You are on page 1of 9

Special Issue Article

Proc IMechE Part F:


J Rail and Rapid Transit
Computational fluid dynamics simulation 2015, Vol. 229(6) 635–643
! IMechE 2014

of rail vehicles in crosswind: Application Reprints and permissions:


sagepub.co.uk/journalsPermissions.nav

in norms and standards DOI: 10.1177/0954409714551013


pif.sagepub.com

Mikael Sima1, Sandor Eichinger2, Ander Blanco3 and Irfan Ali4

Abstract
The application of computational fluid dynamics (CFD) to the determination of aerodynamic coefficients for crosswind
stability in the context of vehicle assessment has been studied as part of the AeroTRAIN project. The work consisted in
establishing best practice guidelines for the use of standard Reynolds-averaged Navier–Stokes (RANS) approaches using a
streamlined and less-streamlined vehicle, project partners applying different computational codes and turbulence models
to a common vehicle, and then application to further vehicles in order to cover a range of different vehicles and yaw
angles. The simulations were complimented with wind tunnel measurements to allow the accuracy of standard RANS
approaches to be judged for various vehicle shapes and yaw angles. This paper summarises the overall results and the
recommendations made for the use of CFD in vehicle assessment of crosswind stability in relation to the EN 14067-6:
2010 standard. The main aspects of the guidelines are reported in a separate paper in this special issue. The considered
standard allows the use of CFD for vehicle speeds up to a maximum of 200 km/h whereas the HS RST TSI (2008) only
allows aerodynamic coefficients to be determined using wind tunnel measurements. The obtained results show that a
well-performed RANS CFD can predict the aerodynamic coefficient of streamlined trains with a relatively high accuracy.
The challenges increase for blunter-shaped trains and may be further influenced by equipment installed on the roof of a
train. Combined with the developed simulation guidelines it is considered that CFD can be used as an alternative to wind
tunnel tests in all cases provided that the accuracy of the approach is validated on a benchmark train with similar features
to those of the simulated train.

Keywords
Rail vehicles, crosswind stability, aerodynamics, CFD, standards and norms, TSI, RANS

Date received: 30 April 2014; accepted: 14 August 2014

comparisons are limited in terms of train shape


Introduction (most studies are of streamlined trains) and yaw
The determination of the aerodynamic characteristics angle range. In some cases uncertainties remain
of a train for use in the assessment of its level of about the actual conditions of the test. A comprehen-
crosswind stability currently depends on static tests sive study of the accuracy that can be achieved with
performed in wind tunnels. Although numerical simu- standard RANS simulations for a range of vehicle
lations are now an integral part in the design of trains shapes and yaw angles has been performed in the
they find more restricted use in vehicle authorisation.
Numerical simulations using Reynolds-averaged 1
Centre of Competence in Aero and Thermodynamics, Bombardier
Navier–Stokes (RANS) approaches are allowed in
Transportation, Sweden
the EN 14067-6:2010 standard for train speeds up 2
qpunkt GmbH, Stuttgart, Germany
to a maximum of 200 km/h, but not at all in the 3
Technology Management, Aerodynamics, CAF, Spain
HS RST TSI (2008).1,2 The use of numerical simula- 4
Rail Systems, Mechanical Systems Engineering, Aerodynamics, Siemens
tions to derive aerodynamic coefficients has been AG, Germany
explored in several studies in recent decades, see, for
Corresponding author:
example, the reviews of Diedrichs.3–5 It has been Mikael Sima, Centre of Competence in Aero and Thermodynamics,
shown that train coefficients can be predicted with a Bombardier Transportation, SE-721 73 Västerås, Sweden.
reasonable and acceptable accuracy. However, the Email: mikael.sima@se.transport.bombardier.com

