Data-Driven Prediction Method for Remaining Useful Life of Rolling Bearings

Abstract As a critical component of maritime vessels, bearings face various extreme environments and challenges, making them susceptible to high loads and adverse working conditions. Predictive analysis of bearing life allows maritime maintenance teams to take proactive measures, such as actively replacing or repairing damaged bearings, thereby preventing sudden failures and ensuring the reliability and stability of the vessel. To achieve accurate prediction of the remaining life of rolling bearings, this study proposes a novel two-stage strategy for life prediction. In the first stage, time-domain and frequency-domain features are extracted from the original vibration signals of rolling bearings. Subsequently, an adaptive feature selection method autonomously determines the dimensions of the feature subset, identifying the optimal feature subset. In the second stage, the optimal feature subset is input into the proposed ResNet-BiLSTM-ATT model for predicting the remaining useful life of bearings. The proposed prediction strategy has been validated on two public datasets. The results indicate that the proposed data-driven prediction method can accurately predict the remaining life of rolling bearings. This study holds significant practicality in ensuring navigation safety, optimizing maintenance management, and reducing energy wastage.