Development of a Data-Driven Wall Model for Large-Eddy Simulation of Gas Turbine Film Cooling Flows

A data-driven wall modeling framework is developed for large-eddy simulation (LES) of gas turbine film cooling systems. High-fidelity near-wall turbulent flow datasets are extracted from wall-resolved LES of a canonical gas turbine film cooling configuration, comprised of a flat plate with a single row of 7-7-7 shaped cooling holes, performed using a high-order spectral element CFD solver Nek5000. Wall-normal distance, near-wall fluid velocity, velocity/pressure gradients, and fluid density are used as input features and wall shear stress is predicted as the output. Two machine learning (ML) models are explored for predicting wall shear stress: Light Gradient Boosting Machine (LGBM) and artificial neural network (ANN). ML training is performed for blowing ratio (BR) of 1, whereas datasets for BRs 0.5, 1.5, and 2 are employed as the test set. It is observed that ANN tends to generalize slightly better than LGBM. Moreover, adding velocity gradient information and incorporating flow feature information from multiple streamwise/normal/spanwise neighbors improves the accuracy and generalizability of the data-driven wall model.