Search NASA⌕ Search

SEARCH · Search NASA

Results for “PROPOSALS”

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 1,657 records · Page 92

Reliable extrapolation of deep neural operators informed by physics or sparse observations

Deep neural operators can learn nonlinear mappings between infinite-dimensional function spaces via deep neural networks. As promising surrogate solvers of partial differential equations (PDEs) for real-time prediction, deep neural operators such as deep operator networks (DeepONets) provide a new simulation paradigm in science and engineering. Pure data-driven neural operators and deep learning models, in general, are usually limited to interpolation scenarios, where new predictions utilize inputs within the support of the training set. However, in the inference stage of real-world applications, the input may lie outside the support, i.e., extrapolation is required, which may result to large errors and unavoidable failure of deep learning models. Here, we address this challenge of extrapolation for deep neural operators. First, we systematically investigate the extrapolation behavior of DeepONets by quantifying the extrapolation complexity, via the 2-Wasserstein distance between two function spaces and propose a new strategy of bias–variance trade-off for extrapolation with respect to model capacity. Subsequently, we develop a complete workflow, including extrapolation determination, and we propose five reliable learning methods that guarantee a safe prediction under extrapolation by requiring additional information—the governing PDEs of the system or sparse new observations. The proposed methods are based on either fine-tuning a pre-trained DeepONet or multifidelity learning. We demonstrate the effectiveness of the proposed framework for various types of parametric PDEs. Furthermore, our systematic comparisons provide practical guidelines for selecting a proper extrapolation method depending on the available information, desired accuracy, and required inference speed.

42 ENGINEERING↗

Scalable Bayesian optimization with randomized prior networks

Several fundamental problems in science and engineering consist of global optimization tasks involving unknown high-dimensional (black-box) functions that map a set of controllable variables to the outcomes of an expensive experiment. Bayesian Optimization (BO) techniques are known to be effective in tackling global optimization problems using a relatively small number objective function evaluations, but their performance suffers when dealing with high-dimensional outputs. To overcome the major challenge of dimensionality, here we propose a deep learning framework for BO and sequential decision making based on bootstrapped ensembles of neural architectures with randomized priors. Using appropriate architecture choices, we show that the proposed framework can approximate functional relationships between design variables and quantities of interest, even in cases where the latter take values in high-dimensional vector spaces or even infinite-dimensional function spaces. In the context of BO, we augmented the proposed probabilistic surrogates with re-parameterized Monte Carlo approximations of multiple-point (parallel) acquisition functions, as well as methodological extensions for accommodating black-box constraints and multi-fidelity information sources. We test the proposed framework against state-of-the-art methods for BO and demonstrate superior performance across several challenging tasks with high-dimensional outputs, including a constrained multi-fidelity optimization task involving shape optimization of rotor blades in turbo-machinery.

97 MATHEMATICS AND COMPUTING↗

Gradient flow based phase-field modeling using separable neural networks

Allen–Cahn equation is a reaction–diffusion equation and is widely used for modeling phase separation. Machine learning methods for solving the Allen–Cahn equation in its strong form suffer from inaccuracies in collocation techniques, errors in computing higher-order spatial derivatives, and the large system size required by the space–time approach. To overcome these challenges, we propose solving the gradient flow of the Ginzburg–Landau free energy functional, which is equivalent to the Allen–Cahn equation, thereby avoiding the second-order spatial derivatives associated with the Allen–Cahn equation. A minimizing movement scheme is employed to solve the gradient flow problem, eliminating the complexities of a space–time approach. We utilize a separable neural network that efficiently represents the phase field through low-rank tensor decomposition. As we use the minimizing movement scheme to numerically solve the gradient flow problem, we thus, refer to the proposed method as the Separable Deep Minimizing Movement (SDMM) method. The evaluation of the functional in the minimizing movement scheme using the Gauss quadrature technique bypasses the inaccuracies associated with collocation techniques traditionally used to solve partial differential equations. A hyperbolic tangent transformation is introduced on the phase field prior to the evaluation of the functional to ensure that it remains strictly bounded within the values of the two phases. For this transformation, theoretical guarantee for energy stability of the minimizing movement scheme is established. Our results suggest that this transformation helps to improve the accuracy and efficiency significantly. The proposed method resolves the challenges faced by state-of-the-art machine learning techniques, outperforming them in both accuracy and efficiency. It is also the first machine learning method to achieve an order of magnitude speed improvement over the finite element method. In addition to its formulation and computational implementation, several case studies illustrate the applicability of the proposed method.

