Dadoulis, GeorgiosGeorgiosDadoulisManolis, GeorgeGeorgeManolisKatakalos, KonstantinosKonstantinosKatakalosDragos, KosmasKosmasDragosSmarsly, KayKaySmarsly2024-11-152024-11-152024-10-29Engineering Structures 322: 119216 (2025)https://hdl.handle.net/11420/51924Damage detection via vibration testing typically relies on damage-sensitive features, which serve as “damage indicators”, and decisions upon the existence of damage are based on comparing the damage indicators retrieved from two distinct structural states. However, the relatively low sensitivity of damage indicators to the onset of structural damage remains an open question, despite the considerable research efforts in vibration testing over the years. Low-sensitivity problems may be particularly exacerbated by the complex dynamic behavior of lightweight structures, such as lightweight bridges subjected to vehicular traffic. In particular, due to material (and, by extension, mass) reduction in lightweight bridges, vehicles essentially act as “traveling masses”, which are comparable to the structural mass and result in a coupled complex dynamic motion problem that may obscure typical damage indicators used in vibration testing. This paper presents a damage detection approach for lightweight bridges with traveling masses, leveraging the powerful feature-extraction capabilities of machine learning (ML). In particular, a convolutional neural network (CNN) is trained to classify acceleration response data, collected from vibration testing, into damage scenarios. The training data for the CNN are created via simulations of damage scenarios, using calibrated analytical models. The damage detection approach is validated in laboratory tests on a continuous beam, showcasing the capability of the CNN to classify damage scenarios of the beam. The outcome of this paper aims to serve as a starting point towards employing ML for damage detection in the context of vibration testing as well as structural health monitoring.en0141-02962024ElsevierAnalytical modelingArtificial intelligenceDamage detectionMachine learningStructural dynamicsVibration testingTechnology::690: Building, ConstructionJournal Article10.1016/j.engstruct.2024.119216Journal Article