Artificial Neural Networks and Multi-Class Support Vector Machines for Classifying Magnetic Measurements in Tokamak Reactors

This paper is mainly concerned with the application ofa novel technique of data interpretation for classifying measurementsof plasma columns in Tokamak reactors for nuclear fusionapplications. The proposed method exploits several concepts derivedfrom soft computing theory. In particular, Artificial Neural Networksand Multi-Class Support Vector Machines have been exploited toclassify magnetic variables useful to determine shape and position ofthe plasma with a reduced computational complexity. The proposedtechnique is used to analyze simulated databases of plasma equilibriabased on ITER geometry configuration. As well as demonstrating thesuccessful recovery of scalar equilibrium parameters, we show thatthe technique can yield practical advantages compared with earliermethods.