基于GAN-LSTM的PMSM逆变器开路故障诊断研究  

Research on Open Circuit Fault Diagnosis of PMSM Inverter Based on GAN-LSTM

作　　者：黄潇 冯莉 罗洪林 HUANG Xiao;FENG Li;LUO Hong-lin(Chongqing Jiaotong University,Chongqing 400074,China)

机构地区：[1]重庆交通大学,交通运输学院,重庆400074

出　　处：《电力电子技术》2024年第12期49-53,共5页Power Electronics

基　　金：重庆市研究生教育教学改革研究项目(yjg213094)。

摘　　要：针对电机驱动系统非平稳时序故障信号数量稀少且特征精细刻画难以及诊断精确度差等问题,提出了一种生成对抗网络(GAN)和长短期记忆(LSTM)网络结合的故障诊断方法。首先,采用信号脉冲控制策略模拟三相逆变器开路故障不同类型。其次,利用多尺度特征提取方法构建高维故障特征数据集。然后,在传统鉴别器上增添了基于LSTM的二层故障诊断鉴别层。最后,基于所提方法完成对永磁同步电机(PMSM)驱动系统逆变器开路故障诊断。实验结果表明,该方法的分类准确率可达98.92%,通过与其他故障诊断方法比较,验证了所提方法的优越性和有效性。A fault diagnosis approach based on generative adversarial network(GAN)and long short-term memory(LSTM)network is proposed to solve the problem in which the number of non-stationary time series fault signals of the motor drive system is rare,the characteristics are difficult to describe and the diagnostic accuracy is poor.Firstly,the different types of open circuit faults of three-phase inverters are simulated by signal pulse control strategy.Secondly,the multiscale feature extraction method is used to construct a high-dimensional fault feature data set.Then,a two-layer fault diagnosis identification layer based on LSTM is added to the traditional discriminator.Finally,based on the proposed method,the open circuit fault diagnosis of the inverter of the permanent magnet synchronous motor(PMSM)drive system is completed.The experimental results show the classification accuracy of this method can reach 98.92%.By comparing with other fault diagnosis methods,the superiority and effectiveness of the proposed method are verified.

关 键 词：永磁同步电机 逆变器 二层故障诊断鉴别层 

分 类 号：TM341[电气工程—电机]