Decentralized Control of Winding Systems: a Hybrid Evolutionary-algebraic Approach

Abstract: This paper considers the problem of controlling a web transport system, using a 3-input 3-output nonlinear model that has been extensively validated against the real plant. There are two difficulties that arise when designing a controller for this industrial plant. The first one is the wide variation in the system dynamics due to the variation of radius and inertia of the rollers. Secondly, the multivariable controller should be decentralized to increase the reliability of the control loop. An approach for designing decentralized controllers that satisfy a H2-performance measure and place the closed loop poles in a prescribed region is proposed. The approach is based on the combined use of Genetic Algorithms and Lyapunov equation solvers. A decentralized controller that meets all design specifications is presented, and the simplicity of the design procedure is demonstrated. Among the attractive features of the proposed technique is its computational efficiency and the fact that it does not need to be initialized. Copyright c 2005 IFAC