Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

ARTICLE IN PRESS

J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020

Contents lists available at ScienceDirect

Journal of Wind Engineering


and Industrial Aerodynamics
journal homepage: www.elsevier.com/locate/jweia

A comparison of different methods to evaluate the wind induced forces on a


high sided lorry
M. Sterling a,, A.D. Quinn a, D.M. Hargreaves b, F. Cheli c, E. Sabbioni c, G. Tomasini c, D. Delaunay d,
C.J. Baker a, H. Morvan e
a
School of Civil Engineering, The University of Birmingham, Edgbaston, Birmingham B15 2TT, UK
b
School of Civil Engineering, The University of Nottingham, Nottingham NG7 2RD, UK
c
Mechanical Department, Politecnico di Milano, Italy
d
Meteodyn, 75 bd A.Oyon, 72100 Le Mans, France
e
School of Mechanical Engineering, The University of Nottingham, Nottingham NG7 2RD, UK

a r t i c l e in f o a b s t r a c t

Article history: This paper examines the wind induced forces and moments experienced by a high sided lorry. Full-scale
Received 30 May 2008 measurements are combined with wind tunnel and CFD simulations in order to gain an insight into the
Received in revised form ow eld around the vehicle. Differences and similarities between the three techniques are noted. It is
22 July 2009
shown that the rolling moment coefcient obtained from full-scale measurements and CFD simulations
Accepted 21 August 2009
Available online 23 September 2009
agree consistently across a wide range of yaw angles. With respect to the side force coefcient, good
agreement between the wind tunnel and full-scale data are achieved. Pressure distributions over
Keywords: selected sections of the lorry reveal that despite good agreement with the overall forces, the localised
Vehicle aerodynamics pressure eld can be signicantly different.
Cross winds
& 2009 Elsevier Ltd. All rights reserved.
Lift force
Side force
Rolling moment

1. Introduction of attention has not been focused on the road sector. In order to
address this issue a research project was funded by the European
The effect of cross winds on the stability of high sided vehicles Union through the Sixth Framework Program (FP6). Within this
is a signicant issue. For example Snbjornsson et al. (2007) state programme, a CRAFT project was supported whose aim was to
that in the developing world, road accidents cause more injuries develop an innovative Wind Early Alarm system for Terrestrial
and casualties than any other man-made or natural hazard. In transportation (WEATHER).
addition, earlier work undertaken by Baker and Reynolds (1992) The WEATHER project started in October 2004 and ran for just
illustrated that during the storms on the 25th January 1990 there over two years. In total six SMEs where involved (Meteodyn,
were around 400 wind-induced accidents involving either death France; Atmos, France; EMI, France; Nublia, Italy; Geonica, Spain
or injury in the UK. As such the impacts of high winds on and Campbell Scientic, UK). In addition to the SMEs, the
transportation has risen up the research agenda in many fundamental research aspects of the project where undertaken
countries. by the Politecnico di Milano (Italy), the University of Nottingham
Weather warning systems have traditionally existed for long (UK) and the University of Birmingham (UK).
span bridges and in some cases are even automated. However, for One of the key elements of this research was the determination
long stretches of road, where the wind velocity can change of the aerodynamic forces and moments acting on a typical high
signicantly due to local topographical effects, their success has sided vehicle, and it is with this that the current paper is
been questionable. Recent signicant developments have been concerned. In calculating the wind induced forces or moments, it
made in the rail industry. For example Delaunay et al. (2003) is important to consider both the wind velocity relative to the
discuss an automated procedure, based on a scientic probabil- vehicle (Vr) in addition to the angle that this velocity makes with
istic approach that is used for the TGVMe diterrane e in France. respect to the vehicle velocity (Vv), i.e. the yaw angle (y) (see
Despite this and other advances in the rail industry, the same level Fig. 1). With reference to Fig. 1, it is relatively straightforward to
obtain the following expressions:
q
 Corresponding author. Tel.: + 44 121 414 5065; fax: + 44 121 414 3675.
Vr Vw2 Vv2 2Vw Vv cosa 1
E-mail address: m.sterling@bham.ac.uk (M. Sterling).

0167-6105/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jweia.2009.08.008
ARTICLE IN PRESS
M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020 11

Fig. 1. A simple sketch illustrating how the wind velocity (Vw), wind direction (a), vehicle velocity (VV) combine to form the wind velocity relative to the vehicle (Vr). y is the
yaw angle.

