ROTOR DIAGNOSTICS OF INDUCTION MOTORS BY MEANS OF NEURAL NETWORKS

Leon Swedrowski

Abstract

The investigation results presented in the paper are related to diagnostics of induction rotor’s cages. The virtual tool shown in the paper was created for investigations as an instrument to measure, to present and to register the stator’s current spectrum characteristics. The neuron classifier was constructed to create an instrument enabling to assign the induction motor under test to one of two groups – faultless or defective, and to prove the effectiveness of applying the neuron network in conjunction with the stator’s current spectrum analysis to find out damage in the rotor’s cage. Two options are described: Kohonen self-organizing feature maps and unidirectional multi-layer perceptrons (MLP). Both the networks have successfully solved the stator’s current spectra classification problem assigned to it, and also the technical diagnostics of the rotor’s cage condition.

Keywords

motor, neural network

Download

PWC-2003-TC10-019.pdf

DOI

-

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings