基于深度度量学习的轴承故障诊断方法  

Bearing fault diagnosis method based on deep metric learning

作　　者：李小娟[1,2] 徐增丙 熊文 王志刚[1,2] 谭俊杰[1,2] LI Xiaojuan;XU Zengbing;XIONG Wen;WANG Zhigang;TAN Junjie(Key Laboratory of Metallurgical Equipment and Control Technology of Ministry of Education,Wuhan 430081,China;Wuhan University of Science and Technology college of machinery and automation,Wuhan 430081,China;Key Laboratory of Waterjet Propulsion Technology,No.708 Research Institute,China State Shipbuilding Corporation,Shanghai 200011,China)

机构地区：[1]冶金装备及其控制教育部重点实验室,武汉430081 [2]武汉科技大学机械自动化学院,武汉430081 [3]中国船舶工业集团公司第708研究所,喷水推进技术重点实验室,上海200011

出　　处：《振动与冲击》2020年第15期25-31,共7页Journal of Vibration and Shock

基　　金：国家自然科学基金(51775391);装备预研基金(6142223180312)。

摘　　要：针对机械大数据因故障类内离散度和类间相似度较大而导致诊断精度低的问题,提出一种深度度量学习故障诊断方法,采用深度神经网络(Deep Neural Network, DNN)对故障特征进行自适应提取,并利用基于欧氏距离的边际Fisher分析(Marginal Fisher Analysis, MFA)方法进行了优选,在构建的深度度量网络(Deep Metric Network, DMN)顶层特征输出层添加BPNN(Back Propagation Neural Network, BPNN)分类器对网络参数进行微调,并实现故障的分类识别。通过对不同类型和严重程度的轴承故障进行了诊断分析,验证了该方法可以有效地对轴承故障进行高精度诊断,效果优于传统深度信念网络(Deep Belief Network, DBN)故障诊断方法以及常用时域统计特征结合支持向量机(Support Vector Machine, SVM)分类的故障诊断方法。Aiming at problems of larger within-class scatter and inter-class similarity in bearing fault data causing lower diagnosis accuracy, a new bearing fault diagnosis method based on deep metric learning was proposed. The deep neural network(DNN) was used to adaptively extract fault features, and the marginal Fisher analysis method based on Euclidean distance was used to optimize the extracted features. Then, the back propagation neural network(BPNN) classifier was added to the top-level feature output layer of the constructed DMN to slightly adjust the network parameters, and realize fault classification and recognition. Finally, diagnosis analysis was performed for different types and severities of bearing faults. It was shown that the proposed method can be used to effectively diagnose bearing faults with high precision;its effect is better than those of the traditional deep belief network(DBN) fault diagnosis method and the common time-domain statistical features combined to support vector machine(SVM) classification one.

关 键 词：深度度量学习 轴承 故障诊断 相似度 

分 类 号：TP165+.3[自动化与计算机技术—控制理论与控制工程] TH133.33[自动化与计算机技术—控制科学与工程]