A Comparison of Custom and Standard Neuron Model Random Walks on the Ornstein-Uhlenbeck Equation for Simplified Turbulence
Explore the source record for details and available documents.
Taylor, Brady Garland [Sandia National Laboratorie↗
SEARCH · Engineering Papers
Results for “model comparison.”
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.
Interpreting machine learning prediction of fire emissions and comparison with FireMIP process-based models
Annual burned areas in the United States have increased 2-fold during the past decades. With more large fires resulting in more emissions of fine particulate matter, an accurate prediction of fire emissions is critical for quantifying the impacts of fires on air quality, human health, and climate. This study aims to construct a machine learning (ML) model with game-theory interpretation to predict monthly fire emissions over the contiguous US (CONUS) and to understand the controlling factors of fire emissions. The optimized ML model is used to diagnose the process-based models in the Fire Modeling Intercomparison Project (FireMIP) to inform future development. Results show promising performance for the ML model, Community Land Model (CLM), and Joint UK Land Environment Simulator-Interactive Fire And Emission Algorithm For Natural Environments (JULES-INFERNO) in reproducing the spatial distributions, seasonality, and interannual variability of fire emissions over the CONUS. Regional analysis shows that only the ML model and CLM simulate the realistic interannual variability of fire emissions for most of the subregions (r >0.95 for ML and r =0.14~0.70 for CLM), except for Mediterranean California, where all the models perform poorly (r =0.74 for ML and r <0.30 for the FireMIP models). Regarding seasonality, most models capture the peak emission in July over the western US. However, all models except for the ML model fail to reproduce the bimodal peaks in July and October over Mediterranean California, which may be explained by the smaller wind speeds of the atmospheric forcing data during Santa Ana wind events and limitations in model parameterizations for capturing the effects of Santa Ana winds on fire activity. Furthermore, most models struggle to capture the spring peak in emissions in the southeastern US, probably due to underrepresentation of human effects and the influences of winter dryness on fires in the models. As for extreme events, both the ML model and CLM successfully reproduce the frequency map of extreme emission occurrence but overestimate the number of months with extremely large fire emissions. Comparing the fire PM 2.5 emissions from the ML model with process-based fire models highlights their strengths and uncertainties for regional analysis and prediction and provides useful insights into future directions for model improvements.
54 ENVIRONMENTAL SCIENCES↗
Testing multiflavored ULDM models with SPARC
We perform maximum likelihood estimates (MLEs) for single and double flavor ultralight dark matter (ULDM) models using the Spitzer Photometry and Accurate Rotation Curves (SPARC) database. These estimates are compared to MLEs for several commonly used cold dark matter (CDM) models. By comparing various CDM models we find, in agreement with previous studies, that the Burkert and Einasto models tend to perform better than other commonly used CDM models. We focus on comparisons between the Einasto and ULDM models and analyze cases for which the ULDM particle masses are: free to vary; and fixed. For each of these analyses, we perform fits assuming the soliton and halo profiles are: summed together; and matched at a given radius. When we let the particle masses vary, we find a negligible preference for any particular range of particle masses, within $10^{-25}\,\text{eV}\leq m\leq10^{-19}\,\text{eV}$, when assuming the summed models. For the matched models, however, we find that almost all galaxies prefer particles masses in the range $10^{-23}\,\text{eV}\lesssim m\lesssim10^{-20}\,\text{eV}$. For both double flavor models we find that most galaxies prefer approximately equal particle masses. We find that the summed models give much larger variances with respect to the soliton-halo (SH) relation than the matched models. When the particle masses are fixed, the matched models give median and mean soliton and halo values that fall within the SH relation bounds, for most masses scanned. When the particle masses are fixed in the fitting procedure, we find the best fit results for the particle mass $m=10^{-20.5}\,\text{eV}$ (for the single flavor models) and $m_1=10^{-20.5}\,\text{eV}$, $m_2=10^{-20.2}\,\text{eV}$ for the double flavor, matched model. We discuss how our study will be furthered using a reinforcement learning algorithm.