 
Vv Vw cosa concerning the aerodynamic forces and moments acting on the
y cot1 2
Vw sina lorry.
where Vw is the velocity of the wind and a is the wind direction
relative to the vehicle velocity. Hence, with reference to Eqs. (1) 2.1. Full-scale experiments
and (2), and Fig. 1 it can be appreciated that for a xed cross wind
(Vw and a constant), then the wind velocity relative to the vehicle A full description of the experimental set-up is given in Quinn
is a direct function of the vehicles speed. Implicit in the derivation et al. (2007). However, in order to aid with interpretation of the
of Eqs. (1) and (2) is the assumption that the vehicle is moving in a results, brief details are presented in this section. Fig. 2 illustrates
straight line. If this is not the case then additional factors need to the high-sided lorry which was used for the purposes of the full-
be taken into account (e.g. slide slip angle). However, for the scale work. The lorry in question was a Leyland DAF 45-130 and
purposes of the current work it is reasonable to disregard these was obtained from a well-known UK publication dedicated to
factors. selling used vehicles. This particular lorry was chosen since it was
If a quasi-steady hypothesis is assumed, (i.e. that uctuations similar to those which typically blow over during periods of high
in the wind induce corresponding forces and moments on the cross winds. The vehicle box was 6.1 m long, 2.5 m wide and 2.5 m
vehicle), then the following equations can be written: tall on a chassis which raises it approximately 1 m off the ground.
The importance of the wind relative to the vehicle was briey
S 12 rACS Vr2 ; L 12 rACL Vr2 ; R 12rAhCR Vr2 3
discussed in the previous section. However, in addition to the
where S is the instantaneous side force, L is the instantaneous lift vehicle motion affecting the relative velocity, it can also modify
force, R is the instantaneous rolling moment, r is the density of the vehicle induced turbulence which in turn can affect the forces
air, A is the side area of the vehicle, h is the height of the vehicle, CS and moments acting on the vehicle. In order to investigate if such
is the mean side coefcient, CL is the mean lift coefcient and CR is phenomena was of importance in the current work a series of
the mean rolling moment coefcient. Hence, provided details of transient (moving vehicle) experiments (phase 1) were under-
these parameters are known, in addition to the relative wind taken. In addition, a second series of experiments (phase 2) were
velocity, the wind induced forces and moments on a given vehicle also undertaken on a stationary vehicle. During the phase 1
can be obtained and an assessment of vehicle stability can be experiments, the pressures on the sides and roof of the cargo
made. In theory, such analysis enables warnings to be given to space were made using pressure tappings. These tapping points
drivers during periods of high winds encouraging them to reduce connected to a differential pressure transducer using exible
their speed. tubing (full details given in Quinn et al., 2007). Integration of the
Section 2 of the paper outlines the experiments that were pressures around the vehicle enabled an evaluation of the rolling
undertaken in order to evaluate the forces and moments acting on moment for a number of yaw angles. Fig. 2(a) also illustrates the
a typical light weight ( o7 ton) lorry. Full-scale, wind tunnel and ultrasonic anemometer and direction vane which were xed the
CFD experiments were undertaken in order to build an extensive lorry in order to measure the wind speed and direction relative to
database of information for a variety of environmental conditions. the vehicle. In phase 2, the rolling moment was measured using
The results of these experiments are examined in Section 3. The four weigh-pad type load cells placed beneath the vehicle and the
results are interpreted in terms of mean aerodynamic coefcients supporting the vehicle chassis directly through four steel posts.
and mean pressure coefcients. Finally, Section 4 draws appro- The length of these posts was chosen to provide a clearance of
priate conclusions from the work presented in the paper. approximately 0.05 m between the wheels and the ground.
Finally, the posts were placed on a solid concrete slab in order
to isolate any ground movement effects on the experiments.
2. Description of experiments One of the main ndings of Quinn et al. (2007), was that the
results obtained from the transient experiments (phase 1) were
For most commercial and research projects it is common to use comparable to those obtained from stationary experiments
either a wind tunnel or a CFD package in order to gain an insight (phase 2). This indicated that the structure of the ow around
into the wind induced forces and moments acting on a vehicle. the vehicle did not signicantly alter with vehicle motion. Hence,
Such experiments are rarely undertaken at full-scale mainly due at full-scale sensible data could be obtained from the stationary
to the cost of the initial experimental set-up and maintenance of experiments and it is this data which will be presented and
the equipment. In addition, there is no guarantee that the discussed subsequently. In the analysis that follows the lift force
appropriate environmental conditions will occur at the required and rolling moment have been obtained directly from load cell
time. Hence, for short term projects (ca. 2 years) such experiments data. The data from the pressure coefcients is used to examine
can be risky to undertake. This project offered the rare the ow structure around the vehicle but is not used to determine
opportunity to compare the results of CFD and wind tunnel the mean coefcients at full-scale. However, the results of Quinn
simulations with data measured at full-scale. This approach was et al. (2007) indicate that the results obtained from the load cell
adopted since it was envisaged that once a successful comparison and pressure data are comparable with one another.
had been achieved with the full-scale data, the CFD and wind One of the benets of undertaking full-scale experiments
tunnel simulations could be used to rapidly investigate a variety relates to the atmospheric boundary layer (ABL). Providing that
of environmental conditions and provide additional information there is a sufcient fetch and no upstream obstacles then the
ARTICLE IN PRESS
12 M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020

Fig. 2. Photographs illustrating the commercial vehicle used for the full-scale experiments: (a) phase 1moving experiments; (b) phase 2static experiments.

