An efficient calculation method of fan-shaped film cooling based on Reynolds-averaged Navier–Stokes simulation and deep learning

Film cooling is extensively utilized to protect the high-pressure turbine blade in advanced aero engines from high-temperature gases. In many of the current studies and design optimizations, artificial intelligence is deployed with deep learning, with datasets often generated based on Computational Fluid Dynamics (CFD) simulations of the cooling effectiveness of film cooling holes. In the current study, first, a variety of aerodynamic and geometric parameters are studied, including the blowing ratio, injection angle, laidback angle, lateral angle, etc. The various design parameters and flow conditions of fan-shaped holes are interrelated, and comprehensive considerations must be carefully carried out. Second, the numerical results are used as input datasets for several neural network models. Different neural network architectures are employed to estimate the film cooling effectiveness. The outcomes demonstrate that a well-designed neural network model with moderate layers can accurately predict film cooling within a tiny fraction of CFD computing time, even with a small dataset. Finally, the sparrow search algorithm is adopted to optimize the parameters of the cooling hole inside the design space. This demonstrates that the neural network approach is a powerful tool with great potential for numerical design and optimization of a massive number of film cooling holes on an entire blade.