Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “conditional random fields”

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 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2016

A well-posed and stable stochastic Galerkin formulation of the incompressible Navier–Stokes equations with random data

We present a well-posed stochastic Galerkin formulation of the incompressible Navier–Stokes equations with uncertainty in model parameters or the initial and boundary conditions. The stochastic Galerkin method involves representation of the solution through generalized polynomial chaos expansion and projection of the governing equations onto stochastic basis functions, resulting in an extended system of equations. A relatively low-order generalized polynomial chaos expansion is sufficient to capture the stochastic solution for the problem considered. We derive boundary conditions for the continuous form of the stochastic Galerkin formulation of the velocity and pressure equations. The resulting problem formulation leads to an energy estimate for the divergence. With suitable boundary data on the pressure and velocity, the energy estimate implies zero divergence of the velocity field. Based on the analysis of the continuous equations, we present a semi-discretized system where the spatial derivatives are approximated using finite difference operators with a summation-by-parts property. With a suitable choice of dissipative boundary conditions imposed weakly through penalty terms, the semi-discrete scheme is shown to be stable. Numerical experiments in the laminar flow regime corroborate the theoretical results and we obtain high-order accurate results for the solution variables and the velocity divergence converges to zero as the mesh is refined.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

Predicting Biomass Yields of Advanced Switchgrass Cultivars for Bioenergy and Ecosystem Services Using Machine Learning

The production of advanced perennial bioenergy crops within marginal areas of the agricultural landscape is gaining interest due to its potential to sustainably produce feedstocks for biofuels and bioproducts while also improving the sustainability and resilience of commodity crop production. However, predicting the biomass yields of this production system is challenging because marginal areas are often relatively small and spread around agricultural fields and are typically associated with various abiotic conditions that limit crop production. Machine learning (ML) offers a viable solution as a biomass yield prediction tool because it is suited to predicting relationships with complex functional associations. The objectives of this study were to (1) evaluate the accuracy of commonly applied ML algorithms in agricultural applications for predicting the biomass yields of advanced switchgrass cultivars for bioenergy and ecosystem services and (2) determine the most important biomass yield predictors. Datasets on biomass yield, weather, land marginality, soil properties, and agronomic management were generated from three field study sites in two U.S. Midwest states (Illinois and Iowa) over three growing seasons. The ML algorithms evaluated in the study included random forests (RFs), gradient boosting machines (GBMs), artificial neural networks (ANNs), K-neighbors regressor (KNR), AdaBoost regressor (ABR), and partial least squares regression (PLSR). Coefficient of determination (R 2 ) and mean absolute error (MAE) were used to evaluate the predictive accuracy of the tested algorithms. Results showed that the ensemble methods, RF (R 2 = 0.86, MAE = 0.62 Mg/ha), GBM (R 2 = 0.88, MAE = 0.57 Mg/ha), and GBM (R 2 = 0.78, MAE = 0.66 Mg/ha), were the most accurate in predicting biomass yields of the Independence, Liberty, and Shawnee switchgrass cultivars, respectively. This is in agreement with similar studies that apply ML to multi-feature problems where traditional statistical methods are less applicable and datasets used were considered to be relatively small for ANNs. Consistent with previous studies on switchgrass, the most important predictors of biomass yield included average annual temperature, average growing season temperature, sum of the growing season precipitation, field slope, and elevation. This study helps pave the way for applying ML as a management tool for alternative bioenergy landscapes where understanding agronomic and environmental performance of a multifunctional cropping system seasonally and interannually at the sub-field scale is critical.

