Search NASA⌕ Search

SEARCH · Search NASA

Results for “Fluid Network”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Coupled Investigation of Fracture Permeability Impact on Reservoir Stress and Seismic Slip Behavior (Final Technical Report)

Enhanced Geothermal Systems (EGS) produce clean energy by circulating fluid through hot rock deep underground and bringing that heat to the surface to generate electricity. For this process to work reliably, fluids must be able to move efficiently through networks of natural or engineered fractures in the rock. Enhancing and maintaining subsurface permeability over time is essential for sustainable energy production. However, fluid injection changes the underground temperature, pressure, rock stress, and chemistry, which can alter permeability and sometimes trigger earthquakes. Predicting these interconnected processes remains a key challenge. To address this, we combined high-temperature laboratory experiments with high-fidelity simulations to better understand how fractures in geothermal reservoirs evolve over time. Our experiments measured how fractures respond to stress, slip, slip rate, and chemical reactions under geothermal conditions. These data were integrated into coupled thermal-hydrological-mechanical-chemical and earthquake (THMC+E) models tailored to the Utah FORGE site. The validated modeling framework improves predictions of reservoir performance and seismic response and helps guide operational decisions. This work reduces technical risk and strengthens the scientific foundation needed to make geothermal energy a reliable and scalable clean energy resource.

15 GEOTHERMAL ENERGY↗
DOE OSTI2026

Monitoring of Liquid Metal Reactor Heater Zones with Recurrent Neural Network Learning of Temperature Time Series

Advanced high-temperature fluid reactors (ARs), such as sodium fast reactors (SFRs) and molten salt cooled reactors (MSCRs) utilize high-temperature fluids at ambient pressure. To melt the fluid during reactor startup and prevent fluid freezing during cooldown, the thermal–hydraulic systems of such ARs include heater zones consisting of specific heaters with controllers, temperature sensors, and thermal insulation. The failure of heater zones due to insulation material degradation or improper installation, resulting in parasitic heat losses, can lead to fluid freezing. The detection of faults using a heat-transfer model is difficult because of a lack of knowledge of the experimental details. Data-driven machine learning of heater zone temperature time series offers a viable alternative. In this study, we benchmarked the performance of recurrent neural networks (RNNs) in an analysis of heat-up transient temperature time series of heater zones installed on a liquid sodium vessel. The RNN models include long short-term memory (LSTM) and gated recurrent unit (GRU) networks, as well as their bi-directional variants, BiLSTM and BiGRU. Anomalous temperature points were designated using a percentile-based threshold applied to residual fluctuations in the detrended temperature time series. Additionally, the impact of the exponentially weighted moving average (EWMA) method on detection accuracy was examined. The RNN models’ performance was assessed using precision, recall, and F 1 score metrics. Results demonstrated that RNN models effectively detect anomalies in temperature time series with the best models for each heater zone achieving F 1 scores of over 93%. To explain the variations in RNN model performance across different heater zones, we used Kullback–Leibler (KL) divergence to quantify the relative entropy between training and testing data, and the Detrended Fluctuation Analysis (DFA) to assess long-range temporal correlations. For datasets with strong long-range correlations and minimal relative entropy between training and testing data, GRU is the best-performing model. When the data exhibits weaker long-term correlations and a significant relative entropy between training and testing distributions, BiGRU shows the best performance. For the data sets with intermediate values of both KL divergence and DFA, the best performance is obtained with LSTM and BiLSTM, respectively.

gated recurrent unit↗
DOE OSTI2026

Differentiable multiphase flow model for physics-informed machine learning in reservoir pressure management

Accurate subsurface reservoir pressure control is extremely challenging due to geological heterogeneity and multiphase fluid-flow dynamics. Predicting behavior in this setting relies on high-fidelity physics-based simulations that are computationally expensive. Yet, the uncertain, heterogeneous properties that control these flows make it necessary to perform many of these expensive simulations, which is often prohibitive. To address these challenges, we introduce a physics-informed machine learning workflow that couples a fully differentiable multiphase flow simulator, which is implemented in the DPFEHM framework with a convolutional neural network (CNN). The CNN learns to predict fluid extraction rates from heterogeneous permeability fields to enforce pressure limits at critical reservoir locations. By incorporating transient multiphase flow physics into the training process, our method enables more practical and accurate predictions for realistic injection-extraction scenarios compared to previous works. To speed up training, we pretrain the model on single-phase, steady-state simulations and then finetune it on full multiphase scenarios, which dramatically reduces the computational cost. We demonstrate that high-accuracy training can be achieved with fewer than three thousand full-physics multiphase flow simulations – compared to previous estimates requiring up to ten million. This drastic reduction in the number of simulations is achieved by leveraging transfer learning from much less expensive single phase simulations.

