Hostname: page-component-848d4c4894-nmvwc Total loading time: 0 Render date: 2024-06-16T07:57:15.021Z Has data issue: false hasContentIssue false
  • English
  • Français

Neural networks: from image recognition to tokamak plasma tomography

Published online by Cambridge University Press:  17 April 2019

Axel Jardin Open the ORCID record for Axel Jardin [Opens in a new window] ,
Jakub Bielecki Open the ORCID record for Jakub Bielecki [Opens in a new window] ,
Didier Mazon ,
Jan Dankowski Open the ORCID record for Jan Dankowski [Opens in a new window] ,
Krzysztof Król ,
Yves Peysson  and
Marek Scholz Open the ORCID record for Marek Scholz [Opens in a new window]
Axel Jardin*
Affiliation:
Institute of Nuclear Physics Polish Academy of Sciences (IFJ PAN), PL-31-342, Krakow, Poland
Jakub Bielecki
Affiliation:
Institute of Nuclear Physics Polish Academy of Sciences (IFJ PAN), PL-31-342, Krakow, Poland
Didier Mazon
Affiliation:
CEA, IRFM F-13108 Saint Paul-lez-Durance, France
Jan Dankowski
Affiliation:
Institute of Nuclear Physics Polish Academy of Sciences (IFJ PAN), PL-31-342, Krakow, Poland
Krzysztof Król
Affiliation:
Institute of Nuclear Physics Polish Academy of Sciences (IFJ PAN), PL-31-342, Krakow, Poland
Yves Peysson
Affiliation:
CEA, IRFM F-13108 Saint Paul-lez-Durance, France
Marek Scholz
Affiliation:
Institute of Nuclear Physics Polish Academy of Sciences (IFJ PAN), PL-31-342, Krakow, Poland
*
Author for correspondence: Axel Jardin, Institute of Nuclear Physics Polish Academy of Sciences (IFJ PAN), PL-31-342, Krakow, Poland. E-mail: axel.jardin@ifj.edu.pl
Get access Rights & Permissions [Opens in a new window]

Abstract

In this paper, the possibility of using neural networks for fast tomographic reconstructions of tokamak plasma soft X-ray (SXR) emissivity is investigated. Indeed, the radiative cooling of heavy impurities like tungsten could be detrimental for the plasma core performances of ITER, thus developing robust and fast SXR diagnostic tools is a crucial issue to monitor the impurities and to mitigate in real-time their central accumulation. As preliminary work, a database of emissivity phantoms with associated synthetic measurements is used to train the neural network to solve the inversion problem. The inversion method, training process, and first tomographic reconstructions are presented with the perspectives about our future work.

Keywords

Neural networkssoft X raystokamaksimpuritiestomography
Type
Research Article
Information
Laser and Particle Beams , Volume 37 , Issue 2 , June 2019 , pp. 171 - 175
Copyright
Copyright © Cambridge University Press 2019 

