Fault Detection Using SVM Based Motor Current Signature Analysis for 3-Phase Induction Motors

Dian Sawitri and Taufik Taufik

Keywords

Induction motor, Interwinding faults, Fault detection

Abstract

This paper proposes an early detection system to identify inter-winding faults in induction motors using the Support Vector Machine (SVM) in Motor Current Signature Analysis (MCSA) technique. Electric current signals from fault occurrence in induction motor were first recorded followed by signal analysis. Recorded signals were then extracted by using wavelet and Principal Component Analysis (PCA) algorithm. The extracted signals were then analyzed in time base and frequency base domains that were later transformed into a set of features using statistic calculations. The SVM in MCSA was also used to classify faults. Based on laboratory test results, the proposed early fault detection system is able to reliably identify faults in particular those that are caused by inter-winding shorts.

Important Links:



Go Back