Analysis and design of heterogeneous control laws for nonlinear chemical processes.

摘要

Most regulatory control paradigms for chemical processes assume that the process remains close to an equilibrium point, and that its behavior about this point is linear. However, modern chemical processes are often required to operate in wide operating regions where the linearity assumption does not apply. There is, therefore, a need for the design of global controllers that are valid over a wide operating region.; For nonlinear processes that have to be operated over a wide operating region where a linear approximation is not appropriate, a direct method of designing global controllers is by using model-based techniques with an explicit nonlinear model of the process. Some important theoretical results in this area have recently been achieved from techniques like feedback linearization, nonlinear internal model control, and nonlinear model predictive control. A different approach that has been widely applied in industry to design global controllers consists of dividing the operating space into different operating regions, designing local controllers in each one of these regions (usually linear controllers), and combining or switching the local control laws as the process evolves from one operating region to another.; In this dissertation a mathematical framework to define this class of controllers is presented, defining them as heterogeneous control laws. This framework is used to analyze fuzzy heuristic control and sliding mode control laws, and to show their similitude. Concepts from feedback linearization are used to identify good candidates to apply fuzzy heuristic control and sliding mode control.; A geometric interpretation of controlled dynamic systems using velocity fields is provided as a tool to help in the analysis of nonlinear systems, and it is applied to explain inverse dynamics, to represent control conditions as non-parametric geometric conditions, to design an heterogeneous controller, and to analyze robustness of some control laws.