Corroded bridges: Drones and AI can reduce risks
Drones and an AI can detect corrosion on steel bridges largely automatically. However, the damage must still be assessed by humans.
A corroded bridge section as seen by the AI: The coloring shows corroded sections. The darker the color, the deeper the corrosion.
(Image: NTT (Screenshot))
The Japanese technology companies NTT Corporation and NTT e-Drone Technology Corporation have jointly developed a method of detecting corroded steel on bridges using drones and image recognition based on artificial intelligence (AI). The method is said to be significantly superior to conventional visual inspections and can detect the need for repairs to steel-based bridges at an early stage. The safety of the bridges is thus better guaranteed and the cost of maintenance would be reduced.
The maintenance of steel-based bridges is difficult because corrosion often occurs in places that are difficult to access. Conventional inspection methods based solely on visual inspection are not sufficient. Although they can detect corrosion, they cannot assess the depth of corrosion. However, it is decisive in determining whether repairs are required and how much.
Corrosion depth detection through AI image recognition
NTT wants to tackle this problem with drones and advanced image recognition based on AI. The drones are equipped with cameras and fly over the bridges. They can also reach areas that are difficult to access. The recorded images are analyzed by an AI. It can not only detect damage, but also provide information on the possible depth of corrosion by evaluating the images. Ultrasound measurements can therefore be omitted. The information is then evaluated by experts and the damage is finally assessed.
Videos by heise
There is no need to erect expensive scaffolding to assess the damage on bridges, and there is no need to use ultrasound probes manually to determine the depth of corrosion. The human error source during detection is also largely eliminated. With conventional corrosion detection, the evaluation of the visual inspection and the measurement results depend heavily on the experience of the assessing engineer. With the new method, on the other hand, assessors only have to make a final assessment of the extent of the damage. This increases the efficiency and accuracy of bridge inspections.
The two NTT companies tested the technology on a steel bridge in Kumagaya City in Japan. The drones independently recorded detailed images of the bridge, which were automatically analyzed by an AI to identify possible corrosion damage. Thanks to the extensive automation, this process is significantly more cost-effective and can therefore be used more frequently than conventional bridge inspections.
The two companies want to further refine the accuracy of the technology. In addition, the efficiency of drone operation is to be increased. The system should then be ready for commercial use from 2025. The companies currently assume that the technology can also be applied to other infrastructure, such as steel towers or crash barriers on roads.
(olb)
