Case 2DrivAer Fastback and Estate 1st Automotive CFD Prediction Workshop
2019-12-11
Petter EkmanLinköping University
Content
• Background about chosen Method
– Time-Step Size Sensitivity Study *
– Turbulence Model Study **
• Chosen Method Case 2
• Simulation Results Case 2
2019-12-14 2Title/Lecturer
* Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.
** Ekman, P., et al. Assessment of Hybrid RANS-LES Methods for Accurate Automotive Aerodynamic Simulation., Submitted to Journal of Wind Engineering & Industrial Aerodynamics
Method – Sensitivity Study
• DrivAer Reference Model – Notchback
– Smooth Underbody
• 𝑅𝑒𝐿 = 3.12 ∙ 106
• 5° of yaw
• Test section included in the simulations
– GroWiKa WT at TU Berlin
• Stationary ground and wheels
Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.
Method – Sensitivity Study
• ANSYS Fluent
• Stress Blended Eddy Simulation (SBES)
– k-ω SST RANS model
– Dynamic Smagorinsky SGS Model
• ∆𝑡 = 1.4 ∙ 10−6𝑠
– 𝐶𝐹𝐿 < 1
• Mesh
– 15-20 prisms layers
– 61, 102 and 158 million cells
Mesh size 𝐂𝐃 𝐂𝐋61 million cells 0.268 -0.120102 million cells 0.266 -0.136158 million cells 0.269 -0.137
Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.
Method – Sensitivity Study
Comparison to Wind Tunnel Measurements – Following Best Practice
Method 𝐂𝐃 𝐂𝐋CFD 0.268 ± 0.002 -0.136 ± 0.001Wind Tunnel 0.272 ± 0.003 -0.119
Measurements performed by TU BerlinWieser, D., et al. Experimental Comparison of the Aerodynamic Behavior of Fastback and Notchback DrivAer Models. No. 2014-01-0613. SAE Int. J. Passeng. Cars, 2014.
Method – Sensitivity Study
• Time-Step Size Investigation Corresponding time-step size for Case 2
CFL Time-step size [s] (𝑳/(∆𝒕 ∙ 𝑼∞))
1 1.4 ∙ 10−6 2085010 1.4 ∙ 10−5 208520 2.8 ∙ 10−5 1042.550 7.0 ∙ 10−5 417
100 1.4 ∙ 10−4 208.5
CFL50
CFL Time-step size [s]
1 1.38 ∙ 10−5
10 1.38 ∙ 10−4
20 2.76 ∙ 10−4
50 6.89 ∙ 10−4
100 1.38 ∙ 10−3
Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.
Results – Sensitivity Study
Forces - Difference against CFL1
• Drag forces relative insensitive
• Lift forces more sensitive
Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.
Results – Sensitivity Study
CFL10 CFL20
CFL50 CFL100
Total Pressure and Skin Friction
Differences Against CFL1
Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.
Results – Sensitivity Study
Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.
Results – Sensitivity Study
SBES vs DDES and IDDES
Notchback Fastback
Ekman, P., et al. Assessment of Hybrid RANS-LES Methods for Accurate Automotive Aerodynamic Simulation., Submitted to Journal of Wind Engineering & Industrial Aerodynamics
Measurements performed by TU BerlinWieser, D., et al. Experimental Comparison of the Aerodynamic Behavior of Fastback and Notchback DrivAer Models. No. 2014-01-0613. SAE Int. J. Passeng. Cars, 2014.
Results – Sensitivity Study
SBES vs DDES and IDDES
Drag difference when increasing yaw angle for 0°
Notchback Fastback
Ekman, P., et al. Assessment of Hybrid RANS-LES Methods for Accurate Automotive Aerodynamic Simulation., Submitted to Journal of Wind Engineering & Industrial Aerodynamics
Measurements performed by TU BerlinWieser, D., et al. Experimental Comparison of the Aerodynamic Behavior of Fastback and Notchback DrivAer Models. No. 2014-01-0613. SAE Int. J. Passeng. Cars, 2014.
Results – Sensitivity StudyNotchback
SBES vs DDES and IDDES
Ekman, P., et al. Assessment of Hybrid RANS-LES Methods for Accurate Automotive Aerodynamic Simulation., Submitted to Journal of Wind Engineering & Industrial Aerodynamics
Results – Sensitivity StudyFastback
SBES vs DDES and IDDES
Ekman, P., et al. Assessment of Hybrid RANS-LES Methods for Accurate Automotive Aerodynamic Simulation., Submitted to Journal of Wind Engineering & Industrial Aerodynamics
Chosen Method – Case 2
• ANSYS Fluent 2019R1
• Stress Blended Eddy Simulation (SBES)
– Dynamic Smagorinsky SGS model
– k-ω SST RANS model
• Δt= 1.375 ∙ 10−4s (corresponding to CFL10)
• 5 Inner Iterations
• Simulation Time: 5+20 Convective Flow Units 𝑡 ∙ Τ𝑈∞ 𝐿
• Mesh = Medium Hexapoly
• Boundary Conditions according to Case 2 description
p-v SIMPLEC
Momentum 2nd order Bounded Central Difference
Turbulence 2nd order Upwind
Pressure 2nd order Central Difference
Temporal 2nd order Bounded Implicit IterativeTime-Advancement
SBES is ~25% more expensive than DDES for the same mesh and numerical settings
• Simulation Cost on 1920 cores• Fastback = 133 658 corehours• Estate = 125 429 corehours
Results - Forces
• Absolute Forces
• Force Difference: Estate - Fastback
Car Body/Method 𝑪𝑫 𝑪𝑳 𝑪𝑳𝑭 𝑪𝑳𝑹
Fastback – SBES 0.229 -0.035 -0.120 0.086
Fastback – WT* 0.243 - - -
Estate - SBES 0.279 -0.198 -0.154 -0.044
Estate – WT* 0.292 - - -
Method ∆𝑪𝑫 ∆𝑪𝑳
SBES 0.050 -0.163
WT* 0.049 -
* Heft, A., et al. Introduction of a New Generic Realistic Car Model for Aerodynamic Investigations. No. 2012-01-0168. SAE Technical Paper, 2012.
Time-Averaging time (20 flow units)
Results - WSS
Results - Pressure
• Comparison to Heft, A., et al. * and
* Heft, A., et al. Introduction of a New Generic Realistic Car Model for Aerodynamic Investigations. No. 2012-01-0168. SAE Technical Paper, 2012.
Results - Pressure
• Comparison to Avadiar, T., et al. *
• Offset of Cp = 0.05
* Avadiar, T., et al. Characterisation of the wake of the DrivAer estate vehicle. Journal of Wind Engineering & Industrial Aerodynamics, 2018.
Conclusions
• Possible to be aggressive with time-step size
– Drag relative insensitive
– Lift more sensitive
• High accuracy achieved with SBES
– Able to capture the complex flow over the rear window
– Base pressure correlate well with measurements
– Good drag prediction for different yaw and car configurations
– Excellent trend prediction
– ~25% more expensive than DDES k-ω SST
Acknowledgements
Thanks to TU Berlin and especially Dirk Wieser for sharing measurement data
Thanks to National Supercomputer Centre at Linköping University for providing computational resources
Thank you!
Extra Material
CFL10 CFL20
CFL50 CFL100
Total Pressure and Skin Friction
CFL1
Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.
Extra Material
Surface Pressure
CFL1
CFL10 CFL20
CFL50 CFL100
Ekman, P., et al. Accuracy and Speed for Scale-Resolving Simulations of the DrivAer Reference Model. No. 2019-01-0639. SAE Technical Paper, 2019.