Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Model Inference”

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 307 records · Page 17
All sourcesNASA NTRSDOE OSTI
DOE OSTI2017

Experimental Observation of a Current-Driven Instability in a Neutral Electron-Positron Beam

Here, we report on the first experimental observation of a current-driven instability developing in a quasineutral matter-antimatter beam. Strong magnetic fields (≥ 1T) are measured, via means of a proton radiography technique, after the propagation of a neutral electron-positron beam through a background electron-ion plasma. The experimentally determined equipartition parameter of ε B ≈ 10 -3 is typical of values inferred from models of astrophysical gamma-ray bursts, in which the relativistic flows are also expected to be pair dominated. The data, supported by particle-in-cell simulations and simple analytical estimates, indicate that these magnetic fields persist in the background plasma for thousands of inverse plasma frequencies. The existence of such long-lived magnetic fields can be related to analog astrophysical systems, such as those prevalent in lepton-dominated jets.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2019

Evolving Energy Efficient Convolutional Neural Networks

As deep neural networks have been deployed in more and more applications over the past half decade and are finding their way into an ever increasing number of operational systems, their energy consumption becomes a concern whether running in the datacenter or on edge devices. Hyperparameter optimization and automated network design for deep learning is a quickly growing field, but much of the focus has remained only on optimizing for the performance of the machine learning task. In this work, we demonstrate that the best performing networks created through this automated network design process have radically different computational characteristics (e.g. energy usage, model size, inference time), presenting the opportunity to utilize this optimization process to make deep learning networks more energy efficient and deployable to smaller devices. Optimizing for these computational characteristics is critical as the number of applications of deep learning continues to expand.

Young, Steven↗
DOE OSTI2017

Deterministic influences exceed dispersal effects on hydrologically-connected microbiomes: Deterministic assembly of hyporheic microbiomes

Subsurface zones of groundwater and surface water mixing (hyporheic zones) are regions of enhanced rates of biogeochemical cycling, yet ecological processes governing hyporheic microbiome composition and function through space and time remain unknown. We sampled attached and planktonic microbiomes in the Columbia River hyporheic zone across seasonal hydrologic change, and employed statistical null models to infer mechanisms generating temporal changes in microbiomes within three hydrologically-connected, physicochemically-distinct geographic zones (inland, nearshore, river). We reveal that microbiomes remain dissimilar through time across all zones and habitat types (attached vs. planktonic) and that deterministic assembly processes regulate microbiome composition in all data subsets. The consistent presence of heterotrophic taxa and members of the Planctomycetes-Verrucomicrobia-Chlamydiae (PVC) superphylum nonetheless suggests common selective pressures for physiologies represented in these groups. Further, co-occurrence networks were used to provide insight into taxa most affected by deterministic assembly processes. We identified network clusters to represent groups of organisms that correlated with seasonal and physicochemical change. Extended network analyses identified keystone taxa within each cluster that we propose are central in microbiome composition and function. Finally, the abundance of one network cluster of nearshore organisms exhibited a seasonal shift from heterotrophic to autotrophic metabolisms and correlated with microbial metabolism, possibly indicating an ecological role for these organisms as foundational species in driving biogeochemical reactions within the hyporheic zone. Taken together, our research demonstrates a predominant role for deterministic assembly across highly-connected environments and provides insight into niche dynamics associated with seasonal changes in hyporheic microbiome composition and metabolism.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2016

Aquifer recharge, depletion, and connectivity: Inferences from GRACE, land surface models, and geochemical and geophysical data

We used data from the Gravity Recovery and Climate Experiment (GRACE) and outputs of the CLM4.5 model to estimate recharge and depletion rates for large aquifers, investigate the connectivity of an aquifer's subbasins, and identify barriers and preferred pathways for groundwater flow within an aquifer system. The Nubian Sandstone Aquifer System and its subbasins (Dakhla, Northern Sudan Platform, and Kufra) in northeast Africa were used for demonstration purposes, and findings were tested and verified against geological, geophysical, remote sensing, geochronologic, and geochemical data. There are four major findings. (1) The average annual precipitation data over recharge areas in the southern Kufra section and the Northern Sudan Platform subbasin were estimated at 54.8 km 3 , and 32.8 km 3 , respectively, and knowing the annual extraction rates over these two areas (similar to 0.40 ± 0.20 km 3 ), recharge rates were estimated at 0.78 ± 0.49 km 3 /yr and 1.44 ± 0.42 km 3 /yr, respectively. (2) GRACEderived groundwater depletion rates over the Dakhla subbasin and the Northern Kufra section were estimated at 4.44 ± 0.42 km 3 /yr and 0.48 ± 0.32 km 3 /yr, respectively. (3) The observed depletion in the southern parts of the Dakhla subbasin is apparently caused by the presence of the east- west- trending Uweinat-Aswan basement uplift, which impedes the south-to-north groundwater flow and hence reduces replenishment from recharge areas in the south. (4) A major northeast- southwest- trending shear zone (Pelusium shear system) is apparently providing a preferred groundwater flow pathway from the Kufra to the Dakhla subbasin. Thus, inte-grated approach provides a replicable and cost-effective model for better understanding of the hydrogeologic setting of large aquifers worldwide and for optimum management of these groundwater resources.

