Contrast Feature-Based Approach for Fault Detection in Wound-Rotor Induction Machines

Fault detection in induction machines has become a topic of great interest for researchers due to the growing utility these are giving to today’s industry. Real time machine learning methods are recently proposed in this area to improve the precision of fault detection. In this paper, a novel methodology based on texture feature estimation using an artificial neural network is proposed to be used for fault detection of induction machines. The method is tested for a typical fault as rotor phase opening of a wound rotor induction machine. For this, the stator current in time domain is used to detect the fault based on the proposed method. The experimental results are shown for several loads and speeds conditions to prove the effectiveness of the method.

I. Introduction

THE proper diagnostic of an existing fault in induction machines has become more and more interesting for researchers due to its usefulness for industrial applications [1]. Induction machines are widely used in industry thanks to their low cost, robustness, easy control, low maintenance and versatility for variable-speed applications. However, due to the different environments to which they are subjected, these machines can present failures that represent economic losses [2].