You are on page 1of 6

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014

Porto, Portugal, 30 June - 2 July 2014


A. Cunha, E. Caetano, P. Ribeiro, G. Mller (eds.)
ISSN: 2311-9020; ISBN: 978-972-752-165-4

Traininduced aerodynamic pressure and its effect on noise protection walls


P. Ampunant1 , F. Kemper2 , I. Mangerig1 , M. Feldmann2
of Steel Construction, UniBw Munich, WernerHeisenbergWeg 37, 85579 Neubiberg, Germany
2 Institute of Steel Construction, RWTH Aachen, MiesvanderRoheStreet 1, 52074 Aachen, Germany
email: Paschraphon.Ampunant@unibw.de, Kemper@stb.rwth-aachen.de,
Ingbert.Mangerig@unibw.de, Feldmann@stb.rwth-aachen.de

1 Institute

ABSTRACT: Traininduced aerodynamic loads on noise protection walls have been a field of research for some time. During
the passage of a highspeed train along noise protection walls, they are exposed to high aerodynamic pressure loads that reduce
the endurance limit. In this work measurements of the excerted pressure are shown and the results of a statistical analysis. The
measurement data are the foundation of a numerical simulation based on computational fluid dynamics (C FD) to simulate the
pressure flow around the passing train. Based on the NAVIERS TOKESequation which are formed to the R EYNOLDS averaged
NAVIERS TOKESequation (R ANSequation) the flow is resolved. The results of the simulation are compared to the measurement
data.
KEY WORDS: S EGES, noise protection walls, highspeed trains, computational fluid dynamics (C FD)
1

INTRODUCTION

Noise protection walls along railroads of high-speed trains are


exposed to an fluctuating aerodynamic pressure field when trains
are passing by. During this passage the flow compression at
the bow of the train induces an overpressure at the barrier
which is immediately followed by an aerodynamic suction
due to a significantly accelerated parallel flow field. These
quickly changing pressure amplitudes represent a dynamic load
impact which lead to dynamic structural movements. The same
pressure fluctuations in opposite arrangement are excited at the
rear of the train. The response of the structure is dependent
on the Eigenfrequencies and the damping behavior of the noise
protection wall. Due to the non-uniform distribution of loads
on the elements of the noise protection walls which is a present
field of research it is the aim to investigate the formation of the
pressure field and its effects on the structure.
2
2.1

Figure 1. Schematic set-up of measurement equipment


recorded, each with a sample duration of T=30 s. In Figure 2
the principal setup of the used sensors is illustrated for a section
of L=5.0 m. The total length of the noise barrier with glass
elements was LT =50.0 m

MEASUREMENTS ON SITE
Measurement set-up

Numerical simulations which aim to determine aerodynamic


pressures on complex surfaces should be validated in order
to ensure the model quality. In case of bluff bodies under
natural wind flow, mainly wind tunnel results are used for this
aim. However, the simulation of the aerodynamic pressure
field induced on a noise barrier by a high speed train is not
easy to investigate on a scaled model. For the validation of
numerical results shown in this paper, pressure data obtained
by means of on-site measurements are used. The corresponding
measurement campaign is described in the following.
During the research project S EGES measurements on noise
barriers made of glass have been performed in the summer of
2012. During a time period of six weeks, train induced pressure
fields and associated structural reactions have been measured. In
total, about 2,700 train events (different train types) have been

Figure 3. Velocity Histogram of all recorded Trains


For the analyzes in the scope of this paper, four differential
pressure transducers have been considered. All sensors were

3739

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014

installed at a height of h=0.0 m above the top of the tracks. The


sensors were located in a maximum longitudinal dimension of
LS =42.2 m between the first and the last sensor.

The results plotted in figure 5 can directly be used to validate


the numerical simulations which are further described in the
following sections.

2.2

Analysis of Pressure Values

Based on the time difference between the pressure signals at the


beginning and the end of the test wall, the individual speed of
each train passage has been analyzed:

vTrain =

s(Sensor1, Sensor2)
42.2m
=
(1)
t( p)

t( pSensor1 ) t( pSensor2 )

The pressure amplitudes have been normalized by the


velocity pressure of the passing train in order to determine the
pressure coefficients:

NUMERICAL SIMULATION

4.1

Theory

In this chapter the theoretical basis of the numerical simulation


is explained. Futher details can be found in e.g. F ERZIGER et
al. [1], L ECHELER [2] and S CHICHTLING [3]. The foundation
of the numerical simulation are the equations
~
~
U + ~F = Q
t

