Skip to main content

Bridge safety: Using artificial intelligence to improve bridge inspections

Mostafa Tazarv
Mostafa Tazarv, an associate professor in the Jerome J. Lohr College of Engineering. 

Each year, the Alaska Department of Transportation and Public Facilities is responsible for inspecting roughly 1,000 bridges throughout the state.

Considering the size and relative remoteness of Alaska—it is larger than Texas, California and Montana combined—manually inspecting each of these bridges is both time consuming and costly.

Mostafa Tazarv, an associate professor in South Dakota State University’s Jerome J. Lohr College of Engineering, and Kwanghee Won, an assistant professor in the Lohr College of Engineering, are developing practical artificial intelligence tools that could revolutionize how bridges are inspected, saving both time and money. 

The two-year, $251,717 project is being funded through the ADOT&PF and the National Center for Transportation Infrastructure Durability and Life Extension, which is a U.S. Department of Transportation University Transportation Center.

Post-event damage assessment tools

Over the past couple years, Tazarv and Won have been working to develop BrDATs, a set of bridge damage assessment tools in the form of an internet application. The app uses computer vision and engineering analyses to scan and pinpoint cracks and other damages in bridge columns and other structural components and to assess the safety of the elements and the bridge.

While the app is still being fully designed and tested, the goal is to determine if a bridge is safe or unsafe simply by uploading pictures from the bridge under inspection. The developers want the app to make safe, consistent and accurate recommendations that can be understood by both engineers and non-engineers.

“Once a picture is uploaded, the AI software is going to find if there’s any cracking, if there is any spalling, if there is any bar exposed,” Tazarv explained. “There it will tell if the bridge is safe or not. The user of the bridge assessment tool doesn’t need AI knowledge, and you don’t need to be a structural engineer. We made an app where everything is being done behind the scenes for you.”

Originally, this assessment tool was designed for post-earthquake response and analysis. In major West Coast cities, situated along the San Andreas Fault, bridge inspections following an earthquake are an absolute necessity, but they can also incredibly time consuming. For example, if Seattle, Washington, were to have an earthquake, all of its bridges in the area would need to be inspected before they are reopened. Prior to Tazarv’s and Won’s project, the only real way to inspect them was to send a certified inspector to each bridge.

“You need hundreds of engineers in the area following an earthquake. Where are you going to get all of those experts?” Tazarv explained. “That is a major challenge. But if you have an application like (the bridge damage assessment tools), even the local Department of Transportation folks (non-experts) could do it.”

Because non-AI experts and non-bridge engineers can easily use and understand the tool, post-earthquake response and assessment would be much more efficient and resourceful.

The Alaska Department of Transportation learned of this assessment tool and asked if AI could be used for routine inspection, which led to Tazarv’s and Won’s current project.

While the project is being funded by the ADOT&PF, the application of AI for routine inspection is applicable to the entire U.S. Every two years, every single bridge in the U.S. must be inspected as part of the country’s routine inspection requirements. Because there are more than 600,000 bridges, more than 300,000 need to be inspected every year.

Considering that previously all inspections were done manually—which included hand-measuring cracks and potholes and spending hours at each site—a tool that helps with routine inspection would create more consistent and efficient results. The use of non-destructive methods such as ground-penetrating radar is gaining interest but is not very common for routine inspections.

“The goal for the ADOT&PF project is to bring AI and expedite damage detection and/or measurements through AI,” Tazarv explained. “That would save a lot of time.”

2023 and beyond

In 2023, two more projects funded by ADOT&PF will begin. The first will be a continuation of the post-earthquake assessment tool but including sub-standard columns, those designed with old codes.

“We need to upgrade our tool for the different types of bridges out there,” Tazarv said. “We started from the current bridge technologies, and we are going back in time. We need to include older columns, non-seismic detailing and all the structural conditions to make sure that after an earthquake, our tool is accurate and working, for all bridges.”

The second project beginning in 2023 will have Tazarv, Won and Marco Ciarcia, an assistant professor in SDSU’s Department of Mechanical Engineering, investigating if drones can help calculate the displacement of a bridge during load testing—used to determine if a bridge is healthy and working correctly. For example, a bridge may have six beams supporting it. When a truck passes over a bridge, all six of those beams should be working in unison to take on the truck’s load. If only one or two beams are supporting the load, the bridge isn’t functioning correctly. Load testing is an experiment to ensure the bridge is working correctly and the bridge’s posted load is accurate.

For load testing, a team of engineers and inspectors will use conventional instruments and sensors to judge the bridge’s performance and calculate the maximum load size. Load testing using conventional sensors usually needs hours of preparation. In this new project, the research team proposes that drones could “watch” bridge girders deform under moving load and translate that into the bridge displacement—again, saving time, money and resources.

“By flying a drone, it will watch those beams for you,” Tazarv said. “When there is a truck, it will lock in a pixel and estimate how much those pixels move up and down, and then we’ll translate it to displacement.”

The team will also travel to Alaska for some on-site inspections and further research.

“This is a true collaboration—combining expertise in structural engineering, computer vision (Won) and aerospace engineering (Ciarcia),” Tazarv said. “We are bringing expertise from basically three different areas and making something that was not feasible individually.”

In the future, Tazarv is hopeful that the bridge damage assessment tools will become standard practice for post-event responses.