The. physical phenomena leading to disruptions are very. complex and non linear and the present state of knowledge is not sufficient to explain the intrinsic structure of the data of interest. One viable way to extract information from the complex multidimensional operational space of a tokamak is to assume that the data which describe this space lie on an embedded, possibly nonlinear, low-dimensional subspace (manifold) within the higher dimensional space.To this purpose, recently, data visualization and. dimensionality reduction methods have been actively investigated. Among nonlinear methods the most popular are the. Self Organizing Map (SOM) and its probabilistic variant, the. Generative. Topographic. Mapping(GTM. ). The SOM has been already employed as disruption predictor at ASDEX Upgrade with good results. In this study, a 2D GTM has been built to represent the 7D ASDEX Upgrade operational space described by means of. a database of disrupted and nondisrupted discharges selected in the shot range 21654- 26891 and performed in ASDEX Upgrade between May 2007 and April 2011. The GTM clearly highlights the presence of a large region with an associated low risk of. disruption and. some. small regions (located in the map margins) with an associated high risk of disruption. The GTM proves to be able to separate nondisruptive. and disruptive states of plasma. Therefore, likewise the SOM ,t.he. GTM can be. used as a disruption predictor. by. track. ing. the temporal sequence of the samples on the map, depicting the. movement of the operating point during a discharge. Following the trajectory in the GTM. , it will be possible to eventually recognize the proximity to an operational. region where the risk of a. n imminent disruption is high.. In this paper, v. arious. criteria have been studied to. associate the risk of disruption of each map region with a disruption alarm threshold. The prediction performance of the proposed predictive system has been evaluated on a test set of discharges coming from experimental campaigns carried out at ASDEX Upgrade from May 2011 to November 2012. The achieved results are encouraging and indicate the appropriateness of the method, also comparing it with those obtained using SOM. Moreover,it is worth emphasizing that, compared to other disruption prediction approaches the GTM map. provides significant additional value. Whereas the tools in the reference papers are black boxes, which provide a prediction but are very difficult to interpret, on the contrary, the map allows to follow the trajectory of the plasma and to study its behavior leading to a d. isruption. So the developed map has the potential to provide much more than a simple prediction in the understanding of the operational space and the causes of the disruptions.

Data visualization and dimensionality reduction methods for disruption prediction at ASDEX Upgrade

SIAS, GIULIANA;ALEDDA, RAFFAELE;CANNAS, BARBARA;FANNI, ALESSANDRA;PAU, ALESSANDRO;
2013-01-01

Abstract

The. physical phenomena leading to disruptions are very. complex and non linear and the present state of knowledge is not sufficient to explain the intrinsic structure of the data of interest. One viable way to extract information from the complex multidimensional operational space of a tokamak is to assume that the data which describe this space lie on an embedded, possibly nonlinear, low-dimensional subspace (manifold) within the higher dimensional space.To this purpose, recently, data visualization and. dimensionality reduction methods have been actively investigated. Among nonlinear methods the most popular are the. Self Organizing Map (SOM) and its probabilistic variant, the. Generative. Topographic. Mapping(GTM. ). The SOM has been already employed as disruption predictor at ASDEX Upgrade with good results. In this study, a 2D GTM has been built to represent the 7D ASDEX Upgrade operational space described by means of. a database of disrupted and nondisrupted discharges selected in the shot range 21654- 26891 and performed in ASDEX Upgrade between May 2007 and April 2011. The GTM clearly highlights the presence of a large region with an associated low risk of. disruption and. some. small regions (located in the map margins) with an associated high risk of disruption. The GTM proves to be able to separate nondisruptive. and disruptive states of plasma. Therefore, likewise the SOM ,t.he. GTM can be. used as a disruption predictor. by. track. ing. the temporal sequence of the samples on the map, depicting the. movement of the operating point during a discharge. Following the trajectory in the GTM. , it will be possible to eventually recognize the proximity to an operational. region where the risk of a. n imminent disruption is high.. In this paper, v. arious. criteria have been studied to. associate the risk of disruption of each map region with a disruption alarm threshold. The prediction performance of the proposed predictive system has been evaluated on a test set of discharges coming from experimental campaigns carried out at ASDEX Upgrade from May 2011 to November 2012. The achieved results are encouraging and indicate the appropriateness of the method, also comparing it with those obtained using SOM. Moreover,it is worth emphasizing that, compared to other disruption prediction approaches the GTM map. provides significant additional value. Whereas the tools in the reference papers are black boxes, which provide a prediction but are very difficult to interpret, on the contrary, the map allows to follow the trajectory of the plasma and to study its behavior leading to a d. isruption. So the developed map has the potential to provide much more than a simple prediction in the understanding of the operational space and the causes of the disruptions.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11584/108346
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact