arXiv Science⌕ Search

arXiv subjects

Daniel Livescu

Publications and source records attributed to Daniel Livescu.

49 records · Page 3Linked to original sources
arXiv2019

High-resolution Navier-Stokes simulations of Richtmyer-Meshkov instability with re-shock

The interaction of a Mach 1.45 shock wave with a perturbed planar interface between sulphur hexafluoride and air is studied through high-resolution two-dimensional (2D) and three-dimensional (3D) shock-capturing adaptive mesh refinement simulations of multi-species Navier-Stokes equations. The sensitivities of time-dependent statistics on grid resolution for 2D and 3D simulations are evaluated to ensure the accuracy of the results. The numerical results are used to examine the differences between the development of 2D and 3D Richtmyer-Meshkov instability during two different stages: (1) initial growth of hydrodynamic instability from multi-mode perturbations after first shock and (2) transition to chaotic or turbulent state after re-shock. The effects of the Reynolds number on the mixing in 3D simulations are also studied through varying the transport coefficients.

physics.flu-dyn↗
arXiv2019

Time-series learning of latent-space dynamics for reduced-order model closure

We study the performance of long short-term memory networks (LSTMs) and neural ordinary differential equations (NODEs) in learning latent-space representations of dynamical equations for an advection-dominated problem given by the viscous Burgers equation. Our formulation is devised in a non-intrusive manner with an equation-free evolution of dynamics in a reduced space with the latter being obtained through a proper orthogonal decomposition. In addition, we leverage the sequential nature of learning for both LSTMs and NODEs to demonstrate their capability for closure in systems which are not completely resolved in the reduced space. We assess our hypothesis for two advection-dominated problems given by the viscous Burgers equation. It is observed that both LSTMs and NODEs are able to reproduce the effects of the absent scales for our test cases more effectively than intrusive dynamics evolution through a Galerkin projection. This result empirically suggests that time-series learning techniques implicitly leverage a memory kernel for coarse-grained system closure as is suggested through the Mori-Zwanzig formalism.

physics.comp-ph↗
arXiv2019

Density Effects on the Post-shock Turbulence Structure and Dynamics

Turbulence structure resulting from multi-fluid or multi-species, variable-density isotropic turbulence interaction with a Mach 2 shock is studied using turbulence-resolving shock-capturing simulations and Eulerian (grid) and Lagrangian (particle) methods. The complex roles density play in the modification of turbulence by the shock wave are identified. Statistical analyses of the velocity gradient tensor (VGT) show that the density variations significantly change the turbulence structure and flow topology. Specifically, a stronger symmetrization of the joint probability density function (PDF) of second and third invariants of the anisotropic velocity gradient tensor, PDF$(Q^\ast, R^\ast)$, as well as the PDF of the vortex stretching contribution to the enstrophy equation, are observed in the multi-species case. Furthermore, subsequent to the interaction with the shock, turbulent statistics also acquire a differential distribution in regions having different densities. This results in a nearly symmetrical PDF$(Q^\ast, R^\ast)$ in heavy fluid regions, while the light fluid regions retain the characteristic tear-drop shape. To understand this behavior and the return to "standard" turbulence structure as the flow evolves away from the shock, Lagrangian dynamics of the VGT and its invariants are studied by considering particle residence times and conditional particle variables in different flow regions. The pressure Hessian contributions to the VGT invariants transport equations are shown to be not only affected by the shock wave, but also by the density in the multi-fluid case, making them critically important to the flow dynamics and turbulence structure.

physics.flu-dyn↗
arXiv2019

Leveraging Bayesian Analysis To Improve Accuracy of Approximate Models

We focus on improving the accuracy of an approximate model of a multiscale dynamical system that uses a set of parameter-dependent terms to account for the effects of unresolved or neglected dynamics on resolved scales. We start by considering various methods of calibrating and analyzing such a model given a few well-resolved simulations. After presenting results for various point estimates and discussing some of their shortcomings, we demonstrate (a) the potential of hierarchical Bayesian analysis to uncover previously unanticipated physical dependencies in the approximate model, and (b) how such insights can then be used to improve the model. In effect parametric dependencies found from the Bayesian analysis are used to improve structural aspects of the model. While we choose to illustrate the procedure in the context of a closure model for buoyancy-driven, variable-density turbulence, the statistical nature of the approach makes it more generally applicable. Towards addressing issues of increased computational cost associated with the procedure, we demonstrate the use of a neural network based surrogate in accelerating the posterior sampling process and point to recent developments in variational inference as an alternative methodology for greatly mitigating such costs. We conclude by suggesting that modern validation and uncertainty quantification techniques such as the ones we consider have a valuable role to play in the development and improvement of approximate models.

