Machine Learning Design Optimization of Inner Shells on Double Shell Target Capsules

The design phase of inertial confinement fusion (ICF) experiments relies on predictive physics simulations. High fidelity predictive physics simulations in xRAGE [1] have many parameters to vary and large computational costs, making systematic parameter scans in search of optimal designs infeasible. A data efficient approach is required to find optimal designs for ICF targets. Machine learning design optimization can identify optimal designs while limiting computational expense, making the application to ICF highly desirable. In this work, Bayesian Optimization with Gaussian processes is applied to the design of double shell targets to maximize neutron yields. Double shell capsules are an alternative, volumetric-based approach to ignition [2]. In this study, double shells with varying inner shell materials are optimized, allowing for comparison of their underlying physics.





References

[1] M. Gittings, R. Weaver, M. Clover, R. Betlach, N. Byrne, R. Coker, E. Dendy, R. Hueckstaedt, K. New, W. Oakes, D. Ranta, and R. Stefan, “The RAGE Radiation-Hydrodynamic Code,” Computational Science Discovery 1 (2008), https://doi.org/10.1088/1749-4699/1/1/015005.



[2] D. Montgomery, W. Daughton, B. Albright, A. Simakov, D. Wilson, E. Dodd, R. Kirkpatrick, R. Watt, M. Gunderson, E. Loomis, et al., “Design considerations for indirectly driven double shell capsules,” Physics of Plasmas 25, 092706 (2018).



[3] N. Vazirani, M. Grosskopf, D. Stark, P. Bradley, B. Haines, E. Loomis, S. England, and W. Scales, “Coupling 1d xrage simulations with machine learning for graded inner shell design optimization in double shell capsules,” Physics of Plasmas 28, 122709 (2021).