INTERNAL MODEL CONTROL DESIGN : A MULTIPLE MODEL APPROACH

Abstract

In this study, two novel IMC design procedures which are capable of controlling dynamic systems with a wide range of operating points are presented. To do so, the concept of operating region decomposition is applied to decompose a dynamic system into a set of operating regimes. The particular structure for local model selected is the transfer function model, which can be obtained from either linearization of the first-principles model or identification from available input-output data. Each local model is assigned with a validity function and then the global multiple model is formed by the weighted sum of local models using these model validity functions. In the first approach, the actual process dynamics in the IMC structure is represented by the global multiple model. By treating the validity functions as time-varying uncertainties in the system, an analytical framework is developed employing a structured singular value test for mixed time-invariant and time-varying uncertainties. Simulation results confirm that the resultant IMC design method is indeed superior to the conventional IMC design. In the other approach, a nonlinear IMC design approach is proposed. In this method, the global multiple model is used in the construction of the model inverse. Compared with other nonlinear IMC design, this method has the advantage of easy implementation in practical application because it has a simple structure and requires moderate computation. Simulation results also verify that the performance of this nonlinear IMC design shows a marked improvement over the conventional IMC design.