Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: This paper presents new results concerning aerodynamics of a bluff body fitted with non-flat underbody diffusers.
Aerodynamics For the current study a computational approach is used, taking the advantage that CFD has become a valuable and
Bluff body important tool in road vehicle aerodynamics. There are analyzed two basic shapes, namely circular and the
Curved underbody diffuser
elliptical surfaces. As in previous investigations performed, angle and length of the diffuser are the parameters
CFD
systematically varied within ranges relevant for a hatchback passenger car. Coefficients of lift and drag are
computed and compared with the values obtained for the flat underbody diffuser. In order to get a better over-
view, 3D maps of both lift and drag coefficients were generated using polynomial interpolation of their values.
The results reveal significant improvements concerning the lift generated by body fitted with curved diffusers,
which are emphasised by an average growth of the downforce larger than 20%. In addition, for small and
moderate angles, and small diffuser lengths, the curved diffusers lead also to smaller values of drag, compared to
the plane diffusers. Thus, short diffusers with curved sections work better from aerodynamic point of view. In
terms of vehicles design, short diffusers have also the advantage of small interferences with other underbody
components of cars, as rear wheelhouses and rear silencer. In addition, they offer a significant larger cargo space.
* Corresponding author.
E-mail address: angel.h@unitbv.ro (A. Huminic).
https://doi.org/10.1016/j.jweia.2020.104300
Received 30 January 2020; Received in revised form 4 July 2020; Accepted 4 July 2020
Available online xxxx
0167-6105/© 2020 Elsevier Ltd. All rights reserved.
A. Huminic, G. Huminic Journal of Wind Engineering & Industrial Aerodynamics 205 (2020) 104300
configurations. Kekus and Angland (2018) also examined an Ahmed type Fig. 2. Ahmed body, 3D view and side view, plane (a), circular (b), and ellip-
tical underbody diffuser (c) (dimensions are in mm).
body fitted with two-steps underbody plane diffuser, and Ehirim et al.
(2018b) studied a plane underbody diffuser having a convex bump near
the end. Mair (1969), Grandemange et al. (2015) and Bonnavion and current study. Thus, several CFD (Computational Fluid Dynamics) ana-
Cadot (2019) also analyzed the effects of underbody diffusers with lyses were performed, as in previous investigation Huminic and Huminic
boat-tail shape on aerodynamics of Ahmed type bluff bodies. All (2017), using the facilities provided by ANSYS CFX. In the first stage, the
mentioned works highlight the positive effect of the underbody diffusers accuracy and validity of the adopted computational model and procedure
on the aerodynamics of the studied bodies. were checked using the experimental results of Strachan et al. (2007).
Following this basic idea, to increase the depression under cars, The main stages of the simulations are described in the following
current paper presents new results concerning the aerodynamics of a sections.
bluff body fitted with non-flat underbody diffusers. There are analyzed
two basic shapes, namely circular and elliptical surfaces. The angle (ϕd)
2.1. CAD and computational models
and the length of the diffuser (ld) are the parameters systematically
varied within ranges relevant for a hatchback passenger car. Coefficients
The simulations were carried out using the bluff body of Ahmed et al.
of lift (cL) and drag (cD) are compared with the values obtained for the
(1984), because its aerodynamics has been widely investigated in many
flat underbody diffuser, and the results reveal significant improvements
studies and it is very well known, as mentioned by Rao et al. (2018). It is
of the aerodynamic characteristics of the body fitted with curved dif-
1044 mm long, 389 mm wide and 288 mm height, and consists of three
fusers, mainly regarding the downforce generated. Thus, current work
sections: a short curved one in front, followed by a long straight middle
represents the first systematic study concerning curved underbody dif-
section and an upper angled (α) rear end. For current study, the angle α ¼
fusers in two well-defined configurations, which reveals not only their
35 was adopted because it corresponds to a low drag (cD ¼ 0.275) and
benefits as in already published research, but also their limitations.
stable aerodynamic configuration, as shown in Fig. 5. Thus, it represents
a generic 1:4 scale model of lower-medium size hatchback vehicles, as
2. Research methodology
depicted in Fig. 2, which show also the lateral view of the bodies
investigated in the current study. Dashed lines show the contours of
Based on amount of work and the large number of models involved
previous diffusers.
