Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “generative diffusion”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,765 records · Page 98

Realizing the Full Potential of Insertion Anodes for Mg-ion Batteries Through the Nano-Structuring of Sn

For next-generation rechargeable batteries, magnesium is of great interest as an alternative to Lithium due to its relative abundance, low toxicity, and bivalent charge (3833 mAh/cm3 and 2205 mAh/g). However, Mg-ion batteries face unique challenges related to the formation of anode passivation layers, anode-electrolyte-cathode incompatibilities, slow solid-state Mg2+ diffusion, and ion trapping. Using analytical (scanning) transmission electron microscopy ((S)TEM) and ab initio modeling, we have investigated Mg2+ intercalation and extraction mechanisms in β-SnSb alloy nanoparticles (NPs). During the first several charge-discharge cycles, the SnSb particles irreversibly break down into a network of pure-Sn and Sb-rich sub-particles, as Mg-ions replace Sn ions in the SnSb lattice. Once the morphology has stabilized, the small Sn NPs (< 20 nm) are responsible for the majority of reversible storage capacity, while the Sb-rich particles trap substitutional-Mg atoms in the lattice and are significantly less active. This result strongly indicates that pure-Sn nanoparticles on a graphene support can act as a high capacity anode for Mg-ion batteries.

Environmental Molecular Sciences Laboratory↗

Evolution of micro-pores in Ni–Cr alloys via molten salt dealloying

Porous materials with high specific surface area, high porosity, and high electrical conductivity are promising materials for functional applications, including catalysis, sensing, and energy storage. Molten salt dealloying was recently demonstrated in microwires as an alternative method to fabricate porous structures. The method takes advantage of the selective dissolution process introduced by impurities often observed in molten salt corrosion. This work further investigates molten salt dealloying in bulk Ni–20Cr alloy in both KCl–MgCl 2 and KCl–NaCl salts at 700 °C, using scanning electron microscopy, energy dispersive spectroscopy, and X-ray diffraction (XRD), as well as synchrotron X-ray nano-tomography. Micro-sized pores with irregular shapes and sizes ranging from sub-micron to several microns and ligaments formed during the process, while the molten salt dealloying was found to progress several microns into the bulk materials within 1–16 h, a relatively short reaction time, enhancing the practicality of using the method for synthesis. The ligament size increased from ~ 0.7 μm to ~ 1.3 μm in KCl–MgCl 2 from 1 to 16 h due to coarsening, while remaining ~ 0.4 μm in KCl–NaCl during 16 h of exposure. The XRD analysis shows that the corrosion occurred primarily near the surface of the bulk sample, and Cr 2 O 3 was identified as a corrosion product when the reaction was conducted in an air environment (controlled amount sealed in capillaries); thus surface oxides are likely to slow the morphological coarsening rate by hindering the surface diffusion in the dealloyed structure. 3D-connected pores and grain boundary corrosion were visualized by synchrotron X-ray nano-tomography. This study provides insights into the morphological and chemical evolution of molten salt dealloying in bulk materials, with a connection to molten salt corrosion concerns in the design of next-generation nuclear and solar energy power plants.

36 MATERIALS SCIENCE↗

Water confinement in nanoporous silica materials

The influence of the surface polarity of cylindrical silica nanopores and the presence of Na{sup +} ions as compensating charges on the structure and dynamics of confined water has been investigated by molecular dynamics simulations. A comparison between three different matrixes has been included: a protonated nanopore (PP, with SiOH groups), a deprotonated material (DP, with negatively charged surface groups), and a compensated-charge framework (CC, with sodium cations compensating the negative surface charge). The structure of water inside the different pores shows significant differences in terms of layer organization and hydrogen bonding network. Inside the CC pore the innermost layer is lost to be replaced by a quasi bulk phase. The electrostatic field generated by the DP pore is felt from the surface to the centre of pore leading to a strong orientation of water molecules even in the central part of the pore. Water dynamics inside both the PP and DP pores shows significant differences with respect to the CC pore in which the sub-diffusive regime of water is lost for a superdiffusive regime.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

