General Atomics Energy Division hosts the DIII-D National Fusion Facility, the largest-operating magnetic confinement fusion device in the United States. The DIII-D and magnetic confinement GA Theory teams are co-located at General Atomics in San Diego, Calif.
Both the DIII-D and GA Theory teams make use of Department of Energy computational resources at Leadership Class Computing Facilities (LCCF) – Argonne, NERSC at Berkeley Lab, and Oak Ridge – to analyze new data and evaluate models of magnetic confinement fusion including equilibrium reconstruction, microscale turbulence, magnetohydrodynamic stability, fueling, heating, and current drive. The use of LCCF ranges from inter-discharge analyses (automatically triggered from DIII-D) to database assembly of data analyses for tens of thousands of discharges, to database generation for hundreds of thousands of simulation runs, to large high-fidelity simulations that consume tens of thousands of node hours. An emerging element of research is the construction of machine learning (ML) models based on databases of raw data, analyzed data, or simulations. The ML models can find hidden relationships in the data and have a much faster evaluation time than the original simulations to provide faster feedback for experimental or machine design.