Optimization of a Diagnostic Suite for Magnetized Target Fusion Experiments

Future Magnetized Target Fusion (MTF) experiments at General Fusion (GF) will attempt to compress a plasma with an imploding liner to 10 keV and approach the Lawson criterion. A suite of diagnostic sensors will measure plasma density n, temperature Te, and energy confinement time at various compression stages. The suite will maximize accuracy on these parameters or their source quantities (magnetic energy, Ohmic power). Other plasma parameters (q and lambda profile) will also be quantified to monitor plasma instabilities and characterize the compression target.

Machine learning algorithms are applied to libraries of Grad-Shafranov (GS) equilibria calculated with GF's custom solver, and to scaling laws from simulated compression cases. The data are processed by a dimensionality reduction algorithm paired with a clustering algorithm, which identifies key sensor positions. The expected error on each parameter is back-calculated with a reconstruction algorithm. A Monte Carlo error propagation algorithm is finally applied to each case, in order to estimate the impact of measurement accuracy on the reconstructed values. Preliminary results show that reliable measurements of magnetic parameters can be achieved with 10 surface probes, with errors on magnetic energy as low as 10%. The same analysis will be expanded to include more diagnostics, and to several stages of compression.

A custom Monte Carlo algorithm paired with physics-based constraints will predict the expected error on the triple product, quantify the accuracy required on each diagnostic system, and provide a reliable confirmation of the main scientific milestones.