Augmentation of predictive turbulence modeling applied to low pressure turbines using adaptive field inversion

Based on the framework of field inversion and machine learning, an adaptive modification for Reynolds-Averaged Navier–Stokes-based turbulence models is proposed for the simulation of low-pressure turbine cascades involving flow separation. This method adjusts the results of turbulence by modifying the source terms correspondingly at different spatial locations. First, the specific regions for modification are obtained by Gaussian mixture models adaptively according to the baseline results and the spatial distribution of the correction term is inferred by an ensemble-based inversion method with effective utilization of high-fidelity data. Then a corrective model form of the flow quantities calculated with the baseline model is established by the Gradient Boosting Decision Tree model and used for the simulation of T106 cascade cases. The results demonstrate that with the adaptive modified turbulence model, reduced deficiency on predicting the load distribution can be obtained. The modified model can also predict a more accurate separation onset by damping eddy viscosity in separated region for case out of the training set. With correction added to the turbulence model solely in a specific region, the computational cost can be reduced compared with full-field inversion, and the method can be possibly applied in simulating the three-dimensional flow considering rotation effects.