ON THE MAGNETIZATION OF COSMIC OUTFLOWS: PLASMA MODES AND INSTABILITIES OF UNMAGNETIZED PLASMA BEAMS

The dissipation of the kinetic energy of cosmic outflows in interactions with ambient collision-free plasmas and the associated generation of electromagnetic plasma turbulence is a fundamental problem of modern astrophysics. Thermalization by elastic two-body Coulomb collisions is orders of magnitudes too slow as compared to interactions with electric and magnetic fields because of the generally low density of cosmic plasmas. Alternative dissipation mechanisms have to be examined such as energy diffusion by second-order Fermi interactions of charged particles with electromagnetic turbulence, which are an intrinsic property of any sufficiently agitated magnetized plasma. We consider the microphysical details of the energy conversion in relativistic and nonrelativistic outflows by investigating the solutions of the linear plasma dispersion relation in an unmagnetized anisotropic beam plasma consisting of two overall-neutral particle beams propagating with arbitrary velocities in the same direction. The general plasma dispersion relation is derived for arbitrary propagation angle {theta} with respect to the beam propagation direction both in the initial laboratory frame and in the counterstream frame of reference. Solutions of the linear dispersion relation are derived for parallel ({theta} = {pi}/2) and perpendicular ({theta} = 0) propagation angle, respectively. For parallel propagation angles, the electrostatic mode is excited and its maximum growth rate depends on the relative bulk Lorentz factor of the flows and their density ratio. For perpendicular propagation angles, the aperiodic filamentation mode is excited and its maximum growth rate depends differently on the relative bulk Lorentz factor of the flow and their density ratio. The respective maximum growth rates indicate that for nonrelativistic flow velocities the electrostatic instability (EI) is excited much faster than the filamentation instability (FI), whereas for relativistic flow velocities the FI has a larger growth rate than the EI. This implies that nonrelativistic flows dissipate their kinetic energy predominantly in electric fields associated with the electrostatic mode, while relativistic flows dissipate their kinetic energy predominantly in magnetic fields. Therefore, initially unmagnetized relativistic flows are more effectively magnetized than initially unmagnetized nonrelativistic flows.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

THREE-DIMENSIONAL HYDRODYNAMICAL SIMULATIONS OF A PROTON INGESTION EPISODE IN A LOW-METALLICITY ASYMPTOTIC GIANT BRANCH STAR

We use the three-dimensional (3D) stellar structure code DJEHUTY to model the ingestion of protons into the intershell convection zone of a 1 M{sub Sun} asymptotic giant branch star of metallicity Z = 10{sup -4}. We have run two simulations: a low-resolution one of around 300,000 zones and a high-resolution one consisting of 2,000,000 zones. Both simulations have been evolved for about 4 hr of stellar time. We observe the existence of fast, downward flowing plumes that are able to transport hydrogen into close proximity to the helium-burning shell before burning takes place. The intershell in the 3D model is richer in protons than the 1D model by several orders of magnitude and so we obtain substantially higher hydrogen-burning luminosities-over 10{sup 8} L{sub Sun} in the high-resolution simulation-than are found in the 1D model. Convective velocities in these simulations are over ten times greater than the predictions of mixing length theory, though the 3D simulations have greater energy generation due to the enhanced hydrogen burning. We find no evidence of the convective zone splitting into two, though this could be as a result of insufficient spatial resolution or because the models have not been evolved for long enough. We suggest that the 1D mixing length theory and particularly the use of a diffusion algorithm for mixing do not give an accurate picture of these events. An advective mixing scheme may give a better representation of the transport processes seen in the 3D models.

79 ASTRONOMY AND ASTROPHYSICS↗

COSMIC-RAY AND X-RAY HEATING OF INTERSTELLAR CLOUDS AND PROTOPLANETARY DISKS

Cosmic-ray and X-ray heating are derived from the electron energy-loss calculations of Dalgarno, Yan, and Liu for hydrogen-helium gas mixtures. These authors treated the heating from elastic scattering and collisional de-excitation of rotationally excited hydrogen molecules. Here we consider the heating that can arise from all ionization and excitation processes, with particular emphasis on the reactions of cosmic-ray and X-ray generated ions with the heavy neutral species, which we refer to as chemical heating. In molecular regions, chemical heating dominates and can account for 50% of the energy expended in the creation of an ion pair. The heating per ion pair ranges in the limit of negligible electron fraction from {approx}4.3 eV for diffuse atomic gas to {approx}13 eV for the moderately dense regions of molecular clouds and to {approx}18 eV for the very dense regions of protoplanetary disks. An important general conclusion of this study is that cosmic-ray and X-ray heating depends on the physical properties of the medium, i.e., on the molecular and electron fractions, the total density of hydrogen nuclei, and, to a lesser extent, on the temperature. It is also noted that chemical heating, the dominant process for cosmic-ray and X-ray heating, plays a role in UV irradiated molecular gas.

