaves very poor when modelling the
boundary conditions, convergence becomes very difficult.
As the Reynolds numbers are quite high we should not use
meshes with y+ values less than 50
An average mesh size of 0.90 m with a smallest size of 0.20 m
at some corners of the surface of the building was selected

Katz_13 / 126 Computational Mechanics

 Isotropic Mesh
Allowing Wind from any direction

Katz_13 / 127 Computational Mechanics

 Steady state solution

False Time Step

Katz_13 / 128 Computational Mechanics

 2D pressures small mesh

Katz_13 / 129 Computational Mechanics

 2D-pressures large mesh

Katz_13 / 130 Computational Mechanics

 Resultant forces Fx
M
X

Fy

Krasnobogatirsky Building I

2,500
2,000
1,500
1,000
0,500 Cx-I
coeff

0,000 Cy-I
-0,500 0 30 60 90 120 150 180 210 240 270 300 330 Cm-I
-1,000
-1,500
-2,000
-2,500
incidence

Katz_13 / 131 Computational Mechanics

 Transient Results of Interaction
1st Building, 60 degree attack

spring forces SPX group 12

spring forces SPY group 12 20000.0

20000.0

15000.0
15000.0

10000.0
10000.0

5000.0
5000.0

0.0 Time 0.0 Time

10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 [sec] 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 [sec]

-5000.0 -5000.0

-10000.0 -10000.0

-15000.0
-15000.0

-20000.0
-20000.0

Katz_13 / 132 Computational Mechanics

 Transient Results of Interaction
2nd Building, 60 degree attack
spring forces SPY group 11

20000.0

15000.0

10000.0

5000.0

0.0 Time
10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 [sec] spring forces SPX group 11
20000.0
-5000.0
15000.0
-10000.0
10000.0

-15000.0
5000.0

0.0 Time
10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 [sec]

-5000.0

-10000.0

-15000.0

-20000.0

Katz_13 / 133 Computational Mechanics

 Dynamic Assessment
1,6
100
f 0.4 60 m H 200 m
H

Estimate of Eigen frequency 0.20 Hertz

Numerical Analysis 0.138 respective 0.171 Hertz
Vortex shedding of 0.079 Hertz and thus a Strouhal number
of 0.11.
We may further estimate a coefficient c-lat of about 20% of
the longitudinal value which would then be for the first
building
clat = 0.18
For the second building we obtained a dynamic coefficient:
clon = clat = 0.45

Katz_13 / 134 Computational Mechanics

 Galopping

Take the
drag
coefficients
from the
grid of
attack
angles and
create an
estimate for
the local
derivatives

Katz_13 / 135 Computational Mechanics

 3D Meshing

2D mesh generator has

created a quadrilateral
surface mesh. From that
mesh a triangular mesh is
generated to become the
advancing front of the
unstructured tetrahedral
mesher. But before
starting the boundary layer
is created by 6 noded
wedge-type volume cells.
numerical effort for a three
dimensional analysis is
much higher (1.3 Mio finite
volumes)
Steady state solution only

Katz_13 / 136 Computational Mechanics

 Resulting drag coefficients along the height

Krasnobogatirsky Building I

1,200
1,000
0,800
Cx-I
coeff

0,600
Cy-I
0,400
0,200
0,000
5 35 65 95 5
12
height

Katz_13 / 137 Computational Mechanics

 Pressure
distribution

Katz_13 / 138 Computational Mechanics

 Observations

The maximum pressure coefficients are close to 1.0, only at

some edges we have local higher values. As it is known,
that the k-e model creates slightly higher pressures, the
results are in an acceptable range.
Maximum suctions are obtained at the edge of the roof up
to 3.80 which is in good agreement with expected values.
On the wind front we have a distribution of pressures
according to the wind profile, but there is a nearly constant
suction along the height at the sides and the rear.
There is only a small disturbance of the flow pattern by the
other buildings.

