Remaining Useful Life Prediction of High-Voltage DC Contactors Based on Acoustic-Electrical Signal Feature Fusion and TCN-Transformer

A remaining useful life (RUL) prediction method for high-voltage (HV) DC contactors in new energy vehicles is proposed, based on acoustic-electrical signal feature fusion and TCN-Transformer. Since the contactors are typically located within vehicle HV boxes or battery packs, insulation degradation is used as the failure criterion. By monitoring the electrical signals of the contacts and acoustic signals during closing throughout the operation of the contactor, on the basis of the conventional features of performance degradation, the features of acoustic signals are explored to enable a diversified analysis of the performance degradation process. Key features reflecting performance degradation are selected and fused using Multidimensional Scaling (MDS), resulting in a performance degradation factor. This factor shows a strong correlation with real-time insulation resistance measurements, supporting degradation analysis that incorporates insulation characteristics. Using hierarchical clustering (HC) algorithm based on the degradation factor, multi-stage degradation analysis is conducted to accurately identify the final stage of degradation. A feature vector for RUL prediction is then constructed by combining key features and the degradation factor, and the TCN-Transformer model is applied to predict the RUL of the contactors at the final stage of degradation. Experimental results demonstrate the effectiveness of the proposed method in quantitatively predicting the RUL of onboard contactors.