Remaining Useful Life Prediction of Rolling Element Bearings Based on an Advanced Two-Stage Degradation Model

Bearings are essential to rotating machinery, and malfunctioning might result in unscheduled downtime, significant financial losses, and serious safety risks. Thus, predictive maintenance depends on precisely estimating the bearings’ remaining useful life (RUL). This study introduces a novel approach for early-stage degradation detection and RUL estimation by integrating a Joseph-form Kalman filter for reliable health indicator denoising and the Akaike information criterion for precise detection of the first change point. The proposed method employs an advanced two-stage approach based on the Wiener process with drift to model degradation. The model parameters are iteratively estimated and updated using the least-squares method to accurately capture the degradation trajectory. Experimental validation on the Pronostia bearing dataset demonstrates that the proposed method achieves the lowest MAE, RMSE, and MAPE values under different operating conditions, confirming its superior prediction accuracy and reliability compared with traditional statistical and model-based methods.