Evaluation of crack locations in beam using artificial neural network‐based modified curvature damage index

Abstract

Although the frequency response‐curvature methodology is commonly used to detect irregularities in mechanical and civil structures, the artificial neural network‐based frequency response‐curvature damage index method may have good efficacy in the detection and localization of structural damages. By utilizing experimental data sets, a novel method is proposed to pinpoint a saw‐cut damage location and the degree of damage in beam models. Using a dynamic data logger, the frequency response function of a beam model is obtained for altering damage levels at different positions. As frequency response data contains environmental and operational noise, the accuracy of obtained results may get reduced. To improve the accuracy by reducing the noise effect, the experimentally obtained frequency response data is trained through an artificial neural network. Using central difference approximation, the sets of trained modal data are utilized to determine the improved mode shape curvature. The curvature damage index is then obtained by using the improved mode shape curvature for different damaged scenarios to ultimately identify structural damages.

A novel sensor placement strategy based on an enhanced damage assessment method in beam‐like structures: Saddle point criteria

“…In terms of structural integrity, the detection of multiple damage locations remains a greater challenge than the detection of single damage in civil infrastructure protection [12]. A new damage sensitive parameter is proposed by upgrading an existing damage-sensitive parameter.…”

“…This study aims to propose a novel sensor deployment procedure based on a novel damage identification method. In the literature, an effort to find an optimal damage indicator for beams has been spent on numerous studies [12]. The damage identification part of this study focuses on sensitive damage location detection with the proposed damage indicator.…”