Nonlinear Dynamic Matrix Control using Local Models

This paper proposes using a Local Model (LM) network representation of a nonlinear chemical process as a basis for nonlinear Dynamic Matrix Control (DMC). The LM network is composed of local linear ARX models and is trained using a hybrid learning approach. The nonlinear DMC uses step responses for different process operating points extracted from the LM network, rather than the single-step response model of linear DMC. Simulation studies of a pH neutralisation process indicate improve ments in both set point tracking and disturbance rejection.