About

This paper considers how to achieve best dynamic response from a multivariable predictive controller. Condition number is used as a main criterion. High condition number reflects collinear dependence of some process inputs and outputs. Another criterion for building an optimal controller is controller performance, which can be best achieved with a squared controller configuration. These criteria lead to an optimal multivariable controller that utilizes a squared controller with minimal condition number. Constraint handling in such a controller is achieved by managing controller set points determined by an LP optimizer. To get faster controller response, which is important for handling constraints, in addition to the condition number, process dynamics too are taken into account in developing the controller configuration. These considerations are used in developing a procedure for industrial controller design. The procedure, along with preliminary test results is outlined in the paper. The tests present controller performance and constraints handling on a simulated distillation unit.

- Downloadable files require Adobe Acrobat Reader.