Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “transfer learning.”

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 595 records · Page 33
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Federated Learning with Frequency Estimation for Smart Meter Systems

Federated learning (FL) is a powerful framework that enables multiple distributed clients to collaborate without the need to transfer their data to a central server. However, FL does not inherently guarantee the level of privacy that clients often require. In our review of recent studies on privacy-enhancing techniques in FL, we found that frequency estimation (FE) methods remain underexplored. To address this gap, we developed and integrated FE techniques on the client side, further examining the effects of incorporating an adaptive range and a shuffled model. We also analyzed the impact of varying hyper-parameters on privacy preservation. Our results provide clear guidance on the algorithms and configurations that are most effective for enhancing privacy in FL, particularly when using long short-term memory (LSTM) architectures.

Kotevska, Olivera [ORNL] (ORCID:0000000316772243)↗
DOE OSTI2026

Stakeholder-guided holistic, Adaptive Framework for enhancing community Energy Resilience (SAFER) (Final Technical Report)

The Stakeholder-guided holistic, Adaptive Framework for enhancing community Energy Resilience (SAFER) project advances resilience science and engineering by addressing challenges in rural Kansas communities where aging infrastructure, extreme weather, and socioeconomic disparities heighten vulnerability to energy disruptions. Traditional approaches often focus on technical performance while overlooking community concerns and priorities. SAFER responds by integrating community perspectives with advanced analytical frameworks to create a holistic model for measuring and improving resilience. Project objectives included developing novel resilience metrics, advancing modeling frameworks that capture interdependencies across infrastructures, and embedding community-centric indicators directly into planning processes for distributed energy resources. The key technical innovations included the creation of self-organizing map (SOM)-based indices for objective resilience quantification, hetero-functional graph theory (HFGT) models linking power, water, transportation, and community assets, and graph neural network (GNN) tools for identifying critical nodes in complex systems. Community-centric energy planning was demonstrated through optimal siting and sizing of (photovoltaic) PV and battery storage, ensuring resilience enhancements also addressed energy burden and energy insecurity. SAFER engaged community partners in Dodge City and Ford County through surveys, focus groups, and workshops, generating more than 600 responses that established baseline measures of energy burden, financial insecurity, and willingness-to-pay to avoid outages. This data, organized in terms of a community capitals framework, informed the development of weighted reliability indices that better reflect community costs than traditional utility metrics. SAFER’s GNN-based critical node identification framework identified expert-labelled critical nodes with over 99% accuracy, while also uncovering additional functionalities essential for proactive resilience planning. The project’s models demonstrated that optimal PV and storage deployment could improve resilience indices by over 11 percent, with dispatch strategies further enhancing outcomes, confirming both the technical effectiveness and economic feasibility of these approaches. Through its combined emphasis on rigorous modeling, community-focused planning, and community engagement, SAFER advances the state of resilience research while delivering direct benefits to rural communities. The project provides tools, guidelines, and resilience heatmaps that help utilities, local governments, and residents better anticipate disruptions, prioritize investments, and strengthen the capacity to withstand and recover from energy-related hazards. Furthermore, the developed HFG and GNN frameworks are designed for transferability, allowing them to be adapted for resilience planning in other communities with minimal retraining. This inductive learning capability provides a scalable pathway to extend the SAFER project’s impact. Thus, creating a foundation for a nationally applicable model of infrastructure resilience. Additionally, the HFG can also be extended to include other FEMA community lifelines.

14 SOLAR ENERGY↗
DOE OSTI2023

Tree-Based Ensemble Learning Models for Wall Temperature Predictions in Post-Critical Heat Flux Flow Regimes at Subcooled and Low-Quality Conditions

Accurately predicting post-critical heat flux (CHF) heat transfer is an important but challenging task in water-cooled reactor design and safety analysis. Although numerous heat transfer correlations have been developed to predict post-CHF heat transfer, these correlations are only applicable to relatively narrow ranges of flow conditions due to the complex physical nature of the post-CHF heat transfer regimes. In this paper, a large quantity of experimental data is collected and summarized from the literature for steady-state subcooled and low-quality film boiling regimes with water as the working fluid in vertical tubular test sections. In addition, a low-quality water film boiling (LWFB) database is consolidated with a total of 22,813 experimental data points, which cover a wide flow range of the system pressure from 0.1 to 9.0 MPa, mass flux from 25 to 2750 kg/m 2 s, and inlet subcooling from 1 to 70 °C. Two machine learning (ML) models, based on random forest (RF) and gradient boosted decision tree (GBDT), are trained and validated to predict wall temperatures in post-CHF flow regimes. The trained ML models demonstrate significantly improved accuracies compared to conventional empirical correlations. To further evaluate the performance of these two ML models from a statistical perspective, three criteria are investigated and three metrics are calculated to quantitatively assess the accuracy of these two ML models. For the full LWFB database, the root-mean-square errors between the measured and predicted wall temperatures by the GBDT and RF models are 5.7% and 6.2%, respectively, confirming the accuracy of the two ML models.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2024

