Toyota Industries AI cuts fracture analysis time up to 90%

Toyota Industries traces its materials research back to a department founded in 1947, giving it roughly 80 years of accumulated know-how in analyzing metals such as iron and aluminum. The new technology draws directly on that history, using image data from the past 10 years of veteran technicians' analyses to train the AI. Fracture-surface analysis matters because it traces the cause of a break at the nanoscale, helping improve design conditions and material selection for metal products such as automobiles. That work had depended heavily on experienced specialists, and the company frames standardization and efficiency as pressing needs as electrification and automation speed up product development.

The system goes beyond judging images. When an observation sample is set on the electron microscope, it automatically carries out the main steps of the analysis, finding traces of the damage cause that may be only about 1 micrometer across within a fracture surface a few centimeters in diameter. Toyota Industries also built what it calls MLOps, a setup that continuously learns from new analysis data gathered during operation to improve the model's reliability. The company says that, according to its own research, this is the first technology in the world to automate the entire fracture analysis workflow.