New dynamic hysteresis model by means of soft computing approach

The study and calculation of magnetic field in electrical machines required dynamic hysteresis model. On the basis of fuzzy clustering and Neuro-fuzzy identification capability of any kind of nonlinear, continuous functions represented by its discrete set of measured data, a new modeling technique for dynamic magnetic hysteresis is presented and compared with measured data. Four technique are study and compared, the first one is based on neuro-fuzzy technique by using an adaptive neuro-fuzzy system identification and the others are based in Gustafson-Kessel, Gath- Geva and EM fuzzy clustering algorithm. Very accurate prediction of dynamic hysteresis loops is observed, proving that the clustering and Neuro-fuzzy techniques are suitable for hysteresis modeling.