Memoire koli

Golem, le monstre de mémoire Colloque des 7 et 8 juin 2018 sur le personnage imaginaire Université Polytechnique Hauts de France

la bourse

2018

Lutte contre la déforestation et protection de la biodiversité : la cuisine solaire ADES, indispensable à tous ! Mémoire en vue de l'obtention du DUT Département « Techniques de commercialisation » IUT DE BORDEAUX UNIVERSITE DE BORDEAUX

The work developed in this paper has a contribution to the study of methods for state reconstruction, detection and localization of faults in nonlinear systems based observer. We linearized the non-linear systems by a multimodel technique to extend the estimation of states and diagnosis of linear systems to nonlinear systems. Then we developed a multiobserver and a multiobserver with unknown input, for state reconstruction and estimation of faults affecting separately the inputs and the outputs of a system. The faults detection is based on the comparison of a residue; which represents the estimated error at the output, to a detection threshold. The localization step is based on the prior generation of a theoretical matrix of signatures to detect defects. The use of a bench of multiobservers allows us to structure the residues and then to form a matrix of experimental signatures. The localization of defects can be obtained by comparing the theoretical and experimental signatures. We applied these methods to an example of a nonlinear system "Three Tank System". The simulation results showed the performances of multimodel approach for the approximation of nonlinear systems and those of the multiobsevers to perform the faults detection and localization.