In the rotating machines, maintenance of the high speed operated bearings is the major problem and is one of the key issues due to excessive vibrations. Hence, the vibration signatures can be used as a feature for the fault diagnosis. This paper presents the Artificial Neural Networks (ANN) based fault analysis, which is used to classify various known faults using the features extracted from the vibration signals. The vibration signals from the piezoelectric accelerometers are being measured for the following conditions — No defect (NOD), Outer race defect (ORD), Inner race defect (IRD), Ball fault (BF) and Combination of above (COMB). The features are extracted from the time domain using the statistical method. These features are filtered using wavelet filter & kernel filter and compiled as the input vectors. The multilayer neural network is trained by these input vectors. The training and testing results show that wavelet and kernel filter can be effective tool in the diagnosis of ball bearing faults using ANN. Results obtained from the ANN predict that the wavelet filter provides good accuracy with reduction in the training time.

This content is only available via PDF.
You do not currently have access to this content.