(4)

with

c p (t) =

p(t)
1/2 v2Train

~ = ~u
U
 ,
e + 21 ~u2

~u
,
~F = ~u ~u + p I

~u h + 12 ~u2 ~u T

0
~ = ~g
.
Q
~g ~u + qS

(2)

As the time scale of the measured amplitudes depends


on the train velocity, all time series have been converted to
displacement related series:
xTrain (t) = t vTrain

(3)

For a better comparability, the first pressure maximum has


been set to a xTrain =100 m. Some typical time series of the
pressure coefficient due to the passage of an ICE 3 train is
plotted in figure 4. In the plot, the significance of the fore
wave and the rear wave can be seen. The inner variation
of pressure is provoked by the wheel sets of each car and
by the gaps between the cars. Whereat the fore wave let
to comparable low fluctuations of the determined pressure
coefficients, the train parallel flow and the flow separation at
the rear part are accompanied by significant variations of the
pressure coefficients.
3

STATISTICAL ENSURED RESULTS

3.1

Statistical ensured results

In order to allow a statistical ensured result for the design


of noise barriers, it is important to take into account a larger
number of data records. As the track is used by different train
types with the analyzes have been focused on velocities above
vTrain = 150 km
h .
Altogether, a number of n=72 independent pressure time
series have been analyzed with respect to their statistical
distribution. In order to allow a more profound analysis, all
pressure series have been normalized and equally discretized.
Therewith, the mean value and the standard deviation could be
analyzed for each train position xTrain .
In figure 5 the results are plotted as mean values. Additionally
the 34% - 68% and the 5% - 95% quantile intervals have been
plotted in this figure. The associated statistic constants for
the double sided delimitation are u0.84 = 0.99 and u0.975 =
1.96. Hence, the pressure quantile intervals of the pressure
coefficients are determined as follows:
c p 0.99 c p c p + 0.99
c p 1.96 c p c p + 1.96 .

3740

Equations (4) are called the NAVIERS TOKESequations.


The mentioned equations are written in differential form and
include the conservation of mass, momentum and energy. In
order to approximate turbulence flow the NAVIERS TOKES
equations are formed to the R EYNOLDS-averaged NAVIER
S TOKESequations (R ANSequations).
The spacial discretization is based on the finitevolumemethod
so that the conservation equations can be solved numerically.
Furthermore the R ANSequations have to be discretisized
in time as timedependent factors play a role. The time
discretization is conducted with a 2ndorder implicit E ULER
method.
The NAVIERS TOKESequations are simplified as we assume
that the fluid is incompressible. Due to the fact that the
simulated train velocity does not exceed 300 km
h and the M ACH
number Ma
Ma =

vtrain
1 300
=

0, 245 0, 3
vsound
340 3, 6

is below 0.3 the fluid can be modelled as incompressible


(YOUNG et al. [4]) .
4.2

Geometry

A 3Dbased numerical simulation is implemented, the geometrical information of an I NTERCITY E XPRESS 3 (I CE 3) is taken
from D INE N 14067-6 [5] in the first step. In figure 6 the head
of the train is shown.
For the numerical simulation the geometry is simplified. The
wheels and the connections for electricity are neglected so that
the number of elements is reduced. As the surrounding air
field is of interest the fluid space around the train is modelled.
The fluid space is subdivided into 4 parts. The first part is

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014

Figure 2. Differential Pressure Sensor (left), Accelerometer (middle) and approaching ICE 3 Train (right)

Figure 4. Normalized series of differential pressure, 18 passages of ICE-3 trains

Figure 5. Mean Values of the Pressure Coefficient c p and Quantile-Intervals

3741

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014

with a constant velocity. The mesh is therefore not deformed


during the whole simulation. The two other dynamic fluid
regions bh and bv are located behind and in front of the region
bm . These movable regions contain hexahedral elements in
a structured way so that the movement of the train can be
compensated by stretching and squeezing of the two mentioned
fluid regions. During this process the angles of the mesh grid,
however, are not changed and keep the perpendicular manner
of the implemented, structured grid. The static fluid region s
which surrounds the noise protection wall is not deformable.
4.4

Figure 6. Geometry of I CE 3 [5]


the static fluid region s that includes the noise protection
wall. The second region contains the train surface and is called
dynamic fluid space bm . Two additional dynamic fluid regions
are modeled in order to implement the movement of the train
geometry. All parts of the fluid region are shown in figure 7

