Feedforward Control for Non-Minimumphase Local Model Networks Using Model Following Control

We consider the feedforward control design for a local model network (LMN) with possibly non-minimumphase dynamics. For a class of LMN with a parametrisation typical for experimental identification we show that the existence of a flat output is a very restrictive assumption. Therefore we propose a model following control (MFC) structure for the online-generation of a feedforward control signal. This method allows for arbitrary reference signals and has good robustness properties with respect to model uncertainties. We verify our approach by a simulation example and compare with a standard method using model inversion. Applying the proposed approach to a nonlinear process with unstable zero-dynamics illustrates the attainable good performance results.