25 ENERGY STORAGE↗
DOE OSTI2024

Dataset for scientific paper "Simulated plant‑mediated oxygen input has strong impacts on fine‑scale porewater biogeochemistry and weak impacts on integrated methane fluxes in coastal wetlands", a modeling study based on field observation at the tidal salt marshes of the Parker River Estuary, Massachusetts, United States

This dataset is the raw and processed data for the paper "Simulated plant ‑ mediated oxygen input has strong impacts on fine ‑ scale porewater biogeochemistry and weak impacts on integrated methane fluxes in coastal wetlands". This study investigated how plant-mediated oxygen input affects subsurface biogeochemical reactions of organic carbon degradation and the resulting methane emissions of coastal wetlands by model simulation. We used the subsurface geochemical simulator PFLOTRAN for the modeling, which produced the simulated changes in porewater chemical substances and methane emissions over 10 days under different scenarios of plant-mediated oxygen input.Specifically, this dataset contains: 1) the input files for PFLOTRAN of all simulation runs conducted in this study. Those files are with an extension of ".in", containing information of the biogeochemical reaction network (stoichiometry, reaction rate, Monod constants, etc), fluid flow rate and oxygen concentration in the fluid which together simulated the plant-mediated oxygen input, the configuration of artificial reactions that simulated the methane fluxes, etc. The PFLOTRAN input files are text files, which can be opened by NotePad, but running these input files will require proper installation of PFLOTRAN (instruction: https://documentation.pflotran.org/user_guide/how_to/installation/installation.html). 2) the raw and processed model output from PFLOTRAN of all simulation runs, and 3) the python scripts used to process the raw model output, including random allocation of root cells, converting raw data into organized formats, calculating the methane fluxes based on the model output, data visualization, etc. The raw and processed model output from PFLOTRAN are in .spydata format, which can be viewed with Python. and 3) the python scripts for data processing and analysis are programming scripts, which can be opened with Python.This modeling work, in particular the model parameterization of root density and initial conditions of porewater concentrations of biogeochemical substances, was based on field measurements at the salt marsh of the Upper Parker River Estuary, Massachusetts, United States.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Do not forget the electrons: Extending moderately-sized nuclear networks for multidimensional hydrodynamic codes

Context.Nuclear networks are widely used coupled with hydrodynamical simulations of explosive scenarios to account for the change of nuclear species and energy generation rate due to nuclear reactions. In this way, there is a feedback mechanism between the hydrodynamical state and the nuclear processes. Unfortunately, the timescale of nuclear reactions is orders of magnitude smaller than the dynamical timescale that drives hydrodynamical simulations. Therefore, these nuclear networks are usually very small, reduced in most cases to a dozen elements, especially when simulations are carried out in more than one dimension. Aims.We present here an extended nuclear network, with 90 species, designed for being coupled with hydrodynamic simulations, which includes neutrons, protons, electrons, positrons, and the corresponding neutrino and anti-neutrino emission. This network is also coupled with temperature, making it extremely robust and, together with its size, unique of its kind. The inclusion of electron captures on free protons makes the network very appropriate for multidimensional studies of Type Ia supernova explosions, especially when the exploding object is a massive white dwarf. Methods.We perform several tests that are relevant to simulate explosive scenarios, such as Type Ia supernovae and core-collapse supernovae. We compare the results of the 90 nuclei network with a standardα-chain network with 14 elements to evaluate the differences in the energy generation rate. We also evaluate the relevance of including the electrons in the network in terms of generated yields and how it affects the pressure of a degenerate fluid such as that of white dwarfs. The results obtained with the 90-nuclei network have been verified with a much larger 2000-nuclei network built from REACLIB (WinNet), in terms of nuclear energy generation rate, pressure, and produced yields. Results.The results obtained with the proposed medium-sized network compare fairly well, to a few percent, with those computed withWinNetin scenarios reproducing the gross physical conditions of current Type Ia supernova explosion models. In those cases where the carbon and oxygen fuel ignites at high density, the high-temperature plateau typical of the nuclear statistical equilibrium regime is well defined and stable, allowing large integration time steps. We show that the inclusion of electron captures on free protons substantially improves the estimation of the electron fraction of the mixture. Therefore, the pressure is better determined than in networks where electron captures are excluded, which will ultimately lead to more reliable hydrodynamic models. Explosive combustion of helium at low density, occurring near the surface layer of a white dwarf, is also better described with the proposed network, which gives nuclear energy generation rates much closer toWinNetthan typical reduced alpha networks. Conclusions.A nuclear network withN= 90 species, including electrons, aimed at multidimensional calculations of supernova explosions is described and verified. The proposed network is suitable for the study of Type Ia supernova explosions because it provides better values of pressure and electron abundance than other existing networks with smaller or even a similar size but without including electron capture processes.

