Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “surrogate approximator”

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
NASA NTRS1997

A Novel, Poly-Etalon, Fabry-Perot for Planetary Research

In an effort to develop a mechanically robust, high throughput and solid state spectrometer several liquid crystal Fabry-Perot etalons were constructed. The etalons were tested for spectral response, radiation resistance and optical transmission. The first year of this project was spent developing and understanding the properties of the liquid crystal etalons; in the second year an intensified all-sky imaging system was developed around a pair of LC etalons. The imaging system, developed jointly with SRI International represents a unique brassboard to demonstrate the use of LC etalons as tunable filters. The first set of etalons constructed in year one of this project were tested for spectral response and throughput while etalon surrogates were exposed to proton radiation simulating the exposure of an object in Low Earth Orbit (LEO). The 2" diameter etalons had a measure finesse of approximately 10 and were tunable over five orders. Liquid crystals exposed to proton irradiation showed no signs of damage. In year two two larger diameter (3") etalons were constructed with gaps of 3 and 5 microns. This pair of etalons is for use in a high resolution, all-sky spectral imager. The WATUMI imager system follows the heritage of all sky, narrow band, intensified imagers however it includes two LC Fabry-Perot etalons to provide tunability and the ability to switch wavelengths rapidly, an import consideration in auroral airglow imaging. This work also resulted in two publications and one poster presentation. The instrument will be uniquely capable, with superior throughput and speed, to measure optical airglow of multiple emission lines in harsh conditions.

Kerr, Robert B.↗
NASA NTRSDate not supplied

Probabilistic Calibration of Expensive Models using Efficiently Trained Surrogates

Calibration of computational models in the presence of uncertainty is often cast as a Bayesian inference problem and solved via sampling methods, e.g., Markov chain Monte Carlo. When the computational model is expensive, this task becomes intractable due to the large number of samples required to accurately estimate the posterior distribution of the calibration parameters. A popular solution to this problem is to use machine learning to develop a faster-to-evaluate, lower-fidelity substitute for the original model to serve as a surrogate while solving the inference problem. Although considered an offline cost, generating training data to construct this surrogate model can still be an expensive task in practice. An active learning algorithm is presented that focuses training on improving surrogate accuracy specifically in and around the bulk of the posterior distribution, as this is where the model is exercised during calibration. Candidate samples are drawn from families of distributions related to an approximation of the posterior. The sample maximizing predictive variance is then selected for evaluation by the original computational model, yielding a label for the training point. Iterating this approach increases efficiency relative to space filling designs (e.g., Latin hypercube sampling) by avoiding low probability points. Practical considerations are discussed, including the benefits of using a sequential Monte Carlo sampling approach, convergence heuristics, and the importance of both exploration and exploitation given that the true posterior is unknown a priori.

uncertainty quantification↗
DOE OSTI2026

Latent Twins

Over the past decade, scientific machine learning has transformed the development of mathematical and computational frameworks for analyzing, modeling, and predicting complex systems. From inverse problems to numerical partial differential equations (PDEs), dynamical systems, and model reduction, these advances have pushed the boundaries of what can be simulated. Yet they have often progressed in parallel, with representation learning and algorithmic solution methods evolving largely as separate pipelines. With Latent Twins, we propose a unifying mathematical framework that creates a hidden surrogate in latent space for the underlying equations. Whereas digital twins mirror physical systems in the digital world, Latent Twins mirror mathematical systems in a learned latent space governed by operators. Through this lens, classical modeling, inversion, model reduction, and operator approximation all emerge as special cases of a single principle. We establish the fundamental approximation properties of Latent Twins for both ordinary differential equations (ODEs) and PDEs and demonstrate the framework across three representative settings: (i) canonical ODEs, capturing diverse dynamical regimes; (ii) a PDE benchmark using the shallow-water equations, contrasting Latent Twin simulations with deep operator network and forecasts with a four-dimensional variational method baseline; and (iii) a challenging real-data geopotential reanalysis dataset, reconstructing and forecasting from sparse, noisy observations. Latent Twins provide a compact, interpretable surrogate for solution operators that evaluate across arbitrary time gaps in a single-shot, while remaining compatible with scientific pipelines such as assimilation, control, and uncertainty quantification. Looking forward, this framework offers scalable, theory-grounded surrogates that bridge data-driven representation learning and classical scientific modeling across disciplines.

