Comparison between Artificial Neural Network and Support Vector Method for a Fault Diagnostics in Rolling Element Bearings

Rolling element bearings are the most crucial part of any rotating machines. The failures of bearing without warning will result catastrophic consequences in many situations. Therefore condition monitoring of bearing is very important. In this paper, artificial intelligence techniques are used to predict and analyses the bearing faults. Experiments were carried out on rolling bearing having localized defects on the various bearing components for wide range of speed and vibration signals were stored. Condition monitoring systems is divided in two important part one feature extraction and second diagnosis through extracted features. Daubechies wavelet is popular for smoothing of signals so, it is chosen for reducing the background noise from vibration signal. Kurtosis, RMS, Creast factor and Peak difference as suitable time domains features are extracted from decompose time velocity signals. Back propagation multilayer neural network was train and tested by 369 pre-treated normliesed features. Support vector machine is also used for the same data for predicting bearing faults. Finally, it is found that Support vector machine techniques gives better results over ANN.