Astronomy & Astrophysics↗
DOE OSTI2026

Recurrent convolutional neural networks for modeling nonadiabatic dynamics of quantum-classical systems

Recurrent neural networks (RNNs) have recently been extensively applied to model the time evolution in fluid dynamics, weather predictions, and even chaotic systems due to their ability to capture temporal dependencies and sequential patterns in data. Here we present an RNN model based on convolutional neural networks for modeling the nonlinear nonadiabatic dynamics of hybrid quantum-classical systems. The dynamical evolution of the hybrid systems is governed by equations of motion for classical degrees of freedom and von Neumann equation for electrons. The Physics-Aware Recurrent Convolution (PARC) neural network structure incorporates a differentiator-integrator architecture that inductively models the spatiotemporal dynamics of generic physical systems. Here, we apply our RNN approach to learn the space-time evolution of a one-dimensional semiclassical Holstein model after an interaction quench. For shallow quenches (small changes in electron-lattice coupling), the deterministic dynamics can be accurately captured using a single-CNN-based recurrent network. In contrast, deep quenches induce chaotic evolution, making long-term trajectory prediction significantly more challenging. Nonetheless, we demonstrate that the PARC-CNN architecture can effectively learn the statistical climate of the Holstein model under deep-quench conditions.

Holstein model↗
DOE OSTI2025

CRN Modeling of Ammonia RQL Combustion using a Partially-Stirred Reactor Approach

Ammonia is a promising alternative to hydrogen with high energy density and favorable storage and transport characteristics. However low flammability and a propensity for high nitrogen oxide (NOx) emissions make direct utilization challenging. Recently, two-stage rich-quench-lean (RQL) combustion strategies have shown promise in achieving low NOx emissions with ammonia. In this approach, the rich stage serves to oxidize a portion of the fuel, while thermally decomposing as much of the remaining ammonia as possible, generating hydrogen. In the second (lean) stage, air is rapidly introduced, burning out the hydrogen and residual ammonia. Two-stage RQL combustion of ammonia has been investigated in the open literature both experimentally and numerically. In general, idealized chemical reactor network (CRN) models predict NOx concentrations below that of 2D/3D computational fluid dynamics models and experiments. The primary drivers of these discrepancies may be largely attributed to finite rate mixing non-adiabatic operation. The typical CRN model is comprised of a perfectly-stirred-reactor (PSR), followed by a plug-flow-reactor (PFR), meant to represent the flame, and post-flame zones, respectively. In the two-stage RQL approach two PSR-PFR networks are arranged sequentially, corresponding to the rich and lean stages, with secondary air injection in between. In the authors’ past work, this arrangement has demonstrated the significant sensitivity of exit NOx to the rich stage equivalence ratio, while the amount of secondary air injection was shown to be less critical. In this paper, the CRN model is extended to (1) include the impacts of heat loss and (2) utilize a partially-stirred-reactor (PaSR) approach to study the impacts of mixing on emissions performance. Varying amounts of heat loss are applied to the rich relaxation zone to understand emissions performance and changes to optimization of equivalence ratio and residence time. Premixed and non-premixed configurations are considered in the rich stage PaSR, with varying degrees of mixing intensity to study the interaction between mixing, transport, and kinetic timescales. Critically, the impact of mixing between hot products and secondary air injection is studied to understand practical injector needs. Results show unburnt ammonia leaving the rich stage as a primary contributor to NOx emissions – driven both by increased heat loss and reduced mixing rates. Furthermore, heat losses have shown to create conditions which are conducive to increased N2O formation in the lean stage. The results of this study will be considered in the context of developing optimized two-stage RQL combustors for ammonia..

advanced gas turbines↗
DOE OSTI2024

Asynchronous GPU-based DEM solver embedded in commercial CFD software with polyhedral mesh support

