Nonlinear model-based control with local linear neuro-fuzzy models.

summary many processes display nonlinear behavior if they are driven over a large operating range. if linear controllers cannot yield satisfactory control performance, nonlinear control techniques have to be employed. this requires the knowledge of nonlinear process models. this paper presents an overview about process model architectures originating from the fields of neural networks and fuzzy systems, based on which nonlinear model-based controllers can be designed. three commonly used model-based control approaches are described. depending on the controller design approach and later controller implementation, different demands on the model architecture arise. these demands concern the exploitation of the linear control techniques, the incorporation of prior process knowledge and the fulfillment of hardware requirements. these issues will be discussed and nonlinear modeling and control of an industrial-scale heat exchanger based on neuro-fuzzy network will be presented as an illustrative example. [ABSTRACT FROM AUTHOR]