79 ASTRONOMY AND ASTROPHYSICS↗

Increased light-emission efficiency in disordered (In, Ga)N through the correlated reduction of recombination rates

Experiments have shown that the light-emission efficiency of indium gallium nitride (InGaN) lightemitting diodes improves with increasing indium concentration. It is widely thought that compositional fluctuations due to indium incorporation suppress diffusion of carriers to non-radiative centers, thus leading to defect-insensitive emission. However, recent experiments have challenged this hypothesis by revealing unexpectedly long diffusion lengths at room temperature. Here, we demonstrate an alternative mechanism involving the correlated reduction in radiative and nonradiative recombination rates that explains the increase in light-emission efficiency of InGaN with increasing indium concentration, without invoking the suppression of carrier diffusion. Furthermore, our analysis challenges the notion that carrier localization gives rise to defect tolerance in InGaN.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Application of Scalar Filtered Density Function to Turbulent Flows Under Supercritical Condition

Abstract The scalar filtered density function (FDF) methodology is extended and employed for large eddy simulation (LES) of turbulent flows under supercritical condition. To describe real fluid behavior, the extended methodology incorporates the generalized heat and mass diffusion models along with real fluid thermodynamic relations which are derived using the cubic Peng–Robinson equation of state. These models are implemented within the stochastic differential equations comprising the scalar FDF transport. Simulations are conducted of a temporally developing mixing layer under supercritical condition and the results are assessed by comparing with data generated by direct numerical simulation (DNS) of the same layer. The consistency of the proposed FDF methodology is assessed. The LES-FDF predictions are shown to agree favorably with the DNS data and exhibit several key features pertaining to supercritical turbulent flows.

Energy & Fuels↗

Dramatic performance of Clostridium thermocellum explained by its wide range of cellulase modalities

Clostridium thermocellum is the most efficient microorganism for solubilizing lignocellulosic biomass known to date. Its high cellulose digestion capability is attributed to efficient cellulases consisting of both a free-enzyme system and a tethered cellulosomal system wherein carbohydrate active enzymes (CAZymes) are organized by primary and secondary scaffoldin proteins to generate large protein complexes attached to the bacterial cell wall. This study demonstrates that C. thermocellum also uses a type of cellulosomal system not bound to the bacterial cell wall, called the “cell-free” cellulosomal system. The cell-free cellulosome complex can be seen as a “long range cellulosome” because it can diffuse away from the cell and degrade polysaccharide substrates remotely from the bacterial cell. The contribution of these two types of cellulosomal systems in C. thermocellum was elucidated by characterization of mutants with different combinations of scaffoldin gene deletions. The primary scaffoldin, CipA, was found to play the most important role in cellulose degradation by C. thermocellum , whereas the secondary scaffoldins have less important roles. Additionally, the distinct and efficient mode of action of the C. thermocellum exoproteome, wherein the cellulosomes splay or divide biomass particles, changes when either the primary or secondary scaffolds are removed, showing that the intact wild-type cellulosomal system is necessary for this essential mode of action. As a result, this new transcriptional and proteomic evidence shows that a functional primary scaffoldin plays a more important role compared to secondary scaffoldins in the proper regulation of CAZyme genes, cellodextrin transport, and other cellular functions.

09 BIOMASS FUELS↗

Insights into PEMFC Performance Degradation from HCl in Air

The performance degradation of a proton exchange membrane fuel cell (PEMFC) is studied in the presence of HCl in the air stream. The cathode employing carbon-supported platinum nanoparticles (Pt/C) was exposed to 4 ppm HCl in air while the cell voltage was held at 0.6 V. The HCl poisoning results in generation of chloride and chloroplatinate ions on the surface of Pt/C catalyst as determined by a combination of electrochemical tests and ex-situ chlorine K-edge X-Ray absorption near-edge structure (XANES) spectroscopy. The chloride ions inhibit the oxygen reduction reaction (ORR) and likely affect the wetting properties of diffusion media/catalyst layer, while the chloroplatinate ions are responsible for enhanced platinum particle growth most likely due to platinum dissolution-redeposition. The chloride ions can cause corrosion of the Pt nanoparticles in the presence of aqueous HCl in air even if no potential is applied. Although the majority of chloride ions are desorbed from the Pt surface by hydrogen treatment of the cathode, they partially remain in the system and re-adsorb on platinum at cell voltages of 0.5-0.9 V. Chloride ions are removed from the system and fuel cell performance at 0.5-0.7 V is restored by multiple exposures to low potentials.