Latent Twins↗
NASA NTRS2000

Uncertainty Analysis of Decomposing Polyurethane Foam

Sensitivity/uncertainty analyses are necessary to determine where to allocate resources for improved predictions in support of our nation's nuclear safety mission. Yet, sensitivity/uncertainty analyses are not commonly performed on complex combustion models because the calculations are time consuming, CPU intensive, nontrivial exercises that can lead to deceptive results. To illustrate these ideas, a variety of sensitivity/uncertainty analyses were used to determine the uncertainty associated with thermal decomposition of polyurethane foam exposed to high radiative flux boundary conditions. The polyurethane used in this study is a rigid closed-cell foam used as an encapsulant. Related polyurethane binders such as Estane are used in many energetic materials of interest to the JANNAF community. The complex, finite element foam decomposition model used in this study has 25 input parameters that include chemistry, polymer structure, and thermophysical properties. The response variable was selected as the steady-state decomposition front velocity calculated as the derivative of the decomposition front location versus time. An analytical mean value sensitivity/uncertainty (MV) analysis was used to determine the standard deviation by taking numerical derivatives of the response variable with respect to each of the 25 input parameters. Since the response variable is also a derivative, the standard deviation was essentially determined from a second derivative that was extremely sensitive to numerical noise. To minimize the numerical noise, 50-micrometer element dimensions and approximately 1-msec time steps were required to obtain stable uncertainty results. As an alternative method to determine the uncertainty and sensitivity in the decomposition front velocity, surrogate response surfaces were generated for use with a constrained Latin Hypercube Sampling (LHS) technique. Two surrogate response surfaces were investigated: 1) a linear surrogate response surface (LIN) and 2) a quadratic response surface (QUAD). The LHS techniques do not require derivatives of the response variable and are subsequently relatively insensitive to numerical noise. To compare the LIN and QUAD methods to the MV method, a direct LHS analysis (DLHS) was performed using the full grid and timestep resolved finite element model. The surrogate response models (LIN and QUAD) are shown to give acceptable values of the mean and standard deviation when compared to the fully converged DLHS model.

Hobbs, Michael L.↗
DOE OSTI2024

Sensitivity analysis, surrogate modeling, and optimization of pebble-bed reactors considering normal and accident conditions

This research provides a valuable tool that streamlines the optimization process while significantly increasing its accuracy. This study creates a robust framework for reactor design optimization by incorporating comprehensive modeling using the Comprehensive Reactor Analysis Bundle, or BlueCRAB, within the Multiphysics Object-Oriented Simulation Environment (MOOSE). BlueCRAB is the United States Nuclear Regulatory Commission's code suite for non-light water reactor analysis and includes the Griffin, Pronghorn, and Bison applications. This not only improves the efficiency of the optimization process but also enhances the reliability of the results. Such a tool is essential for advancing the state-of-the-art in pebble-bed reactor technology and is critical for achieving the goals of Generation IV reactors, which aim for safe, sustainable, and economically viable nuclear energy solutions. This work presents and applies this workflow on pebble-bed reactors while considering both normal and off-normal conditions. A representative gas-cooled pebble-bed reactor at equilibrium core conditions serves as the nominal design specification for normal operation and is based on previous research. The depressurized loss-of-forced-cooling accident is deployed for off-normal conditions in this work. After defining design-related parameters and quantities of interest regarding reactor safety and performance, this multiphysics model is sampled using the MOOSE stochastic tools module. The result is a comprehensive dataset of configurations, enabling sensitivity analysis and the generation of surrogate models. Subsequently, the dataset and surrogate models are employed in two optimization studies aimed at maximizing fuel utilization and economic profit while adhering to safety and operational constraints. Performing the optimization process with fuel utilization as the metric leads to an improvement of approximately 10%, compared to engineering-judgment-based nominal conditions. The optimization on economic profit leads to an estimated increase of ~300 million USD over the lifetime of the reactor.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Quantum Drude oscillators coupled with Coulomb potential as an efficient model for bonded and non-covalent interactions in atomic dimers

