Preisach´s function identification by neural network

Summary form only given. Preisach´s model is a powerful tool in the description of hysteresis. It describes the hysteresis loop by a superposition of Preisach´s particles which are mathematical entities whose magnetization can have two values. The switching from one value to the other is triggered when the external field reaches some specific values. These values are variable and the global magnetization is computed as a superposition of all the Preisach´s particles. The number and the characteristics of the Preisach´s particles is described through a distribution function generally called as Preisach´s Function. Several procedures have been developed in order to identify Preisach´s function, some of them implies the solution of set of differential equations and the knowledge of several parameters In this paper we present a methodology of identify a Preisach´s function by using neural networks. The neural network used is a Kohonen network. It is shown that for narrow hysteresis loops the method has a great capability of reproducing the loops. For broad loops this capability can be assured by using more than two gaussians. In the paper a procedure is also attempted to identify the major loop by having information only on the minor loop. It is shown how the method is able in some cases to reconstruct the major loop. This approach could be useful in order to identify the hysteresis loop of materials that are used in a non-symmetrical (e.g. branches of ferromagnetic circuit containing permanent magnets) way. The method proposed simplifies the identification of Preisach´s function and needs the knowledge of only a major loop.