ベイズ最適化を用いたロケットエンジン用タービンの三次元空力最適設計

抄録

This paper discusses an attempt at three-dimensional aerodynamic optimization of the turbine blade geometry in a two-stage axial tur-bine for a rocket engine turbopump. The optimization was conducted for the turbine originally designed by JAXA (Japan Aerospace Exploration Agency) as part of the turbopump study conducted by DDT (Dynamic Design Team). To minimize computational costs during optimization, Bayesian optimization was applied to explore optimal turbine blade geometries with high aerodynamic performance based on three-dimensional steady-state flow simulations, aiming to maximize stage efficiency. Subsequently, unsteady flow simulations were conducted on both the optimized turbine and the reference turbine, validating the optimization results based on the steady-state simulation. The flow physics behind the improvement in aerodynamic performance was analyzed by comparing the results from both turbines. The results indicate that the optimization led to 3.7-point improvement in efficiency while achieving 93% reduction in computational costs. The main factors contributing to the increased efficiency were the weakening of shock waves in the first-stage rotor blades, as well as the reduction of secondary flows and leakage vortices.