• Citer

Model-based Fault Detection and Isolation for a Class of Discrete-Time Systems

par De Melo Schons, Silvane
Président du jury Bogaerts, Philippe
Promoteur Kinnaert, Michel ;Coutinho, Daniel Ferreira
Publication Non publié, 2023-04-26

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

During the last decades, robust model-based fault detection and isolation (FDI) has received increasing interest in the literature. It is a fact that FDI methods can provide fault indicators for predictive maintenance and, consequently, preventing significant damages to the system components. Following this route, this doctorate thesis develops model-based fault detection and isolation techniques for linear discrete-time systems as well as for linear discrete-time descriptor systems. The proposed approach is based on the design of observer-based residual generators in order to detect and isolate actuator and/or sensor faults. Besides, a reference model is considered to describe the desired behavior from faults to residual, which allows for including requirements on fault decoupling as well as performance issues regarding robustness related to the optimization problem. Therefore, the residual generator is designed such that the system-filter connection approximately follows the fault to residual behavior provided by the reference model. A multi-objective optimization is designed involving different system criteria, such as $H_infty / H_-$ and $ extit{peak-norm} / H_-$ , which depends on the class of inputs of the system (respectively, $ell_2$ or $ell_infty$ signals). The design conditions based on modified Lyapunov dissipation inequalities are cast in terms of Linear Matrix Inequalities (LMI) constraints. In order to illustrate the theoretical developments, numerical examples are provided as well as the application of the method to a simulated lithium-ion battery pack. There are four main contributions for this thesis. The design of FDI techniques integrated with a triangular reference model structure allows the fault detection and isolation when the faults do not occur simultaneously. Besides, the residual generator design conditions described in terms of LMI constraints made possible to consider mixed performance specifications and different classes of systems in a unified mathematical formalism. From a model transformation, the proposed method can also be applied to discrete-time descriptor systems. Finally, the proposed solution in this thesis can be successfully considered for practical-oriented applications, such as Li-ion battery packs.