Search NASA⌕ Search

SEARCH · Search NASA

Results for “multicomponent”

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 739 records · Page 41

The role of reaction affinity and secondary minerals in regulating chemical weathering rates at the Santa Cruz Soil Chronosequence, California

In order to explore the reasons for the apparent discrepancy between laboratory and field weathering rates and to determine the extent to which weathering rates are controlled by the approach to thermodynamic equilibrium, secondary mineral precipitation and flow rates, a multicomponent reactive transport model (CrunchFlow) was used to interpret soil profile development and mineral precipitation and dissolution rates at the 226 ka marine terrace chronosequence near Santa Cruz, CA. Aqueous compositions, fluid chemistry, transport, and mineral abundances are well characterized (White et al., 2008, GCA) and were used to constrain the reaction rates for the weathering and precipitating minerals in the reactive transport modeling. When primary mineral weathering rates are calculated with either of two experimentally determined rate constants, the nonlinear, parallel rate law formulation of Hellmann and Tisser and [2006] or the aluminum inhibition model proposed by Oelkers et al. [1994], modeling results are consistent with field-scale observations when independently constrained clay precipitation rates are accounted for. Experimental and field rates, therefore, can be reconciled at the Santa Cruz site. Observed maximum clay abundances in the argillic horizons occur at the depth and time where the reaction fronts of the primary minerals overlap. The modeling indicates that the argillic horizon at Santa Cruz can be explained almost entirely by weathering of primary minerals and in situ clay precipitation accompanied by undersaturation of kaolinite at the top of the profile. The rate constant for kaolinite precipitation was also determined based on model simulations of mineral abundances and dissolved Al, SiO{sub 2}(aq) and pH in pore waters. Changes in the rate of kaolinite precipitation or the flow rate do not affect the gradient of the primary mineral weathering profiles, but instead control the rate of propagation of the primary mineral weathering fronts and thus total mass removed from the weathering profile. Our analysis suggests that secondary clay precipitation is as important as aqueous transport in governing the amount of dissolution that occurs within a profile because clay minerals exert a strong control over the reaction affinity of the dissolving primary minerals. The modeling also indicates that the weathering advance rate and the total mass of mineral dissolved is controlled by the thermodynamic saturation of the primary dissolving phases plagioclase and K-feldspar, as is evident from the difference in propagation rates of the reaction fronts for the two minerals despite their very similar kinetic rate laws.

54 ENVIRONMENTAL SCIENCES↗

Effect of subsurface soil moisture variability and atmospheric conditions on methane gas migration in shallow subsurface

We report a major concern resulting from the increased use and production of natural gas has been how to mitigate fugitive greenhouse gas emissions (predominantly methane) from natural gas infrastructure (e.g., leaky shallow pipelines). Subsurface migration and atmospheric loading of methane from pipeline leakage is controlled by source configurations and subsurface soil conditions (e.g., soil heterogeneity and soil moisture) and are further affected by atmospheric conditions (e.g., wind and temperature). However, the transport and attenuation of methane under varying subsurface and atmospheric conditions are poorly understood, making it difficult to estimate leakage fluxes from methane concentration measurements at and above the soil surface. Based on a series of controlled bench-scale experiments using a large porous media tank interfaced with an open-return wind tunnel, this study investigated multiphase processes controlling migration of methane from a point source representing a buried pipeline leaking at fixed flow rate (kg/s) under various saturation and soil-texture conditions. In addition, potential effects of atmospheric boundary controls, wind (0.5 and 2.0 m s -1 ) and temperature (22 and 35 °C), were also examined. Results showed the distinct effects of soil heterogeneity and, to a varying degree, of soil moisture on surface methane concentrations. In addition, results also showed the pronounced effects of wind and, to a lesser degree, of temperature on surface methane concentrations in the presence of varying soil and moisture conditions. The observed subsurface methane profiles were simulated using the multiphase transport simulator TOUGH2-EOS7CA. Observed agreement between measured and simulated data demonstrates that for the conditions studied, multiphase migration of a multicomponent gas mixture (including methane) under density-dependent flow can be adequately represented with a Fickian advection-diffusion (or dispersion) model (ADM) framework. The dominant effect of saturation over the soil texture, could also be inferred from numerical characterization.

