Multiple-model based sensor fault diagnosis

Abstract

This research proposes an effective multiple-model (MM) based sensor failure detection and isolation (FDI) system. A nonlinear gas turbine engine model that represents the operational engine model is combined with a variety of piecewise linear (PWL) models to estimate sensor outputs to create the hybrid kalman filters (HKF) that make up the system. By interpolating the PWL models using a Bayesian method, the proposed FDI system is able to identify and isolate persistent sensor bias issues over the full operational regime of the engine.