No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Numerical Investigations of the DrivAer Car Model using Opensource CFD Solver OpenFOAM
Abstract
A lot of the investigations in automotive aerodynamics are still based on strongly
simplified generic bodies such as the Ahmed Body or the SAE body. To close the
gap between these strongly simplified models and highly complex production cars
the new generic DrivAer body model is introduced by the Institute of Aerodynamics
and Fluid Mechanics, Technische Universitat Munchen (TUM). This current study is
focused on three different DrivAer body models namely Fastback, Estateback, and
Notchback and two different underbody types for each model, smooth and detailed.
Hence, total 6 different models are simulated using open source CFD solver
OpenFOAM at two different ground conditions, with ground and without ground effect
(WGS and WoGS). All the models used in simulations are 2.5 scaled down models
as compared with the actual car dimensions. The vehicle velocity considered for this
numerical study is 40 m/s, Reynolds number is 4.87M and turbulence model used is
k-w-SST. The mesh is generated using SnappyHexMesh (SHM) tool of OpenFOAM
and it is around 11 million volume cells for the smooth underbody and 14 million
volume cells for the detailed underbody. The coefficients of drag (C d) values are
within 0.5% to 12% error band as compared against the experimental values
published by the TUM. The coefficients of pressure (Cp) plots are comparable with
experimental results and also the contribution of individual body part in overall C d
values is obtained in this study. All the simulations are carried out using OpenFOAM
2.1.1 on Tata Consultancy Services (TCS) High Performance Computing facility.
Keywords : DrivAer body, External Aerodynamics, OpenFOAM, SnappyHexMesh,
CFD, HPC, TCS.
... • Interpolation scheme (interpolationSchemes): Linear (Second order). • Convective scheme (divScheme): "linearUpwind" (second order) for momentum (div(phi, U) = linearUpwind) and upwind (first order) for both k and omega (Shinde et al. 2013;Shaharuddin et al. 2017). • Turbulence model: K -omega -SST. ...
... As a result, the Cds of the baseline model in the CFD simulation and the wind tunnel experiment are 0.285 and 0.292, respectively. In addition, other comparative results of Cds with respect to the estate-back model between the CFD simulation with the K -omega -SST turbulence model and the wind tunnel experiment were provided in the previous investigations (Shinde et al. 2013;Ashton and Revell 2015. Apparently, an error of 2.4% in our CFD simulation is acceptable for the industrial automotive flows. ...
... Apparently, an error of 2.4% in our CFD simulation is acceptable for the industrial automotive flows. The errors of 2.7% and 7.1% in previous studies (Shinde et al. 2013;Ashton and Revell 2015) verify the rationality of the simulation. Moreover, it should be noted that the RANS models are typically used to evaluate the direction and magnitude of a trend with an affordable computational burden. ...
To efficiently achieve tangible improvements in the aerodynamic objectives of a vehicle based on a computational fluid dynamics (CFD) simulation that produces unavoidable noise, an improved system is proposed. This system, called regression kriging with re-interpolation (RKri)-based efficient global optimization (EGO) with a pseudo expected improvement (PEI) criterion (RKri-EGO-PEI), is used to directly filter out the noise produced by the CFD simulation, maintain a smooth trend of the surrogate model, and conduct point infills in a parallel manner. To guarantee optimization processes for tuning the hyper-parameters of RKri and searching for a solution of appropriate quality to the PEI function, the performance advantages of the optimizers on a parallel EGO algorithm called EGO-PEI are comprehensively investigated. Then, the best is chosen as the optimizer for the RKri-EGO-PEI system. To confirm the performance of the proposed system, RKri-EGO-PEI competes with ordinary kriging-based EGO-PEI (OK-EGO-PEI) and RKri-based EGO (RKri-EGO) systems on a real-world optimization problem of vehicle aerodynamics. The results of the investigation show that the performance of the optimizer with a higher central goal of exploration–exploitation can not only promote a higher-level convergence of the EGO-PEI algorithm within an appropriate number of point infills, but also ensure the same convergence level of the EGO-PEI algorithm as that using other optimizers, with fewer iterations. In addition, RKri-EGO-PEI searches for a lower drag coefficient (Cd) of the vehicle model with a faster speed and smaller wall-clock time cost than those of OK-EGO-PEI and RKri-EGO under an optimization problem with “noisy” computations.
... Thus, the study is reflected to the realistic production car. The design is based on Audi A4 and BMW 3 series (Shinde et al., 2013). Figure 3 shows the DrivAer fastback geometry chosen for the current study. ...
