A data - based model for Thermal SHAx prediction in RFX mod

In the reversed field experiment RFX - mod, the operation at high plasma currents (Ip >= 1.2 MA) has evidenced the formation of large electron thermal structures, delimited by strong Internal Transport Barriers (ITBs), in the plasma core during the quasi - single helicity (QSH) regimes. These structures appear when a helical equilibrium with a single helical axis (SHAx) has been established in the plasma. Despite the strong the oretical and experimental effort some aspects of the ITBs physics are still unclear. One of the open questions is why the ITBs crash in spite of a longer magnetic QSH state. As the presence and the class of ITBs is now recognized with a post - shot investiga tion of the electron temperature profiles, the aim of the work is to realize a predictive system, based on a data - based approach, able to automatically recognize the nature of ITBs and the possible precursor of their crashing phase. The predictive strategy consist in three phases: in the first one a database is built to train the predictor relying on signals belonging to double filter multi - channel diagnostics; the second realizes a clustering of measurements belonging to the same class of ITB by means of l inear programming, while the third performs a projection of the clusters into a low - dimensional space, by means of a probabilistic neighborhood function, in order to recognize safe and crashes areas. Those areas are then used to test new sets of signals wh ile the low - dimensional space allows better results visualization. The procedure has been tested on a subset of shots, including hydrogen and deuterium campaign from 2011 to 2014, and in this paper the results are presented.