Automated Bridge Inspection using Digital Image Correlation and other Vision-based Methods

Building on the work previously performed as part of this study, the methods developed for fatigue crack characterization using digital image correlation (DIC) will be applied to full-scale structures prone to both fatigue and fracture failures. This will include full-scale sign structure components experiencing fatigue loading and subsequent cracking, as well as representative large-scale girder specimens prone to constraint-induced fracture failure. Additionally, optical data will be generated and collected to use in evaluating the potential for machine learning and artificial intelligence methods in fatigue crack identification and characterization. This phase of the project represents deployment of the previously-developed methodologies while still looking forward to other enhanced vision-based tools. Deployment mechanisms will include various hand-held and stationary cameras, unmanned aerial vehicles (UAVs), and/or augmented reality devices such as the Microsoft HoloLens2. It is anticipated this research program will lead to vision-based inspection tools that can potentially be used in automated bridge inspections.

The project is divided into three phases composed of six individual tasks: Phase I: Deployment of Vision-based Inspection Methods to Full-scale Structures Task 1: Data collection and analysis on full-scale sign structure subjected to fatigue loading. Task 2: Application of vision-based methodologies to fracture-prone girder detail. Phase II: Evaluation of Machine Learning Applications with Fatigue Crack Data Task 3: Generation and collection of optical fatigue crack data. Task 4: Application of machine learning to crack inspection scenarios. Task 5: Analysis and interpretation of results. Phase III: Final Reporting Task 6: A final project report detailing the results of Tasks 1 through 5 will be compiled and submitted to MATC.

It is anticipated that at the end of this project a system/method of vision-based crack inspection will have been deployed and evaluated for large-scale structural inspection applications. In addition to the DIC applications previously developed and evaluated, the project will evaluate multiple deployment mechanisms and data analysis tools. This study will create the potential for the system to be used as part of an automated bridge inspection process.

Developing a methodology and implementation process that allows DOTs to move away from traditional, human-centered bridge inspections drastically increases worker safety associated with inspecting aging infrastructure suffering from lack of maintenance and infrastructure damaged during an extreme event. Additionally, since the inspection technique being developed is expected to be largely invisible to the traveling public while it is occurring, the methodology is also expected to produce safety gains for the traveling public (i.e., reducing traffic accidents associated with having bridge snoopers, DOT workers, traffic control on the bridge). A vision-based inspection methodology is applicable to a number of contexts, including: - Detecting fatigue cracks produced by normal traffic loading - Detecting cracking produced after an unexpected event, such as a truck-strike, blast loading, or a seismic event. Any of these events could produce growth in pre-existing fatigue cracks, or initiate new cracks, making inspection a necessary step before confidence in the structure can be restored.

Download the Final Report