Adaptive inferential control

Despite the successful application of advanced predictive control algorithms to many industrial chemical processes satisfactory system performance cannot always be guaranteed. This is often the case where the infrequent measurement of key process outputs is unavoidable due to sampling limitations. In such situations the ability to detect deviations from desired process behaviour is significantly impaired. Inferential estimation techniques employ more easily measured secondary variables to infer the desired primary variable. This facilitates the early detection of disturbances thus improved control performance is to be expected. An adaptive inferential measurement algorithm has been successfully applied to various industrial processes (Lant et al., 1991; Mitchell et al., 1995). This contribution discusses the development of a model based control strategy using the inferential estimation algorithm as a basis. The theoretical development of the adaptive inferential long range predictive control algorithm is outlined. The algorithm offers enhanced control performance when compared to existing model based design strategies