Application of Parameter Evaluation and PSO-BP Neural Network for Relay Contact Life Prediction

In light of the challenge of selecting performance parameters with high degradation sensitivity and accurately predicting the remaining life of relay contacts, this study proposes a method for evaluating performance parameters based on correlation, monotonicity, and dispersion. Multiple electrical performance parameters of AgNi were collected during the contact material test process and preprocessed using wavelet transform to reduce data randomness and improve prediction accuracy. Correlation, monotonicity, and discreteness indicators were employed to evaluate the preprocessed parameters, with the PSO-BP (Particle Swarm Optimization - Back Propagation) neural network used to predict the life of the contacts. The validity of the performance parameter evaluation indicators during life prediction was confirmed by the results obtained. The findings highlight that the three evaluation indices accurately reflect the degradation sensitivity of performance parameters, thus enabling the selection of performance parameters with higher prediction accuracy.