Prediction of major disruptions in tokamak plasmas, analyses of time series data

Yoshino, Ryuji

This paper reviews recent research activities on the prediction of major disruptions observed in tokamak plasmas using neural networks. Disruptions caused by the density limit, the impurity injection, and the external error magnetic field, are driven by the plasma current and can be predicted from their precursors with a prediction success rate of nearly 100%. Disruptions driven by the increase in the plasma pressure has been predicted with a prediction success rate of about 90% evaluating the closeness to the operational limit by neural networks.