Search NASA⌕ Search

SEARCH · Search NASA

Results for “Network Models”

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 703 records · Page 39

Using Real-Time Electricity Prices to Leverage Electrical Energy Storage and Flexible Loads in a Smart Grid Environment Utilizing Machine Learning Techniques

With exposure to real-time market pricing structures, consumers would be incentivized to invest in electrical energy storage systems and smart predictive automation of their home energy systems. Smart home automation through optimizing HVAC (heating, ventilation, and air conditioning) temperature set points, along with distributed energy storage, could be utilized in the process of optimizing the operation of the electric grid. Using electricity prices as decision variables to leverage electrical energy storage and flexible loads can be a valuable tool to optimize the performance of the power grid and reduce electricity costs both on the supply and demand sides. Energy demand prediction is important for proper allocation and utilization of the available resources. Manipulating energy prices to leverage storage and flexible loads through these demand prediction models is a novel idea that needs to be studied. In this paper, different models for proactive prediction of the energy demand for an entire city using different machine learning techniques are presented and compared. The results of the machine learning techniques show that the proposed nonlinear autoregressive with exogenous inputs neural network model resulted in the most accurate predictions. These prediction models pave the way for the demand side to become an important asset for grid regulation by responding to variable price signals through battery energy storage and passive thermal energy storage using HVAC temperature set points.

25 ENERGY STORAGE↗

SympGNNs: Symplectic Graph Neural Networks for identifying high-dimensional Hamiltonian systems and node classification

Existing neural network models to learn Hamiltonian systems, such as SympNets, although accurate in low-dimensions, struggle to learn the correct dynamics for high-dimensional many-body systems. Herein, we introduce Symplectic Graph Neural Networks (SympGNNs) that can effectively handle system identification in high-dimensional Hamiltonian systems, as well as node classification. SympGNNs combine symplectic maps with permutation equivariance, a property of graph neural networks. Specifically, we propose two variants of SympGNNs: (i) G-SympGNN and (ii) LA-SympGNN, arising from different parameterizations of the kinetic and potential energy. We demonstrate the capabilities of SympGNN on two physical examples: a 40-particle coupled Harmonic oscillator, and a 2000-particle molecular dynamics simulation in a two-dimensional Lennard-Jones potential. Furthermore, we demonstrate the performance of SympGNN in the node classification task, achieving accuracy comparable to the state-of-the-art. Finally, we also empirically show that SympGNN can overcome the oversmoothing and heterophily problems, two key challenges in the field of graph neural networks.

Deep learning↗

DESCRIPTION OF MODELING ANALYSES IN SUPPORT OF THE 200-ZP-1 REMEDIAL DESIGN/REMEDIAL ACTION

The Feasibility Study/or the 200-ZP-1 Groundwater Operable Unit (DOE/RL-2007-28) and the Proposed Plan/or Remediation of the 200-ZP-1 Groundwater Operable Unit (DOE/RL-2007-33) describe the use of groundwater pump-and-treat technology for the 200-ZP-1 Groundwater Operable Unit (OU) as part of an expanded groundwater remedy. During fiscal year 2008 (FY08), a groundwater flow and contaminant transport (flow and transport) model was developed to support remedy design decisions at the 200-ZP-1 OU. This model was developed because the size and influence of the proposed 200-ZP-1 groundwater pump-and-treat remedy will have a larger areal extent than the current interim remedy, and modeling is required to provide estimates of influent concentrations and contaminant mass removal rates to support the design of the aboveground treatment train. The 200 West Area Pre-Conceptual Design/or Final Extraction/Injection Well Network: Modeling Analyses (DOE/RL-2008-56) documents the development of the first version of the MODFLOW/MT3DMS model of the Hanford Site's Central Plateau, as well as the initial application of that model to simulate a potential well field for the 200-ZP-1 remedy (considering only the contaminants carbon tetrachloride and technetium-99). This document focuses on the use of the flow and transport model to identify suitable extraction and injection well locations as part of the 200 West Area 200-ZP-1 Pump-and-Treat Remedial Design/Remedial Action Work Plan (DOEIRL-2008-78). Currently, the model has been developed to the extent necessary to provide approximate results and to lay a foundation for the design basis concentrations that are required in support of the remedial design/remediation action (RD/RA) work plan. The discussion in this document includes the following: (1) Assignment of flow and transport parameters for the model; (2) Definition of initial conditions for the transport model for each simulated contaminant of concern (COC) (i.e., carbon tetrachloride, technetium-99, iodine-129, nitrate [as NO{sub 3}], trichloroethene [TCE], total chromium, tritium), plus uranium; (3) Assumptions underlying the predictive simulations, including the phased implementation of the final full remedy; (4) Approximate number, locations, and rates of extraction and injection wells; and (5) Predicted amounts of contaminant mass extracted and influent concentrations at individual extraction wells for each COC and for uranium. This document is a companion report to pre-conceptual design document (DOE/RL-2008-56). Together these documents describe the sequential, progressive development of the modeling analyses and design basis for the 200-ZP-1 OU remedy.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Open Call LDRD: Physically Informed Autoencoders for Galactic Redshift Regression

