Bayesian Equilibrium Reconstruction for General Fusion Demonstration Plant

General Fusion is designing a magnetized target fusion reactor to compress a toroidal plasma inside a liquid metal cavity and heat it to fusion conditions. Plasma properties are inferred using Bayesian statistics by comparing measurements from our spherical tokamak targets to artificial diagnostic signals from a table of precalculated equilibria. Equilibria are assigned probabilities based on the least-squares fit to diagnostics, temporal constraints such as decaying helicity, and the linear stability of the plasma configuration calculated with RDCON. Measurements from Mirnov probes, interferometers, polarimeters, and Thomson scattering are included as fit constraints. Equilibria are generated using the FLAGSHIPS Grad-Shafranov solver as a 2^nd-order finite element solution on an unstructured mesh. Outputs include the current, pressure, temperature, and density profiles with their probability distributions shown as 2D histograms. We present fits to a wide range of test cases and real shots showing a large set of equilibria that fall inside the measurement uncertainties, providing high levels on confidence in the accuracy of the method. We also show how our framework is used to design probe placement in the Fusion Demonstration Plant to facilitate reconstruction accuracy.