09 BIOMASS FUELS↗

TOUGH-RFPA: Coupled thermal-hydraulic-mechanical Rock Failure Process Analysis with application to deep geothermal wells

This paper presents a coupled thermal-hydraulic-mechanical (THM) model for simulating failure processes in geological media. The coupling between stress/strain and heat transfer describing how thermal expansion affects stress is based on the theory of thermo-elasticity. Additionally, the coupling between stress/strain and fluid flow describing how pore-pressure affects stress is governed by Biot's consolidation theory. Stress/strain-dependent porosity/permeability/capillary-pressure defines the influence of stress on fluid flow and thermal fields. The THM model is implemented in the solver module that links the multiphase, multicomponent fluid flow simulator TOUGH2 with the geomechanical simulator RFPA (Rock Failure Process Analysis), which is based on the concept that heterogeneity leads to non-linearity and progressive failure behavior. Capability of TOUGH-RFPA simulator was verified by two benchmark problems related to hydraulic-mechanical and thermal-mechanical response in a hollow cylinder. Then the applicability of TOUGH-RFPA was demonstrated for modeling cooling-induced damage around a very deep geothermal well with mesoscopic heterogeneous properties, which showed complex damage evolution and fracturing during well cooling and fluid pressurization.

42 ENGINEERING↗

Targeting the Y-box Binding Protein-1 Axis to Overcome Radiochemotherapy Resistance in Solid Tumors

Multifunctional Y-box binding protein-1 (YB-1) is highly expressed in different human solid tumors and is involved in various cellular processes. DNA damage is the major mechanism by which radiochemotherapy (RCT) induces cell death. On induction of DNA damage, a multicomponent signal transduction network, known as the DNA damage response, is activated to induce cell cycle arrest and initiate DNA repair, which protects cells against damage. YB-1 regulates nearly all cancer hallmarks described to date by participating in DNA damage response, gene transcription, mRNA splicing, translation, and tumor stemness. YB-1 lacks kinase activity, and p90 ribosomal S6 kinase and AKT are the key kinases within the RAS/mitogen-activated protein kinase and phosphoinositide 3-kinase pathways that directly activate YB-1. Thus, the molecular targeting of ribosomal S6 kinase and AKT is thought to be the most effective strategy for blocking the cellular function of YB-1 in human solid tumors. In this review, after describing the prosurvival effect of YB-1 with a focus on DNA damage repair and cancer cell stemness, clinical evidence will be provided indicating an inverse correlation between YB-1 expression and the treatment outcome of solid tumors after RCT. In the interest of being concise, YB-1 signaling cascades will be briefly discussed and the current literature on YB-1 posttranslational modifications will be summarized. Finally, the current status of targeting the YB-1 axis, especially in combination with RCT, will be highlighted.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Interdiffusion and Reaction between Zr and Al Alloys from 425 degrees to 625 degrees C

Zirconium has recently garnered attention for use as a diffusion barrier between U–Mo nuclear fuels and Al cladding alloys. Interdiffusion and reactions between Zr and Al, Al-2 wt.% Si, Al-5 wt.% Si or AA6061 were investigated using solid-to-solid diffusion couples annealed in the temperature range of 425 degrees to 625 degrees C. In the binary Al and Zr system, the Al3Zr and Al2Zr phases were identified, and the activation energy for the growth of the Al3Zr phase was determined to be 347 kJ/mol. Negligible diffusional interactions were observed for diffusion couples between Zr vs. Al-2 wt.% Si, Al-5 wt.% Si and AA6061 annealed at or below 475 degrees C. In diffusion couples with the binary Al–Si alloys at 560 degrees C, a significant variation in the development of the phase constituents was observed including the thick t1 (Al5SiZr2) with Si content up to 12 at.%, and thin layers of (Si,Al)2Zr, (Al,Si)3Zr, Al3SiZr2 and Al2Zr phases. The use of AA6061 as a terminal alloy resulted in the development of both T1 (Al5SiZr2) and (Al,Si)3Zr phases with a very thin layer of (Al,Si)2Zr. At 560 degrees C, with increasing Si content in the Al–Si alloy, an increase in the overall rate of diffusional interaction was observed; however, the diffusional interaction of Zr in contact with multicomponent AA6061 with 0.4–0.8 wt.% Si was most rapid.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Effect of alloying elements on the $γ$’ antiphase boundary energy in Ni-base superalloys

