Condition monitoring and remaining useful life prediction using switching Kalman filters

Abstract

The use of condition monitoring (CM) data to infer degradation state and remaining useful life (RUL) prediction has grown with increasing use of health monitoring systems. Most degradation modelling requires a detection threshold to be established and can only model a single dynamical behaviour for the degradation. Such approaches have limitations as detection thresholds can vary widely and a single model may not adequately describe a degradation path as it evolves. In this paper, the switching Kalman filter (SKF) is adopted. The SKF uses multiple dynamical models describing different degradation processes and the most probable model is inferred using Bayesian estimation. The advantages are that it does not depend on a fixed threshold for fault detection and can model the different degradation processes as they evolve. The SKF approach is applied to CM data from helicopter gearbox bearings and is shown as a promising tool for maintenance decision-making.

Remaining Useful Life Prediction of Rolling Element Bearings Based on Unscented Kalman Filter

“…Moreover, the Constraint Kalman Filter (CoKF), which is a kind of computationally efficient approach, has been realized for RUL estimation and achieved encouraging accuracy in the application over a battery failure dataset [6]. In addition, the Switching Kalman Filter (SKF) has been also studied for an expanded degradation model [7], being more adequate to describe the threshold and the evolution processes, the healthy serviceable stage, the linear degradation and the exponential degradation.…”