Zonotopic Extended Kalman Filter For RUL Forecasting With Unknown Degradation Behaviors

Abstract

This paper proposes a novel approach for Remaining Useful Life (RUL) forecasting using interval model-based prognostics techniques based on zonotopes without prior knowledge of the degradation behaviors of the system. Although Kalman filtering techniques have proved their estimation ability with Gaussian noises, an interval approach with zonotopic sets technique has been integrated for optimal estimation of parameters with unknown-but-bounded noises. Moreover, the proposed model-based prognostics technique has been applied to a DC-DC converter described as a nonlinear dynamical system affected by degradation behaviors. Thus, the estimated degraded parameters are adopted in the RUL prediction technique that propagates the zonotopic sets until the End-of-Life (EoL) of the system. In general, the technique is split into estimation and prediction phases using Zonotopic Extended Kalman Filter (ZEKF) to deal with the nonlinearities of the system and compute the optimal observer gain. A DC-DC converter case study in simulation is used to illustrate the utilized techniques and the simulation results prove the effectiveness.