Fast diagnosis of motor controller faults based on OCSSA-SVMD with CNN-BiLSTM

In response to the problems of diverse fault types, difficult feature extraction, as well as the low accuracy and efficiency of traditional diagnostic methods for motor controllers under complex working conditions, this study proposes an intelligent fault diagnosis method based on the Osprey–Cauchy–sparrow search algorithm optimized Sequential Variational Mode Decomposition (SVMD) and Convolutional Neural Network–Bidirectional Long Short-Term Memory (CNN-BiLSTM) network model. First, a fault test platform for the motor controller is built to collect multi-sensor operational data under various typical working conditions. Second, an improved sparrow search algorithm, which incorporates logistic chaotic mapping, Osprey optimization algorithm, and Cauchy mutation strategy, is employed to adaptively optimize the key parameters of variational SVMD using the minimum envelope entropy as the criterion, so as to obtain the representative intrinsic mode functions and realize the effective decomposition and feature extraction of complex signals. Finally, a feature vector is constructed by combining the time–frequency domain features, and the feature vector is input into the CNN-BiLSTM network model to realize the fault diagnosis of the motor controller. The experimental results demonstrate that the proposed method achieves an accuracy of 98.50%, a precision of 97.85%, and an average running time of less than 10 s in motor controller fault diagnosis tasks. Compared with the CNN-BiLSTM model, the accuracy and precision are improved by 13% and 12.64%, respectively, while the diagnostic efficiency is also significantly enhanced. The research results confirm the effectiveness of the proposed method in the fault diagnosis of the motor controller.