Damage detection in beams

Abstract

Identification of crack location and depth in structural elements such as beams, columns, and slabs, etc., subjected to time varying loads will help them to retrofit before failure thereby its life and structural capabilities can be greatly improved. It is not always possible to detect the initial cracks by visual inspection. Laboratory test results of isotropic simply supported beam and Cantilever beams with different crack location and crack depths data are collected from literatures as in terms of system natural frequencies and crack locations for single crack and multi cracks. Different models are developed for crack detention using Natural frequencies as Input and Crack Depth and locations as Output. At first an attempt made for simply supported beam having a Single crack detection using Artificial Neural Network (ANN) Model. Different ANN models are developed using different training functions and found Levenberg-Marquartd (LM) fits best among others which is evaluated in terms of R-square value. The Comparison has been made between developed ANN models and regression models. It shows that the ANN models can perform well than regression models. Further the simple equation has been presented for evaluation of crack depth based on the performance of the models developed. Secondly, the problem involves detection of multi crack damage in a Cantilever beam using ANN Model and Response Surface Method (RSM). To achieve better results in RSM higher degree equation is used. The final performances of each model are evaluated in terms of R- value. And, it is found that ANN fits the equation reasonably good. Crack Depth and Crack location can be detected using ANN approaches which may be helpful in retrofitting the structures.