Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Generative Model”

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 865 records · Page 48
All sourcesNASA NTRSDOE OSTI
DOE OSTI2015

Next-generation system analysis model for studying alternative Waste Management Systems - 15232

The U.S. Department of Energy's Office of Nuclear Energy (DOE-NE) is conducting planning activities within the Nuclear Fuels Storage and Transportation Planning Project (NFST). The plans involve (1) laying the groundwork for implementing interim storage (including dealing with any associated transportation that is needed) per the Administration's Strategy for the Management and Disposal of Used Nuclear Fuel and High-Level Radioactive Waste of January 2013 and (2) developing a foundation for a new nuclear waste management organization. An overarching goal is to develop options for designing an integrated system that will be considered by decision-makers. Based on the dynamic system analysis and software specification requirements that must be met, Argonne National Laboratory (Argonne) is leading the development of a consolidated waste management system tool known as the Next-Generation System Analysis Model (NGSAM). The development of NGSAM leverages the approaches and methods used previously to develop the Transportation Storage Logistics (TSL) and Total System Model (TSM) system analysis tools, but the new tool takes advantage of advanced computational capabilities. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2023

Modeling Exhaust-Generated Aerodynamic Pressure Loads on Airfield Matting Repair Systems

Airfield matting systems are commonly used for rapid repair of damaged runways to facilitate continuity of critical operations. Under normal service conditions, matting repair systems are subject not only to wheel loads exerted by airfield traffic but also to aerodynamic pressure loads resulting from high-speed, turbulent exhaust plumes produced by jets during taxi and take-off. Matting systems employed by the U.S. Air Force have been tested with respect to wheel loads, but probabilities of failure because of pressure loads produced by jet exhaust have not yet been established. This study presents a numerical approach for preliminary estimation of worst-case matting anchor forces resulting from exhaust-generated pressure loads. Using a two-dimensional computational fluid dynamics model, system behavior is evaluated by means of a parametric study of six system variables, of which the most significant are: (1) distance between engine and matting, (2) depth of cavity openings at matting edges, and (3) engine exhaust velocity. The results demonstrate that matting systems are likely to experience net uplift in typical service scenarios, driven by the combined effects of flow separation and cavity pressurization. Worst-case anchor pull-out forces, computed according to a tributary-area approach, are estimated to fall in the range of 130 lb (581 N) to 979 lb (4,350 N), depending on assumed load-sharing behavior among anchors and the size of the repair site. Field testing of instrumented matting systems during jet taxi and take-off sequences is recommended as the best next step toward understanding system behavior.

Engineering↗
DOE OSTI2025

Reweighting configurations generated by transferable, machine learned models for protein sidechain backmapping

Multiscale modeling requires the linking of models at different levels of detail, with the goal of gaining accelerations from lower fidelity models while recovering fine details from higher resolution models. Communication across resolutions is particularly important in modeling soft matter, where tight couplings exist between molecular-level details and mesoscale structures. While multiscale modeling of biomolecules has become a critical component in exploring their structure and self-assembly, backmapping from coarse-grained to fine-grained, or atomistic, representations presents a challenge, despite recent advances through machine learning. A major hurdle, especially for strategies utilizing machine learning, is that backmappings can only approximately recover the atomistic ensemble of interest. We demonstrate conditions for which backmapped configurations may be reweighted to exactly recover the desired atomistic ensemble. By training separate decoding models for each sidechain type, we develop an algorithm based on normalizing flows and geometric algebra attention to autoregressively propose backmapped configurations for any protein sequence. Critical for reweighting with modern protein force fields, our trained models include all hydrogen atoms in the backmapping and make probabilities associated with atomistic configurations directly accessible. We also demonstrate, however, that reweighting is extremely challenging despite state-of-the-art performance on recently developed metrics and generation of configurations with low energies in atomistic protein force fields. Through detailed analysis of configurational weights, we show that machine-learned backmappings must not only generate configurations with reasonable energies, but also correctly assign relative probabilities under the generative model. These are broadly important considerations in generative modeling of atomistic molecular configurations.

