Date of Award

Spring 2014

Document Type

Dissertation/Thesis

Degree Name

Master of Technology Management-Engineering Technology

First Advisor

Sri Kolla

Second Advisor

David Border

Third Advisor

Todd Waggoner

Abstract

Induction motor is one of the most important motors used in industrial applications. The operating conditions may sometime lead the machine into different fault situations. The main types of external faults experienced by these motors are over loading, single phasing, unbalanced supply voltage, locked rotor, phase reversal, ground fault, under voltage and over voltage. The machine should be shut down when a fault is experienced to avoid damage and for the safety of the workers. Computer based relays monitor the machine and disconnect it during the faults. The relay logic used to identify these faults requires sophisticated signal processing techniques for fast and reliable operation. Artificial Intelligence (AI) techniques such as Artificial Neural Networks (ANN) have been applied in induction motor relays. Though the ANN based methods are reliable, the selection of the ANN structures and training is time consuming. Recently it is observed that the AI techniques such as Support Vector Machines (SVM) are alleviating some of the limitations of ANN method.

The objectives of this study are to develop a SVM based induction motor external faults identifier and study its performance with real-time induction motor faults data. Data collected from a 1/3 hp, 208 V three-phase squirrel cage induction motor is used in this project. Radial Bases Function Kernel is used to train and test the SVM, though the effect of other Kernel functions was also studied. The proposed SVM method uses RMS values of three-phase voltages and currents as inputs. The testing results showed the efficacy of the SVM based method for identifying the external faults experienced by 3-phase induction motors. It is observed that the performance of the SVM based method is better than the ANN based method both in model creation and testing accuracy.