Abstract This paper presents an efficient technique for early diagnosis of simultaneous faults in different phases of stator winding of a three-phase induction motor due to turn-to-turn short circuit. A real-life motor has been designed and manufactured with fault emulation features in all the phases of stator winding. Phase currents are recorded by a data acquisition system for different fault conditions. Wavelet kernel-based convolutional neural network (WK-CNN) has been employed for identification and classification of the faults using the recorded current signatures. Various mother wavelets have been tested as convolution filters to extract salient features from the recorded current signatures followed by updating the weights of the filter at each epoch by a supervised learning algorithm. The reason to use a deep framework based on CNN is that it eliminates the requirement of feature extraction and classification algorithms separately. The proposed method also shows promising results when signals are contaminated by the noises, which is always a challenge in an industrial environment. Comparative results show the effectiveness of the proposed technique over the state-of-the-art methods.

中文翻译：

---

基于小波核的卷积神经网络识别和分类感应电机定子匝间故障

摘要 本文提出了一种有效的技术，用于早期诊断三相异步电动机定子绕组不同相由于匝间短路引起的同时故障。真实电机的设计和制造在定子绕组的所有阶段都具有故障仿真功能。数据采集​​系统针对不同的故障条件记录相电流。基于小波核的卷积神经网络 (WK-CNN) 已被用于使用记录的电流特征来识别和分类故障。各种母小波已被测试为卷积滤波器，以从记录的当前特征中提取显着特征，然后通过监督学习算法在每个时期更新滤波器的权重。使用基于 CNN 的深度框架的原因是它消除了特征提取和分类算法分开的要求。当信号被噪声污染时，所提出的方法也显示出有希望的结果，这在工业环境中始终是一个挑战。比较结果显示了所提出的技术相对于最先进方法的有效性。