基于SS-EWT的IGBT驱动电路早期故障诊断  

Incipient fault diagnosis of IGBT drive circuit based on SS-EWT

作　　者：吴昊 钱存元 WU Hao;QIAN Cunyuan(Institute of Rail Transit,Tongji University,Shanghai 201804,China)

机构地区：[1]同济大学铁道与城市轨道交通研究院,上海201804

出　　处：《微电子学与计算机》2024年第4期1-11,共11页Microelectronics & Computer

基　　金：“十二五”国家科技支撑计划(2015BAG19B02)。

摘　　要：驱动电路是绝缘栅双极性晶体管(Insulated Gate Bipolar Transistor,IGBT)正常运行的重要保障。本文基于模拟电路早期故障理论,针对以EXB841驱动模块为核心的300 A/1200 V快速型IGBT驱动电路,提出了一种基于尺度空间(Scale Space,SS)改进的经验小波变换(Empirical Wavelet Transform,EWT)和误差反向传播神经网络(Back Propagation Neural Network,BPNN)相结合的早期故障诊断方法。首先在分析驱动电路硬件结构及工作原理的基础上,对其驱动功能和短路保护功能的主要元件进行灵敏度分析,分别选取相应的早期故障诊断元件;其次通过故障注入获取检测点输出数据,采用SS-EWT对数据处理后构造早期故障特征向量;最后采用BPNN对特征向量进行训练以实现对早期故障模式的识别。针对短路保护元件部分故障类别难以区分的问题,采用再训练对诊断方法进行优化。仿真结果表明,采用基于SS-EWT的早期故障诊断方法处理驱动功能元件及短路保护元件早期故障时准确率分别能达到91.04%和96.05%。Drive circuit is the guarantee for the normal operation of Insulated Gate Bipolar Transistor(IGBT).Based on the incipient fault theory of analog circuit,an incipient fault diagnosis method based on Scale Space Empirical Wavelet Transform(SS-EWT)combined with Back Propagation Neural Network(BPNN)for 200 A/1200 V fast IGBT driver circuit based on EXB841 driver module is proposed.Firstly,based on the analysis of the hardware structure and working principle of the drive circuit,the sensitivity analysis of the main components of the drive function part and short-circuit protection function part is carried out to select the test objects.Secondly,the data obtained by fault injection were processed by SS-EWT and constructed into the incipient fault eigenvector.Finally,BPNN is used to train the eigenvector to recognize the incipient fault.In view of the difficulty in distinguishing some fault categories of short-circuit protection elements,retraining is used to optimize the diagnosis method.The simulation results show that the incipient fault diagnosis method based on SS-EWT for driving function elements and short-circuit protection elements can reach the accuracy of 91.04%and 96.05%,respectively.

关 键 词：尺度空间 经验小波变换 BP神经网络 IGBT驱动电路 早期故障诊断 

分 类 号：TN322.8[电子电信—物理电子学]