The three-stage artificial neural network method for damage assessment of building structures

Abstract

Building structures are often huge and composed of a number of elements. It may not be possible to make modal measurements along the large number of degrees of freedom. Structural damage detection therefore becomes much more challenging both in terms of measurement and subsequent analyses. Accordingly, a problem in structural damage detection is requirement of a systematic and effective method. Among the developed damage detection techniques, artifi cial neural networks (ANNs) have become promising tools recently. The main drawback of using ANNs in structural condition monitoring is the requirement of enormous computational effort. To address this issue, a novel technique is proposed using "damage index" derived from frequency response functions (FRFs) with the three-stage ANN method to detect damage. The basic idea of this method is to establish features of damaged structure using FRFs from different measurement points. Then using these features, damage indices of damage cases of the structure are identifi ed. Damage indices corresponding to different damage locations and severities are introduced to ANNs. The effectiveness of the proposed method is validated using the fi nite element model of a 10-storey framed structure. The results show that the principal component analysis based damage index is suitable for structural damage detection.

Experimental research and numerical simulation of influence of pre-stress values on the natural vibration frequency of concrete simply supported beams

“…wherein, is the flexural stiffness of the beam; indicates the equivalent concentrated force of the pre-stressing tendon; means the partial mass of the simply supported beam. Through solving the above formula, the calculation formula of natural vibration frequency of the simplified pre-stressed simply supported beam was obtained as follows [15][16][17]:…”

Transfer Learning for CNN-Based Damage Detection in Civil Structures with Insufficient Data

“…As a subset of arti cial intelligence, machine learning has proven its high e ciency in many engineering applications and has been growing rapidly with great advances in sensor and computer technologies. Over the last decades, many machine learning algorithms have been utilized in the field of SHM, including, but not limited to, artificial neural network (ANN) [6][7][8][9], fuzzy neural network [10][11][12], support vector machine (SVM) [13][14][15], genetic algorithm (GA) [16][17][18], and federated learning [19,20].…”

Potential of neural networks for structural damage localization

“…feedforward nets, self-organizing maps, learning vector quantization) can become a quite effective damage detection tool when used in conjunction with the dynamic properties of a system (e.g., [4][5]) -note that nowadays is quite straight forward the accurate estimation of important dynamic properties (e.g., natural frequencies) of (possibly damaged) built structural systems (by means of accelerometers and/or other simple decices, and existing software -e.g., ARTeMIS Modal 4.0 [6]). According to Bandara et al [7] and Ahmed [8], a clear challenge concerning ANNs is the fact that they typically need structural data of both damaged and intact structures to be able to classify satisfactorily. If the structure is not considered damaged in its current state, the information regarding the damaged state will be unavailable unless detailed structural models are used to generate this information, such as numerical ones based on the Finite Element Method (FEM).…”