Abstract
This paper deals with system identification in the frequency domain using non integer order models given in the pole-zero form. The usual identification techniques cannot be used in this case because of the non integer orders of differentiation which makes the problem strongly nonlinear. A general identification method based on Levenberg-Marquardt algorithm is developed and allows to estimate the (2n + 2m + 1) parameters of the model. Its application to identify the "skin effect" of a squirrel cage induction machine modeling is then presented.