Optimization of supersonic axial turbine blades based on surrogate models

Due to the transition of the energy system to more decentralized sector-coupled technologies, the demand on small, highly efficient and compact turbines is steadily growing. Therefore, supersonic impulse turbines have been subject of academic research for many years because of their compact and low-cost conditions. However, specific loss models for this type of turbine are still missing. In this paper, a CFD-simulation-based surrogate model for the velocity coefficient, unique incidence as well as outflow deviation of the blade, is introduced. This surrogate model forms the basis for an exemplary efficiency optimization of the ""Colclough cascade"". In a first step, an automatic and robust blade design methodology for constant-channel blades based on the supersonic turbine blade design of Stratford and Sansome is shown. The blade flow is fully described by seven geometrical and three aerodynamic design parameters. After that, an automated numerical flow simulation (CFD) workflow for supersonic turbine blades is developed. The validation of the CFD setup with a published supersonic axial turbine blade (Colclough design) shows a high consistency in the shock waves, separation zones and boundary layers as well as velocity coefficients. A design of experiments (DOE) with latin hypercube sampling and 1300 sample points is calculated. This CFD data forms the basis for a highly accurate surrogate model of supersonic turbine blade flow suitable for Mach numbers between 1.1 and 1.6. The throat-based Reynolds number is varied between 1*104 and 4*105. Additionally, an optimization is introduced, based on the surrogate model for the Reynolds number and Mach number of Colclough and no degree of reaction (equal inlet and outlet static pressure). The velocity coefficient is improved by up to 3 %.