Soft Computing and Chaos Theory for Disruption Prediction in Tokamak Reactors

Plasma disruption in a Tokamak reactor is a sudden loss of magnetic
confinement that can cause damage to the machine walls and the
support structures. For this reason early detection of the onset of
such an event is of practical interest. This paper presents a novel
technique for early prediction of plasma disruption in Tokamak
reactors based on chaos theory and a comparison of neural networks;
neuro-fuzzy inference systems are also presented. In particular,
dynamical reconstruction and chaos theory have been considered
for choosing the time window of prediction and to select the set of
inputs for the prediction system. Multi-layer-perceptron nets and
Sugeno’s neuro-fuzzy inference have been exploited for predicting
the onset of disruption. Within the limits of the available database
(disruptive discharges in JET machine) it is possible to predict the
onset of the disruptive event sufficiently in advance to activate the
control system.