A TVD neural network closure and application to turbulent combustion
Not Available
Engineering topics
Publications and source records attributed to Freund, Jonathan B..
Not Available
Evidence indicates that quantities-of-interest in some turbulent flows can be controlled despite the overall chaotic dynamics. It is typically thought that this is via relatively deterministic, larger-scale components of the turbulence. However, finding such controls, if they exist, is challenging because chaos causes sensitivity gradients to explode and the search space to become intractably non-convex. This challenge is analyzed, and a penalty method is introduced to cope with it. In the new approach, the time domain is broken into segments approximately matching the chaos time scales, so that the solution within each segment is both physical and relatively deterministic. The initial condition of each segment is included in an adjoint-based gradient optimization, which temporarily introduces artificial Δq discontinuities in the overall solution. The optimization then proceeds in stages with increasing penalization of Δq. The method is developed and illustrated for a logistic map, the Lorenz Equation, and an advection augmented Kuramoto–Sivashinsky Equation. These examples show how the Δq temporarily increases the search scale prior to the strong Δq → 0 penalization that recovers a physical solution. It is then applied to turbulent Kolmogorov flow, for which it also far outperforms a standard adjoint-based gradient search. Finally, the utility of such an optimized chaotic solution is discussed.
The formation and growth of plasma kernels generated via nano-second mode-beating laser pulses is investigated here via a non-equilibrium self-consistent computational model. Chemically reactive Navier–Stokes equations are used to describe the hydrodynamics, and non-equilibrium effects are taken into account with a two-temperature model. Inverse Bremsstrahlung and multiphoton ionization are included self-consistently in the model via a coupled solution of the plasma governing equations and the radiative transfer equation (that describes the laser beam propagation and attenuation). A self-consistent approach (despite carrying additional challenges) minimizes the empiricism and it allows for a more accurate description since it prevents both the utilization of artificial plasma seeds to trigger the breakdown and the implementation of tuning parameters to simulate the laser-energy deposition. The advantages of this approach are confirmed by the good agreement between the numerically predicted and the experimentally measured plasma boundary evolution and absorbed energy. This also holds true for the periodic plasma kernel structures that, as suggested by the experiments and confirmed by the simulations presented here, are connected to the modulating frequency.
Here, a short focused laser pulse can generate a hot plasma, which expands then contracts and can eject a hot jet, the speed and direction of which is sensitive to the details of the plasma kernel. The coupling of thermal and chemical nonequilibrium plasma mechanisms with this hydrodynamic development is assessed with detailed simulations of a two-temperature, three-species plasma model. Time scales for electron recombination, thermal relaxation, and diffusion are compared to that of the plasma expansion to anticipate conditions in which these mechanisms might affect the vorticity generation that leads to the ultimate flow pattern. The effect of these mechanisms are analyzed through comparison with corresponding inert-gas and equilibrium models. Thermal-nonequilibrium effects are found to be weak due to rapid relaxation of the heavy-particle and electronic temperatures. In contrast, chemical equilibration occurs at a rate comparable to the expansion and thereby enhance both it and subsequent hydrodynamic mechanisms as the energy stored in ion formation during the preceding breakdown is released by electron recombination. Thermal conduction, enhanced by high-temperature free electrons, weakens the ejection.
We report that the weights of a deep neural-network model are optimized in conjunction with the governing flow equations to provide a model for subgrid-scale stresses in a temporally developing plane turbulent jet at Reynolds number Re 0 = 6000 . The objective function for training is first based on the instantaneous filtered velocity fields from a corresponding direct numerical simulation, and the training is by a stochastic gradient descent method, which uses the adjoint Navier-Stokes equations to provide the end-to-end sensitivities of the model weights to the velocity fields. In-sample and out-of-sample testing on multiple dual-jet configurations show that its required mesh density in each coordinate direction for prediction of mean flow, Reynolds stresses, and spectra is half that needed by the dynamic Smagorinsky model for comparable accuracy. The same neural-network model trained directly to match filtered subgrid-scale stresses, without the constraint of being embedded within the flow equations during the training, fails to provide a qualitatively correct prediction. The coupled formulation is generalized to train based only on mean-flow and Reynolds stresses, which are more readily available in experiments. The mean-flow training provides a robust model, which is important, though a somewhat less accurate prediction for the same coarse meshes, as might be anticipated due to the reduced information available for training in this case. The anticipated advantage of the formulation is that the inclusion of resolved physics in the training increases its capacity to extrapolate. This is assessed for the case of passive scalar transport, for which it outperforms established models due to improved mixing predictions.
Not provided.
