Nonlinear internal model control using local model networks

Local model (LM) networks represent a nonlinear dynamical system by a set of locally valid submodels across the operating range. Training such feedforward structures involves the combined estimation of the submodel parameters and those of the interpolation or validity functions. The paper describes a new hybrid learning approach for LM networks comprising ARX local models and normalised gaussian basis functions. Singular value decomposition (SVD) is used to identify the local linear models in conjunction with quasiNewton optimisation for determining the centres and widths of the interpolation functions. A new nonlinear internal model control (IMC) scheme is proposed, based on this LM network model of the nonlinear plant, which has the important property that the controller can be derived analytically. Accuracy and stability issues for the nonlinear feedback control scheme are discussed. Simulation studies of a pH neutralisation process confirm the excellent nonlinear modelling properties of LM networks and illustrate the potential for setpoint tracking and disturbance rejection within an IMC framework.