The quantum Drude oscillator (QDO) model has been widely used as an efficient surrogate to describe the electric response properties of matter as well as long-range interactions in molecules and materials. Most commonly, QDOs are coupled within the dipole approximation so that the Hamiltonian can be exactly diagonalized, which forms the basis for the many-body dispersion method [Phys. Rev. Lett. 108, 236402 (2012)]. The dipole coupling is efficient and allows us to study non-covalent many-body effects in systems with thousands of atoms. However, there are two limitations: (i) the need to regularize the interaction at short distances with empirical damping functions and (ii) the lack of multipolar effects in the coupling potential. In this work, we convincingly address both limitations of the dipole-coupled QDO model by presenting a numerically exact solution of the Coulomb-coupled QDO model by means of quantum Monte Carlo methods. We calculate the potential-energy surfaces of homogeneous QDO dimers, analyzing their properties as a function of the three tunable parameters: frequency, reduced mass, and charge. We study the coupled-QDO model behavior at short distances and show how to parameterize this model to enable an effective description of chemical bonds, such as the covalent bond in the H2 molecule.

Chemistry↗
NASA NTRS2004

CoSSIR: A New Instrument for Exploring the Utility of Submillimeter-wave Radiometry for Earth Observation

The Conical Scanning Submillimeter-wave Imaging Radiometer (CoSSIR) has been developed to study the application of submillimeter-wave radiometry for remote sensing of cirrus clouds and humidity sounding. Measurements of the global distribution of ice cloud mass and particle size are important for understanding the Earth s energy budget and for evaluating global climate models. The spatial variability and the wide variety of cloud particle shapes and sizes make ice clouds particularly difficult to measure. Ice clouds are essentially undetectable at microwave frequencies due to the low dielectric of ice and small size of the particles relative to wavelength. However, submillimeter wavelengths demonstrate significant response to the presence of ice clouds thus this frequency regime is applicable to measuring ice clouds. Another potentially viable application for submillimeter-wave radiometry is humidity and temperature sounding. The principle of sounding at submillimeter wavelengths is similar to that at microwavelengths. Submillimeter-wave radiometry has the advantage of achieving finer spatial resolution using a smaller antenna aperture which is an important consideration for spaceborne observatories. Submillimeter-wave radiometry also offers the potential of sounding over land and as a surrogate measurement for precipitation. CoSSIR is a new instrument to explore these applications. The CoSSIR is designed to fly aboard the ER-2 aircraft and its modest size (approximately 100 kg) permits it to be configured for other aircraft. A dual-axes gimbals mechanism provides conical, across-track, and along-track scanning capability. In its present configuration CoSSIR has fifteen channels between 183 GHz and 640 GHz. Three channels are centered about the 183 GHz water vapor absorption line, four channels are centered about the 380 GHz water vapor absorption line, and three dual-polarized channels are centered about the 487 GHz oxygen absorption line. Two channels are located in atmospheric windows at 220 GHz and 640 GHz. All channels are single-linear polarized with the exception of those near 487 GHz. Calibration is achieved by periodically observing two blackbody radiators; one blackbody is heated to approximately 325 K and the other is approximately 250 K during flight. Details of the instrument design as well as measurements from the Cirrus Regional Study of Tropical Anvils and Cirrus Layers - Florida Area Cirrus Experiment will be presented.

Racette, P. E.↗
DOE OSTI2025

Experimental investigation of unsteady flow dynamics in therminol-66 using edge tone-induced oscillations

Recent research has demonstrated that flow oscillations induced by a wedge positioned opposite the jet outlet in edge tones can be effectively utilized for flow measurement purposes. This highlights the necessity for developing advanced instrumentation capable of precise flow measurements in experimental setups exposed to radiation and in challenging fluids such as molten salts. Here, this study contributes to the existing body of knowledge by conducting experiments using Therminol-66, an effective surrogate for FLiBe under appropriate temperature conditions for nuclear applications, across a broad range of Reynolds numbers from 17,700 to 555,000. Experiments were conducted with Therminol-66 at approximately 368.8 K under atmospheric pressure, with varying flow rates. A reliable correlation was established between the generated frequencies, corresponding flow rates, and device geometry parameters. The data were subjected to non-dimensional analysis to reveal the interrelationships among flow rate, fluid properties, device geometry, and their impact on the acoustic tones produced by the edge tone device. These findings contribute to the development of precise flow measurement instrumentation for use in challenging environments.

