Prediction Method of Concentricity and Perpendicularity of Aero Engine Multistage Rotors Based on PSO-BP Neural Network

This paper proposes a method for predicting concentricity and perpendicularity based on PSO-BP neural network in order to solve the problem of low accuracy for aero engine multistage rotors assembly. The influence factors of error propagation in the assembly are analyzed based on the characteristics of rotor structure and assembly process. And neural networks for predicting concentricity and perpendicularity of multistage rotors assembly are established. The particle swarm algorithm is used to optimize the hyperparameters of the neural network and the optimal hyperparameters can be obtained. In order to verify the effectiveness of the concentricity and perpendicularity prediction method proposed in this paper, experiments are carried out for four rotors assembly with precision rotary measuring instrument. The results show that for the 30 groups of testing samples, the average deviations of concentricity and perpendicularity by PSO-BP neural network prediction method are 1.0 μm and 0.6 μm, respectively. The prediction accuracy of concentricity and perpendicularity of final assembly are improved by 4.5 μm and 2.6 μm, respectively, compared with the traditional assembly method. The proposed method in this paper can be used not only for the guidance of multistage rotors assembly of aero engine, but also for the tolerance allocation in the design process.