experimental situation is an accurate representation of reality. a Pitot and a Cobra probe (Turbulent Flow Instrumentation).
The phase 2 experiments were undertaken at the University of Further details relating to the experimental set-up can be found in
Birminghams Wind Engineering eld site in Bedfordshire, UK. Cheli et al. (2005).
This is a well documented site (Richards et al., 2000; Sterling et al., Fig. 3(a) illustrates the good level agreement between the
2006; Scarabino et al., 2007) and is orientated to face the streamwise velocity data of the wind tunnel data and the
prevailing west-southwesterly winds and has a large unob- corresponding full-scale measurements over the range 0.5 oz/
structed fetch. The site has been proven to have repeatable wind zref o2.5. There is also a similar level of agreement with respect to
conditions and can be characterised by a log-law prole with a the turbulence intensity data (Fig. 3b). At low values of z/zref
roughness length of zo E0.01 m. (o0.5) the turbulence intensities in the wind tunnel rapidly
The full-scale velocity data were obtained using a three decrease. Although unfortunate, this behaviour is not unexpected
component ultrasonic anemometer (Gill Instruments, Lymington, and is a direct consequence of the difculties of simulating the
UK) which recorded data every 0.1 s. Fig. 3(a) illustrates a typical appropriate turbulence conditions at small distances above the
velocity prole for this site. In Fig. 3(a) the reference height (zref is ground. As will be shown later this will affect the ow under
taken as 3 m). Also shown in Fig. 3(a) are data corresponding to the van and have implications for the lift force. However, it will
the wind tunnel and CFD simulations which will be discussed also be shown that in terms of the current project such effects are
later. A logarithmic velocity prole is shown in Fig. 3(a) where it sufciently small as to be considered negligible. The streamwise
has been assumed that a surface roughness length of 0.01 m is velocity spectra at a full-scale equivalent height of 6 m above the
applicable. The agreement between the full-scale data and ground is shown in Fig. 3(c). The spectra has been normalised by
logarithmic prole is encouraging. the square of the mean streamwise velocity (Vw (wind tun-
Fig. 3(b) illustrates the vertical prole of the streamwise nel)= 6.27 m/s). It has become common place to normalise the
turbulence intensity and highlights the rapid increase in this spectra by the variance. However, since the variance is directly
R1
parameter as the ground is approached. Also shown in Fig. 3(b) related to the spectra (i.e. s2i 0 Sii n dn) it cannot be
are the wind tunnel data (which will be discussed subsequently) considered as an independent parameter and as such tends to
and an empirical curve. The latter assumes that the turbulence mask any differences that exist. Fig. 3(c) illustrates that at values
intensity is inversely proportional to the natural logarithm of of normalised frequencies greater than  0.4 there is a good
height above ground divided by the roughness length, (i.e. IVw =1/ agreement between the wind tunnel simulation and the full-scale
ln(z/zo)), and zo is taken as E0.01 m. data. Below these frequencies the agreement is less convincing,
Finally, Fig. 3(c) illustrates the streamwise velocity (SVwVw) which given the physical constraints of the wind tunnel is to be
spectra at 6 m above the ground, i.e. approximately twice the lorry expected. However, in the wind tunnel, the streamwise turbulent
height. In Fig. 3(c) the spectra has been normalised by the square length scale is approximately  1.4 m which is greater than the
of the streamwise velocity component (Vw = 9.5 m/s) and is plotted dimensions of the van and as such any deciencies in the velocity
against the non dimensionalised frequency ( = nz/Vw, where n is spectra can be considered unimportant for the current analysis. It
the actual frequency and z is the height above the ground). This is perhaps worth noting that if the spectra had been normalised
normalisation has been adopted in order to highlight the by the variance, then such deciencies would not be apparent.
differences between the full-scale and simulation data and will Fig. 3(c) also illustrates the von-Karman spectrum for both the
be discussed later. full-scale (FS) and wind tunnel (WT) data. The inclusion of these
curves also illustrate that both sets of data possess the appropriate
2.2. Wind tunnel experiments spectral characteristics and enable further condence to be placed
in these measurements.
The wind tunnel experiments were undertaken at the
Politecnico di Milano on a 1:10 scale model of the lorry shown 2.3. CFD calculations
in Fig. 4. Measurements were undertaken on a scaled model of a
viaduct, on a scaled model of an embankment and on at terrain The CFD experiments that will be described in this paper were
corresponding to the full-scale site outlined in the previous undertaken using an unsteady RANS, ke model and implemented
section. Although, all of the experiments are relevant to the using the software package CFX version 10. The simulations were
overall research programme, it is with the latter that this paper is run as unsteady with a time step of 0.05 s, which was considered
concerned. The aerodynamic forces and moments were measured small enough to be able to resolve the unsteady aerodynamic
using a six component dynamometric balance, positioned under forces. Default wall functions were used and within each time
the lorry and connected to its wheels. The wind speed was step a convergence criteria of 10  4 was selected. The wind
measured at lorry height (full-scale equivalent of 3 m) using both conditions were assumed to follow a logarithmic prole with a
ARTICLE IN PRESS
M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020 13

3.5
full-scale Wind tunnel CFD log-law
3

2.5

z/zref
1.5

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Vw /Vw-ref

3.5
full-scale Wind tunnel log-law
3

2.5

2
z/zref

1.5

0.5

0
0 5 10 15 20 25 30
IVw (%)

0.025
full-scale wind tunnel Von Karman (FS) Von Karman (WT)

0.02
fSVwVw(f)/Vw2

0.015

0.01

0.005

0
0.01 0.1 1 10 100
f

Fig. 3. Velocity data corresponding to full-scale, wind tunnel and CFD simulations: (a) normalised streamwise velocity prole. zref taken as 6 m and Vw  ref = 6.645, 6.645 and
13.89 m/s for the full-scale data, wind tunnel data and CFD data respectively; (b) normalised turbulence intensity with respect to height; (c) normalised streamwise velocity
spectra at 6 m above the ground.

roughness length of 0.01 m. As illustrated in Fig. 3(a), the velocity found in Hargreaves and Morvan (2007). However, for the
prole closely matches the full-scale data for values of z/zref o2.0. purposes of the current paper it is sufcient to note that
For values of z/zref 42.0 the departure between the full-scale data boundary layers were grown from the main surfaces of the
and that of the CFD simulation is still considered to be within trailer and size functions were implemented to grow the mesh
acceptable limits. away from detailed features such as the mirrors. Non-conformal
All force, moment and pressure coefcients presented here are interfaces were used to transition from the unstructured mesh
time-averages of the CFD results obtained after all start-up around the cab and chassis to the structured mesh around the
transients had disappeared and the simulations had entered a trailer. A mesh of 3.5 million hexahedral, tetrahedral and
stationary state. Fig. 5 provides an insight into the structure triangular prism cells was created using the aforementioned
surface mesh adopted. A full description of the simulation can be techniques.
ARTICLE IN PRESS
14 M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020

Fig. 4. Scale model of the lorry in the Politecnico di Milanos wind tunnel: (a) vidaduct; (b) at terrain.

Fig. 5. A photograph of the full-scale lorry compared to the surface mesh of the equivalent section of the CFD model.

The domain around the truck consisted of two parts, i.e. the signicant, (see Quinn et al., 2007). The experimental errors in the
turntable which mimicked the turntable in the wind tunnel and wind tunnel data are assumed to be of the order of 5%.
the fetch. The vertical axis of the turntable was coincident with The general behaviour of the rolling moment coefcient (Fig. 6)
the centre of the truck enabling any yaw angle to be achieved. is consistent across all three of the techniques employed, i.e. a low
Similar to the wind tunnel experiments, simulations were also yaw angles the coefcient tends towards zero and there is an
undertaken for the case of an embankment and viaduct. However, almost linear increase of coefcient with respect to yaw angles for
for the purposes of this paper these are not reported. values of up to 401. Beyond 401 the rate of increase in rolling
As outlined above, full details of the full-scale experiments, moment reduces and peaks around 601 for the wind tunnel
wind tunnel tests and CFD simulations can be found elsewhere measurements and 801 for both the CFD and full-scale measure-
(Quinn et al., 2007; Cheli et al., 2005; Hargreaves and Morvan, ments. Intuitively one may expect the maximum value of CR to
2007, respectively). However, the following section undertakes a occur at 901, however, the formation of unsteady ow structures
comparison between the three different techniques and analyses over the roof and leading edge of the vehicle result in regions of
new data which cannot be found in any of the above references, high suction, which in turn effects the rolling moment. This
i.e. in particular the pressure distributions. behaviour is in keeping with the results of previous studies on
trains (Baker et al., 2004; Bocciolone et al., 2003, 2008).
Overall the agreement between the CFD and full-scale data is
remarkable and will be examined further shortly. However, it is
3. Results noticeable that the wind tunnel results are signicantly higher
than both the full-scale and CFD results. At this stage, it is perhaps
3.1. Comparison of mean coefcients worth noting that two sets of wind tunnel experiments were
undertaken, one in July 2006 and one in September 2006. A close
At low yaw angles a vehicle is more likely to undergo a course agreement between both sets of data was found but for the
deviation as a result of a high cross wind, whereas at high yaw purposes of clarity only the data relating to the September
angles, stability issues are of paramount importance, i.e. a lorry is experiments are shown. This close agreement indicated a good
more likely to overturn. With this in mind, we will initially repeatability in the wind tunnel experiments but does not explain
consider the variation of rolling moment coefcient (CR) with the differences between the wind tunnel data and the full-scale/
respect to yaw angle (Fig. 6). It must be stressed at this point that CFD data.
the data presented relate to mean values of this parameter, and in Unfortunately, due to the setup of the full-scale experiments it
the case of the full-scale data, the experimental scatter can be was not possible to obtain any information with respect to the
ARTICLE IN PRESS
M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020 15

1.4
Full Scale
1.2 Wind Tunnel
CFD

Rolling moment coefficient


1

0.8

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80 90
Yaw angle

Fig. 6. Mean rolling moment with respect to yaw angle.

