A Tokyo-based engineering firm is applying artificial intelligence (AI) to erosion and sinkhole detection throughout Japan.

Kawasaki Geological Engineering (KGE) trained Fujitsu’s Zinrai AI on an underground dataset of abnormalities to detect erosion-causing subsidence and developing sinkholes. During testing, the AI system identified underground cavities with an 82% rate of accuracy, according to the company.

The AI was applied to data generated by KGE’s geophysical ground-penetrating radar, which is capable of penetrating 2 to 3 m below the surface of the road, locating subsidence cavities and groundwater leakage stemming from the country’s aging infrastructure before it degrades into dangerous underground cavities and subsequent structural collapse. Prior to the AI, collected data was printed out on sheets of A3 paper and analyzed by a team of human engineers. Because one stretch of 100 m road generates one sheet of A3 paper, resulting in pages and pages of data, KGE turned to the Zinrai AI platform to expedite the process of analyzing the data.

KGE engineer Toshimune Imai said: “If you look closely, you can see the responses that show there are cavities underground. The slight bump depicts the fact that there is a cavity. But the shape is not limited to just that, and no matter how talented or well-trained they may be, the human engineers do get tired and there are times where they might overlook a cavity."

Although the rate of accuracy — 82% for the AI versus 80% for the team of human engineers — seems negligible, KGE insists that the AI has substantially reduced the time it takes to detect the presence of forming erosion and deterioration. KGE, however, is not planning to get rid of its team of engineers in favor of AI any time soon, suggesting that the technology only enhances human competency.

“Specialist engineers will always be needed,” said KGE CEO Toshihiko Sakagami. “Developing AI and developing engineers are two sides of the same coin for our business.”