Digital Twins for Bridge Health Monitoring & Management

Research Team: PI: Ertugrul Taciroglu Team: Farid Ghahari

About this project:

We propose the development of an innovative integrated solution for damage identification (detection, localization, quantification) of bridge structures using a non-contact image-based measurement scheme. Computer vision techniques will be used to extract information from raw images, which will be used for joint finite element (FE) model updating and vehicular load estimation. The updated model can be maintained as a digital twin for damage diagnosis of the structural system. The digital twin can be utilized for global load rating and operational condition and post-disaster assessments through the life-cycle of the bridge, so as to help with bridge management decision strategies.

The key contributions of this project include the practical development and implementation of:
i. a portable non-contact measurement method using off-the-shelf camera systems and computer-vision techniques, which can be easily and quickly set up without traffic interruptions;
ii. a data assimilation technique that is based on Bayesian finite element model updating for joint vehicular load and system identification; and
iii. an online user interface to communicate data, to maintain the digital twin, and to provide actionable information to the stakeholders.

What problem does this research aim to address?

U.S. has a relatively older inventory of bridges, which makes the monitoring and maintenance of bridge stock a national priority. U.S. has a total of 614,387 bridges, nearly 40% of which are 50 years or older. The average age of America’s bridges keeps climbing and many of them are approaching the end of their nominally 50-year-long service lives. In 2016, one in 11 of bridges were structurally deficient and more than one in 8 were functionally obsolete. Replacement or rehabilitation of such a large number of bridges cannot be achieved overnight. Since structurally deficient and obsolete bridges will likely remain in service for a significant time period, ensuring public safety necessitates close monitoring of the bridges system to expand the service life, guide the inspection process, and prioritize maintenance and rehabilitation decisions. Current monitoring and maintenance is mainly based on visual inspections, which is labor-intensive, costly, time-consuming, and subjective (i.e., prone to human errors). There is a clear need for developing and implementing low-cost, practical, and technology-based solutions for bridge assessment to facilitate the inspection and management process, and to reduce associated costs and risks.

What are the expected impacts and benefits of the research?

The specific technical objectives of Phase I will be:
i. Design and implementation of a computer-vision technique to extract the deformation time-histories from raw images. Verification and validation of the technique using the image data obtained from realistic finite element model of an iconic bridge.
ii. Selection of a deep learning approach to detect and track vehicles, to identify their locations and types using (regular quality) image data. In this phase, the focus will be on finding suitable training and testing dataset and narrowing down appropriate features and methodology that are suitable for this application.
iii. Implementation and refinement of the Bayesian model updating technique for joint system and vehicular load identification.
Other deliverables include the final project report, archival journal articles, conference proceedings and presentations, and data from field and laboratory experiments. These data will be freely available to the public through the website, which is maintained by the PI’s research group.

By |October 3, 2019|Categories: Sustainable & Resilient Transportation
By |2019-10-03T17:13:08-07:00October 3rd, 2019|