Application of Time-domain Features with Neural Network for Bearing Fault Detection

The normal functioning of induction motors depend largely on ball bearing. Proper functioning and operation of ball bearing is the result of its maintenance by condition monitoring. Out of the various measures of condition monitoring vibration monitoring is the most extensively used and economical technique to detect, identify and distinguish fault in induction motors. In this research paper induction motor faults have been detected using time domain features of vibration signal. For the automated diagnosis of faults neural network has been designed using values of nine statistical features as input parameters which are: peak value, root mean square value, crest factor, kurtosis, skewness, clearance factor, impulse factor, shape factor and standard deviation. The artificial neural network designed was applied for three conditions of bearing i.e. for healthy bearing, bearing with inner raceway defect and bearing with outer raceway defect. The experimental observation shows that the proposed method is effective and is able to detect the faulty condition with high accuracy.