Fault Diagnosis and Fault Tolerant Control of Switched Dynamical Systems

Abstract

Modern technological systems consist of many components with strong interactions between them. Faults may cause an unacceptable loss of the system functionality, instability, or fatality. Systems capable of automatically detecting, diagnosing faults, and maintaining the overall functionality are desirable. Fault diagnosis (FDD) is a process that detects, locates, and nds nature of fault. Fault tolerant control (FTC) system has the ability to tolerate faults. Among dynamical systems, switched systems (SS) have numerous applications in control of robotics, automotive industry, aircraft and air tra c control, industrial electronics (power converters) etc. A typical SS is composed of a family of subsystems and a rule that governs the switching among them. Ideally, the FDD/FTC systems are designed on the basis of assumption that it is switching synchronously with corresponding subsystems of SS; that is; FDD/FTC system switches exactly at the time of switching in the switched system to be monitored. However, in practice, the switching in FDD/FTC system lags the switching of the switched system to be monitored. This creates a particular interest in the design of FDD/FTC systems especially when there is event-based switching. In this dissertation, the term\asynchronous" is used to illustrate this situation. This thesis studies the design of FDD and FTC systems of SS under asynchronous switching scenario, in the presence of disturbances and noise (unknown inputs). In the rst part, a framework for fault detection and isolation (FDI) is proposed. The residual (symptom signal) is so generated that it is sensitive to faults and robust against disturbances. A multi-objective problem is formulated based on H􀀀=H1 ltering. Using the average dwell time approach and the piecewise Lyapunov function technique, su cient conditions are suggested in terms of linear matrix inequalities (LMIs) to guarantee the stability and desired performance. In addition, the proposed framework has also been extended to design FDI strategy for uncertain SSs. A norm-bounded uncertainty is considered. To improve the FDD capability adaptive threshold scheme is developed. xiii In the second part, fault estimation (FE) and FTC schemes are proposed. The proposed framework is based on unknown input observer (UIO) and H1 optimization. On the basis of FE, recon guring control law approach is utilized to tolerate faults. To this end, an integrated approach for FE/FTC is proposed for SSs. The last part of this dissertation addresses another very important problem of highly practical interest; that is, the design of fault detection (FD) scheme for switched system with state delays, under asynchronous switching. The tools from robust control theory, Lyapunov stability theory, and linear matrix inequality are used to propose the schemes. To demonstrate the e ectiveness of the proposed schemes, the algorithms have been tested on the dynamics of highly maneuverable aircraft technology (HiMAT) and battery converter unit (BCU) of hybrid electric vehicle.