Richtmyer-Meshkov initiated, vortex accelerated inhomogeneous turbulence: High performance computing and visiometrics.

摘要

The compressible stratified turbulent flow initiated by Richtmyer-Meshkov (RM) instability environment, i.e., a shock wave traversing through a density inhomogeneity, is studied numerically in this thesis. We focus on the physics from early time instability to late time turbulence under vortex paradigm, via advanced computing and data visualization technique. The insight we obtained will lead to direct impact on turbulent mixing and mass transport in astrophysics, inertial confinement (laser) fusion and internal combustion.; Upon examination of shock waves interacting with different geometries of density inhomogeneity, we investigate how unstable vortex bilayer (VBL) configuration leads to the presence of vortex projectiles (VP) and consequent topological change of the initial geometry. We further quantify the secondary baroclinic process due to the vortex acceleration and density gradient intensification. This secondary baroclinic process serves as the intrinsic forcing to 2D inhomogeneous turbulence. We generalize this type of flow as accelerated inhomogeneous flow and examine the turbulent decay and mixing in 2D and 3D.; We use Piecewise Parabolic Method (PPM) to solve our compressible Euler system. High-resolution requirements for resolving as many scales as possible in a turbulence study makes the data set significantly larger, and more complex if AMR scheme is invoked. In this thesis, visiometrics is proved to be an design a pipeline of feature-based analysis to extract regions of interest, then visualize, track, isolate and quantify their evolution. We address quantitatively the spatial and temporal diffusivity of the mixing zone, and illustrate the first time the correlation of mass and momentum diffusivity. We encapsulate the comprehensive visiometrics pipeline innovatively into an optimization loop to quantify the error in the simulation in a feedback manner. We are able to expose the errors and correlate these errors to initial physical or numerical parameters during the experimental or numerical investigation. Excellent agreements are obtained between our Navier-Stokes simulations of a shock bubble interaction and the experiments performed at Los Alamos National Lab (LANL). Our results outperformed both qualitatively and quantitatively the simulation at LANL. We believe this methodology could be generalized to other disciplines.