SEARCH · Engineering Papers

Results for “sparse data”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

1,361 records · Page 76

All sources NASA NTRS DOE OSTI

DOE OSTI2017

Embedded ensemble propagation for improving performance, portability, and scalability of uncertainty quantification on emerging computational architectures

In this study, quantifying simulation uncertainties is a critical component of rigorous predictive simulation. A key component of this is forward propagation of uncertainties in simulation input data to output quantities of interest. Typical approaches involve repeated sampling of the simulation over the uncertain input data, and can require numerous samples when accurately propagating uncertainties from large numbers of sources. Often simulation processes from sample to sample are similar and much of the data generated from each sample evaluation could be reused. We explore a new method for implementing sampling methods that simultaneously propagates groups of samples together in an embedded fashion, which we call embedded ensemble propagation. We show how this approach takes advantage of properties of modern computer architectures to improve performance by enabling reuse between samples, reducing memory bandwidth requirements, improving memory access patterns, improving opportunities for fine-grained parallelization, and reducing communication costs. We describe a software technique for implementing embedded ensemble propagation based on the use of C++ templates and describe its integration with various scientific computing libraries within Trilinos. We demonstrate improved performance, portability and scalability for the approach applied to the simulation of partial differential equations on a variety of CPU, GPU, and accelerator architectures, including up to 131,072 cores on a Cray XK7 (Titan).

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2026

HARMONY: Large-Scale Architecture Search for Efficient Hybrid Language Models

As large language models scale to trillions of parameters, their computational and memory requirements present critical challenges for efficient training and deployment. While Mixture of Experts (MoE) architectures enable efficient scaling through sparse parameter activation, and state-space models like Mamba offer linear-time complexity, principled methods for combining these paradigms remain undeveloped. We introduce HARMONY (Hybrid Architecture Research for Mamba, Optimized with Neural efficiencY), a multi-objective evolutionary neural architecture search framework for discovering efficient hybrid language models that integrate Transformer attention mechanisms, Mixture-of-Experts routing, and Mamba state-space components. Through large-scale distributed search using 16,384 MI250X GPUs on the Frontier supercomputer, HARMONY explores a comprehensive design space encompassing six attention variants (MHA, MQA, GQA, MLA, SWA, and Mamba-2), variable MoE configurations with both routed and shared experts, and extensive Mamba hyperparameters. Our framework discovers heterogeneous architectures that balance training performance with computational efficiency through multi-objective optimization incorporating latency penalties and fitness-based selection. Analysis of discovered architectures reveals that optimal hybrid designs favor heterogeneous component mixing rather than homogeneous patterns, with Mamba-2 and Multi-Head Latent Attention (MLA) emerging as preferred mechanisms. Discovered architectures demonstrate superior training efficiency: our best configuration achieves a final perplexity of 1.0874 with 2.38B parameters while processing 4,320 tokens/second, outperforming significantly larger manually designed models. Full-scale evaluation shows HARMONY's top architectures achieve better loss trajectories than equivalently-sized models using state-of-the-art configurations including Mixtral, Jamba, and Samba. Additionally, we demonstrate 91% weak scaling efficiency when training discovered 36B-parameter models across 1,024 GPUs. HARMONY is released as an open framework with comprehensive tools for building and training hybrid models using expert-data-pipeline parallelism, democratizing access to automated architecture design for next-generation language models.

Herron, Emily [ORNL] (ORCID:0000000273008172)↗

DOE OSTI2026

Sensor Reduction for Diversion Detection in a Realistic Heat Pipe Microreactor Using Supervised Machine Learning

Microreactors are designed as a smaller, cheaper, and safer alternative to traditional nuclear power plants. Their non-traditional characteristics and prospect of mass production and deployment will likely require new approaches to nuclear safeguards. The primary proliferation concern with microreactors is the diversion of fuel material. Such diversion may produce measurable defects in key physical attributes like neutron flux, which may in turn be detectable using machine learning models. Preliminary work has demonstrated this ability for modeled nominal and diversion scenarios using large quantities of energy integrated neutron flux data. In practice, the number of available sensors for such measurements will be limited and energy integrated flux information will not be available. This work explores the ability of tree-based gradient boosted ensemble models to classify a given microreactor core is nominal or diversion, and determine the number of fuel pins diverted in the case of diversion with reduced numbers of sensors and more realistic detector responses. Classification accuracy of greater than 98% and regression errors as low as 5% of the total number of fuel pins were achieved with as few as 15 sensors, compared to 99% and 4.1% with a maximum of 240 sensors.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2020

Scaling the training of particle classification on simulated MicroBooNE events to multiple GPUs

