Development of a neoclassical transport database by neural network fitting in LHD

WAKASA Arimitsu, MURAKAMI Sadayoshi1), YAMADA Hiroshi2),
YOKOYAMA Masayuki2), MAASSBERG Hening3),
BEIDLER Craig D.3), WATANABE Kiyomasa2), OIKAWA Shun-ichi
and LHD Experimental Group2)

Graduate School of Engineering, Hokkaido University, Sapporo 060-8628, Japan
1)Department of Nuclear Engineering, Kyoto University, Kyoto 606-8501, Japan
2)National Institute for Fusion Science, Toki, Gifu, 509-5292, Japan
3)Max-Planck-Institute für Plasmaphysik, EURATOM Ass., Greifswald, Germany

A database for the neoclassical transport coefficients of the Large Helical Device (LHD) is developed based on the neural network fitting methods. The normalized mono-energetic diffusion coefficients are evaluated by the Monte Carlo simulation code; DCOM (Diffusion Coefficient calculator by the Monte Carlo method). We evaluate the diffusion coefficients for various radial electric fields, the collision frequencies and the radial positions. The neural network fitting method is applied to take convolutions for the given distribution function, e.g. Maxwellian. The neural network fitting developed in the present work considerably reduces the number of points necessary for the smooth interpolation of the diffusion coefficient. The developed database is benchmarked with the results using other numerical models and is applied to the analysis of the LHD experimental results.