International Journal of Heat and Fluid Flow 98 (2022) 109052

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow

journal homepage: www.elsevier.com/locate/ijhff

Flow features of the Ahmed body at a low Reynolds number

Naseeb Ahmed Siddiqui *, Martin Agelin-Chaab
Ontario Tech University, Oshawa, Ontario, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: This study investigates a low Reynolds number flow over the Ahmed body using the unsteady IDDES simulations
Ahmed body at a low Reynold number of 1.4 × 104 based on the model height. The study is motivated by the lack of infor­
Detached eddy simulation mation in the literature on the Ahmed body for slant angles between 25◦ and 30◦ , especially for low Reynolds
Flow transition
numbers. Therefore, the present study documents for the first time the basic flow characteristics in the transition
Low Reynolds number
angle range (TAR) for slant angles from 25◦ to 30◦ using time-averaged and instantaneous flow velocities. The
results show that the width of the recirculation region on the slant surface decreases with increasing slant angle.
Additionally, the length of the recirculation region on the slant surface increases with the slant angle, such that at
30◦ , it is 95 % of the slant length compared to 76 % at 25◦ . Furthermore, the vortex identification documents the
complex system of the vortical structures and the changes due to the slant angles. Overall, the results confirm
that low Reynolds number flows can reproduce the basic flow features of the Ahmed body for slant angles from
25◦ to 30◦ and help elucidate the mechanism of flow transition at a lower cost compared to that at high Reynolds
numbers.

comprises three components at the rear end: a slant separation bubble, a

pair of counter-rotating streamwise vortices at the side edges, and a
1. Introduction
recirculation bubble in the wake (Krajnović, 2014). The interaction
between these components is linked to the aerodynamic drag forces,
The flow around bluff bodies has many industrial applications. The
which affect the overall vehicle performance (Beaudoin and Aider,
most frequent cases encountered in the transport industry are ground
2008; Hanfeng et al., 2016). For the standard Ahmed body with the
vehicles like trains, buses, and cars. The design constraint requires body
rectangular afterbody shape, the slant angle remains the prime factor
bluffness which inevitably brings complex three-dimensional flow to the
that distinguishes this model into high and low drag regimes (Haffner
afterbody (Choi et al., 2014; Hucho and Sovran, 1993). The afterbody
et al., 2020; Rao et al., 2018). Below 10◦ angle (low drag region), the
shape of ground vehicles significantly alters the forces exerted by the
flow is characterized by a large recirculation region (RR) at the after­
oncoming flow, which can sacrifice the vehicle’s stability. Additionally,
body dominated by low pressure. Between the slant angles, 12.5◦ and
the growing energy-constrained and soiling mitigation requirements
30◦ are considered a high drag regime due to a significant rise in the drag
demand a better understanding of the wake dynamics and force in­
coefficient. In this region, the flow has a three-dimensional structure
tensities around road vehicles. The aim is to develop mechanisms
(TDS) constituting a slant separation bubble, C-vortices, and wake
capable of reducing undesired effects using geometric optimization or
recirculation. At a critical angle of around 30◦ , the C-vortices add to a 50
active control devices (Grandemange, 2013; Kasai et al., 2018a). Several
% increase in the pressure drag compared to the 0◦ model (Beaudoin and
optimization tools are being used to devise unique vehicle shapes to
Aider, 2008). However, beyond the critical angle, the C-vortices lose
reduce drag. Geometric optimization requires a great deal of attention to
energy, and the flow undergoes a massive separation at a 35◦ angle.
systematically study the parameters that can initiate a framework of a
Consequently, the flow converts to quasi-axisymmetric separated (QAS)
paradigm shift in the conception of the solid–fluid interface.
flow which reduces the drag coefficient substantially (Ahmed et al.,
Ahmed body by (Ahmed et al., 1984) is suggested as the most widely
1984; Kohri et al., 2014). It should be noted that in a high-drag regime
investigated generic vehicle body. Similar to the idea of critical geom­
with TDS, 64 % of the pressure drag is contributed by the rear window,
etries stated by (Morel, 1978), the Ahmed body is also considered a
30 % by the vertical base, and 6 % by the forebody (Ahmed et al., 1984;
critical geometry along with other bodies such as the slanted cylinders
Evstafyeva et al., 2017; Grandemange et al., 2013). However, while the
(Zigunov et al., 2020). The literature shows that the Ahmed body

* Corresponding author at: Ontario Tech University, Oshawa, Ontario L1G0C5, Canada.
E-mail address: [email protected] (N.A. Siddiqui).

https://doi.org/10.1016/j.ijheatfluidflow.2022.109052
Received 17 June 2022; Received in revised form 12 September 2022; Accepted 15 September 2022
Available online 11 October 2022
0142-727X/© 2022 Elsevier Inc. All rights reserved.
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

found that the transition from TDS to QAS happens due to the occa­
Nomenclature sionally enlarged vortices in the unsteady phenomena that create a
large-scale separated vortex (LSV). Another related study (Guilmineau,
TAR Transition angle range 2018) performed IDDES simulations of 25◦ , 30◦ , 32◦ , 33◦ , and 35◦ slant
RR Recirculation region angles at a high Reynolds number of 7.68 × 105 . Guilmineau (2018)
RANS Reynolds-averaged Navier-stokes equation characterized the flow features and studied the effect of the slant angle.
LES Large Eddy simulation Thus, it can be seen that in the high-drag region, the slant angle range,
WMLES Wall-modelled large eddy simulation 25◦ < α ≤ 30◦ , has not been explored in great detail at low Reynolds
DES Detached eddy simulation numbers, such as examining the parameters responsible for the transi­
DDES Delayed detached eddy simulation tion from TDS to QAS for the time-averaged flow and the detailed flow
IDDES Improved delayed detached eddy simulation structure at low Reynolds numbers. These high-drag angles will be
TDS Three-dimensional structure referred to as the transition angle range (TAR): 25◦ ≤ α ≤ 30◦ . There­
QAS Quasi-axisymmetri separated fore, a systematic study of the flow development in the TAR is needed.
RL Recirculation length Furthermore, most of the existing studies in the TAR are limited to
RR Recirculation region experiments with a few numerical studies. The early simulations were
RH Recirculation height done (Han, 1989) around the Ahmed body. Accordingly, the k-ᶓ tur­
TKE Turbulent kinetic energy bulence model under-predicted the base pressure when used with
incompressible turbulence models. Later, the flow structure over the
Ahmed body was analyzed in detail by (Makowski and Kim, 2000) using
the Reynolds-Averaged Navier-Stoked (RANS) equations. More recently,
(Guilmineau, 2008) noted that the RANS turbulence models accurately
Table 1 predict the flow behavior around the 35◦ Ahmed body due to its de­
Summary of scaled model low Reynolds number studies over the Ahmed body. tached separation; however, the RANS turbulence model could not
Method Reh Title Slant Model capture the three-dimensional structure associated with the 25◦ Ahmed
angle scale body. The best numerical approach to capture the unsteady flow char­
1 Exp. 8300 (Spohn and Gillieron, 25◦ 1/3.5. acteristic is a direct numerical simulation (DNS), but it is still too
2002) expensive to employ for industrial applications (Hinterbergeret al.,
2 Num. 8275 (Bruneau et al., 2008) 0◦
1/0.288 2004). Large Eddy Simulation (LES) has produced results comparable
3 Num. 8275 (Bruneau et al., 2010) 0◦ 1/0.288
with experiments for front end and afterbody separation (Krajnović,
4 Num. 8322 (Minguez et al., 2009) 25◦ 1
5 Exp. 340, (Grandemange et al., 0◦ 1/11.0. 2014; Minguez et al., 2008a, 2009). It was also found that even after
410 2012a,b) reducing the Reynolds number to around ReH = 8,322, the primary flow
6 Exp. + 14,800 (Tunay et al., 2013) 25◦ 1/4. topology remains the same. However, it also comes with a substantial
Num. numerical cost for the TDS of industrial flows under a high Reynolds
7 Exp. 14,800 (Tunay et al., 2014) 25 , 30 ,
◦ ◦
1/4
35◦
number (Delassaux et al., 2021). Therefore, a hybrid model combining
8 Num. 10,000 (Podvin et al., 2020) 0 1 RANS and LES, as proposed by Spalart (1997), called Detached Eddy
9 Exp. 14,000 (Tunay et al., 2018) 25◦ , 35◦ 1/4. Simulation (DES), is gaining popularity as a compromised choice. Some
10 Num. 14,000 (Kang et al., 2021) 0◦ 1/5.33 fundamental studies have established that DES is an attractive tool to
11 Exp. + 14,800 (Tunay et al., 2016) 25◦ 1/4.
simulate the 25◦ Ahmed body, representing a high-drag slant angle
Num.
12 Exp. 17,000 (Cadot et al., 2015) 0◦ 1/1/4. (Serre et al., 2013). Moreover, Shur et al. (2008) further developed the
13 Exp. 17,000 (Essel et al., 2020) 0◦ , 25◦ , 1/5.3 DES models by combining SST k-ω and Delayed Detached Eddy Simu­
35◦ lation into the so-called Improved-Delayed Detached Eddy Simulation
14 Exp. 26,000 (Venning et al., 2022) 25◦ 1/4. (IDDES). This particular model has shown superiority in capturing the
15 Exp. 26,000 (Venning et al., 2015a) 25◦ 1/4.
Num. 14,000 Present 25◦ , 26◦ , 1/4.
flow field in both the TDS and QAS regions (Delassaux et al., 2021).
27◦ , Hence the IDDES method is employed for the current investigation.
28◦ , 29◦ The effect of a scaled model has been studied in detail by (Tunay
30◦ et al., 2013, 2014, 2016a, 2018). A scaled model with a Reynolds
Num. – numerical; Exp. – experimental. number of 0.14 × 105 was compared with a full-scale model at a Rey­
nolds number of 7.68 × 105 (Lienhart et al., 2002). It was concluded that
time-averaged flow structures are well understood in this region, the the wake parameters compared well with the full-scale Ahmed body at
wake dynamics are significantly scattered. the high Reynolds number. Similarly, investigations by (Gosse et al.,
According to the review by (Zhou and Zhang, 2021), most experi­ 2006; Minguez et al., 2008b; Spohn and Gillieron, 2002; Vino et al.,
mental and numerical studies over the Ahmed body are concentrated on 2005) also suggest that with minor differences such as velocity magni­
the slant angle of 0◦ , 25◦ , and 35◦ only to grasp time-averaged and time- tude and recirculation size, the flow structure is close to that of a high
dependent flow structures. Such an approach restricts a comprehensive Reynolds number. Hence, the time-averaged large-scale structure
study of fastback vehicles and the lack of control of drag forces, espe­ (Venning et al., 2022) and transient nature of the flow (Avadiar et al.,
cially in high-drag slant angles. However, (Sims-Williams and Duncan, 2019) compare well with the full-scale high-Reynolds number. Further,
2003; Sims-Williams and Dominy, 1998) studied the slant angles 27.5◦ Avadiar et al. (2019) studied the effect of a low Reynolds number over
and 30◦ and pointed out that the proximity of the slant angle with the the DrivAer model and found that though the wake structures are close
critical angle is more crucial than the absolute angle since the wake to those of a high Reynolds number (≤ 0.2 × 106 ), the drag coefficients
structure characteristic frequency is sensitive to the slant angle. Still, are unlikely to be the same. Instead, the scaled model permits a
they did not discuss the transition mechanism from high to low drag, comprehensive understanding of the time-mean and transient behavior.
which was later carried out by (Kasai et al., 2018a, Kasai et al., 2018b; Nonetheless, the relation between the Reynolds number and the drag
Kobayashi et al., 2016; Kohri et al., 2014, Kohri et al., 2016). In those coefficient is not well documented at the low Reynolds numbers. Ac­
studies, a variety of slant angles, ranging from 0◦ to 25◦ , as well as 27.5 ,
◦
cording to (Meile et al., 2012), the drag coefficient shows a 13 % in­
29 , 29.5 , 30 , 32.5 , and 35 have been studied. In their study, they
◦ ◦ ◦ ◦ ◦
crease from a Reynolds number of 0.7 × 106 to2.7 × 106 and then