Measurements in Liquid Argon Time Projection Chamber (LArTPC) neutrino detectors, such as the MicroBooNE detector at Fermilab, feature large, high fidelity event images. Deep learning techniques have been extremely successful in classification tasks of photographs, but their application to LArTPC event images is challenging, due to the large size of the events. Events in these detectors are typically two orders of magnitude larger than images found in classical challenges, like recognition of handwritten digits contained in the MNIST database or object recognition in the ImageNet database. Ideally, training would occur on many instances of the entire event data, instead of many instances of cropped regions of interest from the event data. However, such efforts lead to extremely long training cycles, which slow down the exploration of new network architectures and hyperparameter scans to improve the classification performance. We present studies of scaling a LArTPC classification problem on multiple architectures, spanning multiple nodes. The studies are carried out on simulated events in the MicroBooNE detector. We emphasize that it is beyond the scope of this study to optimize networks or extract the physics from any results here. Institutional computing at Pacific Northwest National Laboratory and the SummitDev machine at Oak Ridge National Laboratory’s Leadership Computing Facility have been used. To our knowledge, this is the first use of state-of-the-art Convolutional Neural Networks for particle physics and their attendant compute techniques onto the DOE Leadership Class Facilities. We expect benefits to accrue particularly to the Deep Underground Neutrino Experiment (DUNE) LArTPC program, the flagship US High Energy Physics (HEP) program for the coming decades.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2023

Evaluating 3 decades of precipitation in the Upper Colorado River basin from a high-resolution regional climate model

Abstract. Convection-permitting regional climate models (RCMs) have recently become tractable for applications at multi-decadal timescales. These types of models have tremendous utility for water resource studies, but better characterization of precipitation biases is needed, particularly for water-resource-critical mountain regions, where precipitation is highly variable in space, observations are sparse, and the societal water need is great. This study examines 34 years (1987–2020) of RCM precipitation from the Weather Research and Forecasting model (WRF; v3.8.1), using the Climate Forecast System Reanalysis (CFS; CFSv2) initial and lateral boundary conditions and a 1 km × 1 km innermost grid spacing. The RCM is centered over the Upper Colorado River basin, with a focus on the high-elevation, 750 km2 East River watershed (ERW), where a variety of high-impact scientific activities are currently ongoing. Precipitation is compared against point observations (Natural Resources Conservation Service Snow Telemetry or SNOTEL), gridded climate datasets (Newman, Livneh, and PRISM), and Bayesian reconstructions of watershed mean precipitation conditioned on streamflow and high-resolution snow remote-sensing products. We find that the cool-season precipitation percent error between WRF and 23 SNOTEL gauges has a low overall bias (x^ = 0.25 %, s = 13.63 %) and that WRF has a higher percent error during the warm season (x^ = 10.37 %, s = 12.79 %). Warm-season bias manifests as a high number of low-precipitation days, though the low-resolution or SNOTEL gauges limit some of the conclusions that can be drawn. Regional comparisons between WRF precipitation accumulation and three different gridded datasets show differences on the order of ± 20 %, particularly at the highest elevations and in keeping with findings from other studies. We find that WRF agrees slightly better with the Bayesian reconstruction of precipitation in the ERW compared to the gridded precipitation datasets, particularly when changing SNOTEL densities are taken into account. The conclusions are that the RCM reasonably captures orographic precipitation in this region and demonstrates that leveraging additional hydrologic information (streamflow and snow remote-sensing data) improves the ability to characterize biases in RCM precipitation fields. Error characteristics reported in this study are essential for leveraging the RCM model outputs for studies of past and future climates and water resource applications. The methods developed in this study can be applied to other watersheds and model configurations. Hourly 1 km × 1 km precipitation and other meteorological outputs from this dataset are publicly available and suitable for a wide variety of applications.

Geology↗

DOE OSTI2013

Minimal spanning tree algorithm for γ-ray source detection in sparse photon images: cluster parameters and selection strategies

We present that the minimal spanning tree (MST) algorithm is a graph-theoretical cluster-finding method. We previously applied it to γ-ray bidimensional images, showing that it is quite sensitive in finding faint sources. Possible sources are associated with the regions where the photon arrival directions clusterize. MST selects clusters starting from a particular “tree” connecting all the point of the image and performing a cut based on the angular distance between photons, with a number of events higher than a given threshold. In this paper, we show how a further filtering, based on some parameters linked to the cluster properties, can be applied to reduce spurious detections. We find that the most efficient parameter for this secondary selection is the magnitudeM of a cluster, defined as the product of its number of events by its clustering degree. We test the sensitivity of the method by means of simulated and real Fermi-Large Area Telescope (LAT) fields. Our results show that √M is strongly correlated with other statistical significance parameters, derived from a wavelet based algorithm and maximum likelihood (ML) analysis, and that it can be used as a good estimator of statistical significance of MST detections. Finally, we apply the method to a 2-year LAT image at energies higher than 3 GeV, and we show the presence of new clusters, likely associated with BL Lac objects.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2026