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016


636 Proc IMechE Part F: J Rail and Rapid Transit 229(6)

EU FP7 research project AeroTRAIN.6 Based on for vehicle authorisation. The definition of the aero-
these findings, proposals have been made to revise dynamic coefficients follows EN 14067-6:2010.
the current European Norm (EN) and Technical
Specification for Interoperability (TSI) with respect
Simulation of the ICE 3 reference train
to the determination of vehicle aerodynamic co-
efficients using numerical simulations as part of The geometry of an ICE 3 end car given in EN 14067-
the assessment of crosswind stability for vehicle 6:2010 was simulated for the standard single track
authorisation. ballast with rails (STBR) configuration. As the details
As part of the project, the aerodynamic forces of of the wind tunnel used to obtain the reference aero-
several vehicles were measured using the Jules Verne dynamic coefficients were not available, the vehicle
Climatic Wind Tunnel in the Scientific and Technical was simulated in a generalised domain as shown in
Centre for Building (CSTB), Nantes, France. Figure 1 with distances given in Table 1, using a
Additional data in the form of pressure taps and block inlet flow profile. The Reynolds number was
flow field measurements were collected on a stream- 550,000, the model scale 1:25, inlet air speed 70 m/s
lined train and a blunter train (Regional Train) for the and a turbulent intensity of 1% was used with a
purpose of the validation of computational fluid length scale corresponding to 1 m in full-scale.
dynamics (CFD) simulations. Extensive studies on The reproducibility requirements for CFD in
the two types of trains were performed at the EN14067-6: 2010 are similar those for using a wind
Technical University (TU) Berlin to derive best prac- tunnel.1 There are three reference vehicles, an ICE 3
tice RANS simulation guidelines, which are sum- end car, a TGV Duplex power car and an ETR 500
marised in the accompanying paper.7 As part of the power car, for which benchmark aerodynamic
work a suitable setup to simulate the CSTB wind
tunnel was derived. The draft guidelines were applied
by the participating partners (TU Berlin, Bombardier Table 1. ICE 3 generalised domain dimensions following
Transportation, CAF, Siemens) in the CFD crosswind Figure 1. Non-dimensional distances are obtained by dividing by
project work task on the ICE 3 end car, one of the the reference length 3 m at full scale.
reference trains considered in EN 14067–6:2010. This
served as a practical application of the guidelines as Distance
well as a comparison between partners’ codes and Name Non-dim 1:25 (mm) 1:1 (m)
approaches. After this, all partners except TU Berlin
applied the guidelines to other vehicles, thereby L 70.0 8400 210
increasing the range of studied shapes. Simulating W 58.3 7000 175
the vehicle in the wind tunnel allowed the accuracy H 23.3 2800 70.0
of the simulation to be quantified against the measured Ltr,u 18.9 2263 56.6
forces and moments. A summary of these studies are Wtr,u 22.8 2741 68.5
reported together with proposals for the EN and TSI LSTBR,u 11.2 1343 33.6
in relation to the use of CFD to determine aero- LSTBR 7.7 920 23.0
dynamic coefficients for crosswind vehicle assessment

Figure 1. ICE 3, generalised domain.

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016


Sima et al. 637

coefficients are known. The requirement to use a wind aerodynamic coefficients are listed in Table 2. It is
tunnel or CFD is that for one of the reference vehicles clear that the benchmark requirements for CFD in
and for yaw angles relevant for the characteristic wind EN 14067-6:2010 are fulfilled by applying the simula-
curve (CWC) calculation tion guidelines. The results show that individual coef-
  ficients can vary relative to benchmark values more
CMx,lee,test  CMx,lee,bmk than CMx,lee. The most important coefficients are the
max 5 "max
ð1Þ
C Mx,lee,bmk side (Cy) and lift (Cz) forces and the rolling moment
CMx, whereas the drag force (Cx) makes no contribu-
 
