SENSOR-LESS, NEURONAL-NETWORK-BASED CAGE DEFECT DIAGNOSIS FOR INDUCTION MACHINES

This paper presents a PC-based diagnosis system for cage defects in sq uirrel cage induction motors. Second-order effects, like field and vol tage harmonics, core currents and torque ripple caused current sideban ds are integrated in a novel machine model for cage defects. In additi on, the harmonic analysis display a way how the machine speed call be calculated our of the current spectrum. The experimental test of the d iagnosis system reveals not only the ability to diagnose a defect cage , but also to distinguish different kind of defects.