1.40

Wind Tunnel
1.20
CFD

1.00
Side force coefficient

0.80

0.60

0.40

0.20

0.00
0 10 20 30 40 50 60 70 80 90
Yaw angle

Fig. 7. Mean side force with respect to yaw angle.

side force. Hence, Fig. 7 presents information relating to only the the lift force, side force and rolling moment are interrelated such
wind tunnel and CFD simulations. It is interesting to note that in statements are tentative at best. These relationships will be
this case a good agreement between the wind tunnel and CFD examined further in the following section.
data is found.
Since both the CFD and wind tunnel data agree with respect to
the side force coefcient but disagree with respect to the rolling 3.2. Comparison of pressure distributions
moment coefcient, it is reasonable to assume that there will be a
disagreement in terms of the lift force coefcient, since both the In order to understand the differences in force and moments
lift and side forces contribute to the rolling moment. Fig. 8 coefcients obtained from the different simulations outlined
illustrates that this is indeed the case. From the CFD simulations, above, attention will now be focused on the distribution of
the lift force coefcient appears to gradually rise to a maximum pressure over the vehicle. Due to the nature of the full-scale
value at a yaw angle of 451 and then gradually decreases for conditions experienced during phase 2 only pressure data
higher yaw angles; the trend is almost symmetrical about a yaw of corresponding to yaw angles of 501, 701 and 851 were obtained.
451. Full-scale and wind-tunnel data illustrate a gradual increase For the purposes of the current work the mean pressure
in the lift force coefcient for yaw angles greater than 351 and 401 coefcient (Cp) is dened as:
respectively, whereas below these values the lift coefcient is Pi  P
small and predominately negative. Cp 4
0:5rVw 32
The lack of disagreement between the trend of the CFD data
and that obtained from the other experimental methods perhaps where Pi is the pressure at tap i, P is the ambient (reference)
suggests an inability of the CFD simulation to reproduce the pressure, r is the density of air and Vw(3) is the mean streamwise
correct form of the pressure distribution over the roof of the velocity at 3 m above the ground. Fig. 9 illustrates a sketch of the
vehicle, since it is assumed that the ow over the roof inuences vehicle illustrating the coordinate system adopted in the current
the lift force more than the oor under the vehicle. However, since analysis and will aid in the discussion that follows.
ARTICLE IN PRESS
16 M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020

1.4
Full Scale
1.2 Wind Tunnel
CFD
1

Lift force coefficient


1.8

0.6

0.4

0.2

-0.2
0 10 20 30 40 50 60 70 80 90
Yaw angle

Fig. 8. Mean lift force with respect to yaw angle.

W However, over the majority of the trailer, i.e. 0.3 oy/Lo0.8, the CL
and LE measurements are similar indicating that the ow of the
roof of the vehicle results in a reasonably uniform pressure
H distribution at these two locations. It is interesting to note that CL
measurements are less negative over this region (0.3oy/Lo0.8)
whereas the opposite is true for the LE data.
The pressure coefcient obtained from the wind tunnel
experiments is consistently smaller than that of the full-scale
z experiments over the leading edge of the roof. However, over the
x
centre line of the vehicle for values of 0.4 oy/Lo0.7, the
y L agreement is reasonable. Fig. 8, illustrates that the mean lift
coefcient corresponding to this particular yaw angle obtained
from the full-scale data has a larger magnitude that that
Fig. 9. A sketch illustrating the coordinate system adopted: (a) 851 yaw; (b) 701 corresponding to the equivalent wind tunnel data. This appears
yaw; (c) 501 yaw.
to be contradictory to the results of Fig. 10(a), i.e. lower values of
Cp would imply a greater lift force. There are a number of possible
Fig. 10 illustrates the pressure coefcient data corresponding to reasons which may explain this contradiction. Firstly, Fig. 10(a)
a variety of yaw angles. In this gure, y is the distance along the illustrates data corresponding to two locations and as such does
roof of the trailer (y=0 corresponding to the back of the cab) and L not give an indication of the pressure eld over the entire roof.
is the length of the trailer (6.1 m), x corresponds to the distance However, it is not unreasonable to assume that these two
from the leading edge of the windward face of the trailer and W is locations give a reasonable indication of the respective differences
the overall width of the trailer (2.5 m). Data corresponding to between the pressure eld in the two experiments. Therefore, it is
measurements obtained at the centre line of the trailer (CL) are unlikely that additional data would provide a signicant insight.
shown in addition to those at the leading edge (LE). It should be The second, and perhaps more likely explanation relates to the
noted that the LE data are not necessarily located at the same ow under the vehicle. By the very nature of the two experiments
position. For example, the full-scale measurements, wind tunnel and the difculties in correctly generating the appropriate
measurements and CFD simulations take a values of x/W= 0.16, 0.1 turbulence conditions close to the ground it is unlikely that the
and 0.1 respectively. Notwithstanding these differences, the pressure eld underneath the vehicle would be the same in both
location of the pressure taps are considered sufciently close simulations. Since the ow under the vehicle as well as over it
together in order to enable comparisons to be made. contributes the lift force, it perhaps explains why the differences
Fig. 10(a) illustrates data corresponding to a yaw angle of 851. occur. Hence, this is thought to be the major reason for the
The full-scale data illustrates that the CL measurements are differences between outlined above.
consistently lower than the corresponding values at the leading The CFD simulations illustrate a reasonably uniform distribu-
edge, indicating a higher degree of suction. This suggests that at tion along the leading edge and centre line of the roof for both the
full-scale the separation zone associated with the ow over the 851 yaw angle. In general, the CFD results illustrate larger values of
roof of the vehicle extends beyond the CL. In reality such a zone Cp than either the full-scale or wind tunnel simulations. These
would be highly unsteady, but the values of Cp suggest that on differences between the CFD data and those of the other
average this zone is relatively large. The LE data illustrates a slight experiments begin to explain the differences in lift coefcient
downwards trend with increasing values of y/L, whereas the same observed in Fig. 8.
trend is true for the CL data for values of y/L o0.4, after Cp Figs. 10(b) and (c) illustrate the mean pressure coefcient over
remains reasonably constant. The increased values of Cp at low the roof of the vehicle for the full-scale and wind tunnel
values of y/L can possibility be attributed to the cab exerting a experiments. For the 701 yaw case similar trends to those
small inuence on the local ow eld. illustrated in Fig. 10(a) are apparent, i.e. the wind tunnel centre
At the two ends of the trailer the trend illustrated in the full- line values are lower than corresponding full-scale values. Also
scale data can also be observed in the wind tunnel experiments. over a large proportion of the roof of the vehicle the wind tunnel
ARTICLE IN PRESS
M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020 17