A novel graphical processing unit-based discrete element method solver is introduced to improve stability, performance, and provide seamless integration into commercial or open-source computational fluid dynamics software. A key innovation is eliminating a need for network communication between solvers, which was previously required for cross-platform coupling. This is accomplished by a direct coupling method that employs dynamic-linked libraries. Furthermore, the solver optimizes memory usage by streamlining the particle-cell search algorithm by eliminating the cells' searching grid. This ensures the solver is compatible with a wide range of mesh types, providing high geometric flexibility. The approach simplifies the simulation process by directly incorporating computational fluid dynamics mesh information into the discrete element method solver. The performance analysis indicates about sixteen times boost in computational speed compared to benchmark central processing unit-based solvers. Finally, the solver's compatibility with polyhedral meshes, a vital advantage for complex geometries, is tested against a referenced study regarding the simulation of an immersed-tube fluidized bed.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Chemical Reactor Network Modeling of Ammonia Rich-Quench-Lean Combustion Using a Partially Stirred Reactor Approach

Ammonia is a promising alternative to hydrogen with high energy density and favorable storage and transport characteristics. However, low flammability and a propensity for high nitrogen oxide (NO x ) emissions make direct utilization challenging. Recently, two-stage rich-quench-lean (RQL) combustion strategies have shown promise in achieving low NO x emissions with ammonia. In this approach, the rich stage serves to oxidize a portion of the fuel while thermally decomposing as much of the remaining ammonia as possible, generating hydrogen. In the second (lean) stage, air is rapidly introduced, burning out the hydrogen and residual ammonia. Two-stage RQL combustion of ammonia has been investigated in the open literature both experimentally and numerically. In general, idealized chemical reactor network (CRN) models predict NO x concentrations below those of 2D/3D computational fluid dynamics models and experiments. The primary drivers of these discrepancies may be largely attributed to finite rate mixing nonadiabatic operation. The typical CRN model is comprised of a perfectly-stirred-reactor (PSR), followed by a plug-flow-reactor (PFR), meant to represent the flame, and postflame zones, respectively. In the two-stage RQL approach two PSR-PFR networks are arranged sequentially, corresponding to the rich and lean stages, with secondary air injection in between. In the authors' past work, this arrangement has demonstrated the significant sensitivity of exit NO x to the rich stage equivalence ratio, while the amount of secondary air injection was shown to be less critical. In this paper, the CRN model is extended to (1) include the impacts of heat loss and (2) utilize a partially-stirred-reactor (PaSR) approach to study the impacts of mixing on emissions performance. Varying amounts of heat loss are applied to the rich relaxation zone to understand emissions performance and changes to optimization of equivalence ratio and residence time. Premixed and nonpremixed configurations are considered in the rich stage PaSR, with varying degrees of mixing intensity to study the interaction between mixing, transport, and kinetic timescales. Critically, the impact of mixing between hot products and secondary air injection is studied to understand practical injector needs. Results show unburnt ammonia leaving the rich stage as a primary contributor to NO x emissions – driven both by increased heat loss and reduced mixing rates. Furthermore, heat losses have been shown to create conditions that are conducive to increased N 2 O formation in the lean stage. In conclusion, the results of this study will be considered in the context of developing optimized two-stage RQL combustors for ammonia.

Combustion↗
DOE OSTI2024

Design and assembling status of the Mu2e electromagnetic calorimeter mechanical structures

The muon-to-electron conversion (Mu2e) experiment at Fermilab will search for the Charged Lepton Flavour Violating neutrino-less coherent conversion of a muon into an electron in the field of an aluminum nucleus. The observation of this process would be the unambiguous evidence of physics beyond the Standard Model. The detector has been designed as a state-of-the-art crystal calorimeter and employs 1348 pure Cesium Iodide (CsI) crystals readout by UV-extended silicon photosensors and fast front-end and digitization electronics. A design consisting of two identical annular matrices (named disks ) positioned at the relative distance of 70 cm, downstream the aluminum target along the muon beamline, satisfies the Mu2e physics requirements. The hostile Mu2e operational conditions, in terms of radiation levels (total ionizing dose of 12 krad and a neutron fluence of 5 10^{10} n/cm2 @ 1 MeVeq (Si)/y), magnetic field intensity (1 T) and vacuum level (10−4 Torr) have posed tight constraints on the design of the detector mechanical structures and materials choice. The support structure of the two 674 crystal matrices employs two aluminum hollow rings and parts made of open- cell vacuum-compatible carbon fiber. The photosensors and service front-end electronics for each crystal are assembled in a unique mechanical unit inserted in a machined copper holder. The 674 units are supported by a machined plate made of vacuum-compatible plastic material. The plate also integrates the cooling system made of a network of copper lines flowing a low temperature radiation-hard fluid and placed in thermal con- tact with the copper holders. The data acquisition electronics is hosted in aluminum custom crates positioned on the external lateral surface of the two disks. The crates also integrate the electronics cooling system. In this poster we will review the constraints on the calorimeter mechanical structures, the technological choices, and the status of assembling at Fermilab.

