A Two-Stage Adaptive Method for Predicting the Remaining Useful Life of Winders

Winder is a key equipment in the chemical fiber production line. It is crucial to predict the remaining useful life (RUL) of the winder to ensure its stable operation. However, the RUL prediction of the winder is faced with the challenges of early weak failures and time-varying operating speed. This paper proposes a two stage RUL prediction method combing winder feature enhancement and attention mechanism. Specifically, it includes the degradation point identification method for enhancing the features and the attention mechanism enhancement method for predicting the RUL. When performing continuous wavelet transform on the vibration signal of the winder, a time-frequency enhancement factor is added to the wavelet coefficient, which can be adaptively adjusted the sensitivity of degradation point identification. An attention module with global average pooling and global max pooling to the time sequence feature extraction model is added to enhance adaptability to variable speeds. Finally, the two-stage RUL prediction method is used to predict the RUL at different stages. The experimental results show that the proposed method can accurately predict the RUL, indicating its strong adaptability to early weak faults and time-varying speeds of the winder.