Researchers at Tohoku University and the National Institute for Materials Science have developed an automated robotic system that integrates metal heat treatment, X-ray diffraction, and mechanical testing, slashing testing times and laying the groundwork for artificial intelligence-driven materials discovery.
Developing stronger structural metals requires sorting through a massive spectrum of processing conditions. Even when working with an identical alloy composition, a change in heat treatment can drastically alter a material’s internal microstructure and physical strength. Pinpointing the ideal treatment parameters manually demands tedious laboratory work, especially for solid bulk metals that resist standard automation techniques designed for liquids, powders, or thin films.
Researchers Build Automated System for Metal Experiments
A collaborative research group tackled this bottleneck by building a unified, automated experimental platform. The team brought together researchers from the Institute for Materials Research at Tohoku University—including Yulin Xie, Sayaka Sekida, and Goro Miyamoto. They also included Tadashi Furuhara, formerly at the institute and now at the National Institute for Materials Science. They partnered with TsubameLab to design the specialized hardware.
The resulting apparatus combines an induction heating furnace, an X-ray diffractometer, and a micro-impact testing machine. A robotic arm manages the physical transfer of solid metal samples between each instrument. The entire sequence unfolds under centralized computer control without human intervention. The setup was detailed in a study published online in Science and Technology of Advanced Materials: Methods on September 18, 2026.
Team Tempers Steel Ten Times Faster With Robotic Platform
To put the new platform through its paces, the team tested it on one- and two-step tempering routines using a commercial JIS SCM435H low-alloy steel. Transporting specimens, capturing X-ray diffraction measurements, conducting impact tests, and returning the samples for recovery took approximately 800 seconds per sample, excluding the duration of the heat treatment itself.
The automated hardware matches the precision of traditional, manual measurements while running through the sequence more than 10 times faster. By generating synchronized datasets covering processing methods, internal structures, and mechanical properties all at once, the platform produces the exact kind of information required to train artificial intelligence models for metallurgy.
Platform Integrates Artificial Intelligence via NIMS System
Beyond speeding up physical tests, the researchers integrated an artificial intelligence optimization tool directly into the NIMS Orchestration System, known as NIMO. This step automated the adjustment of furnace control parameters, establishing a technical foundation for closed-loop materials exploration.
In future iterations, artificial intelligence will evaluate data from completed runs, select the next experimental conditions to explore, and direct the automated platform to execute the trial. This continuous feedback loop will allow systems to scour massive sets of processing variables autonomously.
This work demonstrates that the entire experimental workflow can be automated, even for tricky bulk structural metallic materials.
Goro Miyamoto, Institute for Materials Research, Tohoku University
By combining this platform with AI that proposes the next experimental conditions, we aim to accelerate the autonomous discovery of innovative processing routes and structural materials.
Goro Miyamoto, Institute for Materials Research, Tohoku University