Rolling bearing remaining useful life prediction method based on vibration signal and mechanism model

Variations in operating conditions and usage environments, bearings often exhibit multiple degradation modes (DMs). Existing degradation models fail to inadequately capture the various degradation trends of bearings, resulting in low accuracy in predicting the remaining useful life (RUL). To address this challenge, a RUL prediction method based on the vibration signal and a mechanism model is proposed. Firstly, signal processing and feature fusion techniques are employed to construct a nonlinear composite health indicator (CHI) as a measure of bearing life degradation. Then, an adaptive degradation starting point (DSP) identification method is employed to segment the bearing’s full life cycle into distinct states. Based on the results of state division, DMs are classified into three types, including the abrupt DM, the progressive DM and the self-healing DM. Secondly, the degradation mechanisms of bearings are analyzed, and a mechanism model is proposed to describe multiple DMs simultaneously. This model is compared with typical life degradation models, demonstrating superior adaptability to the self-healing DM. Finally, a novel RUL prediction method is developed based on the proposed mechanism model. This method allows for predicting the RUL of bearings under various DMs. Compared to other prediction methods, the proposed approach reduces the mean absolute relative error by at least 58.72% and improves the score by at least 363.21%.