2
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

becomes almost constant. (Thacker et al., 2012) indicated 5 % increase 2. Methodology

in drag from 0.5 × 106 to 4 × 106 . However, for the low Reynolds
number at 7.31 × 104 (Kohri et al., 2014) experimentally found around 2.1. Governing equations and turbulence modeling
24 % in drag variation at the 25◦ slant angle compared to the experi­
ments of (Ahmed et al., 1984) conducted at the height-based Reynolds The numerical simulation is carried out using the Ansys Fluent
number of 1.14 × 106 at the same slant angle. This drag discrepancy commercial software. The simulation employs the transient three-
between high and low-order Reynolds numbers is caused by increased dimensional continuity and momentum equations based on the finite
contribution from the skin friction while the pressure drag count re­ volume method. The algebraic equations found through the discretiza­
mains almost constant (Kohri et al., 2014). Consequently, Avadiar et al. tion were solved using bounded second-order implicit time. The gov­
(2019) concluded, based on the low Reynolds number study on a Driv­ erning equations are discretized in time and space. The general form of
eAer vehicle, that drag discrepancy at low order of Reynolds number the Equation can be written as:
exists for the Ahmed body. The LES simulation by Minguez et al. (2012) ∂
∫ ∮ ∮ ∫
and Tunay et al. (2016b) over the 25◦ Ahmed body also confirmed that ρφdV + n.(ρφu)dA = n.(Γφ ∇φ )dA + Sφ dV (1)
∂t CV
the Reynolds number has no significant effect on the flow topology;
A A CV

nonetheless, it affects the magnitude of the integral flow parameters, Here ρ, u, n, Γφ , ∇φ corresponds to density, velocity, unit vector
such as the drag and lift forces. normal to the surface, diffusion coefficient, and gradient operator,
Table 1 summarizes existing studies at low Reynolds numbers. As respectively (Kang et al., 2021). A bounded central differencing (BCD)
shown, the studies by (Cadot et al., 2015; Essel et al., 2020; Kang et al., scheme approximates the convective fluxes, consisting of a blend of a
2021; Tunay et al., 2013) have comparable Reynolds numbers to the pure central difference scheme and a second-order upwind scheme.
present study. On the other hand, the studies by (Grandemange et al., Pressure and velocity are coupled using the SIMPLEC algorithm, and a
2012a,b) have much lower Reynolds numbers. However, the flow fea­ second-order implicit scheme is used to achieve the temporal resolution,
tures found over the Ahmed body at low Reynolds numbers have been assuring a stable numerical scheme (Delassaux et al., 2021).
observed on real vehicles with significantly higher Reynolds numbers. The IDDES, which is a combination of Delayed Detached Eddy
For example, the bistability at low Reynolds numbers over the square Simulation (DDES) and Wall Modeled Large Eddy simulation, is
back Ahmed body is also found on two full-scale vehicles (Bonnavion employed to perform the current investigation due to its superior pre­
et al., 2017; Grandemange et al., 2013). Thus, it motivates further diction quality (Delassaux et al., 2021; Xiao et al., 2015). IDDES pre­
research to understand the flow structures at low Reynolds numbers. cludes an undue reduction of the Reynolds stresses generally found in
As a result of the limited number of studies conducted at low Rey­ the vicinity of the RANS- LES interface, known as the log-layer mismatch
nolds numbers, the time-averaged and unsteady flow characteristics of (He et al., 2021). For RANS simulations, the TKE equation can be written
the Ahmed body are not fully understood. For example, Minguez et al. as (Chen et al., 2022; Xiao et al., 2015):
(2008a) and Serre et al. (2013) used LES simulations to capture all the ( ) [( ]
∂(ρk) ∂ ρuj k ∂ μ ∂k σk1.5
principal flow features. The drag coefficient, however, was over­ + = μ + t) + τij Sij − (2)
∂t ∂xj ∂xj ∂k ∂xj LRANS
estimated by 40 % compared to the experiments at a high Reynolds
number (Serre et al. 2013). The presence of an SSB was not detected at where the time, TKE, density, velocity, molecular viscosity, turbulent
the Reynolds number of 2 × 106 based on unsteady simulations (Rao viscosity, tensor of stress, and mean strain rate is represented by t, k, ρ,
et al., 2018). The QAS was found without the SSB, and this was attrib­ uj , μ, μt , τij andSij , respectively. The LRANS is the turbulent length scale for
uted to low Reynolds number effects (Corallo et al., 2015; Kohri et al.,
RANS. However, in the IDDES, LRANS length scale is replaced by LIDDES
2016; Tunay et al., 2014; Venning et al., 2015b). The SSB did not appear
which is the IDDES turbulent length scale (Xiao et al., 2015), and is
in any of these studies, contrary to some previous findings (Ahmed et al.,
written as:
1984; Rossitto, 2016; Thacker et al., 2012). As mentioned earlier, the
existence of the SSB was confirmed by Minguez et al. (2008) and Serre Ấ
( Ấ )
et al. (2013) based on full-body experiments at high Reynolds numbers.
Ấ
LIDDES = f d (1+f e )LRANS + 1 − f d LLES (3)
Furthermore, there are only two studies (Tunay et al., 2014; Tunay et al.,
2016) that investigated the Strouhal number at low Reynolds numbers.
Ấ
In addition, model decomposition methods such as proper orthogonal where f d is blending function,fe is an elevating function, LRANS is the
decomposition (POD) are rarely applied to study wake flow features at length scale for RANS and LLES is for LES and are defined as:
low Reynolds numbers, which can provide important flow information.
Thus, it is necessary to document the characteristics of the Ahmed body k1/2
LLES = CDES ΔandLRANS = (4)
flows at low Reynolds numbers. β* ω
Ấ

It is clear from the foregoing sections that some basic flow features
have been reported in bits and pieces in some existing studies; however, Here β* = 0.09 is a constant in the SST K-ω, Δ = min[max{Cw Δmax ,
there is no comprehensive study on the TAR at the low Reynolds num­ Cw d, Δmin }, Δmax ] is the sub grid length-scale between Δmin = min{Δx, Δy,
ber. This study aims to document the basic flow characteristics in the Δz} and Δmax = max{Δx, Δy, Δz}. Cw is the empirical constant, d is the
TAR for a variety of slant angles from 25◦ to 30◦ at a low Reynolds Ấ
nearest wall distance. The blending function is defined as f d = max{
number. This study is the first of its kind to document these features in a
(1 − f dt ), f B } where f B is empirical blending function. When f e is equal to
comprehensive and consistent manner for a low Reynolds number and
zero, Equation (3) can be written as:
over the range of slant angles considered. The study hopes to provide
some insight into the relationship between the effect of slant angles on Ấ
( Ấ )
the size of flow recirculation which can help understand the flow Ấ
LIDDES = LDDES = f d LRANS + 1 − f d LLES (5)
transition.
The remainder of the paper is structured as follows: The model ge­
Ấ
ometry, domain, numerical method, and mesh validation are discussed whereas, when f e is higher than zero and f d is equal to f B , Equation (3)
in Section 2. In Section 3, the results are discussed based on time- becomes:
averaged and instantaneous flow features. Finally, Section 4 provides
concluding remarks on the study with further recommendations. LIDDES = LWMLES = f B (1 + f e )LRANS + (1 − f B )LLES (6)

3
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 1. Basic dimensions of the 1/4 scale Ahmed body, where h = 72 mm is the height of the model, α is slant angle, and other dimensions are non-dimensionalized
with the height.