during investigation, a computational approach was adopted for the
For the first body, a circular sector of radius rd was taken into account
2
A. Huminic, G. Huminic Journal of Wind Engineering & Industrial Aerodynamics 205 (2020) 104300
The size of the first layer of cells around it was set to 5∙10-4 m in order to
fulfil the condition 30 yþ100 for computations using the logarithmic
wall function approach, and a growth rate of 1.1 was assigned. This
approach represents a good compromise between the accuracy and cost-
efficiency of CFD simulations, and produces consistent results at high
Reynolds numbers, as in road vehicle aerodynamics. Although a low y þ
approach (yþ1) is desired because it resolves almost whole spectrum of
eddies, the effort and computational resources needed are still too large
for many practical applications. In addition, the low y þ wall treatment
has a more restricted range of application, being suitable mainly for low
Reynolds turbulence models, since it assumes that the viscous sublayer is
properly resolved, as stated by Fu et al. (2018). Ten layers were gener-
ated also for the surface representing the ground. Since simulations are
performed using moving ground, as mentioned in the next paragraph, its
influence on the velocity profile is not significant, since the velocity of
ground equals the velocity of free stream. A detail of mesh around body
Fig. 3. Computational domain (not to scale) and Mesh detail around body. in in the symmetry plane is shown in Fig. 3. The number of elements for
the symmetry plane. entire computational domain is larger than 8600000, of which more than
190000 elements are on surfaces of the body with elliptical diffusers. The
for the underbody diffuser, having the beginning tangent (T) to the flat results concerning the independence of the solutions with the size of
surface of the underbody, computational domain and mesh are presented in the section devoted to
validation of the CFD procedure.
rd ¼ ld2þhd22hd, (1)
where hd is the height of the diffuser 2.2. Simulations setup and boundary conditions
hd ¼ ld tgϕd. (2) The used code solves the full Reynolds-Averaged Navier-Stokes
equations in their conservation form: continuity equation and mo-
For the second one, a quarter elliptical shape was considered. Thus,
mentum equation. Unsteady analyses (URANS method) were performed
both ends of the diffuser are tangent to the flat surfaces of the body with
for incompressible, isothermal and fully turbulent flows (T ¼ 0.2%) in
which they are in connection. Hence, the length ld and the height hd of
conditions of the International Standard Atmosphere: p∞ ¼ 101325 N/
the diffuser are the semi-minor axis of an ellipse. Therefore, shapes of
m2 p∞ ¼ 1 At.p∞ ¼ 1At.(reference pressure) and tair ¼ 15 C T∞ ¼
both curved diffusers can be expressed mathematically. The length and
288.15 KT∞ ¼ 288.15 K (reference temperature of the air). The solver
angle are considered in the same way as for the plane diffuser.
uses these parameters to compute density and viscosity of the air, ρair ¼
In the next step, CAD models were integrated into a rectangular
1.225 kg/m3 and μair ¼ 1.78938∙10-5 kg/(m∙s) respectively. Shear-
enclosure computational domain, as shown in Fig. 3. Generally, the size
Stress-Transport (SST) closure model of Menter (1994) was employed
of computational domains depends on many factors, as complexity of
to perform current simulations. Using a blending function to switch from
geometry and simulation, computational approach and code used to
k-ω model in the near-wall region to k-ε model in the far field of flow, it
simulate the studied phenomena, computational grid and concatenation
combines the advantages of both mentioned turbulence models. This
of elements in areas of interests. Even though there is no predefined
results in a major improvement concerning the prediction of the aero-
solution concerning this aspect, studies involving this bluff body, or
dynamic loads and flow separation. The performance of SST turbulence
similar to it, as those of Guilmineau (2008), Cheng et al. (2011) and Rao
model has been tested successfully in various studies involving fluid
et al. (2018), are performed using rectangular computational domains
dynamics phenomena, including external aerodynamics, as the studies
having dimensions within the following ranges: 2 l–3 l upstream the body
performed by Fu et al. (2018), Urquhart et al. (2018) and Mohottia et al.
and 3 l–6.8 l downstream, 0.7 l–1.8 l on both sides of the body and 1.1
(2019).
l–1.59 l above, relative to the length (l) of the body.
