Three Los Alamos National Laboratory Projects Use Artificial Intelligence And Machine Learning To Propel Science


LANL News:

  • Multiyear efforts include partnerships with national laboratories and universities

Projects that seek to deploy artificial intelligence and machine learning for science and energy have been bolstered by Department of Energy Office of Science funding in recent weeks. Three Los Alamos National Laboratory (LANL) projects use the cutting-edge tools to improve the performance of accelerators, devices that charge particles to high speeds for research and development on a variety of subjects. Another Los Alamos project uses deep learning — a method of artificial intelligence — to tackle fusion reactor design challenges.

“The promise of artificial intelligence, in its various manifestations, is that we can create systems capable of learning and making decisions on their own,” said Mark Chadwick, acting deputy Laboratory director for Science, Technology and Engineering at Los Alamos. “The projects supported by the Office of Science here showcase the various ways in which Los Alamos researchers are harnessing the transformative power of artificial intelligence to develop solutions that drive science forward.”

Research and development accelerator efforts

The Office of Science’s Accelerator Stewardship and Accelerator Development program recently announced $16 million in funding for “accelerator technology that advances science, healthcare, the economy, and our security, and supports public-private partnerships to strengthen domestic suppliers of accelerator technology,” according to the DOE. Los Alamos National Laboratory is one of 35 institutions to play a role in the funded projects.

Alexander Scheinker, research and development engineer at Los Alamos, is leading a project, “Advanced Adaptive Control Systems for Compact Accelerators,” which features a novel approach that will help enhance performance and improve efficiency for compact accelerators. Compact accelerators and their beams are complex systems that require manual setup and continuous iterative tuning, or adjustments, by beam physics experts. Funded with $900,000 over three years, Scheinker’s collaboration with Lawrence Berkeley National Laboratory will further develop an approach that couples adaptive feedback control algorithms, deep convolutional neural networks and physics-based models in one large feedback loop to make better, noninvasive predictions that enable autonomous control of compact accelerators.

Nuclear Physics accelerator projects

Under the Office of Science Nuclear Physics program’s $16 million Artificial Intelligence and Machine Learning for Autonomous Optimization and Control of Accelerators and Detectors awards, Los Alamos researchers are leading or participating in two of 15 funded projects. The awards support the use of artificial intelligence to address “a variety of technical challenges in simulations, control, data acquisition and analysis” that are faced by instrumentation, such as accelerators.

Ming Xiong Liu, physicist at Los Alamos, leads a two-year, $1.6 million project “Intelligent Experiments Through Real-time AI: Fast Data Processing and Autonomous Detector Control for sPHENIX and Future EIC detectors.” At the sPHENIX experiment at the Relativistic Heavy Ion Collider at Brookhaven National Laboratory, Liu will lead a collaborative team to pursue pioneering, artificial intelligence- and machine-learning-driven experimental methods to detect and record extremely rare occurrences of heavy quark production in high-energy heavy ion collisions, offering insight into quark-gluon plasma and shedding light on nuclear matter under extreme conditions and on the evolution of the early universe. The team will also develop an advanced AI-based readout and control system for the ePIC detector at the future Electron Ion Collider (EIC) experiment at Brookhaven. The project is in collaboration with the Massachusetts Institute of Technology, Fermi National Accelerator Laboratory, the New Jersey Institute of Technology, Oak Ridge National Laboratory and the Georgia Institute of Technology.

Scheinker will also spearhead Los Alamos’ participation in a project called “Online Autonomous Tuning of the FRIB Accelerator Using Machine Learning.” The Michigan State University-led project will use machine learning to better tune the beamline at the Facility for Rare Isotope Beams (FRIB), at MSU, which produces exotic isotopes rarely seen on Earth. In addition to healthcare and industrial applications, rare isotopes may help reveal the nature of atoms and the elements in the universe.

Fusion Energy Sciences supports AI for fusion energy

The Office of Science Fusion Energy Sciences program has also sponsored a $29 million round of funding for “Research on Machine Learning, Artificial Intelligence, and Data Resources for Fusion Energy Sciences.” The funding supports seven projects with a total of 19 collaborating institutions that “will apply advanced and autonomous algorithms to address high-priority research opportunities in fusion and plasma sciences.”

At Los Alamos, physicist Xianzhu Tang leads a three-year, $3.3 million collaboration called “DeepFusion Accelerator for Fusion Energy Sciences in Disruption Mitigation.” The project aims to develop and apply fundamental deep learning and artificial intelligence methodologies that enable predictive design and control in fusion energy systems. The research focuses immediately on tokamak disruption mitigation, but the resulting methodologies and tools would be broadly applicable to fusion science and engineering.

The University of Texas, the University of Florida and Pennsylvania State University are partners on the project.