Prediction of bearing failure trend based on composite multiscale divergence entropy and EMD-TCN-GRU

To effectively predict the health status of rolling bearings, a prediction method for the degradation trend of rolling bearings based on composite multi-scale divergence entropy and EMD-TCN-GRU is proposed. This method adopts the composite multi-scale divergence entropy as the foundation to construct degradation features, maps the degradation features to the bearing life in chronological order, and introduces the novel TCN-GRU network in the field of trend prediction to realize the prediction of the degradation trend of rolling bearings. The PHM2012 dataset is introduced for case analysis, and the results indicate that the RMSE of this method can reach 4.8%,the MAE of this method can reach 3.0%, the R² of this method can reach 97.4%. Compared with the traditional composite neural networks, it has greater advantages in terms of computing efficiency and accuracy.