Feature extraction of ball bearings in time-space and estimation of fault size with method of ANN

Abstract

Faults in bearings used in machines cause downtime and leads to catastrophic results on the machining operations. In this study, specific sizes of the artificial bearings defects are created and vibration signals were obtained from a shaft-bearing system. The purpose of this study is to diagnose the size of the defects occurring in bearings by using Artificial Neural Networks(ANN) model. Features of vibration data are extracted in real time and are multiplied with specific weights; then they were given as input to the ANN model. Statistical properties of bearings faults are observed that their values vary depending on fault dimensions in real-time. These features are examined by using ANN and the size of the defects occurring in bearings are classified with 100% success, on the other hand the prediction permonfance of actual error for a ANN model is found 2%.