Abstract:Prediction and likelihood identification of high-beta disruption in JT-60U has been discussed by means of feature extraction based on sparse modeling. In disruption prediction studies using machine learning, the selection of input parameters is an essential issue. A disruption predictor has been developed by using a linear support vector machine with input parameters selected through an exhaustive search, which is one idea of sparse modeling. The investigated dataset includes not only global plasma parameters … Show more
“…So far, all the main families of existing real-time compatible classifier technologies have been explored, including support vector machines, artificial neural networks, generative topographic mapping, fuzzy logic and deep learning. They have also been applied to the data of many tokamaks of various generations: ADITYA (India) 47 , ASDEX Upgrade (Germany) 48 , DIII-D (US) 49 – 51 , J-TEXT (China) 52 , NSTX (US) 53 , ALCATOR C-MOD (US) 54 , JT-60U (Japan) 55 , EAST (China) 56 – 58 , HL-2A (China) 59 and JET (UK) 7 . Up to now only three machine-learning-based predictors have been implemented in JET’s real-time network, APODIS 60 , SPAD 61 , and the centroid-based method 21 .…”
The objective of thermonuclear fusion consists of producing electricity from the coalescence of light nuclei in high temperature plasmas. The most promising route to fusion envisages the confinement of such plasmas with magnetic fields, whose most studied configuration is the tokamak. Disruptions are catastrophic collapses affecting all tokamak devices and one of the main potential showstoppers on the route to a commercial reactor. In this work we report how, deploying innovative analysis methods on thousands of JET experiments covering the isotopic compositions from hydrogen to full tritium and including the major D-T campaign, the nature of the various forms of collapse is investigated in all phases of the discharges. An original approach to proximity detection has been developed, which allows determining both the probability of and the time interval remaining before an incoming disruption, with adaptive, from scratch, real time compatible techniques. The results indicate that physics based prediction and control tools can be developed, to deploy realistic strategies of disruption avoidance and prevention, meeting the requirements of the next generation of devices.
“…So far, all the main families of existing real-time compatible classifier technologies have been explored, including support vector machines, artificial neural networks, generative topographic mapping, fuzzy logic and deep learning. They have also been applied to the data of many tokamaks of various generations: ADITYA (India) 47 , ASDEX Upgrade (Germany) 48 , DIII-D (US) 49 – 51 , J-TEXT (China) 52 , NSTX (US) 53 , ALCATOR C-MOD (US) 54 , JT-60U (Japan) 55 , EAST (China) 56 – 58 , HL-2A (China) 59 and JET (UK) 7 . Up to now only three machine-learning-based predictors have been implemented in JET’s real-time network, APODIS 60 , SPAD 61 , and the centroid-based method 21 .…”
The objective of thermonuclear fusion consists of producing electricity from the coalescence of light nuclei in high temperature plasmas. The most promising route to fusion envisages the confinement of such plasmas with magnetic fields, whose most studied configuration is the tokamak. Disruptions are catastrophic collapses affecting all tokamak devices and one of the main potential showstoppers on the route to a commercial reactor. In this work we report how, deploying innovative analysis methods on thousands of JET experiments covering the isotopic compositions from hydrogen to full tritium and including the major D-T campaign, the nature of the various forms of collapse is investigated in all phases of the discharges. An original approach to proximity detection has been developed, which allows determining both the probability of and the time interval remaining before an incoming disruption, with adaptive, from scratch, real time compatible techniques. The results indicate that physics based prediction and control tools can be developed, to deploy realistic strategies of disruption avoidance and prevention, meeting the requirements of the next generation of devices.
“…So far, all the main families of existing real-time compatible classifier technologies have been explored, including support vector machines, artificial neural networks, generative topographic mapping, fuzzy logic and deep learning. They have also been applied to the data of many tokamaks of various generations: ADITYA (India) [47], ASDEX Upgrade (Germany) [48], DIII-D (US) [49][50][51], J-TEXT (China) [52], NSTX (US) [53], ALCATOR C-MOD (US) [54], JT-60U (Japan) [55], EAST (China) [56][57][58], HL-2A (China) [59] and JET (UK) [7]. Up to now only three machinelearning based predictors have been implemented in JET's real-time network, APODIS [60], SPAD [61], and the centroid based method [62].…”
Section: M1 Brief History Of Machine Learning Based Disruption Predic...mentioning
The objective of thermonuclear fusion consists of producing electricity from the coalescence of light nuclei in high temperature plasmas. The most promising route to fusion envisages the confinement of such plasmas with magnetic fields, whose most studied configuration is the tokamak. Disruptions are catastrophic collapses affecting all tokamak devices and one of the main potential showstoppers on the route to a commercial reactor. Deploying new analysis methods on thousands of JET discharges, covering the isotopic compositions from hydrogen to full tritium and including the last major D-T campaign, the nature of the various forms of collapse is investigated. A new approach to proximity detection has been developed, which allows determining both the probability of and the time interval remaining before an incoming disruption, with adaptive, from scratch, real time compatible techniques. The results indicate that physics based prediction and control tools can be developed, to deploy realistic strategies of disruption avoidance and prevention, meeting the requirements of the next generation of devices.
“…Approaches relying on a mixture of time/frequency domains, including wavelet decompositions, have also been pursued [17][18][19] . With regard to classifier technologies, real-time compatible predictors have typically been based on artificial neural networks, support vector machines, fuzzy logic, generative topographic mapping and deep learning and have been studied on a broad range of tokamaks, including ADITYA (India) 20 , ASDEX Upgrade (Germany) 21 , DIII-D (United States) [22][23][24] , J-TEXT (China) 25 , NSTX (United States) 26 , ALCATOR C-MOD (United States) 27 , JT-60U (Japan) 28 , EAST (China) [29][30][31] , HL-2A (China) 32 and JET (United Kingdom) [33][34][35] .…”
Section: And Jet Contributors*mentioning
confidence: 99%
“…The classifiers employed in the studies on tokamaks mentioned above [20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35] were developed using real-time valid solutions, which guarantee response times within a specified time window. The predictors discussed in the remainder of this work have been tested offline with real-time compatible technologies and using only real-time available signals.…”
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.