ADVANCED MATERIALS & PROCESSES | NOVEMBER/DECEMBER 2024 10 MACHINE LEARNING | AI The Argonne Leadership Computing Facility (ALCF) at the DOE’s Argonne National Laboratory is offering researchers access to a collection of AI machines called the ALCF AI Testbed where AI accelerators handle a variety of tasks including AI model training, inference, large language models, and more. anl.gov. BRIEF AI-DRIVEN SYSTEM SPEEDS MANUFACTURING Researchers at the University of Virginia (UVA), Charlottesville, report a significant advancement in manufacturing technology by developing an AI-driven system that could transform how factories run. Using multi-agent reinforcement learning, the team created a more efficient way to optimize manufacturing systems, improving speed and quality while reducing waste. Their approach integrates AI agents that work together to optimize production processes. By coordinating multiple agents to manage tasks in real time, the system adjusts automatically, improving performance over time. The team’s algorithms, credit- assigned multi-agent actor-attention-critic and physics-guided multi-agent actor-attention-critic, were key in making this advancement. These algorithms allow the system to account for both the physical constraints of machinery and unpredictable production disruptions. Their work has shown notable improvements in both productivity and system robustness. The research was conducted in collaboration with General Motors, a key industry partner that provided valuable insights and real-world applications for the AI system. GM’s involvement helped ensure that the technology addresses the practical challenges of modern manufacturing. “Our collaboration with UVA allowed us to explore innovative solutions that could transform production efficiency across the automotive industry,” says GM researcher Hua-Tzu Fan. virginia.edu. MACHINE LEARNING FOR POLYCRYSTALLINE MATERIALS Researchers at the University of California, Irvine and other international institutions achieved atomic-scale observations of grain rotation in polycrystalline materials, reportedly for the first time. Using state-of-the-art microscopy tools at the UC Irvine Materials Research Institute, scientists heated samples of platinum nano- crystalline thin films and observed the mechanism driving grain rotation in exceptional detail. The study used advanced techni- ques such as 4D Researchers at UVA and GM are working together on AI-driven systems for smarter, faster, and more adaptable automotive manufacturing. scanning transmission electron microscopy (STEM) and high-angle annular dark-field STEM. To address the challenge of interpreting the large 4D-STEM datasets, the authors developed a machine learning-based algorithm to extract critical information. These powerful imaging and analysis tools provided direct, real-time views of the atomic processes involved, specifically highlighting the role of disconnections at grain boundaries. The team discovered that grain rotation in these substances occurs through the propagation of disconnections, line defects with both step and dislocation characteristics, along the grain boundaries. This insight significantly advances understanding of the microstructural evolution in nanocrystalline materials. With the machine learning-assisted data analysis, the study also revealed for the first time a statistical correlation between grain rotation and grain growth or shrinkage. This relationship arises from shear- coupled grain boundary migration driven by disconnection motion, as confirmed by STEM observations and supported by atomistic simulations. Researchers say this finding is pivotal as it not only illuminates the fundamental mechanisms of grain rotation but also offers insights into the dynamics of nanocrystalline materials. uci.edu. Measurement of residual strain at the Σ11 GB before and after GB migration. Courtesy of Science, 2024, doi.org/10.1126/science.adk6384.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==