ARC Damage Detection and Quantification using Infrastructure Digital Twins

Overview

This project develops a scalable, data-driven structural health monitoring framework that fuses computer vision, vibration testing, and physics-based modelling within a 3D digital-twin environment. Field imagery (UAV/handheld) is processed to detect and map defects (e.g. cracking, spalling, corrosion) and to reconstruct asset-level 3D models. These observations are linked to changes in stiffness, load-carrying capacity, and damage propagation using numerical simulations updating informed by ambient/forced vibration data. The proposed methods will be validated and adapted for critical infrastructures like bridges accounting for different geometries, boundary conditions, and environmental variability. The outcome is a quantitative, repeatable condition-assessment tool that converts visual and field measurements into decision-ready indicators for asset owners, enabling risk-based maintenance, prioritised interventions, and reduced lifecycle costs across large infrastructure networks.

Objectives

1. Build a robust computer-vision method (detection / segmentation) for field imagery to automatically identify and size defects relevant to bridges and transport structures.
2. Generate 3D digital twins that aggregate 2D annotations and enable traceable, metric-accurate defect quantification.
3. Conduct vibration testing and implement numerical model updating to estimate structural damage induced stiffness loss and modal changes.
4. Develop numerical models to simulate defect initiation/ propagation and relate observed damage patterns to residual capacity and serviceability/safety margins.
5. Create surrogate and deep-learning models trained on synthetic and field data to predict structural condition directly from images/twin features.
6. Establish correlation mappings from visual/dynamic features to quantitative KPIs (e.g., stiffness reduction, capacity ratio, reliability index).
7. Package the methods into an operational toolkit with standardised data schemas, reporting templates, and API hooks for asset-management systems.

Industry Outcomes

This project aims to deliver the following key industry outcomes:

1. Quantitative, repeatable condition ratings that link observed defects to structural performance (beyond qualitative visual grades).
2. Portfolio-scale triage and prioritisation, enabling owners to rank by performance margin, and predicted deterioration.
3. Reduced inspection cost and downtime through UAV-enabled surveys and automated analytics, focusing human effort on high-value diagnostics.
4. Decision support for maintenance, including “what-if” simulations on the digital twin to test retrofit and load-limit scenarios.
5. Standardised deliverables (GIS/twin layers, dashboards, and concise factsheets) for transparent communication with stakeholders and regulators.
6. Technology transfer package (workflows, datasets, trained models) to integrate with existing asset-management platforms and field-data vendors.
7. Capability uplift for partners via training materials, protocols, and benchmark datasets to sustain ongoing deployment beyond the project period.