Progress on Machine Learning Informed Design Optimization for Double Shell Capsule Graded Density Inner Layer Targets

Abstract

Advances in machine learning have the ability to cut down design costs and enhance the design process for high-energy-density experiments. Machine learning can use expensive simulation data to search the design space efficiently for an optimal design of inertial confinement fusion experiments with a minimal number of simulations. Here we present work focused on indirectly driven double shell capsultes. The focus is optimizing graded density inner shell of the double shell target while reducing hydrodynamic instabilities and maintaining high yields. Graded layer inner shell targets have a vast parameter space that is too large to fully map out, which is why efficient exploration of the design space is not only beneficial but also necessary. Machine learning methods can use predictive physics simulations to identify graded layer designs with high predicted performance as well as novel designs with high uncertainty in performance that may hold unexpected promise. Here we present the application of cutting-edge Bayesian optimization to one dimensional design optimization of double shell graded inner shell targets. By applying machine learning tools to the simulation design, we aim to optimize the target geometry to mitigate the hydrodynamic instabilities and improve yield. Here we present our progress on laying the groundwork for a useful machine learning informed design tool for future NIF experiments with graded layer inner shell targets.