Distilling particle knowledge for fast reconstruction at high-energy physics experiments

Knowledge distillation is a form of model compression that allows artificial neural networks of different sizes to learn from one another. Its main application is the compactification of large deep neural networks to free up computational resources, in particular on edge devices. In this article, we consider proton-proton collisions at the High-Luminosity Large Hadron Collider (HL-LHC) and demonstrate a successful knowledge transfer from an event-level graph neural network (GNN) to a particle-level small deep neural network (DNN). Our algorithm, DistillNet, is a DNN that is trained to learn about the provenance of particles, as provided by the soft labels that are the GNN outputs, to predict whether or not a particle originates from the primary interaction vertex. The results indicate that for this problem, which is one of the main challenges at the HL-LHC, there is minimal loss during the transfer of knowledge to the small student network, while improving significantly the computational resource needs compared to the teacher. This is demonstrated for the distilled student network on a CPU, as well as for a quantized and pruned student network deployed on a field programmable gate array. Our study proves that knowledge transfer between networks of different complexity can be used for fast artificial intelligence (AI) in high-energy physics that improves the expressiveness of observables over non-AI-based reconstruction algorithms. Such an approach can become essential at the HL-LHC experiments, e.g. to comply with the resource budget of their trigger stages.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2023

Revealing the Galaxy–Halo Connection through Machine Learning

Abstract Understanding the connections between galaxy stellar mass, star formation rate, and dark matter halo mass represents a key goal of the theory of galaxy formation. Cosmological simulations that include hydrodynamics, physical treatments of star formation, feedback from supernovae, and the radiative transfer of ionizing photons can capture the processes relevant for establishing these connections. The complexity of these physics can prove difficult to disentangle and obfuscate how mass-dependent trends in the galaxy population originate. Here, we train a machine-learning method called Explainable Boosting Machines (EBMs) to infer how the stellar mass and star formation rate of nearly 6 million galaxies simulated by the Cosmic Reionization on Computers project depend on the physical properties of halo mass, the peak circular velocity of the galaxy during its formation history v peak , cosmic environment, and redshift. The resulting EBM models reveal the relative importance of these properties in setting galaxy stellar mass and star formation rate, with v peak providing the most dominant contribution. Environmental properties provide substantial improvements for modeling the stellar mass and star formation rate in only ≲10% of the simulated galaxies. We also provide alternative formulations of EBM models that enable low-resolution simulations, which cannot track the interior structure of dark matter halos, to predict the stellar mass and star formation rate of galaxies computed by high-resolution simulations with detailed baryonic physics.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

High–Resolution Maps of Near–Surface Permafrost for Three Watersheds on the Seward Peninsula, Alaska Derived From Machine Learning

Permafrost soils are a critical component of the global carbon cycle and are locally important because they regulate the hydrologic flux from uplands to rivers. Furthermore, degradation of permafrost soils causes land surface subsidence, damaging infrastructure that is crucial for local communities. Regional and hemispherical maps of permafrost are too coarse to resolve distributions at a scale relevant to assessments of infrastructure stability or to illuminate geomorphic impacts of permafrost thaw. Here we train machine learning models to generate meter–scale maps of near–surface permafrost for three watersheds in the discontinuous permafrost region. The models were trained using ground truth determinations of near–surface permafrost presence from measurements of soil temperature and electrical resistivity. We trained three classifiers: extremely randomized trees (ERTr), support vector machines (SVM), and an artificial neural network (ANN). Model uncertainty was determined using k–fold cross validation, and the modeled extents of near–surface permafrost were compared to the observed extents at each site. At–a–site near–surface permafrost distributions predicted by the ERTr produced the highest accuracy (70%–90%). However, the transferability of the ERTr to the sites outside of the training data set was poor, with accuracies ranging from 50% to 77%. The SVM and ANN models had lower accuracies for at–a–site prediction (70%–83%), yet they had greater accuracy when transferred to the non–training site (62%–78%). These models demonstrate the potential for integrating high–resolution spatial data and machine learning models to develop maps of near–surface permafrost extent at resolutions fine enough to assess infrastructure vulnerability and landscape morphology influenced by permafrost thaw.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

