Enhancing historical electron temperature data with an artificial neural network in the C-2U FRC

The electron temperature is a vital parameter in understanding the dynamics of fusion plasmas, helping to determine basic properties of the system, stability, and fast ion lifetime. We present a method for improving the sampling rate of historical Thomson scattering data by a factor of 103 on the decommissioned beam-driven C-2U field reversed configuration device by utilizing an artificial neural network. This work details the construction of the model, including an analysis of input signals and the model hyperparameter space. The model's performance is evaluated on both a random subset and selected ensemble of testing data and its predictions are found to agree with the Thomson measurements in both cases. Finally, the model is used to reconstruct the effect of the micro-burst instability in C-2U, which is then compared to more recent results in C-2W, showing that the effects of the micro-burst on core electron temperature have been mitigated in C-2W.