Modeling of chemical reactor dynamics by nonlinear principal components

The modelling of nonisothermal continuous stirred chemical reactor dynamics by linear and nonlinear principal components methods is investigated. The derived models are analysed with respect of their ability to predict the ex istence of the reactor multiple steady states and their use for adaptive on -line process control. The time evolution of the state variables is approxi mated by a single-step finite difference prediction equation. Nonlinear pri ncipal components are determined by a feedforward neural network with a sin gle hidden layer. Input and output patterns are jointly projected to a two dimensional surface yielding an implicit process model. The ability of impl icit models to predict controlled and manipulative variables without the ne ed for separate model development for the direct and inverse models makes t hem ideally applicable in adaptive internal model control loops. The model correctly predicts the existence of three steady states and provides an exc ellent fit to untrained samples of patterns under various dynamic condition s. The linear models based on a partial least squares algorithm can correct ly model behaviour under unsteady conditions, but they fail to predict mult iple steady states in chemical reacting systems. Since accurate model of st eady-state properties is essential for process control, linear principal co mponent models are inadequate when multiple steady states exist. (C) 1999 E lsevier Science B.V. All rights reserved.