Electron Temperature Estimate in C-2 FRC Using Neural Network

The electron temperature ($T_e$) of the plasma inside the C-2 Experiment [1] is measured using a Thomson Scattering (TS) diagnostic. The high energy required for the TS laser pulses limits the number of measurements that can be taken during the lifetime of the C-2 plasma; typically, 2 measurements are made by the TS system during each discharge. In order to get an estimate of the temporal evolution of average $T_e$ inside the separatrix throughout a shot, a novel method has been developed using Neural Networks. Using other diagnostic measurements, which are known to be dependent on $T_e$, we train a neural network to produce an empirical mapping from these diagnostics to the TS-measured $T_e$. We tested this network on TS pulses which were excluded from the neural network training, and we consistently get values for $T_e$ within 10 eV of the TS measurements, agreeing within the error of the TS diagnostic. This allows us to confidently produce a trace for each shot with estimates of the average $T_e$ at a rate of 500 kHz. Further work has been done to use a similar technique to estimate a radial $T_e$ profile. While not yet as successful as the average $T_e$ estimate, this has shown great promise.\\[4pt] [1] M. Tuszewski et al., Phys. Rev. Lett. 108, 255008 (2012)