Work

An Integrated Computer Vision–Digital Twin Framework for Structural Health Monitoring and Comparative Environmental Risk Assessment of Reinforced Concrete Bridges Toward Sustainable Urban Development

Year: 2026
Type: article
Abstract: The progressive deterioration of reinforced concrete (RC) bridges driven by aging, increasing traffic demands, and aggressive environmental exposure presents significant challenges to structural reliability, infrastructure resilience, and sustainable urban development. This study presents an integrated computer vision–digital twin (CV–DT) framework incorporating Mask R-CNN with a ResNet-50 feature pyramid network (FPN) for automated defect segmentation, structural condition assessment, and comparative environmental risk evaluation. A dataset comprising 6,000 annotated images collected from 12 RC bridges was utilized, representing cracking, concrete spalling, exposed reinforcement, and intact concrete surfaces. The proposed deep learning model was trained and evaluated using a 70:15:15 dataset partition, enabling pixel-level defect localization and quantitative damage characterization. The model achieved a precision of 93.1%, recall of 90.4%, and mean average precision ([email protected]) of 94.2%, while image-based crack-width estimation demonstrated a mean absolute error of 0.18 mm. Subsequently, segmented defects and associated damage indicators were integrated into a bridge information modeling–enabled digital twin to facilitate spatial deterioration mapping, condition visualization, and risk-informed maintenance planning. Comparative environmental risk assessment was performed using a multi-criteria decision-making approach based on the Analytic Hierarchy Process (AHP), incorporating chloride exposure, flood susceptibility, temperature variability, and traffic-related pollution. The assessment identified substantial environmental vulnerability differences, with high-exposure bridges exhibiting an average composite risk index of 0.76 compared with 0.39 for low-exposure bridges. Furthermore, a scenario-based life-cycle assessment indicated that preventive maintenance strategies informed by integrated structural and environmental risk indicators could reduce greenhouse gas emissions by approximately 18% relative to conventional reactive maintenance practices. Cross-bridge validation and sensitivity analyses were considered to evaluate model transferability and environmental risk-ranking robustness. The findings highlight the potential of combining instance segmentation, digital twin technology, and environmental risk modeling to improve structural deterioration assessment and infrastructure decision-making. The proposed framework provides a scalable approach for intelligent bridge asset management, optimized maintenance prioritization, reduced environmental impacts, and the long-term sustainability of urban transportation networks.
Language: English
Open Access status: hybrid
OpenAlex ID: W7235678331
FWCI:
Cites: 37
Cited by:
Related to: 10