Bayesian optimization of disruption scenarios with fluid-kinetic models

Paper in proceeding, 2023

Tokamak disruptions can damage the machine due to localized heat
loads, mechanical stresses and impact of energetic runaway
electron beams. We use a Bayesian optimization framework to
optimize massive material injection of deuterium and neon in an
ITER-like tokamak set up. The optimization is performed using both
fluid and kinetic plasma models. The fluid model allows the
exploration of a large parameter space. Once promising parameter
regions are located, these are studied in higher physics fidelity using
kinetic simulations. The kinetic model predicts more optimistic results
regarding the success of the disruption mitigation.