电工技术学报  2016, Vol. 31 Issue (17): 102-112    DOI:
电力电子与电力传动 |
基于小波包分析和概率神经网络的电磁法三电平变换器故障诊断方法
于生宝,何建龙,王睿家,李刚,苏发
吉林大学仪器科学与电气工程学院 长春 130026
Fault Diagnosis of Electromagnetic Three-Level Inverter Based on Wavelet Packet Analysis and Probabilistic Neural Networks
Yu Shengbao, He Jianlong, Wang Ruijia, Li Gang, Su Fa
College of Instrumentation & Electrical Engineering Jilin University Changchun 130026 China
全文: PDF (3201 KB)  
输出: BibTeX | EndNote (RIS)      
摘要 针对基于三电平变换器的电磁法发射机中功率开关器件开路故障特点和复杂工作环境,提出了针对性的故障诊断方法。该方法以变换器输出电压为原始信号,利用变采样频率的小波包分析方法提取特征向量,以提高对信号频率的分辨准确度。然后利用核主成分分析对特征向量进行降维,可以简化分类器的结构,提高诊断时间。采用概率神经网络建立故障分类器,可以提高诊断方法的鲁棒性。在一台5 kW电磁法三电平变换器实验样机上进行实验和分析,实验结果表明该方法可以准确地进行故障诊断,有较好的诊断准确度、实时性和较强的鲁棒性,具有一定的工程应用价值。
服务
把本文推荐给朋友
加入我的书架
加入引用管理器
E-mail Alert
RSS
作者相关文章
于生宝
何建龙
王睿家
李刚
苏发
关键词 三电平变换器电磁法发射机小波包分析核主成分分析概率神经网络故障诊断    
Abstract:According to the fault characteristics and complex working environment of electromagnetic transmitters,a novel fault diagnosis method is proposed to solve the open-circuit fault of three-level inverters based electromagnetic transmitters.the wavelet packet analysis (WPA) is adopted to extract the feature vector of the open-circuit fault output voltage which is used as the original signal.The sample frequency is set to be variable to improve the frequency resolution of WPA.The kernel principal component analysis (KPCA) is employed to reduce the dimension of feature vector,which can simplify the structure of classifier and decrease the time of diagnosis.The probabilistic neural network (PNN) has strong fault tolerance and can be used to establish the fault classifier.A 5 kW laboratory prototype has been built.Experimental results show that the method can effectively diagnose the open-circuit fault and realize real-time fault diagnosis,which has good diagnosis accuracy and strong robustness.And the proposed method has some contribution in practical applications.
Key wordsThree-level inverter    electromagnetic transmitter    wavelet packet analysis    kernel principal component analysis    probabilistic neural networks    fault diagnosis   
收稿日期: 2015-05-13      出版日期: 2016-09-18
PACS: TP277  
基金资助:国家高技术研究发展(863)计划资助项目(2012AA09A20103)。
作者简介: 于生宝 男,1963年生,教授,博士生导师,研究方向为功率源技术与应用。E-mail:yushengbao@jlu.edu.cn;李 刚 男,1981年生,讲师,研究方向为电力电子变换器控制及其在电磁探测中的应用。E-mail:ligang2013@jlu.edu.cn(通信作者)
引用本文:   
于生宝,何建龙,王睿家,李刚,苏发. 基于小波包分析和概率神经网络的电磁法三电平变换器故障诊断方法[J]. 电工技术学报, 2016, 31(17): 102-112. Yu Shengbao, He Jianlong, Wang Ruijia, Li Gang, Su Fa. Fault Diagnosis of Electromagnetic Three-Level Inverter Based on Wavelet Packet Analysis and Probabilistic Neural Networks. Transactions of China Electrotechnical Society, 2016, 31(17): 102-112.
链接本文:  
https://dgjsxb.ces-transaction.com/CN/Y2016/V31/I17/102