AP-Net: An atomic-pairwise neural network for smooth and transferable interaction potentials

Intermolecular interactions are critical to many chemical phenomena, but their accurate computation using ab initio methods is often limited by computational cost. The recent emergence of machine learning (ML) potentials may be a promising alternative. Useful ML models should not only estimate accurate interaction energies but also predict smooth and asymptotically correct potential energy surfaces. However, existing ML models are not guaranteed to obey these constraints. Indeed, systemic deficiencies are apparent in the predictions of our previous hydrogen-bond model as well as the popular ANI-1X model, which we attribute to the use of an atomic energy partition. As a solution, we propose an alternative atomic-pairwise framework specifically for intermolecular ML potentials, and we introduce AP-Net—a neural network model for interaction energies. The AP-Net model is developed using this physically motivated atomic-pairwise paradigm and also exploits the interpretability of symmetry adapted perturbation theory (SAPT). We show that in contrast to other models, AP-Net produces smooth, physically meaningful intermolecular potentials exhibiting correct asymptotic behavior. Initially trained on only a limited number of mostly hydrogen-bonded dimers, AP-Net makes accurate predictions across the chemically diverse S66x8 dataset, demonstrating significant transferability. On a test set including experimental hydrogen-bonded dimers, AP-Net predicts total interaction energies with a mean absolute error of 0.37 kcal mol−1, reducing errors by a factor of 2–5 across SAPT components from previous neural network potentials. The pairwise interaction energies of the model are physically interpretable, and an investigation of predicted electrostatic energies suggests that the model “learns” the physics of hydrogen-bonded interactions.

Glick, Zachary L. (ORCID:0000000309002849)↗
DOE OSTI2022

Fed-DeepONet: Stochastic Gradient-Based Federated Training of Deep Operator Networks

The Deep Operator Network (DeepONet) framework is a different class of neural network architecture that one trains to learn nonlinear operators, i.e., mappings between infinite-dimensional spaces. Traditionally, DeepONets are trained using a centralized strategy that requires transferring the training data to a centralized location. Such a strategy, however, limits our ability to secure data privacy or use high-performance distributed/parallel computing platforms. To alleviate such limitations, in this paper, we study the federated training of DeepONets for the first time. That is, we develop a framework, which we refer to as Fed-DeepONet, that allows multiple clients to train DeepONets collaboratively under the coordination of a centralized server. To achieve Fed-DeepONets, we propose an efficient stochastic gradient-based algorithm that enables the distributed optimization of the DeepONet parameters by averaging first-order estimates of the DeepONet loss gradient. Then, to accelerate the training convergence of Fed-DeepONets, we propose a moment-enhanced (i.e., adaptive) stochastic gradient-based strategy. Finally, we verify the performance of Fed-DeepONet by learning, for different configurations of the number of clients and fractions of available clients, (i) the solution operator of a gravity pendulum and (ii) the dynamic response of a parametric library of pendulums.

Moya, Christian↗
DOE OSTI2022

Constructing A New CHF Look-Up Table Based on the Domain Knowledge Informed Machine Learning Methodology

Accurate prediction of CHF under various fluid flow conditions continues to be required for design, operation and safety analysis of light water reactor rod bundles. Due to the lack of in-depth physical understanding as well as limited high-resolution data in the micro-scale flow and heat transfer, the existing models feature a sub-optimal uncertainty band. In this study, driven by the prior domain knowledge information obtained, an improved CHF look-up table is developed through unified machine learning algorithms for the vertical flow conditions within tube and annulus geometry. The Groeneveld 2006 look-up table is used as the domain knowledge to train machine learning process against tube and annulus CHF data for both DNB and DO type. The new look-up table shows improved accuracy for conditions relevant to PWRs and BWRs. In addition, its domain knowledge informed nature ensures that a rationale prediction can be made, thus accounting for previous valuable information in the machine learning model training process.

