Robust Control System Design for a Fixed-Bed Catalytic Reactor

Citation

Mandler, Jorge Anibal (1987) Robust Control System Design for a Fixed-Bed Catalytic Reactor. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/czeb-7c12. https://resolver.caltech.edu/CaltechETD:etd-03212008-092922

Abstract

The design of control systems in the face of model uncertainty is addressed. A methodology for the design of robust control schemes is outlined, which employs the Structured Singular Value as an analysis tool and Internal Model Control as the synthesis framework. This methodology is applied to the design of control systems for a fixed-bed, laboratory, catalytic methanation reactor. The design procedure allows a clear insight on the fundamental limits to closed-loop performance and provides controllers with explicit stability and performance guarantees for the case of plant-model mismatch.

The overall controller design effort is initiated with a careful mathematical modeling of the system. The original nonlinear partial differential equations are converted through collocation techniques into a nonlinear ordinary-differential/algebraic equation system amenable to dynamic simulation. Interactive software is developed for the open- and closed-loop simulation of general nonlinear differential-algebraic systems, which provides an efficient means to simulate the reactor model. Linearization and control-relevant model reduction techniques are applied to arrive at models appropriate for the control studies.

Both the single-input single-output and the multivariable case are addressed. Three different control configurations are investigated in the context of the single- pass operation of the reactor. In each case-study presented, the controller design procedure is divided into four steps: first, the definition of the control objectives, which not only leads to the selection of the appropriate control configuration but also determines the most adequate design techniques to employ; second, a nominal design step, in which the system-inherent limitations to the closed-loop performance are highlighted; third, a characterization of the uncertainty and the use of this information in the design of robust controllers; and, fourth, the evaluation of the designs through nonlinear simulations.

The thesis describes the first application of structured singular value-based analysis techniques to a chemical reactor system and is in essence the first comprehensive study of the application of robust control to fixed-bed reactors. The power of the new mathematical theory for robust control system design is demonstrated. It is shown that the design of control systems for complex, distributed systems such as the methanation reactor can be addressed in a practical way, and low-order controllers be adequately obtained, which possess near-optimal characteristics when applied in a realistic environment of uncertainty and unavailability of measurements.

Thesis Files