72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
Contextual Active Online Model Selection with Expert Advice
How can we collect the most useful labels to learn a model selection policy, when presented with arbitrary heterogeneous data streams? In this paper, we formulate this task as a contextual active model selection problem, where at each round the learner receives an unlabeled data point along with a context. The goal is to output the best model for any given context without obtaining an excessive amount of labels. In particular, we focus on the task of selecting pre-trained classifiers, and propose a contextual active model selection algorithm (CAMS), which relies on a novel uncertainty sampling query criterion defined on a given policy class for adaptive model selection. In comparison to prior art, our algorithm does not assume a globally optimal model. We provide rigorous theoretical analysis for the regret and query complexity under both adversarial and stochastic settings. Our experiments on several benchmark classification datasets demonstrate the algorithm’s effectiveness in terms of both regret and query complexity. Notably, to achieve the same accuracy, CAMS incurs less than 10% of the label cost when compared to the best online model selection baselines on CIFAR10.
Liu, Xuefeng↗
Comparison of Socio-Technical Threat Models
Given the adoption of emerging technologies and the increasing complexity of managing such systems with a lifecycle much shorter than that of critical infrastructure systems, there is a practical need to be able to analyze sociotechnical dependencies and their associated evolving risks. Threat models based on social influence techniques can be used to implement adversarial tactics analogous to the cyber kill chain and attested to within the MITRE ATT&CK for ICS framework including Initial Access, Persistence, Collection, and Impact. Furthermore, as with cyber disruptions, the impact of social influence threat models can have an asymmetric impact that is not spatially-localized. Finally, unlike cyber attacks with a reasonably short duration (ransomware takes days to months), social influence based attacks have the potential to persist for much longer as they are based on long-term strategic infrastructure investments within the private sector. Given the increased importance of electric vehicle charging stations as a long-term, strategic infrastructure investment within the Energy and Transportation Sectors, we provide initial results that compare the impact of a Loss of Availability (T0826) realized through cyber and social influence based threat models. The analysis employs techniques from automated reasoning and measures of network complexity to understand evolving dominance of EV payment and charging networks within geographic region of interest. Within this context, we compare the impact of a loss of availability due to ransomware versus that of loss of support due to a merger and acquisition. Results across several different metro areas will be provided.
29 ENERGY PLANNING, POLICY, AND ECONOMY↗
Sub-Pin Fission Product Sensitivity to Assembly and Core Models on High Burnup Fuel and Comparison to Experimental Measurements
Explore the source record for details and available documents.
Conant, Andrew↗
A Comparison of Custom and Standard Neuron Model Random Walks on the Ornstein-Uhlenbeck Equation for Simplified Turbulence
An initial investigation into neuron models aimed at efficient stochastic simulation of turbulent flows.
Smith, John Darby [Sandia National Laboratories (S↗
Speed of Sound Measurements of Binary Mixtures of Hydrofluorocarbons [Pentafluoroethane (R-125), 1,1-Difluoroethane (R-152a), or 1,1,1,2,3,3,3-Heptafluoropropane (R-227ea)] with Hydrofluoroolefins [2,3,3,3-Tetrafluoropropene (R-1234yf) or trans-1,3,3,3-Tetrafluoropropene (R-1234ze(E))]
In this article, speed of sound data measured using a dual-path pulse-echo instrument are reported for three binary refrigerant mixtures, R-125/1234yf, R-1234yf/152a, and R-1234ze(E)/227ea, at compositions of (0.33/0.67) and (0.67/0.33) mole fraction. The speed of sound was studied at temperatures ranging from 230 K to 345 K from pressures slightly above the bubble point curve up to 20 MPa for the mixtures containing R-1234yf and 49 MPa for the R-1234ze(E)/227ea mixtures. The relative combined expanded speed of sound uncertainty ranged from 0.039 % to 0.317 % with a mean uncertainty over all state points of less than 0.10 %. The reported data are compared to the most recent mixture models in REFPROP for each blend studied. Comparisons of available mixture models for the R-125/1234yf and R-1234yf/152a blends exhibit average absolute deviation values ranging from 0.10 % to 0.27 %, and the average absolute deviations for R-1234ze(E)/227ea blends range from 0.62 % to 0.94 %. The comparisons show that only minor adjustments are required to the R-125/1234yf and R-1234yf/152a mixture models to represent the speed of sound data within its uncertainty. However, significant adjustments are needed to improve the current mixture models for the R-1234ze(E)/227ea blend. Deficiencies with the R-1234ze(E)/227ea mixture model are not unexpected since it (1) utilizes a pure-fluid EOS R-1234ze(E) that has been shown to inaccurately represent R-1234ze(E) speed of sound values and (2) uses binary interaction parameters for the chemically similar mixture of R-1234yf/227ea.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Climate Change Projection in the Twenty-First Century Simulated by NIMS-KMA CMIP6 Model Based on New GHGs Concentration Pathways