CMx,lee,test  CMx,lee,bmk tion to an overturning event. However, that individual
mean 5 "mean
ð2Þ components have higher variations than CMx,lee is
C Mx,lee,bmk
also the case for the variation between wind tunnels,
CMx,lee is the lee rail roll coefficient, where test as is exemplified for a non-specified vehicle model
denotes the wind tunnel measurements or CFD simu- tested in two different EN compliant wind tunnels in
lations of the reference vehicle; bmk denotes the bench- Table 3.
mark values for the same vehicle provided in the norm. Some small revisions were made to the guidelines
The maximum and mean deviation allowed are resulting from this work. The study confirmed that
"max ¼ 0.15 and "mean ¼ 0.1. As part of the EN14067- similar results can be attained with an alternative
6:2010 requirements, the train geometry, blockage code (Fluent) and other turbulence models. The
ratio, Reynolds number and wind profile were taken ccmþ quadratic k-" and Fluent realizable k-"
to be the same as those for the wind tunnel. approaches produced positive results. The application
The ICE 3 end car simulation results obtained at showed that the standard k-" model should not be
TU Berlin using the CD Adapco code ccmþ used as it is unable to fully resolve the flow field in
(v5.06.007) with polyhedral cells and the k-! SST tur- the lee side, particularly the lee-side vortex, whereas
bulence model are shown in Figure 2, and the differ- by chance and error cancellation CMx,lee might match
ences between the obtained results and the EN for some yaw angle. A comparison was made for 30
and 50 yaw angles. At 50 there was more variation
than at 30 , suggesting a more challenging flow field.
At 30 there was a clear lee-side vortex created by the

Table 3. Variation (in percent) of aerodynamic coefficients for


one car (same model) between two different EN compliant
wind tunnels.

Yaw
angle
(deg) CFx CFy CFz CMx CMy CMz CMx,lee

0 0 56 128 229 45 139 599


10 10 0 18 6 80 7 8
20 10 3 6 0 31 1 1
30 46 15 7 1 32 45 2
40 161 22 10 10 8 60 6
Figure 2. TU Berlin simulated aerodynamic coefficients for 50 3191 21 18 7 10 68 3
the ICE 3 end car together with EN 14067-6 wind tunnel 60 174 17 14 15 82 104 11
coefficients. This figure is available in colour format online.

Table 2. Difference (in percent) between the TU Berlin ICE 3 simulation results and the EN 14067-6:2010 values.

Yaw angle
(deg) CFx_EN CFy_EN CFz_EN CMx_EN CMy_EN CMz_EN CMx,lee_EN

10 9.0 10.1 6.7 9.5 220.0 5.3 8.8


20 45.1 2.5 21.0 3.0 5.8 6.5 8.7
30 163.7 1.0 18.3 0.8 49.3 6.9 6.4
40 110.9 1.0 14.1 2.3 40.6 8.9 5.8
50 1.5 1.1 20.6 1.4 13.0 8.1 6.6
60 13.0 0.7 20.4 0.7 19.8 8.4 4.3

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016


638 Proc IMechE Part F: J Rail and Rapid Transit 229(6)

flow over the front and roof, whereas at 50 the lee- streamlined (IC4) and Regional Train (VT 612).
side vortex moved further away from the train and the A separate simulation domain for each yaw angle
flow has started transitioning into what at 90 is more and measured vehicle was required, corresponding
like two-dimensional flow around a cylinder. to the situation in the wind tunnel. There is a small
yaw angular dependence of the flow in the wind
Accuracy in the prediction of CMx,lee tunnel that was corrected in the measurements by
the use of an effective yaw angle. In the simulations
for different trains and yaw angles
there was no yaw component in the incoming flow.
Simulations of various trains, shown in Figure 3, were The effect of the supports was found to influence the
made to assess the accuracy of the standard RANS flow above the splitter plate. As the number of sup-
approach for a range of vehicle shapes and yaw ports in the wind tunnel was large it was not possible
angles, when employing the derived simulation guide- to model them individually; therefore, the effect on
lines.7 The trains were all tested in the CSTB wind the flow was modelled using a porous media volume
tunnel as part of the AeroTRAIN project. The simu- under the splitter plate that provided a representative
lation domain represents the wind tunnel, Figure 4, as pressure drop. As the presence of a train influences the
derived by TU Berlin in detailed studies of the flow around the splitter plate, the pressure drop was

Figure 3. Simulated vehicles.

Figure 4. Simulated wind tunnel domain left, CSTB test section right.

