Detection et isolation des fautes des systemes a sauts markoviens (French text)

The following work addresses the application of the fault detection and isolation (FDI) procedure to Markov jump linear systems (MJLS).; Two main approaches are considered in this work: the first one considers the Hinfinity filtering and the other one uses the unknown input observer (UIO). Hinfinity approach consists in estimating the fault vector in MJL systems by minimizing the Hinfinity norm of the estimation error. This is a way to build a residual that is robust to perturbations and sensitive to faults. The obtained conditions for the existence and the stability of such a filter are expressed in linear matrix inequalities (LMI) form.; The second approach is based on using an unknown input observer for the fault detection and isolation issue. In fact, the unknown input observer (UIO) is an observer that allows the estimated state to approach asymptotically the real state of the system while rejecting perturbation and noise that are considered as unknown inputs. Consequently, the advantages it gives to observe the system state (thus the output) and to generate a residual completely decoupled from unknown entries allows to handle the problem of fault detection in a straightforward manner. (Abstract shortened by UMI.)...