Application of vibration analysis and data mining techniques for bearing fault diagnosis in two stroke IC engine gearbox

Abstract

This paper is about monitoring of ball bearing used in the IC engine gearbox using condition monitoring techniques. Experiments are conducted on two stroke IC engine which is driven by the 3HP DC motor. Vibration signals are acquired from the gearbox with triaxial accelerometer. Ball bearing with good and induced faulty (outer race fault, inner race fault, ball fault, inner and outer race fault) conditions were used in the analysis. Fault diagnosis of the ball bearing has been carried out using data mining (DM) techniques. In DM there are three stages viz.; feature extraction, feature selection and feature classification. For all the conditions of bearing, statistical and empirical mode decomposition (EMD) features are extracted from the vibration signals. Decision tree technique (J48 algorithm) is used in the analysis for selecting significant features from the feature vector. From the chosen features, ball-bearing conditions are classified using random forest algorithm. Results obtained from the different classifiers were compared, and a better classification algorithm with a decision tree will be suggested for condition monitoring of the rotating components.