42 ENGINEERING↗

Numerical simulation of emulsified fuel spray combustion with puffing and micro-explosion

The purpose of this study was to develop numerical simulation of spray combustion of emulsified fuel with considering puffing and micro-explosion. First, a mathematical model for puffing was proposed. In the proposed puffing model, the rate of mass change of a droplet during puffing was expressed by the evaporation rate of dispersed water and the mass change rate due to fine droplets spouted from the droplet surface. The mass change rate due to fine droplets was related to the evaporation rate of the dispersed water and each liquid content. This model had only one experimental parameter. The essential feature of this model was that it was simple to apply to numerical simulation of spray combustion. First, the validity of the proposed puffing model was investigated with the experimental results for a single droplet. The calculated results for a single droplet with the experimental parameter varying from 5.0 to 10 were in good agreement with the experimental results. Moreover, numerical simulation of spray combustion of emulsified fuel was carried out. The occurrence of puffing and micro-explosion was determined by the inner droplet temperature. When micro-explosion occurred, a droplet changed to vapor rapidly. When the proposed puffing model was used in numerical simulation of spray combustion, the experimental parameter in the puffing model was determined for each droplet by random numbers within the range 5.0-10. The calculated results of spray combustion of emulsified fuel without considering puffing or micro-explosions were different from the experimental results even where combustion reactions were almost terminated. Meanwhile, the calculated results when considering puffing and micro-explosions were in good agreement with experimental results at the same location. (author)

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Modeling soot formation from solid complex fuels

A detailed model is proposed for predicting soot formation from complex solid fuels. The proposed model resolves two particle size distributions, one for soot precursors and another for soot particles. The precursor size distribution is represented with a sectional approach while the soot particle-size distribution is represented with the method of moments and an interpolative closure method is used to resolve fractional methods. Based on established mechanisms, this model includes submodels for precursor coagulation, growth, and consumption, as well as soot nucleation, surface growth, agglomeration, and consumption. The model is validated with comparisons to experimental data for two systems: coal combustion over a laminar flat-flame burner and biomass gasification. Results are presented for soot yield for three coals at three temperatures each, and for soot yield from three types of biomass at two temperatures each. These results represent a wide range of fuels and varying combustion environments, demonstrating the broad applicability of the model.

97 MATHEMATICS AND COMPUTING↗

Graph neural networks for CO 2 solubility predictions in Deep Eutectic Solvents

Deep Eutectic Solvents (DESs) are a promising class of solvents for CO 2 capture. DESs are complex mixtures that can be designed to optimize CO solubility and overall capture process efficiency. However, the vast design landscape of DES mixtures makes experimental investigation prohibitive; as such, there is a need for computational models that can quickly and efficiently navigate the design space and inform data collection efforts. In this work, we propose Graph Neural Network (GNN) models for predicting CO 2 solubility for DESs; the GNN leverages a mixture graph representation that captures the molecular structure of the DES components as well as their intermolecular interactions. Here, we compare the GNN framework against alternative architectures (neural networks, graph convolution networks, and random forests) and data representations (molecular fingerprints, sigma profiles, and graphs). We show that the proposed approach offers superior predictive performance; specifically, we show that solubility can be predicted reliably directly from molecular structure (without the need of using sigma profiles as proposed in previous studies). This result is important, as obtaining sigma profiles requires expensive density functional theory computations. We also explored the ability of GNNs to predict solubility for new DES mixtures and operating conditions. We found that the model extrapolates across temperature reliably. However, we also found deficiencies in the ability of the model to predict solubility for DES mixtures, pressures, and molar ratio not included in the training sets; we show that this is due to an inherent lack of chemical diversity in datasets available in the literature. The proposed computational capabilities can thus help navigate the design space of DES and inform data collection efforts. Our models, data, and benchmarks are shared as Python code implemented in Jupyter notebooks.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Hydro-mechanical coupled modeling of hydraulic fracturing using the mesh fragmentation technique

