Decentralized Control of Large-scale Interconnected Systems with Application to Web Processing Machines

Abstract

The thesis investigates decentralized control of a class of large-scale nonlinear systems, which contain unmatched, linear and nonlinear interconnections between constituent subsystems. A large experimental platform, which mimics most of the features of an industrial web process line, is used for experimentation. It is shown that decentralized control schemes can be developed for a class of unmatched large-scale systems. Sufficient conditions are developed that will result in closed-loop stability. A model reference decentralized adaptive controller is developed for systems with unmatched, linear interconnections. To circumvent the unmatched problem, a new reference model that exchanges reference information between subsystems is proposed. Improved dynamic models for unwind/rewind rolls are obtained and bsed on that accurate inertia compensation terms are computed for on-line compensation. Extensive experiments with the proposed decentralized controllers and an often used industrial decentralized PI controller show substantial reduction in web tension error with the proposed decentralized controllers.