New signal processing methods and information technologies for the real time control of JET reactor relevant plasmas
Autor: | A. Murari, J. Vega, D. Mazon, G.A. Rattá, M.Gelfusa, A. Debrie, C. Boulbe, B. Faugeras, JET-EFDA Contributors |
---|---|
Rok vydání: | 2011 |
Předmět: |
Physics
Jet (fluid) Signal processing Moore's law Nonextensive entropy Genetic Algorithms SVM Mechanical Engineering media_common.quotation_subject Real-time computing Image processing Real time equilibrium reconstruction Support vector machine Range (mathematics) Nuclear Energy and Engineering Real-time Control System General Materials Science Civil and Structural Engineering media_common |
Zdroj: | Fusion engineering and design 86 (2011): 544–547. doi:10.1016/j.fusengdes.2010.12.061 info:cnr-pdr/source/autori:A. Murari; J. Vega; D. Mazon; G.A. Rattá; M.Gelfusa; A. Debrie; C. Boulbe; B. Faugeras; JET-EFDA Contributors/titolo:New signal processing methods and information technologies for the real time control of JET reactor relevant plasmas/doi:10.1016%2Fj.fusengdes.2010.12.061/rivista:Fusion engineering and design/anno:2011/pagina_da:544/pagina_a:547/intervallo_pagine:544–547/volume:86 |
ISSN: | 0920-3796 |
Popis: | A general trend in the experimental programmes of present day Tokamaks, and of JET in particular, is the constant increase in the number of parameters to be controlled in real time, to satisfy the machine protection requirements on the one hand and to improve performance on the other. Since the amount of data collected is also increasing at least at a rate compatible with the Moore law, significant developments are required in the field of real time algorithms particularly for magnetic reconstructions, disruption prediction and image processing. A new real time equilibrium code called EQUINOX, using internal and external measurements of the magnetic fields, has been qualified on JET. It can provide reconstructed accurate equilibria about every 50 ms on a 2 GHz PC. An advanced disruption predictor, based on machine learning tools, has been deployed using inputs selected with a genetic algorithm. Its success rate remains of the order of 94% for up to 170 ms before the occurrence of the disruption. Nonextensive entropies, which are more sensitive to long range correlations, seem to be useful in detecting vibrations in the videos of JET cameras, both visible and infrared. |
Databáze: | OpenAIRE |
Externí odkaz: |