基于机器学习的篦齿封严流动传热性能预测OA

The study focuses on the fine-grained and rapid prediction of the flow and heat transfer performance of labyrinth seals with smooth bushings using machine learning methods.Input parameters are selected for the machine learning model include the labyrinth seal clearance,tooth spacing,forward angle of the labyrinth tooth,inlet total pressure,and rotational speed.Learning samples required for the machine learning model are established based on Latin hypercube sampling and computation fluid dynamic(CFD)simulation methods.Radial basis function neural networks are independently constructed at each node of the machine learning grid(with a number of nodes significantly fewer than the CFD grid)to directly predict the total pressure,total temperature,and velocity at the corresponding locations.Comparison between the radial basis function neural network and CFD calculation results reveals that the mean absolute percentage errors for total pressure,total temperature,and velocity are 0.07%,0.17%,and 0.14%,with root mean square errors of 1.18 kPa,1.8 K,and 6.45 m/s,respectively.A detailed analysis of the numerical relationships between the input and output parameters of the machine learning model is conducted.The research findings provide a new technical approach for efficient and low-cost modeling of the heat transfer performance of labyrinth seals.