A D V A N C E D M A T E R I A L S & P R O C E S S E S | A P R I L 2 0 2 3 1 0 MACHINE LEARNING | AI MACHINE LEARNING AUTOMATES X-RAY DIFFRACTION ANALYSIS Researchers at the National Institute forMaterials Science (NIMS), Japan, have automated the process of ana- lyzing results from x-ray diffraction studies on crystalline materials. The team developed a robotic process automation (RPA) system that uses machine learning to perform Rietveld analysis automatically, reducing human costs and speeding up data analysis. The new RPA system is run on a personal computer and can be combined with numerous graphical user interface applications used to calculate a material’s properties, control experimental equipment, or analyze material data, creating a closed-loop system to design and analyze materials with minimal human intervention. The scientists verified the accuracy of their procedure by analyzing samples of powdered compounds with known crystal structures. They explain that the ability to determine the structures from powdered samples is a strength of Rietveld analysis, avoiding the need to grow large single crystals, which can be difficult for some materials. The team now plans to refine the procedure for more complex crystal structures and explore the use of machine learning RPA strategies for general applications in materials science. Some of the possibilities include various simulation methods used to calculate material properties, as well as applications for controlling experimental equipment. www.nims.go.jp. AI TRANSFORMS NUCLEAR MATERIALS RESEARCH At the DOE’s Argonne National Laboratory, Lemont, Ill., scientist WeiYing Chen in the nuclear materials group is using artificial intelligence (AI) to revolutionize the way scientists analyze videos of their experiments. Chen is using a deep learning-based multi-object tracking (MOT) algorithm to extract data from videos of experiments in an effort to help the U.S. improve advanced nuclear reactor designs. The experiments are being conducted at the Intermediate Voltage Electron Microscope (IVEM) facility, a transmission electron microscope with ion beam accelerator capabilities. Re- sults from the MOT algorithm are helping scientists study the effect of ion irradiation on materials and defects within a picosecond timeframe, Graphical abstract. Courtesy of Sci. Technol. Adv. Mat., 2022. Wei-Ying Chen uses computer vision to collect data about material defects and structural voids in the same way facial recognition software looks for unique faces. which is too fast for manual tracking. With computer vision tools, Chen can extract data from every video frame, which will enable scientists to discover materials that are resistant to higher temperatures and irradiation doses. Traditional methods of capturing data during experiments can only provide a limited number of data points, but the MOT algorithm can track every defect and structural void. This is enabling Chen to build a reliable collection of information about material properties such as temperature resistance, irradiation resilience, microstructural defects, and material lifetimes. As a result, this enhanced data can lead to better models and experiments, allowing scientists to design advanced reactors that can run at higher tem- peratures and generate more clean electricity. Using AI and computer vision also provides a more efficient use of time, as scientists can make on-the-spot adjustments to their experiments and capture important information. “Computer vision can provide information that, from a practical standpoint, was unavailable before,” says Chen. “It’s exciting that we now haveaccess to somuchmore rawdataof unprecedented statistical significance and consistency.” anl.gov.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==