46 - INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AN↗
NASA NTRS2004

Long-Lag, Wide-pulse Gamma-Ray Bursts

Currently, the best available probe of the early phase of gamma-ray burst (GRB) jet attributes is the prompt gamma-ray emission, in which several intrinsic and extrinsic variables determine GRB pulse evolution. Bright, usually complex bursts have many narrow pulses that are difficult to model due to overlap. However, the relatively simple, long spectral lag, wide-pulse bursts may have simpler physics and are easier to model. In this work we analyze the temporal and spectral behavior of wide pulses in 24 long-lag bursts, using a pulse model with two shape parameters - width and asymmetry - and the Band spectral model with three shape parameters. We find that pulses in long-lag bursts are distinguished both temporally and spectrally from those in bright bursts: the pulses in long spectral lag bursts are few in number, and approximately 100 times wider (10s of seconds), have systematically lower peaks in vF(v), harder low-energy spectra and softer high-energy spectra. We find that these five pulse descriptors are essentially uncorrelated for our long-lag sample, suggesting that at least approximately 5 parameters are needed to model burst temporal and spectral behavior. However, pulse width is strongly correlated with spectral lag; hence these two parameters may be viewed as mutual surrogates. We infer that accurate formulations for estimating GRB luminosity and total energy will depend on several gamma-ray attributes, at least for long-lag bursts. The prevalence of long-lag bursts near the BATSE trigger threshold, their predominantly low vF(v) spectral peaks, and relatively steep upper power-law spectral indices indicate that Swift will detect many such bursts.

Norris, J. P.↗
DOE OSTI2024

Data-scarce surrogate modeling of shock-induced pore collapse process

Understanding the mechanisms of shock-induced pore collapse is of great interest in various disciplines in sciences and engineering, including materials science, biological sciences, and geophysics. However, numerical modeling of the complex pore collapse processes can be costly. To this end, a strong need exists to develop surrogate models for generating economic predictions of pore collapse processes. Here, in this work, we study the use of a data-driven reduced-order model, namely dynamic mode decomposition, and a deep generative model, namely conditional generative adversarial networks, to resemble the numerical simulations of the pore collapse process at representative training shock pressures. Since the simulations are expensive, the training data are scarce, which makes training an accurate surrogate model challenging. To overcome the difficulties posed by the complex physics phenomena, we make several crucial treatments to the plain original form of the methods to increase the capability of approximating and predicting the dynamics. In particular, physics information is used as indicators or conditional inputs to guide the prediction. In realizing these methods, the training of each dynamic mode composition model takes only around 30 s on CPU. In contrast, training a generative adversarial network model takes 8 h on GPU. Moreover, using dynamic mode decomposition, the final-time relative error is around 0.3% in the reproductive cases. We also demonstrate the predictive power of the methods at unseen testing shock pressures, where the error ranges from 1.3 to 5% in the interpolatory cases and 8 to 9% in extrapolatory cases.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Characterization of a multi-kW, large-aperture gas-cooled Faraday rotator

We characterize a large-aperture gas-cooled Faraday rotator (FR) designed to mitigate stress-induced depolarization in high-energy, high-power laser systems. The rotator, based on ceramic TGG, was tested using a 150 mW probe beam at 1047 nm and a pump beam at 1070 nm. Compensating for the birefringence induced by the surrogate depolarization plate at 3.3 kW of effective pump power, the rotator reduces the spatially in-homogeneous depolarized light containing linear, circular, and elliptical polarization states to nearly linear with approximately 0.8% (or −21 dB) of residual depolarized energy. The FR demonstrated effective depolarization compensation across its 58×58 mm 2 aperture.

47 OTHER INSTRUMENTATION↗
DOE OSTI2025

Equation-Free Coarse Control of Distributed Parameter Systems via Local Neural Operators

The control of high-dimensional distributed parameter systems (DPS) remains a challenge when explicit coarse-grained equations are unavailable. Classical equation-free (EF) approaches rely on fine-scale simulators treated as black-box timesteppers. However, repeated simulations for steady-state computation, linearization, and control design are often computationally prohibitive, or the microscopic timestepper may not even be available, leaving us with data as the only resource. We propose a data-driven alternative that uses local neural operators, trained on spatiotemporal microscopic/mesoscopic data, to obtain efficient short-time solution operators. These surrogates are employed within Krylov subspace methods to compute coarse steady and unsteady-states, while also providing Jacobian information in a matrix-free manner. Krylov-Arnoldi iterations then approximate the dominant eigenspectrum, yielding reduced models that capture the open-loop slow dynamics without explicit Jacobian assembly. Both discrete-time Linear Quadratic Regulator (dLQR) and pole-placement (PP) controllers are based on this reduced system and lifted back to the full nonlinear dynamics, thereby closing the feedback loop.

