INTEGRATING COMPUTER VISION AND DIGITAL TWIN TECHNOLOGY FOR AUTOMATED CRACK DETECTION AND STRUCTURAL SAFETY ASSESSMENT OF REINFORCED CONCRETE BRIDGES

Authors

  • Imran Iqbal
  • Sabaar Alam
  • Tanveer Hussain
  • Fahad Ahmad

Keywords:

Reinforced Concrete Bridges; Computer Vision; Digital Twin Technology; Automated Crack Detection; Deep Learning; Structural Health Monitoring; Finite Element Analysis; Structural Safety Assessment.

Abstract

The progressive deterioration of reinforced concrete (RC) bridges, particularly through crack initiation and propagation, presents significant challenges to structural integrity, serviceability, and long-term infrastructure resilience. This study develops an integrated framework combining deep learning-based computer vision and digital twin (DT) technology for automated crack detection, quantitative damage characterization, and structural safety assessment of RC bridges. A comprehensive dataset comprising 12,000 annotated concrete surface images was employed, with 70%, 15%, and 15% allocated to training, validation, and testing, respectively. A YOLO-based object detection model was implemented for automated crack localization, while a U-Net segmentation architecture enabled precise extraction of crack geometries. Experimental results demonstrated robust detection performance, achieving a precision of 94.2%, recall of 92.6%, and mean average precision (mAP@0.5) of 95.1%. Furthermore, the segmentation model achieved an intersection over union (IoU) of 87.4% and a Dice similarity coefficient of 93.3%, confirming its effectiveness in identifying complex crack patterns. Quantitative crack characterization, incorporating image calibration and geometric analysis, yielded a mean absolute crack-width measurement error of 0.12 mm compared with reference measurements. Subsequently, the detected cracks were spatially integrated into a three-dimensional digital twin of the bridge, enabling damage visualization, condition monitoring, and finite-element-based structural performance evaluation. Structural analysis revealed an 8.7% reduction in effective flexural stiffness and a 6.3% increase in maximum service-load deflection compared with the undamaged structural configuration, indicating measurable deterioration in structural response. The integration of computer vision outputs with physics-based digital twin modeling established a systematic relationship between surface-level crack characteristics and component-level structural performance. The proposed framework enhances the objectivity, consistency, and efficiency of conventional bridge inspection methodologies while facilitating data-driven maintenance prioritization and informed structural safety evaluations. The findings demonstrate the potential of integrating artificial intelligence and digital twin technologies to advance automated structural health monitoring, support predictive infrastructure management, and improve the long-term reliability and sustainability of reinforced concrete bridge systems.

Downloads

Published

2026-11-16

How to Cite

Imran Iqbal, Sabaar Alam, Tanveer Hussain, & Fahad Ahmad. (2026). INTEGRATING COMPUTER VISION AND DIGITAL TWIN TECHNOLOGY FOR AUTOMATED CRACK DETECTION AND STRUCTURAL SAFETY ASSESSMENT OF REINFORCED CONCRETE BRIDGES. Spectrum of Engineering Sciences, 4(11), 1–36. Retrieved from https://www.thesesjournal.com/index.php/1/article/view/4092