Monroe, Jacob I. [Univ. of Arkansas, Fayetteville,↗
DOE OSTI2024

High-Resolution Meteorology with Climate Change Impacts from Global Climate Model Data Using Generative Machine Learning

As renewable energy generation increases, the impacts of weather and climate on energy generation and demand become critical to the reliability of the energy system. However, these impacts are often overlooked. Global climate models (GCMs) can be used to understand possible changes to our climate, but their coarse resolution makes them difficult to use in energy system modelling. Here we present open-source generative machine learning methods that produce meteorological data at a nominal spatial resolution of 4 km at an hourly frequency based on inputs from 100 km daily-average GCM data. These methods run 40 times faster than traditional downscaling methods and produce data that have high-resolution spatial and temporal attributes similar to historical datasets. We demonstrate that these methods can be used to downscale projected changes in wind, solar and temperature variables across multiple GCMs including projections for more frequent low-wind and high-temperature events in the Eastern United States.

climate change↗
DOE OSTI2010

Modeling of leachate generation from MSW landfills by a 2-dimensional 2-domain approach

The flow of water through Municipal Solid Waste (MSW) landfills is highly non-uniform and dominated by preferential pathways. Thus, concepts to simulate landfill behavior require that a heterogeneous flow regime is considered. Recent models are based on a 2-domain approach, differentiating between channel domain with high hydraulic conductivity, and matrix domain of slow water movement with high water retention capacity. These models focus on the mathematical description of rapid water flow in channel domain. The present paper highlights the importance of water exchange between the two domains, and expands the 1-dimensional, 2-domain flow model by taking into account water flows in two dimensions. A flow field consisting of a vertical path (channel domain) surrounded by the waste mass (matrix domain) is defined using the software HYDRUS-2D. When the new model is calibrated using data sets from a MSW-landfill site the predicted leachate generation corresponds well with the observed leachate discharge. An overall model efficiency in terms of r{sup 2} of 0.76 was determined for a simulation period of almost 4 years. The results confirm that water in landfills follows a preferential path way characterized by high permeability (K{sub s} = 300 m/d) and zero retention capacity, while the bulk of the landfill (matrix domain) is characterized by low permeability (K{sub s} = 0.1 m/d) and high retention capacity. The most sensitive parameters of the model are the hydraulic conductivities of the channel domain and the matrix domain, and the anisotropy of the matrix domain.

42 ENGINEERING↗
DOE OSTI2016

Building energy analysis tool

A building energy analysis system includes a building component library configured to store a plurality of building components, a modeling tool configured to access the building component library and create a building model of a building under analysis using building spatial data and using selected building components of the plurality of building components stored in the building component library, a building analysis engine configured to operate the building model and generate a baseline energy model of the building under analysis and further configured to apply one or more energy conservation measures to the baseline energy model in order to generate one or more corresponding optimized energy models, and a recommendation tool configured to assess the one or more optimized energy models against the baseline energy model and generate recommendations for substitute building components or modifications.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2019

Analyzing machine learning models to accelerate generation of fundamental materials insights

Abstract Machine learning for materials science envisions the acceleration of basic science research through automated identification of key data relationships to augment human interpretation and gain scientific understanding. A primary role of scientists is extraction of fundamental knowledge from data, and we demonstrate that this extraction can be accelerated using neural networks via analysis of the trained data model itself rather than its application as a prediction tool. Convolutional neural networks excel at modeling complex data relationships in multi-dimensional parameter spaces, such as that mapped by a combinatorial materials science experiment. Measuring a performance metric in a given materials space provides direct information about (locally) optimal materials but not the underlying materials science that gives rise to the variation in performance. By building a model that predicts performance (in this case photoelectrochemical power generation of a solar fuels photoanode) from materials parameters (in this case composition and Raman signal), subsequent analysis of gradients in the trained model reveals key data relationships that are not readily identified by human inspection or traditional statistical analyses. Human interpretation of these key relationships produces the desired fundamental understanding, demonstrating a framework in which machine learning accelerates data interpretation by leveraging the expertize of the human scientist. We also demonstrate the use of neural network gradient analysis to automate prediction of the directions in parameter space, such as the addition of specific alloying elements, that may increase performance by moving beyond the confines of existing data.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Data-scarce surrogate modeling of shock-induced pore collapse process

Understanding the mechanisms of shock-induced pore collapse is of great interest in various disciplines in sciences and engineering, including materials science, biological sciences, and geophysics. However, numerical modeling of the complex pore collapse processes can be costly. To this end, a strong need exists to develop surrogate models for generating economic predictions of pore collapse processes. Here, in this work, we study the use of a data-driven reduced-order model, namely dynamic mode decomposition, and a deep generative model, namely conditional generative adversarial networks, to resemble the numerical simulations of the pore collapse process at representative training shock pressures. Since the simulations are expensive, the training data are scarce, which makes training an accurate surrogate model challenging. To overcome the difficulties posed by the complex physics phenomena, we make several crucial treatments to the plain original form of the methods to increase the capability of approximating and predicting the dynamics. In particular, physics information is used as indicators or conditional inputs to guide the prediction. In realizing these methods, the training of each dynamic mode composition model takes only around 30 s on CPU. In contrast, training a generative adversarial network model takes 8 h on GPU. Moreover, using dynamic mode decomposition, the final-time relative error is around 0.3% in the reproductive cases. We also demonstrate the predictive power of the methods at unseen testing shock pressures, where the error ranges from 1.3 to 5% in the interpolatory cases and 8 to 9% in extrapolatory cases.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2014