09 BIOMASS FUELS↗
DOE OSTI2018

Using Machine Learning to Analyze Factors Determining Cycle-to-Cycle Variation in a Spark-Ignited Gasoline Engine

In this work, we have applied a machine learning (ML) technique to provide insights into the causes of cycle-to-cycle variation (CCV) in a gasoline spark-ignited (SI) engine. The analysis was performed on a set of large eddy simulation (LES) calculations of a single cylinder of a four-cylinder port-fueled SI engine. The operating condition was stoichiometric, without significant knock, at a load of 16 bar brake mean effective pressure (BMEP), at an engine speed of 2500 revolutions per minute. A total of 123 cycles was simulated. Of these, 49 were run in sequence, while 74 were run in parallel. For the parallel approach, each cycle is initialized with its own synthetic turbulent field to generate CCV, as part of another work performed by us. In the current work, we used 3D information from all 123 cycles to compute flame topology and pre-ignition flow-field metrics. We then evaluated correlations between these metrics, and peak cylinder pressure (PCP) employing an ML technique called random forest. The computed metrics form the inputs to the random forest model, and PCP is the output. This model captures the effect of all inputs, as well as interactions between them owing to its decision-tree structure. The goal of this work is to demonstrate (as a first step) that ML models can implicitly learn complex relationships between pre-ignition flow-fields, flame shapes, and the eventual outcome of the cycle (whether a cycle will be a high or a low cycle).

42 ENGINEERING↗
DOE OSTI2025

Failure Analysis–Informed Risk Assessment Framework for Geological Carbon Storage Using Numerical Simulation and Machine Learning

Geological carbon storage (GCS) is recognized as a critical technology for achieving large-scale reductions in anthropogenic carbon dioxide (CO 2 ) emissions. Ensuring long-term containment and safety requires robust risk assessment frameworks that account for geological uncertainty and identify potential failure scenarios. Among various indicators, the area of review (AoR) serves as a key metric for evaluating storage performance, regulatory compliance, and monitoring design, as it delineates the spatial extent impacted by pressure buildup and plume migration. However, conventional AoR-based risk assessments typically perturb parameters within narrow uncertainty bounds, potentially overlooking rare but high-impact events arising from extreme geological conditions. In this study, we present a failure analysis–informed risk assessment framework for large-scale GCS projects to improve site prescreening and monitoring design. A suite of 300 numerical simulations was generated using stochastic geological models that vary five key parameters: net-to-gross ratio, anisotropy azimuth, porosity multiplier, permeability multiplier, and vertical-to-horizontal permeability ratio. Among these, 200 realizations represent normal geological uncertainty, while 100 additional cases explore extreme yet plausible conditions for failure-case analysis. The AoR was simulated and computed from pressure and CO 2 saturation fields, where the baseline AoR boundary, representing the extent predicted under typical geological uncertainty, was defined as the union of 200 normal-range simulations, and failure was identified when extreme-range cases exceeded this baseline. Results show that incorporating broader parameter uncertainty produces significantly larger AoR extents, underscoring the potential underestimation of risk under conventional uncertainty ranges. Furthermore, spatial probability maps derived from failure-induced AoR exceedance identify regions requiring enhanced monitoring attention. Various machine learning (ML)–based classifiers were developed to predict failure occurrence from geological parameters, with the random forest model achieving the highest performance (F1-score of 0.986). Consistent findings from correlation coefficient, feature importance, and Sobol sensitivity analyses reveal that low net-to-gross ratios and permeability multipliers are the dominant risk drivers, reflecting reduced reservoir connectivity and limited pressure dissipation. Altogether, these results provide a novel framework for risk-informed site prescreening and monitoring design that explicitly considers rare but high-impact geological scenarios in GCS projects.

25 ENERGY STORAGE↗
DOE OSTI2021

On the origin of internal obstacles to dislocation glide in single-phase NiFe random alloys

