The Many Faces of Magnetized Liner Inertial Fusion: A Data-Driven Approach to Model-Free Characterization of Image Metric Sensitivities for Stagnation Imaging Applications

Magnetized Liner Inertial Fusion (MagLIF) is a magneto-inertial fusion concept studied on Sandia National Laboratories Z-machine. The stagnated fuel plasma produced in MagLIF experiments exhibits structural variability and is observed using spherical-crystal imaging (CXI). We demonstrate a model-free, experimental-data-driven approach to understanding sensitivities of image metrics applied to CXI diagnostics. A convolutional neural network segmentation, registration, and background subtraction routine was developed to automatically preprocess images enabling this study. A suitably chosen metric is computed for over 100 preprocessed images, including multiple scans and differences in spectral filtering, from over 50 experiments. Comparisons are made by computing differences between images in the metric space. We find it essential to account for texture introduced by digitizing image plates to obtain a metric sensitive to structure of the stagnated fuel-liner system. This enables large-scale experimental searches for significant differences in the metric without need to explore images by hand. We demonstrate sensitivity to fuel morphology, the presence of cold liner emission, and underlying conditions such as mix or liner areal density. We also show that it is easy to distinguish between experiment and analytic models. This may enable optimization of model inputs to match the distribution of experimental data, providing understanding of the most critical characteristics of image morphology.