08 HYDROGEN↗

Carrier Transport and Related Effects in Detectors of the Cryogenic Dark Matter Search

The Cryogenic Dark Matter Search (CDMS) is searching for weakly-interacting massive particles (WIMPS), which could explain the dark matter problem in cosmology and particle physics. By simultaneously measuring signals from deposited charge and the energy in nonequilibrium phonons created by particle interactions in intrinsic germanium crystals at a temperature of 40 mK, a signature response for each event is produced. This response, combined with phonon pulse-shape information, allows CDMS to actively discriminate candidate WIMP interactions with nuclei from electromagnetic radioactive background which interacts with electrons. The challenges associated with these techniques are unique. Carrier scattering is dominated by the spontaneous emission of Luke-Neganov phonons due to zeropoint fluctuations of the lattice ions. Drift fields are maintained at only a few V/cm, else these emitted phonons would dominate the phonons of the original interaction. The dominant systematic issues with CDMS detectors are due to the effects of space charge accumulation. It has been an open question how space charge accrues, and by which of several potential recombination and ionization processes. In this work, we have simulated the transport of electrons and holes in germanium under CDMS conditions. We have implemented both a traditional Monte Carlo technique based on carrier energy, followed later by a novel Monte Carlo algorithm with scattering rates defined and sampled by vector momentum. This vector-based method provides for a full anisotropic simulation of carrier transport including free-fight acceleration with an anisotropic mass, and anisotropic scattering rates. With knowledge of steady state carrier dynamics as a function of applied field, the results of our Monte Carlo simulations allow us to make a wide variety of predictions for energy dependent processes for both electrons and holes. Such processes include carrier capture by charged impurities, neutral impurities, static dipoles, and capture forming "anion" (D - /A + ) states. We also generate predictions for impact ionization of shallow impurities and of impact "neutralization" of D - /A + states. We use measurements of carrier capture performed on CDMS detectors to validate a plausible model for electron and hole capture due to neutral shallow impurities and their charged D - /A + states. This model, along with carrier drift and diffusion parameters from Monte Carlo simulation, can be used as the foundation for simulations of space charge evolution in CDMS detectors, simultaneously solving continuity equations with Poisson's equation.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Initial Demonstration of New Griffin Technologies for Simulating the Running-In Phase of Pebble Bed Reactors

Griffin is a reactor multiphysics modeling application based on MOOSE (Multiphysics Object-Oriented Simulation Environment) and specifically targeting transient modeling of advanced reactors. Griffin has been used recently to model pebble-bed reactors for the Nuclear Regulatory Commission (NRC) Office of Nuclear Regulatory Research and the Advanced Reactor Technology program. This modeling work has focused thus far on the direct calculation of equilibrium cores. This report documents an initial demonstration of a new running-in simulation capability. The new running-in capability is verified using the existing direct equilibrium core calculation capability. A simplified pebble-bed reactor model is then used to demonstrate the running-in simulation capability. This demonstration shows that Griffin is able to simulate years of operation during the running-in phase efficiently with each depletion step taking only several seconds. Two new technologies are also presented in this report which have been developed in Griffin that will be essential for improved accuracy both of the direct equilibrium core computation and the new running-in simulation capability. The first technology is an online cross section generation capability specifically targeted for pebble-bed reactors. This will improve the accuracy of the depletion calculation as the cross sections are generated at the exact core status. This also avoids the difficult step of pre-generating a separate standalone multigroup cross section set. Secondly, a newly implemented discretization for discontinuous finite element method (DFEM) SN transport in cylindrical (RZ) coordinates, which can be solved efficiently using the existing SN sweep solver, is discussed and some results are shown demonstrating the usefulness of the additional accuracy transport provides over a diffusion approximation.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

