Intelligent discrimination method for the closed state of relay contacts in high-speed rail trains

Abstract

To ensure the operational safety of high-speed trains, accurate assessment of relay contact closure is critical. This paper proposes a physical- data fusion approach for intelligent monitoring and diagnosis to overcome the inefficiency and safety risks of manual inspection. A wireless embedded system with a 16 × 16 flexible pressure sensor array is developed to capture real-time pressure distribution during contact closure. A feature extraction method combining spatiotemporal processing and a damped oscillation model is used to derive multidimensional features from raw pressure data. A 3D convolutional neural network and a support vector machine with handcrafted features are constructed for contact status classification. Experimental results show that the 3DCNN achieves 99.5% accuracy and the SVM achieves 94.8%, demonstrating the effectiveness of both models in distinguishing good from poor contact conditions. This study provides a new technical pathway for efficient and intelligent relay maintenance in high-speed rail applications.