A tiny crack in a bridge, dam or building can result in catastrophic consequences, however, a system being developed that uses artificial intelligence to detect these anomalies may soon be able to avert disasters.
The potential for unidentified, under-identified or misidentified structural damage is enormous throughout the built world. Until now, inspection of structures and trawling through videos captured by automatic crack detection systems has been time-consuming and reliant on humans not to make errors.
In the US, Purdue University's College of Engineering is working on an AI system that detects cracks captured in videos and analyses them frame by frame. And the algorithm used is capable of tracking cracks from one frame to the next.
Assistant Professor Mohammad R Jahanshahi believes it is a "giant leap for inspection technology", and is set to lead to reductions in accidents, deaths and maintenance costs.
"It lets the computer do the hard work," he said. "[It] provides a human operator with quantitative information about the crack, such as the thickness and the length of the crack."
The operator is then able to review the video and go to the specific frames referenced by the software to examine the cracks and determine what action to take. In its research, the engineering team has tested the software on 20 nuclear power plan inspection videos. The results showed the method was more robust than any other approach, with the system offering many potential applications such as detecting cracks on large buildings, roads and wind turbines.
Professor Jahanshahi said the system is smart and adaptive, can be reprogrammed to detect cracks within various structures and different materials, and the operator can use their own data. He believes the system will become even more useful as robots and drones are deployed to collect large amounts of visual data.