Mechanical and fatigue performance of γ-γ' Ni-base superalloys are strongly affected by the antiphase boundary energy (APBE) of the γ' precipitates which, in turn, is dictated by the alloy's composition. Due to the multicomponent character of these alloys, establishing composition-APBE relationships are challenging, even though the qualitative effect of individual solutes on the APBE may be known. This work attempts to utilize density functional theory-based cluster expansion calculations to systematically assess the effect of composition on the APBE of the γ' phases in Ni-base superalloys. We aim to elucidate the influence of not only one single element but also multiple coexisting alloying elements on the γ' APBE. By explicit consideration of configurational disorder via Monte Carlo sampling, the effect of temperature on the APBE has also been analyzed. Overall, this work reveals that (1) effects of individual solute element M on the APBE energy obtained in an isolated, ternary condition (i.e. in Ni 3-x Al 1-y M x+y ) does not directly translate to a multi-solute case and that (2) the mutual synergistic interactions among different solute elements are not negligible. Based on the present results, an empirical master equation that predicts the APBE based on the composition of the γ' phases has been obtained.

36 MATERIALS SCIENCE↗

Uncertainty quantification of a deep learning fuel property prediction model

Deep learning models are being widely used in the field of combustion. Given the black-box nature of typical neural network based models, uncertainty quantification (UQ) is critical to ensure the reliability of predictions as well as the training datasets, and for a principled quantification of noise and its various sources. Deep learning surrogate models for predicting properties of chemical compounds and mixtures have been recently shown to be promising for enabling data-driven fuel design and optimization, with the ultimate goal of improving efficiency and lowering emissions from combustion engines. In this study, UQ is performed for a multi-task deep learning model that simultaneously predicts the research octane number (RON), Motor Octane Number (MON), and Yield Sooting Index (YSI) of pure components and multicomponent blends. The deep learning model is comprised of three smaller networks: Extractor 1, Extractor 2, and Predictor, and a mixing operator. The molecular fingerprints of individual components are encoded via Extractor 1 and Extractor 2, the mixing operator generates fingerprints for mixtures/blends based on linear mixing operation, and the predictor maps the fingerprint to the target properties. Two different classes of UQ methods, Monte Carlo ensemble methods and Bayesian neural networks (BNNs), are employed for quantifying the epistemic uncertainty. Combinations of Bernoulli and Gaussian distributions with DropConnect and DropOut techniques are explored as ensemble methods. All the DropConnect, DropOut and Bayesian layers are applied to the predictor network. Aleatoric uncertainty is modeled by assuming that each data point has an independent uncertainty associated with it. The results of the UQ study are further analyzed to compare the performance of BNN and ensemble methods. Although this study is confined to UQ of fuel property prediction, the methodologies are applicable to other deep learning frameworks that are being widely used in the combustion community.

33 ADVANCED PROPULSION SYSTEMS↗

Solid-solid phase transition via the liquid in a Pd 43 Cu 27 Ni 10 P 20 bulk metallic glass under conventional conditions

The microscopic mechanism of solid-solid phase transitions is a long-standing fundamental issue in materials science. Here we directly report the experimental evidence of the existence of an intermediate liquid during the solid-solid phase transition in a bulk Pd 43 Cu 27 Ni 10 P 20 alloy by in situ high temperature high energy X-ray diffraction, combined with the differential scanning calorimetry at slow heating rates down to 5 K/min. We elucidate that this intermediate liquid is attributed to the melting of the low-melting-point crystalline phases, rather than induced by the elastic stress energy along the interface proposed in literatures. This could also be a general feature in multicomponent alloys if a low-melting-point crystalline phase is formed together with other crystalline phases during heating. Furthermore, all results obtained here are scientifically sound and provide new insight into the phase transition theory, which will trigger more studies on the microscopic mechanism of solid-solid phase transition via an intermediate liquid.

