Machine-Aided Bridge Deck Crack Condition State Assessment Using Artificial Intelligence

Area of Science:

• Civil Engineering
• Artificial Intelligence
• Computer Vision

Background:

• Federal Highway Administration (FHWA) mandates biannual bridge inspections, recorded in the National Bridge Inventory (NBI).
• Increasing complexity of inspection specifications, including element-level assessments, demands more inspector training and field time.
• Current methods face challenges in efficiently assessing detailed bridge element conditions.

Purpose of the Study:

• To develop and evaluate a machine-aided bridge inspection method using artificial intelligence (AI).
• To automate the condition state assessment of cracking in reinforced concrete bridge deck elements.
• To assist bridge inspectors in meeting new FHWA requirements for element-level inspections.

Main Methods:

• Utilized a deep learning-based workflow integrating image classification and semantic segmentation.
• Employed a deep neural network to extract critical information from bridge images.
• Developed a system to evaluate crack condition states according to FHWA specifications.

Main Results:

• Demonstrated the effectiveness of the AI workflow for assessing crack conditions in bridge decks.
• The AI method accurately extracts information required by bridge inspection manuals.
• The system enables objective condition state determination for cracks.

Conclusions:

• The AI-based method enhances the efficiency and accuracy of bridge deck crack inspections.
• This approach helps balance costs and risks associated with AI in infrastructure management.
• Departments of Transportation can implement this AI tool to improve bridge asset management and community service.