Research on Temperature Distribution of a Turbine Blade Based on Simulation Using PINNs

Simulation analysis of turbulence is one of the key technologies for design, development, use and maintenance of fluid machinery. In this paper, aiming at the heat transfer problem of the steady-state turbulent flow field of the turbine blades, the 2-D model of the flow field of the high-pressure turbine blades is established, and then the inlet, outlet and boundary conditions are set. The two step approach is proposed to solve the temperature distribution of turbine blades using the PINN method which the loss function including the residual of the RANS equations. This approach considers both the turbulent characteristics and the computational cost. The approach show that PINN method has good applicability for RANS simulation and can avoid the problem caused by the Reynolds stress terms. The PINN results show good agreement with that of the FEM, for the temperature distribution of the fluid domain, the relative error of the maximum temperature is less than 10%, and for the blade, The relative error of the maximum temperature is less than 6.0%. The method proposed has some reference significance for the flow simulation of turbulence based on machine learning methods.