Equilibrium Reconstruction on the START Spherical Tokamak Using a Neural Network

The neural network post-shot reconstruction analysis, as implemented on the START spherical tokamak at Cul- ham, is described. This used a combination of magnetic and soft X-ray diagnostic inputs to evaluate selected plasma parameters as a function of time throughout the shot. An in-built “novelty detector “was used to detect inadequate database coverage associated with plasma configurations not represented in the database. The method is illustrated by results from the high beta campaign on START.