Physical constraints have been suggested to make neural network models more generalizable, act scientifically plausible, and be more data-efficient over unconstrained baselines. In this report, we present preliminary work on evaluating the effects of adding soft physical constraints to computer vision neural networks trained to estimate the conditional density of redshift on input galaxy images for the Sloan Digital Sky Survey. We introduce physically motivated soft constraint terms that are not implemented with differential or integral operators. We frame this work as a simple ablation study where the effect of including soft physical constraints is compared to an unconstrained baseline. We compare networks using standard point estimate metrics for photometric redshift estimation, as well as metrics to evaluate how faithful our conditional density estimate represents the probability over the ensemble of our test dataset. We find no evidence that the implemented soft physical constraints are more effective regularizers than augmentation.

97 MATHEMATICS AND COMPUTING↗

Decoding crystallography from high-resolution electron imaging and diffraction datasets with deep learning

While machine learning has been making enormous strides in many technical areas, it is still massively underused in transmission electron microscopy. To address this, a convolutional neural network model was developed for reliable classification of crystal structures from small numbers of electron images and diffraction patterns with no preferred orientation. Diffraction data containing 571,340 individual crystals divided among seven families, 32 genera, and 230 space groups were used to train the network. Despite the highly imbalanced dataset, the network narrows down the space groups to the top two with over 70% confidence in the worst case and up to 95% in the common cases. As examples, we benchmarked against alloys to two-dimensional materials to cross-validate our deep-learning model against high-resolution transmission electron images and diffraction patterns. We present this result both as a research tool and deep-learning application for diffraction analysis.

97 MATHEMATICS AND COMPUTING↗

Ab initio Modeling of the Electronic Structures and Physical Properties of a‐Si 1− x Ge x O 2 Glass ( x = 0 to 1)

The amorphous silica (a‐SiO 2 ) and germania (a‐GeO 2 ) have a wide range of applications in glass industry. Based on a previously constructed near‐perfect continuous random network model with 1296 atoms and periodic boundary conditions, we extend our study to amorphous Si 1− x G e x O 2 models of homogeneous random substitution of Si by Ge with x ranging from 0 to 1. We have calculated the structural, electronic, mechanical, and optical properties for the series by using the first‐principles density functional theory methods. The x ‐dependence of the variations in the properties is analyzed and critically compared with available experimental data. The mass density, volume, total bond order density, bulk mechanical properties, and refractive index are found to vary linearly as a function of x . For x = 0.5, we have also constructed six different kinds of particle immersion models to test the effect of inclusion of spherical particles of one glass of different sizes in the medium of the other glass on their physical properties. It is shown that particle sizes do affect the properties of particle immersion. Our calculations provide deep insight on the properties of mixture and nanocomposites of a‐SiO 2 and a‐GeO 2 glasses.