93B52, 93C20, 47N70, 65J15, 65M32, 68T07, 68T20, 6↗
NASA NTRS2014

Raising Nuclear Thermal Propulsion (NTP) Technology Readiness Above 3

NTP development is currently supported by the NASA program office "Advanced Exploration Systems". The concept is a main propulsion option being considered for human missions to Mars in the 2030's. Major NTP development took place in the 1960's and 1970's under the Rover/NERVA program. The technology had matured to TRL 6 and was preparing to go to TRL 7 with a prototype flight engine before the program was cancelled. Over the last 40 years, a variety of continuations started, but only lasted a few years each. The Rover/NERVA infrastructure is almost all gone. The only remains are a few pieces of hardware, final reports and a few who worked the Rover/NERVA. Two types of nuclear fuel are being investigated to meet the current engine design specific impulse of 900 seconds compared to approximately 850 seconds demonstrated during Rover/NERVA. One is a continuation of composite fuel with new coatings to better control mid-band corrosion. The other type is a CERMET fuel made of Tungsten and UO2. Both fuels are being made from Rover/NERVA lessons learned, but with slightly different recipes to increase fuel endurance at higher operating temperatures. The technology readiness level (TRL) of these current modified reactor fuels is approximately TRL 3. To keep the development cost low and help mature the TRL level past 4 quickly, a few special non-nuclear test facilities have been made to test surrogate fuel, with depleted uranium, as coupons and full length elements. Both facilities utilize inductive heating and are licensed to handle depleted uranium. TRL 5 requires exposing the fuel to a nuclear environment and TRL 6 requires a prototype ground or flight engine system test. Currently, three different NTP ground test facility options are being investigated: exhaust scrubber, bore hole, and total exhaust containment. In parallel, a prototype flight demonstration test is also being studied. The first human mission to Mars in the 2030's is currently 2033. For an advanced propulsion concept to be seriously considered for use, the engine development plans need to show it is feasible and affordable to reach TRL 8 by 2027 and can be qualified for human mission use.

Gerrish, Harold P., Jr.↗
DOE OSTI2023

A COUPLED DETERMINISTIC TRANSPORT CALCULATION IN COMSOL USING PROPER ORTHOGONAL DECOMPOSITION

A coupled physics problem consisting of radiation transport and heat transfer was modeled using COMSOL Multiphysics to simulate a criticality accident. Reduced order models (ROM) were created to accelerate the radiation transport calculation through Proper Orthogonal Decomposition (POD). Additionally, on-the-fly neutron cross sections were generated from a non-linear function through a feed forward neural network. The Livelink for MATLAB module allowed for data transfer between an otherwise independent solver and in-house modules. A data driven approach to POD was developed to act as a surrogate model to couple with heat transfer, with the POD module written in MATLAB coupled to an isolated heat transfer solve in COMSOL. A coupled transient solution resulted in a good approximation of the flux and temperature, with the maximum error being 3E-2 and 2.5E-3 respectively.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2024

Surrogate Distributed Radiological Sources—Part II: Aerial Measurement Campaign

In this second part of a multipaper series, we present results from outdoor aerial measurements of surrogate distributed gamma-ray sources. Here we detail the design, manufacture, and testing of 300 individual ~7 mCi Cu-64 sealed sources at the Washington State University (WSU) research reactor and their deployment in various source patterns (each comprising up to 100 point sources) during the aerial measurement campaign. We show the results of two such measurements, in which approximate source shapes and qualitative source intensities can be seen from the count rate versus position plots, even without performing reconstructions. We also detail our efforts in ground-truthing the deployed sources and comparing measured gamma-ray data to model predictions. In particular, we compare measured versus expected count data using the Poisson deviance formalism of Part I to evaluate whether the fielded surrogate point-source arrays “look like” their truly continuous distributed source analogs. More generally, we find that the point-source array technique provides high source placement accuracy, relative ease of quantifying the true source configuration, scalability to source dimensions of ≲100 m, ease of reconfiguration and removal, and relatively low dose to personnel. Finally, we consider potential improvements and generalizations of the point-source array technique for future measurement campaigns.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2024