Pasciuto, D.↗
DOE OSTI2025

Deep Learning–Assisted Multiobjective Optimization of Geological CO 2 Storage Performance under Geomechanical Risks

In geological CO 2 storage, designing the optimal well control strategy for CO 2 injection to maximize CO 2 storage while minimizing the associated geomechanical risks is not trivial. This challenge arises due to pressure buildup, CO 2 plume migration, the highly nonlinear nature of geomechanical responses to rock-fluid interaction, and the high computational cost associated with coupled flow and geomechanics simulations. In this paper, we introduce a novel optimization framework to address these challenges. The optimization problem is formulated as follows: maximize total CO 2 storage while minimizing geomechanical risks by adjusting the injection schedules within bounded constraints. The geomechanical risks are primarily driven by injection-induced pressure build-up, which is characterized by ground displacement and the induced microseismicity. We used the Fourier neural operator (FNO)-based deep learning model to construct surrogate models, replacing the time-consuming coupled flow and geomechanics simulations for evaluating the aforementioned objective functions. The developed surrogate models have been incorporated into a multiobjective optimization framework through a genetic algorithm to reduce the computational burden. The proposed optimization framework reduces the computational cost from approximately 2,400 hours, when using objective function evaluations based on physics-based simulations, to around 20 minutes. A set of Pareto-optimal solutions of the proposed workflow yields nontrivial optimal decisions, reducing the microseismicity potential and the vertical displacement. This Pareto front highlights the optimal trade-offs between CO 2 storage amount, safety, and ground displacement, emphasizing the need for careful optimization and management of injection strategies to achieve a balanced outcome. The novelty of this work is twofold. First, we demonstrate the importance of incorporating the minimization of the geomechanical risks as objective functions into the CO 2 storage optimization workflow to mitigate the potential risk of induced microseismicity and ground displacement. Second, we leverage the FNO-based surrogate models to optimize a real-field CO 2 storage operation.

42 ENGINEERING↗
DOE OSTI2024

Symbolic diagnostics to interpret and analyze neural network models

Embedded machine-learned models (EMLMs) have the promise to improve the predictive accuracy of engineering simulators in environments of national interest. EMLMs often comprise complex input-output maps (e.g., neural networks), which make them unamenable to rigorous analysis and generally difficult to interpret. In the face of decades of theory, this lack of interpretability is a significant barrier to building confidence in these models. This work outlines an approach to interpret EMLMs using sparse polynomial regression for comparison with theoretical understanding. To do so, we build on the concept of Locally Interpretable Model-agnostic Explanations (LIME) using physics-informed clustering, prototype selection, and library construction. While general, we demonstrate our method on tensor-basis neural networks used in Reynolds-Averaged Navier-Stokes simulations of hypersonic fluid flows. Results are presented for a simulated toy model and for direct numerical simulations (DNS) of turbulent flows over a flat plate.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Fracture Network Quantification for Enhanced Reservoir Characterization in CO2 Storage Sites

Presentation slides on “Fracture Network Quantification for Enhanced Reservoir Characterization in CO2 Storage Sites” for the CCUS 2025 conference held in Houston, Texas March 3-5, 2025. Accurate reservoir characterization is fundamental for the safe and efficient operation of carbon capture, utilization, and storage (CCUS) projects. This process is not only critical during initial site selection but also plays a key role in continuously updating reservoir geomodels throughout injection operations to effectively manage subsurface storage sites. Our primary objective is to develop a comprehensive understanding of fracture networks within CO2 injection reservoirs, which are critical for predicting the behavior of injected fluids, ensuring storage integrity post-injection, and mitigating risks such as leakage or induced seismicity.

artifical intelligence / machine learning (AI/ML)↗
DOE OSTI2025

Heterogeneous Dynamics in Shear Thickening Colloids Revealed by Intrinsic Heterodyne Correlation Spectroscopy

