Inferring time resolved electron temperature of imploded capsules using Convolutional Neural Networks

Here, we illustrate the application of a deep neural network structure to aid in understanding the results from fusion experiments at the National Ignition Facility. X-ray data generated from capsule implosion experiments can be used to infer the temperature of the hot core within the implosion, which can reach several millions of degrees! In order to get the temperature, the measured x-ray data is used in a forward fit algorithm that compares the measurement to synthetic signal based on several models. Since the resulting temperature depends heavily on the model used, there is some uncertainty in this technique. As an alternative, a deep neural network is developed using thousands of 2-D and 3-D hydrodynamic simulations. Several experiments with known electron temperatures will be used as a bridge from simulations to data. This presentation will describe the deep learning technique employed, as well as the parameters and strategy used to match the simulations. The results from this approach will be compared with that obtained with analytical models.