Fig. 2. Flow domain of the simulation.

Detailed discussions about the equations and coefficients can be residuals of continuity, momentum, TKE, and scalar dissipation rate fall
found in (Chen et al., 2022; Shur et al., 2008; Xiao et al., 2015). below 10-4, which is consistent with the study by (He et al., 2021).
For all the simulations carried out, the time step is set to 5 × 10− 4
that ensures a Courant number of less than 1.5 (Kang et al., 2021). The 2.2. Geometry and boundary conditions
total simulation flow time is t* = tUin /h = 67, and the flow statistics are
collected over the lastt* = 58 with ten iterations/time-step because the The Ahmed body used in this investigation is an ¼ scaled-down
lift coefficient was less fluctuating while the drag coefficient became version of the standard Ahmed body (Ahmed et al., 1984) with a
stable. Therefore, the time averaging of the last 1.25 s is reliable to avoid height of h = 72 mm. The overall dimensions of the model are shown in
contamination The solution is considered to have converged when Fig. 1., Fig. 2.

4
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 3. Computational mesh (a) Front view of the mesh at symmetry (Y = 0), (b) Enlarged view with refinement boxes, (c) enlarged view at the front-end, (d)
Enlarged view at the slanted end, and (e) Isometric view.

flow which was similar to the low-drag regime Ahmed body found above
Table 2
the 30◦ slant angle. On the other hand, other LES simulations (Minguez
Relevant length scale normalized by the model height.
et al., 2008; Minguez et al., 2012) captured the flow features of the 25◦
L/h λT /h ηk /h Ahmed body similar to that of the present study. Therefore, the angles
3.625 4.52 × 10− 2
8.86 × 10− 4 studied in the present study are still not understood and point to the
need for further studies.
The computational domain follows the ERCOFTAC workshop on
Table 3 refined turbulence modeling (Manceau and Bonnet, 2002), where it
Mesh sensitivity analysis summary. extends 7.250 h in front of the model, 18.125 h in the back, with a width
of 6.493 h and height of 4.861 h, as shown in Fig. 3. The blockage ratio of
Grid Cellcount/106 Cd Cl
such a domain equals 4.28 %, which plays an essential role in aero­
I 3.6 0.449 0.356 dynamic drag prediction (Guilmineau et al., 2018; Keogh et al., 2016).
II 7.6 0.448 0.359
The Reynolds number used is 1.4 × 104 based on the model height
III 13.5 0.448 0.358
corresponding to a freestream velocity of U∞ = 3.1m/s. The constant
pressure outlet is applied at the outlet, and the left, right, and top sur­
faces are set as the symmetry plane. A no-slip boundary condition was
Table 4
employed for the model, and the road is considered non-moving.
Comparison of the drag coefficient.
Reference C¯d Reynolds number Method

(Ahmed et al., 1984) 0.285 1.38 × 106 Experimental 2.3. Meshing and validation
(Meile et al., 2012) 0.290 2.78 × 106 Experimental
(Thacker et al., 2012) 0.405 1 × 106 Experimental Estimating length scales is essential to capture the Ahmed body wake
(Serre et al., 2013) 0.431 7.68 × 105 LES correctly. Since the mesh grid size depends on these length scales. In
(Guilmineau et al., 2018) 0.380 7.68 × 105 IDDES Table 2, the model length to height (L/h) ratio, Taylor (λT ) and the
(Kohri et al., 2014) 0.365 7.31 × 104 Experiment
Kolmogorov (ηk ) length scales associated with the smallest turbulent
Present study 0.448 1.4 × 104 IDDES
length normalized by the model height are given for the 25◦ Ahmed
body (Howard et al., 2002; Tunay et al., 2013).
A total of six different slant angles, namely, 25◦ , 26◦ , 27◦ , 28◦ , 29◦ ,
The simulation accuracy is confirmed by performing mesh sensitivity
and 30◦ , are investigated in the standard Ahmed body. The selection
and time-averaged velocity validation using the 25◦ Ahmed body slant
represents the TAR specified in the current paper, providing a compre­
angle. A structured polyhedral mesh was applied in the simulation. It has
hensive database of the flow structure. This is important since there is no
been proven in several studies that polyhedral mesh requires less
comprehensive study of the slant angle between 25◦ and 30◦ at a low
number of iterations for simulation convergence. The quality of the
Reynolds number, and also, the effect of the slant angle on the di­
mesh is higher with a reduced element number which significantly
mensions of the recirculation region as well as the flow transition
minimizes the cost and time associated with the simulation (Corallo
mechanism is not understood. The closest study (Guilmineau, 2018)
et al., 2015; Rao et al., 2018; Sosnowski, 2018). Along with the domain,
performed IDDES simulations of 25◦ , 30◦ , 32◦ , 33◦ , and 35◦ at a high
two different refinement boxes around the model accurately capture the
Reynolds number of 7.68 × 105 . The present study complements the
length scale mentioned in Table 1. The grid resolution criteria for the
work of Guilmineau (2018) in terms of the recirculation lengths over the
smallest cell size for meshingλT /Δ > 1 was achieved (Guilmineau et al.,
slant surface and the center points of the wake recirculation bubbles.
2018; Kang et al., 2021; Minguez et al., 2008), where Δ is the sub-grid
However, Guilmineau (2018) did not investigate the slant angles be­
length scale mentioned in Eqs. (4) (Shur et al., 2008). Consequently,
tween 25◦◦ and 30◦ . Furthermore, Guilmineau (2018) found the flow
as specified in Table 3, three different meshes are applied for the grid
transition above 32◦ slant angle, which is not consistent with the
independence study. To capture the boundary layer over the model,
experimental studies both at high and low Reynolds numbers (Ahmed
inflation is used with a first layer height of 8.4 × 10− 6 with 30 layers.
et al., 1984; Tunay et al., 2014) that show the transition at a 30◦ angle.
Fig. 3 shows the mesh used in the study.
Furthermore, the 28◦ slant angle has been investigated using LES at
It can be seen that Mesh II and III provide reasonably close drag lift
4.29 × 106 (Howard and Pourquie, 2011). They found a fully detached
along with mean velocities. Furthermore, the y+ < 1 is achieved for both

5
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 4. (a) Streamwise velocity streamlines at the symmetry plane where RR = Recirculation region. (b), (c) and (d) are normalized velocity profile validation with
the experiment of (Lienhart et al., 2002). Where (b) Streamwise (c) spanwise, and (d) wall-normal direction.

Fig. 5. Time-averaged streamwise velocity streamlines in X-Z planet at Y = 0. (a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

6
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 6. Isosurface of the mean velocity at Isovalue of 0, where (a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles. Where RL: Recirculation length, RH:
Recirculation height, RW: Recirculation width. RW1 is measured at X = − 0.28 and RW2 at X = − 0.35. Isovalue 0 is taken to capture various RR’s.

Mesh II, and Mesh III, which confirms the grid resolution sufficiently et al., 2012) at a high Reynolds number.
captures the viscous sublayer over the walls. Therefore, Mesh II is There are existing studies (Corallo et al., 2015; Kohri et al., 2016;
selected for the simulation for reduced computational time. Tunay et al., 2014; Venning et al., 2015b) that were unable to capture
The mesh sensitivity analysis and validation with the existing liter­ the RR2 at a low Reynolds number at the slant surface. The non-
ature have been done using the time-averaged velocity profile and the existence of RR2 was attributed to the low Reynolds number effect.
mean drag coefficients. In Table 4, the drag coefficient is compared with However, (Rao et al., 2018) also found the QAS without RR2 at the high
the current data. However, as Avadiar et al. (2019) stated, the scaled- Reynolds number of 2 × 106 using the unsteady simulation. Therefore,
down model drag coefficient is unlikely to match the full-scale high the absence of RR2 cannot be attributed to the low Reynolds number
Reynolds number. effect. Since the RR2 has been proven by (Minguez et al., 2008b; Serre
Based on Table 4, the drag coefficient is scattered even at a high et al., 2013), using LES at a low Reynolds number is consistent with a
Reynolds number due to various reasons. An accurate LES simulation at full-scale Ahmed body experiment at a high Reynolds number. Hence,
a low Reynolds number captured all the major flow features. (Minguez the flow topology of the current simulation generally agrees with the
et al., 2008b; Serre et al., 2013). Nonetheless, their drag coefficient is experimental and LES data. Therefore, the IDDES of the present study
overestimated by 40 % compared to the experimental data (Ahmed capture the flow field accurately and are hence considered validated.
et al., 1984). However, it was found that although the time-averaged Furthermore, the time-averaged velocity profile is validated against the
and turbulence quantities matched the experiment well, the drag coef­ experimental data of (Lienhart et al., 2002), as shown in Fig. 4. It pre­
ficient did not. (Serre et al., 2013) argued that the drag variation at the sents the streamwise, spanwise, and wall-normal velocity profiles.
low Reynolds number is due to the large RR at the front end, specifically
for the low Reynolds number. This front-end recirculation is also re­ 3. Results & discussion
ported by (Spohn and Gillieron, 2002), as shown in Fig. 4 as Recircu­
lation Region 1 (RR1), along with three other primary separation The flow characteristics are investigated using both time-averaged
bubbles: Recirculation Region 2 (RR2), Recirculation Region 3 (RR3) and time-depended flow variables. The streamwise, spanwise, and
and Recirculation Region 4 (RR4). These RRs are consistent with the wall-normal coordinates are normalized with the model height h:
experimental studies by (Ahmed et al., 1984; Rossitto, 2016; Thacker

