l/sub 1/ identification applied to a fluid dynamics problem

An application of robust identification techniques to a fluid dynamics problem is presented. The experimental data proceeds from a Taylor-Couette instability process. Its dynamics is usually modeled by a linear partial differential equation which does not describe adequately certain oscillatory behavior. To this end we apply an identification technique to produce a more suitable description. The dynamics of the problem is excited by a tracer impulse and step injection. The output consists of the tracer concentration at the outlet of the experimental setup. The process presents a delay which is Identified parametrically. For a given Reynolds number, the nondelayed dynamics can be considered as linear and infinite dimensional. It is identified nonparametrically by means of a tuned asymptotically optimal /spl epsiv//sub 1/ robust identification procedure. Several experiments are performed on this fluid dynamics process for different Reynolds numbers and inputs. The identification procedure is applied to this experimental data to obtain linear delayed models in each case.