FAULT DETECTION OF CRACKED BEAMS LIKE STRUCTURE USING ARTIFICIAL NEURAL NETWORK (ANN) APPROACH

Rabinarayan Sethi, D.R.K. Parhi, S.K.Senapati

Abstract: This research work presents to train an artificial neural network to damage prediction of the beam with various possible fractured conditions. An experimental investigation has been conducted on the cantilever beam structures to measure the vibration responses, both in the fractured and un- fractured conditions. A study of the dynamic behavior of a damaged and un- damaged conditions beams expose that, there exists a close relationship among the relative natural frequencies at different modes at different relative crack location and depth. The entire data obtain from experimental procedure have been used as the basis to formulate the technique based on Artificial Neural Network (ANN). These outputs of the experimental analysis were mainly used as inputs data for training and analysis the artificial neural network. The Radial basis function (RBF) network implemented using MATLAB had engaged in this study. The spotlight of this work has been to study the probability of using an artificial neural network trained with only natural frequency data to assess of the damage in fractured cantilever beam. The developed neural network system can predict the location and intensity of the crack in a close proximity to the actual results.

Keywords: Vibration, Damage, Relative Natural Frequency, Artificial Neural Network.

DOI: https://doi.org/10.15623/ijret.2016.0525007