Shear induced frictional networks have been proposed to be responsible for the emergence of discontinuous shear thickening (DST) in complex fluids. However, little experimental evidence exists to support this model directly. Here, in this study, using x-ray photon correlation spectroscopy (XPCS), we show the existence of an intrinsic heterodyne feature during shear cessation, which originates from the relative motion of mobile particles against an aggregated or jammed network induced by shear thickening. Upon removing the shear, the shear stress dissipates rather quickly in a two-step fashion, whereas the heterogeneous particle dynamics persist much longer with the relative velocity decaying slowly with time as 𝑡 −1 . More importantly, both continuous shear thickening (CST) and DST show similar heterodyne features, indicating the intrinsic mechanisms causing shear thickening are similar in nature

Complex fluid↗
DOE OSTI2025

Transformers and Long Short-Term Memory Transfer Learning for GenIV Reactor Temperature Time Series Forecasting

Automated monitoring of the coolant temperature can enable autonomous operation of generation IV reactors (GenIV), thus reducing their operating and maintenance costs. Automation can be accomplished with machine learning (ML) models trained on historical sensor data. However, the performance of ML usually depends on the availability of large amount of training data, which is difficult to obtain for GenIV, as this technology is still under development. We propose the use of transfer learning (TL), which involves utilizing knowledge across different domains, to compensate for this lack of training data. TL can be used to create pre-trained ML models with data from small-scale research facilities, which can then be fine-tuned to monitor GenIV reactors. In this work, we develop pre-trained Transformer and long short-term memory (LSTM) networks by training them on temperature measurements from thermal hydraulic flow loops operating with water and Galinstan fluids at room temperature at Argonne National Laboratory. The pre-trained models are then fine-tuned and re-trained with minimal additional data to perform predictions of the time series of high temperature measurements obtained from the Engineering Test Unit (ETU) at Kairos Power. The performance of the LSTM and Transformer networks is investigated by varying the size of the lookback window and forecast horizon. The results of this study show that LSTM networks have lower prediction errors than Transformers, but LSTM errors increase more rapidly with increasing lookback window size and forecast horizon compared to the Transformer errors.

LSTM↗
DOE OSTI2024

Iodine speciation in basaltic melts at depth

The speciation of iodine in basalts has been investigated by combining in situ X-ray diffraction at high pressures and temperatures up to 4.9 GPa and 1600 °C, and Raman spectroscopy on recovered high pressure glasses at ambient conditions. Both methods point to iodine being oxidized in basalts, whether molten or quenched as glasses. Observed interatomic distances and Raman vibrational modes are consistent with iodine being dissolved as complex iodate groups alike polyiodates or periodates, not as IO$^-_3$ groups. Iodine speciation in basalts therefore seems to reflect a trend amongst halogens, with lighter chlorine bonding to network modifying cations, and bromine changing affinity from network modifying cations to oxygen anions under pressure. In the absence of a fluid aqueous phase, iodine could thus reach the Earth’s surface in basaltic magmas as an oxide, not as a reduced species.

58 GEOSCIENCES↗
DOE OSTI2025

Learning to Simulate Aerosol Dynamics with Graph Neural Networks

Aerosol effects on climate, weather, and air quality depend on characteristics of individual particles, which are tremendously diverse and change in time. Particle-resolved models are the only models able to capture this diversity in particle physiochemical properties, and these models are computationally expensive. As a strategy for accelerating particle-resolved microphysics models, we introduce Graph-based Learning of Aerosol Dynamics (GLAD) and use this model to train a surrogate of the particle-resolved model PartMC-MOSAIC. GLAD implements a Graph Network-based Simulator (GNS), a machine learning framework that has been used to simulate particle-based fluid dynamics models. In GLAD, each particle is represented as a node in a graph, and the evolution of the particle population over time is simulated through learned message passing. Here, we demonstrate our GNS approach on a simple aerosol system that includes condensation of sulfuric acid onto particles composed of sulfate, black carbon, organic carbon, and water. A graph with particles as nodes is constructed, and a graph neural network (GNN) is then trained using the model output from PartMC-MOSAIC. The trained GNN can then be used for simulating and predicting aerosol dynamics over time. Results demonstrate the framework's ability to accurately learn chemical dynamics and generalize across different scenarios, achieving efficient training and prediction times. We evaluate the performance across four scenarios, highlighting the framework's robustness and adaptability in modeling aerosol microphysics and chemistry.

aerosol chemistry dynamics↗