Abstract - IPAM

Abstract

Machine Learning of Local Confinement and Transport from the International Multi-Tokamak Confinement Profile Database

Hong Qin

Princeton University

With the accelerated push toward commercial magnetic DT fusion, tritium self-sufficiency has emerged as a critical challenge. One of the key quantities governing tritium self-sufficiency is the tritium burn fraction, which depends critically on particle confinement time in magnetically confined plasmas. Theoretically, particle confinement is controlled by transport coefficients that are functions of plasma and device parameters local to each flux surface. Much of the historical effort to characterize these coefficients has relied on empirical scaling laws, typically developed for individual devices.

We present a Gaussian process regression study of local confinement and transport coefficients based on the 2008 public release of the International Multi-Tokamak Confinement Profile Database, which includes 344 discharges from TFTR, JET, DIII-D, Tore Supra, T-10, ITER simulations, JT-60U, ASDEX Upgrade, RTP, MAST, Alcator C-Mod, FTU, and TEXTOR. Preliminary results will be discussed, together with their implications for cross-device confinement modeling.
No video available
Back to Workshop III: Fusion Device Design and Engineering