Detecting model-plant mismatch of nonlinear multivariate systems using mutual information

Abstract For model based control systems, such as MPC, the model plays a key role in controller design and tuning. For linear systems, many different approaches have been well developed for MPM detection while nonlinearity was not considered. This article describes mutual information (MI) as a general correlation measure to detect model-plant mismatch (MPM) in nonlinear systems. First, MI between dithering signals and model residuals is estimated to reveal process MPM regardless of change in disturbance dynamics. Then for MIMO systems, we propose a method to locate the sub-sets of mismatch so that only sub-systems will be perturbed for model updating. The validity of the proposed method for nonlinear systems is illustrated through a numerical example and an industrial polypropylene(PP) process.