In order to accelerate the convergence of the analyses, steady state
These dimensions allowed blockage ratio smaller than 2%, which is
simulations were conducted in the first stage. Their solutions were used
within the recommended value of 5% in wind tunnel testing of road
in the next stage as initial conditions for transient analyses. These were
vehicles, as stated by Hucho and Sovran (1993). Based on studies
performed for a time of 1 s with constant time steps of 0.02 s. Thus, the
mentioned above, current computational domain has the following
duration of transient analyses is almost 2.5 times larger than physical
dimensions:
timescale of simulations (0.40455 s), which is the time taken for a point
in the flow to make its way through the fluid domain. It is computed as
- 2.5 l in front of the body,
length of the computational domain divided by the velocity of free
- 12 l behind the body,
stream. The time step was considered in such way that it is smaller than
- 1.5 l for top and lateral boundaries.
the time necessary for a point in the flow to cover a distance equal to the
characteristic length of the flow (length of the body) with the velocity v∞
According to previous works Huminic et al. (2012) and Huminic and
(0.0261 s). At each time step, flow variables were evaluated for cycles of
Huminic (2017), a larger length was considered behind the body, which
10 iterations. In order to achieve high precision results, residual target of
is placed 50 mm above the ground, as in the original study of Ahmed,
RMS was set up to 1e-006. In addition, history of both drag and lift co-
isolated insight the computational domain. In this way, no unnecessary
efficients was also monitored, and it was found that their variations
interference affects the flow around it. The blockage ratio for current
become insignificant during final stage of simulations. The results were
study is 1.7%, significantly smaller than in a wind tunnel test chamber, to
collected for the final step of each simulation.
avoid negative pressure effect between body and domain boundaries.
According to the frame in Fig. 3, the following boundary conditions
A multi-block scheme was used to generate the mesh of the compu-
were considered:
tational domain. It consists on tetrahedral and pentahedral elements in
- a uniform and constant velocity at inlet, vx ¼ v∞, vy ¼ vz ¼ 0, and
the proximity of the solid surfaces, in order to solve accurately the flow
also for ground, as for moving walls, in order to simulate the relative
inside the boundary layer, which consists of 30 layers around the body.
motion between vehicle and road; in this way, all the on-road boundary
3
A. Huminic, G. Huminic Journal of Wind Engineering & Industrial Aerodynamics 205 (2020) 104300
Table 1
Change in drag and lift with size of the computational domain.
lb/l 6 9 12 15
ΔcD – 0.0008 0.0004 0.0002
ΔcL – 0.001 0.0005 0.0004
Table 2
Change in drag and lift with size of the mesh on surfaces of the body for lb/l ¼ 12.
No. of elements 81628 125784 163424
2.3. Validation of the CFD procedure computational domain parameters, a ten times smaller value ΔcDffi 10-4
was considered the limit between two CFD analyses in row. Based on
Accuracy of the results provided by the computational model and CFD above results it was concluded that simulations do not depend with the
procedure used in the current study have been checked against the size and mesh of the computational domain, as mentioned in the para-
experimental results provided Strachan et al. (2007) with moving ground graph devoted to CAD and computational models.
(MG). Thus, several simulations were performed in similar conditions The results concerning aerodynamic coefficients are depicted in
employed in these experiments. The independence of solutions with the Figs. 4 and 5, for all angles (α) of the upper slanted rear surface of the
size of computational domain and mesh was checked for several di- Ahmed body. Current CFD results concerning drag coefficient are also
mensions of the length behind the Ahmed body with α ¼ 35 , and the compared with the values of Graysmith et al. (1994) and Ahmed et al.
results are shown in Tables 1 and 2, where ΔcD ¼ cDi-cDi-1 and ΔcL ¼ (1984) obtained for stationary ground (SG).
cLi-cLi-1 are the changes in drag and lift due to variations of lb and mesh As shown, current values of the lift coefficient estimate closely the
size on the surfaces of body, which is the most influential parameter. experiments of Strachan for angle up to α ¼ 25 and the experiments of
Sumantran and Sovran (1996) point out that CFD simulations must be Graysmith for larger angles. The maximum relative deviation (δ) for all
able to discriminate a change in drag as small as ΔcD ¼ 0.002, as in the experiments is smaller than 10%, and they are due to the presence of
case of experiments in wind tunnels. In order to fulfil this requirement, sting in the experimental setups, which induces errors as proven by
and make the variation of cD independent by changes of the Hetherington and Sims-Williams (2006).
4
A. Huminic, G. Huminic Journal of Wind Engineering & Industrial Aerodynamics 205 (2020) 104300
Fig. 7. Variation of vz/v∞ on (yOz) plane, (x/l) ¼ 0, (z/l) ¼ 0.26 and α ¼ 25 , Fig. 8. 3D maps of cL ðld ; ϕd Þ for plane (P), circular (C) and elliptical
adapted from Strachan et al. (2007). (E) diffusers.
