Fault detection lter and fault accommodation controller design for uncertain systems.

• Ano de defesa: 2022
• Autor(a) principal: Carvalho, Leonardo de Paula
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Fault-detection; Fault-tolerant control; Linear matrix inequality; Linear parameter varying; Markovian jump linear system; Modelos não lineares; Sistemas dinâmicos; Teoria de sistemas e controle
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/3/3139/tde-28032022-133506/

Model-based Fault Detection (FD) and Fault Accommodation (FA) approaches have been applied in a variety of cases. We propose several techniques to include uncertainties in the design process. First, we focus on the design of the Fault Detection Filter (FDF) and Fault Accommodation Controller (FAC) for Markovian Jump Linear Systems (MJLS). The MJLS framework allows us to include the network behavior (packet loss) during the design of the FDF and FAC. Second, we propose an FDF and FAC design for the MJLS, under the assumption that the Markov chain mode is not directly accessible. Since we are using the MJLS framework to model the network behavior, the assumption that the network state is not instantly accessible is useful because from a practical standpoint this is a truthful assumption. Third, from the results presented for the MJLS framework, we provided follow-up results using Lur\'e Markov Jump System. This is compelling since on some occasions the nonlinear behavior cannot be ignored. Therefore, applying the Lur\'e MJS framework allows us to consider the same assumptions from MJLS, but now adds the nonlinearities. Fourth, we propose the design Gain-Scheduled FDF and FAC for Linear Parameter Varying (LPV) systems, under the assumption that the schedule parameter is not directly acquired. We assume that the schedule parameter is subject to additive noise. This imprecision is included during the design, using change of variables and multi-simplex techniques. Finally, throughout the thesis, we provide some numerical examples to illustrate the viability of the proposed approaches.