Over the past decade, a new class of unconventional alloys, where several constitutive elements are presented in close to equal concentrations, is considered as promising structural materials in many applications. Properties and behavior of many concentrated solid solution alloys (CSSAs) under different conditions, including mechanical loading, differs from that of conventional alloys and has been the subject of intensive studies by different techniques. A key feature of these materials is that there is no clear distinction between solute and solvent species of atoms, which is the conventional approach to considering dislocation motion in alloys. To understand this regime, we have recently reported on the glide of a screw dislocation glide in a simple equiatomic NiFe alloy and reported the two-mode character. At low applied stress, the dislocation motion is similar to a glide through a field of obstacles. In this work, we report the results of a detailed study on the nature of these internal obstacles. We demonstrate that these are not conventional localized “obstacles” but is the result of significantly different energies of instantaneous dislocation configurations over the alloy bulk. The strong pinning positions are associated with composition changes in the partial dislocation cores, while observing no significant composition changes along the stacking faults. Local energies can vary significantly, as demonstrated using simulations of short-periodicity dislocation lines. Therefore, the critical stress to move a dislocation under such conditions is not a constant Peierls stress but represents a wide spectrum of barriers between different non-straight valleys that the dislocation line accommodates in the bulk at particular stress/strain conditions.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Observational benchmarks inform representation of soil organic carbon dynamics in land surface models

Abstract. Representing soil organic carbon (SOC) dynamics in Earth system models (ESMs) is a key source of uncertainty in predicting carbon–climate feedbacks. Machine learning models can help identify dominant environmental controllers and establish their functional relationships with SOC stocks. The resulting knowledge can be integrated into ESMs to reduce uncertainty and improve predictions of SOC dynamics over space and time. In this study, we used a large number of SOC field observations (n=54 000), geospatial datasets of environmental factors (n=46), and two machine learning approaches (namely random forest, RF, and generalized additive modeling, GAM) to (1) identify dominant environmental controllers of global and biome-specific SOC stocks, (2) derive functional relationships between environmental controllers and SOC stocks, and (3) compare the identified environmental controllers and predictive relationships with those in models used in Phase 6 of the Coupled Model Intercomparison Project (CMIP6). Our results showed that the diurnal temperature, drought index, cation exchange capacity, and precipitation were important observed environmental predictors of global SOC stocks. While the RF model identified 14 environmental factors that describe climatic, vegetation, and edaphic conditions as important predictors of global SOC stocks (R2=0.61, RMSE = 0.46 kg m−2), current ESMs oversimplify the relationships between environmental factors and SOC, with precipitation, temperature, and net primary productivity explaining > 96 % of the variability in ESM-modeled SOC stocks. Further, our study revealed notable disparities among the functional relationships between environmental factors and SOC stocks simulated by ESMs compared with observed relationships. To improve SOC representations in ESMs, it is imperative to incorporate additional environmental controls, such as the cation exchange capacity, and refine the functional relationships to align more closely with observations.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Empirical relationships between environmental factors and soil organic carbon produce comparable prediction accuracy as the Machine Learning

Accurate representation of environmental controllers of soil organic carbon (SOC) stocks in Earth System Model (ESM) land models could reduce uncertainties in future carbon-climate feedback projections. Using empirical relationships between environmental factors and SOC stocks to evaluate land models can help modelers understand prediction biases beyond what can be achieved with the observed SOC stocks alone. In this study, we used 31 observed environmental factors, field SOC observations (n = 6,213) from the continental US, and two Machine Learning approaches [Random Forest (RF) and Generalized Additive Modeling (GAM)] to (1) select important environmental predictors of SOC stocks, (2) derive empirical relationships between environmental factors and SOC stocks, and (3) use the derived relationships to predict SOC stocks and compare the prediction accuracy of simpler model developed with the machine learning predictions. Out of the 31 environmental factors we investigated, 12 were identified as important predictors of SOC stocks by the RF approach. In contrast, the GAM approach identified six (of those 12) environmental factors as important controllers of SOC stocks: potential evapotranspiration, normalized difference vegetation index, soil drainage condition, precipitation, elevation, and net primary productivity. The GAM approach showed minimal SOC predictive importance of the remaining six environmental factors identified by the RF approach. Our derived empirical relations produced comparable prediction accuracy as the GAM and RF approach using only a subset of environmental factors. The empirical relationships we derived using the GAM approach can serve as important benchmarks to evaluate environmental control representations of SOC stocks in ESMs, which could reduce uncertainty in predicting future carbon-climate feedbacks.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Co-Simulation of Electric Power Distribution Systems and Buildings including Ultra-Fast HVAC Models and Optimal DER Control

