A Digital Twin System for Two-Stage PMSM Rotor and Bearing Faults Identification Based on Deep Learning and Improved-RGB Acoustic Image

This article proposes a diagnosis method for six types of rotor and bearing faults of permanent magnet synchronous motor based on acoustic signals, and builds a digital twin system for two-stage diagnosis on the basis of an Internet of Things component and the cloud. The first stage is located at the edge of the motor, only short-term Fourier transform and convolutional neural network are used to distinguish fault state from healthy state. By comparing the performance of different frequency bands, a scheme with the lowest hardware requirements is optimized and designed, reducing the number of network nodes by 100 times and the sampling frequency by 12 times. The second stage is located in the cloud, which fully utilizes the data-efficient image transformer to recognize the multidimensional images converted from acoustic signals, thereby achieving the identification of different faults. These images contain ten acoustic features, which are specifically layered and arranged in the Red, Green, and Blue channels of the image to improve the accuracy of fault identification and anti-interference ability. Finally, experimental verification shows that the diagnostic scheme exhibits excellent performance in terms of accuracy and computational resource utilization.