Model-reference Based Robust Adaptive Control And Adaptive Fault-tolerant Control

Based on Model Reference design method, this dissertation focuses on two aspects: Model Reference Robust Adaptive Control and Model Reference Adaptive Fault-tolerant Control.?) For multivariable adaptive control systems, the required assumptions are very stringent to ensure stabilization of the closed-loop system. Therefore, to expand the ap-plication of multivariable adaptive control, how to relax its strict restriction is one of the theoretical issues to be solved. For a multivariate controlled system that does not meet the required assumptions, its system model can be simplified properly. The simplified model can be considered as the nominal part of the controlled system, if it meets system as-sumptions. The dynamical parts ignored in the process of the model reduction, constitute unmodelled dynamics of the controlled system. Because a simplified model and its orig-inal system model are, in principle, close as far as possible, some appropriate constraints can be given for the unmodelled dynamics. For a large class of discrete-time multivariable systems with unmodeled dynamics and bounded disturbances, a direct scheme of robust model reference adaptive control is reinvestigated, in an input-output approach. Com-pared with the existing results, in virtue of a high-frequency gain matrix factorization, the assumptions required in the scheme are further relaxed without overmuch complicat-ing the controller structure, and a permissible range of the gain parameter in a modified adaptive algorithm is clearly specified. Moreover, paralleling the continuous-time theo-retic framework, robust stability and robust tracking performance are analyzed by using the multivariable versions of some important technical lemmas, such as an exponentially weighted norm-relation lemma and two swapping lemmas.The contributions in this part are as follows.(1) An crucial corollary of the high frequency gain matrix decomposition lemma is established. By virtue of this corollary, a range of the adaptive gain matrix can be quan-titatively determined. The original model reference adaptive control scheme, based on a decomposition of the high frequency gain matrix of multivariable discrete time systems, is difficult to determine the range, for the adaptive gain matrix is subject to a matrix in-equality constraint that relies on the unknown factor matrix of the high frequency gain matrix.(2) The multivariable versions of some significant technical lemmas, such as an ex-ponentially weighted norm-relation lemma and two swapping lemmas are established, which pay the foundation for analyzing robust stability and robust tracking performance of the closed-loop system.II) During a long time running in practical control system, because of the aging, wear and tear, wrong operation, some unpredictable system faults can occur. For mul-tiple inputs single output linear systems with uncertain parameters and unknown actu-ator faults, based on the jump-parameter based swapping lemma established by us, the standard model reference adaptive design method is extended to linear uncertain system with unknown actuator faults. A model reference adaptive fault-tolerant control scheme is presented for multi-input single-output linear systems with unknown system parame-ters and unpredictable actuator fault. The scheme requires a weaker hypothesis than the existing relevant results and it removes some assumptions on faulty systems. This is ob-tained based on a novel design idea that every functional actuator shares a part of the pre-specified control task as well as the negative effect of actuator faults. Adaptive failure accommodation controllers can guarantee systems internal stability and asymptotically tracking.A contribution in this part is that an equivalent model reference adaptive fault-tolerant control scheme is established. The scheme only requires the transfer functions between inputs and the output are minimum phase, but the existing results require that the transfer functions between any system input combination and system output are minimum phase. Therefore, system assumption is relaxed. The weakening assumptions expand the application range of the model reference adaptive fault-tolerant control.