A framework is introduced that leverages known physics to reduce overfitting in machine learning for scientific applications. The partial differential equation (PDE) that expresses the physics is augmented with a neural network that uses available data to learn a description of the corresponding unknown or unrepresented physics. Training within this combined system corrects for missing, unknown, or erroneously represented physics, including discretization errors associated with the PDE's numerical solution. For optimization of the network within the PDE, an adjoint PDE is solved to provide high-dimensional gradients, and a stochastic adjoint method (SAM) further accelerates training. Additionally, the approach is demonstrated for large-eddy simulation (LES) of turbulence. High-fidelity direct numerical simulations (DNS) of decaying isotropic turbulence provide the training data used to learn sub-filter-scale closures for the filtered Navier–Stokes equations. Out-of-sample comparisons show that the deep learning PDE method outperforms widely-used models, even for filter sizes so large that they become qualitatively incorrect. It also significantly outperforms the same neural network when a priori trained based on simple data mismatch, not accounting for the full PDE. Measures of discretization errors, which are well-known to be consequential in LES, point to the importance of the unified training formulation's design, which without modification corrects for them. For comparable accuracy, simulation runtime is significantly reduced. A relaxation of the typical discrete enforcement of the divergence-free constraint in the solver is also successful, instead allowing the DPM to approximately enforce incompressibility physics. Since the training loss function is not restricted to correspond directly to the closure to be learned, training can incorporate diverse data, including experimental data.
Not provided.
A filtering operation, based on B-spline discretizations, is introduced to target weakly growing mesh-scale oscillations that can arise in high-fidelity turbulence simulations. This is a spectral regularization that can be described using the singular values of a banded matrix operator, with the filtering strength set by a scalar- or vector-valued penalty parameter. The penalty parameter can be specified though it can also be advantageously selected to minimize the generalized cross validation (GCV) measure of distance between the pre- and post-filtered solutions. Efficient algorithms are developed to compute both the scalar and vector penalty parameters. The B-spline filter has a sharper localization to high-wavenumber than compact or explicit filters of the same stencil width and is demonstrated for solutions of the Burgers' equation, decaying Burgers' turbulence, and compressible Navier–Stokes turbulent channel flow. Furthermore, these simulations confirm the scheme's numerical stability and ability to narrowly target the high wavenumber components of numerical solutions. An advantage over finite-difference filters is that these B-spline filters are stable on bounded domains and even preserve formal order of accuracy.
In this paper, a detailed three-dimensional computational model is developed to assess the potential of micron-scale field emission dielectric barrier discharge (FE-DBD) plasma actuators in non-premixed microburners with channel heights in the range of L z = 0.25–0.75 mm. Results for H 2 –O 2 combustion with Reynolds number Re = 100 are studied for different configurations of plasma actuator arrays. Without plasma actuation, results agree with corresponding experiments, although in this case the burner typically suffers from incomplete combustion due to thermal quenching, radical quenching, and slow diffusive mixing. For the range of conditions simulated, plasma actuation increases combustion completeness from 25%–70% to 40%–80% by enhancing mixing, generating radicals, and Joule heating. Ionic wind resulting from the FE-DBD plasma disrupts the otherwise simple laminar flow in the burner, accelerating the growth of a mixing layer and enhancing heat transfer to channel walls. The larger, lower temperature flames reduce thermal gradients (and thereby thermal stresses) in the walls. Radical generation due to electron-impact dissociation extends the reaction zone in the diffusion flame and increases combustion completeness. In all cases, the additional heat release with plasma actuation (100–190 W) is significantly larger than the power needed to sustain the plasma (≲0.1 W), suggesting that plasma actuation may provide a significant advantage in energy generation using realistic portable micro-combustion devices.
A non-equilibrium model for laser-generated plasmas is used to represent collinear dual nano-second-pulse interactions. In the specific case considered, the breakdown is initiated with an ultraviolet (UV) laser pulse at 266 nm, which pre-ionizes the gas, and it is followed by a near-infrared (NIR) pulse at 1064 nm, which deposits significant energy into the ionized mixture. The model is validated against corresponding experiments, and simulation results are interrogated to understand key features of the plasma-kernel dynamics and the post-discharge hydrodynamics. The hydrodynamics of the non-equilibrium plasma is governed by the two-temperature Navier–Stokes equations accounting for both multiphoton ionization and inverse bremsstrahlung. The interaction between the laser beam and the plasma is modeled based on the radiative transfer equation. The temporal and spatial offsets of the two pulses can generate ionization kernels with different topology and dynamics. It is shown that the UV pre-ionization pulse can tailor the plasma region, leading to a larger ionized volume than would occur for a single-pulse breakdown, and it increases the efficiency of the energy deposition for the following NIR discharge. Vorticity in the early post-discharge phase is generated via baroclinic torque from the misalignment of the radial gradient of density (sudden gas expansion) and the strong pressure gradient (initiated by energy deposition from the laser beam). It is predicted that the collinear dual pulse can be designed to specify the plasma kernel size, shape, and maximum temperature in the breakdown phase, and the initial post-breakdown vorticity and plasma-core decay.