The inclusion of special interface elements between standard finite elements is becoming a common technique to model discontinuities in porous media using the finite element method (FEM). Recently, it was proved that high aspect ratio (HAR) elements can reproduce very satisfactorily the main features of behavior observed in a single hydraulically-induced fracture. In this paper, the mesh fragmentation technique (MFT) is proposed to extend the application of HAR elements to simulate more general problems involving the formation of hydraulic fractures in rocks by introducing this type of element in-between the standard FE of a typical mesh. The proposed framework makes use of standard FE techniques. Therefore, it is relatively easy to upgrade an existing FE program for continuous porous media to deal with evolving discontinuities. The proposed approach is verified and validated against available analytical and numerical solutions. The crack patterns predicted by the MFT are also compared with the ones observed in the laboratory. The formation of multiple fractures is also analyzed. In all cases, the proposed technique was able to properly capture the main features of rock behavior subjected to hydraulic fracturing.

58 GEOSCIENCES↗

Implementation of extrinsic cohesive zone model (ECZM) in 2D finite-discrete element method (FDEM) using node binding scheme

The combined finite-discrete element method (FDEM) has been widely used for rock fracturing simulations. Conventionally, FDEM is realized using the intrinsic cohesive zone model (ICZM); however, it has the drawback of artificial compliance and high computational expense. As a complement, the extrinsic cohesive zone model (ECZM) is seen to be realized in FDEM recently, whereas the node splitting scheme utilized is cumbersome. Here, within the framework of ICZM-based FDEM, we propose a node binding scheme to efficiently bind the pre-discretized finite elements and thus guarantee the continuum behavior of materials in the elastic stage. The yield surfaces, controlled by ECZM, are dynamically embedded by invoking the pre-inserted cohesive elements. The effectiveness and efficiency of the proposed approach are validated and tested by performing a suite of numerical experiments. Compared with ICZM-based FDEM, the proposed approach can correctly capture material deformation and reduce the computation cost. In contrast to the existing ECZM-based FDEM, the proposed approach can overcome the frequent and complex element topology updating. Finally, this work provides a novel perspective that fully inherits the advantages of both ICZM and ECZM, but circumvents their shortcomings, which guarantees a more efficient and effective simulation of brittle material evolution from continuum to discontinuum.

58 GEOSCIENCES↗

A thermodynamically consistent discontinuous Galerkin formulation for interface separation

Our paper describes the formulation of an interface damage model, based on the discontinuous Galerkin (DG) method, for the simulation of failure and crack propagation in laminated structures. The DG formulation avoids common difficulties associated with cohesive elements. Specifically, it does not introduce any artificial interfacial compliance and, in explicit dynamic analysis, it leads to a stable time increment size which is unaffected by the presence of stiff massless interfaces. This proposed method is implemented in a finite element setting. Convergence and accuracy are demonstrated in Mode I and mixed-mode delamination in both static and dynamic analyses. Significantly, numerical results obtained using the proposed interface model are found to be independent of the value of the penalty factor that characterizes the DG formulation. By contrast, numerical results obtained using a classical cohesive method are found to be dependent on the cohesive penalty stiffnesses. The proposed approach is shown to yield more accurate predictions pertaining to crack propagation under mixed-mode fracture because of the advantage. Furthermore, in explicit dynamic analysis, the stable time increment size calculated with the proposed method is found to be an order of magnitude larger than the maximum allowable value for classical cohesive elements.

