Linearization based and local model based controller design

This paper deals with the comparison of two design methodologies for nonlinear fuzzy model based controllers. A straightforward approach to design nonlinear controllers is to linearize the nonlinear model at the current operating point and to design a linear controller. If the design of the linear controller, however, requires the optimization of parameters, this design procedure can become infeasible for real-time application because of the high computational effort. Local linear models, however, provide the possibility to design such controllers off-line for each submodel separately. During the control process, it is interpolated between the linear controllers according to the current operating point. These two design strategies are discussed and the control performance is compared by a simulation example. Finally, both strategies are utilized to design a predictive controller for temperature control of a heat exchanger.