Abstract

Faults are minor malfunctions that deteriorate the performance of a system. In a safety critical situation such as the control of an airplane, compounding faults may cascade into a catastrophic event if not properly compensated. Active Fault Tolerant Control (AFTC) addresses the fault accommodation problem - the reliability and robustness of the system in faults - beyond the conventional stability and performance requirements for a normally operating plant. This thesis studies the AFTC of an electro-hydraulic driven elevator, which serves as a primary control surface of an airplane. The proposed AFTC system consists of three components: (1) A Fault Detection and Estimation (FOE) component is designed based on two robust adaptive observers. (1). Adaptive Unknown Input Observer: a disturbance decoupled observer utilizing the geometry property and measurement redundancy of the system; (2). H x /H _ adaptive observer: an optimization based observer to maximize the system's response to faults and minimize that to disturbances. The H x /H _ adaptive observer is constructed with the technique of Unitary System, which is defined as a linear system whose singular values of transfer matrix are equal. (2) A fuzzy Proportional-Integral (PI) controller is designed based on the fuzzy Tagaki-Sugeno (TS) model of a nonlinear system, which consists of different linear models at different operating points. (3) The reconfiguration is carried out based on the fault information available from FDE. To reduce the time needed for the online computation, multiple controllers are designed offline for different faults scenarios. A new controller is constructed online as a fuzzy combination of these controllers to meet the post-fault stability and performance requirements. Simulation results show that, with the proposed AFTC, occurring faults are detected promptly and estimated accurately with the FDE component. The performance of the post-fault elevator is quickly restored after the reconfiguration