42 ENGINEERING↗

An extrapolation method to remove spurious stress concentration in micromechanical analyses of composites using pixel-based meshes

It is easier to generate a finite element mesh of complicated geometries with voxel-based meshes compared to conformal meshes. However, the local stresses in voxel-based meshes are higher owing to artificial stress concentration caused due to jagged lines that approximate a curve. Here, we propose a method that accounts for this anomaly which would lead to accurate prediction of maximum von Mises stress in voxel-based meshes. The method is based on the variation of reciprocal of stresses and associated numerical gradient. The prediction of the proposed method is compared against the results of a conformal mesh by considering four two-dimensional example problems. The accuracy of the proposed method is verified using different mesh densities, material properties, and integration schemes. The proposed method was also implemented on a conformal mesh to verify that it filters out just the spurious stress concentration and not the realistic stress concentration.

36 MATERIALS SCIENCE↗

Prediction of heat transfer coefficients for forced convective boiling of N 2 -hydrocarbon mixtures at cryogenic conditions using artificial neural networks

A key problem faced in the design of heat exchangers, especially for cryogenic applications, is the determination of convective heat transfer coefficients in two-phase flow such as condensation and boiling of non-azeotropic refrigerant mixtures. Here, this paper proposes and evaluates three models for estimating the convective coefficient during boiling. These models are developed using computational intelligence techniques. The performance of the proposed models is evaluated using the mean relative error (mre), and compared to two existing models: the modified Granryd’s correlation and the Silver-Bell-Ghaly method. The three proposed models are distinguished by their architecture. The first is based on directly measured parameters (DMP-ANN), the second is based on equivalent Reynolds and Prandtl numbers (eq-ANN), and the third on effective Reynolds and Prandtl numbers (eff-ANN). In conclusion, the results demonstrate that the proposed artificial neural network (ANN)-based approaches greatly outperform available methodologies. While Granryd's correlation predicts experimental data within a mean relative error mre = 44% and the S-B-G method produces mre = 42%, DMP-ANN has mre = 7.4% and eff-ANN has mre = 3.9%. Considering that eff-ANN has the lowest mean relative error (one tenth of previously available methodologies) and the broadest range of applicability, it is recommended for future calculations. Implementation is straightforward within a variety of platforms and the matrices with the ANN weights are given in the appendix for efficient programming.

42 ENGINEERING↗

Leveraging large language models to address data scarcity in machine learning for graphene synthesis

Machine learning in experimental materials science faces significant challenges due to the scarcity of data, which are costly and time-consuming to generate, particularly when relying on in-house experiments. Literature data mining offers a potential solution but introduces issues like mixed data quality, inconsistent formats, and non-uniform reporting of synthesis parameters, resulting in partially missing and heterogeneous features across the dataset. Here, we propose data imputation and feature engineering methods that employ pre-trained large language models (LLMs) to enhance machine learning performance on scarce, heterogeneous datasets, demonstrated on graphene CVD synthesis data and the ML-HydPARK hydrogen storage dataset. GPT models perform data imputation via tailored prompting and semantic normalization of inconsistently reported features through embeddings, for example, to harmonize the complex nomenclature of CVD substrates. Beyond yielding more diverse and richer feature representations than traditional methods such as K-nearest neighbors (KNN) and Multivariate Imputation by Chained Equations (MICE), LLM-based data imputation is evaluated against dataset characteristics and prompting strategies. We vary the level of autonomy granted to the LLM, from generic prompting that leverages pre-trained knowledge for autonomous data generation to data-informed prompting that constrains outputs using target-specific information, and demonstrate which level of autonomy yields superior imputation performance across datasets and feature types. The proposed data engineering methods markedly improve downstream performance; for example, in graphene layer number classification using a support vector machine (SVM), binary accuracy increases from 39% to 65% and ternary accuracy from 52% to 72%. Fine-tuning experiments on both datasets show that combining our proposed LLM-based data imputation and feature encoding methods with numerical machine learning predictors outperforms standalone fine-tuned LLM predictors in data-scarce settings. The proposed strategies emphasize data enhancement techniques rather than refining learning architectures or regularizing loss functions, offering a broadly applicable framework for improving machine learning performance on scarce, inhomogeneous datasets.

