Performance evaluation of an inferential predictive control algorithm

Within the chemical and process industries it is common to find plants suffering from the infrequent measurement of key process outputs due to sampling limitations. This contribution describes an “integrated” inferential predictive controller that can overcome measurement limitations while offering the benefits of predictive control. The results of a rigorous performance evaluation of the inferential predictive controller are described when applied to a nonlinear simulation of a distillation column. A direct application of the inferential controller is considered against a cascaded implementation. The inferential controller is tuned to meet robust performance objectives. The results of the simulation studies are used to develop heuristics for the implementation of the inferential controller. It is found that no extra benefits are obtained through the use of a cascade implementation