부산대학교 도서관

In this paper, a modified multivariable disturbance observer-based (MDOB) design method is proposed to improve disturbance rejection performance for multiple-input-multiple-output (MIMO) processes with multiple time delays. Two important issues in MDOB control are studied here: computation of the inverse model and design of the filter. First, to reduce the influence of time delays, the nominal process is factorized as a diagonal time delay matrix and a remaining transfer function matrix, and the remaining part is used to design the inverse model in MDOB. Considering the difficulty of matrix inversion, the Nyquist set is used to describe the dynamic of the inverse model, and accordingly, the simplified inverse model is obtained by the curve fitting technique. In addition, to guarantee the properness and stability of the simplified inverse model, a compensator is introduced. Rules for selecting compensator parameters are given. Finally, with the proposed inverse model and compensator, the filter Q (s) is optimized by minimizing the H∞ norm of the disturbance estimation error. The filter constants are determined by fully consider the disturbance rejection performance and robustness of the system. Simulation examples are presented to illustrate the effectiveness and merits of the proposed method. [ABSTRACT FROM AUTHOR]