Katz_13 / 139 Computational Mechanics

 Comparison 2D and 3D @ 100m

Cx - I Cy-I Cx - II Cy - II

2D 0.889 0.390 1.037 0.871

3D 0.981 0.357 1.073 0.654

From that it may be justified to make the global analysis of the

structure based on the 2D drag coefficients, and to use the 3D
pressure coefficients only for local effects or a refined analysis.
A transient 3D analysis has not been established therefore.
Analysis Time between several hours and 2 days for the
transient analysis of 12 directions on a standard Windows AMD
Computer

Katz_13 / 140 Computational Mechanics

 Conclusions for bluff bodies

CFD Analysis is better than expected!

The may produce a better feeling for the qualitative
effects of the flow.
A wind tunnel test it still necessary in many cases!
To what extend you want to know the wind loads ?
There is no single solution, we have to cope with
limit cases.
A non linear transient analysis with Fluid-Structure-
Interaction may be used for new design concepts,
not available with a wind tunnel alone.

Katz_13 / 141 Computational Mechanics

 Mesh Generation Issues
(Rhino STL Export NETGEN Tetrahedral meshing in SOFiMSHC)

January jj / 142 Wind Analysis

 Finite Volume Types

Tetrahedral:
easy meshing
Very poor numerics
Hexahedral:
difficult meshing
good numerics
Polyhedral:
Best of both ?

January jj / 143 Wind Analysis

 Why polyhedral cells?

Minimize the total number of faces / cells

More neighbours (ca 14 instead of 4 or 6)

January jj / 144 Wind Analysis

 Performance Test

Tetrahedral Mesh: 6168 cells

Polyhedral Mesh: 1336 cells

Hexahedral Mesh: 1331 cells

January jj / 145 Wind Analysis

 Performance Test

Residual is not the error of the solution!

Residual

Residual

Residual

Residual
v-x v-y v-z

mass
[m/sec [m/se [m/se
v-x

v-y

v-z
] c] c]
715 Tetrah.
800 7.8E-6 7.1E-7 3.0E-6 5.6E-5 11.4 0.66 0.96
Iterations
211
Polyhedra
6.3E-8 7.6E-9 9.7E-9 2.9E-7 10.0 0.005 0.02
800
Iterations

Reference 10.0 0.0 0.0

January jj / 146 Wind Analysis

 Convergence of larger meshes

Analysis Convergence
1,000E+00
0 10 20 30 40 50 60 70 80 90
1,000E-01

1,000E-02

1,000E-03
Ux residuum

1,000E-04

1,000E-05

1,000E-06

1,000E-07

1,000E-08
Iteration

Tetrahedral Mesh Polyhedral Mesh Hexahedral Mesh

January jj / 147 Wind Analysis

 Silsoe Cube

January jj / 148 Wind Analysis

 Reference: Measurements and LES-Analysis

January jj / 149 Wind Analysis

 Tetrahedral - Polyhedral k-e-RNG Pressures

January jj / 150 Wind Analysis

 Wind on Bridges
Famous Example of Tacoma Narrows Bridge
Other Examples Great Belt Eastern Bridge

Katz_13 / 151 Computational Mechanics

 Wind dynamics for bridges

Gust-response
(Buffeting)
Vortex excitation
(Karman)
Aero elastic damping
(positive or negative)
Torsional divergence
Galloping
Flutter
Interferences

Katz_13 / 152 Computational Mechanics

 gust effects
the Eurocode 1991-1-4 procedure
S F (f ) f T
2F S x H x SF H x
Last-Spektrum

ln f
H
2 f1 f2
rF

HxF1
HxT HxF2
Mechanische bertragung

S (f ) f

F
2 .. f1 f2
ln f
S
(f ) f
F

2r
Hx HxF1 HxF2
Last-Spektrum
..
Antwort-Spektrum
ln f
H
2 f1 f2
rF ln f
f1 f2

Mechanische bertragung

ln f
2x Sx (f) df
f1 f2 0
Sr ( f ) f
2r
Katz_13 / 153 Computational Mechanics
 Aero elastic damping
cl()
U U

v cd()

v U

A body in the wind moves in transverse direction

Resulting in a virtual change of the angle of attack by
= atan(v/U)
The resultant vertical drag coefficient is
cv() = cl()/cos() + cd() tan()/cos()
For small angles one may use cos()=1 and tan()=
cv() = cl() + cd() (v/U)

Katz_13 / 154 Computational Mechanics

 Aero elastic damping

Dependant on the shape of the section the curve of the

drag coefficient may be quite different, especially the
derivatives may vary to a great extend.

