Vibration Based Fault Detection of Deep Groove Ball Bearing Using Data Mining Algorithm

Abstract

Deep groove ball bearing is a heart of rotating machinery. So, early fault detection of bearing can prevent failures of the machineries. Vibration signals collected from bearing carries useful information about its health. This paper presents a methodology to identify various faults in deep groove ball bearing from vibration signals acquired from different bearing condition. Features such as RMS, Variance, Mean, Crest Factor, Kurtosis and Skewness are calculated from time domain for various bearing conditions such as normal bearing, fault at inner race, fault at outer race, and fault on ball. The dataset of the various bearing condition is applied on five classifiers such as Naive Bayes (NB), Multi-Level Perceptron (MLP), K-Star, J-Rip, and J-48 using data mining algorithm WEKA. The distribution of training and testing dataset is carried out using WEKA. In a result, statistical parameters generated from classification algorithms are compared to determine the correctly classified instances and to find the efficient classification algorithm among five algorithms. Result shows that K-Star gives highest accuracy for training as well as for testing among all classification algorithms.