Boundary conditions

The outer boundaries of the fluid regions are modeled as noslip


walls. The size of the fluid space perpendicular to the moving
direction of the train is 20 times the size of the train to ensure
that the farfield boundaries do not have an influence on the
resulting pressure field. The surfaces of the noise protection wall
and the train are modeled as noslip walls and the behaviour of
all noslip walls are smooth. The surface in front of the head of
the train is defined as outlet whereas the suface behind the tail
of the train is modeled as inlet.
The movement at the boundaries is set to zero and the moving
velocity of the region bm is set to a constant velocity. The mesh
velocity of the regions between the constantly moving mesh and
the static boundaries move are linearized from zero to the train
velocity.
4.5

Solution and comparison

All simulation runs are conducted with an amount of


approximately 1.7 million elements. In order to ensure a
dimensionless wall distance y+ of
y+ 100

Figure 7. Geometry of fluid space


Due to the symmetric properties of the train geometry and
the low influence of the assymetric positioning of the noise
protection walls to the passing train, only half the size of the
flow field is implemented in order to reduce the amount of grid
elements.
4.3

Moving Mesh

The translational movement of the train can be modeled as


movement of the mesh grid. When conducting a moving
mesh the mesh deforms and therefore the quality of the mesh
deteriorates. In order to keep the quality of the mesh the same
during the whole simulation a structured mesh is necessarily
implemented in all parts of the fluid regions. The fluid region
bm which contains the geometry of the train moves translatory

3742

the smallest spatial discretization is 4x = 0.001 m. Due to the


movement of the train body the simulation is transient, the time
discretization 4t does not exceed 0.001 s. The approximation
of the turbulence layer was realized by implementing the S ST
kmodell by M ENTER [6].
When the train passes the noise protection wall at a certain
point the bow of the train induces an overpressure that is
immediately followed by an aerodynamic suction. In figure 8
the pressuretimecurve of a passing train is shown at rail level
for a train velocity of vtrain = 160 km
h . The measured pressure
and the numerical results are both shown. By comparing the
two curves a good matching between the two data samples can
be seen especially at the bow of the train. As the rear part
of the train passes the monitoring point at the wall a higher
pressure is calculated than measured. Although the pressure
difference is more than 100 % there is a qualitative similarity
the pressure curve. The difference can be explained by the
significant variations of the pressure measured at the rear part
as documented in section 2.2.
The fluctuating pressure can be simulated and are similar to the
measurements. The difference of the highest values of pressure
between measurement and numerical results is underneath 5 %
for the pressure when the bow of the trains passes by.
In figure 9 the pressure at crosssection and the pressure in
longitudinal direction are presented. The left picture shows a

Proceedings of the 9th International Conference on Structural Dynamics, EURODYN 2014

REFERENCES

Figure 8. Comparison between measurement and numerical


results [7]

[1] Joel H. Ferziger and Milovan Peric,


Numerische
Stromungsmechanik, Springer, 2002.
[2] Stefan Lecheler,
Numerische Stromungsberechnung,
Vieweg+Teubner, 2009.
[3] Hermann Schichtling and Klaus Gersten, GrenzschichtTheorie, Springer, 2006.
[4] Donald F. Young, Bruce R. Munson, Theodore H. Okiishi,
and Wade W. Huebsch, A Brief Introduction to Fluid
Mechanics, Wiley & Sons, 2007.
[5] DINEN 140676 Railway applications Aerodynamics
Part 6: Requirements and test procedures for cross wind
assessment, May 2010.
[6] Florian R. Menter, Zonal Two Equation k turbulence
models for aerodynamics flows, AIAA Paper 93-2906,
1993.
[7] Projekt S EGES SchallschutzElemente aus Glas an
EisenbahnStrecken, Final report, 2012.

nonlinear distribution at the crosssection, the picture on the


right-hand side presents a drag area in front of the bow which
is followed by a suction area.

Figure 9.
pressure at cross-section (left), pressure in
longitudinal direction (right)

SUMMARY

This work shows the results of measurements when highspeed trains go past noise protection barriers.
These
measurements were made during the research project S EGES.
The measurement data were analyzed and the results were
described. The fluctuating pressure which can be described as
timedependent and a quick change of drag and suction when
the bow of the train passes with high velocity were measured
and a numerical simulation based on C FD was implemented to
reproduce the pressure field which act on the noise protection
walls. The theoretical background and the setting of the
numerical simulation was explained. A comparison between the
measurements and the results of the numerical simulation was
made. It was shown that there is a good matching between the
measured data sample and the numerical results. The difference
between the pressure measured and the calculated numerical
results are underneath 5 % when the bow of the train passes the
noise protection wall. The pressure field can be well reproduced.

3743

You might also like