基于CNN与SVM的变压器振动故障自动诊断  

Automatic Diagnosis of Transformer Vibration Fault Based on CNN and SVM

作　　者：薛广华 XUE Guanghua(Qingdao Technician College,Qingdao,Shandong 266000,China)

机构地区：[1]青岛市技师学院,山东青岛266000

出　　处：《自动化应用》2025年第2期62-64,共3页Automation Application

摘　　要：针对变压器故障诊断问题,提出了一种基于一维卷积神经网络(1D-CNN)和支持向量机(SVM)的自动诊断方法。该方法首先通过位移传感器采集变压器工作时的振动信号,随后,采用1D-CNN对振动信号进行特征提取,并利用SVM对提取的特征向量进行故障分类。结果表明,该方法在准确率、召回率和F1分数等指标上均具有良好表现,验证了其在实际应用中的有效性和可靠性。This study proposes an automatic diagnosis method for transformer faults based on One-Dimensional Convolutional Neural Network(1D-CNN)and Support Vector Machines(SVM).This method first collects the vibration signal of the transformer during operation through displacement sensors.Then,1D-CNN is used to extract features from the vibration signal,and SVM is used to classify the extracted feature vectors for faults.The results show that the method has good performance in accuracy,recall,and F1 score,verifying its effectiveness and reliability in practical applications.

关 键 词：卷积神经网络 支持向量机 变压器 振动信号 故障诊断 

分 类 号：TM411[电气工程—电器]