How Can AI Discover Structural Defects in Bridges?

Bridge failures are disastrous and often deadly. The conventional method for detecting defects is also much less involved than many people might think. Sounding is a non-destructive way to find flaws in concrete structures — including bridges — by asking individuals to interpret what they hear. A more advanced option is to use artificial intelligence. The techniques with which AI identifies bridge defects vary.

When relying on sounding, people hit or drag chains across the concrete, listening for any flat or hollow sounds. Those could indicate defects since flaw-free concrete should have an associated ringing sound when engaged in these ways. However, this method is time-consuming and subjective. Using AI could result in a revolution that makes people safer and improves current processes.

More Than 200,000 U.S. Bridges in Poor Condition

The specifics of bridge maintenance vary depending on which authorities oversee the structures. However, it’s primarily a reactive method. The process starts with a periodic check by an inspector, but as many as 72 months can pass between inspections in the United States. The later parts of the process depend on what that professional finds. How fast defects get addressed depends on their severity and the transportation authority’s budget, too.

However, letting so much time elapse between bridge checks could prove problematic. A 2021 report from the American Road & Transportation Builders Association found 36% of bridges in the U.S. need significant repairs or replacement. The total figure is almost 224,000. Even more worrisome is 167.5 million crossings by car occur on bridges classified as “structurally deficient.”

The good news was a 3.2% drop in bridges in poor condition from 2020 to 2021. However, the researchers said it would take more than three decades to fix all those needing work at the current rate.

Evidence also suggests current inspection methods fall short. Consider the case of a massive crack in a bridge across the Mississippi River — snapshots from amateur photographers showed the defect had existed since at least 2016. However, annual inspections failed to spot it until years later and work to fix the issue only began in 2021.

One inspector lost his position due to the egregious oversight. If AI identifies bridge defects instead, could it do a better job? Some researchers and commercial technology providers think so. These efforts are still in the early stages, but they’re well worth exploring.

Australian Authorities Test New AI Bridge Inspection Method

Current methods of checking bridges can take days and require closing lanes of traffic. However, infrastructure officials in New South Wales, Australia, believe letting AI help could accelerate the inspection process. They recently completed a three-week trial of a method relying on AI-enabled drones. The organization using the AI oversees 4,000 bridges across New South Wales. Not surprisingly, decision-makers are always looking for effective and feasible ways to make the maintenance process faster, safer and less disruptive.

In this case, drones can fly close to the bridges, taking 4K images of them. The drones also make 3D maps of the bridges’ surroundings. Leaders believe this technology will play a vital role in evaluating bridges after natural disasters. The people involved in the project also trained water-resistant drones within the 20-vehicle fleet. They say this will let them continue the inspections in rainy weather and check submerged parts of structures.

AI Identifies Bridge Defects Faster Than People

One of the main advantages of AI is that it can pick up on things humans may initially miss. It often does this by analyzing gigantic quantities of data in relatively short timeframes. A company called Dynamic Infrastructure capitalizes on this concept by having AI algorithms evaluate past inspection reports to find changing defects and emerging maintenance risks.

People at the organization recently tested the tech on bridges in the United States, Sweden and Australia. They compared the AI’s results with that of human engineers. The algorithm found nearly 91% of the actual defects and identified more than 99% of those categorized as poor or severe condition states. It got those outcomes faster than humans could.

People can analyze the results within minutes, giving them the details they need to decide how to prioritize certain bridge issues. This controlled assessment involved using AI to examine steel and concrete bridges of various ages and climates and confirmed it identifies bridge defects with consistent success. Thus, it’s a viable option for decision-makers who are ready to inspect bridges differently and may struggle with a lack of people to do the job.

AI isn’t the right choice in every case. After all, it typically takes a lot of data and time to train models properly. However, it’s still worth exploring, especially when infrastructure managers know their current methods need improvement.

Using AI Along With Other Technological Advancements

Some people thinking about using AI to help with bridge inspections may already have other advanced technologies on those structures. In such cases, successfully deploying AI with these could make people increasingly open to the possibilities.

Poor lighting can raise safety issues and cause productivity decreases. That’s true in workplaces worldwide, but it can affect travel across bridges, too. The earliest examples of bridge lighting were primarily to ensure drivers did not have accidents while crossing them. However, starting a couple of decades ago, they got more advanced. For example, some bridge lights become more or less intense depending on the time of day.

Many lights also have internet connectivity, so their statuses feed into a central dashboard. Then, if a weather event damages a bridge light or it simply stops working, those in charge of monitoring the infrastructure will immediately know.

This improved oversight aligns well with exploring how AI identifies bridge defects. In the United Kingdom, officials want a future where bridges remain well-maintained without closures and believe they can reach that goal with the help of AI. People working on the project have trained their algorithm with images of different bridge problems. After one year, researchers collected 25,000 pictures to use as training data. They hope to rely on AI to find defects and classify them by severity.

The connected bridge lights mentioned above help people react faster to outages. Similarly, this project could help people become more aware of when to perform maintenance to minimize disruption. It’s a three-year endeavour, so it’ll be a while before those involved know the outcomes. Even so, the effort seems promising.

AI Could Keep the World’s Bridges Safer

Bridges are crucial for providing the infrastructure that helps people get where they need to go. Conventional methods of inspecting these structures often require road closures and people sometimes miss significant problems. Thus, some transportation experts are looking at how AI might provide much-needed improvements. Such efforts remain in the early stages, but they’ll undoubtedly offer valuable lessons for the future.