Modeling of Amplitude and Frequency of Vibration for Rotor Bearing System using Ann and Taguchi Methods

Abstract

Vibrations in rotor bearing systems are mostly produced by misalignment, imbalance, mechanical looseness, cracks and other malfunctions. Now-a-days, diagnostics of rotor defects gaining importance day by day. This paper introduces a new methodology to detect cracks based on amplitude and frequency of vibration data. Experiments were conducted at different levels of crack depth, its position and shaft rotational speed on a shaft which was held between two bearings. Experimental results of amplitude and frequency of vibration were measured in axial, vertical and horizontal directions at both the bearings. Signal to noise ratios were calculated for the experimental results using Taguchi method to identify significant parameters which are having effect on the experimental results. Neural network models were developed for the experimental results to predict them. The neural network was trained with the experimental data and predicted the responses. In addition to that, Taguchi method was also used to predict the responses. A comparison was carried out among the experimental results, artificial neural networks (ANN) predicted values and Taguchi predicted values. The comparison reveals that there is good agreement among them and the ANN and Taguchi methods can be used to predict the amplitude and frequency of responses.