Jin, Yue↗
DOE OSTI2024

NeuroSEM: A hybrid framework for simulating multiphysics problems by coupling PINNs and spectral elements

Multiphysics problems that are characterized by complex interactions among fluid dynamics, heat transfer, structural mechanics, and electromagnetics, are inherently challenging due to their coupled nature. While experimental data on certain state variables may be available, integrating these data with numerical solvers remains a significant challenge. Physics-informed neural networks (PINNs) have shown promising results in various engineering disciplines, particularly in handling noisy data and solving inverse problems in partial differential equations (PDEs). However, their effectiveness in forecasting nonlinear phenomena in multiphysics regimes, particularly involving turbulence, is yet to be fully established. Here, this study introduces NeuroSEM, a hybrid framework integrating PINNs with the highfidelity Spectral Element Method (SEM) solver, Nektar++. NeuroSEM leverages the strengths of both PINNs and SEM, providing robust solutions for multiphysics problems. PINNs are trained to assimilate data and model physical phenomena in specific subdomains, which are then integrated into the Nektar++ solver. We demonstrate the efficiency and accuracy of NeuroSEM for thermal convection in cavity flow and flow past a cylinder. The framework effectively handles data assimilation by addressing those subdomains and state variables where the data is available. We applied NeuroSEM to the Rayleigh-B´enard convection system, including cases with missing thermal boundary conditions and noisy datasets. Finally, we applied the proposed NeuroSEM framework to real particle image velocimetry (PIV) data to capture flow patterns characterized by horseshoe vortical structures. Our results indicate that NeuroSEM accurately models the physical phenomena and assimilates the data within the specified subdomains. The framework’s plug-and-play nature facilitates its extension to other multiphysics or multiscale problems. Furthermore, NeuroSEM is optimized for efficient execution on emerging integrated GPU-CPU architectures. This hybrid approach enhances the accuracy and efficiency of simulations, making it a powerful tool for tackling complex engineering challenges in various scientific domains.

42 ENGINEERING↗
DOE OSTI2026

Prediction and Experimental Verification of Electrolyte Solvation Structure from an OMol25-Trained Interatomic Potential

A molecular-level understanding of electrolyte solvation structure and ion–ion correlations is critical to developing next-generation battery chemistries. Atomistic simulation capabilities with sufficient accuracy, speed, and transferability to deliver reliable structural insights while avoiding arduous system-specific reparameterization are thus highly desirable. Machine learning interatomic potentials (MLIPs) trained on large, chemically diverse data sets are revolutionizing computational chemistry, enabling molecular dynamics simulations of battery electrolytes with near-DFT accuracy over 10,000× faster than DFT. While previous MLIP training data sets with suitable elemental coverage for electrolytes have been based on inorganic materials, the Open Molecules 2025 (OMol25) data set provides large-scale molecular DFT MLIP training data with broad elemental coverage and specifically samples tens of millions of electrolyte configurations. Here, we integrate computational modeling with experimental validation to systematically assess the ability of large-scale MLIPs pretrained on materials data or on OMol25 to accurately resolve nanoscale structural organization and ion-solvation characteristics in Na-ion battery electrolytes across diverse physicochemical conditions and compositional regimes. We find that the OMol25-trained Universal Model of Atoms (UMA-OMol) predicts experimentally measured densities and X-ray structure factors in substantially better agreement compared to state-of-the-art models trained only on inorganic materials data. Using UMA-OMol, we further analyze systematic trends in solvation structure as a function of cation identity, anion chemistry, salt concentration, and solvent topology. We observe that increasing system temperature amplifies the heterogeneity within the solvation environment, perturbing cation–solvent interactions and promoting the formation of contact ion pairs (CIPs). Moreover, subtle variations in the solvent topology of glyme-based electrolytes cause pronounced changes in ion correlations and solvation structure. The experimental agreement and microscopic insights shown here position OMol25-trained MLIPs as a practical route to predictive, high-throughput electrolyte simulations beyond the limits of classical force fields and direct DFT molecular dynamics, serving as a powerful tool for accelerating the design of next-generation Na-ion battery electrolytes and beyond.