7
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 7. Variation in the center point of (a) RR1 (b) RR2, and (c) RR3 in X and Z axis direction with slant angle. Where angle is in Degrees and X and Z are normalized
with height.

x y
X = ,Y = ,Z =
z
(7) Furthermore, the rear end of the Ahmed body is subject to pressure drag
h h h caused by the separation bubbles along the slant, the C-vortices by the
The velocity component in the x, y, and z-direction is defined as: side edges, and the RR in the wakes (Ahmed et al., 1984; Yu and Bingfu,
2021). These flow features are the focus of our analysis and discussion.
u v w
Vx = , Vy = , Vz = (8)
U∞ U∞ U∞
3.1.1. Flow separation and recirculation regions
where U∞ is the freestream velocity. Similarly, the Q-criterion is defined The time-averaged velocity streamlines in the symmetry plane are
as: shown in Fig. 5. The geometry of the front-end causes an adverse pres­
sure gradient, resulting in an initial separation of the boundary layer
Q = Qh2 /U∞ 2 (9) from the front top roof, creating RR1 (shown in Fig. 5(a)). The nature of
the flow around the vehicle’s front end is not exhaustively studied. A few
Q is the second invariant of the velocity tensor. Finally, the pressure
studies are: (Aleyasin et al., 2021; Krajnović and Davidson, 2005). Since
coefficient (Cp) is defined as:
most of the studies at the high and low Reynolds number only concen­
Cp =
Pi − P∞
(10) trate on the rear slant surface flow, and data at the front-end separation
1/2ρU2∞ is scarce. The RR1 can be seen at all the slant angles in Fig. 5. After the
creation of RR1 at the front-end, the flow travels towards the rear end
Here, Pi is the pressure at the surface, p∞ is the static reference
and separates at the upper edge of the slant surface. It reattaches before
pressure, ρ is the medium density. Note that all the data reported in the
the rear end of the slant surface, creating a RR2 that can be visualized at
tables have an accuracy of ± 0.5%in the normalized form.
all the slant angles. Following this, the flow separates again at the rear
end of the slant surface and forms RR3 and RR4. The RR4 begins to
3.1. Base flow
weaken after a 28◦ slant angle.
The isosurface of the time-averaged velocity with an isovalue of zero
According to Yu and Bingfu 2021, the pressure drag over the 25◦
to capture the various RRs is shown in Fig. 6. The RR1 is defined as RL1,
Ahmed body is primarily generated by the slant surface (64 %) and the
RH1 and RW1, RR2 as RL2, RH2 and RW2, and RR3 & RR4 with RL3,
vertical base (30 %), with only 6 % coming from the front end.

8
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 8. Variation in the recirculation lengths with slant angle. (a) RL1, RH1, and RW1 of front-end RR1 while (b) Variation in RL2 with slant angle, (c) Variation in
RH2 with slant angle, and (d) Variation in RW2 with slant angle. Where RL2, RH2, and RW2 are parameters of RR2. The angle is in Degrees, and % change are in
accordance with the slant length.

RH3, and RW3, as indicated in Fig. 6. It can be visualized from Fig. 6 that by (Guilmineau, 2018), the position of RR2 in high-drag regime angles is
the shape of RR1 for all slant angles appears similar except at 25◦ and almost constant at high Reynolds numbers. Additionally, no flow tran­
28◦ , where it is comparatively larger. On the contrary, the shape of RR2 sition from TDS to QAS is reported even until the 32◦ slant angle because
increases consistently in length with the slant angles. At 30◦ , it almost the flow is reattached at the rear of the slant surface. In contrast, Tunay
touches the rear end of the slant surface. Finally, the width of RR3 looks et al. (2014) found attached flow over the slant surface at the 25◦ angle
nearly the same as the model width, and it means there is no change in and flow transitions at a 30◦ slant angle with a low Reynolds number.
the wake width with the slant angle. However, it is not possible to Secondly, In the present study, as the slant angle increased towards
visualize all the RR parameters from Fig. 6. Therefore, the detailed the critical 30◦ , the position of RR3 moved closer to the vertical base and
features of these RRs are discussed by analyzing the recirculation length caused a low-pressure region along that base, which is consistent with
(RL), height (RH), and width (RW), as shown in Fig. 6(a). It is discussed the results reported by (Rossitto et al., 2016). However, it shows minor
in subsequent paragraphs. variation in the Z direction. The location of the center point of RR3 from
In conjunction with Fig. 5 and the discussions about the RR’s, the (Guilmineau, 2018) and Tunay et al. (2014) is compared with the pre­
center points of the RRs are plotted against the slant angle in Fig. 7. The sent simulation at 25◦ and 30◦ in Fig. 7 (c). The location reported by
left axis of Fig. 7 shows the movement of the center point in the (Guilmineau, 2018) is close to the current simulation in X and Z di­
streamwise direction (X-axis), and the right axis indicates movement in rections for both angles. Nonetheless, (Tunay et al., 2014) show agree­
the wall-normal direction (Z-axis). Fig. 7 (a) displays the change in RR1 ment at 25◦ in the X direction, it displays a large offset at 30◦ . Yet, the
center point, Fig. 7(b) of RR2, and Fig. 7 (c) of RR3. differences in the Z direction are comparatively less at both angles. The
At first, the center point of RR1 in Fig. 7(a) remains constant in the reason behind this trend is a detached flow found by Tunay et al. (2014)
wall-normal direction (Z-axis) but oscillates at an average streamwise at a 30◦ angle while attached flow at 25◦ . The detached flow shifts the
(X-axis) position of X = − 2.79. In contrast, the center point of RR2 top shear layer far downstream of the body, shifting the center point of
(Fig. 7 (b) moves in the streamwise direction until 29◦ and then slightly RR3. However, the present simulation has no detached flow at a 30◦
rolls back in the upstream direction at 30◦ while oscillating at an average angle. Finally, the RR4 appears largest at 26◦ (Fig. 5) and then decreases.
wall-normal value of Z = 1.09. However, based on an IDDES simulation As discussed in Figs. 6 and 7, the parameters of RRs are plotted in

9
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 9. Variation in the RL3 and RH3 with slant angle in terms of slant length. Angle is in degrees where RL3 and RH3 are the length and height of the wake
recirculation region.

Table 5
Dimensions of the recirculation regions.
Dimensions of RR1 Dimensions of RR2 Dimensions of RR3
% change in terms of model slant length (Guilmineau, 2018) (Rossitto et al., 2016) % change in terms of (Zhang et al., 2015) Liu et al. (2021)
model slant length

% change in terms of model slant length

Angle RL1 RH1 RW1 RL2 RH2 RW2 RL3 RH3 RW3

25◦ 116 10 105 75 3.2 137.5 77 & 72 75 76 43 159 83 71

26◦ 116 6 96 78.3 3.8 118
27◦ 121 9 93.4 79.1 4.8 110.2
28◦ 121 9 101 85.6 5.8 112.8
29◦ 112 7.7 90.8 92.1 7.7 105
30◦ 120 8 93.4 94.8 9.2 103.7 87

Figs. 8 and 9. Fig. 8 (a) shows the effect of slant angle on the RL1, RH1, a low Reynolds number of 8322 and 0.7 h at a high Reynolds number
and RW1 parameters of RR1. The percentage variation at the left axis is 7.68 × 105 . Furthermore, the RL1 at a 30◦ slant angle is reported in an
with respect to the roof length. Similarly, in Fig.s 8 (b), (c), and (d), the experimental study by (Sims-Williams and Dominy, 1998) using the full
change in RL2, RH2, and RW2 are depicted as feature RR2, with the scale 30◦ Ahmed model at a Reynolds number of 1.71 × 105 . They re­
percentage variation on the left axis being relative to the slant surface ported a length of 0.2175 h, while the present IDDES simulation found a
length. value of 0.93 h. It can be seen that the front-end recirculation region
The RL1 (Fig. 8 (a)) shows that the slant angle has no significant shows large differences. However, these differences can be attributed to
effect on the parameters of RR1 at the front end. The parameters RL1, the Reynolds numbers, as reported by Minguez et al. (2009), that a low
RH1, and RW1 do not vary much with the change in slant angle. The Reynolds number provides a higher RL1 compared to a high Reynolds
RH1 varies only 0.1 % of the roof length, while RL1 & RW1 show a number.
variation of around 0.5 % of the roof length. These results are apparent Most of the existing studies only focus on and discuss the flow around
because RL2 and RW2 will be higher than RH1 due to geometrical dif­ the rear end of the Ahmed body (Yu and Bingfu, 2021). Fig. 8 (b), (c),
ferences at the front-end and rear-end. The existing literature does not and (d) show the effect of slant angle on the parameters of RR2. There
report front-end recirculation lengths for every slant angle. However, are a few important conclusions that can be drawn here. Firstly, there is
the RL1 is reported by (Krajnović and Davidson, 2005) with full-scale a consistent increase in the RL2 with the slant angle. At 25◦ , the reat­
LES simulation around the 25◦ Ahmed model at Reynolds number of 2 × tachment point of RL2 is located at 76 % of the slant length. This change
105 using different grid sizes. The RL1 decreases when the grid size is consistent with 77 % of the slant length reported from IDDES simu­
increases from coarse to fine. While a coarse mesh shows the RL1 as lation and 72 % by experiment at 7.68 × 105 Reynolds number (Guil­
0.23 h, the fine mesh is significantly reduced to 0.08 h (Krajnović and mineau, 2018). Rossitto et al. (2016) also reported the reattachment
Davidson, 2005). The present IDDES simulation found RL1 to be 0.91 h. point to be 75 % of the slant length. The RL2 further increases with slant
This length is close to that of (Minguez et al., 2009), who found 0.95 h at angle, and at the critical angle of 30◦ , the reattachment point is 95 % of

10
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 10. The distribution of pressure coefficient over the Ahmed body where (a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ . Where Cp1 is taken at X = − 0.28,Y =
0, Z = 1.17, Cp2 at X = − 0.40, Y = 0, Z = 1.03, Cp3 at X = 0, Y = 0, Z = 0.86, Cp4 at X = 0, Y = 0, Z = 0.3 and Cp5 at X = 0.48,Y = 0.65,Z = 1.05.