physics.comp-ph↗
arXiv2019

Compressed Convolutional LSTM: An Efficient Deep Learning framework to Model High Fidelity 3D Turbulence

High-fidelity modeling of turbulent flows is one of the major challenges in computational physics, with diverse applications in engineering, earth sciences and astrophysics, among many others. The rising popularity of high-fidelity computational fluid dynamics (CFD) techniques like direct numerical simulation (DNS) and large eddy simulation (LES) have made significant inroads into the problem. However, they remain out of reach for many practical three-dimensional flows characterized by extremely large domains and transient phenomena. Therefore designing efficient and accurate data-driven generative approaches to model turbulence is a necessity. We propose a novel training approach for dimensionality reduction and spatio-temporal modeling of the three-dimensional dynamics of turbulence using a combination of Convolutional autoencoder and the Convolutional LSTM neural networks. The quality of the emulated turbulent fields is assessed with rigorous physics-based statistical tests, instead of visual assessments. The results show significant promise in the training methodology to generate physically consistent turbulent flows at a small fraction of the computing resources required for DNS.

physics.flu-dyn↗
arXiv2019

Reaction analogy based forcing for incompressible scalar turbulence

We present a novel reaction analogy (RA) based forcing method for generating stationary scalar fields in incompressible turbulence. The new method can produce more general scalar PDFs (e.g. double-delta) than current methods, while ensuring that scalar fields remain bounded. Such features are useful for generating initial fields in non-premixed combustion or for studying non-Gaussian scalar turbulence. The RA method mathematically models hypothetical chemical reactions that convert reactants in a mixed state back into its pure unmixed components. Various types of chemical reactions are formulated and the corresponding mathematical expressions derived such that the reaction term is smooth in the scalar space and is consistent with mass conservation. For large values of the scalar dissipation rate, the method produces statistically steady double-delta scalar PDFs. Quasi-uniform, Gaussian, and stretched exponential scalar statistics are recovered for smaller values of the scalar dissipation rate. The shape of the scalar PDF can be further controlled by changing the stoichiometric coefficients of the reaction. The ability of the new method to produce fully developed passive scalar fields with quasi-Gaussian PDFs is also investigated, by exploring the convergence of the third order mixed structure function to the "four-thirds" Yaglom's law.

physics.flu-dyn↗
arXiv2018

Effects of Isothermal Stratification Strength on Vorticity Dynamics for Single-Mode Compressible Rayleigh-Taylor Instability

The effects of isothermal stratification strength on vorticity dynamics for single-mode Rayleigh-Taylor instability (RTI) are examined using two dimensional fully compressible wavelet-based direct numerical simulations. The simulations model low Atwood number ($A=0.04$) RTI development for four different stratification strengths, corresponding to Mach numbers from 0.3 (weakly stratified) to 1.2 (strongly stratified), and for three different perturbation Reynolds numbers, from 5,000 to 20,000. All simulations use adaptive wavelet-based mesh refinement to achieve very fine spatial resolutions at relatively low computational cost. For all stratifications, the bubble and spike go through the exponential growth regime, followed by a slowing of the RTI evolution. For the weakest stratification, this slow-down is then followed by a re-acceleration, while for stronger stratifications the suppression of RTI growth continues. Bubble and spike asymmetries are observed for weak stratifications, with bubble and spike growth rates becoming increasingly similar as the stratification strength increases. For the range of cases studied, there is relatively little effect of Reynolds number on bubble and spike heights, although the formation of secondary vortices becomes more pronounced as Reynolds number increases. The underlying dynamics are analyzed in detail through an examination of the vorticity transport equation, revealing that incompressible baroclinicity drives RTI growth for small and moderate stratifications, but increasingly leads to the suppression of vorticity production and RTI growth for stronger stratifications. These variations in baroclinicity are used to explain the suppression of RTI growth for strong stratifications, as well as the anomalous asymmetry in bubble and spike growth rates for weak stratifications.

physics.flu-dyn↗
arXiv2018

Two-point spectral model for variable-density homogeneous turbulence

We present a study of buoyancy-driven variable-density homogeneous turbulence, using a two-point spectral closure model. We compute the time-evolution of the spectral distribution in wavenumber $k$ of the correlation of density and specific-volume $b(k)$, the mass flux $\bm{a}(k)$, and the turbulent kinetic energy $E(k)$, using a set of coupled equations. Under the modeling assumptions, each dynamical variable has two coefficients governing spectral transfer among modes. In addition, the mass flux $\bm{a}(k)$ has two coefficients governing the drag between the two fluids. Using a prescribed initial condition for $b(k)$ and starting from a quiescent flow, we first evaluate the relative importance of the different coefficients used to model this system, and their impact on the statistical quantities. We next assess the accuracy of the model, relative to Direct Numerical simulation of the complete hydrodynamical equations, using $b$, ${\bm a}$ and $E$ as metrics. We show that the model is able to capture the spectral distribution and global means of all three statistical quantities at both low and high Atwood number for a set of optimized coefficients. The optimization procedure also permits us to discern a minimal set of four coefficients which are sufficient to yield reasonable results while pointing to the mechanisms that dominate the mixing process in this problem.