On the Origin of Ultra High Energy Cosmic Rays

Turbulence-driven plasma accelerators produced by magnetized accretion disks around black holes are proposed as the mechanism mainly responsible for observed cosmic ray protons with ultra high energies 10{sup 19}-10{sup 21} eV. The magnetized disk produces a voltage comparable to these cosmic ray energies. Here we present a Poynting model in which this voltage provides all of the energy to create the jet-like structures observed to be ejected from accretion disks, and this voltage also accelerates ions to high energies at the top of the expanding structure. Since the inductive electric field E = -v x B driving expansion has no component parallel to the magnetic field B, ion acceleration requires plasma wave generation - either a coherent wave accelerator as recently proposed, or instability-driven turbulence. We find that turbulence can tap the full inductive voltage as a quasi-steady accelerator, and even higher energies are produced by transient events on this structure. We find that both MHD modes due to the current and ion diffusion due to kinetic instability caused by the non-Maxwellian ion distribution contribute to acceleration. We apply our results to extragalactic giant radiolobes, whose synchrotron emissions serve to calibrate the model, and we discuss extrapolating to other astrophysical structures. Approximate calculations of the cosmic ray intensity and energy spectrum are in rough agreement with data and serve to motivate more extensive MHD and kinetic simulations of turbulence that could provide more accurate cosmic ray and synchrotron spectra to be compared with observations. A distinctive difference from previous models is that the cosmic ray and synchrotron emissions arise from different parts of the magnetic structure, thus providing a signature for the model.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Separation of CO2 from flue gas using electrochemical cells

Past research with high temperature molten carbonate electrochemical cells has shown that carbon dioxide can be separated from flue gas streams produced by pulverized coal combustion for power generation. However, the presence of trace contaminants, i.e., sulfur dioxide and nitric oxides, will impact the electrolyte within the cell. If a lower temperature cell could be devised that would utilize the benefits of commercially-available, upstream desulfurization and denitrification in the power plant, then this CO2 separation technique can approach more viability in the carbon sequestration area. Recent work has led to the assembly and successful operation of a low temperature electrochemical cell. In the proof-of-concept testing with this cell, an anion exchange membrane was sandwiched between gas-diffusion electrodes consisting of nickel-based anode electrocatalysts on carbon paper. When a potential was applied across the cell and a mixture of oxygen and carbon dioxide was flowed over the wetted electrolyte on the cathode side, a stream of CO2 to O2 was produced on the anode side, suggesting that carbonate/ bicarbonate ions are the CO2 carrier in the membrane. Since a mixture of CO2 and O2 is produced, the possibility exists to use this stream in oxy-firing of additional fuel. From this research, a novel concept for efficiently producing a carbon dioxide rich effluent from combustion of a fossil fuel was proposed. Carbon dioxide and oxygen are captured from the flue gas of a fossilfuel combustor by one or more electrochemical cells or cell stacks. The separated stream is then transferred to an oxy-fired combustor which uses the gas stream for ancillary combustion, ultimately resulting in an effluent rich in carbon dioxide. A portion of the resulting flow produced by the oxy-fired combustor may be continuously recycled back into the oxy-fired combustor for temperature control and an optimal carbon dioxide rich effluent

Keywords: Carbon dioxide sequestration Carbon capt↗

FULLY CONVECTIVE MAGNETO-ROTATIONAL TURBULENCE IN LARGE ASPECT-RATIO SHEARING BOXES

We present a numerical study of turbulence and dynamo action in stratified shearing boxes with both finite and zero net magnetic flux. We assume that the fluid obeys the perfect gas law and has finite thermal diffusivity. The latter is chosen to be small enough so that vigorous convective states develop. The properties of these convective solutions are analyzed as the aspect ratio of the computational domain is varied and as the value of the mean field is increased. For the cases with zero net flux, we find that a well-defined converged state is obtained for large enough aspect ratios. In the converged state, the dynamo can be extremely efficient and can generate substantial toroidal flux. We identify solutions in which the toroidal field is mostly symmetric about the mid-plane and solutions in which it is mostly anti-symmetric. The symmetric solutions are found to be more efficient at transporting angular momentum and can give rise to a luminosity that is up to an order of magnitude larger than the corresponding value for the anti-symmetric states. In the cases with a finite net flux, the system appears to spend most of the time in the symmetric states.

79 ASTRONOMY AND ASTROPHYSICS↗

Reinvigorating Reverse-Osmosis Membrane Technology to Stabilize the V 2 O 5 Lithium-Ion Battery Cathode

V 2 O 5 is deemed as one of the most promising cathode materials for next-generation high-capacity lithium-ion batteries (LIBs). It possesses a theoretical capacity of 294 mAh g -1 , which is much higher than conventional cathodes. But, there are many issues to be solved before its practical use, including poor cycle life and unsatisfactory rate performance, mainly owing to its low electronic conductivity and ionic diffusivity, as well as structural instability. Our work reports three types of V 2 O 5 asymmetric membranes synthesized by using an adapted reverse-osmosis membrane technology combined with sol-gel chemistry, aiming to stabilize the cyclability and improve the rate performance. V 2 O 5 asymmetric membrane cathodes prepared using graphene as the conductive additives have a specific capacity of approximately 160 mAh g -1 at a current density of 100 mA g -1 with no capacity degradation after 380 cycles. It is also found that the annealing temperature and the choice of conductive additives can affect the morphology of V 2 O 5 nanoparticles and the overall electrode cyclability. Furthermore, we find that a lower annealing temperature (300 vs. 400 °C) and the addition of graphene are beneficial to long-term cycling performance.