Baral, Khagendra↗

Using an Advanced Distribution Management System Test Bed to Evaluate the Impact of Model Quality on Volt/VAR Optimization: Preprint

In this paper, we present a test bed for evaluating existing and future advanced distribution management system (ADMS) applications in a realistic laboratory setting, including other utility management systems and field equipment. We present an example of using it to evaluate the impact of the ADMS network model quality on a Volt/VAR optimization (VVO) application. The test bed integrates a commercial ADMS with a real-time simulation model of a utility distribution feeder. Representative power and controller hardware are integrated through hardware-in-the-loop (HIL) techniques. The performance of the ADMS VVO application is also evaluated for different levels of measurement density. Initial results indicate that a higher model quality achieves the highest possible energy savings while avoiding voltage violations, whereas a lower model quality results in increased energy savings but at the expense of more voltage violations.

ADMS↗

Steady State Heat Deposits Modeling in the Nb3Sn Quadrupole Magnets for the Upgrade of the LHC Inner Triplet

In hadron colliders such as the LHC, the energy deposited in the superconductors by the particles lost from the beams or coming from the collision debris may provoke quenches detrimental to the accelerator operation. In previous papers, a Network Model has been used to study the thermodynamic behavior of magnet coils and to calculate the quench levels in the LHC magnets for expected beam loss profiles. This model was subsequently used for thermal analysis and design optimization of Nb{sub 3}Sn quadrupole magnets, which LARP (US LHC Accelerator Research Program) is developing for possible use in the LHC luminosity upgrade. For these new magnets, the heat transport efficiency from the coil to the helium bath needs to be determined and optimized. In this paper the study of helium cooling channels and the heat evacuation scheme are presented and discussed.

43 PARTICLE ACCELERATORS↗

Supercritical fluids behave as complex networks

Abstract Supercritical fluids play a key role in environmental, geological, and celestial processes, and are of great importance to many scientific and engineering applications. They exhibit strong variations in thermodynamic response functions, which has been hypothesized to stem from the microstructural behavior. However, a direct connection between thermodynamic conditions and the microstructural behavior, as described by molecular clusters, remains an outstanding issue. By utilizing a first-principles-based criterion and self-similarity analysis, we identify energetically localized molecular clusters whose size distribution and connectivity exhibit self-similarity in the extended supercritical phase space. We show that the structural response of these clusters follows a complex network behavior whose dynamics arises from the energetics of isotropic molecular interactions. Furthermore, we demonstrate that a hidden variable network model can accurately describe the structural and dynamical response of supercritical fluids. These results highlight the need for constitutive models and provide a basis to relate the fluid microstructure to thermodynamic response functions.

42 ENGINEERING↗

A graph-based network for predicting chemical reaction pathways in solid-state materials synthesis

Abstract Accelerated inorganic synthesis remains a significant challenge in the search for novel, functional materials. Many of the principles which enable “synthesis by design” in synthetic organic chemistry do not exist in solid-state chemistry, despite the availability of extensive computed/experimental thermochemistry data. In this work, we present a chemical reaction network model for solid-state synthesis constructed from available thermochemistry data and devise a computationally tractable approach for suggesting likely reaction pathways via the application of pathfinding algorithms and linear combination of lowest-cost paths in the network. We demonstrate initial success of the network in predicting complex reaction pathways comparable to those reported in the literature for YMnO 3 , Y 2 Mn 2 O 7 , Fe 2 SiS 4 , and YBa 2 Cu 3 O 6.5 . The reaction network presents opportunities for enabling reaction pathway prediction, rapid iteration between experimental/theoretical results, and ultimately, control of the synthesis of solid-state materials.

36 MATERIALS SCIENCE↗

Toward transferable empirical valence bonds: Making classical force fields reactive

The empirical valence bond technique allows classical force fields to model reactive processes. However, parametrization from experimental data or quantum mechanical calculations is required for each reaction present in the simulation. We show that the parameters present in the empirical valence bond method can be predicted using a neural network model and the SMILES strings describing a reaction. This removes the need for quantum calculations in the parametrization of the empirical valence bond technique. In doing so, we have taken the first steps toward defining a new procedure for enabling reactive atomistic simulations. This procedure would allow researchers to use existing classical force fields for reactive simulations, without performing additional quantum mechanical calculations.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Codon2Vec v1.0

