Evolution of an industrial nonlinear model predictive controller

Motivated by a specific manufacturing problem in 1990, Exxon Chemical Company embarked on the development of a nonlinear multivariable model-based predictive controller. The controller’s evolution has included collaboration among academic researchers, engineers from industry, and process control software vendors. The resulting control algorithm was patented by Exxon Chemical Company and commercialized by Dynamic Optimization Technology Products, Inc. At the same time, several other academic interactions produced results supporting the implemention of these controllers in our manufacturing facilities. This paper chronicles the evolution of the controller development, and presents the details of the control algorithm. The control algorithm features are discussed, and where applicable, compared to other model predictive control (MPC) algorithms. Finally, two industrial examples are presented to illustrate the methodology