Chemical vapor deposition↗

Social vulnerability indicators as a sustainable planning tool

In the face of global warming and environmental change, the conventional strategy of resource centralization will not be able to cope with a future of increasingly extreme climate events and related disasters. It may even contribute to inter-regional disparities as a result of these events. To promote sustainable development, this study offers a case study of developmental planning in Chiayi, Taiwan and a review of the relevant literature to propose a framework of social vulnerability indicators at the township level. The proposed framework can not only be used to measure the social vulnerability of individual townships in Chiayi, but also be used to capture the spatial developmental of Chiayi. Seventeen social vulnerability indicators provide information in five dimensions. Owing to limited access to relevant data, the values of only 13 indicators were calculated. By simply summarizing indicators without using weightings and by using zero-mean normalization to standardize the indicators, this study calculates social vulnerability scores for each township. To make social vulnerability indicators more useful, this study performs an overlay analysis of social vulnerability and patterns of risk associated with national disasters. The social vulnerability analysis draws on secondary data for 2012 from Taiwan's National Geographic Information System. The second layer of analysis consists of the flood potential ratings of the Taiwan Water Resources Agency as an index of biophysical vulnerability. The third layer consists of township-level administrative boundaries. Analytical results reveal that four out of the 18 townships in Chiayi not only are vulnerable to large-scale flooding during serious flood events, but also have the highest degree of social vulnerability. Administrative boundaries, on which social vulnerability is based, do not correspond precisely to “cross-administrative boundaries,” which are characteristics of the natural environment. This study adopts an exploratory approach that provides Chiayi and other government agencies with a foundation for sustainable strategic planning for environmental change. The final section offers four suggestions concerning the implications of social vulnerability for local development planning. -- Highlights: • This study proposes a framework of social vulnerability indicators at the township level in Chiayi County, Taiwan. • Seventeen social vulnerability indicators are categorized into four dimensions. • This study performs a three-layer overlay analysis of social vulnerability and natural disaster risk patterns. • 4 out of the 18 townships not only have potential for large-scale flooding, but also high degree of social vulnerability. • This study provides a foundation for sustainable strategic planning to deal with environmental change. • Four suggestions are proposed regarding the implications of social vulnerability for local development planning.

54 ENVIRONMENTAL SCIENCES↗

An Analytic Network Process approach for the environmental aspect selection problem — A case study for a hand blender

Life Cycle Assessment is a tool to assess, in a systematic way, the environmental aspects and its potential environmental impacts and resources used throughout a product's life cycle. It is widely accepted and considered as one of the most powerful tools to support decision-making processes used in ecodesign and sustainable production in order to learn about the most problematic parts and life cycle phases of a product and to have a projection for future improvements. However, since Life Cycle Assessment is a cost and time intensive method, companies do not intend to carry out a full version of it, except for large corporate ones. Especially for small and medium sized enterprises, which do not have enough budget for and knowledge on sustainable production and ecodesign approaches, focusing only on the most important possible environmental aspect is unavoidable. In this direction, finding the right environmental aspect to work on is crucial for the companies. In this study, a multi-criteria decision-making methodology, Analytic Network Process is proposed to select the most relevant environmental aspect. The proposed methodology aims at providing a simplified environmental assessment to producers. It is applied for a hand blender, which is a member of the Electrical and Electronic Equipment family. The decision criteria for the environmental aspects and relations of dependence are defined. The evaluation is made by the Analytic Network Process in order to create a realistic approach to inter-dependencies among the criteria. The results are computed via the Super Decisions software. Finally, it is observed that the procedure is completed in less time, with less data, with less cost and in a less subjective way than conventional approaches. - Highlights: • We present a simplified environmental assessment methodology to support LCA. • ANP is proposed to select the most relevant environmental aspect. • ANP deals well with the interdependencies between aspects and impacts. • The methodology is less subjective, less complicated, and less time–money consuming. • The proposed methodology is suitable for use by SMEs.