Palm: Easing the Burden of Analytical Performance Modeling

Analytical (predictive) application performance models are critical for diagnosing performance-limiting resources, optimizing systems, and designing machines. Creating models, however, is difficult because they must be both accurate and concise. To ease the burden of performance modeling, we developed Palm, a modeling tool that combines top-down (human-provided) semantic insight with bottom-up static and dynamic analysis. To express insight, Palm defines a source code modeling annotation language. By coordinating models and source code, Palm's models are `first-class' and reproducible. Unlike prior work, Palm formally links models, functions, and measurements. As a result, Palm (a) uses functions to either abstract or express complexity (b) generates hierarchical models (representing an application's static and dynamic structure); and (c) automatically incorporates measurements to focus attention, represent constant behavior, and validate models. We discuss generating models for three different applications.

annotation languages↗
DOE OSTI2025

Using Machine Learning to Generate a GISS ModelE Calibrated Physics Ensemble (CPE)

A neural network (NN) surrogate of the NASA GISS ModelE atmosphere (version E3) is trained on a perturbed parameter ensemble (PPE) spanning 45 physics parameters and 36 outputs. The NN is leveraged in a Markov Chain Monte Carlo (MCMC) Bayesian parameter inference framework to generate a second posterior constrained ensemble coined a “calibrated physics ensemble,” or CPE. The CPE members are characterized by diverse parameter combinations and are, by definition, close to top-of-atmosphere radiative balance, and must broadly agree with numerous hydrologic, energy cycle and radiative forcing metrics simultaneously. Global observations of numerous cloud, environment, and radiation properties (provided by global satellite products) are crucial for CPE generation. The inference framework explicitly accounts for discrepancies (or biases) in satellite products during CPE generation. We demonstrate that product discrepancies strongly impact calibration of important model parameter settings (e.g., convective plume entrainment rates; fall speed for cloud ice). Structural improvements new to E3 are retained across CPE members (e.g., stratocumulus simulation). Notably, the framework improved the simulation of shallow cumulus and Amazon rainfall while not degrading radiation fields, an upgrade that neither default parameters nor Latin Hypercube parameter searching achieved. Analyses of the initial PPE suggested several parameters were unimportant for output variation. However, many “unimportant” parameters were needed for CPE generation, a result that brings to the forefront how parameter importance should be determined in PPEs. From the CPE, two diverse 45-dimensional parameter configurations are retained to generate radiatively-balanced, auto-tuned atmospheres that were used in two E3 submissions to CMIP6.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Hypothesis testing via AI: Generating physically interpretable models of scientific data with machine learning (Full Technical Report)

Deep learning has demonstrated an exceptional ability to solve complex tasks (an engineering success); however, it has done so at the expense of the ability to generate new knowledge (a scientific failure). We propose an alternative framework—entitled Deep Symbolic Regression (DSR)—in which artificial neural networks (NNs) rapidly generate hypotheses about physical relationships among inputs. This framework bypasses the need to interpret an NN altogether, while still leveraging the representational power of deep learning. The resulting models are tractable mathematical expressions, which are inherently and readily human interpretable and can provide insights into underlying physical phenomena. Further, we fold this methodology into the scientific process by allowing the scientist to directly integrate a priori knowledge and beliefs to accelerate learning. We demonstrate this methodology on symbolic regression—the problem of rediscovering underlying expressions describing a dataset—and achieve state-of-the-art performance across a wide variety of symbolic regression problems. Further, we generalize our DSR framework to apply to the more general class of symbolic optimization problems, in which one seeks to optimize a sequence of symbols or “tokens” under a black-box reward function. Examples of other symbolic optimization problems include neural architecture search and computational antibody design. Our generalized tool, Deep Symbolic Optimization (DSO), has been demonstrated on the task of learning symbolic control policies for reinforcement learning environments, and has been adopted as an enabling capability for computational antibody design.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2010

Higgs boson pair production in new physics models at hadron, lepton, and photon colliders

