MEASUREMENT SELECTION AND CONTROL SYSTEM DESIGN FOR MULTIVARIABLE INTERACTING PROCESSES (DISTILLATION, ROBUSTNESS, UNCERTAINTY)

A cornerstone of process control system development is the issue of pairing of input and output variables. In many cases multivariable processes can be adequately controlled by multiple single input-single output controllers. A general framework is developed to aid the process engineer in determining which output variables should be paired with which manipulated variables. In particular a new technique is presented to predict a priori when a setpoint change in one loop will affect the performance of another loop.; Distillation is the most important separation process in the chemical process industries. It is an energy intensive process that requires a large percentage of the energy consumed in a typical plant. Dual composition control in distillation minimizes the energy consumption for a given set of product specifications. The most common method that has been applied to dual composition control is input decoupling. It was previously suggested that implicit decoupling can be achieved by the proper selection of process measurements, which as the sum or difference of temperature measurements. This idea has motivated the research on non-interacting control system design methods in distillation. This work provides a rigorous mathematical formulation as well as a physical basis for implicit decoupling. An interesting result is that many of the non-interacting control system design methods suggested in the literature result in the same type of control strategy for the material balance control scheme in distillation.; A critical requirement of a process control system is that it remain stable over a wide range of operating conditions, while maintaining desirable performance characteristics. The primary limitation to high performance distillation column control is the nonlinear nature of the process. The nonlinearities have been represented as a region of linear systems, in order to perform a robust stability analysis. Control systems were designed for both conventional control and material balance control utilizing the structured singular value methodology. Unstructured uncertainty techniques were then used to compare the various control strategies with respect to sensitivity to unstructured perturbations.