基于卷积神经网络的牵引电机定子绕组匝间短路故障诊断

Fault Diagnosis of Stator Winding Inter-turn Short Circuit in Traction Motors Based on Convolutional Neural Network

为实现牵引电机定子绕组匝间短路故障诊断,提出一种基于一维卷积神经网络(one-dimensional convolutional neural network, 1D-CNN)的故障诊断方法。首先对电机健康状态、不同相发生匝间短路故障及不同故障严重程度下的定子电流进行三层小波分解,得到小波分解高频系数和低频系数;求取小波分解系数的二范数,作为电机电流的特征;设计并训练1D-CNN,将训练好的1D-CNN作为分类器,实现牵引电机定子绕组匝间短路故障“端到端”的智能诊断。设计并搭建异步电机定子绕组匝间短路故障诊断实验平台。实验结果表明:所提方法可以准确有效诊断出轻微的匝间短路故障。在闭环控制下,电机发生1匝短路故障时,诊断正确率达到90.5%,并能够有效区分故障相。

In order to realize the diagnosis of inter-turn short circuit faults in traction motor stator winding,an intelligent fault diagnosis method was proposed based on one-dimensional convolutional neural network(1D-CNN).Firstly,a discrete three-layer wavelet transform was applied to the motor stator currents in both healthy status of the motor and the cases of inter-turn short circuit faults in different phases and at different fault severity levels using Daubechies-8 wavelet,to obtain high-frequency and low-frequency wavelet decomposition coefficients.Next,the L 2 norm of the coefficients was calculated to represent the features of the traction motor currents.Lastly,the 1D-CNN model was designed,trained,and used as a classifier to achieve“end-to-end”intelligent fault diagnosis of inter-turn short circuit in the traction motor stator winding.A test platform was designed and built for the diagnosis of inter-turn short circuit faults in induction motor stator winding.The results demonstrate that the method can accurately and effectively diagnose minor inter-turn short circuit faults in the stator winding.Under closed-loop control,in the case of one-turn short-circuit fault in the motor,the diagnostic accuracy reaches 90.5%,with the fault phase being effectively distinguished.