摘要
针对传统滚动轴承故障诊断方法过度依赖专家经验,故障特征提取及选取困难的问题,提出一种基于集成深度小波神经网络和深度小波支持向量机的滚动轴承故障诊断方法。首先,利用不同的小波函数设计不同的改进小波自编码器,并构造相应的深度小波神经网络;然后,将轴承振动信号输入各深度小波神经网络进行无监督特征学习并进行微调;最后,将每个深度小波神经网络的顶层特征融合,输入深度小波支持向量机分类器实现对轴承故障的自动识别。试验结果表明,该方法能够对滚动轴承进行多工况及多种故障程度的有效识别,特征提取能力和识别能力优于浅层人工神经网络、支持向量机等传统方法以及深度信念网络、深度稀疏自编码器等深度学习模型。
The traditional fault diagnosis methods for rolling bearings have such shortcomings as largely dependent on expertise,difficulty in fault feature extraction and selection,a fault diagnosis method for rolling bearings is proposed based on ensemble deep wavelet neural network(EDWNN)and deep wavelet support vector machine(DW-SVM).Firstly,the different improved wavelet auto-encoders(WAEs)are designed by using different wavelet functions,the corresponding deep wavelet neural networks(DWNNs)are constructed.Then,the vibration signals of the bearings are input into each DWNN for unsupervised feature learning and fine tuning.Finally,the top features of each DWNN are fused,and the DW-SVM classifier is input to realize automatic recognition of bearing faults.The experiment results show that the method recognizes multiple fault severities of rolling bearings under multiple operating conditions effectively,the ability of feature extraction and recognition is superior to that of traditional methods such as shallow artificial neural network,SVM and deep learning models such as deep belief network,deep sparse auto-encoder and so on.
作者
杜小磊
陈志刚
张楠
许旭
DU Xiaolei;CHEN Zhigang;ZHANG Nan;XU Xu(College of Mechanical Electronic and Vehicle Engineering,Beijing University of Civil Engineering and Architecture,Beijing 100044,China;Beijing Engineering Research Center of Monitoring for Construction Safety,Beijing 100044,China)
出处
《轴承》
北大核心
2019年第11期60-67,共8页
Bearing
基金
国家自然科学基金项目(51605022)
北京市教育委员会科技计划一般项目(SQKM201710016014)
北京市优秀人才培养项目(2013D005017000013)
关键词
滚动轴承
故障诊断
深度小波神经网络
深度小波支持向量机
rolling bearing
fault diagnosis
deep wavelet neural network
deep wavelet support vector machine
