Remaining life prediction of automotive high voltage relay based on LSTM

As an essential component of new energy vehicles, the lifespan of automotive high-voltage relays directly impacts the reliability and driving safety of automotive electrical and electronic products. This paper presents a methodology for predicting the lifespan of automotive high-voltage relays. To address the significant noise issues, present in the measured high-voltage relay parameters from experimental data, this study proposes the use of a wavelet analysis algorithm for parameter pre-processing. A standard one-dimensional convolutional neural network (CNN) algorithm is initially employed to conduct lifespan prediction experiments for high-voltage relays. Additionally, considering the temporal characteristics of the actual degradation data, a Long Short-Term Memory (LSTM) network algorithm is utilized to extract these temporal features for more accurate lifespan prediction. To fully verify the performance of these two algorithms in predicting the lifespan of high-voltage relays, comparative experiments and analyses were conducted using root-mean-square and other errors as evaluation metrics. The results demonstrate that the LSTM network significantly outperforms the CNN in this application.