58 GEOSCIENCES↗
DOE OSTI2019

Scalable Extended Dynamic Mode Decomposition Using Random Kernel Approximation

The Koopman operator is a linear, infinite-dimensional operator that governs the dynamics of system observables; extended dynamic mode decomposition (EDMD) is a data-driven method for approximating the Koopman operator using functions (features) of the system state snapshots. This paper investigates an approach to EDMD in which the features used provide random approximations to a particular kernel function. The objective of this is computational economy for large data sets: EDMD is generally ill-suited for problems with large state dimension, and its dual kernel formulation (KDMD) is well-suited for such problems only if the number of data snapshots is relatively small. We discuss two specific methods for generating features: random Fourier features and the Nÿstrom method. The first method is a data-independent method for translation-invariant kernels only and involves random sampling in feature space; the second method is a data-dependent empirical method that may be used for any kernel and involves random sampling of data. We first discuss how these ideas may be applied in an EDMD context, as well as a means for adaptively adding random Fourier features. As a result, we demonstrate these methods on two example problems and conclude with an analysis of the relative benefits and drawbacks of each method.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Performance Study of CXL Memory Topology

This paper presents a comprehensive evaluation of the performance impact of various Compute Express Link (CXL) memory topologies, with a particular emphasis on CXL switches, in the context of High- Performance Computing (HPC) and Large Language Model (LLM) inference workloads. Our study unveils significant performance variations across different topologies, demonstrating that certain configurations yield superior performance for specific workloads. These findings underscore the critical importance of tailored topol- ogy selection in optimizing system performance. Additionally, we address the inherent challenges associated with integrating CXL switches, including overhead considerations and routing complex- ities. Our research highlights the necessity for thorough evalua- tion methodologies to fully leverage CXL technology’s potential in contemporary computing environments. These insights provide valuable guidance for system architects and data center operators in designing and optimizing CXL-based infrastructures for diverse workload requirements.

CXL, memory, Artificial Intelligence (AI), HPC↗
DOE OSTI2022

Contributing to Sunny open source code

The Sunny package is an existing open source software to support the simulation of magnetic spin systems. It is available online under the MIT license here: https://github.com/SunnySuite/Sunny.jl The package aims to include the following features: - Analysis for symmetry-allowed interactions - Monte Carlo algorithms for simulating equilibrium configurations. - Schemes for integrating spin dynamics. - Inference of model Hamiltonians - Support for generalized SU(N) spins - Tools for interpreting experimental data, such as diffraction data, magnetization curves, and heat capacity data

Barros, Kipton↗
DOE OSTI2026

Simulation-Based Inference for Neutrino Parameter Tuning

This code trains a simulation-based inference (SBI) model for neutrino interaction parameter tuning and subsequently evaluates its performance on MicroBooNE, T2K, and NuWro datasets.

Tame-Narvaez, Karla [Fermi National Accelerator La↗
DOE OSTI2021

2020 ETI Annual Summer School: Data Science and Engineering

The Consortium for Enabling Technologies & Innovation (ETI) was established in 2019 to address emerging technologies within the context of nuclear nonproliferation. ETI creates a research and education environment to support cross-cutting technologies across three core disciplines: 1) computer and engineering science research specifically in a form of machine learning and high performance computing (HPC), 2) advanced manufacturing, and 3) nuclear detection technologies. For outreach and development, ETI hosted the first of three summer schools from August 24-28, 2020 with the theme of “Data Science and Engineering”. The school was hosted in an on-line format and had over 200 participants. The recorded content is available on-line as a resource for students. The summer school had four modules: 1) Fundamentals of data Applications, 2) Computational Machine Learning, 3) Bayesian Modeling and Inference, and 4) Data Science for Safeguards. Modules contained both lectures as well as student exercises. Poll Everywhere was utilized in some modules as an on-line method to engage large groups of students. Upcoming ETI Summer Schools include Novel Instrumentation in 2021 and Advanced Manufacturing in 2022.

Biegalski, Steven R.↗
DOE OSTI2022

Semantic Regression for Disentangling Beam Losses in the Fermilab Main Injector and Recycler

Fermilab’s Main Injector enclosure houses two accelerators: the Main Injector (MI) and the Recycler (RR). In periods of joint operation, when both machines contain high intensity beam, radiative beam losses from MI and RR overlap on the enclosure’s beam loss monitoring (BLM) system, making it difficult to attribute those losses to a single machine. Incorrect diagnoses result in unnecessary downtime that incurs both financial and experimental cost. In this work, we introduce a novel neural approach for automatically disentangling each machine’s contributions to those measured losses. Using a continuous adaptation of the popular UNet architecture in conjunction with a novel data augmentation scheme, our model accurately infers the machine of origin on a per-BLM basis in periods of joint and independent operation. Crucially, by extracting beam loss information at varying receptive fields, the method is capable of learning both local and global machine signatures and producing high quality inferences using only raw BLM loss measurements.