54 ENVIRONMENTAL SCIENCES↗

The social vulnerability approach for social impact assessment

Highlights: • Social vulnerability is a process in two stages: the Structural and the Acquired.• SVA is apt for the diagnosis and management of social impacts caused by projects. • SVA is based on the Natural disaster and the Assets Based approaches. • Structural Vulnerability influences the very nature of the project. • Understanding the project as a possible hazard improves its analysis. - Abstract: The Social Impact Assessment (SIA) literature has highlighted the need to focus attention on the most vulnerable groups to improve the management of socio-environmental risk. However, methodological proposals to introduce vulnerability in SIAs are still incipient. The aim of this article is to present a proposal to analyze the role of social vulnerability in the production of risks generated by large infrastructure projects. Taking the theoretical-methodological advances made from the social vulnerability and natural disasters approach, and from the social vulnerability, development and poverty approaches, we have designed a Social Vulnerability Approach (SVA) which we aim at incorporating into the SIA of large infrastructure projects. To illustrate our methodological proposal, the case of the socio-environmental conflict generated by the HydroAysen Project (which entailed the construction of a set of hydroelectric dams on the Baker and Pascua rivers in the Chilean Patagonia region) has been selected. Our proposal allows us to identify elements of vulnerability throughout the whole project cycle and, therefore, windows for intervention in order to reduce social vulnerability as well as other kind of hazards that large infrastructure projects may generate on affected populations.

54 ENVIRONMENTAL SCIENCES↗

A modeling framework for optimal energy management of a residential building

Residential buildings are currently equipped with energy production facilities, e.g., solar rooftops and batteries, which in conjunction with smart meters, can function as smart energy hubs coordinating the loads and the resources in an optimal manner. This paper presents a mathematical model for the optimal energy management of a residential building and proposes a centralized energy management system (CEMS) framework for off-grid operation. The model of each component of the hub is integrated within the CEMS. The optimal decisions are determined in real-time by considering these models with realistic parameter settings and customer preferences. Model predictive control (MPC) is used to adapt the optimal decisions on a receding horizon to account for the deviations in the system inputs. Simulation results are presented to demonstrate the feasibility and effectiveness of the proposed CEMS framework. Results show that the proposed CEMS can reduce the energy cost and energy consumption of the customers by approximately 17% and 8%, respectively, over a day. In conclusion, using the proposed CEMS, the total charging cycles of the ESS were reduced by more than 50% in a day.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Generating realistic building electrical load profiles through the Generative Adversarial Network (GAN)

Building electrical load profiles can improve understanding of building energy efficiency, demand flexibility, and building-grid interactions. Current approaches to generating load profiles are time-consuming and not capable of reflecting the dynamic and stochastic behaviors of real buildings; some approaches also trigger data privacy concerns. In this study, we proposed a novel approach for generating realistic electrical load profiles of buildings through the Generative Adversarial Network (GAN), a machine learning technique that is capable of revealing an unknown probability distribution purely from data. The proposed approach has three main steps: (1) normalizing the daily 24-hour load profiles, (2) clustering the daily load profiles with the k-means algorithm, and (3) using GAN to generate daily load profiles for each cluster. The approach was tested with an open-source database – the Building Data Genome Project. We validated the proposed method by comparing the mean, standard deviation, and distribution of key parameters of the generated load profiles with those of the real ones. The KL divergence of the generated and real load profiles are within 0.3 for majority of parameters and clusters. Additionally, results showed the load profiles generated by GAN can capture not only the general trend but also the random variations of the actual electrical loads in buildings. We report the proposed GAN approach can be used to generate building electrical load profiles, verify other load profile generation models, detect changes to load profiles, and more importantly, anonymize smart meter data for sharing, to support research and applications of grid-interactive efficient buildings.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