A Remaining Useful Life Prediction Method for Rolling Bearing Based on TCN-Transformer

Abstract

Predicting the remaining useful life (RUL) of rolling bearings is crucial to ensure the stable operation of equipment. In recent years, predictive methodologies that leverage intelligent models have witnessed widespread development, significantly enhancing the precision of equipment prognostication. However, operating environments are inherently complex and can cause stochastic fluctuations in the characteristic indicators extracted during the rolling bearing degradation stage, leading to uncertainty in prediction outcomes. This study presents a TCN-transformer model and a two-stage degradation feature optimization methodology to address these challenges. The first stage uses Kalman filtering to suppress abnormal noise in the degradation index. In the second stage, a nonlinear smoothing algorithm based on degradation trends was constructed to improve the performance of degradation indicators. The proposed method constructs more stable and reliable degradation indicators. Additionally, to improve prediction accuracy, a TCN-transformer rolling bearing lifespan prediction model is proposed. Probability prediction and interval prediction are incorporated into rolling bearing RUL prediction to enhance the reliability of the model. Finally, the effectiveness of the proposed method is validated on the publicly available dataset XJTU-SY.