Riemann-Liouville Derivative Kernel Recursive-Least-Square Filtering for Remaining Useful Life Prediction on Rolling Bearings

The state of rolling bearings greatly determines the machinery health in modern rotating machinery related production. Hence finding effective online methods to estimate the remaining-useful-life (RUL) utilizing the measurement data via accessible sensors, is of utmost practical significance. The majority of data-oriented approaches are heavily dependent on the assumption that the dynamics of the condition health indexes (HIs) has a relationship with the RUL of a rolling bearing. Motivated by this concern, this paper introduces a novel algorithm for rolling-bearing RUL estimation based on a fractional-order derivative kernel recursive-least-square filtering (FrKRLS) strategy. It leverages the idea that evolving faults imply changing states in the evolution trend of HIs sensitive to RUL. The innovation lies in combining the Riemann—Liouville fractional derivative with kernel recursive least squares to effectively and rapidly capture HI state variations, enabling high-performance RUL estimation. Experimental results showcasing its superiority when compared to traditional adaptive filters.