Creating stronger and tougher structural metals such as steel requires searching through a huge number of processing regimes. Even with the same alloy composition, heat treatment strongly changes the microstructure - the internal structure of the material - and its mechanical properties. Automation would help find optimal conditions more efficiently, but experiments with bulk metal specimens are difficult to automate: solid samples must be moved and large specialized equipment must be integrated.
A research team led by Yulin Xie, Sayaka Sekida and Goro Miyamoto of the Institute for Materials Research at Tohoku University, together with Tadashi Furuhara (formerly at the same institute, now at the National Institute for Materials Science, NIMS) and the company TsubameLab, developed an automated experimental system. It combines heat treatment, structural analysis and mechanical testing of bulk metals. The work was published in the journal Science and Technology of Advanced Materials: Methods.
Automating experiments with bulk structural metals is more difficult than with liquids, powders or thin films: a solid specimen must be reliably transported, positioned, heated, examined and mechanically tested. The scientists overcame these difficulties by creating special tools for handling specimens and integrating separate instruments under unified computer control.
The system includes an induction furnace, an X-ray diffractometer (XRD - an instrument that determines a material's structure from X-ray scattering) and a machine for micro-impact testing. A robotic arm automatically transfers specimens between stages. On the new platform, the team conducted automated experiments on single- and two-step tempering (a type of heat treatment) of industrial low-alloy steel JIS SCM435H. Excluding the heating time itself, the full cycle - specimen transfer, XRD measurement, mechanical testing and return of the specimen - takes about 800 seconds per specimen.
The system evaluates mechanical properties with accuracy comparable to conventional manual measurements, but performs the entire sequence more than 10 times faster. In addition, it simultaneously collects data on the processing regime, structure and properties. Such datasets are necessary for applying artificial intelligence in materials development. The researchers also integrated an AI-based optimization tool into the NIMO (NIMS Orchestration System) and demonstrated automatic optimization of furnace control parameters, laying the technological foundation for future closed-loop materials search.
In the future, the autonomous system will analyze the results of each experiment and propose the next heat treatment conditions. The platform will automatically carry out the proposed regime, characterize the structure and properties, and return new data to the AI. Repeating such a cycle will make it possible to efficiently find promising regimes among a huge number of possibilities.
"This work shows that the entire experimental process can be automated even for complex bulk structural metallic materials," Miyamoto said. "By combining the platform with an AI that proposes the next experimental conditions, we aim to accelerate the autonomous discovery of innovative processing routes and structural materials."
The development is an important step toward autonomous laboratories for structural metals. Rapidly obtaining data on the "processing - structure - properties" relationship could significantly expand the range of heat treatment regimes studied and accelerate the creation of advanced metallic materials.