System identification through neuro-fuzzy methodologies

Abstract

The aim of the present work is to propose a way to identify the behaviour of an induction motor supplied by using a DC/AC converter controlled through a pulse width modulation (PWM) technique. Although a mathematical description of the motor is well-known in literature, the model is sensitive to parameters variations. Moreover it is impossible to modelize in a mathematical way the system composed by the motor and the inverter together. A neuro fuzzy network, trained with a set of I/O measures, it is able to identify the whole system. The results proposed show how the behaviour of the identified system matches the real one.