A Remaining Useful Life Prediction Method for Rolling Bearing Based on Multi-channel Fusion Hierarchical Vision Transformer

This study proposes an RUL prediction method based on a multi-channel fusion hierarchical vision transformer to effectively improve the remaining useful life (RUL) prediction accuracy and timely replace the critical components of rotating machinery represented by rolling bearings. In this innovative network architecture, a multi-channel of hierarchical vision transformer is used to extract high degradation correlation features efficiently. The introduction of multi-headed attention channel fusion can effectively reduce the redundancy effect of multi-channel depth features and construct a better regression mapping relationship to improve the accuracy of prediction results. Through experimental verification using rolling bearing samples under different operating conditions, the improved method is able to show higher prediction accuracy in more degradation processes. It can improve prediction accuracy by up to 28.48% compared to the classical RUL method. This indicates that the improved method can be better applied to the rolling bearing RUL prediction problem in practical engineering.