Statistical Correlation of Heliostat Pointing Deviation With Wind

This work was carried out as part of the Heliostat Consortium (HelioCon) Field Deployment subtask with the aim to develop a reduced order model framework for correlating wind speed and pointing deviation of a heliostat facet. There are only sparse field measurements of heliostat pointing deviations and accompanying wind conditions published in the literature. Heliostat test standards, such as IEC 62862-4-3, propose a suite of tests including laser pointing repeatability at wind speeds below 4 m/s, and provide technical requirements for heliostat slope and tracking deviations in coarse average wind speed bins of 4 m/s, 6 m/s, and 8 m/s. In addressing the gap of the variation of heliostat pointing deviation with wind speed, field measurements of laser pointing on a grid target and wind conditions were analyzed in this study at the Third-Party Metrology Platform at the National Laboratory of the Rockies (NLR) Flatirons Campus. Horizontal pointing deviations were found to follow a logarithmic relationship with peak wind speed, whereas vertical pointing deviations follow an exponential relationship with peak wind speed. Both horizontal and vertical pointing deviations also follow a second order polynomial relationship, as expected from the proportionality of elastic loads and deformations with the square of wind speed. The results indicate that heliostat facet pointing deviations in the vertical direction increase at a faster rate than in the horizontal direction with increasing wind speed over the tested range, however these are dependent on the heliostat structural design. Next steps are recommended for additional field measurements to confirm a linear relationship of pointing deviation with applied moment on a heliostat facet, and to distinguish between gravity-induced and wind-induced pointing deviations at different elevation angles. The derived correlations in the preliminary analysis in this report serve as a case study for heliostat developers and plant operators to estimate the wind-induced pointing deviations and their variation with peak gust wind speed. Next steps in future work would recommend higher resolution and longer duration datasets for different elevation angles and wind directions to reduce uncertainties and variance of collected laser beam spot data and their correlations with bin-averaged wind speed.

17 WIND ENERGY↗

DOE OSTI2017

Exploring performance and energy tradeoffs for irregular applications: A case study on the Tilera many-core architecture

High performance, parallel applications with irregular data accesses are becoming a critical workload class for modern systems. In particular, the execution of such workloads on emerging many-core systems is expected to be a significant component of applications in data mining, machine learning, scientific computing and graph analytics. However, power and energy constraints limit the capabilities of individual cores, memory hierarchy and on-chip interconnect of such systems, thus leading to architectural and software trade-os that must be understood in the context of the intended application’s behavior. Irregular applications are notoriously hard to optimize given their data-dependent access patterns, lack of structured locality and complex data structures and code patterns. We have ported two irregular applications, graph community detection using the Louvain method (Grappolo) and high-performance conjugate gradient (HPCCG), to the Tilera many-core system and have conducted a detailed study of platform-independent and platform-specific optimizations that improve their performance as well as reduce their overall energy consumption. To conduct this study, we employ an auto-tuning based approach that explores the optimization design space along three dimensions - memory layout schemes, GCC compiler flag choices and OpenMP loop scheduling options. We leverage MIT’s OpenTuner auto-tuning framework to explore and recommend energy optimal choices for different combinations of parameters. We then conduct an in-depth architectural characterization to understand the memory behavior of the selected workloads. Finally, we perform a correlation study to demonstrate the interplay between the hardware behavior and application characteristics. Using auto-tuning, we demonstrate whole-node energy savings and performance improvements of up to 49:6% and 60% relative to a baseline instantiation, and up to 31% and 45:4% relative to manually optimized variants.