Smart homes and virtual power plant (VPP) controls are growing fields of research with potential for improved electric power grid operation. A novel testbed for the co-simulation of electric power distribution systems and distributed energy resources (DERs) is employed to evaluate VPP scenarios and propose an optimization procedure. DERs of specific interest include behind-the-meter (BTM) solar photovoltaic (PV) systems as well as heating, ventilation, and air-conditioning (HVAC) systems. The simulation of HVAC systems is enabled by a machine learning procedure that produces ultra-fast models for electric power and indoor temperature of associated buildings that are up to 133 times faster than typical white-box implementations. Hundreds of these models, each with different properties, are randomly populated into a modified IEEE 123-bus test system to represent a typical U.S. community. Advanced VPP controls are developed based on the Consumer Technology Association (CTA) 2045 standard to leverage HVAC systems as generalized energy storage (GES) such that BTM solar PV is better utilized locally and occurrences of distribution system power peaks are reduced, while also maintaining occupant thermal comfort. An optimization is performed to determine the best control settings for targeted peak power and total daily energy increase minimization with example peak load reductions of 25+%.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2016

Generation of quasi-monoenergetic protons from a double-species target driven by the radiation pressure of an ultraintense laser pulse

In laser-driven proton acceleration, generation of quasi-monoenergetic proton beams has been considered a crucial feature of the radiation pressure acceleration (RPA) scheme, but the required difficult physical conditions have hampered its experimental realization. As a method to generate quasi-monoenergetic protons under experimentally viable conditions, we investigated using double-species targets of controlled composition ratio in order to make protons bunched in the phase space in the RPA scheme. From a modified optimum condition and three-dimensional particle-in-cell simulations, we showed by varying the ion composition ratio of proton and carbon that quasi-monoenergetic protons could be generated from ultrathin plane targets irradiated with a circularly polarized Gaussian laser pulse. The proposed scheme should facilitate the experimental realization of ultrashort quasi-monoenergetic proton beams for unique applications in high field science.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

Detonation structure in the presence of mixture stratification using reaction-resolved simulations

Many investigations of detonation-based combustors have identified reactant mixture inhomogeneity as having a leading-order impact on wave dynamics and combustion efficiency. To examine this phenomenon in a simplified context, an array of two- and three-dimensional channel detonation simulations are conducted in the present work. The reactant mixture consists of stratified fuel and air, wherein the randomly distributed equivalence ratio field features a characteristic stratification length scale. Detailed chemical kinetics are implemented in an adaptive mesh refinement solution framework where the region near the shock front is resolved with Ο (100) cells per representative ZND induction length. The results show that in comparison to baseline cases with uniform reactant mixtures, reactant stratification has a marked impact on the detonation structure. Increasing the stratification length scale increases the size and irregularity of the detonation cells, yielding larger variations in wave speed. Triple point collisions in fuel-rich regions lead to local wave speeds above the notional mean CJ speed, but wave passage through inert regions causes the local wave speed and strength to diminish. Further, conditional statistics show that increasing the stratification length scale increases the variance in pressure and temperature in the primary reaction zone, as well as the variance in heat release over a range of mixture conditions. In addition to the reactant mixture, the impact of the boundary condition behind the detonation is also investigated. The results show that an inflow boundary condition acts to over-drive the wave, leading to higher peak pressures, smaller detonation cells, and increased reactant consumption. On the other hand, cases with a wall behind the wave exhibit weaker waves with lower peak pressures and heat release rates, as well as greater variance in conditional quantities. Comparisons between complementary two- and three-dimensional simulations show reasonable qualitative agreement in wave structure, speed, and conditional statistics.

