AI-Based Inverse Design for Control of Shock Wave Dynamics
Please login to view abstract download link
As a singular attractor of the conservation equations underlying Navier-Stokes hydrodynamic flow, shockwaves are ubiquitous in nature. Found in natural phenomena such as earthquakes, lightning and stellar supernova, as well as unnatural phenomena, e.g. aerospace and inertial confinement fusion applications, an important facet of shockwaves is being able to understand, predict and (possibly) control their effect on material and fluid interfaces. The interaction of shockwaves at material interfaces of differing impedance results in an inherently unstable dynamical trajectory, with non-linear growth of perturbations and subsequent jetting occurring in chaotic fashion owing to the well-known Richtmyer-Meshkov instability (RMI). RMI is sustained via baroclinic torque arising from the misalignment of density and pressure gradients, as angular momentum is deposited at material interfaces by passage of the shockwave. There are very few known solutions to preventing the formation of RMI and even partial progress toward its stabilization would enable many important applications in physics and engineering. Through application of new artificial intelligence and machine learning workflows, coupled to simulated hydrodynamics, we show that there exist solutions whereby RMI can be controlled, and made to be either suppressed or enhanced at will via engineering special initial conditions. It has been found that the solution space for stabilization of RMI is surprisingly large – ML techniques have enabled the focusing of massive supercomputing resources to attain this understanding through hydrodynamic design optimization. This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under contract DE-AC52-07NA27344. Lawrence Livermore National Security, LLC. We gratefully acknowledge the LLNL Lab Directed Research and Development Program for funding support of this research under Project No. 21-SI-006 and from the Advanced Simulation and Computing program. LLNL-ABS-870431