Background: Codon2Vec is an embedding neural network that predicts 'high' or 'low' gene expression directly from the protein-coding sequences. Embedding neural networks are commonly used for natural language processing (NLP) applications. Analogous to how an English sentence is a string of words, a gene can be thought of as a string of codons. Similar to how NLP neural networks model English sentences as a non-random sequence of words, we considered a coding sequence as a non-random non-overlapping array of codons (k-mers of length = 3). Value Proposition: - Codon2Vec achieved a high median AUC-ROC score of 83.8% when trained and applied to transcriptomic data from 300 fungal species - Unlike Codo2Vec, conventional methods predicting for expression based on codon usage rely on a priori knowledge of optimal codons or a set of reference genes. - Unlike Codon2vec, these methods do not account for the effect of codon order on gene expression. - Codon2Vec neural network bypasses the need for artisanal feature selection step that is necessary for traditional machine learning models.

Wint, Rhondene↗

Generic Crystalline Disposal Reference Case

A generic reference case for disposal of spent nuclear fuel and high-level radioactive waste in crystalline rock is outlined. The generic cases are intended to support development of disposal system modeling capability by establishing relevant baseline conditions and parameters. Establishment of a generic reference case requires that the emplacement concept, waste inventory, waste form, waste package, backfill/buffer properties, EBS failure scenarios, host rock properties, and biosphere be specified. The focus in this report is on those elements that are unique to crystalline disposal, especially the geosphere representation. Three emplacement concepts are suggested for further analyses: a waste packages containing 4 PWR assemblies emplaced in boreholes in the floors of tunnels (KBS-3 concept), a 12-assembly waste package emplaced in tunnels, and a 32-assembly dual purpose canister emplaced in tunnels. In addition, three failure scenarios were suggested for future use: a nominal scenario involving corrosion of the waste package in the tunnel emplacement concepts, a manufacturing defect scenario applicable to the KBS-3 concept, and a disruptive glaciation scenario applicable to both emplacement concepts. The computational approaches required to analyze EBS failure and transport processes in a crystalline rock repository are similar to those of argillite/shale, with the most significant difference being that the EBS in a crystalline rock repository will likely experience highly heterogeneous flow rates, which should be represented in the model. The computational approaches required to analyze radionuclide transport in the natural system are very different because of the highly channelized nature of fracture flow. Computational workflows tailored to crystalline rock based on discrete transport pathways extracted from discrete fracture network models are recommended.

Environmental Protection↗

CeMbio - The Caenorhabditis elegans Microbiome Resource

The study of microbiomes by sequencing has revealed a plethora of correlations between microbial community composition and various life-history characteristics of the corresponding host species. However, inferring causation from correlation is often hampered by the sheer compositional complexity of microbiomes, even in simple organisms. Synthetic communities offer an effective approach to infer cause-effect relationships in host-microbiome systems. Yet the available communities suffer from several drawbacks, such as artificial (thus non-natural) choice of microbes, microbe-host mismatch (e.g., human microbes in gnotobiotic mice), or hosts lacking genetic tractability. Here we introduce CeMbio, a simplified natural Caenorhabditis elegans microbiota derived from our previous meta-analysis of the natural microbiome of this nematode. The CeMbio resource is amenable to all strengths of the C. elegans model system, strains included are readily culturable, they all colonize the worm gut individually, and comprise a robust community that distinctly affects nematode life-history. Several tools have additionally been developed for the CeMbio strains, including diagnostic PCR primers, completely sequenced genomes, and metabolic network models. With CeMbio, we provide a versatile resource and toolbox for the in-depth dissection of naturally relevant host-microbiome interactions in C. elegans.

59 BASIC BIOLOGICAL SCIENCES↗

Distributed output-feedback model predictive control for multi-agent consensus

We introduce a distributed output-feedback model predictive control approach for achieving consensus among multiple agents. Each agent computes a distributed control action based on an output-feedback measurement of a local neighborhood tracking error and communicates information only to its neighbors, according to a communication network modeled as a directed graph. Each agent computes its distributed control action by solving a local min–max optimization problem that simultaneously computes a local state estimate and control input under worst-case assumptions on unmeasured input disturbances and measurement noise. Under easily verified controllability and observability assumptions, this distributed output-feedback model predictive control approach provides an upper bound on the group consensus error, thereby ensuring practical consensus in the presence of unmeasured disturbances and noise. A numerical example with four agents connected in a directed graph is demonstrated to illustrate the results.