Abstract The National Institute of Meteorological Sciences-Korea Meteorological Administration (NIMS-KMA) has participated in the Coupled Model Inter-comparison Project (CMIP) and provided long-term simulations using the coupled climate model. The NIMS-KMA produces new future projections using the ensemble mean of KMA Advanced Community Earth system model (K-ACE) and UK Earth System Model version1 (UKESM1) simulations to provide scientific information of future climate changes. In this study, we analyze four experiments those conducted following the new shared socioeconomic pathway (SSP) based scenarios to examine projected climate change in the twenty-first century. Present day (PD) simulations show high performance skill in both climate mean and variability, which provide a reliability of the climate models and reduces the uncertainty in response to future forcing. In future projections, global temperature increases from 1.92 °C to 5.20 °C relative to the PD level (1995–2014). Global mean precipitation increases from 5.1% to 10.1% and sea ice extent decreases from 19% to 62% in the Arctic and from 18% to 54% in the Antarctic. In addition, climate changes are accelerating toward the late twenty-first century. Our CMIP6 simulations are released to the public through the Earth System Grid Federation (ESGF) international data sharing portal and are used to support the establishment of the national adaptation plan for climate change in South Korea.
54 ENVIRONMENTAL SCIENCES↗
Toward electron temperature profiles in hot-dense plasmas from x-ray spectral ensembles
High repetition rate laser systems enable new strategies for diagnosing plasma behavior with large datasets. Here, we define an ensemble technique that relies on randomized targeting of x-ray tracer micro-stripes. On each shot, a high-intensity laser pulse is focused on a solid target with Ti tracer stripes embedded in an Al foil, randomly targeting a micro-stripe, a portion of a stripe, or a gap between stripes. High-resolution, time-integrated x-ray spectrometers capture line emission from the portion of the micro-stripe that is heated to sufficiently high electron temperatures. Accumulation of many such cases is used to construct ensemble distributions of x-ray line intensities that encompass all relative offsets of the laser focus to the micro-stripe centers. Synthetic intensity distributions are likewise generated using collisional-radiative modeling. Bayesian fitting of modeled to measured intensity distributions establishes the most likely radial temperature profiles, enabling comparison to hydrodynamic models and calling into question the cylindrical symmetry of these micro-stripe-embedded systems. Ensemble techniques have significant potential for high-energy-density plasma diagnostics, especially with the advent of high repetition rate experiments.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Bistable dynamics and Hopf bifurcation in a refined model of early stage HIV infection
Recent clinical studies have shown that HIV disease pathogenesis can depend strongly on many factors at the time of transmission, including the strength of the initial viral load and the local availability of CD4+ T-cells. In this article, a new within-host model of HIV infection that incorporates the homeostatic proliferation of T-cells is formulated and analyzed. Due to the effects of this biological process, the influence of initial conditions on the proliferation of HIV infection is further elucidated. The identifiability of parameters within the model is investigated and a local stability analysis, which displays additional complexity in comparison to previous models, is conducted. The current study extends previous theoretical and computational work on the early stages of the disease and leads to interesting nonlinear dynamics, including a parameter region featuring bistability of infectious and viral clearance equilibria and the appearance of a Hopf bifurcation within biologically relevant parameter regimes.
59 BASIC BIOLOGICAL SCIENCES↗
Thermal Extraction and Delivery from Advanced Reactor Systems
This document presents preliminary thermodynamic analysis of thermal extraction and delivery from advanced nuclear reactors. Light-water, high temperature gas, sodium fast, and molten salt reactors are templated and analyzed for thermal extraction from nominal steam production conditions. Uncontrolled, controlled, and partial admission steam extraction methods are described and analyzed to show the differences in system impact from the various methods. Thermal delivery modeling results show comparisons that have become possible through the computational capabilities of the thermal delivery model. Due to the significant degrees of freedom of the problem, specific cases are used to compare the thermodynamic costs of heat transport using various heat transfer fluids. All this work is done using Modelica models, allowing for the future application of models in dynamic analyses beyond the steady-state calculations presented in this report. Appendices describe additional capabilities and methods that can help users analyze complex problems using HYBRID models developed in Modelica.