physics.flu-dyn↗
arXiv2004

Small scale structure of homogeneous turbulent shear flow

The structure of homogeneous turbulent shear flow is studied using data generated by Direct Numerical Simulations (DNS) and a linear analysis for both compressible and incompressible cases. At large values of the mean shear rate, the Rapid Distortion Theory (RDT) limit is approached. Analytical solutions are found for the inviscid compressible RDT equations at long times. The RDT equations are also solved numerically for both inviscid and viscous cases. The RDT solutions, confirmed by the DNS results, show that the even order transverse derivative moments of the dilatational and solenoidal velocity fields are anisotropic, with the dilatational motions more anisotropic than their solenoidal counterparts. The results obtained for the incompressible case are similar to those obtained for the solenoidal motions in the compressible case. The DNS results also indicate an increase in the anisotropy of the even order transverse derivative moments with the order of the moment, in agreement with the RDT predictions. Although the anisotropy decreases with Reynolds number, it is likely that for higher even order moments it will persist at large values of the Reynolds number, in contrast with the postulate of local isotropy. The RDT solutions also predict that the normalized odd order transverse derivative moments of the solenoidal velocity for the compressible case and of the velocity for the incompressible case should approach a constant different than zero at large times. This prediction is supported by the DNS data. For higher odd order normalized moments, the RDT analysis suggests that the anisotropy may persist at large values of the Reynolds number, in agreement with the existent experimental data.

physics.flu-dyn↗
arXiv2003

Compressibility effects on the Rayleigh-Taylor instability growth between immiscible fluids

The linearized Navier-Stokes equations for a system of superposed immiscible compressible ideal fluids are analyzed. The results of the analysis reconcile the stabilizing and destabilizing effects of compressibility reported in the literature. It is shown that the growth rate $n$ obtained for an inviscid, compressible flow in an infinite domain is bounded by the growth rates obtained for the corresponding incompressible flows with uniform and exponentially varying density. As the equilibrium pressure at the interface $p_\infty$ increases (less compressible flow), $n$ increases towards the uniform density result, while as the ratio of specific heats $γ$ increases (less compressible fluid), $n$ decreases towards the exponentially varying density incompressible flow result. This remains valid in the presence of surface tension or for viscous fluids and the validity of the results is also discussed for finite size domains. The critical wavenumber imposed by the presence of surface tension is unaffected by compressibility. However, the results show that the surface tension modifies the sensitivity of the growth rate to a differential change in $γ$ for the lower and upper fluids. For the viscous case, the linearized equations are solved numerically for different values of $p_\infty$ and $γ$. It is found that the largest differences compared with the incompressible cases are obtained at small Atwood numbers. The most unstable mode for the compressible case is also bounded by the most unstable modes corresponding to the two limiting incompressible cases.

physics.flu-dyn↗
arXiv2002

Compressibility effects on the scalar mixing in reacting homogeneous turbulence

The compressibility and heat of reaction influence on the scalar mixing in decaying isotropic turbulence and homogeneous shear flow are examined via data generated by direct numerical simulations (DNS). The reaction is modeled as one-step, exothermic, irreversible and Arrhenius type. For the shear flow simulations, the scalar dissipation rate, as well as the time scale ratio of mechanical to scalar dissipation, are affected by compressibility and reaction. This effect is explained by considering the transport equation for the normalized mixture fraction gradient variance and the relative orientation between the mixture fraction gradient and the eigenvectors of the solenoidal strain rate tensor.

physics.flu-dyn↗
arXiv2002

Non-premixed Flame-Turbulence Interaction in Compressible Turbulent Flow

Nonpremixed turbulent reacting flows are intrinsically difficult to model due to the strong coupling between turbulent motions and reaction. The large amount of heat released by a typical hydrocarbon flame leads to significant modifications of the thermodynamic variables and the molecular transport coefficients and thus alters the fluid dynamics. Additionally, in nonpremixed combustion, the flame has a complex spatial structure. Localized expansions and contractions occur, enhancing the dilatational motions. Therefore, the compressibility of the flow and the heat release are intimately related. However, fundamental studies of the role of compressibility on the scalar mixing and reaction are scarce. In this paper we present results concerning the fundamental aspects of the interaction between non-premixed flame and compressible turbulence.

physics.flu-dyn↗