Validation of reduced-order models for closed loop applications

Abstract

This work is concerned with the validation of reduced-order models which are intended for closed-loop applications. The main objective is to point out some issues arising in the use of such models in closed-loop applications. It is shown that in validating simplified models, special attention should be given to the accuracy of the approximation at crossover frequencies. This observation naturally leads to the conclusion that, in some cases, open-loop model reduction techniques can be successfully used in deriving models intended for closed-loop applications thus avoiding optimal methods which are far more time-consuming. A second objective is to show how a reduced-order model could have varying degrees of accuracy in closed-loop when controller parameters are varied. Numerical examples which use the model of an actual fuel control system are included to illustrate the main points of the paper.<>