Crack detection in beam-like structures using a wavelet-based neural network

Abstract

This article proposes a method for crack identification using a wavelet-based neural network (NN; wave-net). The input data for the wave-net training are both global and local in-plane vibrational parameters of beam-like structures. In this study, the vibrational parameters of intact and damaged beams are obtained using the finite element method. Different cracks are introduced in the span of the beam with different locations and depths to obtain necessary data for training NN. The identification results are compared with those of some convectional NNs including radial basis function and multilayer perceptron ones. Results show good accuracy and efficiency of the proposed NN method.