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016


Sima et al. 639

adjusted to match the measured velocity near the roof simulation guidelines and using their tool and
immediately above the measured train. It was also approach previously validated on the ICE 3 train.
found essential to refine the mesh around the leading The IC4, VT 612 and AGC trains were simulated
edge of the splitter plate to resolve the mean flow and with CD-Adapco’s ccmþ code using polyhedral cells
to closely match the measured axial flow on the split- and the k-! SST turbulence model; the ADR train
ter plate (measured without the vehicle). The was simulated using Ansys Fluent, polyhedral cells
approach was chosen to minimise these influences; and realizable k-" turbulence model; the RevCo
however, it adds some uncertainty to the comparison train was simulated using Ansys Fluent, hexcore
of results. cells (Spider 2.0) and realizable k-" turbulence
Table 4 shows the accuracy in the prediction of model. Second-order upwind convective discretisation
CMx,lee for the simulated trains in a comparison with was used in all the considered cases. Figures 5 and 6
the wind tunnel measurements. The simulations were show plots of CMx,lee versus yaw angle for the ADR
performed by the various partners by following the and AGC train, respectively. The IC4 end car had
two variants, one with roof corrugations representing
the actual train and another version that did not have
the roof corrugations (smooth). As the IC4 and AGC
Table 4. Deviation (in percent) of CMx,lee (CFD-WT)/WT
are articulated trains the second cars were measured
(WT ¼wind tunnel) results relative to wind tunnel
and simulated. Also the RevCo has roof corrugations
measurements.
although different from those on the IC4 train.
Yaw angle (deg) Vari. WT The last two columns in Table 4 represent the
variability in the wind tunnel measurements meas-
Vehicle 10 20 30 40 50 60 Mean Max
urements including 7/8 Remax, Remax, the positive
IC4 - end corrug. 3 2 3 5 2 11 3 4 yaw angle sweep and negative yaw angle sweep
IC4 - end smooth 5 2 3 4 (up to 40 ), where Remax had a value of 600,000.
IC4 - second car 13 5 6 Some additional simulations are included that used
VT 612 5 6 10 9 14 24 4 6 a generalised domain with a block inlet flow profile
(GB). The generalised domain was similar to that used
AGC end car 6 3 21 12 3 7 4 6
for ICE 3 shown in Figure 1, but at the scale 1:15.
AGC second car 0 13 8 9 3 4 4 9
It had 70 m (full-scale) lateral and axial upstream
RevCo (corrug.) 9 5 3 6
lengths of no-slip ground, measured from the origin
ADR 4 4 8 6 3 3 for the first vehicle, which ensured that the boundary
With generalised domain and block inlet low profile layer developed along the no-slip ground was not
IC4 - end corrug. 11 6 5 4 21 higher than the STBR. In the wind tunnel domain
IC4 - second car 1 3 5 8 11 the train was rotated for each yaw angle, similar to
AGC end car 13 30 25 20 the real wind tunnel, and then re-meshed. The flow
inlet was always placed in the front boundary. With
AGC second car 19 3 5 2
the GB the same mesh could be used for all wind

Figure 5. CMx,lee of ADR normalised with wind tunnel value at 30 .

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016


640 Proc IMechE Part F: J Rail and Rapid Transit 229(6)

(a) (b)

Figure 6. CMx,lee of AGC normalised by the wind tunnel value at þ30 . This figure is available in colour format online.

