基于GAPSO-RBFNN的动车组电机吊架多目标稳健优化设计  

GAPSO-RBFNN-Based Multi-Objective Robust Optimal Design for Motor Hanger of EMU

作　　者：李永华[1] 盛自强 宫琦[1] LI Yonghua;SHENG Ziqiang;GONG Qi(College of Locomotive and Rolling Stock Engineering,Dalian Jiaotong University,Dalian Liaoning 116028,China;School of Mechanical Engineering,Dalian Jiaotong University,Dalian Liaoning 116028,China)

机构地区：[1]大连交通大学机车车辆工程学院,辽宁大连116028 [2]大连交通大学机械工程学院,辽宁大连116028

出　　处：《中国铁道科学》2020年第3期103-110,共8页China Railway Science

基　　金：国家自然科学基金资助项目(51875073);国家重点研发计划项目(2016YFB1200504);辽宁省自然科学基金资助项目(20170540129)。

摘　　要：基于遗传粒子群(GAPSO)算法获取最优平滑系数,从而改进径向基神经网络(RBFNN);通过电机吊架的灵敏度分析筛选出对其总质量和自然频率等质量特性影响较大的关键设计变量;结合正交试验设计与有限元分析得出电机吊架各质量特性值及对应的信噪比,将试验数据作为输入、信噪比作为输出用于GAPSO-RBFNN的训练和测试,并对比分析预测精度;基于GAPSO-RBFNN构建电机吊架的多目标稳健优化模型,采用NSGA-II多目标优化算法对其寻优求解,并与传统设计方案进行对比。结果表明:GAPSO-RBFNN的预测误差远低于传统RBFNN;优化后电机吊架各质量特性信噪比得到提高,实现了对电机吊架的多目标稳健优化,降低了电机吊架总质量,提高了其自然频率。Based on the genetic particle swarm optimization(GAPSO) algorithm, the optimal smooth coefficient was obtained, thus the radial basis function neural network(RBFNN) was improved. Through the sensitivity analysis of the motor hanger, the key design variables, which had great influence on the quality characteristics such as total mass and natural frequency, were selected. The quality characteristic values and the corresponding signal-to-noise ratio of the motor hanger were obtained with the orthogonal test design of experiment and finite element analysis. The obtained experimental data were taken as the input and the signal-to-noise ratio as the output for training and testing the GAPSO-RBFNN. The prediction accuracy was compared and analyzed. The multi-objective robust optimization model of motor hanger was constructed based on GAPSO-RBFNN. The multi-objective optimization algorithm NSGA-Ⅱ was used to solve the optimization model, which was compared with the traditional design scheme. Results show that the prediction error of GAPSO-RBFNN is much lower than that of traditional RBFNN. The signal-to-noise ratio for each quality characteristic of the motor hanger is improved after optimization, which means that the multi-objective robust optimization of the motor hanger is realized, the total mass of the motor hanger is reduced, and its natural frequency is increased.

关 键 词：径向基神经网络 遗传粒子群 信噪比 电机吊架 多目标稳健优化 

分 类 号：U270.2[机械工程—车辆工程]