一种基于稀疏自编码器的电机故障诊断方法  

A Motor Fault Diagnosis Method based on Sparse Autoencoders

作　　者：王黎阳 杜翀[2] 汪欣[2] 翟旭平[1] WANG Liyang;DU Chong;WANG Xin;ZHAI Xuping(Laboratory of Specialty Fiber Optics and Optical Access Networks,Shanghai University,Shanghai 200072,China;Shanghai Research Center of Urban Public Safety,Shanghai Advanced Research Institute of Chinese Academic of China,Shanghai 201210,China)

机构地区：[1]上海大学特种光纤与光接入网重点实验室,上海200072 [2]中国科学院上海高等研究院城市公共安全研究中心,上海201210

出　　处：《噪声与振动控制》2019年第5期197-202,249,共7页Noise and Vibration Control

基　　金：中国科学院重点部署资助项目(KFZD-SW-310)

摘　　要：在电机故障诊断研究领域,基于人工智能技术和现代信号处理方法相结合的故障诊断技术正逐步成为目 前的研究热点。一般的模式识别方法往往对信号的数据采集和处理有较高要求,且往往因模型泛化能力有限而受到 制约。为解决这一问题,提出一种基于稀疏自编码器的故障诊断方法,利用Hilbert包络谱信号训练稀疏自编码器,自 适应将大数据的内在特征提炼为简单的特征函数,通过特征函数表达实现电机状况的智能诊断。实验结果表明,相比 BP(Back propagation,BP)神经网络和支持向量机分类算法,本方法可快速、有效地提高故障分类的准确度,对电机故障 精准诊断具有重要意义。In the field of motor fault diagnosis,the fault diagnosis based on artificial intelligence technology and modern signal processing methods is gradually becoming a research hotspot.General pattern recognition methods usually have high requirements for data acquisition and signals processing,and are constrained by the limited generalization ability of models.In order to solve this problem,a fault diagnosis method based on sparse autoencoder is proposed.Hilbert envelope spectrum signal is used to train the sparse autoencoder.The intrinsic features of large data are adaptively refined into simple feature functions,and the intelligent diagnosis of motor status is realized through the expression of feature functions.The experimental results show that compared with back propagation(BP)neural network and support vector machine classification algorithm,this method can improve the accuracy of fault classification quickly and effectively,and is of great significance for accurate diagnosis of motor faults.

关 键 词：故障诊断 稀疏自编码 模式识别 特征提取 

分 类 号：TP206[自动化与计算机技术—检测技术与自动化装置]