Short circuit incipient fault detection and supervision in a three-phase induction motor with a SOM-based algorithm

Abstract

An offline two-dimensional SOM neural based algorithm is used in order to supervise a three-phase squirrel-cage induction motor and detect short-circuit incipient fault condition. A sinusoidal PWM inverter is used to feed the motor and some components of the current motor frequency spectrum are used as input variables. A special electrical structure was built to emulate incipient short-circuit at the stator windings of the induction motor. The data were acquired with the motor operating under different frequencies, load level and fault extent. Through the generated data base, the algorithm was tested and a high mean success rate combined with a good visualization of the problem was achieved. In near future, this algorithm can be used as base for an online supervisory system for this kind of motor failure. © 2013 Springer-Verlag.