Fig. 11. (a) Variation in the pressure coefficient with slant angle at symmetry plane (Y = 0). Where Cp1 is taken at X = − 0.28,Z = 1.17, Cp2 at X = − 0.40,Z = 1.03,
Cp3 at X = 0,Z = 0.86, Cp4 at X = 0,Z = 0.3 and Cp5 at X = 0.48,Y = 0.65,Z = 1.05. (b) % Change between Cp1 & Cp2 denoted as %Cp1-2, Cp2 & Cp3 denoted as %
Cp2-3.

the slant length. This is on the verge of transition from the slant surface. to QAS but is close to transition. Secondly, along with the RL2, the RH2
However, according to (Guilmineau, 2018), the RL2 is 87 % at 30◦ and also increases with the slant angle. The RH2 increases from 25◦ with 3.2
95 % at the 32◦ angle of the slant length. Consequently, (Guilmineau, % of the slant length to 9.2 % at 30◦ . Thus, the slant angle increases both
2018) found flow transition at higher slant angles above 32◦ . The cur­ RL2 and RH2. Such an increase in the RL2 and RH2 of the RR2 is ad­
rent study at a low Reynolds number achieves a reattachment point of vantageous for transition. Thirdly, RW2 decreases with the slant angle
95 % of the slant length at a 30◦ angle; the flow does not show transition except at 28◦ , where it slightly increases. All these conclusions can also

11
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 12. Isosurface of the pressure at an Isovalue of − 2, where (a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

Fig. 13. Normalized time-averaged streamwise velocity contours at the slant surface separation located at X = − 0.35. The solid black line shows the Zero velocity
contour line to capture the RR.

be visualized in Fig. 6. The RR2 can grow in length and decrease in (Zhang et al., 2015), ~0.67 by Wang et al. (2015), ~ 0.65 by (Sellappan
height with the slant angle. Compared with 25◦ , the isosurface of the et al., 2018) using volumetric PIV, and ~ 0.55 reported by Liu et al.
velocity at 30◦ demonstrated the facts presented above. (2021). The slant angle generally decreases the RL3 and RH3, reaching a
Furthermore, Fig. 9 shows the percentage change in the RL3 and RH3 minimum of around 28◦ angle. The 28◦ slant angle has been investigated
with respect to the slant angle. The RL3 for the 25◦ is ~ 0.58, which is by (Howard & Pourquie, 2011) using LES at 4.29 × 106 . They found a
76 % of the slant length. This is close to the value of ~ 0.64 reported by fully detached flow which was similar to the post-critical angle (30◦ )

12
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 14. Normalized time-averaged streamwise velocity contours at the slant rear end located at X = 0.

Fig. 15. Normalized time-averaged streamwise velocity contours at X = 0.28, where (a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

low-drag regime Ahmed body. Yet, other LES simulations by (Minguez 2019). Accordingly, an increased RL3 and reduced RH3 help reduce the
et al., 2008; Minguez et al., 2012) captured the flow features of the 25◦ pressure drag. Furthermore, the reduction in wake width is a strong
Ahmed body similar to the present study. Thus, the 28◦ Ahmed flow indication of base pressure recovery (Barros et al., 2016) . For the TAR,
features are not completely understood. The present study as well found the RW3 does not change and implies it does not contribute to drag
the variations at the 28◦ angle, which motivates studies to critically reduction Fig. 9. Table 5 provides the summary of recirculation di­
examine the physics behind such variations. Moreover, the effect of mensions (RL, RH and RW) in terms of slant length percentage for more
length and height of the RR3 on the pressure drag is discussed by some clarity.
existing studies (Barros et al., 2016; Mariotti et al., 2015; Pavia et al.,

13
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 16. Three-dimensional Isosurface of the Q-criterion coloured by the normalized time-averaged streamwise velocity at an Isovalue of − 15, where (a) 25◦ (b) 26◦
(c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

3.1.2. Pressure The flow shows a pressure recovery between Cp1 to Cp2 and is highest at a
The pressure coefficient (Cp) over the Ahmed body is shown in 30◦ angle with a 16 % recovery.
Fig. 10, according to Equation (9). The change in the slant angle does not On the contrary, %Cp2-3 highlights the comparative pressure recov­
affect the negative Cp area for the RR1. However, the Cp of RR2 was ery between a point inside RR2 and just outside RR2. At 25◦ , Cp2 =
reduced with a slant angle. The formation of C-vortices can also be seen, − 0.58 and Cp3 = − 0.17 provided a recovery of around 70 %. Yet at 30◦
and the increase in slant angle leads to pressure recovery in the C- the Cp recovery drops to around 50 % between the points at Cp2 = − 0.63
vortices, as shown in Fig. 10(f) and 11(a). and Cp3 = − 0.31. Furthermore, the inside the C-vortices Cp5 shows
However, to further investigate the changes in the pressure inside the consistent recovery with the slant angle. At 25◦ , the Cp5 around the C-
RR’s and C-vortices Cp is plotted in Fig. 11 (a) at the five different lo­ vortices has an approximate value of Cp5 = − 0.94, and at 30◦ , Cp5 =
cations indicated in Fig. 10(a), where Cp1 is a point inside RR1 (as − 0.66. It shows a Cp5 recovery of almost 30 % at 30◦ compared to 25◦ .
defined in the caption for Fig. 11). Also, Cp2 & Cp3 within RR2, Cp4 shows The impact of Cp5 is emphasized in the next section. Similarly, from
the point at the vertical base, and Cp5 indicates the C-vortices (See Fig. 11(a), the change in the Cp can be seen for any particular slant angle
Fig. 10(a)). as well. At a 25◦ angle, the Cp shows continuous pressure recovery from
Fig. 11 (a) shows that Cp1 increases with the slant angle up to 27◦ , Cp1 to Cp4. All the other angles follow this trend, and a significant
and decreases at 28◦ before rising again at 29◦ . However, the difference pressure recovery is found after Cp2 for all the angles.
in the Cp1 values is not significant. Similar trends are exhibited in the All these RRs can also be observed in Fig. 12, which shows the iso­
Cp2, Cp3 and Cp4. In all these cases, a slight pressure recovery is surface of the pressure at an isovalue of Cp = − 2 to display the necessary
demonstrated at 28◦ , which is not consistent with the results of existing RRs and C-vortices. Along with the existence of low-pressure separation
studies that show a consistent rise in drag coefficient with the slant bubbles (RR1 & RR2), C-vortices coming out of the side edges are also
angle. Consequently, the Cp at the rear end is expected to also decrease captured for all the slant angles. Especially the size of the RR2 appears to
with a slant angle. The low pressure at the Cp1 is due to the flow sepa­ increase with slant angle except at the 28◦ angle. Another critical piece
ration at the front-end, followed by the slant surface separation where of information is the increased size of the C-vortices with slant angles. As
the Cp2 slightly increases compared to the front-end. However, the C-vortices grow in size, it merges with the RR2. The C-vortices
compared to the Cp1 and Cp2, a considerable pressure recovery is dis­ constrain the flow at the slant surface to spread spanwise. Consequently,
played at Cp3 and Cp4. Inside the RL2, the Cp2 = − 0.6 and just after the the increased C-vortices at a 30◦ angle merged with the RR2 and reduced
RL2, it increased to Cp2=− 0.2. Fig. 11 (b) shows the percentage change the RW2. This explanation is supported by Fig. 8(f), where RW2 de­
between the Cp values inside RR1 and RR2. The %Cp1-2 indicates the Cp creases with the slant and is further discussed in the next section.
recovery between a point inside RR1 and RR2, as shown in Fig. 11 (b).