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

Anton, M, Weisen, H, Dutch, MJ, Von Der Lindenz, W, Buhlmann, F, Chavan, R, Marletaz, B, Marmillod, P and Paris, P (1996) X-ray tomography on the TCV tokamak. Plasma Physics and Controlled Fusion 38, 1849.Google Scholar
Bielecki, J, Giacomelli, L, Kiptily, V, Scholz, M, Drozdowicz, K, Conroy, S, Craciunescu, T, Kempenaars, M and JET EDFA Contributors (2015) Phillips-Tikhonov regularization with a priori information for neutron emission tomographic reconstruction on Joint European Torus. Review of Scientific Instruments 86, 093505.Google Scholar
Cannas, B, Fanni, A, Pautasso, G, Sias, G and the ASDEX Upgrade Team (2011) Disruption prediction with adaptive neural networks for ASDEX upgrade. Fusion Engineering and Design 86, 1039–1044.Google Scholar
Casson, FJ, Angioni, C, Belli, EA, Bilato, R, Mantica, P, Odstrcil, T, Pütterich, T, Valisa, M, Garzotti, L, Giroud, C, Hobirk, J, Maggi, CF, Mlynar, J and Reinke, ML, JET EFDA contributors and ASDEX Upgrade team (2015) Theoretical description of heavy impurity transport and its application to the modelling of tungsten in JET and ASDEX upgrade. Plasma Physics and Controlled Fusion 57, 014031.Google Scholar
Cichy, RM, Khosla, A, Pantazis, D, Torralba, A and Oliva, A (2016) Comparison of deep neural networks to spatio-temporal cortical dynamics of human visual object recognition reveals hierarchical correspondence. Scientific Reports 6, 27755.Google Scholar
Citrin, J, Breton, S, Felici, F, Imbeaux, F, Aniel, T, Artaud, JF, Baiocchi, B, Bourdelle, C, Camenen, Y and Garcia, J (2015) Real-time capable first principle based modelling of tokamak turbulent transport. Nuclear Fusion 55, 092001.Google Scholar
Ferreira, DR, Carvalho, PJ and Fernandes, H (2018) Full-pulse tomographic reconstruction with deep neural networks. Fusion Science and Technology 74, 47–56.Google Scholar
Glorot, X and Bengio, Y (2010) Understanding the difficulty of training deep feedforward neural networks. PMLR 9, 249–256.Google Scholar
Hansen, PC (1992) Analysis of discrete ill-posed problems by means of the L-curve. SIAM Review 34, 561.Google Scholar
Imrisek, M, Mlynar, J, Loffelmann, V, Weinzettl, V, Odstrcil, T, Odstrcil, M and Tomes, M (2016) Optimization of soft X-ray tomography on the COMPASS tokamak. Nukleonika 61, 403–408.Google Scholar
Jardin, A (2017) Soft X-ray measurements for impurity transport studies in tokamak plasmas. PhD thesis. Aix-Marseille University.Google Scholar
Jardin, A, Mazon, D and Bielecki, J (2016 a) Comparison of two regularization methods for SXR tomography at Tore Supra. Physica Scripta 91, 044007.Google Scholar
Jardin, A, Mazon, D, O'Mullane, M, Mlynar, J, Loffelmann, V, Imrisek, M, Chernyshova, M, Czarski, T, Kasprowicz, G, Wojenski, A, Bourdelle, C and Malard, P (2016 b) Tomographic capabilities of the new GEM based SXR diagnostic of WEST. Journal of Instrumentation 11, C07006.Google Scholar
Lecun, Y, Bottou, L, Bengio, Y and Haffner, P (1998) Gradient-based learning applied to document recognition. Proceedings of the IEEE 86, 2278–2324.Google Scholar
Loffelmann, V, Mlynar, J, Imrisek, M, Mazon, D, Jardin, A, Weinzettl, V and Hron, M (2016) Minimum Fisher Regularization adapted to real-time tomography. Fusion Science and Technology 69, 505.Google Scholar
Matos, FA, Ferreira, DR and Carvalho, PJ (2017) Deep learning for plasma tomography using the bolometer system at JET. Fusion Engineering and Design 114, 18–25.Google Scholar
Mazon, D, Vezinet, D, Malard, P, Pacella, D, Gabelieri, L, Romano, A, Piergotti, V, Benedetto, T, Murtas, F, Dagabov, S and Corradi, G (2012) Soft x-ray tomography for real-time applications: present status at Tore Supra and possible future developments. Review of Scientific Instruments 83, 063505.Google Scholar
Mikszuta, K (2018) Comparison of the criterions of selection the regularization parameter, implemented to deconvolution methods application in plasma diagnostics. 14th Kudowa Summer School “Toward Fusion Energy”. Kudowa-Zdrój.Google Scholar
Mlynar, J, Weinzettl, V, Bonheure, G and Murari, A (2010) Inversion techniques in the soft-X-ray tomography of fusion plasmas: toward real-time applications. Fusion Science and Technology 58, 733–741.Google Scholar
Nielsen, MA (2015) Neural Networks and Deep Learning. Determination Press.Google Scholar
Odstrcil, T, Pütterich, T, Odstrcil, M, Gude, A, Igochine, V, Stroth, U and ASDEX Upgrade Team (2016) Optimized tomography methods for plasma emissivity reconstruction at the ASDEX upgrade tokamak. Review of Scientific Instruments 87, 123505.Google Scholar
Pütterich, T, Neu, R, Dux, R, Whiteford, AD, O'Mullane, MG, Summers, HP and the ASDEX Upgrade Team (2010) Calculation and experimental test of the cooling factor of tungsten. Nuclear Fusion 50, 025012.Google Scholar
Rumelhart, DE, Hinton, GE and Williams, RJ (1986) Learning representations by back-propagating errors. Nature 323, 533–536.Google Scholar