... Similar treatment has been used by Guilmineau (2014), where the floor under the model is made moving with the freestream velocity. A higher mean D C is obtained when the floor under the model is made as no-slip condition (Shinde et al., 2013;Ashton et al., 2016), i.e., the boundary layer is allowed to developed upstream of the model, where up to 7.0% difference in mean D C is observed. ...
... Dimensions of DrivAer modelSource: Taken fromShinde et al. (2013) ...
Interior noise of a production car is a total contribution mainly from engine, tyres and aerodynamics. At high speed, wind noise can dominate the total interior noise. Wind noise is associated with the unsteadiness of the flow. For most production cars, A-pillar and side view mirror are the regions where the highly separated and turbulent flows are observed. This study quantifies the wind noise contribution from A-pillar and side view mirror with respect to the interior noise of a generic realistic model, DrivAer. The noise sources are obtained numerically from the flow-structure interactions based on the unsteady Reynolds averaged Navier stokes (URANS) while the noise propagation is estimated using Curle's equation of Lighthill acoustic analogy. The sound pressure frequency spectrum of the interior noise is obtained by considering the sound transmission loss from the side glass by using the mass law for transmission loss. The study found that the noise from the A-pillar is higher than the noise from the side view mirror in the whole frequency range. Near the end of the A-pillar component contributes the highest radiated noise level with up to 20 dB louder than that at the front part of the A-pillar.
... Thus, the study is reflected to the realistic production car. The design is based on Audi A4 and BMW 3 series (Shinde et al., 2013). Figure 3 shows the DrivAer fastback geometry chosen for the current study. ...
... Similar treatment has been used by Guilmineau (2014), where the floor under the model is made moving with the freestream velocity. A higher mean D C is obtained when the floor under the model is made as no-slip condition (Shinde et al., 2013;Ashton et al., 2016), i.e., the boundary layer is allowed to developed upstream of the model, where up to 7.0% difference in mean D C is observed. ...
... Dimensions of DrivAer modelSource: Taken fromShinde et al. (2013) ...
Interior noise of a production car is a total contribution mainly from engine, tyres and aerodynamics. At high speed, wind noise can dominate the total interior noise. Wind noise is associated with the unsteadiness of the flow. For most production cars, A-pillar and side view mirror are the regions where the highly separated and turbulent flows are observed. This study quantifies the wind noise contribution from A-pillar and side view mirror with respect to the interior noise of a generic realistic model, DrivAer. The noise sources are obtained numerically from the flow-structure interactions based on the unsteady Reynolds averaged Navier stokes (URANS) while the noise propagation is estimated using Curle's equation of Lighthill acoustic analogy. The sound pressure frequency spectrum of the interior noise is obtained by considering the sound transmission loss from the side glass by using the mass law for transmission loss. The study found that the noise from the A-pillar is higher than the noise from the side view mirror in the whole frequency range. Near the end of the A-pillar component contributes the highest radiated noise level with up to 20 dB louder than that at the front part of the A-pillar.
... Thus, the study is reflected to the realistic production car. The design is based on Audi A4 and BMW 3 series (Shinde et al., 2013). Figure 3 shows the DrivAer fastback geometry chosen for the current study. ...
... Similar treatment has been used by Guilmineau (2014), where the floor under the model is made moving with the freestream velocity. A higher mean D C is obtained when the floor under the model is made as no-slip condition (Shinde et al., 2013;Ashton et al., 2016), i.e., the boundary layer is allowed to developed upstream of the model, where up to 7.0% difference in mean D C is observed. ...
... Dimensions of DrivAer modelSource: Taken fromShinde et al. (2013) ...
... Several numerical works have also been carried out with the DrivAer model, particularly RANS simulations (Shinde et al., 2013;Guilmineau, 2014a,b;Peters et al., 2015;Ashton et al., 2016;Jakirlic et al., 2014) and hybrid approaches such as detached eddy simulations (DES) (Guilmineau, 2014a;Ashton et al., 2016) or partial averaged Navier-Stokes (PANS) (Jakirlic et al., 2014). Other numerical approaches have also been used, such as the lattice Boltzmann method (LBM) (Pasquali et al., 2015). ...
... Guilmineau (2014b) identified the interactions present between the rotating wheels and the underbody of the car, leading to a reduction in both the drag and lift coefficient (with respect to the same geometry with stationary wheels). This change in force coefficients was also observed experimentally by Heft et al. (2012) and numerically by Shinde et al. (2013) among others. Other observed effects include a change in the vortices stemming from the wheels, particularly those forming around the front wheels. ...
