A Digital Twin Framework for Bridge Structural Health Monitoring with Physics-Informed Graph Networks
PDF

Keywords

Digital Twin
Structural Health Monitoring
Physics-Informed Graph Networks
Bridge Engineering

Abstract

The aging of global infrastructure necessitates advanced methodologies for real-time Structural Health Monitoring (SHM) to ensure public safety and optimize maintenance schedules. Digital Twin technology, characterized by the seamless integration of physical assets and their virtual replicas, offers a transformative approach to predictive maintenance. However, traditional data-driven models for Digital Twins often suffer from poor generalization in data-scarce regimes, while high-fidelity finite element models are computationally prohibitive for real-time application. This paper proposes a novel framework: a Physics-Informed Graph Network (PIGN) designed specifically for the Digital Twin of bridge structures. By representing the bridge topology as a graph, where sensor nodes communicate via edges defined by physical connectivity, the model explicitly captures spatial dependencies. Furthermore, the integration of governing physical laws directly into the loss function ensures that predictions adhere to fundamental mechanical principles. We demonstrate that this hybrid approach significantly outperforms purely data-driven baselines in trajectory prediction and damage identification, particularly under noisy sensor conditions. The proposed framework reduces the dependency on massive labeled datasets while maintaining the computational efficiency required for continuous monitoring.

PDF
Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright (c) 2025 Wei Chen, John A. Smith (Author)

Downloads

Download data is not yet available.