基于PCA和ACA改进ELM算法的齿轮箱故障诊断  

Gearbox Fault Diagnosis Based on PCA and ACA Improved ELM Algorithm

作　　者：余震[1] 于春霞[1] 张建国[2] YU Zhen;YU Chunxia;ZHANG Jianguo(Department of Computer Science,Yellow River Institute of Science and Technology,Zhengzhou 450000,China;School of Mechanical Engineering,Henan Polytechnic University,Zhengzhou 450000,China)

机构地区：[1]黄河科技学院计算机系,郑州450000 [2]河南理工大学机械工程学院,郑州450000

出　　处：《机械设计与研究》2022年第6期94-97,共4页Machine Design And Research

基　　金：河南省民办高等学校品牌专业建设项目(ZLG201903)。

摘　　要：齿轮箱故障诊断对保障传动效率具有重要的意义。为了弥补极限学习机(ELM)收敛速度慢且易陷入局部最优的缺点,设计了一种基于核主成分分析(PCA)和蚁群算法(ACA)改进ELM方法。选择PCA实现特征矩阵降维处理,并构建以PCA特征提取和蚁群优化ELM方法,对齿轮箱故障诊断。本方法选择蚁群算法优化ELM随机生成输入层权值和偏置,从而获得更准确的预测结果。研究结果表明:通过降维处理去除其中的无效数据,经降维处理可以使ELM分类精度由89%上升至93%。通过ACA优化ELM权值能够更精确判断点磨和磨损的故障情况,使ELM达到更高精度。ACA-ELM模型是通过ELM优化,实现了诊断精度的明显提升。ACA-ELM模型在预测过程中只形成了很低幅度的波动,具备更高诊断精度,实现ELM稳定性和分类性能的优化,满足实时性控制条件。Gearbox fault diagnosis plays an important role in ensuring transmission efficiency.In order to make up for the disadvantages of extreme learning machine(ELM),such as slow convergence rate and tendency to fall into local optimum,an improved ELM method based on kernel principal component analysis(PCA)and ant colony algorithm(ACA)is designed.PCA is chosen to reduce the dimension of the feature matrix,and the ELM method based on PCA feature extraction and ant colony optimization is constructed to diagnose the gearbox fault.This method selects ant colony algorithm to optimize ELM to randomly generate input layer weights and bias,so as to obtain more accurate prediction results.The results show that the classification accuracy of ELM can be increased from 89%to 93%by removing invalid data through dimension reduction.ACA optimization of ELM weight can more accurately judge the fault conditions of point wear and wear,so that ELM can achieve higher accuracy.By optimizing ELM,a significant improvement in diagnostic accuracy is achieved.The ACA-ELM model only forms very low amplitude fluctuation in the prediction process,which has higher diagnostic accuracy,realizes the optimization of ELM stability and classification performance,and meets the real-time control conditions.

关 键 词：齿轮箱 故障诊断 核主成分分析 蚁群算法 极限学习机 

分 类 号：TH162[机械工程—机械制造及自动化]