... Heft (2014), 2 :Guilmineau (2014b),3 Peters et al. (2015), 4 :Shinde et al. (2013), 5 :Strangfeld et al. (2013), 6 : Guilmineau (2014a), 7 :Ashton et al. (2016) coefficient results vary between C D = 0.2254(Guilmineau, 2014b) to C D = 0.268(Ashton et al., 2016). These differences can be attributed to two main reasons: Low frequency behavior of some quantities requiring a longer integration time, and differences in experimental set up and CFD. ...
Numerical simulations are carried out on the flow over a realistic generic car geometry, the DrivAer-fastback car model. Pure large eddy simulations (LES) and wall-modeled large eddy simulations (WMLES) are used and com
pared to numerical and experimental results to assess the validity of these approaches when solving the flow field around complex automotive geometries.
Results show a 70% CPU time reduction when using the wall model. Drag coefficient results show the influence of the wall model on coarser meshes is positive, reducing the difference on those obtained using finer meshes. Pressure profiles exhibit mixed results. The wall model used works well in adverse pressure gradients and smooth geometry changes. Results worsen in sections where the flow detaches and experiences large pressure drops. Flow structures and unsteady effects around the car are also analyzed, obtaining several characteristic frequencies for the different flow structures encountered. It should be noted that the present investigation shows how WMLES helps to reduce the computing cost and response vs pure LES, while providing high-quality
unsteady data, although computational cost remains high. Results show potential in the introduction of this tool as a competitive simulation strategy for complex geometries.
... 5.6 Asmo car lift coefficients. . [34], 2 : Guilmineau [32], 3 Peters et al. [99], 4 : Shinde et al. [123], 5 : Strangfeld et al. [128], 6 : Guilmineau [31], 7 : Ashton et al. [ ...
... Several numerical works have also been carried out with the drivAer model, particularly RANS simulations [5,31,32,43,99,123] and hybrid approaches such as detached eddy simulations (DES) [5,31] or partial averaged Navier-Stokes (PANS) [43]. Other numerical approaches have also been used, such as the lattice Boltzmann method (LBM) [94]. ...
... Guilmineau [32] identified the interactions present between the rotating wheels and the underbody of the car, leading to a reduction in both the drag and lift coefficient (with respect to the same geometry with stationary wheels). This change in force coefficients was also observed experimentally by Heft et al. [34] and numerically by Shinde et al. [123] amongst others. Other observed effects include a change in the vortices stemming from the wheels, particularly those forming around the front wheels. ...
Aerodynamic analysis has become one of the most important tools in many engineering applications. In this sense, this thesis work is aimed at performing aerodynamic analysis of different geometries, expanding the available knowledge and obtaining valuable insight from the obtained results. Aerodynamic analysis can be carried out, principally, in two ways: Experimental research and Computational Fluid Dynamics (CFD). The former makes use of prototypes, wind tunnels and test tracks, making it a very expensive option. On the other hand, CFD makes use of numerical tools to solve the Navier-Stokes equations within a computational discretized domain. This latter approach is essentially limited by the available computational power and by the aerodynamicist's experience.
This work comprises eight chapters. The first one is an introduction to the type of flows and geometries considered, as well as, the general methodology followed in the posterior studies. The following six chapters are the core of this dissertation, and encompass the numerical resolution of the Navier-Stokes equations in selected geometries, ordered by complexity level. In particular, the contents of these seven chapters have been submitted or published in international journals and conferences. For this reason, they are self contained and few changes have been made. The reader might find that some concepts are repeated along them. The last chapter contains concluding remarks. Finally, appendix 1 describes some applications of aerodynamic studies to some related projects and appendix 2 comprises a list of publications done during the PhD.
... This model is based on the mockup of the Audi A4 and the BMW 3 Series, leading to a medium sized car, with three main shapes: fastback, notchback and estate shapes. Many numerical investigations have been done on the three shapes ( [9], [19], [3]). The most problematic shape remains the estate one, due to the massive separation at the back and the interactions with longitudinal vortices generated at the front of the vehicle (A-pillar vortices and mirror screens). ...
The aim of the “Models for Vehicle Aerodynamics” (MOVA) Project is to develop, refine, and validate the latest generation of turbulence models for selected examples encountered in vehicle aerodynamics. The validation of turbulence models requires the availability of detailed experimental data. These quantitative data should cover the most critical flow regions around a bluff car-shaped body and they should give physical quantities that can directly be correlated to the results of numerical simulations. Such experimental data were measured in the LSTM low speed wind tunnel using a 2-component laser-Doppler anemometer (LDA) mounted on a traversing system and a simplified model of a car (Ahmed model). Measurements were made for two rear vehicle body slant angles (25° and 35°) at a bulk air velocity of 40 m/s. This paper serves as a synopsis of the major results of this experimental investigation.
