Intelligent fault classification of rolling bearings using neural network and discrete wavelet transform

This paper is about diagnosis and classification of bearing faults using Neural Networks (NN), employing nondestructive tests. Vibration signals are acquired by a bearing test machine. The acquired signals are preprocessed using discrete wavelet analysis. Standard deviation of discrete wavelet coefficient is chosen as the distinguishing feature of the faults. This feature vector is given to the design network as inputs. The input vector is normalized prior to be applied to neural network. There are four output neurons each of which corresponds to: 1) bearing with inner race fault, 2) bearing with outer race fault, 3) bearing with ball defect, and 4) normal bearing. The structure of NN is 6:20:4 and with 99 % performance