22 GENERAL STUDIES OF NUCLEAR REACTORS↗
Bioenergy Technologies in Long-Run Climate Change Mitigation: Results from the EMF-33 Study
Bioenergy is expected to play an important role in long-run climate change mitigation strategies as highlighted by many integrated assessment model (IAM) scenarios. These scenarios, however, also show a very wide range of results, with uncertainty about bioenergy conversion technology deployment and biomass feedstock supply. To date, the underlying differences in model assumptions and parameters for the range of results have not been conveyed. Here we explore the models and results of the 33rd study of the Stanford Energy Modeling Forum to elucidate and explore bioenergy technology specifications and constraints that underlie projected bioenergy outcomes. We first develop and report consistent bioenergy technology characterizations and modeling details. We evaluate the bioenergy technology specifications through a series of analyses—comparison with the literature, model intercomparison, and an assessment of bioenergy technology projected deployments. We find that bioenergy technology coverage and characterization varies substantially across models, spanning different conversion routes, carbon capture and storage opportunities, and technology deployment constraints. Still, the range of technology specification assumptions is largely in line with bottom-up engineering estimates. We then find that variation in bioenergy deployment across models cannot be understood from technology costs alone. Important additional determinants include biomass feedstock costs, the availability and costs of alternative mitigation options in and across end-uses, the availability of carbon dioxide removal possibilities, the speed with which large scale changes in the makeup of energy conversion facilities and integration can take place, and the relative demand for different energy services.
bioenergy↗
Demonstration of RELAP5-3D for transient analysis of a dual coolant lead lithium fusion blanket concept
Our research aims to build a multiphysics framework for transient analysis of the Dual- Coolant Lead-Lithium (DCLL) blanket design of the proposed Fusion Nuclear Science Facility (FNSF) for all Onticipated Operational Occurrences (AOOs) using Reduced Order Models (ROMs). This paper presents an assessment of using RELAP5-3D for transient thermal-hydraulic blanket analysis of the DCLL design. Preliminary validation studies for the RELAP5-3D properties were conducted using a representative vertical flow loop model and a comparison study using experimental data from the thermal convection corrosion loops at ORNL. These studies inspected the thermal hydraulic response of Lead-Lithium Eutectic (PbLi) systems within the RELAP5-3D code; mainly pressure drop and heat transfer, as compared to analytical calculations and experimental data. The Magnetohydrodynamic (MHD) pressure drop effect for liquid metal under the effects of a uniform magnetic field was implemented into our models using an equivalent forms loss coefficient based on the phasic momentum equations. We verified our implementation of MHD pressure drop within RELAP5-3D using a systematic approach consisting of two studies. Models were developed to perform comparison studies based on experimental data from the Argonne Liquid metal EXperiment (ALEX) facility and the Magnetohydrodynamic PbLi Experiment (MaPLE) facility. Using the developed validation basis, we built a simplified DCLL channel model within RELAP5-3D, inspected MHD pressure drop and performed transient temperature analysis. The DCLL channel model used radial heating data from the International Thermonuclear Experimental Reactor (ITER) which was normalized to the FNSF design. Here, a representative startup transient was performed using a relative power curve from the initial ramp up of a long-term power pulse of the Tore Supra reactor. This study demonstrated that RELAP5-3D ROMs are capable of accurate transient analysis of the DCLL blanket. We have since utilized MCNP to develop heating profiles based on the FNSF design and are working toward implementing them into our DCLL model.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Dynamics of Aqueous Electrolyte Solutions: Challenges for Simulations
This Perspective article focuses on recent simulation work on the dynamics of aqueous electrolytes. It is well-established that full-charge, nonpolarizable models for water and ions generally predict solution dynamics that are too slow in comparison to experiments. Models with reduced (scaled) charges do better for solution diffusivities and viscosities but encounter issues describing other dynamic phenomena such as nucleation rates of crystals from solution. Polarizable models show promise, especially when appropriately parametrized, but may still miss important physical effects such as charge transfer. First-principles calculations are starting to emerge for these properties that are in principle able to capture polarization, charge transfer, and chemical transformations in solution. Finally, while direct ab initio simulations are still too slow for simulations of large systems over long time scales, machine-learning models trained on appropriate first-principles data show significant promise for accurate and transferable modeling of electrolyte solution dynamics.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Understanding the Drivers of Atlantic Multidecadal Variability using a Stochastic Model Hierarchy
