Learning Neural-Network-Based Turbulence Models for External Transonic Flows Using Ensemble Kalman Method

This paper presents a neural-network-based turbulence modeling approach for transonic flows based on the ensemble Kalman method. The adopts tensor-basis neural network Reynolds-stress representation, with modified inputs to consider fluid compressibility. normalization of input features is also investigated avoid feature collapsing in presence shock waves. Moreover, turbulent heat flux accordingly estimated model gradient diffusion hypothesis. method used train experimental data velocity and wall pressure due its derivative-free nature. proposed framework tested two canonical configurations, that is, two-dimensional over RAE2822 airfoils three-dimensional ONERA M6 wings. Numerical results demonstrate capability learning accurate models external flows.