angles and the inlet flow was applied with the correct the leeward rail roll moment. The extent varies with
yaw angle on the upwind boundaries. Furthermore, both vehicle shape and the yaw angle. Depending on
the aerodynamic coefficients of the first and second whether this is predicted to be lower or higher with
cars were evaluated from the same simulation when the wind tunnel setup, the GB prediction may get
that was applicable. The GB represents, in some closer or further away from the measured value.
sense, the ideal conditions that are the aim in the Looking closer at the vehicle shapes in Figure 3,
wind tunnels. the following points can be made regarding the lead-
The predictions are close to the measured values ing vehicles.
for the IC4 end car, more or less within the variability
in the measurements excluding 60 . A similar accu- 1. The most accurate predictions are for streamlined
racy is observed without corrugations, even though vehicles or vehicles with sloping fronts. The
the corrugated end car CMx,lee is significantly higher RevCo has a lower slope in the upper part,
than without corrugations, on average 10% higher for above a bluff lower part.
yaw angles between 10 and 60 with the largest dif- 2. Correspondingly the more challenging to predict
ference being 16% at þ40 . This shows that the VT 612 and AGC are blunter trains.
streamlined vehicle is closely predicted and also that 3. Corrugations, although increasing CMx,lee, do not
the effect of a corrugated roof is captured. Following significantly impact the accuracy of the prediction.
the guidelines and good practice the corrugation However, resolving the corrugation feature
required additional mesh refinement. The ADR is requires a higher grid resolution.
also well predicted as is the RevCo (although results 4. ADR and RevCo have underfloor boxes. Whether
were only obtained for two yaw angles). The predic- or not the underfloor is covered does not particu-
tion is less accurate for VT 612, suggesting a more larly seem to affect the accuracy of the prediction.
complex flow field. Least accurate is the prediction 5. VT 612 has a noticeable roof box that has an influ-
for the AGC, particularly at 30 yaw angle. The ence on the flow field over the roof.
AGC is the only train with a roof fairing. For both 6. AGC has roof fairings that make the flow different
AGC and VT 612 CMx,lee is primarily predicted to be to that of other trains. The fairing fixes the separ-
higher than the measured values. The second cars are ation point to the windward side with the flow
potentially more difficult to predict, as in principle it going over (30 yaw angle and more) or into the
relies on a correctly predicted flow around the roof recess (20 yaw angle and less).
upstream vehicle, which forms the upstream flow
condition for the second vehicle. For the articulated By studying the flow around the trains it is possibly
trains studied here, the forces on the second car are to isolate some features of the flow. For more stream-
lower than on the first car, and are therefore contri- lined trains, e.g. the IC4, ADR and ICE 3, the char-
buting less to the overall overturning prediction. The acteristic lee-side vortex is initiated by the flow
variability of the wind tunnel measurements are passing over the front of the train. For blunt trains,
mostly higher for vehicles that are also more challen- there is a separation bubble at the lee side of the front
ging for the CFD, and are comparatively low for the from flow passing around it, and the lee-side vortex is
streamlined vehicles. generated by the flow over the roof. This separation
Using a GB rather than a CSTB wind tunnel region for blunter trains that transitions to a lee-side
domain has the effect of increasing the coefficient of vortex further downstream results in the flow and

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016


Sima et al. 641

Figure 7. Plots of velocity magnitude and in plane vectors close after front for: (a) ICE3 (5.9 m from front tip without buffers;
max scale ¼ 1.43 Vin); (b) IC4 (5.5 m; 1.76 Vin); (c) ADR (4.8 m; 2.20 Vin); (d) AGC (4.3 m; 1.42 Vin); (e) VT612 (4.0 m; 1.54 Vin); and
(f) RevCo (4.5 m; 1.28 Vin) at 30 yaw angle. Varying scales but in all cases blue is low speed and high speed is red, purple and/or white.
This figure is available in colour format online.

forces being more difficult to predict. RevCo has a significant impact on the flow features, in contrast
blunt shape for the lower half of its front and a with the VT 612 where the box is protruding more
lower slope at the upper half that makes it more like (250 mm). It is difficult to say if the higher roof box
a streamlined vehicle; therefore, the flow has simila- makes the flow more challenging to predict, but it
rities with both blunt and streamlined trains. Figure 7 does influence the flow and adds flow features. It is
shows the flow through a section just after the front reasonable that if you can accurately predict the
for all vehicles at a 30 yaw angle. forces of a vehicle with roof boxes, you will also be
Roof boxes can disrupt the flow over the roof and able to accurately predict a similar vehicle without
the lee-side vortex if they are large enough. It is also roof boxes. A roof fairing generates a particular
likely that there is an influence of streamlining the box flow that is more difficult to predict with RANS and
edges. There are low protrusions on the roof of ADR possibly a more advanced modelling technique is
(57 mm) and IC4 (85 mm), but these do not have a required. As the flow is different from other trains,