36 MATERIALS SCIENCE↗

Revisit the VEC criterion in high entropy alloys (HEAs) with high-throughput ab initio calculations: A case study with Al-Co-Cr-Fe-Ni system

Valence electron concentration (VEC) was treated as a useful parameter to predict the stability of solid solution phases. However, the available experimental data to support this criterion is far from enough. In the current study, the high-throughput ab initio modeling is applied to investigate the relative stability of FCC and BCC single crystals of the Al-Co-Cr-Fe-Ni high entropy alloys (HEAs) by using the special quasi-random structure (SQS) approach. Furthermore, the predictions start with pure elements of the Al-Co-Cr-Fe-Ni system and are continued with binaries, ternaries, and quaternary compositions, which come up with 180 compositions (360 structures). After that, the reliability of the VEC criterion is testified. The results show that the VEC criterion not only works for the stable structure but also works effectively for metastable structure when both FCC and BCC are not thermodynamic stable. However, it is found that the old VEC criterion proposed by Guo et al. fails to work effectively for compositions containing high concentrations of light-weight metals such as Al at VEC<5. To solve this problem, the present work proposes a new VEC rule to define the stability of FCC and BCC structures at the ground state. With the implementation of the new VEC rule, the effectiveness of the VEC rule (EVEC) of both FCC and BCC structures is enhanced, especially for pure elements and binary compositions, indicating that this rule does not only work effectively for multicomponent systems but also works for low-order systems.

36 MATERIALS SCIENCE↗

Heterogeneous phase transformation pathways in additively manufactured Al-Ce-Mn alloys

Heat treatment of additively manufactured Al-Ce based multicomponent alloys leads to complex microstructure evolution. In this research, the ability to extend the phase transformation theories involving nucleation of a product phase from a heterogeneous multi-phase microstructure typical to that of additively manufactured samples is explored. The Al-10Ce-8Mn (wt%) was used as a model alloy system. Under additive manufacturing conditions different solidification microstructures were obtained due to spatial and temporal variations of thermal gradients (G) and liquid-solid interface velocities (R) within a given melt pool. Near the melt pool boundary (high G and low R, referred as MPB region), initially, Al 20 Mn 2 Ce forms from the liquid followed by a eutectic of FCC Al and Al 11 Ce 3 . In the melt pool interiors (low G and high R referred as ES region) a eutectic structure between FCC Al and Al 20 Mn 2 Ce is observed. During subsequent heat treatments, the MPB and ES regions transform into different sets of microstructures. In the MPB region, a fine globular microstructure containing FCC Al, Al 11 Ce 3 , Al 6 Mn, and Al 12 Mn results from the decomposition of Al 20 Mn 2 Ce. In the ES region a faceted Al 51 Mn 7 Ce 4 plate phase results from the decomposition of Al 20 Mn 2 Ce. The formation of the Al 51 Mn 7 Ce 4 phase within the eutectic microstructure at the boundaries of FCC Al and Al 20 Mn 2 Ce has not been reported in the literature. Further, these two distinct phase transformation pathways are rationalized based on the role of driving force on the nucleation of (Al 6 Mn) and/or metastable intermetallic (Al 51 Mn 7 Ce 4 ) phases at the interface of aluminum (FCC) and the non-equilibrium intermetallic (Al 20 Mn 2 Ce) phases.

36 MATERIALS SCIENCE↗

Electron and ion transport equations in computational weakly-ionized plasmadynamics

A new set of ion and electron transport equations is proposed to simulate steady or unsteady quasi-neutral or non-neutral multicomponent weakly-ionized plasmas through the drift–diffusion approximation. The proposed set of equations is advantaged over the conventional one by being considerably less stiff in quasi-neutral regions because it can be integrated in conjunction with a potential equation based on Ohm's law rather than Gauss's law. The present approach is advantaged over previous attempts at recasting the system by being applicable to plasmas with several types of positive ions and negative ions and by not requiring changes to the boundary conditions. Several test cases of plasmas enclosed by dielectrics and of glow discharges between electrodes show that the proposed equations yield the same solution as the standard equations but require 10 to 100 times fewer iterations to reach convergence whenever a quasi-neutral region forms. Further, several grid convergence studies indicate that the present approach exhibits a higher resolution (and hence requires fewer nodes to reach a given level of accuracy) when ambipolar diffusion is present. Because the proposed equations are not intrinsically linked to specific discretization or integration schemes and exhibit substantial advantages with no apparent disadvantage, they are generally recommended as a substitute to the fluid models in which the electric field is obtained from Gauss's law as long as the plasma remains weakly-ionized and unmagnetized.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Modeling weakly-ionized plasmas in magnetic field: A new computationally-efficient approach

