P-타입 반복 학습 제어의 강인성, 안정성에 대한 연구 및 연속 주조 공정에의 적용

Abstract

In this thesis, we consider a general nonlinear output tracking problem with P-type learning rule. In particular, the P-type high order learning rule is studied. In general, the high order learning rule shows a better performance for convergence speed and robustness than conventional learning rule in presence of the periodic disturbances. The robustness and convergence is analyzed and verified by computer simulations. However, when we use the iterative learning control (ILC) in real world, an unexpected huge overshoot can be observed easily. Therefore we study these phenomenon with various point of view and propose a new learning rule to avoid this unexpected situation. The PD-type and P-type normalized learning rules are proposed and their convergence is proved. Simulation results show that the proposed learning rule is effective for the unexpected huge overshoot. Moreover, the robust optimal design is presented when there is an uncertainty interval. The new robust performance indices are introduced and studied. This design method is very useful when we apply the ILC to a real world application. The optimal design has better performance for convergence speed and global uniform bound than conventional design method. Finally, the ILC is applied to a continuous casting process. To reduce a periodic bulging disturbances, a proposed learning rule with forgetting factor and switching mechanism is applied to the continuous casting system. A 1:4 hardware simulator is used to verify performance of the proposed method.