Katz_13 / 155 Computational Mechanics

 Aero elastic damping

For the change of the drag force within time we replace the
function of the coefficient by its Taylor series and remove
all constant parts:

U 2 v cl
DF H cd
2 U 0
As this force is now proportional to the speed of the section
it may be introduced in the equation of motion by an
additional damping. (m = mass of oscillator)

U cl
D Dstruct H cd
2m 0

Katz_13 / 156 Computational Mechanics

 Galloping

Dependant on the sign of the derivative we have a positive

or negative aero-elastic damping.
For the latter case we have a certain critical wind speed
where the total damping becomes negative. Then the
oscillation becomes instable and the amplitudes grow
steadily.
If exchanging the modal damping by the logarithmic
decrement , we get the stability criteria:

2 m
U crit
c f
H
0

Katz_13 / 157 Computational Mechanics

 Torsional divergence

There is a second type of instability given for the case that

the change of the torsional moment per change of rotation
becomes larger than the structural torsional stiffness.
Again we have a critical wind speed for that type of failure:

2 k
U crit
c
B 2 m
0

Katz_13 / 158 Computational Mechanics

 Flutter

If the coupling of the rotation with the transverse

displacement is taken into account, we have forces and
moments depending on displacements, rotations and their
derivatives in time.
Scanlan has suggested 8 coefficients to describe this
highly non linear effects in a linearized form
By that process the height H is replaced by the width B and
from the Eigen frequency of the structure we get a reduced
flutter frequency k:

U 2 f B
U red ; k
f B U

Katz_13 / 159 Computational Mechanics

 Scanlan Derivativa

Katz_13 / 160 Computational Mechanics

 Flutter diagram according Klppel-Thiele

Katz_13 / 161 Computational Mechanics

 Flutter of a plate

Katz_13 / 162 Computational Mechanics

 A Bridge with a great height (Millau)

Wind tunnel tests have been performed with a

very low turbulence and an intensity of 9-10 %.
For a terrain roughness z0 of 0.30 m a turbulence
intensity of 9.4 % is obtained in 200 m height.
CFD Analysis with constant inflow conditions

Katz_13 / 163 Computational Mechanics

 Integral Length Scale

Length scale for CFD would be 81.0 [m] in 200 m

height.
For these height Eurocode defines a longitudinal
measure of 300 [m], DIN of 260 [m], Counihan ca.
250 [m] as well. With 20 % for the vertical length
scale this yields a value between 50 and 60 m.
Vortices of this size will be disturbed by a section
with a height of h = 4.50 [m]!

Katz_13 / 164 Computational Mechanics

 Section of Millau Bridge

Sections in steel and in concrete

With or without wind shields
Wind tunnel tests for a section with intradosed
fixing of cables, but not the cables itself
Wind shields as perforated plates with 50 %
permeability
Katz_13 / 165 Computational Mechanics
 Flow field at wind side

Katz_13 / 166 Computational Mechanics

 Pressure distribution

Katz_13 / 167 Computational Mechanics

 Angle of attack =
rotating the section

Katz_13 / 168 Computational Mechanics

 Forces on bridge

Katz_13 / 169 Computational Mechanics

Katz_13 / 170 Computational Mechanics
Katz_13 / 171 Computational Mechanics
 Remarks

Results, especially the slope of the curves is

obtained with good quality
The quantity fits within the range of the wind
tunnel experiments, however the drag and lift
coefficients are less than those of the tests
There are deviations between Test and Model
Reynolds number
geometric scale: 1:75, wind speed 1:2
Surface modelling
Modelling of wind shields