Despite its success at simulating accurately both non-neutral and quasi-neutral weakly-ionized plasmas, the drift-diffusion model has been observed to be a particularly stiff set of equations. Recently, it was demonstrated that the stiffness of the system could be relieved by rewriting the equations such that the potential is obtained from Ohm's law rather than Gauss's law while adding some source terms to the ion transport equation to ensure that Gauss's law is satisfied in non-neutral regions. Although the latter was applicable to multicomponent and multidimensional plasmas, it could not be used for plasmas in which the magnetic field was significant. This paper hence proposes a new computationally-efficient set of electron and ion transport equations that can be used not only for a plasma with multiple types of positive and negative ions, but also for a plasma in magnetic field. Because the proposed set of equations is obtained from the same physical model as the conventional drift-diffusion equations without introducing new assumptions or simplifications, it results in the same exact solution when the grid is refined sufficiently while being more computationally efficient: not only is the proposed approach considerably less stiff and hence requires fewer iterations to reach convergence but it yields a converged solution that exhibits a significantly higher resolution. The combined faster convergence and higher resolution is shown to result in a hundredfold increase in computational efficiency for some typical steady and unsteady plasma problems including non-neutral cathode and anode sheaths as well as quasi-neutral regions.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

A finite-volume HLLC-based scheme for compressible interfacial flows with surface tension

Shock waves are often used in experiments to create a shear flow across liquid droplets to study secondary atomization. Similar behavior occurs inside of supersonic combustors (scramjets) under startup conditions, but it is challenging to study these conditions experimentally. In order to investigate this phenomenon further, a numerical approach is developed to simulate compressible multiphase flows under the effects of surface tension forces. The flow field is solved via the compressible multicomponent Euler equations (i.e., the five equation model) discretized with the finite volume method on a uniform Cartesian grid. The solver utilizes a total variation diminishing (TVD) third-order Runge–Kutta method for time-marching and second order TVD spatial reconstruction. Surface tension is incorporated using the Continuum Surface Force (CSF) model. Fluxes are upwinded with a modified Harten–Lax–van Leer Contact (HLLC) approximate Riemann solver. An interface compression scheme is employed to counter numerical diffusion of the interface. The present work includes modifications to both the HLLC solver and the interface compression scheme to account for capillary force terms and the associated pressure jump across the gas–liquid interface. A simple method for numerically computing the interface curvature is developed and an acoustic scaling of the surface tension coefficient is proposed for the non-dimensionalization of the model. The model captures the surface tension induced pressure jump exactly if the exact curvature is known and is further verified with an oscillating elliptical droplet and Mach 1.47 and 3 shock-droplet interaction problems. The general characteristics of secondary atomization at a range of Weber numbers are also captured in a series of simulations.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

LPSE : A 3-D wave-based model of cross-beam energy transfer in laser-irradiated plasmas

The new component of the “laser–plasma simulation environment” (LPSE) described here is a practical numerical model that solves the coupled vector equations for the propagation of nearly monochromatic, coherent, lectromagnetic waves in inhomogeneous unmagnetized plasmas. It operates efficiently in the numerically challenging semiclassical regime where characteristic plasma scale lengths are many times greater than the wavelength of the light and solutions are highly oscillatory. Solutions can be obtained in one, two, or three spatial dimensions and time. Additionally, the model includes the effects of nonlinear coupling of electromagnetic waves to the low-frequency plasma perturbations (i.e., ion-acoustic response) that are responsible for stimulated Brillouin scattering. Induced plasma perturbations are assumed to be imposed on a prescribed large-scale inhomogeneous background that includes spatially varying plasma density and flow. Our code is directly relevant to the problem of cross-beam energy transfer in laser-driven inertial confinement fusion. Lastly, it may also be applicable in other areas where eikonal solutions of multicomponent wave equations (or coupled wave equations) are insufficient, such as optical scattering from ultrasound, electron dynamics in quantum devices or in nanoscale light–matter interactions.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Assessment of diffuse-interface methods for compressible multiphase fluid flows and elastic-plastic deformation in solids

