Diagnosis Method of PMSM Inter-turn Short Circuit Based on Dual Feature Fusion

Abstract

The early fault features of the stator inter-turn short circuit in permanent magnet synchronous motors (PMSM) are weak. A single electrical or mechanical signal is difficult to fully characterize the fault state, and the traditional threshold diagnosis method has poor generalization ability under different working conditions and inaccurate fault calibration values. This paper proposes a weighted fusion diagnosis method based on dual features and attention mechanism. First, the stator current and casing vibration signals of the motor are collected. Then, the improved sparrow search algorithm is used to optimize the variational mode decomposition for the adaptive decomposition of the dual feature signals, and the intrinsic mode functions are extracted respectively. Next, a two-branch convolutional neural network is constructed, and the modal fusion attention mechanism is introduced to dynamically weight and fuse the dual features. The fully connected layer and output layer of the bidirectional long short-term memory network at the back are used to classify the fused features of the current and vibration. The experimental results show that the diagnostic accuracy of the proposed method reaches 95.1% under five fault conditions, which is better than the single-feature diagnosis model and other fusion models. The research verifies the effectiveness of the proposed method in improving the fault diagnosis accuracy.