Date Thesis Awarded
Honors Thesis -- Access Restricted On-Campus Only
Bachelors of Science (BS)
In this study we investigate how the plasma density changes over time during the transition from Low confinement mode (L-Mode) to High confinement mode(H-mode). During this transition (L-H transition) the plasma self-organizes resulting in a localized density rise and the formation of steep density gradients near the edge of the plasma. We recreate the density rise of L-H transitions using a fluid model that separates radial transport into diffusive and convective terms. During this transition, radial transport perpendicular to magnetic flux surfaces is responsible for the density rise. The model uses numerical differential equation solvers and multi-dimensional optimization techniques to determine transport coefficients to match simulated density profiles to experimental measurements from the DIII-D tokamak. The model predicts key features of the L-H transition including the density rise, pedestal structure, and higher core densities compared with edge densities. This study investigates the $~2.5$ times faster density rise in hydrogen plasmas compared to deuterium plasmas. This study shows that diffusion driven electron transport in deuterium plasma accounts for the slower density rise compared to the stronger convection present in hydrogen plasmas. In addition to analysis based on optimization techniques to determine transport coefficients, we built and trained a Convoluted Neural Network (CNN) to find these transport parameters with experimental electron density profiles. Using synthetic density profiles generated from the model we trained the CNN to a final accuracy of $89\%$. The CNN’s predictions did over predict the optimization’s results, but still captured the key characteristics of the L-H transition.
Chiriboga, Javier, "Investigating Electron Transport in Tokamaks Using Computer Simulation and Machine Learning" (2023). Undergraduate Honors Theses. William & Mary. Paper 1974.
On-Campus Access Only