14
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 17. The Top view of the Ahmed body Isosurface of the Q-criterion coloured by the normalized time-averaged streamwise velocity at an Isovalue of − 15, where
(a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

3.2. C-vortices Along with C-V1, there are small negative C-vortices indicated as C-V2
in Fig. 14(a). Additionally, at X = 0, a positive streamwise velocity rear
The discussion in the previous sections highlighted the effect of slant bubble can be seen, which exists outside the RR2. As the slant angle
angle on the geometric parameters of the RR’s. However, the issue of increases, C-V1 grows in size and becomes weaker. At 30◦ , C-V1 merges
transition from the slant surface, which is supposed to occur at a critical with the positive streamwise velocity rear bubble in Fig. 14(a). Along
angle of 30◦ , needs further investigation (Guilmineau, 2018; Tunay with C-V1, C-V2 also grows in size with a slant angle. However, C-V2 is
et al., 2014). In the current study, the slant angle is reported to increase dominated by negative streamwise velocities. Furthermore, the C-
the RL2, which at the 30◦ angle extends to almost 95 % of the slant vortices restrict the flow on the slant from spreading in the spanwise
length. This increase has a corresponding increase in RH2 but a reduc­ direction. This is one of the reasons found in the present paper for
tion in RW2 with the slant angle (Guilmineau, 2018). The proceeding reducing RW2 as the slant angle increases.
information helps to provide some insight into the effect of the slant In Fig. 15, time-averaged streamwise velocity contours are shown
angle on the C-vortices. Figs. 13-16 show the development of C-vortices just after the end of the slant surface inside the wake recirculation. At
at cross-section planes. this location, the C-vortices can still be seen up to a slant angle of 28◦ ,
Fig. 13 shows the time-averaged normalized streamwise velocity (x- after which they merge with the shear layer from the top. Nonetheless,
direction) contours inside the RL2 for all the slant angles. The separating the change in slant angle does not affect the wake width (RW3), which
contour line at the core shows the zero streamwise velocity indicating remains consistent for all the slant angles and implies that it has no
the RR over the slant surface. The existence of the C-vortices pair is contribution to the pressure drag. Furthermore, according to Fig. 9, the
found, and RH2 and RW2 are the parameters of RR2, as indicated in percentage change in RL3 and RH3 is relatively small, and except for
Fig. 13(a). At first, a consistent reduction in the RW2 value with 28◦ , all of them are similar. It indicates that the shift in RL2, RH2, and
increasing slant angle can be observed, with the lowest RW2 value at RW2 also depends on the slant height.
30◦ . Additionally, RH2 increases with the slant angle, reaching a
maximum of 30◦ . These two images in Fig. 13 agree with Fig. 8(e) & (f). 3.3. Vorticial structures
At 25◦ , however, the C-vortices show the highest streamwise velocity,
but this value continuously reduces with the slant angle. For example, Instantaneous vortices can help better understand the unsteady 3D
from the slant angles of 25◦ to 30◦ , the streamwise velocity decreased wake structure around the Ahmed body. A variety of methods can be
from 0.9 to 0.1, an 89 % reduction. used, including the Q-criterion based on the complex eigenvalues of the
Although the C-vortices lose energy with increasing slant angles, velocity gradient tensor, Δ-criterion which is the second invariant of the
they grow in size with downstream distance. Fig. 14 shows this more velocity gradient tensor, the swirling strength λci and the λ2 criterion
evidently at the slant rear edge. Fig. 14(a) displays the time-averaged (Kang et al., 2021). In this study, both Q and λ2 criteria are used to
streamwise velocity contour at the slant rear end X = 0 for all the analyze the vortical structures around the Ahmed body.
slant angles. The pair of C-vortices are shown with C-V1 in Fig. 14(a).

15
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 18. The underbody of the Ahmed body Isosurface of the Q-criterion coloured by the normalized time-averaged streamwise velocity at an Isovalue of − 15, where
(a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

3.3.1. Q-criterion front end travel the body length to merge with the shear layer coming
Here, rotational flow structures are visualized using isosurfaces from the top and creates interesting structures of the vortices in the wake
calculated based on the Q-criterion (Jeong and Hussain 1995). Vortex of the body.
identification in this manner uses the tensor of velocity gradient ∇u, Similarly, In Fig. 17, the top view illustrates the hairpin-like vortices
which can be broken down into two parts: Sij, which is called strain over the roof. The existence of RR1 (visualized with negative velocity)
tensor, and second Ωij that is, the antisymmetric rotational part of the can be seen due to flow separation at the front end at all angles. These
tensor. A qualitative analysis of turbulent flows is possible with the Q- vortices are joined by two small vortices (shown with a black circle in
criterion (Saeedi et al., 2019). The isovalue can also be compared to Fig. 17(a)) coming from the side edges of the front end in all the cases.
contour lines over a topographic map, showing the structure of surfaces The hairpin vortices look almost symmetric to the Y = 0 plane over the
defined by a scalar field. The mathematical Equation for this criterion roof with slant angles. Afterward, the flow separates at the slant surface
can be found in the references (Dubief and Delcayre, 2000; Gohlke et al., to form RR2, which the negative velocities can recognize at all the angles
2008; Hunt et al., 1988; Jeong and Hussain, 1995). shown in Fig. 17 (a)). Finally, the shear vortices in the wake are sym­
Figs. 16, 17, and 18 show the isosurface of the Q-criterion colored by metrically distributed in the downstream direction.
normalized time-averaged streamwise velocity for all the slant angles. Fig. 17 captures the underbody vortical structures with slant angles.
Fig. 16 shows the three-dimensional view, Fig. 17 displays the top view, The vortices due to the cylindrical wheels can be seen at the front and
and Fig. 18 exhibits the underbody vortical structures. For all the Figs, then in the middle of the underbody. The vortices around the first pair of
an isovalue of − 15 is chosen to extract the salient flow features from the wheels have more negative vortices compared to the second pair
data. downstream. These vortical structures move towards the rear end but
In Fig. 16, due to the separated shear layer, the vortices at the front bifurcate into two underbody corner vortices (UCV), shown as UCV-1
end are dominated by the Kelvin-Helmholtz (KH) instability which can and UCV-2 (Fig. 17 (a)). The UCV-1 flows inwardly towards the rear
be seen at all the slant angles (indicated in Fig. 16(a)) (Kang et al., wake structure as indicated by black arrows and merges. Consequently,
2021). As the flow reattached to the roof, hairpin-like vortices (Fig. 16 the UCV-2 remains corner vortices coming out of the underbody along
(a) - Omega vortex) emerge due to boundary layer development with the with the side corners. However, it should be noted that UCV-1 has
corner vortices along the side edges. Similarly, the emergence of vortical negative vortices compared to UCV-2, which has positive ones. Such a
structures can also be seen surrounding the side surface of the body. development displays a V shape flow (shown with arrows) under the
Especially the prominent distinguishing feature, the RR2 at the slant and body, caused by the presence of the cylindrical wheels. The vortices at
pair of counter-rotating C-vortices, is also evident in all the cases. It can the wake RR can be seen for all the slant angles and are symmetric
be seen that compared to the 25◦ , the C-vortices are dominated by less around the symmetry plane.
energetic vortices; as the slant angle increases, the angle and, at 30◦ ,
merge with the RR2. The underbody vortical structures that begin at the

16
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 19. Three-dimensional Isosurface of the λ2 -criterion coloured by the normalized time-averaged streamwise velocity at an Isovalue of − 1500, where (a) 25◦ (b)
26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

3.3.2. λ2 Criterion cylinders along with the UCV-1 and UCV-2, similar to the Q-criterion.
The λ2 criterion extracts the coherent vortical structures based on Thus, the comparison between the Q-criterion and λ2 criterion shows
regions where the second largest eigenvalue λ2 of the tensor a general behavior in a similar manner. However, λ2 criterion is more
∂U
Sik Skj +Ωik Ωkj is negative (Jeong and Hussain 2006). Here Sij = (∂∂Uxji + ∂xij ) precise in extracting the front-end separation, flow separation at the
∂Uj upper slant edge, and complex system of C-vortices development at all
/2 and Ωij = (∂∂Uxji − ∂xi )/2 are the symmetric and antisymmetric parts of angles.
the velocity gradient tensor ∂∂Uxji . Figs. 19–21 show the isometric, top view
and bottom view, respectively, based on the λ2 criterion. 4. Conclusion
The general flow structures captured by both Q and λ2 criteria are the
same which is also mentioned by (Kang et al., 2021). However, the λ2 This study investigates a low Reynolds number flow over the Ahmed
criterion provides a more accurate vortical structure (Fig. 19 (a) high­ body using the unsteady IDDES simulations at a low Reynold number of
lighted with a circle at the front-end) of the K-H instability at the front 1.4 × 104 based on the model height. The study is motivated by the lack
part, which is not fully captured by the Q-criterion. All the slant angles of information in the literature on the Ahmed body for slant angles be­
show this trend of the matured effect of the front-end separation-reat­ tween 25◦ and 30◦ , especially for low Reynolds numbers. Therefore, the
tachment phenomena. This effect is also visible on the side faces of the present study documents for the first time the basic flow characteristics
models, which show complex vortical structures associated with the in the transition angle range (TAR) for slant angles from 25◦ to 30◦ using
front-end separation (Fig. 19 and Fig. 20). Furthermore, although Q- time-averaged and instantaneous flow velocities at a low Reynolds
criterion can show the existence of C-vortices (Fig. 16) for all the slant number. The main observation from the study are summarized as
angles, yet the λ2 criterion provides the more complex development of follows:
the C-vortices over the slant surface highlighted with a circle in (Fig. 19
(a)). On the other hand, the Q-criterion indicates vortical structures 1. The width of the recirculation region on the slant surface decreases
some distance downstream of the slant surface while λ2 criterion dis­ with increasing slant angle.
plays starting with the upper slant edge. It can be seen in both the iso­ 2. The C-vortices become larger with increasing slant angle and, as a
metric (Fig. 19 highlighted with a circle at the rear end) and top view result, restrict the flow on the slant from spreading in the spanwise
(Fig. 20 highlighted with a circle) of the λ2 criterion. Finally, at the direction.
underbody (Fig. 21), λ2 criterion captures the vortical structures due to

17
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Fig. 20. The Top view of the Ahmed body Isosurface of the λ2 -criterion coloured by the normalized time-averaged streamwise velocity at an Isovalue of − 1500,
where (a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

Fig. 21. The underbody of the Ahmed body Isosurface of the λ2 -criterion coloured by the normalized time-averaged streamwise velocity at an Isovalue of − 1500,
where (a) 25◦ (b) 26◦ (c) 27◦ (d) 28◦ (e) 29◦ and (f) 30◦ slant angles.

