Correlating CFD and experimental vehicle aerodynamics using multi-fidelity simulation approaches

The automotive industry faces several challenges requiring faster product development, where numerical simulations and digitalization are key enablers to reduce time to market and development costs. Numerical methods require both short turnaround times and high-fidelity results. Capturing small differences across vehicle designs, by accurately predicting complex flow phenomena is crucial for aerodynamics optimization. The disruptive and fast development of GPGPU computing hardware, promising accelerated turnaround times at lower costs, found its natural position in this landscape. This paper describes simulation approaches with increasing fidelity applied to a set of variants of a Stellantis estate production car: these include geometrical, yaw angle and ride height changes, and all were tested in wind tunnel test facilities. Correlation between aerodynamics CFD simulations using Simcenter STAR-CCM+ and wind tunnel measurements is verified by comparing drag trends, pressure probes and wake visualizations. From a numerical perspective, several simulation approaches are analyzed with increasing complexity. First, all design variants were simulated in an open road domain, using both steady state RANS and transient, scale-resolving DDES. The same approaches were applied in a digital wind tunnel domain corresponding to the real-world test conditions. Domain and modeling strategy impacts are analyzed and correlated to test results. All simulations were run on both CPU and GPU computing facilities, providing an answer in terms of price/performance differences and results consistency. To confirm the validity of the highest fidelity approach identified, an additional SUV model was simulated in a different wind tunnel environment and drag deltas were compared to test.