Frequency-dependent approach to model validation for iterative identification and control schemes

Classical validation methods 'accept' or 'reject' a model as a valid representation of a plant for some intended use. However, this binary decision has several problems. First, models are neither good nor bad but have a certain valid frequency range and secondly the procedure gives no insight into why the model is not useful or how to improve the model. Moreover within the framework of iterative identification and control design the model validation issue arises the following requirements: (i) Is it possible to improve an existing model? (ii) How can the model be improved? and (iii) What performance increase can provide the designed controller? These facts question the suitability of traditional model validation schemes in general and their suitability for iterative control schemes in particular. We present a new validation procedure that overcomes these problems by performing a frequency-dependent model validation. The validation procedure is then more informative because of its frequency information content. As a result the same model can be validated for some frequency band and invalidated for a distinct frequency range.