München 2019 – wissenschaftliches Programm
P 18.103: Poster
Donnerstag, 21. März 2019, 16:30–18:30, Foyer Audimax
Artificial Neural Networks for Plasma Edge Analysis in Wendelstein 7-X — •Marko Blatzheim1,2, Daniel Böckenhoff1, Roger Labahn2, and Thomas Sunn Pedersen1 for the The Wendelstein 7-X Team collaboration — 1Max Planck Institute for Plasma Physics, Greifswald, Germany — 2Institute for Mathematics, University of Rostock, Rostock, Germany
Wendelstein 7-X (W-7X) is a stellarator type nuclear fusion experiment. The plasma facing components show heat load patterns detectable by infrared cameras due to the contact with hot plasma. Artificial neural networks can be trained with observations of heat load pattern based on Field Line Diffusion simulations to reconstruct plasma properties in real time. Different types of neural networks from feed-forward fully-connected and convolutional neural networks to deep residual inception networks can be assigned with that task. The advantages and disadvantages for each of these neural network architerctures are investigated with all results generally satisfactory.