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016


642 Proc IMechE Part F: J Rail and Rapid Transit 229(6)

it is difficult to say that if a train with a roof fairing is Table 5. Classification of benchmark trains.
closely predicted then the prediction for trains with
Roof
other features will have the same or better accuracy.
In general, it can be concluded that the accuracy of No or low Roof Roof
the prediction is higher for streamlined trains than for Front protrusions box(es) fairing
blunt trains. Non-smooth roof arrangements compli-
Blunt 1-A 1-B 1-C
cate the flow and are likely to reduce the accuracy of
the prediction. Therefore, an accurate prediction for Streamlined 2-A 2-B 2-C
one of the EN14067-6:2010 benchmark vehicles,
which are all streamlined and do not have protrusions
on the roof, does not mean that a similar prediction
accuracy will be achieved for all vehicles. Further vehicle is that it has similar flow features (similar
appropriate benchmark vehicles are required to shape) as the new vehicle that is to be quantified.
ensure the accuracy of the prediction. Benchmark vehicles are classified into six different
The current definition of accuracy (equations (1) groups as shown in Table 5. There are two main
and (2)) is probably suitable for application to other classes (relating to the front shape of the leading
benchmark trains. However, it can be seen in the vari- vehicle):
ability of the measured values of CMx,lee that it is
higher for AGC and VT 612, and also for the . 1: blunt;
second car of IC4 and AGC. This suggests that the . 2: streamlined.
variation in measurements for different wind tunnels
is also higher for the trains that are more challenging
for CFD than for the current EN14067-6: 2010 bench- Within these classes there are three different cate-
mark trains. Further checks on blunt trains with non- gories that depend on the roof geometry or layout:
smooth roof arrangements in different wind tunnels
could give more insight to the inherent sensitivity and . A: no or low protrusions (<100 mm);
whether it could be argued that less strict require- . B: roof box or boxes;
ments are required on the prediction accuracy for . C: roof fairing.
these trains.
As far as yaw angles are concerned, the most Low protrusions extend less than 100 mm around
challenging seems to be within the most relevant for the surrounding roof. All roof boxes are considered to
determining CWCs, i.e. 15–50 . It appears that the be equal such that any roof configuration that is not
accuracy is less for 60 , but on the other hand it is category A or C is category B.
only in few cases, if any, that 60 is relevant for the There is a need to define streamlined and blunt
calculation of the CWCs. vehicles. Streamlined vehicles have a lee-side vortex
starting at the front at a 30 yaw angle. This should
be verified, e.g. by plotting vectors of in-plane vel-
Recommendations for standards
ocity for cross-sections normal to the train direction,
The results for the AeroTRAIN investigations sum- at several positions starting ahead of the front and
marised in the last sections lead to recommendations continuing to the middle of the first car. No signifi-
for amendments to the CEN/TSI methodology for cant separation zone with low speeds should be vis-
the use of CFD to quantify the aerodynamic coeffi- ible in the velocity magnitude or axial velocity.
cients relating to crosswind stability.8 The main Examples are ICE 3, IC4, ADR and RevCo. Even
approach in EN 14067-6: 2010 is maintained, how- though the RevCo front shows blunt characteristics
ever, further benchmarks for the validation of the in the lower half and streamlined characteristics in
method are recommended to reflect the fact that the upper part, the CFD results suggest that the
vehicles with blunt or streamlined front shape as dominant feature is the shape of the upper half of
well as geometrical features in the roof region the front as regards slope and curvature. Examples
create different challenges for the flow prediction. of blunt vehicles are VT 612 and AGC. If in doubt,
This is complemented with guideline recommenda- the shape of the benchmark vehicle and the new
tions for the current industry standard RANS simu- vehicle should have similar-shaped fronts and roof
lation methodology to minimise the variability in areas. The simulation flow fields should have similar
predictions by different users. In addition, it is rec- features. Furthermore, to be conservative, the bench-
ommended that CFD can be used as an alternative mark vehicle, if in doubt, should be considered to be
to wind tunnel testing in all cases, given the fulfil- streamlined, whereas the new vehicle, if in doubt,
ment of the accuracy requirements. should be considered blunt.
The following is a recommendation for classifying If the simulation approach is validated for a blunt
benchmark vehicles taking the simulation results into vehicle it is also valid for a streamlined vehicle of the
account. A general requirement of a benchmark same roof category. Similarly, if the simulation

