基于深度网络的三相异步电动机故障检测  

Fault Detection of Three-phase Asynchronous Motor Based on Deep Network

作　　者：史玉芳 SHI Yufang(Xuzhou Electromechanical Technician College,Xuzhou 221131,Jiangsu,China)

机构地区：[1]徐州机电技师学院,江苏徐州221131

出　　处：《电气传动自动化》2023年第5期44-46,11,共4页Electric Drive Automation

摘　　要：本文提出了一种深度网络特征的三相异步电动机运行故障自动检测方法。首先利用基于SIFT的关键点匹配算法在热图中检测出关注区域,之后,这些图像基于一个预先训练的卷积神经网络,被转换成具有代表性的特征向量。利用K-means将训练矢量样本聚类为冷簇和热簇,对于每个聚类,训练一个基于SVM的分类器,测试特征向量样本采用相应训练后的SVM分类器进行聚类并映射成类。通过对包括实际热图像在内的数据集进行评估,表明该算法能够100%地检测出异步电动机的故障。This article proposes a deep network feature based automatic fault detection method for three-phase asynchronous motors.Firstly,a key point matching algorithm based on SIFT is used to detect the region of interest in the heat map.The images are then transformed into representative eigenvectors based on a pre-trained convolutional neural network.Then,the training vector samples are clustered into cold and hot clusters by K-means.For each cluster,a SVM-based classifier is trained.The test feature vector samples are clustered and mapped into classes by the corresponding trained SVM classifier.Through the evaluation of the data set including the actual thermal image,it is shown that the fault of induction motor can be detected 100% by this algorithm.

关 键 词：深层网络 三相异步电机 热图像 故障检测 

分 类 号：TM506[电气工程—电器] TP273[自动化与计算机技术—检测技术与自动化装置]