IEEE International Electric Machines and Drives Conference (IEMDC 2009), Florida, United States Of America, 3 - 06 May 2009, pp.1408-1413
In this work, the mechanical faults that occur in induction motors, which are widely used in industry due to their simple and rugged construction and cost effective pricing, are analyzed. The certain fault modes are generated synthetically in induction motors and motors are run by a voltage controlled frequency inverter with four different speed references and three different load conditions to gather the digitized stator current data in laboratory environment. By applying wavelet packet decomposition, wavelet packet coefficients are obtained from stator current and these coefficients are used to obtain feature vectors by using statistical properties of the data. Fault detection is performed by classifying these feature vectors with different types of classifiers. As a result, correct classification rates are compared not only with faulty and healthy motors, but also with different fault modes. The effects of the speed reference, stator current level, selected filter type in wavelet packet decomposition and type of the classifier in fault detection is also observed.