Particle-in-cell (PIC) simulation methods are attractive for representing species distribution functions in plasmas. However, as a model, they introduce uncertain parameters, and for quantifying their prediction uncertainty it is useful to be able to assess the sensitivity of a quantity-of-interest (QoI) to these parameters. Such sensitivity information is likewise useful for optimization. However, computing sensitivity for PIC methods is challenging due to the chaotic particle dynamics, and sensitivity techniques remain underdeveloped compared to those for Eulerian discretizations. This challenge is examined from a dual particle–continuum perspective that motivates a new sensitivity discretization. Two routes to sensitivity computation are presented and compared: a direct fully-Lagrangian particle-exact approach provides sensitivities of each particle trajectory, and a new particle-pdf discretization, which is formulated from a continuum perspective but discretized by particles to take the advantages of the same type of Lagrangian particle description leveraged by PIC methods. Since the sensitivity particles in this approach are only indirectly linked to the plasma-PIC particles, they can be positioned and weighted independently for efficiency and accuracy. The corresponding numerical algorithms are presented here in mathematical detail. The advantage of the particle-pdf approach in avoiding the spurious chaotic sensitivity of the particle-exact approach is demonstrated for Debye shielding and sheath configurations. In essence, the continuum perspective makes implicit the distinctness of the particles, which circumvents the Lyapunov instability of the N -body PIC system. The cost of the particle-pdf approach is comparable to the baseline PIC simulation.
The paper presents velocity measurements of shock-induced flow field, leading to vortex generation and plasma deformation in air. Femtosecond-laser electronic excitation tagging (FLEET) velocimetry was performed at the times Δt = 3, 20, 50, and 100 µs post laser-induced breakdown. Emissions over Δt = 3–10 µs showed the propagations of the initially elliptic shock, transitioning into a spherical front. The shock emanated along the laser axis causes the flow outward, and then the pressure gradient generated by the rarefaction wave drives the inward flow at later moments, with the velocity magnitude approaching a steady-state value of 40 m s -1 . Temporal velocity evolution was compared with non-self-similar solutions behind the propagating shock, which are sensitive to the size of the energy deposition, and the use of the measured initial plasma diameter reproduced the experiment. There establishes a region of uniform velocity around 35 m s -1 in the air-flow running through the plasma, which triggers the roll-up of the plasma surface by a large-scale vortex, providing the detail of flow field evolution from the shock propagation to the plasma deformation. A collective Thomson scattering and hybrid fs/ps pure rotational coherent anti-Stokes Raman scattering (CARS) were also performed to gain insight into the high-temperature plasma. An effective electron–ion recombination rate of 2 x 10 -12 cm 3 s -1 was measured at Δt = 0.5–10 µs, during a dynamic plasma expansion and compression. When the shock resides close to the plasma at Δt = 0.5–1 µs, the temperature distributions were found to follow the similarity law.
A multi-physics numerical model for laser-induced optical breakdown and laser-plasma interaction in a non-equilibrium gas is presented, accounting for: production of priming electrons via multi-photon ionization, energy absorption, cascade ionization, induced hydrodynamic response, and shock formation and propagation. The gas is governed by the Navier-Stokes equations, with non-equilibrium effects taken into account by means of a two-temperature model. The space-time dependence of the laser beam is modeled with a flux-tube formulation for the Radiative Transfer Equation. The flow governing equations are discretized in space using a second-order finite volume method. The semi-discrete equations are marched in time using an implicit-explicit (IMEX) dual time-stepping strategy, where diffusion and chemistry are solved implicitly, whereas convection is explicit. Application to a 20 ns long 50 mJ pulse laser-induced breakdown in quiescent O 2 shows the advantages of this temporal discretization during and just after the laser pulse, while a less-expensive symmetric Strang splitting (with implicit chemistry) is sufficient for the post-breakdown gas dynamics after ≃ 0.1 μs. The integrated model is shown to reproduce key features of corresponding experiments.
The behavior of nonpremixed edge flames to low and high strains in a wedge-shaped region are simulated and studied in detail. The chemistry is modeled by a one-step mechanism and transport by constant coefficient Fickian diffusion, as a qualitative model describing the physical mechanisms in the thermochemistry of combustion of hydrocarbon fuels. The reactive zero-Mach-number Navier-Stokes equations are used to describe the flow, both in a thermodiffusive regime (constant density) and a hydrodynamically-coupled regime (variable density). This configuration is an adaptation of the experimental opposed inclined jet geometry long used at the University of Southern California, but made stationary by the addition of a sink at the apex of the wedge. This crucial modification enables the stable presence of stationary edge flames, and it allows the study of advancing and retreating edge flames in the low and high strain regimes. The dependence of the edge flames speed on strain rate, stoichiometry, and Lewis numbers is discussed and compared to previous studies.