Automated Bridge Inspection Using Digital Image Correlation Part III: Examination Alternative Vision-Based Methods and Deployment Mechanisms for Field Implementation

Cracks in structures, and specifically fatigue cracks in metallic materials, present a serious problem in highway infrastructure due to the fact that they are difficult to detect and can potentially cause structural failure. Current fatigue crack inspection methods are very manual in nature, putting both inpsectors and the traveling public in danger. This project aims to develop an automated, vision-based inspection tool used to identify and characterize fatigue cracks that can be deployed in the field to reduce the amount of human interaction needed during inspections.

This research aims to expand upon the vision-based inspection methodology developed in previous phases of the project. Having developed a digital image correlation crack characterization methodology, the research now focuses on ways to make the process more robust and reliable. This is needed in order to deploy the developed inspection tool under in-service field conditions. Stability of the hardware used and accuracy of the methodology is to be examined within the context of what is necessary for a field deployment.

Researchers and industry practitioners are interested in moving towards automated inspection methodologies for infrastructure applications. Many are examining the use of robots and unmanned aerial vehicles for inspections, but there are currently no technologies that can be used with these inspection tools to identify realistic fatigue cracks. Most automated inspection technologies are only capable of identifying gross structural damage. Therefore, the focus on robotics or drones for inspection purposes is limited until technological tools are developed that can be employed on these devices. The project has developed a vision-based methodology for characterizing distortion-induced fatigue cracks, those most prevalent on highway bridges. The research now focuses on making the vision-based system more robust and reliable, allowing for future implementation.