δ ¼ 100cexp-cCFD/cexp %. (3)
In the first stage, the impact of plane diffuser was studied in order to where ai are the coefficients of the polynomial fitting function. In addi-
have a reference. Then, the effect of the curved underbody diffusers was tion, the lift coefficient varies linearly with the angle ϕd for the plane (P)
examined. Additional simulations have been carried out for elliptical and circular (C) diffusers, and it is a second order polynomial function for
5
A. Huminic, G. Huminic Journal of Wind Engineering & Industrial Aerodynamics 205 (2020) 104300
Table 3
Coefficients of cL ðld ; ϕd Þ interpolation function.
Plane diffuser (P) Circular diffuser (C) Elliptical diffuser (E)
Table 4
Coefficients of cD ðld ; ϕd Þ interpolation function.
Fig. 10. Variation of cD ðϕd Þ for ld ¼ 0:1 l
Plane diffuser (P) Circular diffuser (C) Elliptical diffuser (E)
X
n
cL ðϕd Þ ¼ aj ϕjd ; (6)
j¼0
X
3
cL ðld ; ϕd Þ ¼ aij lid ϕjd ; ði þ j 3Þ; (7) Fig. 11. Variation of cD ðϕd Þ for ld ¼ 0:2 l
i;j¼0
where the values of the coefficients aij together with RMSE (Root Mean
Square Error) of the fitting functions are shown in Table 3.
As results reveal, there is a substantial growth of the downforce
generated by underbody diffusers, as concluded also in the review of
Ehirim et al. (2018a). In addition, the lift coefficients of body with curved
diffusers are significantly lower than that of body with plane diffusers.
Lift decreases continuously with both diffuser angle and length and
reaches the following minimum values for studied configurations: cLðPÞ ¼
0:645, cLðCÞ ¼ 0:814, and cLðEÞ ¼ 0:941. The average relative
decreasing of lift for all configurations is ΔcLðCÞ ¼ 25 % for circular
diffusers, and ΔcLðEÞ ¼ 53 % for elliptical diffuser. References for the lift
variations are the values of the plane diffuser
cLðC; EÞ cLðPÞ
ΔcLðC; EÞ ¼ 100 ½%: (8)
cLðPÞ
Concerning drag, the results show a more complex variation with the
parameters of diffusers. Thus, both cD ðld Þ and cD ðϕd Þ are third order
polynomial functions for all studied configurations
X
3
cD ðld Þ ¼ bi lid ; (9) Fig. 12. Variation of cD ðϕd Þ for ld ¼ 0:3 l
i¼0
X
3 and the interpolation function of cD ðld ; ϕd Þ is expressed similarly as in
cD ðϕd Þ ¼ bj ϕjd ; (10) Eq. (7)
j¼0
6
A. Huminic, G. Huminic Journal of Wind Engineering & Industrial Aerodynamics 205 (2020) 104300
7
A. Huminic, G. Huminic Journal of Wind Engineering & Industrial Aerodynamics 205 (2020) 104300
Table 5 an inflexion point. The pressure coefficient reaches the minimum value
Values of cD min ðld ; ϕd Þ more quickly for circular diffuser due to its faster increase of the curva-
cD min ld =l ½ ϕd ½ ture dz=dx.
The second acceleration of the air flow through the curved diffusers is
Plane diffuser 0.237 0.4 6
Circular diffuser 0.247 0.4 4 due to the natural tendency of fluids to adhere to surfaces of bodies on
Elliptical diffuser 0.251 0.3 2 which they flow. Thus, following the curvature of diffusers, the air ve-
locity increases mainly due to occurrence and development of the vertical
component vz , which is negligible in the flat section of the underbody.
This aspect is emphasised in Figs. 17 and 18, which show the profiles of
dimensionless velocity components vz =v∞ and vx =v∞ in five sections
along the diffusers, from inlet (x=l ¼ 0:8) to exit (x=l ¼ 1).
For small values of curvature, the vertical component growths as long
as the curvature increases. Because the elliptical diffuser has a significant
change of the curvature only at its end, it has a larger growth of vz . In
addition, as depicted in detail of Fig. 16, there is a third sudden accel-
eration of the flow around the edge at the underbody end for plane and
circular diffusers, which is emphasised also by vz =v∞ variation for x=l ¼
1:0. No acceleration of the air happens at the end of the elliptical diffuser
because its surface is tangent to the vertical end surface of the body.