Sound propagation in realistic interactive 3D scenes with parameterized sources using deep neural operators

We address the challenge of acoustic simulations in three-dimensional (3D) virtual rooms with parametric source positions, which have applications in virtual/augmented reality, game audio, and spatial computing. The wave equation can fully describe wave phenomena such as diffraction and interference. However, conventional numerical discretization methods are computationally expensive when simulating hundreds of source and receiver positions, making simulations with parametric source positions impractical. To overcome this limitation, we propose using deep operator networks to approximate linear wave-equation operators. This enables the rapid prediction of sound propagation in realistic 3D acoustic scenes with parametric source positions, achieving millisecond-scale computations. By learning a compact surrogate model, we avoid the offline calculation and storage of impulse responses for all relevant source/listener pairs. Our experiments, including various complex scene geometries, show good agreement with reference solutions, with root mean squared errors ranging from 0.02 to 0.10 Pa. Notably, our method signifies a paradigm shift as—to our knowledge—no prior machine learning approach has achieved precise predictions of complete wave fields within realistic domains.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2023

Toward DMC Accuracy Across Chemical Space with Scalable Δ-QML

In the past decade, quantum diffusion Monte Carlo (DMC) has been demonstrated to successfully predict the energetics and properties of a wide range of molecules and solids by numerically solving the electronic many-body Schrödinger equation. With O(N 3 ) scaling with the number of electrons N, DMC has the potential to be a reference method for larger systems that are not accessible to more traditional methods such as CCSD(T). Assessing the accuracy of DMC for smaller molecules becomes the stepping stone in making the method a reference for larger systems. We show that when coupled with quantum machine learning (QML)-based surrogate methods, the computational burden can be alleviated such that quantum Monte Carlo (QMC) shows clear potential to undergird the formation of high-quality descriptions across chemical space. We discuss three crucial approximations necessary to accomplish this: the fixed-node approximation, universal and accurate references for chemical bond dissociation energies, and scalable minimal amons-set-based QML (AQML) models. Numerical evidence presented includes converged DMC results for over 1000 small organic molecules with up to five heavy atoms used as amons and 50 medium-sized organic molecules with nine heavy atoms to validate the AQML predictions. Finally, numerical evidence collected for Δ-AQML models suggests that already modestly sized QMC training data sets of amons suffice to predict total energies with near chemical accuracy throughout chemical space.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Prediction of vacancy defect diffusion paths in high entropy alloys via machine learning on molecular dynamics data

Identifying the diffusion path of point defects is a critical step in understanding their evolution and the mechanisms of related phenomena. Defect diffusion occurs at small length and time scales, with impacts on material properties that may continue to evolve over ns to μs, ms, and the continuum scale (s, min, etc., and cm, m, etc.). The time scale accessible to molecular dynamics (MD) simulations is limited by small step sizes, typically in the fs range. Thus, surrogate models of MD simulations through machine learning (ML)-based algorithms are of great interest, especially for complex systems such as high entropy alloys (HEAs). In this work, dynamics governing vacancy migration in HEA were approximated with graph convolutional network (GCN) models as ansatzes for kinetic Monte Carlo (KMC) rate catalogs. Network design considered that diffusion in crystalline solids generally depends on interactions between defects and their immediate neighbor atoms. Graphs represented the vacancy surroundings, MD-generated trajectories provided training and comparison datasets, and unsupervised GCN models approximated interatomic dynamics governing vacancy migration in HEAs as ansatzes for KMC. A proof-of-concept model trained on MD data for the Fe, Ni, Cr, Co, and Cu HEA environment was used with two different neighbor interactions to assess the feasibility of training a GCN to predict vacancy defect transition rates in the HEA environment. The resulting setup rapidly generated MD-formatted synthetic trajectories based on dynamics learned from the MD training set, with a time acceleration of roughly two orders of magnitude and a similar diffusion coefficient to MD observations. Additionally, Nudged Elastic Band (NEB) calculations were performed on randomly generated FeNiCrCoCu HEA structures to determine vacancy migration barriers across nearest-neighbor sites. Transition probabilities for each jump, categorized by atomic type, were extracted from these calculations. NEB-based and GCN-based approaches led to similar outcomes.

Reimer, C↗