Reconstruction of Plasma Shape Parameters in the T-15MD Tokamak Using Neural Networks

The work is aimed at solving the problem of reconstructing the shape parameters of the plasma column (configuration) from magnetic diagnostics data of the T-15MD device using neural networks based on a multilayer perceptron with a backpropagation algorithm. The neural network was trained and tested on limiter D-shaped plasma configurations with positive or negative triangularity. A study was carried out of the influence of noise in the signals of magnetic diagnostics on the reconstruction of geometric parameters by the trained network. The issue of parameter reconstruction in the event of the disconnection of a number of sensors of the magnetic diagnostics system is also considered.