18
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

3. The length of the recirculation region on the slant surface increases Cadot, O., Evrard, A., Pastur, L., 2015. Imperfect supercritical bifurcation in a three-
dimensional turbulent wake. Phys. Rev. E: Stat. Nonlinear Soft Matter Phys. 91,
with the slant angle, such that at 30◦ , it is 95 % of the slant length
063005.
compared to 76 % at 25◦ . Chen, G., Li, X.B., Liang, X.F., 2022. IDDES simulation of the performance and wake
4. The height of the recirculation region on the slant surface also dynamics of the wind turbines under different turbulent inflow conditions. Energy
increased from 3.2 % at 25◦ to 9.2 % of the slant length at 30◦ . 238, 121772.
Choi, H., Lee, J., Park, H., 2014. Aerodynamics of Heavy Vehicles. Annu. Rev. Fluid
5. It is found that the extension of the geometric slant height with the Mech. 46, 441–468.
slant angle increases the space over the slant surface, which is Corallo, M., Sheridan, J., Thompson, M.C., 2015. Effect of aspect ratio on the near-wake
occupied by reverse flow due to the pressure difference around the flow structure of an Ahmed body. J. Wind Eng. Ind. Aerodyn. 147, 95–103.
Delassaux, F., Mortazavi, I., Itam, E., Herbert, V., Ribes, C., 2021. Sensitivity analysis of
reattachment point. hybrid methods for the flow around the Ahmed body with application to passive
6. The vortex identification indicates that the corner vortices at the control with rounded edges. Comput. Fluids 214, 104757.
underbody separate into two, with one merging with the wake and Dubief, Y., Delcayre, F., 2000. On coherent-vortex identification in turbulence. J. Turbul.
1, N11.
the other attaching to the side edges. Essel, E., Das, S., Balachandar, R., 2020. Effects of Rear Angle on the Turbulent Wake
Flow between Two in-Line Ahmed Bodies. Atmosphere 2020, Vol. 11, Page 328 11,
The study provides systematic documentation of the flow features in 328.
Evstafyeva, O., Morgans, A.S., Dalla Longa, L., 2017. Simulation and feedback control of
terms of dimensions (such as positions, length, width or height, and the Ahmed body flow exhibiting symmetry breaking behaviour. J. Fluid Mech. 817,
pressure) which is not available in the literature. Therefore, it is hoped R2.
that the results of the current study will be used as a reference in future Gohlke, M., Beaudoin, J.F., Amielh, M., Anselmet, F., 2008. Thorough analysis of vortical
structures in the flow around a yawed bluff body. J. Turbul. 9, N15.
studies. The flow transition at 30◦ was not observed, so future work
Gosse, K., Paranthoën, P., Patte-Rouland, B., Gonzalez, M., 2006. Dispersion in the near
should extend this beyond 32◦ to examine the transition phenomena at a wake of idealized car model. Int. J. Heat Mass Transf. 49, 1747–1752.
low Reynolds. Further, the study confirms that the Ahmed body at low Grandemange, M., Cadot, O., Gohlke, M., 2012a. Reflectional symmetry breaking of the
Reynolds numbers can reproduce the basic flow features and enable separated flow over three-dimensional bluff bodies. Phys. Rev. E: Stat. Nonlinear
Soft Matter Phys. 86, 035302.
researchers to understand the transition phenomena. This can poten­ Grandemange, M., Gohlke, M., Parezanović, V., Cadot, O., 2012b. On experimental
tially reduce the costs of experiments and simulations associated with sensitivity analysis of the turbulent wake from an axisymmetric blunt trailing edge.
high Reynolds number studies. Phys. Fluids 24, 035106.
Grandemange, M., Gohlke, M., Cadot, O., 2013. Turbulent wake past a three-dimensional
blunt body. Part 1. Global modes and bi-stability. J. Fluid Mech. 722, 51–84.
CRediT authorship contribution statement Grandemange, M., 2013. Analysis and Control of Three-dimensional Turbulent Wakes:
from Axisymmetric Bodies to Road Vehicles. École Polytechnique — ENSTA
ParisTech.
Naseeb Ahmed Siddiqui: Conceptualization, Data curation, Formal Guilmineau, E., 2008. Computational study of flow around a simplified car body. J. Wind
analysis, Validation, Visualization, Writing – original draft. Martin Eng. Ind. Aerodyn. 96, 1207–1217.
Agelin-Chaab: . Guilmineau, E., Deng, G.B., Leroyer, A., Queutey, P., Visonneau, M., Wackers, J., 2018.
Assessment of hybrid RANS-LES formulations for flow simulation around the Ahmed
body. Comput. Fluids 176, 302–319.
Declaration of Competing Interest Guilmineau, E., 2018. Effects of Rear Slant Angles on the Flow Characteristics of the
Ahmed Body by IDDES Simulations. SAE Technical Papers 2018.
Haffner, Y., Boreé, J., Spohn, A., Castelain, T., 2020. Mechanics of bluff body drag
The authors declare that they have no known competing financial reduction during transient near-wake reversals. J. Fluid Mech. 894, A14.
interests or personal relationships that could have appeared to influence Han, T., 1989. Computational analysis of three-dimensional turbulent flow around a
bluff body in ground proximity. AIAA J. 27, 1213–1219.
the work reported in this paper. Hanfeng, W., Yu, Z., Chao, Z., Xuhui, H., 2016. Aerodynamic drag reduction of an Ahmed
body based on deflectors. J. Wind Eng. Ind. Aerodyn. 148, 34–44.
Data availability He, K., Minelli, G., Wang, J., Gao, G., Krajnović, S., 2021. Assessment of LES, IDDES and
RANS approaches for prediction of wakes behind notchback road vehicles. J. Wind
Eng. Ind. Aerodyn. 217, 104737.
Data will be made available on request. Hinterberger, C., García-Villalba, M., Rodi, W., 2004. Large eddy simulation of flow
around the Ahmed body. Springer, Berlin, Heidelberg, pp. 77–87.
Howard, R.J.A., Lesieur, M., Bieder, U., 2002. Structured and Non-Structured Large Eddy
Acknowledgement Simulations of the Ahmed Reference Model. Fluid Mech. Appl. 65, 185–198.
Howard, R.J.A., Pourquie, M., 2011. Large eddy simulation of an Ahmed reference
The authors would like to acknowledge the financial support from model. J. Turbul. 3, N12.
Hucho, W., Sovran, G., 1993. Aerodynamics of Road Vehicles. Annu. Rev. Fluid Mech.
the Natural Science and Engineering Research Council of Canada (Grant 25, 485–537.
#: RGPIN-2018-05369). Also, the support of computational resources Hunt, J.C.R., Wray, A.A., Moin, P., 1988. Eddies, streams, and convergence zones in
from the Shared Hierarchical Academic Research Computing Network turbulent flows.
Jeong, J., Hussain, F., 1995. On the identification of a vortex. J. Fluid Mech. 285, 69.
(SHARCNET) and Compute Canada is gratefully acknowledged.
Kang, N., Essel, E.E., Roussinova, V., Balachandar, R., 2021. Effects of approach flow
conditions on the unsteady three-dimensional wake structure of a square-back
References Ahmed body. Phys. Rev. Fluids 6, 034613.
Kasai, A., Shiratori, S., Kohri, I., Kobayashi, Y., Katoh, D., Nagano, H., Shimano, K.,
2018b. Large-Scale Separated Vortex Generated in a Wake Flow of Ahmed’s Body.
Ahmed, S.R., Ramm, G., Faltin, G., 1984. Some Salient Features of The Time-Averaged
Flow, Turbulence and Combustion 2018 102:2 102, 373–388.
Ground Vehicle Wake. SAE Transactions 840300.
Keogh, J., Barber, T., Diasinos, S., Doig, G., 2016. The aerodynamic effects on a cornering
Aleyasin, S.S., Tachie, M.F., Balachandar, R., 2021. Characteristics of flow past elongated
Ahmed body. J. Wind Eng. Ind. Aerodyn. 154, 34–46.
bluff bodies with underbody gaps due to varying inflow turbulence. Phys. Fluids 33,
Kobayashi, Y., Kohri, I., Kasai, A., Nasu, T., Katoh, D., Hashizume, Y., 2016. Numerical
125106.
Analysis on the Transitional Mechanism of the Wake Structure of the Ahmed Body.
Avadiar, T., Thompson, M.C., Sheridan, J., Burton, D., 2019. The influence of reduced
SAE Technical Papers 1592.
Reynolds number on the wake of the DrivAer estate vehicle. J. Wind Eng. Ind.
Kohri, I., Yamanashi, T., Nasu, T., Hashizume, Y., Katoh, D., 2014. Study on the Transient
Aerodyn. 188, 207–216.
Behaviour of the Vortex Structure behind Ahmed Body. SAE Int. J. Passenger Cars
Barros, D., Borée, J., Noack, B.R., Spohn, A., Ruiz, T., 2016. Bluff body drag
Mech. Syst. 7, 586–602.
manipulation using pulsed jets and Coanda effect. J. Fluid Mech. 805, 422–459.
Kohri, I., Kobayashi, Y., Kasai, A., Nasu, T., Katoh, D., Hashizume, Y., 2016.
Beaudoin, J.F., Aider, J.L., 2008. Drag and lift reduction of a 3D bluff body using flaps.
Experimental Analysis on the Transitional Mechanism of the Wake Structure of the
Exp. Fluids 44, 491–501.
Ahmed Body. SAE Int. J. Passenger Cars Mech. Syst. 9, 612–624.
Bonnavion, G., Cadot, O., Évrard, A., Herbert, V., Parpais, S., Vigneron, R., Délery, J.,
Krajnović, S., 2014. Large eddy simulation exploration of passive flow control around an
2017. On multistabilities of real car’s wake. J. Wind Eng. Ind. Aerodyn. 164, 22–33.
Ahmed body. J. Fluids Eng., Trans. ASME 136.
Bruneau, C.H., Creusé, E., Depeyras, D., Gilliéron, P., Mortazavi, I., 2010. Coupling
Krajnović, S., Davidson, L., 2005. Flow Around a Simplified Car, Part 1: Large Eddy
active and passive techniques to control the flow past the square back Ahmed body.
Simulation. J. Fluids Eng. 127, 907–918.
Comput. Fluids 39, 1875–1892.
Bruneau, C.-H., Mortazavi, I., Gilliéron, P., 2008. Passive Control Around the Two-
Dimensional Square Back Ahmed Body Using Porous. J. Fluid Eng. 130, 1–12.

