A Physical Model-based Methodology for RUL Prognostics of Rolling Element Bearings

Prognostics of rolling element bearings (REBs) is one of the most challenging tasks in condition monitoring (CM), especially in aerospace applications, such as aircraft engines, helicopters, and fuel and hydraulic systems. Effective maintenance, repairs, and replacement schedules must be strategically planned to avoid economic losses, sudden stops and potential casualties. In this study, a physical model-based methodology combined with an extended Kalman filter (EKF) estimator is developed to estimate the remaining useful life (RUL) of bearings. The study consists of three main components: (1) Calculation of an offline RMS map based on a lumped parameter model, which describes the relationship between a quasi-defect area (defined by its depth and length of the arc) and RMS. The key assumption is that the energy, generated by real defects and captured by measurements, can be represented by RMS values simulated by a quasi-defect physical model, ensuring a monotonically increasing health indicator (HI) for improved RUL estimation. (2) Online data acquisition and fault detection, using a proposed fault detection indicator based on the Enhanced Envelope Spectrum (EES) estimated by spectral correlation. (3) An RMS map search strategy, initiated after a fault is detected, where the quasi-defect size will increase monotonically, to identify the corresponding quasi-defect size at the current time, and to estimate the RUL using the EKF. The proposed methodology is tested on a dataset including three bearings. The main contribution of this study is the development of a systematic RUL prognostics methodology that can be applied to the online maintenance scheduling of aerospace REBs. This methodology integrates signal processing technologies, a lumped parameter model of REBs, and a statistical estimator, resulting in interpretable outcomes.