85 yaw
0

-0.5

-1

Cp
-1.5

full-scale (LE) wind tunnel (LE) CFD (LE)


-2
full-scale (CL) wind tunnel (CL) CFD (CL)

-2.5
0 0.2 0.4 0.6 0.8 1
y/L

70 yaw
0

-0.5

-1
Cp

-1.5

full-scale (LE) wind tunnel (LE) CFD (LE)


-2
full-scale (CL) wind tunnel (CL) CFD (CL)

-2.5
0 0.2 0.4 0.6 0.8 1
y/L
50 yaw
0

-0.5

-1
Cp

-1.5

-2 full-scale (LE) wind tunnel (LE) CFD (LE)


full-scale (CL) wind tunnel (CL) CFD (CL)
-2.5
0 0.2 0.4 0.6 0.8 1
y/L

Fig. 10. Pressure distributions over the roof of the vehicle. (LE: full-scale, x/W = 0.16, wind tunnel, z/W= 0.1; W = 2. 5m and 262 mm at full-scale and wind tunnel scale
respectively): (a) 851 yaw; (b) 701 yaw; (c) 501 yaw.

CL and LE data are similar. The full-scale measurements along LE these yaw angles (Cook, 1985; Coleman and Baker, 1994).
are reasonably uniform along the entire length of the trailer Interestingly, this behaviour is not observed at full-scale. The
whereas the corresponding CL measurements appear to have a data corresponding to the full-scale CL and LE measurements
local minimum at approximately half way along the trailer. It is illustrates similar trends and magnitudes, i.e. a slight increase for
interesting to note, that unlike the 801 yaw angle case the small values of y/L, remaining reasonably constant across a large
agreement between the full-scale and CFD LE results for values of proportion of the roof (y/L40.2).
0 oy/Lo0.4 is encouraging. As the yaw angle reduces to 501, the Fig. 11 illustrates a selection of pressure coefcient data
agreement between the full-scale and the CFD simulations corresponding to the windward face of the trailer. In Fig. 11, CL
increases over a wider range of y/L values. corresponds to the horizontal centreline on the face of the van and
The inuence of the cab can clearly be seen in the CL UE represents the upper edge. Similar to the results presented in
measurements corresponding to the wind tunnel. This effect can Fig. 10 the pressure data corresponding to the UE are not
be more clearly illustrated in Fig. 10(c) where the yaw angle is 501. coincident, i.e. for the full-scale data, wind tunnel data and CFD
The wind tunnel data corresponding to the CL, clearly illustrates data z/H is equal to 0.83, 0.88 and 0.88 respectively. Also, it is
evidence of a high suction zone extending over a large proportion worth noting that the wind tunnel data corresponding to CL was
of the trailer. It is hypothesised that these low values of Cp arise as measured at z/h =0.63 as opposed to 0.5. Notwithstanding these
a result of a delta-wind type vortex which is known to exist at differences, the location of the pressure taps are considered
ARTICLE IN PRESS
18 M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020

85 yaw
1

0.8

0.6

Cp 0.4
full-scale (UE, z/H=0.83) wind tunnel (UE, z/H=0.88)
0.2 CFD (UE, z/H = 0.83) full-scale (CL, z/H=0.5)
wind tunnel (CL, z/H=0.63) CFD (CL, z/H = 0.63)

0
0 0.2 0.4 0.6 0.8 1
y/L

70 yaw
1

0.8

0.6
Cp

0.4
full-scale (UE, z/H=0.83) CFD (UE, z/H=0.88)
0.2 wind tunnel (UE, z/H=0.88) full-scale (CL, z/H=0.5)
wind tunnel (CL, z/H=0.63) CFD (CL, z/H=0.63)

0
0 0.2 0.4 0.6 0.8 1
y/L
50 yaw
1
full-scale (UE, z/H=0.83) wind tunnel (UE, z/H=0.88)
0.8 CFD (UE, z/H=0.88) full-scale (CL, z/H=0.5)
wind tunnel (CL, z/H=0.63) CFD CL, z/H=0.63)

0.6
Cp

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
y/L

Fig. 11. Pressure distributions over the windward face the vehicle: (a) 851 yaw; (b) 701 yaw; (c) 501 yaw.