19
N.A. Siddiqui and M. Agelin-Chaab International Journal of Heat and Fluid Flow 98 (2022) 109052

Lienhart, H., Stoots, C., Becker, S., 2002. Flow and Turbulence Structures in the Wake of Sims-Williams, D.B., Dominy, R.G., 1998. Experimental Investigation into Unsteadiness
a Simplified Car Model (Ahmed Modell). In: New Results in Numerical and and Instability in Passenger Car Aerodynamics. SAE Technical Papers.
Experimental Fluid Mechanics III. Springer, Berlin Heidelberg, pp. 323–330. Sims-Williams, D.B., Duncan, B.D., 2003. The Ahmed Model Unsteady Wake:
Liu, K., Zhang, B., Zhou, Y., 2021. Correlation between drag variation and rear surface Experimental and Computational Analyses. Journal of passenger car: mechanical
pressure of an Ahmed body. Exp. Fluids 62, 124. systems journal 112, 1385–1396.
Makowski, F.T., Kim, S.-E., 2000. Advances in External-Aero Simulation of Ground Sosnowski, M., 2018. The influence of computational domain discretization on CFD
Vehicles Using the Steady RANS Equations. SAE Technical Paper 0484. results concerning aerodynamics of a vehicle. J. Appl. Math. Comput. Mech. 17,
Manceau, R., Bonnet, J. (Eds.), 2002. 10th joint ERCOFTAC/IAHR/QNET-CFD Workshop 79–88.
on Refined Turbulence Modelling. France. SPALART, PR, 1997. Comments on the feasibility of LES for wings, and on a hybrid
Mariotti, A., Buresti, G., Salvetti, M.V., 2015. Connection between base drag, separating RANS/LES approach. Proceedings of first AFOSR international conference on DNS/
boundary layer characteristics and wake mean recirculation length of an LES.
axisymmetric blunt-based body. J. Fluids Struct. 55, 191–203. Spohn, A., Gillieron, P., 2002. Flow separations generated by a simplified geometry of an
Meile, W., Brenn, G., Reppenhagen, A., Lechner, B., Fuchs, A., 2012. Experiments and automotive vehicle. IUTAM Symposium. Unsteady Separated Flows, Toulouse.
numerical simulations on the aerodynamics of the Ahmed body 3, 32–39. Thacker, A., Aubrun, S., Leroy, A., Devinant, P., 2012. Effects of suppressing the 3D
Minguez, M., Pasquetti, R., Serre, E., 2008. High-order large-eddy simulation of flow separation on the rear slant on the flow structures around an Ahmed body. J. Wind
over the “Ahmed body” car model. Phys. Fluids 20, 095101. Eng. Ind. Aerodyn. 107–108, 237–243.
Minguez, M., Pasquetti, R., Serre, E., 2009. High-order LES of the flow over a simplified Tunay, T., Sahin, B., Akilli, H., 2013. Experimental and numerical studies of the flow
car model. Eur. J. Comput. Mech. 18, 627–646. around the Ahmed body. Wind Struct. 17, 515–535.
Morel, T., 1978. The Effect of Base Slant on the Flow Pattern and Drag of Three- Tunay, T., Sahin, B., Ozbolat, V., 2014. Effects of rear slant angles on the flow
Dimensional Bodies with Blunt Ends 191–226. characteristics of Ahmed body. Exp. Therm Fluid Sci. 57, 165–176.
Pavia, G., Passmore, M., Varney, M., 2019. Low-frequency wake dynamics for a square- Tunay, T., Yaniktepe, B., Sahin, B., 2016. Computational and experimental investigations
back vehicle with side trailing edge tapers. J. Wind Eng. Ind. Aerodyn. 184, of the vortical flow structures in the near wake region downstream of the Ahmed
417–435. vehicle model. J. Wind Eng. Ind. Aerodyn. 159, 48–64.
Podvin, B., Pellerin, S., Fraigneau, Y., Evrard, A., Cadot, O., 2020. Proper orthogonal Tunay, T., Firat, E., Sahin, B., 2018. Experimental investigation of the flow around a
decomposition analysis and modelling of the wake deviation behind a squareback simplified ground vehicle under effects of the steady crosswind. Int. J. Heat Fluid
Ahmed body. Phys. Rev. Fluids 5, 064612. Flow 71, 137–152.
Rao, A., Minelli, G., Basara, B., Krajnović, S., 2018. On the two flow states in the wake of Venning, J., Lo Jacono, D., Burton, D., Thompson, M., Sheridan, J., 2015a. The effect of
a hatchback Ahmed body. J. Wind Eng. Ind. Aerodyn. 173, 262–278. aspect ratio on the wake of the Ahmed body. Exp. Fluids 56, 126.
Rossitto, G., Sicot, C., Ferrand, V., Borée, J., Harambat, F., 2016. Influence of afterbody Venning, J., Lo Jacono, D., Burton, D., Thompson, M., Sheridan, J., 2015b. The effect of
rounding on the pressure distribution over a fastback vehicle. Exp. Fluids 57, 1–12. aspect ratio on the wake of the Ahmed body. Exp. Fluids 56, 1–11.
Rossitto, G., 2016. Influence of afterbody rounding on the aerodynamics of a fastback Venning, J., McQueen, T., Jacono, D.L., Burton, D., Thompson, M., Sheridan, J., 2022.
vehicle. Ecole Nationale Superieure de Mecanique et d’Aerotechnique,. Aspect ratio and the dynamic wake of the Ahmed body. Exp. Therm Fluid Sci. 130,
Saeedi, M., Tarokh, A., Derny, A., Liu1, K., 2019. On the aerodynamic performance of tail 110457.
devices for drag reduction of road heavy vehicles, in: Proceedings of The Joint Vino, G., Watkins, S., Mousley, P., Watmuff, J., Prasad, S., 2005. Flow structures in the
Canadian Society for Mechanical Engineering and CFD Society of Canada near-wake of the Ahmed model. J. Fluids Struct. 20, 673–695.
International. CSME-CFDSC Congress 2019, London, p. 8. Xiao, L., Xiao, Z., Duan, Z., Fu, S., 2015. Improved-Delayed-Detached-Eddy Simulation of
Sellappan, P., McNally, J., Alvi, F.S., 2018. Time-averaged three-dimensional flow cavity-induced transition in hypersonic boundary layer. Int. J. Heat Fluid Flow 51,
topology in the wake of a simplified car model using volumetric PIV. Experiments in 138–150.
Fluids 2018 59:8 59, 1–13. Yu, Z., Bingfu, Z., 2021. Recent Advances in Wake Dynamics and Active Drag Reduction
Serre, E., Minguez, M., Pasquetti, R., Guilmineau, E., Deng, G.B., Kornhaas, M., of Simple Automotive Bodies. Appl. Mech. Rev. 73.
Schäfer, M., Fröhlich, J., Hinterberger, C., Rodi, W., 2013. On simulating the Zhang, B.F., Zhou, Y., To, S., 2015. Unsteady flow structures around a high-drag Ahmed
turbulent flow around the Ahmed body: A French-German collaborative evaluation body. J. Fluid Mech. 777, 291–326.
of LES and DES. Comput. Fluids 78, 10–23. Zhou, Y., Zhang, B., 2021. Recent Advances in Wake Dynamics and Active Drag
Shur, M.L., Spalart, P.R., Strelets, M.K., Travin, A.K., 2008. A hybrid RANS-LES approach Reduction of Simple Automotive Bodies. Appl. Mech. Rev. 73 (6), 060801.
with delayed-DES and wall-modelled LES capabilities. Int. J. Heat Fluid Flow 29, Zigunov, F., Sellappan, P., Alvi, F., 2020. Reynolds number and slant angle effects on the
1638–1649. flow over a slanted cylinder afterbody. J. Fluid Mech. 893, A11.

20