Fault Diagnosis of Misalignment Based on Wavelet and Attention Convolutional Neural Network

Servo motor and transmission device are important equipment in modern industrial production. However, after long-term use, it is inevitable that faults will occur. Therefore, identifying faults in equipment holds significant importance in preventing unforeseen losses. In this paper, we present a data-driven intelligent fault diagnosis approach for parallel misalignment based on a novel continuous wavelet transform-attention convolutional neural network (CWT-ACNN) model. Firstly, the current signals are collected and labeled as 8 fault types, and then Morlet wavelet transform is used to convert the current signals of time series into time-frequency feature images. After changing the size of the images, they are fed into a convolutional neural network with an attention mechanism to identify fault types. Experimental results show that the CWT-ACNN proposed in this paper can achieve 99.79% accuracy, which is better than the existing methods for fault classification.