Adjoint-Based Model Tuning and Machine Learning Strategy for Turbulence Model Improvement 2022-01-0899

As turbulence modeling has become an indispensable approach to perform flow simulation in a wide range of industrial applications, how to enhance the prediction accuracy has gained increasing attention during the past years. Of all the turbulence models, RANS is the most common choice for many OEMs due to its short turn-around time and strong robustness. However, the default setting of RANS is usually benchmarked through classical and well-studied engineering examples, not always suitable for resolving complex flows in specific circumstances.

Many previous researches have suggested a small tuning in turbulence model coefficients could achieve higher accuracy on a variety of flow scenarios. Instead of adjusting parameters by trial and error from experience, this paper introduced a new data-driven method of turbulence model recalibration using adjoint solver, based on Generalized k-ω (GEKO) model, one variant of RANS. In this algorithm, the model coefficients are tuned locally to match the measured drag force and surface pressure of DrivAer generic cars as closely as possible. After well-tuned process, neural network strategy was utilized with a training dataset to learn the correlation between calibrated coefficients and certain flow features. Once trained, the improved model can be employed in similar cases, like geometrical modifications and varying working conditions. Results show that through tuning and machine learning, prediction of drag and surface pressure presented better agreement with test data, while with only C_D as the target, the trained model application scope is quite limited, thus, a more comprehensive input of accessible data is needed to obtain a more generalized model.

Adjoint-based GEKO tuning, validated by the cases of this paper and correlated with wind tunnel test, proved to be very helpful in turbulence model optimization. Hence, this study indicated the promising potential of turbulence model tuning and machine learning in vehicle aerodynamic application, and at the same time, discussed the related issues encountered during this work.