The large wind tunnel at the Technische Universität München was upgraded by integrating a modular single-belt system, which enables the simulation of moving ground conditions for ground vehicle testing. Central part of this system is its large belt that moves at a maximum speed of 50 m/s. This belt not only simulates the relative motion between the model vehicle under investigation and the floor, but also drives the model's wheels. Due to its size, the wind tunnel facility is suited for testing 40 % scaled models of typical passenger cars, which are held in place by a newly designed model support system consisting of five struts: One strut to support the body of the model and four struts to hold the model's wheels on top of the moving belt. Another crucial step in upgrading the wind tunnel was to install a boundary layer scoop system to reduce the thickness of the boundary layer approaching the moving belt. All new systems were designed such that they can be moved into and out of the test section of the wind tunnel to be able to restore the old setting of the test section floor. This report addresses the key challenges we had to face during the process of upgrading the wind tunnel facility and introduces its special features and its most important sub-systems. The current report also contains results of tests we conducted to measure the distribution of the static pressure along the test section and the size of the boundary layer at different locations on top of the moving belt. Even though we did not obtain perfect conditions regarding the size of the boundary layer thickness on top of the moving belt, we still were successful in improving the scope of the wind tunnel facility in terms of ground vehicle testing.
State of the art aerodynamic research of vehicles often employs strongly simplified car models, such as the Ahmed and the SAE body, to gain general insights. As these models exhibit a high degree of abstraction, the obtained results can only partly be used for the aerodynamic optimization of production vehicles. Aerodynamic research performed on specific vehicles is on the other hand often limited due to their short life span and restricted access. A new realistic generic car model for aerodynamic research-the DrivAer body-is therefore proposed to close this gap. This paper focuses on the development of the model and the first experimental results, namely force and pressure measurements of the different configurations. The experiments were performed in the recently updated Wind Tunnel A of the Institute of Aerodynamics and Fluid Mechanics at the Technische Universität München.
Unsteady aerodynamic flow phenomena are investigated in a wind tunnel by oscillating a realistic 50% scale model around the vertical axis. Thus the model is exposed to time-dependent flow conditions at realistic Reynolds and Strouhal numbers. Using this setup unsteady aerodynamic loads are observed to differ significantly from quasi steady loads. In particular, the unsteady yaw moment exceeds the quasi steady approximation significantly. On the other hand, side force and roll moment are over predicted by quasi steady approximation but exhibit a significant time delay. Part 2 of this study proves that a delayed and enhanced response of the surface pressures at the rear side of the vehicle is responsible for the differences between unsteady and quasi steady loads. The pressure changes at the vehicle front, however, are shown to have similar amplitudes and almost no phase shift compared to quasi steady flow conditions. The difference between unsteady and quasi steady yaw moment proves to be independent of oscillation amplitudes between 2deg and 4deg. It is assumed that the intensity of the unsteady flow phenomena is determined by the interaction of the time scale of the model rotation and the time scale of the delayed wake flow, described by the Strouhal number. The largest magnification factors for the yaw moment are found at 140kph and 2Hz for the notchback geometry, which results in a Strouhal number of Sr=0.12. It is finally shown that the yaw moment overshoot is less pronounced for a fastback and especially for a fullback geometry, which is explained by smaller unsteady pressure variations at the rear side of the fullback.
The unsteady flow structures at the rear end of a car and in its wake strongly influence its handling characteristics. State of the art research on this topic often employs strongly simplified car models, such as the Ahmed body. Due to their high degree of ab- straction, however, the insights gained cannot be fully transferred to the development of production cars. To close this gap a new reference car model for aerodynamic research - the DrivAer body - is proposed in this paper. Its geometry and validation data will be published to encourage independent studies. The investigations of both the Ahmed body and the DrivAer model are conducted both numerically and experimentally. The transient numerical simulations were carried out using Large Eddy Simulation (LES) and the Scale-Adaptive Simulation Shear-Stress Transport (SAS SST) turbulence model of the open source software OpenFOAM®. These results are compared to time-accurate surface pressure measurements and force measurements. Initial results show good agreement with existing experimental data.
