Modularized Observer Design and Analysis for a Class of Cascade-Feedback Connected Nonlinear Systems

• Main Authors: Yu-Wei Lin, 林郁偉
• Other Authors: J.-W. John Cheng
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/19173680544697741037

博士 === 國立中正大學 === 機械工程所 === 98 === Most nonlinear observer designs employ the Lie-derivative based nonlinear coordinate transformation as the first design step, and design complexity becomes an issue when dimension of the dynamic system is high. In this dissertation, this design issue for high dimensional systems is addressed. Motivated by the intrinsic structure of the most dynamic systems, the formulation of a class of nonlinear systems composed of cascade-feedback connected sub-systems was proposed and its observability was addressed. Exploiting the cascade-feedback structure, a modularized approach to observer design was developed. The modularized approach reduces the design complexity related to high dimensional systems by delegating one observer to each sub-system with lower dimension To assess applicability and effectiveness of the proposed observer designs, observers for the melt-front position of the injection molding process were developed and evaluated on a commercial injection molding machine. The melt-front observers were shown to have good convergence performance in estimation, and through an empirical study, the robustness of the observers was studied against environmental disturbance and parametric uncertainty.