sufciently close together in order to enable comparisons to be obtained (Fig. 11(b)). However, in contrast the difference between
made. the CL and UE values for the CFD results remains almost constant.
The data in Fig. 11(a) illustrate the same general trend for all of As the yaw angle decreases further to 501 (Fig. 11(c)), the values
the results, i.e. a reasonably uniform distribution across the of Cp obtained from the wind tunnel are larger than their
majority of the windward face with a reduction at either end of corresponding full-scale and CFD counterparts. Fig. 11(c) illus-
the trailer. The full-scale values are generally larger than the trates that irrespective of which experiment is considered, there is
corresponding wind tunnel data. The CFD results corresponding to a general of decrease Cp with increasing y/L.
the CL are larger than both the wind tunnel and full-scale data. The pressures on the windward face of the vehicle directly
However, with respect to the UE, the CFD results illustrate good contribute to the wind induced side force. For a yaw angle of 501,
agreement with the wind tunnel data over a wide range of y/L the wind tunnel pressures are larger than those predicted by CFD
values. and support the ndings illustrated in Fig. 7. However, for yaw
As would be expected, the values of CL across all of the angles of 701 and 801, the same ndings are not apparent which
experiments are larger than their counterpart values of UE. This would suggest behaviour contradictory to that illustrated in Fig. 7.
trend is also observed for two yaw angles shown in Figs. 10(b) and To explore this issue further would require a detailed analysis of
(c). However, as the yaw angle reduces from 851 to 701, the the pressure distribution over the entire windward and leeward
difference between full-scale data and wind tunnel data reduces faces of the vehicle. Unfortunately, such data was not obtained at
and over a large proportion of the trailer similar values of Cp are the time of the experiments and as a result it is difcult to
ARTICLE IN PRESS
M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020 19

85 yaw
1
full-scale (y/L=0.13)
0.8
Wind tunnel (y/L=0.81)
full-scale (y/L=0.94)
0.6
CFD (y/L=0.19)

z/H
Wind tunnel (y/L=0.19)
0.4
CFD (y/L=0.81)

0.2

0
0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
CP

70 yaw
1

0.8

0.6 full-scale (y/L=0.13)


z/H

Wind tunnel (y/L=0.81)


0.4 full-scale (y/L=0.94)
CFD (y/L=0.19)
0.2 Wind tunnel (y/L=0.19)
CFD (y/L=0.81)
0
0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
Cp
50 yaw
1
full-scale (y/L=0.13)
0.8 Wind tunnel (y/L=0.81)
full-scale (y/L=0.94)
0.6 CFD (y/L=0.19)
z/H

Wind tunnel (y/L=0.19)


CFD (y/L=0.81)
0.4

0.2

0
0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
Cp

Fig. 12. Pressure distributions over the leeward face the vehicle. (a) 851 yaw; (b) 701 yaw; (c) 501 yaw.

speculate as to why Figs. 7 and 10(a) and (c) appear to illustrate Fig. 12(a) illustrates that the pressure distribution between the
contradictory information. front and rear of the vehicle is reasonably consistent across the
Fig. 12 illustrates the pressure distributions over the leeward height of the vehicle in the wind tunnel. Similarly the CFD results
face of the vehicle. Data is only available for a selected number of illustrate an almost uniform distribution at the front of the vehicle
locations since the experimental resource was focused on and a reasonably uniform distribution at the rear of the vehicle for
obtaining data relating to the windward side and roof of the values of z/H40.4. The magnitude of the pressure coefcients are
vehicle. With respect to the leeward face data it is worth noting self consistent with respect to the wind tunnel and CFD results, i.e.
that the location of each set of pressure taps are similar to one both the front and rear coefcients are similar in magnitude with
another but not identical (with the exception of the wind tunnel respect to location but differ with respect to the simulation
and CFD results). For example, data relating to the front of the technique. However, although the full-scale data illustrate self
vehicle was collected at normalised distances of y/L equal to 0.13 consistent trends there is a larger degree of scatter between the
and 0.19 corresponding to full-scale and wind tunnel/CFD results obtained for the front and rear of the vehicle. In the full-
respectively. Whereas the rear pressure taps were positioned at scale experiments, the pressure at the rear of the vehicle is
a normalised distances of 0.94 and 0.81 (full-scale and wind consistently more negative than that predicted by either the wind
tunnel/CFD respectively). Despite these differences the locations tunnel or CFD experiments. However, as the yaw angle decreases
are considered close enough together to enable general this observation does not hold, and for the 501 case (Fig. 12(c)) the
comparisons to be made. full-scale data are bounded by the wind tunnel results. It is
ARTICLE IN PRESS
20 M. Sterling et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 1020