Katz_13 / 172 Computational Mechanics

 Influence of turbulence (high / low)

Katz_13 / 173 Computational Mechanics

 Drag coefficients

Katz_13 / 174 Computational Mechanics

 Revised drag coefficients

Katz_13 / 175 Computational Mechanics

 The flow for a reasonable large length scale

Katz_13 / 176 Computational Mechanics

 Vortex particle method

Katz_13 / 177 Computational Mechanics

 Great Belt East Bridge

Very simple geometry

Wind tunnel tests available
CFD-Analysis available in literature
V = 40 m/sec, Re >1 E7

Katz_13 / 178 Computational Mechanics

 Select the mesh size !

Shear velocity u approx. 1.0 m/sec

(boundary layer theory for plate)
=> y = 0.016 mm
but from adaptive run y =0.5 to 0.8 mm
So
either use a mesh size below y+ without the wall
function
Or use the wall function and select the minimum size of
the mesh to become between 40 and 300 y+
We have selected a mesh for the 2nd case with
the wall function

Katz_13 / 179 Computational Mechanics

 FE / FV - Mesh

Katz_13 / 180 Computational Mechanics

Katz_13 / 181 Computational Mechanics
Katz_13 / 182 Computational Mechanics
 There are no wakes !

Even with a
transient analysis
only small wakes
are obtained

Katz_13 / 183 Computational Mechanics

 Parameters to play with:

Mesh Size
Turbulence model
None
k-e standard
k-e RNG
k-e MMK
k-
Inflow
Turbulence Intensity
Dissipation rate
Boundary condition
Wall
Simple no slip

Katz_13 / 184 Computational Mechanics

 Results

Compressible flow increases the quality of pressures

The RNG and the MMK behave better with respect to the
pressure values than the Standard model
Draft and lift are still far away

Katz_13 / 185 Computational Mechanics

 There is something fishy

Comparison with an other CFD code yield same effect

There is a strong influence of the Reynolds number for
this section (see high pressure results by Schewe)
Published results in literature are rather doubtful
the parameters used are seldom described
But there are hints that the wall function is a problem
So we try to run the model with turbulence but without
wall function (which is extremely wrong for a CFD
expert)

Katz_13 / 186 Computational Mechanics

 Pressure distribution

Distribution is acceptable
The drag coefficient is now to high, but if we
discard the high shear stress we obtain cd = 0.64 !

Katz_13 / 187 Computational Mechanics

 Another Result with lower turbulences

Katz_13 / 188 Computational Mechanics

 Results (k, mt, p)

Katz_13 / 189 Computational Mechanics

 CFD + Wind Tunnel Tests (A. Sarkic)
-0.41

U =-6

pressure (+)
Measurement suction (+)
pressure (+)
U = 5.0 m/s Simulation suction (+)

Katz_13 / 190 Computational Mechanics

 A Note on the Mesh Size

A coarse mesh analysis:

Katz_13 / 191 Computational Mechanics

 Turbulent Viscosity and Lift Coefficient

Katz_13 / 192 Computational Mechanics

 Mesh size reduced by a factor of 5

Katz_13 / 193 Computational Mechanics

 Conclusion

It is possible to have results even with rather

coarse mesh
If we look deeper, we find quite a lot of problems
However using CFD in the design phase gives
valuable insight for many details
It can not been used for replacing the wind tunnel
If we know better about the discrepancies we
might reach the state of the racing car industry:
We know that the results are wrong by 40 percent, so we
add such a factor to our numerical results.