Large eddy simulations (LES) were made of flows around a generic ground vehicle with sharp edges at the rear end (an Ahmed body with a 25 degrees angle of the rear slanted surface). Separation of the flow at the rear results in large regions with recirculating flow. As the separation is determined by the geometry, the Reynolds number effects are minimized. Resolution requirements of this recirculating flow are smaller than those in LES of wall attached flows. These two consequences of the geometry of the body are used to predict the experimental flow at relatively high Reynolds number Recommendations are presented for the preparation and realization of LES for vehicle flows. Comparison of the LES results with the experimental data shows good agreement.
For a basic ground vehicle type of bluff body, the time averaged wake structure is analysed. At a model length based reynolds number of 4.29 million, detailed pressure measurements, wake survey and force measurements were done in a wind tunnel. Some flow visualisation results were also obtained. Geometric parameter varied was base slant angle. A drag breakdown revealed that almost 85% of body drag is pressure drag. Most of this drag is generated at the rear end. Wake flow exhibits a triple deck system of horseshoe vortices. Strength, existence and merging of these vortices depend upon the base slant angle. Characteristic features of the wake flow for the low drag and high drag configurations is described. Relevance of these phenomena to real ground vehicle flow is addressed.
This paper presents a Spalart-Allmaras based Detached-Eddy Simulation (DES) of the Ahmed reference car model with 25° and 35° slant angles using unstructured grids and the solverCobalt. Comparisons are made to experimental laser doppler velocity measurements as well as total and surface pressure integrated drag. The Reynolds number based on body length was 2.78 ×106, making the boundary layers approaching the slant fully turbulent. The flow over the base slant in the experiments is attached at 25° and separated at 35°. This causes a large drop in the drag with the increased slant angle as the vortices on the side of the slant are weakened due to the separation. These cases stress turbulence models due to the need to accurately predict the boundary layer separation over the slant as well as predict the pressures in the massively separated base region accurately. The DES results are compared to the experiments as well as the Spalart-Allmaras RANS model. DES is seen to predict separation at 25◦ slant angle, in contrast to the experiments. Drag is relatively close to the experiments, but the distribution of drag is more on the rear than on the slant due to the separation. At the 35°
slant angle, DES is in good agreement to the experimental drag, with the correct distribution, while RANS over-predicts the drag.
The Ahmed reference body represents a simplified car geometry that can be used to investigate the main flow features in the wake of vehicles. The present work presents unsteady flow simulations at the rear slant angles 25° and 35° using the PowerFLOW 4.0 D3Q19 lattice Boltzmann model. The flow of the wake is discussed and distributions of averaged pressure and velocities are compared with available experimental findings. The resolution requirement is investigated in terms of computational requirement and accuracy achieved. The predictive capability and the feasibility of the very large eddy simulation (VLES) approach within the lattice Boltzmann framework is demonstrated.
Some salient features of the time averaged ground vehicle wake. SAE paper no. 840300. the wake of a simplified car model (Ahmed Model)
- S R Ahmed
- G Ramm
- G Faltin
S. R. Ahmed, G. Ramm, and G. Faltin, 1984. Some salient features of the time averaged ground vehicle wake. SAE paper no. 840300. the wake of a simplified car model (Ahmed Model). In DGLR Fach Symp. der AGSTAB.
A Turbulence based Computational Study on the Drag Breakdown of Ahmed Body
- R Pandya
- S Pavitran
- K Nikam
- A Singh
Pandya, R., Pavitran, S., Nikam, K., Singh, A., A Turbulence based
Computational Study on the Drag Breakdown of Ahmed Body, IUTAM
Symposium on Bluff Body Flows, December 12-16, 2011, IIT Kanpur, India.
Pg. 337-340.
Investigation of Unsteady Flow Structures in the Wake of a Realistic Generic Car Model On the Use of Reference Models in Automotive Aerodynamics, SAE Technical Paper
- A I Heft
- T Indingery
- N A Adams
A. I. Heft, T. Indingery, and N. A. Adams, Investigation of Unsteady Flow Structures in the Wake of a Realistic Generic Car Model, 29th AIAA Applied Aerodynamics Conference, 27 -30 June 2011, Honolulu, Hawaii 10. Le Good, G. M., and Garry, K. P., On the Use of Reference Models in Automotive Aerodynamics, SAE Technical Paper 2004-01-1308, 2004, doi: 10.4271/2004-01-1308.
The Ground-Simulation Upgrade of the TUM Wind Tunnel
- S Mack
- T Indinger
- N A Adams
- P Unterlechner
Mack, S., Indinger, T., Adams, N. A., and Unterlechner, P., " The Ground-Simulation Upgrade of the TUM Wind Tunnel, " SAE Technical Paper 2012-01-0299, 2012, doi:10.4271/2012-01-0299.