A multi-perspective architecture for high-speed train fault diagnosis based on variational mode decomposition and enhanced multi-scale structure

摘要

The performance degradation and failure of high-speed train bogie would directly threaten the safe long-term operation of the vehicle. The fault diagnosis based on vibration signals is encountering difficulties as nonlinearity, high complexity, strong coupling, and high uncertainty. To address these challenges, this paper proposes a multi-perspective architecture for fault diagnosis, based on variational mode decomposition and enhanced multi-scale convolutional neural network. The proposed method provides multiple perspectives for the multi-channel and multi-component signal analysis, including perspectives from channel, component and time scale, with low input dimension and reduced model complexity. Signal features under different perspectives can be adaptively extracted. The effectiveness of the proposed method is validated on high-speed train fault data and rolling element bearings dataset. The experimental results show that the proposed scheme not only improves the accuracy of fault diagnosis but also has superior noise robustness which could be valuable for practical applications of complex systems, especially in dynamic environments.