Identification of noncircular plasma equilibria using a neural network approach

A technique using a single hidden layer backpropagation neural network is described to establish a nonlinear mapping between a set of magnetic flux measurements and some shaping parameters of a noncircular plasma. The technique has been applied for the identification of limiter and X point equilibria in the ASDEX Upgrade geometry; the dataset of equilibria required for training and testing the neural network has been generated by means of an integrated use of a fixed and a free boundary MHD code. The average accuracy of the identification procedure is quite good, with a further improvement if a linear connection between the input and output layers is introduced. A procedure is also proposed for the selection of the optimum location of a limited number of sensors. The relationship existing between the behaviour of the neural network and some statistical parameters of the dataset is analysed and discussed