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016


Sima et al. 643

approach is proved for a vehicle with roof category B, Acceptance for the Interoperable Network (www.trio-
it is also valid for roof category A. This means that if train.eu). AeroTRAIN has been a collaborative project
the simulation approach is validated for: medium-scale focused research project supported by the
European 7th Framework Programme co-ordinated by
UNIFE, contract number: 233985.
. 1-B, it is also valid for 1-A, 2-B and 2-A;
. 2-B, it is also valid for 2-A;
References
. 1-C is also valid for 2-C;
. 1-A is also valid for 2-A. 1. EN14067-6: 2010. Railway applications - aerodynamics -
part 6: requirements and test procedures for cross wind
assesment.
For assessing a new vehicle a generalised domain
2. EU TS1 2008/232/EC: 2008. Technical specification for
with block inlet flow profile should be used that interoperability relating to the ‘rolling stock’ sub-system
includes more detailed requirements on the domain. of the Trans-European High-Speed Rail System.
For validating the approach using a benchmark vehi- 3. Diedrichs B. Computational methods for crosswind stabil-
cle, a representation of the wind tunnel geometry and ity of railway trains, a literature survey. Report TRITA
conditions when the measurements were made is suit- AVE 2005: 27, 2005. Stockholm, Sweden, Royal Institute
able to minimise setup influences on the comparison. of Technology.
However, since it may be difficult to accurately repre- 4. Diedrichs B. Aerodynamic calculations of crosswind sta-
sent the wind tunnel and the aim of the wind tunnel is bility of a high-speed train using control volumes of arbi-
to mimic block profile conditions, a generalised trary polyhedral shape. In: Zasso A, Belloli M, Cheli F,
domain with a block profile may be used. et al. (eds) The BBAA VI international colloquium on bluff
bodies aerodynamics & applications, Milano, Italy, 20–24
Simulations replace wind tunnel tests; as such the gen-
July 2008, pp.1–16.
eral parameters as for example Reynolds number 5. Dipartimento di Meccanica – CIRIVE, Politechnico
should be chosen to be representative of the wind Milano; Possibly editors (a cura di): Alberto Zasso,
tunnel test conditions. It is also considered that very Marco Belloli, Federico Cheli, Giorgio Diana, Sara
accurate numerical simulations could provide valid- Muggiasca, Daniele Rocchi; 1–16. ISBN 88-901916-4-3.
ation data, i.e. provide numerical benchmarks. 6. Diedrichs B. Aerodynamic crosswind stability of a regio-
nal train model. Proc IMechE, Part F: J Rail Rapid
Transit 2010; 224: 580–591.
Conclusions 7. Sima M, Grappein E, Weise M, et al. Presentation of
A structured procedure has been employed to assess the EU FP7 AeroTRAIN project and first results. In:
the accuracy that can be achieved with standard The ninth world congress on rail research, Lille, France,
22–26 May 2011.
RANS simulation approaches to assess aerodynamic
8. Eichinger S, Sima M and Thiele F. Numerical simulation
coefficients relating to crosswind stability for a range of regional train in cross-wind. Proc IMechE, Part F:
of vehicle shapes and yaw angles. It was made pos- J Rail Rapid Transit 2015; 229(6): 625–634.
sible by many partners considering a large range of 9. Sima M, Eichinger S, Blanco A, et al. WP3 – T3.3 CFD
vehicles within a project combining both wind tunnel simulations output document. AeroTRAIN project publi-
tests and simulations, which has not previously been cation D6.3(1), A.3.3, 2012.
done to this extent for crosswind studies of rail vehi-
cles. The work resulted in the proposal of best practice
simulation guidelines to minimise the variability in
predictions by different users and recommendations Appendix 1
for amendments to the CEN/TSI methodology
relating to the use of CFD. It is recommended
Notation
to allow simulations instead of wind tunnel tests CMx rolling moment coefficient (around
in all cases the accuracy requirement is fulfilled. On longitudinal axis between rails)
the other hand, the benchmark assessment of accur- CMx,lee rolling moment coefficient around lee
acy is extended to reflect the finding that different rail
vehicle shapes provide different flow features and Cx drag force coefficient
challenges. Cy side force coefficient
Cz lift force coefficient
Funding Re Reynolds number
This paper describes work undertaken in the context of the Vin inlet velocity
AeroTRAIN project, Aerodynamics: Total Regulatory

Downloaded from pif.sagepub.com at PENNSYLVANIA STATE UNIV on May 11, 2016

You might also like