Utilizing mode shape curvature and artificial neural networks for structural damage assessment in beam-like structures

Abstract Damage within a structure refers to changes in both its geometric and material characteristics, resulting in a drop in the stiffness that impacts the structure’s performance adversely. This decrease in stiffness causes alterations in modal parameters, including natural frequencies and mode shapes. Utilizing modal analysis allows for the extraction of modal frequencies and mode shapes, facilitating the analysis of mode shape curvature to detect structural damage. In recent years, artificial neural networks (ANNs) have achieved significant application, mainly for their exceptional capability in pattern recognition, which proves invaluable for identifying structural damage. This article proposes a novel method based on mode shape curvature and ANNs for detecting damage in beam-like structures. Experimental study is conducted to analysis damaged and undamaged structural modal behaviours. A feed-forward neural network with two hidden layers, trained on damage indices from mode shape data, is used to accurately pinpoint damage locations within the structure. The proposed approach for damage detection is validated and proves its ability to precisely pinpoint the location of damage. The results of this study demonstrate that ANNs trained with modal curvatures hold significant promise for identifying structural damage, enabling early detection in beam-like structures and contributing to ensuring their safe operation.