42 ENGINEERING↗
DOE OSTI2018

Investigation of scour adjacent to submerged geotextiles used for shore protection

This study presents the results of an experimental investigation of morphology change in the vicinity of submerged geotextiles placed within the surf zone. The study was motivated by the emerging use of submerged geotextile tubes for shore protection, shoreline stabilization, and surf amenity enhancement and the need to understand the mechanisms responsible for scour in the vicinity of these structures to preserve their structural integrity and reduce their structural failure. A movable bed physical model experiment was conducted at the U.S. Army Engineer Research and Development Center’s Large-scale Sediment Transport Facility (LSTF) to develop empirical formulations to predict the mean scour depth adjacent to geotextiles under oblique wave-breaking conditions as a function of the maximum Keulegan-Carpenter, Shields, and Reynolds numbers. The observed scour in the vicinity of the geotextiles was also compared to a previous study of scour in the vicinity of submerged cylinders. Formulations developed by Cataño-Lopera and García (2006) relating the Keulegan-Carpenter, Shields, and Reynolds numbers to the scour depth were used to predict the scour observed during the LSTF experiment. Results show that the formulations of Cataño-Lopera and García (2006) over-predict the observed scour when calculated using the maximum Keulegan-Carpenter, Shields, and Reynolds numbers. New, modified expressions of Cataño-Lopera and García (2006) were developed for use in oblique random wave fields.

Gorton, Alicia M.↗
DOE OSTI2017

Overview of Nuclear Integration Activities for Mitigation of Radiation Loads to ITER Superconducting Magnets

The Nuclear Integration Unit (NIU) is a trans-departmental team from the ITER Organization and Domestic Agencies, responsible for all nuclear integration aspects of the facility and for ensuring that systems and components are suitable for their radiation environment and satisfy all related performance requirements. This is a very challenging task namely due to the large number of systems, their still evolving designs, the intricate and far-reaching impacts of these in nuclear responses everywhere, and the complex world-wide distribution of procurement responsibilities. A sound systems engineering approach that involves all relevant experts and stakeholders, and trades-off between different performance requirements in an optimum way, needs to be taken and sustained during design evolution and construction, always with the objective of final overall project success. One of the most important current occupations of the NIU is the evaluation of, and protection from, radiation conditions to superconducting magnets during plasma operations, and particularly nuclear heating. This requires neutronic analyses in massive computer resources, with state-of-the-art methods and complex, extensive computer models. The main ITER superconducting magnet systems subject to significant nuclear loads are the toroidal field coils (TFCs) and the poloidal field coils (PFCs). These are illustrated in Fig. 1. Nuclear responses in these depend mostly on the design of all in-vessel (blankets, manifolds, divertor, in-vessel coils, port plugs), vessel and thermal shield components. Sources to consider are the 500 MW DT burning plasma (14.1 MeV neutrons) and the activated water in the in-vessel cooling system (photons up to 7 MeV from N-16 and neutrons up to 3.5 MeV from N-17). This paper discusses a snapshot of the present understanding of nuclear heat conditions in both the TFC and PFC magnets. This includes integral values and distributions, as well as best estimates from computations and conservative upper estimates by consideration of systematic effects and random uncertainties. The computer models, methods and assumptions used for evaluation of these responses are described. Nuclear integration aspects and ongoing efforts to study and implement additional shielding and other mitigating measures are also discussed.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

AdS 3 wormholes from a modular bootstrap

In recent work we computed the path integral of three-dimensional gravity with negative cosmological constant on spaces which are topologically a torus times an interval. Here we employ a modular bootstrap to show that the amplitude is completely fixed by consistency conditions and a few basic inputs from gravity. This bootstrap is notably for an ensemble of CFTs, rather than for a single instance. We also compare the 3d gravity result with the Narain ensemble. The former is well-approximated at low temperature by a random matrix theory ansatz, and we conjecture that this behavior is generic for an ensemble of CFTs at large central charge with a chaotic spectrum of heavy operators.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

A review on recent machine learning applications for imaging mass spectrometry studies

