Optimal disturbance observer design using an H-infinity approach

Abstract

In this dissertation, a new design method and analysis for optimal disturbance observer formulated in the transfer function domain is suggested. The presence of external disturbances, model uncertainties and nonlinear effects are inevitable in almost all control system. In addition, the unexpected effects such as undesired fault or model uncertainty may destroy and corrupt the system. However, conventional disturbance observer control scheme cannot be internally stabilized for non-minimum phase plants. For such reasons, a general design of disturbance observer that can be applied to all linear plants has received considerable attention. Since the Linear Matrix Inequalities (LMI) based H-infinity optimization problem is originated from convex optimization, the proposed disturbance observer for uncertain single-input single-output (SISO) linear systems is designed as an H-infinity approach. Verification of the proposed design method is achieved for minimum phase plant, non-minimum phase plant and an unstable plant as the numerical examples for reference tracking problem. The proposed disturbance observer using an H-infinity approach reveals several important contributions: 1) the disturbance observer can be systematically designed; 2) the disturbance observer can be applied to all linear plants, in other words, non-minimum phase plants and unstable plants as well as minimum phase plant. 3) the all weighted objectives for estimating unknown input, reducing the effect of a control input and attenuating measurement noise can be optimally met; 4) three weighting functions allows for the flexible design. At first, a configuration and a design procedure of optimal disturbance observer using an H-infinity approach are newly proposed in case of disturbance estimation problem. In order to design the disturbance observer systematically, a class of single-input single-output linear system is considered. The concept of proposed disturbance observer design is that the de...