Particle physics experiments

LLNL team wins top prize for ML-based approach to ICF experiments

August 5, 2022 – The IEEE Nuclear and Plasma Science Society (NPSS) has announced that a team from Lawrence Livermore National Laboratory (LLNL) has been awarded its 2022 Transactions on Plasma Science (TPS) Best Paper Award for its work applying machine learning to inertia. fusion by confinement (ICF) experiments.

In the paper, lead author Kelli Humbird and co-authors propose a novel technique for calibrating ICF experiments by combining machine learning with experimental data via “transfer learning” (TL), a method in which models are trained on a task and then partially recycled. to solve a separate but related task with limited data.

The figure plots the actual values ​​(x axis) against the predicted values ​​(y axis) for five measured quantities of interest in the Omega ICF experiments – neutron yield, bang time, burn width, ion temperature and the surface density.

The team, which includes Luc Peterson and Brian Spears from LLNL and Ryan McClarren from the University of Notre Dame, introduced the concept of hierarchical TL, where neural networks trained on low-fidelity models are calibrated on high-fidelity models. , then applied to experimental experiments. The data. The researchers applied the technique to a database of ICF simulations and experiments performed at the Omega Laser Facility at the University of Rochester, finding that combining deep neural networks with experiments resulted in models of ICF experiments better and more predictive than simulations alone.

Typically, ICF experiments are designed using computer simulations that approximate physical modeling and must be calibrated to predict experimental observations, usually through a statistics-based Bayesian method.

“In this paper, we were really able to pioneer the use of transfer learning as a means of calibrating scientific simulations,” Humbird said. “Transfer learning is a fundamentally different approach to calibrating a simulation-based model to experimental data, and it works well in some cases where the Bayesian approach struggles.”

In addition to the award, Humbird and his team received a plaque, certificates and a cash prize of $1,500 divided equally among team members. The work is part of a larger effort by the Department of Energy to discover, design, manufacture and deploy artificial intelligence and machine learning at every stage of a project. With a more reliable, AI-driven simulator, researchers could predict future experiments at the National Ignition Facility and find new optimal implosion designs, Humbird explained.

“Simulators are good maps, but transfer learning helps you take that good map and identify the actual experimentally observed features by taking all of the experimental data we’ve collected in our ICF implosions and incorporating it into the model,” Humbird said. “The transfer learning model also gets smarter over time, constantly updating its idea of ​​what the ‘NIF map’ looks like as we pull more and more shots at the facility. So when it comes to design optimization, using this data-driven model to research designs holds promise, as we believe the map we use to chart our path to high performance is precise.

In addition to fusion research, transfer learning could have impacts beyond ICF, including any experiment that uses approximate simulations to model complex systems and requires the generation of a predictive model with only a small amount of experimental data.

The annual award is the fourth of its kind given by the IEEE TPS journal, which covers the theory and application of plasma science and engineering and disseminates technical information for the Plasma Science and Applications Committee of the IEEE NPSS, the Pulsed Power Science and Technology Committee, and the Fusion Technology Committee.

The journal determined the winner based on criteria such as: quantifiable usefulness to the community, number of downloads, article citations, quality, clarity of presentation, originality, importance and contributions to the field.

The award will be presented at the Plasma Science Conference in Santa Fe, New Mexico in May 2023.


Source: LLNL