Imaging mass spectrometry (IMS) is a powerful analytical technique widely used in biology, chemistry, and materials science fields that continue to expand. IMS provides a qualitative compositional analysis and spatial mapping with high chemical specificity. The spatial mapping information can be 2D or 3D depending on the analysis technique employed. Due to the combination of complex mass spectra coupled with spatial information, large high-dimensional datasets (hyperspectral) are often produced. Therefore, the use of automated computational methods for an exploratory analysis is highly beneficial. The fast-paced development of artificial intelligence (AI) and machine learning (ML) tools has received significant attention in recent years. These tools, in principle, can enable the unification of data collection and analysis into a single pipeline to make sampling and analysis decisions on the go. There are various ML approaches that have been applied to IMS data over the last decade. Here, in this review, we discuss recent examples of the common unsupervised (principal component analysis, non-negative matrix factorization, k-means clustering, uniform manifold approximation and projection), supervised (random forest, logistic regression, XGboost, support vector machine), and other methods applied to various IMS datasets in the past five years. The information from this review will be useful for specialists from both IMS and ML fields since it summarizes current and representative studies of computational ML-based exploratory methods for IMS.

47 OTHER INSTRUMENTATION↗
DOE OSTI2021

Estimating basis functions in massive fields under the spatial mixed effects model

Abstract Spatial prediction is commonly achieved under the assumption of a Gaussian random field by obtaining maximum likelihood estimates of parameters, and then using the kriging equations to arrive at predicted values. For massive datasets, fixed rank kriging using the expectation–maximization algorithm for estimation has been proposed as an alternative to the usual but computationally prohibitive kriging method. The method reduces computation cost of estimation by redefining the spatial process as a linear combination of basis functions and spatial random effects. A disadvantage of this method is that it imposes constraints on the relationship between the observed locations and the knots. We develop an alternative method that utilizes the spatial mixed effects model, but allows for additional flexibility by estimating the range of the spatial dependence between the observations and the knots via an alternating expectation conditional maximization algorithm. Experiments show that our methodology improves estimation without sacrificing prediction accuracy while also minimizing the additional computational burden of extra parameter estimation. The methodology is applied to a temperature dataset archived by the United States National Climate Data Center, with improved results over previous methodology.

Pazdernik, Karl↗
DOE OSTI2017

Dynamics of Nanoparticles in Entangled Polymer Solutions

The mean square displacement < r 2 > of nanoparticle probes dispersed in simple isotropic liquids and in polymer solutions is interrogated using fluorescence correlation spectroscopy and single-particle tracking (SPT) experiments. Probe dynamics in different regimes of particle diameter (d), relative to characteristic polymer length scales, including the correlation length (ξ), the entanglement mesh size (a), and the radius of gyration (R g ), are investigated. In simple fluids and for polymer solutions in which d >> R g , long-time particle dynamics obey random-walk statistics < r 2 >:t, with the bulk zero-shear viscosity of the polymer solution determining the frictional resistance to particle motion. In contrast, in polymer solutions with d < R g , polymer molecules in solution exert noncontinuum resistances to particle motion and nanoparticle probes appear to interact hydrodynamically only with a local fluid medium with effective drag comparable to that of a solution of polymer chain segments with sizes similar to those of the nanoparticle probes. Under these conditions, the nanoparticles exhibit orders of magnitude faster dynamics than those expected from continuum predictions based on the Stokes–Einstein relation. SPT measurements further show that when d > a, nanoparticle dynamics transition from diffusive to subdiffusive on long timescales, reminiscent of particle transport in a field with obstructions. This last finding is in stark contrast to the nanoparticle dynamics observed in entangled polymer melts, where X-ray photon correlation spectroscopy measurements reveal faster but hyperdiffusive dynamics. As a result, we analyze these results with the help of the hopping model for particle dynamics in polymers proposed by Cai et al. and, on that basis, discuss the physical origins of the local drag experienced by the nanoparticles in entangled polymer solutions.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2024

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↗