Learning pedestal dynamics via training reduced models against experimental tokamak plasmas

Abhilash Mathews

MIT

Tuesday, November 12, 2019

5:00pm

NW17-218

PSFC Student Seminars

The outer edge region of high confinement tokamak plasmas, known as the pedestal, is associated with the formation of transport barriers. This strongly influences energy and particle confinement, and in turn the energy gain of tokamaks which is crucial for upcoming devices (e.g. SPARC, ITER), yet a fully predictive model of pedestal structure is currently lacking. Pedestal pressure is constrained by magnetohydrodynamic limits due to edge localized modes (ELM), but a general model of pedestal density and temperature in ELM-suppressed regimes is absent. Therefore, this work explores potential methods for evaluating reduced plasma transport models across the pedestal against experiment. Towards this goal, an adaptive Gaussian process regression routine for automating time-dependent​ analysis of the pedestal has been developed and will be outlined. This tool can assist with interpretive modelling, improving inputs for simulations sensitive to gradients, validation efforts, and generating training data for supervised machine learning. ​