深度学习在电机故障诊断中的研究现状综述  

Review of Status of Research on Deep Learning in Motor Fault Diagnosis

作　　者：陈海宇 贺珂珂 王惠中[2,3] Chen Haiyu;He Keke;Wang Huizhong(Department of Public Basic Courses,Zhaoqing Medical College,Zhaoqing Guangdong 526020,China;College of Electrical and Information Engineering,Lanzhou University of Technology,Lanzhou Gansu 730050,China;Gansu Province Key Laboratory for Advanced Control of Industrial Processes,Lanzhou Gansu 730050,China)

机构地区：[1]肇庆医学高等专科学校公共基础部,广东肇庆526020 [2]兰州理工大学电气工程与信息工程学院,甘肃兰州730050 [3]甘肃省工业过程先进控制重点试验室,甘肃兰州730050

出　　处：《电气自动化》2020年第3期1-3,共3页Electrical Automation

基　　金：2019年广东省高等教育学会高职高专云计算与大数据委员会考研规划开题(GDYJSKT19-18);2019年肇庆市科技创新指导类项目(201904030401),“基于深度学习的唇裂修复预测算法模型研究”。

摘　　要：深度学习模型在特征提取与模型拟合方面凸显优势,模型基于多层神经网络的层次结构,可以自动学习表达数据本质与隐含规律的特征,同时克服传统学习方法手工设计特征算子的局限性,显示其优势。因此,将深度学习应用于电机故障诊断领域有一定意义。为此,详细介绍了几种典型模型的原理及在故障诊断领域的研究现状,指出了深度学习存在的问题与未来发展趋势。Deep learning model has prominent advantages in the field of feature extraction and model fitting.The model,developed with a hierarchical structure based on the multi-layer neural network,has such advantages as automatic learning of the characteristics expressing data essence and implicit laws as well as overcoming of limitations of manual design of characteristic operators in the traditional leaning method.Therefore,it is of certain significance to apply deep leaning to the field of motor fault diagnosis.The principles of several typical models and the current status of research on their application in the field of fault diagnosis were introduced in detail,and some problems with deep learning and its future development trend were pointed out.

关 键 词：深度学习 电机故障诊断 神经网络 特征提取 

分 类 号：TH165.3[机械工程—机械制造及自动化] TP183[自动化与计算机技术—控制理论与控制工程]