The relative importance of ocean and atmospheric dynamics in generating Atlantic Multidecadal Variability (AMV) remains an open question. Comparisons between climate models with SLAB and fully-dynamic (FULL) ocean components are often used to explore this question, but cannot reveal how individual ocean processes generate these differences. We build a hierarchy of physically interpretable stochastic models to investigate the contribution of two upper-ocean processes to AMV: the role of seasonal variation and mixed-layer entrainment. This interpretability arises from the stochastic model’s simplified representation of sea surface temperature (SST), considering only the local upper ocean response to white-noise atmospheric forcing and its impact on surface heat exchange. We focus on understanding differences between SLAB and FULL non-eddy resolving pre-industrial control simulations of the Community Earth System Model 1 (CESM), and estimate the stochastic model parameters from each respective simulation. Despite its simplicity, the stochastic model reproduces temporal characteristics of SST variability in the SPG, including reemergence, seasonal-to-interannual persistence and power spectra. Furthermore, unrealistically persistent SST of the CESM-SLAB ocean simulation is reproduced in the equivalent stochastic model configuration where the mixed-layer depth (MLD) is constant. The stochastic model also reveals that vertical entrainment primarily damps SST variability, thus explaining why SLAB exhibits larger SST variance than FULL. Here, the stochastic model driven by temporally stochastic, spatially coherent forcing patterns reproduces the canonical AMV pattern. However, the amplitude of low-frequency variability remains underestimated, suggesting a role for ocean dynamics beyond entrainment.
54 ENVIRONMENTAL SCIENCES↗
SCITUNE: Aligning Large Language Models with Human-Curated Scientific Multimodal Instructions
Instruction finetuning is a popular paradigm to align large language models (LLM) with human intent. Despite its popularity, this idea is less explored in improving the LLMs to align existing foundation models with scientific disciplines, concepts and goals. In this work, we present SciTune as a tuning framework to improve the ability of LLMs to follow scientific multimodal instructions. To test our methodology, we use a human-generated scientific instruction tuning dataset and train a large multimodal model LLaMA-SciTune that connects a vision encoder and LLM for science-focused visual and language understanding. LLaMA-SciTune significantly outperforms the state-of-the-art models in the generated figure types and captions in multiple scientific multimodal benchmarks. In comparison to the models that are fine-tuned with machine generated data only, LLaMA-SciTune surpasses human performance on average and in many sub-categories on the ScienceQA benchmark.
• Artificial intelligence (AI) / machine learning ↗
Deep Learning enabled spectral energy conversion for in situ exposure measurements
A detector-specific deep learning (DL) approach is presented for spectra-to-exposure conversion using large-format sodium iodide (NaI(Tl)) detectors deployed for in situ environmental radiation measurements in emergency response scenarios. Accurate determination of exposure from NaI spectra is challenging due to poor energy resolution, partial energy absorption, and the strong sensitivity of traditionally deployed analytical conversion methods to calibrated source geometry and pre-deployment assumptions. Here, to address these limitations, a multi-layer perceptron model was trained on a hybrid in situ /Monte Carlo dataset constructed to span a broad range of photon energies, spatial extents, and realistic deployment variability, representative of general in situ emergency response conditions. The DL model was evaluated against commonly fielded analytical approaches under matched simulation conditions, including a single-factor method, a G-function method, and a modeled pressurized ion chamber (PIC) baseline. This study was intentionally computational in scope to enable controlled, like-for-like comparisons between conversion techniques while minimizing confounding real-world variability. Comparison to the modeled PIC provides contextual benchmarking and is not intended as a field inter-comparison with deployed instruments. Across the evaluated 20 keV to 3 MeV energy range, the DL approach consistently exhibited higher accuracy and reduced variance relative to the analytical methods against a deterministically calculated exposure. This may indicate improved robustness to spectral complexity without reliance on source-, geometric-, or spectral region-specific optimization. While results do not represent real-world validation, the presented work demonstrates that deep learning may effectively learn the nonlinear detector response-to-exposure relationship for asymmetric NaI(Tl) detectors and offers a promising pathway for improving in situ exposure estimation using spectroscopic systems already integrated into initial real-time emergency response operations.
61 RADIATION PROTECTION AND DOSIMETRY↗