A Data Driven Approach for Prediction of Overloading in Ball Bearing

Abstract

The study is to develop a data driven approach for the prediction of overloading in ball bearing, which is the major cause for the bearing faults next to improper lubrication. Prediction of overloading helps us to take remedial action and thus the faults can be prevented. In this study, a SKF 6204 ball bearing was chosen and it was overloaded beyond its fatigue load limit by creating unbalance with the help of rotor. Experiment was conducted at different load ranges such as light load, medium load, and safe load and over load and the vibration data (acceleration) was acquired at each condition. From the raw data, features were extracted and a classification model has been developed with four classes and it was classified using decision tree and support vector machine algorithm.