Machine learning guided discovery and design for inertial confinement fusion

Kellil Humbird, Research Scientist, Lawrence Livermore National Lab

Inertial confinement fusion (ICF) experiments and their corresponding computer simulations produce an immense amount of rich data. However, quantitatively interpreting that data remains a grand challenge. Design spaces are vast, data volumes are large, and the relationship between models and experiments may be uncertain.

We propose using machine learning to aid in the design and understanding of ICF implosions by integrating simulation and experimental data into a common framework. We will present a novel deep learning algorithm, “deep jointly-informed neural networks” (DJINN) [1], which enables non-data scientists to quickly train neural network models on their own datasets. DJINN enables the creation of models which combine simulation and experimental data into a common, predictive framework. Specifically, we will demonstrate a novel method for model calibration with deep neural networks which produces models that are more predictive of ICF experiments at the Omega Laser Facility than simulations alone. We use these models to study the discrepancies between simulations and experiments, and to search for high performing experimental designs.
*Prepared by LLNL under Contract DE-AC52-07NA27344. LLNL-ABS- 763781.