Bremen 2017 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
P: Fachverband Plasmaphysik
P 28: Helmholtz Graduate School III
P 28.6: Vortrag
Donnerstag, 16. März 2017, 16:05–16:30, HS 1010
Acceleration of Bayesian Model Based Data Analysis for W7-X Density and Temperature Profiles — •Humberto Trimino Mora1, Jakob Svensson1, Andreas Werner1, Oliver Ford1, Sergey Bozhenkov1, Dirk Timmermann2, Robert Wolf1, and W7-X Team1 — 1MPI für Plasmaphysik — 2Rostock University
Density estimation for plasma analysis and control is a crucial element in magnetic confinement devices. Most of these have redundant density diagnostics to compare or calibrate; meaning that data from two different diagnostics measuring the same plasma parameter are available. Although the data is typically analyzed separately, a good solution for data fusion of two or more diagnostics is Bayesian data analysis. This allows estimation of specific parameters and their uncertainties for non-linear inverse problems in a strictly mathematical way.
The computation time and power required for the aforementioned problem is usually long, making it a post-processing technique that cannot be used in real time with rare exceptions.
This contribution proposes a design to accelerate data fusion of the W7-X Dispersion Interferometer’s line integrated electron density with the Thomson Scattering’s electron density and temperature along congruent lines of sight. This in order to provide in real time, a temperature profile and a more reliable density profile. The proposed design is implemented with reconfigurable hardware taking advantage of application specific circuits and parallelism to improve its processing time. An acceleration of Bayesian analysis for inverse problems is often necessary and would prove generally valuable for scientific inference.