25 ENERGY STORAGE↗

Bridging the Transition from Mesoscale to Microscale Turbulence in Numerical Weather Prediction Models

With a focus towards developing multiscale capabilities in numerical weather prediction models, the specific problem of the transition from the mesoscale to the microscale is investigated. For that purpose, idealized one-way nested mesoscale to large-eddy simulation (LES) experiments were carried out using the Weather Research and Forecasting model framework. It is demonstrated that switching from one-dimensional turbulent diffusion in the mesoscale model to three-dimensional LES mixing does not necessarily result in an instantaneous development of turbulence in the LES domain. However, very large fetches are needed for the natural transition to turbulence to occur. The computational burden imposed by these long fetches necessitates the development of methods to accelerate the generation of turbulence on a nested LES domain forced by a smooth mesoscale inflow. To that end, four new methods based upon finite amplitude perturbations of the potential temperature field along the LES inflow boundaries are developed, and investigated under convective conditions. Each method accelerated the development of turbulence within the LES domain, with two of the methods resulting in a rapid generation of production and inertial range energy content associated to microscales that is consistent with non-nested simulations using periodic boundary conditions. The cell perturbation approach, the simplest and most efficient of the best performing methods, was investigated further under neutral and stable conditions. Successful results were obtained in all the regimes, where satisfactory agreement of mean velocity, variances and turbulent fluxes, as well as velocity and temperature spectra, was achieved with reference non-nested simulations. In contrast, the non-perturbed LES solution exhibited important energy deficits associated to a delayed establishment of fully-developed turbulence. The cell perturbation method has negligible computational cost, significantly accelerates the generation of realistic turbulence, and requires minimal parameter tuning, with the necessary information relatable to mean inflow conditions provided by the mesoscale solution.

17 WIND ENERGY↗

Classification of machine learning frameworks for data-driven thermal fluid models

Thermal fluid processes are inherently multi-physics and multi-scale, involving mass-momentum-energy transport phenomena at multiple scales. Thermal fluid simulation (TFS) is based on solving conservative equations, for which – except for “first-principles” direct numerical simulation – closure relations (CRs) are required to provide microscopic interactions or so-called sub-grid-scale physics. In practice, TFS is realized through reduced-order modeling, and its CRs as low-fidelity models can be informed by observations and data from relevant and adequately evaluated experiments and high-fidelity simulations. This paper is focused on data-driven TFS models, specifically on their development using machine learning (ML). Five ML frameworks are introduced including physics-separated ML (PSML or Type I ML), physics-evaluated ML (PEML or Type II ML), physics-integrated ML (PIML or Type III ML), physics-recovered (PRML or Type IV ML), and physics-discovered ML (PDML or Type V ML). The frameworks vary in their performance for different applications depending on the level of knowledge of governing physics, source, type, amount and quality of available data for training. Notably, outlined for the first time in this paper, Type III models present stringent requirements on modeling, substantial computing resources for training, and high potential in extracting value from “big data” in thermal fluid research. Here the current paper demonstrates and investigates ML frameworks in three examples. First, we utilize the heat diffusion equation with a nonlinear conductivity model formulated by convolutional neural networks (CNNs) and feedforward neural networks (FNNs) to illustrate the applications of Type I, Type II, Type III, and Type V ML. The results indicate a preference for Type II ML under deficient data support. Type III ML can effectively utilize field data, potentially generating more robust predictions than Type I and Type II ML. CNN-based closures exhibit more predictability than FNN-based closures, but CNN-based closures require more training data to obtain accurate predictions. Second, we illustrate how to employ Type I ML and Type II ML frameworks for data-driven turbulence modeling using reference works. Third, we demonstrate Type I ML by building a deep FNN-based slip closure for two-phase flow modeling. The results show that deep FNN-based closures exhibit a bounded error in the prediction domain.

42 ENGINEERING↗