A Hybrid ST-Bilstm Based Prediction Method for Estimating the Life of Ball Mill Rolling Bearings

Aiming at the problems of low prediction accuracy and slow convergence speed of existing ball mill rolling bearing Remaining Useful Life (RUL) prediction models, an RUL prediction method using Temporal Convolutional Network (TCN) and Bidirectional Long and Short-term Memory (BiLSTM) is proposed. Firstly, the Savitzky-Golay (SG) filter is used to remove the interference of random noise in the original vibration signals; then the short-term features of the sequences are extracted using TCN and capture the long-term correlations in the data with the help of BiLSTM; finally, the attention mechanisms assign different weights to features in order to achieve the goal of being able to more accurately extract information about the health state of the device. Theoretical analysis and data validation demonstrate that the method improves the accuracy and adaptability of model series predictions, and the IEEE PHM 2012 dataset is used to verify the effectiveness of the proposed method.