Hinf Based Fault Detection and Isolation for Markovian Jump Systems

Authors:	Akouz Kaoutar, Ecole Polytechnique de Montreal, Canada Boukas El-Kebir, Ecole Polytechnique de Montreal, Canada
Topic:	6.4 Safeprocess
Session:	Safety and Structure Analysis for Fault Diagnosis/ Diagnosis of Hybrid and Discrete-Event Systems
Keywords:	fault detection and isolation, Hinf filtering, linear matrix inequality (LMI), Markov jump linear systems

Abstract

Most of industrial processes are a combination of continuous and discrete dynamics known as hybrid systems. Markov jump linear systems (MJLS) as a special class of this family are characterized by their switching from a mode to another. It is desirable that such systems be reliable and that their running be the most efficient and safe possible. Thus, the monitoring of MJLS is an interesting area to explore. This paper addresses the fault detection and isolation issue for markovian jump linear systems. The error between the fault vector and its estimate is minimized using Hinf approach. The resulting estimate could be taken as a residual to be evaluated to show fault occurence.