A methodology for integration of control and fault diagnostics

In an attempt to realize fault tolerant control, a methodology for integrated design of control and fault diagnostic is investigated in this dissertation. Reliable operation of a system with automatic control should depend, not only on the robustness of the control system but also on the systems fault diagnostics capabilities. A comprehensive strategy therefore needs to be implemented for fault tolerant control. Such a control strategy would comprise of three modules: a control module, a fault diagnostics module and a controller reconfiguration module. Design of these modules is a nontrivial issue. There exists significant inter modular interaction. However, conventional design approaches hitherto did not consider such interactions during the design process. This dissertation focuses on a comprehensive integration of these three modules and develops a methodology for an integrated design of fault diagnostics and control.; Specifically, the integrated design of control and diagnostics is achieved by combining the Integral sliding mode control methodology, observers with hypothesis testing and controller reconfiguration. Information obtained from integral sliding mode control and from observers with hypothesis testing is utilized so that a fault can be detected, isolated and compensated for. In summary, the role of the control module and the diagnostics module can be stated as follows. (1) Control Module. (a) Perform primary control task; (b) Detection of fault using additive control terms (equivalent value of integral control term); (c) Input (actuator) fault compensation. (2) Diagnostics Module. (a) Generation of analytic redundancy (estimation using observers with hypothesis testing); (b) Fault isolation. (i) Input (actuator) fault is isolated by observers using both output and estimated fault. (ii) Output (sensor) fault is isolated by decoupling a faulty sensor or by direct estimation of the fault. In the latter case, information obtained from controller (time derivative of sensor fault is obtained using integral sliding mode control term).; As an application example, the air and fuel dynamics of IC engine is considered. A mean value engine model is developed and implemented in Simulink. The air and fuel dynamics of the engine are identified using experimental data. The proposed algorithm for integration of control and diagnostics is then validated using the identified engine model.; For both control and diagnostic application, two novel observer design methods are proposed. An observer using a binary sensor (HEGO sensor) is developed for the estimation of fuel dynamics and air to fuel ratio dynamics. A nonlinear input observer is also developed using a sliding mode methodology and applied to the indicated torque estimation problem for an IC engine.