AI Selects Ultrafine Defects Like a 'Jjokjipge' Expert

Kang Seonghun, Principal Researcher at Korea Institute of Materials Science,

[Asia Economy Reporter Kim Bong-su] A domestic research team has developed a technology that automatically detects ultra-fine defects on surfaces with complex shapes using artificial intelligence (AI).

The Korea Institute of Materials Science (KIMS) announced on the 19th that Dr. Kang Seong-hoon’s research team from the Materials AI and Big Data Laboratory developed a technology that can accurately detect ultra-fine defects on product surfaces that are difficult for humans to distinguish using AI.

Micro defects on products appear in various forms, and visually inspecting and checking them requires a lot of time and cost. Especially, surfaces of parts with complex hierarchical structures have three-dimensional shapes, making it difficult to accurately identify defects using only microscopes or the naked eye. In particular, ultra-fine defects, which are about half the thickness of a human hair (50㎛), cannot be accurately read by the naked eye, so workers had to go through a process of taking dozens of photos per specimen using a microscope and checking each one individually.

The research team created an AI that automatically recognizes the hierarchical structure of the product surface through images that simplify the three-dimensional shape into two dimensions and detects micro defects that deviate from the reference image. The AI automatically learns the characteristics of defect factors and then performs defect detection and classification. It is possible to accurately detect defect factors in about one second. In addition to defect detection, the team applied a Class Activation Map (CAM) neural network so that users can understand which surface features the AI used as criteria to select defects, thereby increasing the reliability of ultra-fine defect detection.

Dr. Kang Seong-hoon, the lead researcher, said, “Recently, process automation and smart factory construction have become social issues,” adding, “If the developed technology is applied to various automated process lines, it will be possible to detect micro defects that are difficult for humans to interpret in a short time.”

This content was produced with the assistance of AI translation services.