MLIPs↗
DOE OSTI2020

State-of-the-art on research and applications of machine learning in the building life cycle

Fueled by big data, powerful and affordable computing resources, and advanced algorithms, machine learning has been explored and applied to buildings research for the past decades and has demonstrated its potential to enhance building performance. This study systematically surveyed how machine learning has been applied at different stages of building life cycle. By conducting a literature search on the Web of Knowledge platform, we found 9579 papers in this field and selected 153 papers for an in-depth review. The number of published papers is increasing year by year, with a focus on building design, operation, and control. However, no study was found using machine learning in building commissioning. There are successful pilot studies on fault detection and diagnosis of HVAC equipment and systems, load prediction, energy baseline estimate, load shape clustering, occupancy prediction, and learning occupant behaviors and energy use patterns. None of the existing studies were adopted broadly by the building industry, due to common challenges including (1) lack of large scale labeled data to train and validate the model, (2) lack of model transferability, which limits a model trained with one data-rich building to be used in another building with limited data, (3) lack of strong justification of costs and benefits of deploying machine learning, and (4) the performance might not be reliable and robust for the stated goals, as the method might work for some buildings but could not be generalized to others. Finally, findings from the study can inform future machine learning research to improve occupant comfort, energy efficiency, demand flexibility, and resilience of buildings, as well as to inspire young researchers in the field to explore multidisciplinary approaches that integrate building science, computing science, data science, and social science.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Regional-scale soil carbon predictions can be enhanced by transferring global-scale soil–environment relationships

Accurate modelling and mapping soil organic carbon are crucial for supporting soil health restoration and climate change mitigation at both regional and global scales. However, regional soil predictions often suffer from data scarcity and high prediction uncertainty. Utilizing a pre-trained global-to-regional soil carbon predictive model can be a potential solution to address this challenge. Despite its promise, how to construct and apply the global-scale model to enhance regional-scale soil carbon mapping remains largely unexplored. Here, we propose the Global Soil Carbon Pre-trained Model (GSoilCPM), a deep-learning-based domain adaptative model, to enhance regional-scale soil carbon predictions. Based on large amount of environmental covariate data and 106,167 soil samples across the globe, we verify our hypothesis of the effectiveness of this 'global-to-regional' modelling strategy. The pre-trained model can be then transferred and fine-tuned to bridge the regional- and global-scale soil–environment relationships. We applied and validated this modelling strategy in four regional-scale study areas, three in the Northern Hemisphere and one in the Southern Hemisphere, each with distinct environmental background. Compared to traditional modelling approaches as a baseline, four case studies all demonstrated significant improvement in prediction accuracy across diverse environments and varying data availabilities. The average percentage improvement across all regions is 10.93% (absolute values decreased by 1.20 g kg−1 averagely) in MAE and 29.04% (absolute values increased by 0.10 averagely) in CCC. The applicability and future horizons of using GSoilCPM were further discussed. We further reveal that regions with fewer soil samples or lower baseline accuracy benefit more from the pre-trained global model. Our findings highlight the advantages of leveraging the generalized knowledge from global models to enhance specifically localized soil modelling, positioning a potential paradigm shift in digital soil mapping, and far-reaching implications for soil monitoring and land management.

Deep learning↗
DOE OSTI2024

Manufacturing of Fabric Electrodes using a High-Throughput Screening Platform for Redox Flow Batteries

The objective of this project is to establish a new manufacturing methodology with machine learning- based high-throughput screening for the design and development of hierarchical structured, high-performance fabric electrodes for redox flow batteries (RFBs). The end goal of the project is to design and manufacture fabric electrodes for RFB applications that can provide 250 mA/cm2 current density operation for 100-cycles with 80% average energy efficiency. This was accomplished by first examining the structure-performance-property linkages of the electrodes provided by our partner, AvCarb. The electrodes’ microstructure was characterized by determining their pore size distribution, tortuosity, specific surface area, and porosity. The ohmic, charge transfer and mass transfer resistances were then calculated using electrochemical impedance spectroscopy. Carbon cloth electrodes showed the greatest resistance, which was dominated by charge transfer resistance, which we believe is related to the surface functionalization. Full cell cycling was used in order to determine the area specific resistance and energy efficiency of the cells. All of this experimental data and the results of the mathematical model (to increase the amount of inputs with parametric sweeping) were used to develop a machine learning-based model for the design of high-performance fabric electrodes. Using the results from the machine learning tool, optimized electrodes were fabricated by AvCarb. The ohmic, charge transfer and mass transfer resistances for these new electrodes were measured, and both performed better than any of the initial samples which had been provided by AvCarb.