This work describes three diffuse-interface methods for the simulation of immiscible, compressible multiphase fluid flows and elastic-plastic deformation in solids. The first method is the localized-artificial-diffusivity approach of Cook, Subramaniam et al., and Adler and Lele, in which artificial diffusion terms are added to the individual phase mass fraction transport equations and are coupled with the other conservation equations. The second method is the gradient-form approach that is based on the quasi-conservative method of Shukla et al., in which the diffusion and sharpening terms (together called regularization terms) are added to the individual phase volume fraction transport equations and are coupled with the other conservation equations. The third approach is the divergence-form approach that is based on the fully conservative method of Jain et al., in which the regularization terms are added to the individual phase volume fraction transport equations and are coupled with the other conservation equations. In the present study, all three diffuse-interface methods are used in conjunction with a four-equation, multicomponent mixture model, in which pressure and temperature equilibria are assumed among the various phases. The primary objective of this work is to compare these three methods in terms of their ability to: maintain constant interface thickness throughout the simulation; conserve mass, momentum, and energy; and maintain accurate interface shape for long-time integration. The second objective of this work is to consistently extend these methods to model interfaces between solid materials with strength. To assess and compare the methods, they are used to simulate a wide variety of problems, including (1) advection of an air bubble in water, (2) shock interaction with a helium bubble in air, (3) shock interaction and the collapse of an air bubble in water, and (4) Richtmyer–Meshkov instability of a copper–aluminum interface. The current work focuses on comparing these methods in the limit of relatively coarse grid resolution, which illustrates the true performance of these methods. In conclusion, this is because it is rarely practical to use hundreds of grid points to resolve a single bubble or drop in large-scale simulations of engineering interest.

97 MATHEMATICS AND COMPUTING↗

A new re-redistribution scheme for weighted state redistribution with adaptive mesh refinement

State redistribution (SRD) is a recently developed technique for stabilizing cut cells that result from finite-volume embedded boundary methods. SRD has been successfully applied to a variety of compressible and incompressible flow problems. When used in conjunction with adaptive mesh refinement (AMR), additional steps are needed to preserve the accuracy and conservation properties of the solution if the embedded boundary is not restricted to a single level of the mesh hierarchy. In this work, we extend the weighted state redistribution algorithm to cases where cut cells live at or near a coarse-fine interface within the domain. Here, we present numerical results that demonstrate that the algorithm is conservative when the coarse-fine interface intersects the embedded boundary. Additionally we compare the numerical solution of the Sod shock tube problem in an inclined cylinder with the analytic solution, and we compare the simulation of a shock hitting a cylindrical obstacle with experimental data. Finally we demonstrate the methodology for simulation of the multicomponent compressible Navier-Stokes equations in a piston-bowl geometry, and discuss the computational efficiency gained by not requiring the entire embedded boundary to be defined at the finest level.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

GrainGNN: A dynamic graph neural network for predicting 3D grain microstructure

We propose GrainGNN, a surrogate model for the evolution of polycrystalline grain structure under rapid solidification conditions in metal additive manufacturing. High fidelity simulations of solidification microstructures are typically performed using multicomponent partial differential equations (PDEs) with moving interfaces. The inherent randomness of the PDE initial conditions (grain seeds) necessitates ensemble simulations to predict microstructure statistics, e.g., grain size, aspect ratio, and crystallographic orientation. Here, currently such ensemble simulations are prohibitively expensive and surrogates are necessary.In GrainGNN, we use a dynamic graph to represent interface motion and topological changes due to grain coarsening. We use a reduced representation of the microstructure using hand-crafted features; we combine pattern finding and altering graph algorithms with two neural networks, a classifier (for topological changes) and a regressor (for interface motion). Both networks have an encoder-decoder architecture; the encoder has a multi-layer transformer long-short-term-memory architecture; the decoder is a single layer perceptron.We evaluate GrainGNN by comparing it to high-fidelity phase field simulations for in-distribution and out-of-distribution grain configurations for solidification under laser power bed fusion conditions. GrainGNN results in 80%–90% pointwise accuracy; and nearly identical distributions of scalar quantities of interest (QoI) between phase field and GrainGNN simulations compared using Kolmogorov-Smirnov test. GrainGNN's inference speedup (PyTorch on single x86 CPU) over a high-fidelity phase field simulation (CUDA on a single NVIDIA A100 GPU) is 150×–2000× for 100-initial grain problem. Further, using GrainGNN, we model the formation of 11,600 grains in 220 seconds on a single CPU core.

36 MATERIALS SCIENCE↗