A novel frequency-domain health indicator for bearing RUL estimation using adaptive Wiener process degradation modeling

Abstract

Accurate and reliable estimation of bearing health conditions requires the construction of a suitable Health Indicator (HI). In this study, the Modified Total Harmonic Distribution (MTHD) health indicator is developed based on advanced frequency domain analysis to describe the bearing health status effectively. It has also been validated that MTHD demonstrated desirable properties of monotonicity, robustness, and trendability. To accurately identify the First Prediction Time (FPT), a linear combination of the mean and variance of the MTHD curve is employed. However, due to variations in operating conditions and loading, the degradation process of bearings may differ. As a result, a single fixed model cannot accurately characterize the occurrence of different degradation processes. To address this issue, an adaptive Wiener model is proposed. In this framework, the Remaining Useful Life (RUL) prediction is achieved using either an appropriate linear or nonlinear Wiener model selected through a model adaptive algorithm. Finally, the effectiveness of the proposed model is validated using the XJTU-SY bearing dataset as well as the laboratory's own generated dataset. © 2026 Elsevier Ltd