Fault Diagnosis of Suspension Controller for High-Speed Maglev Train Based on MTNet

Potential faults in the suspension controller of high-speed maglev trains can affect the stability and reliability of the entire vehicle's operation. Due to the diversity of data in the suspension system, it is challenging to effectively diagnose faults in the suspension controller. This paper proposes a fault diagnosis model that integrates a Multi-scale Transformer Networks (MTNet). The model designs a multi-scale feature extraction network that uses multiple convolutional kernels to extract temporal dependencies at different scales in multivariate time series, capturing the temporal characteristics of suspension system data. Then, the Transformer model enhances the contextual relationships within the time series in sliding windows, ultimately achieving fault diagnosis for the suspension controller. Experimental results demonstrate that the proposed method can accurately diagnose faults in the suspension controller.