Vehicle health monitoring system using multiple-model adaptive estimation

Abstract

In this thesis, we propose two failure detection and identification (FDI) approaches based on the multiple-model estimation algorithm to monitor the health of vehicles, specifically aircraft applications. They detect and identify failing components of the vehicle, and the system variations. The dynamics of the vehicle are modeled as a stochastic hybrid system with uncertainty-unknown model structure or parameters. FDI performance is evaluated for each approach. We demonstrate the reliability, validity of these approaches by applying them to simulate aircraft machinery experiencing component failures or structural variations. The approaches that we surveyed are: (i) Multiple-Hypothesis Kalman Filter, and (ii) Interacting Multiple-Model (IMM) Estimator. By coupling the fault detection and identification (FDI) scheme with the reconfigurable controller design scheme, a fault-tolerant control system based on the multiple-model estimation algorithm is defined.