Fault Diagnosis of Electric Motors Using Vibration Signal Analysis

Abstract

Electric motors find their application in almost every industry or plant. As the processes of various plants rely on these machines, presence of any fault in the motor is a matter of great concern and in such situations an efficient fault diagnosing technique must be developed to locate the fault and rectify it in least possible duration. As motor bearing faults are the most abundant faults present in an electric motor, this research work presents fault diagnosis methods of these faults using vibration signal analysis. A public domain vibration database containing vibration signals acquired both for normal operation as well as motor’s operation with inner and outer raceway, rolling element faults present was used for analysis. The presence of bearing faults in the motor was detected using time domain analysis of statistical features of vibration data. Further the fault location was fetched using cepstrum analysis of vibration signals. Then lastly, fault classification was performed using Support Vector Machine (SVM), Artificial Neural Network (ANN) and KNearest Neighbour (KNN) classifiers. Two different cases of faults were considered, first with faults of uniform dimensions and second with faults of varying dimensions introduced in the motor bearing components. A hybrid model for bearing fault diagnosis and severity level classification has been proposed in this research work. The cepstrum analysis technique presented possesses the capability to locate any type of bearing fault present. The classifiers used have produced high accuracy for bearing fault classification and also detected the severity level of the fault present.