Katz_13 / 194 Computational Mechanics

 A Recent Project: Vortex & Galloping

January jj / 195 Wind Analysis

 Transient Drag and Lift in Free Field

Federsumme SPX Gruppe [kN] Federsumme SPY Gruppe [kN]

FORCES IN TIME
Federsumme SPX Gruppe 3

0.2
FORCES IN TIME
6.0
Federsumme SPY Gruppe 3
5.0
0.1
4.0

3.0
Zeit

2.0 Zeit 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 [sec]

5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 [sec]

-0.1

-0.2

-0.3

January jj / 196 Wind Analysis

 The influence of the ground

January jj / 197 Wind Analysis

Design: Marc Mimram, Paris
Wind Engineer: PSP-Technologie GmbH,
Aachen
Checks by: Leonhardt Andr &
Partner, Stuttgart

Katz_13 / 198 Computational Mechanics

 Characteristics

Spans of main bridge:

43.72 + 183.37 + 43.72
Thickness of pathways t = 15 cm

Katz_13 / 199 Computational Mechanics

 Eigenfrequencies (undamped)

Katz_13 / 200 Computational Mechanics

 Dynamic Sensitivity

Sharp edges of section

Interference
Small torsional stiffness
Negative aero elastic
Damping
Torsional Galloping
Pedestrians
(1 Hertz horizontal!)
Tuned Mass Dampers
are mandatory !

Katz_13 / 201 Computational Mechanics

 Numerical Simulation

Wind climate (the global wind)

Topology (local wind)
10 - min wind (Mean value + Gust + Spectra)
Aerodynamics (Drag coefficients)
Wind tunnel tests / CFD
Dynamics (Response of the structure)
Non linear transient time history analysis
Aero-Elasticity (Damping, Galloping, Flutter)
Design of the structure
100 year life time

Katz_13 / 202 Computational Mechanics

 Wind climate (global + local)
(10 Minutes of 100 Year-Wind)

25 years of
measurements of
meteorological station
Lahr

v = 27.2 m/sec

Atmospheric wind from

long term
measurements and a
global wind map an a
surface roughness of
z = 0.01 m

v = 25.3 m/sec

Katz_13 / 203 Computational Mechanics

 Wind spectra + Time Histories

Spectra Spektrum der diskreten Turbulenzkomponenten in LF 910

in Frequenzabstnden von df = 0.001 Hz
f S lon 4 X lon in Hhen von hmin und hmax = 5 / 90 m

2lon 1 70.8 X 2
lon
5/6

X lon f lon L lon / V 10 min longitudinal

1.0

0.5
hmax
hmin
m/s

0.2

Artificial
0.1

Wind longitudinal

Histories
0.2 0.4 0.6 0.8 1.0 Hz

Katz_13 / 204 Computational Mechanics

 Analysis procedure for wind part I

For each wind direction one base wind case

Global atmospheric wind + topology
Subtask I
Create wind profile
Subtask II
Create wind spectra in all points of the structure
accounting for coherences
Subtask III
Create wind histories

Katz_13 / 205 Computational Mechanics

 Aero dynamics

Wind
tunnel
CFD
(Vortex
Particle
Method)

Katz_13 / 206 Computational Mechanics

 CFD + Measurements (Sarkic)

H*1 H*2

H*3 H*4

A*1 A*2

A*3 A*4

Ured Ured
Force meas.; Pressure wr; Pressure wo; Simulation wr; Simulation wo
(wr-with, wo-without subtracting the reference measurement)

Katz_13 / 207 Computational Mechanics

 Drag Coefficient Variants

Values found in Literature for

Circular sections, Simple Bluff bodies
Roofs and walls
Rolled steel shapes
User defined factors for element groups
User defined local variants of factors similar
to load patterns (e.g. suction at edges)

Katz_13 / 208 Computational Mechanics

 Analysis procedure for wind part II

Deformations and velocities at

time ti Windhistory

Extrapolated deformations and Wind speed at time

velocities at time ti+1 ti+1

Non linear loading at time ti+1

Implicit solution of equilibrium at time ti+1

Katz_13 / 209 Computational Mechanics

 Instability
(constant wind)

Katz_13 / 210 Computational Mechanics

 Effectiveness of TMD

Katz_13 / 211 Computational Mechanics

 Response of deformations

Katz_13 / 212 Computational Mechanics

 Tingkau Bridge, Hongkong

Checking Analysis by
Schlaich, Bergermann and Partner

Katz_13 / 213 Computational Mechanics

 Cross Section

Katz_13 / 214 Computational Mechanics

 Wind profile
Hhenprofil Windgeschwindigkeit, Turbulenz und effektive Wellenlngen LF 8700