auto tuning↗

DOE OSTI2013

IMAGE ANALYSIS FOR COSMOLOGY: RESULTS FROM THE GREAT10 STAR CHALLENGE

We present the results from the first public blind point-spread function (PSF) reconstruction challenge, the GRavitational lEnsing Accuracy Testing 2010 (GREAT10) Star Challenge. Reconstruction of a spatially varying PSF, sparsely sampled by stars, at non-star positions is a critical part in the image analysis for weak lensing where inaccuracies in the modeled ellipticity e and size R {sup 2} can impact the ability to measure the shapes of galaxies. This is of importance because weak lensing is a particularly sensitive probe of dark energy and can be used to map the mass distribution of large scale structure. Participants in the challenge were presented with 27,500 stars over 1300 images subdivided into 26 sets, where in each set a category change was made in the type or spatial variation of the PSF. Thirty submissions were made by nine teams. The best methods reconstructed the PSF with an accuracy of {sigma}(e) Almost-Equal-To 2.5 Multiplication-Sign 10{sup -4} and {sigma}(R {sup 2})/R {sup 2} Almost-Equal-To 7.4 Multiplication-Sign 10{sup -4}. For a fixed pixel scale, narrower PSFs were found to be more difficult to model than larger PSFs, and the PSF reconstruction was severely degraded with the inclusion of an atmospheric turbulence model (although this result is likely to be a strong function of the amplitude of the turbulence power spectrum).

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2022

Creep and Fatigue Characterization of High Strength Alloy Thin Sections in Advanced CO2 Heat Exchangers

The objective of this work was to characterize and model elevated temperature creep and fatigue behavior for thin sheet and foil forms of gamma-prime strengthened alloys in wrought form and as-processed folded and brazed constructions. This work was motivated by the demanding temperature and pressure service conditions of the GEN3 Concentrated Solar Power (CSP) and supercritical CO2 (sCO2) power cycle working fluid. More specifically, the possibility of leveraging the superior creep strength of gamma-prime alloys in folded-fin and brazed-plate heat exchanger constructions. Gamma-prime alloys represent a step-change in raw-material strength over solid-solution strengthened alloys. And the folded-fin and brazed-plate heat exchanger architecture is lightweight and leverages cost-effective material stock forms. The investigation contained two parallel paths. (1) The first is referred to as a fundamental investigation where Oak Ridge National Laboratory conducts uniaxial creep testing on thin sheet and foil in wrought form. This effort aimed to serve as a benchmark against a relatively sparse existing database and a baseline comparison for path number 2. (2) The second path is referred to as the practical investigation where Brayton Energy manufactures plate-fin heat exchangers and performs pressurized creep and fatigue testing. This effort aimed to de-risk heat exchanger manufacturing process for service under sCO2 CSP conditions. A total of 14 uniaxial creep tests were completed using Haynes 282 thin sheet and foil. A variety of heat treatments were specified to coincide with path number 2. Baseline metallography of test samples and creep strength performance are contained. Benchmarks relatively to existing thick-form Haynes 282 are made, as well as to other thin-form Nickel-based superalloys. Description of a wrought-form modeling approach for thin gamma-prime alloys is also discussed. A total of 11 pressurized creep and fatigue tests were completed successfully with Haynes 282 heat exchanger prototypes. Manufacturing processing details, testing details, testing results, and failure analysis are discussed. Additionally, creep modeling techniques to predict failure are discussed, and modeling to support technological-to-market. In conclusion, Haynes 282 foils were demonstrated to yield rupture two-to-three orders of magnitude higher than Haynes 230 foils under similar conditions. And the manufactured heat exchanger prototypes demonstrated strength similar to the wrought constituents. Both of which contribute to elevated performance potential or cost savings in practice. Discussion is included.

14 SOLAR ENERGY↗

DOE OSTI2026

Out With the Old: Empirical Trends in U.S. Land‐Based Wind Turbine Decommissioning and Repowering

A growing number of wind turbines (WTs) across the globe are now reaching or exceeding their expected service lifetime; WT decommissioning is on the rise. Accordingly, questions pertaining to WT end-of-life have risen in importance in policy and practice. Yet, research on the various factors relating to WT decommissioning is relatively sparse. Moreover, the key assumptions underpinning that prior research (e.g., the lifespan of WTs, characteristics of WTs being decommissioned, and whether the site is repowered with new WTs) have never been empirically tested across a large set of decommissioned WTs. Leveraging a uniquely comprehensive and spatially explicit dataset of decommissioned WTs in the United States, this research analyzes spatial, technological, and temporal trends in WT decommissioning and develops a novel predictive model for WT decommissioning. Our analysis pinpoints more than 12,400 WTs that have been fully decommissioned in the United States., the majority of which have been relatively old (> 30 years) and small (< 200 kW). While a WT's age alone is a good predictor of the likelihood of decommissioning, other factors such as the size of the WT and recent performance are also important and significant predictors. Most sites where decommissioning has occurred have seen subsequent repowering, with repowered plants featuring substantially fewer WTs (−86 on average) and higher rated plant capacity (+62 MW on average). Many existing WTs in the U.S. are approaching the end of their expected life with roughly 7500 being 20 or more years old. Findings can help policymakers and stakeholders begin preparing for this potential wave of future decommissioning and repowering.

Decommissioning / End-of-life↗