Multi-objective Robust Optimization of EMU Brake Module

In the robust optimization of Electrical Multiple Units (EMU) brake module, a multi-objective robust optimization mathematical model of the brake module based on Radial Basis Function Neural Network (RBFNN) that is improved by Genetic Particle Swarm Optimization (GPSO-RBFNN) is proposed, considering the complex mathematical mapping relationship between the design variables and the target responses. For improving the calculation efficiency, an Improved Non-Dominated Sorted Genetic Algorithm-III (INSGA-III) is proposed. Firstly, the key design variables to total mass and natural frequency are selected by the sensitivity analysis for the brake module. Secondly, the sample value of each design variable of the brake module and its corresponding quality characteristic value are obtained through the orthogonal test, and then the Signal-to-Noise Ratio (SNR) of each quality characteristic can be calculated. The sample value obtained from the orthogonal test design is used as the input and the corresponding signal-to-noise ratio is used as the output for the training and testing of GPSO-RBFNN. Finally, establish the multi-objective robust optimization mathematical model based on GPSO-RBFNN, and the INSGA-III is used to solve the optimization model. Then compare the robust optimization result with the traditional design scheme. The result shows that the prediction accuracy of GPSO-RBFNN is much higher than that of RBFNN; the computational efficiency of INSGA-III is higher than that of NSGA-III. After optimization, the SNR of each quality characteristic of the brake module is improved, and the multi-objective robust optimization of the EMU brake module is realized.