基于GA-PSO混合优化SVM的机载EHA故障诊断

Fault Diagnosis of Airborne EHA Based on GA-PSO Hybrid Optimization SVM

针对机载电静液作动器(Electro-Hydrostatic Actuator,EHA)的典型故障,详细分析了故障原理并在MATLAB/Simulink中搭建了仿真模型。为了高效准确识别故障类型,提出一种用遗传算法(Genetic Algorithm,GA)和粒子群算法(Particle Swarm Optimization,PSO)混合优化支持向量机(Support Vector Machine,SVM)的故障诊断算法。GA鲁棒性好且全局搜索能力强但收敛速度慢,PSO对样本规模不敏感且具有记忆功能但易陷入局部最优,故融合两种算法寻找SVM的最优参数。另外,为了解决传统SVM多分类方法“一对多”和“一对一”易出现不可分的问题,建立一种偏二叉树结构的SVM多分类模型。对于采集的原始数据高度重合的情况,引入时域特征统计量进一步提升模型的分类性能。实验结果表明,提出的混合优化算法寻优速度更快、所寻参数更佳,同时用该算法优化的SVM分类模型相比于其他5类常用的机器学习模型分类效果更好,故障识别正确率可达97.7%。

Aiming at the typical faults of airborne electro-hydrostatic actuator(EHA),the fault principle was analyzed and the simulation model was established in MATLAB/Simulink.To identify fault types efficiently,a fault diagnosis algorithm based on genetic algorithm(GA)and particle swarm optimization(PSO)hybrid optimization support vector machine(SVM)was proposed.GA has good robustness and global search ability,but its convergence speed is slow.PSO is insensitive to sample size and has memory function,but it’s easy to fall into local optimization.Thus,the two algorithms were combined to find the optimal parameters of SVM.In addition,the traditional multi-classification methods of SVM‘one to many’and‘one to one’are easy to be inseparable,to solve the problem,a method with partial binary tree structure was proposed.For solving the collected original data were highly coincident,time domain feature statistics were introduced to improve the classification performance.The experimental results show the proposed GA-PSO has faster optimization speed and better parameters.At the same time,the classification effect of SVM optimized by this algorithm is better than other five commonly used machine learning models,and the accuracy of fault identification is 97.7%.