Concerning horizontal component, the velocity profiles vx =v∞ are
similar at the inlet of diffuser for all configurations. After a slight growth
of vx for curved underbodies, it follows a continuous diminution, which
happens earlier for circular layout. For plane diffuser, vx only decreases
along the body. At the end of the body, vx =v∞ tends again to a similar
Fig. 16. Variation of cp in the symmetry plane for ϕd ¼ 6 and ld ¼ 0:2 l profile for all diffuser. Thus, due to the velocity growth, the depression
increases under body, which also leads to a larger downforce generated
small angles, the shape of diffuser has a smaller influence. by the curved diffusers.
The combinations of parameters for which each diffuser has a mini-
mum drag are depicted in Fig. 15, which represents the projection on 4. Conclusions
horizontal plane (ld ; ϕd ) of cD maps in Fig. 9. As shown, a small strip area
corresponding to cD min of the circular diffusers intercalates between the In this paper, a bluff body fitted with circular and the elliptical un-
sections of elliptical and plane diffusers. The values of minimum drag for derbody diffusers is studied, the angle and the length of diffuser being the
all studied cases are in Table 5. parameters which are systematically varied within ranges relevant for a
Pressure and velocity variations in the symmetry plane were also hatchback passenger car. The results are compared with those obtained
studied to gain insight of phenomena that lead to improvement of for the plane diffuser, and they reveal a significant reduction of the lift
aerodynamics for the bodies fitted with circular and elliptical diffusers. coefficients of body with curved diffusers due to the pair of counter-
As shown in Fig. 16, for ϕd ¼ 6 ld ¼ 0:2 l, there are two stages of rotating vortices induced on the sides of the diffusers, which contribute
pressure recovery beneath the body for curved diffusers, as for the plane to the increase of the pumping effect, and consequently to a larger
diffuser. Since there is no edge to generate a sudden growth of the air downforce, as stated by Cooper et al. (1998). The negative effect of these
velocity, as for plane diffusers, pressure variation takes place continu- vortices is the grow of drag for diffusers with large lengths and angles.
ously, and the previously mentioned cusp point of cp variation turns into The reduction of lift is also emphasised by the pressure coefficient
Fig. 17. Profiles of the dimensionless velocity vz =v∞ along the diffusers in the symmetry plane.
8
A. Huminic, G. Huminic Journal of Wind Engineering & Industrial Aerodynamics 205 (2020) 104300
Fig. 18. Profiles of the dimensionless velocity vx =v∞ along the diffusers in the symmetry plane.
variation, which takes place continuously, with no cusp point as for plane Ehirim, O.H., Knowles, K., Saddington, A., 2018a. A review of ground-effect diffuser
aerodynamics. J. Fluid Eng. 141, 020801, 19 pages.
diffusers. The average reduction of lift for all configurations studied is Δ
Ehirim, O.H., Knowles, K., Saddington, A.J., Finnis, M.V., 2018b. Aerodynamics of a
cLðCÞ ¼ 25 % for circular diffusers and ΔcL(E) ¼ 53% for elliptical convex bump on a ground-effect diffuser. J. Fluid Eng. 140, 091102, 11 pages.
diffuser. Fu, C., Uddin, M., Robinson, C.A., 2018. Turbulence modeling effects on the CFD
predictions of flow over a NASCAR Gen 6 racecar. J. Wind Eng. Ind. Aerod. 176,
Based on the values of aerodynamic coefficients, polynomial inter-
98–111.
polation functions c(ld, ϕd) are established for both lift and drag. 3D Grandemange, M., Cadot, O., Courbois, A., Herbert, V., Ricot, D., Ruiz, T., Vigneron, R.,
maps show the lift decreases continuously with both diffuser angle and 2015. A study of wake effects on the drag of Ahmed’ s squareback model at the
industrial scale. J. Wind Eng. Ind. Aerod. 145, 282–291.
length for studied parameters. Concerning drag, the results show a more
Graysmith, J.L., Baxendale, A.J., Howell, J.P., Haynes, T., 1994. Comparisons between
complex variation with the parameters of diffusers. Thus, cD decreases in CFD and Experimental Results for the Ahmed Reference Model. RAeS conference on
the first stage and reaches a minimum value, and then it starts to rise, vehicle aerodynamics, Loughborough, pp. 30.1–30.11.
significantly swift for the curved diffuser. For the elliptical setup, cD even Guilmineau, E., 2008. Computational study of flow around a simplified car body. J. Wind
Eng. Ind. Aerod. 96, 1207–1217.
exceed the drag of the body without diffuser if ϕd>4 and ld > 0:1 l, and Hetherington, B., Sims-Williams, D., 2006. Support strut interference effects on passenger
the gain on downforce generated becomes less relevant. The curved and racing car wind tunnel models. SAE Technical Paper No. 2006-01-0565.
diffusers generate smaller values of drag than the plane diffuser for small Hucho, W.-H., Sovran, G., 1993. Aerodynamics of road vehicles. Annu. Rev. Fluid Mech.