V-BOE
LVER

LLAT

TVER

LLON
TLAT

TLON
m

V
300

250

200

150

100

50

0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170m/s V, V-BOE

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 m/s TLON,TLAT,TVER

0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600 1700 m LLON,LLAT,LVER

Wind profile given as 1 h mean values + gust speeds =>

Conversion.
Katz_13 / 215 Computational Mechanics
 Safety concept

Safety factors provided by client:

gf = 1.9 for mean values
gf = 1.4 for wide band wind
gf = 1.2 for narrow band wind
It is better to increase the wind speed by g
than to multiply the loads (aerodynamic damping)
Turbulence (vgust vmean) receives its own factor
from

qult = 1.9 * qmean + 1.4 * (qgust - qmean )

Narrow band wind is accounted for by enlarging

the damping.

Katz_13 / 216 Computational Mechanics

 Advantages of
numerical solution
Non linear transient analysis copes for
Changes of the wind flow
Aerodynamic effects (Damping)
Aerodynamic effects (Flutter)
Aerodynamic effects (Torsional galloping)
Nonlinear material (Hysteresis = Damping)
Tuned mass dampers
Variants of tuned mass dampers
Where to install
Effectiveness, Malfunction
10 to 20 % Savings in forces within structures

Katz_13 / 217 Computational Mechanics

 Errors with the simplified approach

Every section moves individually in the wind

Transient movements of the air are not included.
Especially the time needed to establish a different flow field
is not accounted for. Phase differences between
movements of the bridge and the wind forces are not
included.
A complete FSI might
solve all these problems
but the numerical effort
is much to high.

Katz_13 / 218 Computational Mechanics

 Work Flow I Wind tunnel

Identify sensitive parts

Measurements of drag forces of a prismatic bar
which is moved periodically in the wind tunnel
This gives transient wind forces and the scanlan-
derivativa derived from those.
Wind forces are dependant on the angle of attack,
different series of measurements are required.
Reactions depend on the amplitudes!

Katz_13 / 219 Computational Mechanics

 Workflow II - Analysis

The history of the displacements h(t) and (t) will

be approximated by a harmonic function.
This identifies a current matching frequency,
amplitude and angle of attack.
The wind forces are evaluated based on the
Scanlan H*1 H*4 resp A*1 A*4.
There are also solutions with 18 instead of 8
derivativa to account for the movement in the
wind direction.

Katz_13 / 220 Computational Mechanics

 Other possibilities

Scanlan Derivativa are semi empiric

No unique measurement / analysis ?
How to judge the curves to be correct ?
We need a frequency
Alternative Indicial Functions ?

1 2 2 CM
t

M (t ) u 2 B X X M (t ) d
2 t 0

Katz_13 / 221 Computational Mechanics

 Example Brandangersund-Bridge

Katz_13 / 222 Computational Mechanics

 Parametre of Brandangersund-Bridge

Arch
Span width 240 m
Height 30 m
Section
width only 7.6 m
Wind parameter:
an art of its own
Stability limits

Katz_13 / 223 Computational Mechanics

 Eigenfrequencies

Katz_13 / 224 Computational Mechanics

 Eigenfrequencies

Katz_13 / 225 Computational Mechanics

 Stability check

Katz_13 / 226 Computational Mechanics

 Analysis

Katz_13 / 227 Computational Mechanics

 Assessment

This bridge established extreme requests (3rd check)

We are very close to the stability limits
Forces caused by wind load where higher than for the stable
condition, a secondary effect from the vicinity to the stability
limits
Development of torsional deformations more than proportional
Classical analysis methods could not be applied, a nonlinear
analysis including secondary effects was necessary
Horizontal bending moments from gusts exceeded the value of the
mean wind by a factor of 4. The most important part was the
secondary effect of the torsional divergence (ca 20 %) while the
effect of the derivativa was only 6 to 8 % in that case.

Katz_13 / 228 Computational Mechanics