We study Higgs boson pair production processes at future hadron and lepton colliders including the photon collision option in several new physics models; i.e., the two-Higgs-doublet model, the scalar leptoquark model, the sequential fourth generation fermion model and the vectorlike quark model. Cross sections for these processes can deviate significantly from the standard model predictions due to the one-loop correction to the triple Higgs boson coupling constant. For the one-loop induced processes such as gg{yields}hh and {gamma}{gamma}{yields}hh, where h is the (lightest) Higgs boson and g and {gamma} respectively represent a gluon and a photon, the cross sections can also be affected by new physics particles via additional one-loop diagrams. In the two-Higgs-doublet model and scalar leptoquark models, cross sections of e{sup +}e{sup -}{yields}hhZ and {gamma}{gamma}{yields}hh can be enhanced due to the nondecoupling effect in the one-loop corrections to the triple Higgs boson coupling constant. In the sequential fourth generation fermion model, the cross section for gg{yields}hh becomes very large because of the loop effect of the fermions. In the vectorlike quark model, effects are small because the theory has decoupling property. Measurements of the Higgs boson pair production processes can be useful to explore new physics through the determination of the Higgs potential.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2025

High harmonic generation in Haldane model quantum dots

We study theoretically the nonlinear electron dynamics of Haldane model quantum dots placed in the field of an ultrashort optical pulse. One of the tuning parameters of the Haldane model is the phase accumulated by an electron during its transfer between the next nearest neighbor sites. We study how this parameter affect the nonlinear electron dynamics and the generation of high optical harmonics. With increasing the phase from its zero value to 90°, the low-energy electron states in the conduction and valence bands become more localized near the edges of the quantum dot resulting in suppression of the band gap, enhancement of the dipole inter-band coupling, and strong suppression of the average low-energy electron density of the states. As a result, the nonlinear response of the electron system of a Haldane model quantum dot is the strongest at intermediate values of the phase, ≈ 30− 40°. At these values of the phase, the electron dynamics is irreversible and a few first high-order harmonics have the largest intensities. Here, when the phase approaches 90° value the generation of high harmonics is strongly suppressed.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2022

Adaptive Generation of Training Data for ML Reduced Model Creation

Machine learning proxy models are often used to speed up or completely replace complex computational models. The greatly reduced and deterministic computational costs enable new use cases such as digital twin control systems and global optimization. The challenge of building these proxy models is generating the training data. A naive uniform sampling of the input space can result in a non-uniform sampling of the output space of a model. This can cause gaps in the training data coverage that can miss finer scale details resulting in poor accuracy. While larger and larger data sets could eventually fill in these gaps, the computational burden of full-scale simulation codes can make this prohibitive. In this paper, we present an adaptive data generation method that utilizes uncertainty estimation to identify regions where training data should be augmented. By targeting data generation to areas of need, representative data sets can be generated efficiently. The effectiveness of this method will be demonstrated on a simple one-dimensional function and a complex multidimensional physics model.

Cianciosa, Mark↗
DOE OSTI2019

Benchmarking the accuracy of coverage-dependent models: adsorption and desorption of benzene on Pt (111) and Pt 3 Sn (111) from first principles

Bimetallic catalysts have demonstrated properties favorable for upgrading biofuel through catalytic hydrodeoxygenation. However, the design and optimization of such bimetallic catalysts requires the ability to construct accurate, predictive models of these systems. To generate a model that predicts the kinetic behavior of benzene adsorbed on Pt (111) and a Pt 3 Sn (111) surface alloy (Pt 3 Sn (111)), the adsorption of benzene was studied for a wide range of benzene coverages on both surfaces using density functional theory (DFT) calculations. The adsorption energy of benzene was found to correlate linearly with benzene coverage on Pt (111) and Pt 3 Sn (111); both surfaces exhibited net repulsive lateral interactions. Through an analysis of the d-band properties of the metal surface, it was determined that the coverage dependence is a consequence of the electronic interactions between benzene and the surface. The linear coverage dependence of the adsorption energy allowed us to quantify the influence of the lateral interactions on the heat of adsorption and temperature programmed desorption (TPD) spectra using a mean-field model. A comparison of our simulated TPD to experiment showed that this mean-field model adequately reproduces the desorption behavior of benzene on Pt (111) and Pt 3 Sn (111). In particular, the TPD correctly exhibits a broadening desorption peak as the initial coverage of benzene increases on Pt (111) as well as and a low temperature desorption peak on Pt 3 Sn (111). Furthermore, due to the sensitivity of the TPD peak temperature to the desorption energy, precise alignment of experimental and theoretical TPD spectra demands an accurate calculation of the adsorption energy. Therefore, an analysis of the effect of the exchange-correlation functional on TPD modeling is presented. Through this work, we show the necessity of incorporating lateral interactions into theoretical models in order to correctly predict experimental behavior.

09 BIOMASS FUELS↗