摘要
针对机械设备故障数据大容量、多样性的特点,提出一种基于堆叠稀疏自编码(SSAE)的滚动轴承故障智能诊断方法。使用自动编码器(AE)逐层训练网络,从海量数据中自适应地学习各类故障的特征表达,再通过有监督的反向传播算法优化整个网络,最终将特征输入softmax分类器实现滚动轴承健康状况精确诊断。在动力传动故障诊断试验台采集了5类轴承故障数据进行测试。试验结果表明:SSAE算法能够有效地提取故障特征,且故障诊断效果优于传统智能诊断方法。
Aiming at high capacity and diversity of fault data of mechanical equipment,an intelligent fault diagnosis method for rolling bearings is proposed based on Stacked Sparse Autoencoder( SSAE). The network is trained layer by layer with autoencoder( AE),and the feature expression for all kinds of fault is learned self-adaptively from massive data. Then the entire network is optimized through supervised back propagation algorithm. The feature is inputted into softmax classifier to realize accurate diagnosis of health condition of rolling bearings. Five kinds of bearing fault data are collected to test from power drive fault diagnosis test bed. The result indicates that the SSAE algorithm is able to effectively extract fault feature,and the fault diagnosis effect is better than that of traditional intelligent diagnosis methods.
出处
《轴承》
北大核心
2018年第3期49-54,60,共7页
Bearing
基金
国家自然科学基金项目(51505234
51405241)
关键词
滚动轴承
深度学习
堆叠稀疏自编码算法
故障诊断
rolling bearing
deep learning
stacked sparse autoencoder algorithm
fault diagnosis