interesting to note that in general the CFD results tend to be for FS and WT agree, they differ in magnitude at a number of
consistently less negative than the either the full-scale or wind yaw angles. It is likely that these differences arise as a result of
tunnel results at the same location. Overall, the results illustrated the difculties in correctly simulating the near ground ow
in Figs. 11(b) and (c) highlight that as the yaw angle decreases the conditions.
variation of pressure coefcient with respect to z/H tends to  An investigation of the pressure coefcients along four sections
become more uniform, irrespective of location. illustrate differences in the results obtained from the three
The pressures on the windward face and roof of the vehicle simulation methods. These results give an indication of the
combine to contribute to the wind induced moment experienced differences in ow eld that exist around the lorry.
by the vehicle. The differences observed in Fig. 6 can also be better
understood with reference to Figs. 9 and 10, i.e. the differences
between the wind tunnel and full-scale data arise as a result of
differences in the simulation of the ow over and under the vehicle. Acknowledgements
It is unlikely that these differences could be resolved, given the
fundamental restrictions imposed by scale model testing. Funding for the research outlined in this paper was provided
Examining the mean coefcient data obtained in Figs. 69 by the EU under the auspices of the Sixth Framework programme
together with that obtained from Figs. 1012 it is clear that the for European research contract CT-2004-512862. The nancial
ow over the vehicle is complex. Furthermore, it appears as if support is gratefully acknowledged.
there are signicant disagreements between the wind tunnel and
CFD simulations. However, it transpires that if these pressure
References
distributions are area averaged (in order to calculate the mean
force/moment coefcients) then in general, the disagreements
Baker, C.J., Reynolds, S., 1992. Wind-induced accidents of road vehicles. Accident
between the simulations are reduced. In other words it is perhaps Analysis & Prevention 24 (6), 559575.
not unreasonable to say that the agreement between the CFD and Baker, C.J., Jones, J., Lopez-Calleja, F., Munday, J., 2004. Measurements of the cross
full-scale data in terms of the rolling moment is somewhat wind forces on trains. Journal of Wind Engineering and Industrial Aero-
dynamics 92 (78), 547563.
fortuitous. Bocciolone, M., Cheli, F., Corradi, R., Diana, G., Tomasini, G., 2003. Wind tunnel tests
for the identication of the aerodynamic forces on rail vehicles. In: 11th
International Conference on Wind Engineering, Lubbock, Texas, USA, 25 June.
4. Conclusions Bocciolone, M., Cheli, F., Corradi, R., Muggiasca, S., Tomasini, G., 2008. Crosswind
action on rail vehicles: wind tunnel experimental analyses. Journal of Wind
Engineering and Industrial Aerodynamics 96 (5), 584610.
This paper has examined the wind induced forces and Cook, N.J., 1985. The designers guide to wind loading of building structures. Part 1,
moments experienced by a high sided lorry using three different background, damage survey, wind data and structural classication. Butterworths
(for) Building Research Establishment, Department of the Environment, London.
approaches: full-scale measurements (FS); wind tunnel measure-
Coleman, S.A., Baker, C.J., 1994. The reduction of risk for high sided road vehicles in
ments (WT) and CFD simulations. The main conclusions arising cross winds. Journal of Wind Engineering and Industrial Aerodynamics
from the work can be summarised as follows: 44 (13), 401429.
Cheli, F., Belforte, P., Melzi, S., Sabbioni, E., Tomasini, G., 2005. A numerical-
experimental approach for evaluating cross wind aerodynamic effects on
 All three techniques illustrate that the mean streamwise heavy vehicles. In: XIX IAVSD International Association for Vehicle System
velocity prole is correctly simulated. Dynamics, 29 August2 September, Milano, Italy.
 There is reasonable agreement between the WT and FS Delaunay, D., Cle on, L.M., Sacre , C., Sourget, F., Gautier, P.E., 2003. Designing a wind
alarm system for the TGV-Me diterrane e. In: 11th International Conference on
turbulence intensities across a wide range of heights. However, Wind Engineering, Lubbock, Texas, USA, 25 June.
as the ground is approached differences in the data are Hargreaves, D.M., Morvan, H.P., 2007. Towards the validation of crosswind effects on
noticeable. Although these differences exists, they can be a static high-sided vehicle. NAFEMS International Journal of CFD Case Studies 7.
Quinn, A.D., Sterling, M., Robertson, A.P., Baker, C.J., 2007. An investigation of the
considered negligible in terms of evaluating the overall force wind induced rolling moment on a commercial vehicle in the atmospheric
and rolling moment coefcients on the lorry. boundary layer. Proceedings of the Institution of Mechanical Engineers, Part D,
 An excellent agreement between the FS and CFD results are Journal of Automobile Engineering 221 (11), 13671379.
Richards, P.J., Hoxey, R.P., Short, J.L., 2000. Spectral models for the neutral
obtained across a wide range of yaw angles for the rolling
atmospheric surface layer. Journal of Wind Engineering and Industrial
moment coefcient, while the WT data are larger than that of Aerodynamics 87 (23), 167185.
either the FS or CFD data. Scarabino, A., Sterling, M., Richards, P.J., Baker, C.J., Hoxey, R.P., 2007. An
investigation of the structure of ensemble averaged extreme wind events.
 The side force coefcient illustrates similar trends between the
Wind and Structures 10 (2), 135151.
WT and CFD data. When interpreted in conjunction with the Snbjornsson, J.Th., Baker, C.J., Sigbjornsson, R., 2007. Probabilistic assessment of
rolling moment coefcient this implies that there are differ- road vehicle safety in windy environments. Journal of Wind Engineering and
ences in the lift force simulation between the two methods. Industrial Aerodynamics 95 (911), 14451462.
Sterling, M., Baker, C.J., Richards, P.J., Hoxey, R.P., Quinn, A.D., 2006. An
 There are large differences in the trend of lift force coefcient investigation of the wind statistics and extreme gust events at a rural site.
between the CFD data and the WT/FS data. Although the trends Wind and Structures 9 (3), 193215.

You might also like