25 ENERGY STORAGE↗
DOE OSTI2026

Federated learning for 2D synchrotron x-ray diffractometry: a cross-institutional approach for phase quantification of Ti–6Al–4V alloy

High-energy Two dimensional (2D) synchrotron x-ray diffractometry provides important insights into the atomistic structure and phase evolution of materials, yet traditional analysis methods remain complex, knowledge-intensive, and computationally demanding. Deep-learning models offer a powerful alternative for automating their analysis. Institutions that hold these datasets may be unwilling to share their data due to privacy and security policies, as well as the challenges associated with large-scale data transfer. As a result, models trained on local datasets often perform well only on their own data but exhibit bias and poor generalization across different instruments or facilities. To overcome these limitations, we explore federated learning (FL) for 2D synchrotron diffractograms, enabling collaborative model training without exchanging raw data. In this study, 2D synchrotron diffractograms of Ti–6Al–4V alloy collected from two independent facilities are used to train convolutional neural networks for predicting the β-phase volume fraction. Experimental results show that federated global models significantly outperform locally trained models in terms of generalization and achieve accuracy comparable to centralized trained models. These findings demonstrate the potential of FL to enable secure, cross-institutional collaboration and enhance the scalability of deep-learning-based materials characterization.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Machine-learning force-field models for dynamical simulations of metallic magnets

We review recent advances in machine-learning (ML) force-field methods for Landau–Lifshitz–Gilbert simulations of itinerant electron magnets, focusing on their scalability and transferability. Built on the principle of locality, a deep neural-network model is developed to efficiently and accurately predict electron-mediated forces governing spin dynamics. Symmetry-aware descriptors constructed through a group-theoretical approach ensure rigorous incorporation of both lattice and spin-rotation symmetries. The framework is demonstrated using the prototypical s-d exchange model widely employed in spintronics. ML-enabled large-scale simulations reveal novel nonequilibrium phenomena, including anomalous coarsening of tetrahedral spin order on the triangular lattice and the freezing of phase-separation dynamics in lightly hole-doped, strong-coupling square-lattice systems. These results establish ML force-field frameworks as scalable, accurate, and versatile tools for modeling nonequilibrium spin dynamics in itinerant magnets.

Artificial neural networks↗
DOE OSTI2024

Machine Learning Based Prediction of Airflow Maldistribution in A-Type Heat Exchangers

Airflow maldistribution is one of the primary causes of performance degradation in air-to refrigerant heat exchangers (HX) and has been shown to decrease heat transfer by as much as 35%. As a result, many units are oversized to meet the target capacity, resulting in increased system cost and refrigerant charge. Several studies have explored how characteristics like package type and HX geometry impact the flow profile, but results are restricted to a limited range of parameters and cannot be extrapolated to new designs. In this work, a machine learning (ML) model is trained to predict the inlet flow profile of dry air entering A-type HXs across a broad range of geometries and conditions. Flow profiles are generated using a porous media CFD model and used to train an Artificial Neural Network (ANN) which exhibits maximum and average relative L2 norm errors of 0.48 and 0.05. Additionally, these predictions take less than a second to generate resulting in a speed up factor of 2.42E5 compared to CFD. Component-level simulations are conducted to determine the performance degradation resulting from the predicted airflow maldistribution profiles. The new ML model will enable rapid and accurate prediction of performance degradation resulting from airflow maldistribution in A-type HXs, allowing for more accurate and cost-effective HX design.

42 ENGINEERING↗
DOE OSTI2023

Reinforcement Learning for Load-balanced Parallel Particle Tracing

We explore an online reinforcement learning (RL) paradigm to dynamically optimize parallel particle tracing performance in distributed-memory systems. Our method combines three novel components: (1) a work donation algorithm, (2) a high-order workload estimation model, and (3) a communication cost model. First, we design an RL-based work donation algorithm. Our algorithm monitors workloads of processes and creates RL agents to donate data blocks and particles from high-workload processes to low-workload processes to minimize program execution time. The agents learn the donation strategy on the fly based on reward and cost functions designed to consider processes' workload changes and data transfer costs of donation actions. Second, we propose a workload estimation model, helping RL agents estimate the workload distribution of processes in future computations. Third, we design a communication cost model that considers both block and particle data exchange costs, helping RL agents make effective decisions with minimized communication costs. We demonstrate that our algorithm adapts to different flow behaviors in large-scale fluid dynamics, ocean, and weather simulation data. Our algorithm improves parallel particle tracing performance in terms of parallel efficiency, load balance, and costs of I/O and communication for evaluations with up to 16,384 processors.

Distributed and parallel particle tracing↗