43 PARTICLE ACCELERATORS↗
DOE OSTI2019

Data and Models: A Framework for Advancing AI in Science

On June 5, 2019, the Office of Science (SC) organized a one-day roundtable to focus on enhancing access to high-quality and fully traceable research data, models, and computing resources to increase the value of such resources for artificial intelligence (AI) research and development and the SC mission.1 In this report, we consider AI to be inclusive of, for example, machine learning (ML), deep learning (DL), neural networks (NN), computer vision, and natural language processing (NLP). We consider “data for AI” to mean the digital artifacts used to generate AI models and/or employed in combination with AI models during inference. In part, this roundtable was motivated by the recognition that a large portion of science data currently are not well suited for AI.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Absorber Column CFD model validation against PNNL’s device-scale absorber column on the LCFS unit

Absorber column has been widely used for CO 2 capture in the coal-fired power plants. High-fidelity CFD models play an important role in absorber column design and solvents optimization, which help enhance the CO 2 capture efficiency and reduce the operation cost. This report provides a comprehensive description of the development of CFD absorber models at two different levels, the design and implementation of PNNL’s device-scale absorber column experiment, and the methodology to combine the CFD results, experiment data, and Aspen model for a better understanding of the interface area in packed column. A composite model is firstly proposed in the Discrete Element Method (DEM) packing process, which can model the complex geometry of various packing elements. This generates a realistic packing pattern and accurate packing porosity ε and specific area a p compared to the actual values for absorber column used in LCFS. Two level of CFD absorber models were developed, namely the full-size column model (FCM) to simulate the entire packed column with a focus on the wall/entrance effects, and the representative column model (RCM) to simulate a section of column with a focus on the sensitivity study of interface area. A Design of Experiment (DoE) plan was developed to guide the collection of 100 run CFD data and 12 experiment runs. The 100 CFD runs were carried out in the RCM with Pro-Pak packing and cover a wide operation range and solvent properties. Impact of influential factors on the interface area in packed column were investigated in details. The CFD interface area was then combined with the experimental data and Aspen model to infer some information of the effective contact angle in the column. The accuracy and uncertainties in the interface area and contact angle can be quantified.

01 COAL, LIGNITE, AND PEAT↗
DOE OSTI2016

Optical and near-infrared observations of SN 2013DX associated with GRB 130702A

We present optical and near-infrared (NIR) light curves and optical spectra of SN 2013dx, associated with the nearby (redshift 0.145) gamma-ray burst GRB 130702A. The prompt isotropic gamma-ray energy released from GRB 130702A is measured to be E{sub γ,iso}=6.4{sub −1.0}{sup +1.3}×10{sup 50} erg (1 keV to 10 MeV in the rest frame), placing it intermediate between low-luminosity GRBs like GRB 980425/SN 1998bw and the broader cosmological population. We compare the observed g{sup ′}r{sup ′}i{sup ′}z{sup ′} light curves of SN 2013dx to a SN 1998bw template, finding that SN 2013dx evolves ∼20% faster (steeper rise time), with a comparable peak luminosity. Spectroscopically, SN 2013dx resembles other broad-lined SNe Ic, both associated with (SN 2006aj and SN 1998bw) and lacking (SN 1997ef, SN 2007I, and SN 2010ah) gamma-ray emission, with photospheric velocities around peak of ∼ 21,000 km s{sup −1}. We construct a quasi-bolometric (g{sup ′}r{sup ′}i{sup ′}z{sup ′}yJ) light curve for SN 2013dx, only the fifth GRB-associated SN with extensive NIR coverage and the third with a bolometric light curve extending beyond Δt>40 days. Together with the measured photospheric velocity, we derive basic explosion parameters using simple analytic models. We infer a {sup 56}Ni mass of M{sub Ni}=0.37±0.01 M{sub ⊙}, an ejecta mass of M{sub ej}=3.1±0.1 M{sub ⊙}, and a kinetic energy of E{sub K}=(8.2±0.43)×10{sup 51} erg (statistical uncertainties only), consistent with previous GRB-associated supernovae. When considering the ensemble population of GRB-associated supernovae, we find no correlation between the mass of synthesized {sup 56}Ni and high-energy properties, despite clear predictions from numerical simulations that M{sub Ni} should correlate with the degree of asymmetry. On the other hand, M{sub Ni} clearly correlates with the kinetic energy of the supernova ejecta across a wide range of core-collapse events.

79 ASTRONOMY AND ASTROPHYSICS↗