42 ENGINEERING↗

A Unified Workflow for Sensitivity-Based Kinetic Analysis in Microkinetic Models

Degrees of rate control (DRC), apparent activation energies, and apparent reaction orders are established local sensitivity diagnostics for interpreting microkinetic models, but applying them routinely to large mechanisms often requires substantial reaction-specific bookkeeping, perturbation design, and postprocessing. Here, in this study, we present a unified derivative-based workflow that evaluates these quantities from a single compiled reaction-network model and target-rate definition. For any user-provided microkinetic model, the workflow compiles the mechanism into stoichiometrically consistent mass-action rate equations, solves the surface dynamics, and uses automatic differentiation to compute sensitivities with respect to rate constants, temperature, and gas partial pressures. By combining their calculations in the same framework, the workflow clearly demonstrates the relationships between different DRCs and the apparent activation energy. Using existing examples of propylene partial oxidation and methane oxidation on Pd(100), we verify expected transient redistribution of rate control, distinguish net Campbell DRCs from one-sided directional sensitivities, and show how apparent activation energy can be reconstructed either from one-sided DRCs or from state-based DRCs while critical mechanistic insights are obtained consistently. In the methane oxidation case, a pathway-subset test further illustrates how a simplified mechanism preserves key kinetic signatures of a full model, showing the potential of our user-friendly tool for model construction beyond kinetic analysis.

36 MATERIALS SCIENCE↗

Predicting wind-driven spatial deposition through simulated color images using deep autoencoders

Abstract For centuries, scientists have observed nature to understand the laws that govern the physical world. The traditional process of turning observations into physical understanding is slow. Imperfect models are constructed and tested to explain relationships in data. Powerful new algorithms can enable computers to learn physics by observing images and videos. Inspired by this idea, instead of training machine learning models using physical quantities, we used images, that is, pixel information. For this work, and as a proof of concept, the physics of interest are wind-driven spatial patterns. These phenomena include features in Aeolian dunes and volcanic ash deposition, wildfire smoke, and air pollution plumes. We use computer model simulations of spatial deposition patterns to approximate images from a hypothetical imaging device whose outputs are red, green, and blue (RGB) color images with channel values ranging from 0 to 255. In this paper, we explore deep convolutional neural network-based autoencoders to exploit relationships in wind-driven spatial patterns, which commonly occur in geosciences, and reduce their dimensionality. Reducing the data dimension size with an encoder enables training deep, fully connected neural network models linking geographic and meteorological scalar input quantities to the encoded space. Once this is achieved, full spatial patterns are reconstructed using the decoder. We demonstrate this approach on images of spatial deposition from a pollution source, where the encoder compresses the dimensionality to 0.02% of the original size, and the full predictive model performance on test data achieves a normalized root mean squared error of 8%, a figure of merit in space of 94% and a precision-recall area under the curve of 0.93.

54 ENVIRONMENTAL SCIENCES↗

Deep learning the properties of inorganic perovskites

We report the ability to accurately and quickly predict the stability of materials and their structural and electronic properties remains a grand challenge in materials science. Density functional theory is widely used as a means of predicting these material properties, but is known to be computationally expensive and scales as the cube of the number of electrons in the material’s unit cell. In this article, for a previously published dataset of inorganic perovskites, we show that a single neural network model using only the elemental properties of the compounds’ constituents can predict lattice constants to within 0.1 Å, heat of formation to within 0.2 eV, and band gaps to within 0.7 eV RMSE. We also compare the performance of the trained network to two widely used regression techniques, namely random forest and Kernel ridge regression, and find that the neural network’s predictions are more accurate for each of the properties. The simultaneous accurate prediction of multiple key properties of technologically relevant materials is promising for rational design and optimization in known and novel chemical spaces.

36 MATERIALS SCIENCE↗