Model validation approaches for nonlinear feedback systems using frequency response measurements

We present a model validation problem, motivated by the modeling of a combustion process under investigation at United Technologies Research Center (UTRC). Both linear and non-linear versions of the UTRC combustion model are presented and discussed. The system excitation is well modeled by a stochastic source and this dominates the response. The number of data points that must be recorded to capture the system behavior precludes the use of time-domain model validation techniques. A critical analysis of the results of the linear model validation indicates the limitations of the linear model validation techniques in the context of nonlinear feedback systems. The need for new model validation techniques that take into account nonlinear model perturbations and the effect of the driving broad band disturbance is indicated