Development of digital twin concept to detect changes in structural behaviour

Abstract

The rapid advancement of digital technologies and data-driven approaches in structural engineering has opened new avenues for real-time damage identification in infrastructure. Digital Twins (DTs), which are digital replicas of physical assets, have emerged as powerful tools for monitoring structural health. This study focuses on the development of a DT for real-time structural damage identification through advanced modelling techniques, such as the Reduced Basis (RB) method, and artificial intelligence technologies, particularly Deep Learning (DL), while accounting for model uncertainties. A hybrid DT model is con- structed by integrating mathematical formulations based on structural mechanics into a machine learning framework. An accelerated RB Finite Element (FE) model is employed to analyse linear elastic structures subjected to static loading, with stiffness reduction of structural elements introduced as an indicator of dam- age. During the offline phase, the RB model is used to train an adaptive Neural Network (NN) to detect and classify deviations in the structural behaviour from the undamaged state. In the monitoring phase, the trained NN analyses sensor data from a sensor array, optimally placed on the physical structure, to identify potential damage scenarios. The RB model is then utilised to assess the severity of the detected damage. A systematic experimental programme was conducted to validate the methodology on a laboratory-scale truss structure subjected to static loading. Artificial damage was introduced by reducing the cross-sectional area of selected truss elements. Real-time synchronisation between the physical structure and the virtual model was established using Internet of Things (IoT) technology, facilitated by a custom-designed Printed Circuit Board (PCB) and supported by communication protocols and a cloud server. The developed DT mirrors the real-time structural behaviour and provides accurate online assessment of structural damage. This methodology enhances the efficiency of real-time damage detection, offering a pathway to proactive maintenance and extended infrastructure lifespan. Keywords: Damage diagnosis, Deep Learning (DL), Digital Twin (DT), Internet of Things (IoT), Pre- dictive maintenance, Reduced Basis (RB), Structural Health Monitoring (SHM).