Accelerating Design Space Exploration of Film Cooling with Parametric Machine Learning Based on High-Fidelity CFD

Shaped film cooling holes are a key element in cooling turbine components to increase their operating temperature while preventing component damage. There is therefore a need for designers to rapidly explore options and identify the optimum film cooling hole configuration. In this study, we propose a machine learning (ML) approach to evaluate the performance of a given film cooling hole shape in a matter of seconds using the publicly available 777-shaped hole design. The machine learning model was trained using data generated with a Lattice Boltzmann Method (LBM) numerical solver as part of a Design of Experiment with four different parameters: the forward expansion angle, the lateral expansion angle, the slot injection angle, and the blowing ratio. Blind predictions from inference of the ML model were compared to results predicted by LBM and showed a mean error of ~3% for the area averaged film cooling effectiveness while preserving the design relative ranking. This suggests that the method described in this paper is viable by industry standards.