AR model of the torque signal for mechanical induction motor faults detection and diagnosis

Abstract

Mechanical faults present a large portion of induction motor failures, if left undetected, it can lead to partial or total breakdown of the machine. This paper propose a scheme to detect and diagnose mechanical faults in an induction motor by the AR model coefficients of the Torque signal. First, the torque signal obtained from experiment in different conditions: healthy condition, motor with dynamic eccentricity fault, and motor with misalignment fault are normalized to exclude the load effect. Then the AR model coefficients are extracted as features to reduce the dimension data while keeping the effective information. Finally, the Self Organizing Map neural network is used for classification of the different conditions. The experimental results show the effectiveness of the proposed method, were both eccentricity and misalignment faults might be easily detected and discriminated from each other.