25, 485–537.
and moderate angles and short lengths of diffusers. Huminic, A., Huminic, G., Soica, A., 2012. Study of aerodynamics for a simplified car
Hence, it is concluded that short diffusers with curved section works model with the underbody shaped as a Venturi nozzle. Int. J. Vehicle Des. 58, 15–32.
better from aerodynamic point of view. In terms of vehicles design, short Huminic, A., Huminic, G., Carauleanu, M., Ciolofan, C., 2016. Comparative study on the
performances of aerodynamic devices used in decreasing of the automobiles lift force.
diffusers have also the advantage of small interferences with other un- In: CONAT 2016 International Congress of Automotive and Transport Engineering,
derbody components of cars, as rear wheelhouses and rear silencer. In pp. 48–54. https://doi.org/10.1007/978-3-319-45447-4_5.
addition. they offer a significant larger cargo space. Huminic, A., Huminic, G., 2017. Aerodynamic study of a generic car model with wheels
and underbody diffuser. Int. J. Automot. Technol. 18, 397–404.
Kekus, P., Angland, D., 2018. Automatic wind tunnel-based optimisation of an automotive
Credit author statement underbody diffuser. AIAA Aerospace Sciences Meeting. https://doi.org/10.2514/
6.2018-0045.
Lienhart, H., Soots, C., Becker, S., 2002. Flow and turbulence structures in the wake of a
The authors contributed equally to the work.
simplified car model (Ahmed Model). In: Proceedings DGLR Fach. Symp. der AG
STAB Stuttgart University.
Declaration of competing interest Mair, W.A., 1969. Reduction of base drag by boat-tailed afterbodies in low-speed flow.
Aeronaut. Q. 20 (4), 307–320.
Menter, F.R., 1994. Two-equation eddy-viscosity turbulence models for engineering
The authors declare that they have no known competing financial applications. AIAA J. 32, 1598–1605.
interests or personal relationships that could have appeared to influence Mohottia, D., Wijesooriya, K., Dias-da-Costa, D., 2019. Comparison of Reynolds
the work reported in this paper. Averaging Navier-Stokes (RANS) turbulent models in predicting wind pressure on tall
buildings. J. Build. Eng. 21, 1–17.
Rao, A., Minelli, G., Basara, B., Krajnovic, S., 2018. On the two flow states in the wake of a
References hatchback Ahmed body. J. Wind Aerod. Ind. Aerod. 173, 262–278.
Rossitto, G., Sicot, C., Ferrand, V., Boree, J., Harambat, F., 2017. Aerodynamic
Ahmed, S., Ramm, G., Faltin, G., 1984. Some Salient Features of the Time-Averaged performances of rounded fastback vehicle. Proc. Inst. Mech. Eng. D: J. Automob. Eng.
Ground Vehicle Wake. SAE Technical Paper 840300. 231, 1211–1221.
Bonnavion, G., Cadot, O., 2019. Boat-tail effects on the global wake dynamics of a flat- Strachan, R., Knowles, K., Lawson, N., 2007. The vortex structure behind an Ahmed
backed body with rectangular section. J. Fluid Struct. 89, 61–71. reference model in the presence of a moving ground plane. J. Exp. Fluids 42,
Cheng, S.Y., Tsubokura, M., Nakashima, T., Nouzawa, T., Okada, Y., 2011. A numerical 659–669.
analysis of transient flow past road vehicles subjected to pitching oscillation. J. Wind Sumantran, V., Sovran, G., 1996. Vehicle aerodynamics, PT-49. SAE International.
Eng. Ind. Aerod. 99, 511–522. Urquhart, M., Sebben, S., Sterken, L., 2018. Numerical analysis of a vehicle wake with
Cooper, K., Bertenyi, T., Dutil, G., Syms, J., 1998. The aerodynamic performance of tapered rear extensions under yaw conditions. J. Wind Eng. Ind. Aerod. 179,
automotive underbody diffusers, developments in vehicle aerodynamics SAE SP- 308–318.
1318, pp. 5–36.