Search NASA⌕ Search

SEARCH · Search NASA

Results for “SURROGATE 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 1,243 records · Page 69

Exploring gasoline oxidation chemistry in jet stirred reactors

Recent decades have seen increasingly restrictive regulations applied to gasoline engines. Gasoline combustion chemistry must be investigated to achieve a better understanding and control of internal combustion engine efficiency and emissions. In this work, several gasoline fuels, namely the FACE (Fuel for Advanced Combustion Engines) gasolines, were selected as targets for oxidation study in jet-stirred reactors (JSR). The study is facilitated by formulating various gasoline surrogate mixtures with known hydrocarbon compositions to represent the real gasolines. Surrogates included binary mixtures of n-heptane and iso-octane, as well as more complex multi-component mixtures. Here, the oxidation characteristics of FACE gasolines and their surrogates were experimentally examined in JSR-1 and numerically simulated under the following conditions: pressure 1 bar, temperature 500–1050 K, residence time 1.0 and 2.0 s, and two equivalence ratios (Φ= 0.5 and 1.0). In the high temperature region, all real fuels and surrogates showed similar oxidation behavior, but in the low temperature region, a fuel’s octane number and composition had a significant effect on its JSR oxidation characteristics. Low octane number fuels displayed more low temperature reactivity, while fuels with similar octane number but a larger number of n-alkane components were more reactive. A gasoline surrogate kinetic model was examined with FACE gasoline experiments either measured in JSR-2, or taken from previous work under the following conditions: pressure 10 bar, temperature 530–1200 K, residence time 0.7 s, and three equivalence ratios (Φ = 0.5, 1.0 and 2.0). Comparison between FACE gasoline experimental results with surrogate model predictions showed good agreement, demonstrating considerable potential for surrogate fuel kinetic modeling in engine simulations.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

WE-AB-303-11: Verification of a Deformable 4DCT Motion Model for Lung Tumor Tracking Using Different Driving Surrogates

Purpose: Tumor tracking is an advanced technique to treat intra-fractionally moving tumors. The aim of this study is to validate a surrogate-driven model based on four-dimensional computed tomography (4DCT) that is able to predict CT volumes corresponding to arbitrary respiratory states. Further, the comparison of three different driving surrogates is evaluated. Methods: This study is based on multiple 4DCTs of two patients treated for bronchial carcinoma and metastasis. Analyses for 18 additional patients are currently ongoing. The motion model was estimated from the planning 4DCT through deformable image registration. To predict a certain phase of a follow-up 4DCT, the model considers for inter-fractional variations (baseline correction) and intra-fractional respiratory parameters (amplitude and phase) derived from surrogates. In this evaluation, three different approaches were used to extract the motion surrogate: for each 4DCT phase, the 3D thoraco-abdominal surface motion, the body volume and the anterior-posterior motion of a virtual single external marker defined on the sternum were investigated. The estimated volumes resulting from the model were compared to the ground-truth clinical 4DCTs using absolute HU differences in the lung volume and landmarks localized using the Scale Invariant Feature Transform (SIFT). Results: The results show absolute HU differences between estimated and ground-truth images with median values limited to 55 HU and inter-quartile ranges (IQR) lower than 100 HU. Median 3D distances between about 1500 matching landmarks are below 2 mm for 3D surface motion and body volume methods. The single marker surrogates Result in increased median distances up to 0.6 mm. Analyses for the extended database incl. 20 patients are currently in progress. Conclusion: The results depend mainly on the image quality of the initial 4DCTs and the deformable image registration. All investigated surrogates can be used to estimate follow-up 4DCT phases, however uncertainties decrease for three-dimensional approaches. This work was funded in parts by the German Research Council (DFG) - KFO 214/2.

60 APPLIED LIFE SCIENCES↗

Further examples of the failure of surrogates to properly model the structural and hydrothermal chemistry of transuranium elements: Insights provided by uranium and neptunium diphosphonates

In situ hydrothermal reduction of Np(VI) to Np(IV) in the presence of methylenediphosphonic acid (C1P2) results in the crystallization of Np[CH2(PO3)2](H2O)2 (NpC1P2-1). Similar reactions have been explored with U(VI) resulting in the isolation of the U(IV) diphosphonate U[CH2(PO3)2](H2O) (UC1P2-1), and the two U(VI) diphosphonates (UO2)2[CH2(PO3)2](H2O)3 H2O (UC1P2-2) and UO2[CH2(PO3H)2](H2O) (UC1P2-3). Single crystal diffraction studies of NpC1P2-1 reveal that it consists of eight-coordinate Np(IV) bound by diphosphonate anions and two coordinating water molecules to create a polar three-dimensional framework structure wherein the water molecules reside in channels. The structure of UC1P2-1 is similar to that of NpC1P2-1 in that it also adopts a three-dimensional structure. However, the U(IV) centers are seven-coordinate with only a single bound water molecule. UC1P2-2 and UC1P2-3 both contain U(VI). Nevertheless, their structures are quite distinct with UC1P2-2 being composed of corrugated layers containing UO6 and UO7 units bridged by C1P2; whereas, UC1P2-3 is found as a polar three-dimensional network structure containing only pentagonal bipyramidal U(VI). Fluorescence measurements on UC1P2-2 and UC1P2-3 exhibit emission from the uranyl moieties with classical vibronic fine-structure.

Nelson, Anna-Gay D↗

Land-use change and greenhouse gas emissions from corn and cellulosic ethanol

Background: The greenhouse gas (GHG) emissions that may accompany land-use change (LUC) from increased biofuel feedstock production are a source of debate in the discussion of drawbacks and advantages of biofuels. Estimates of LUC GHG emissions focus mainly on corn ethanol and vary widely. Increasing the understanding of LUC GHG impacts associated with both corn and cellulosic ethanol will inform the on-going debate concerning their magnitudes and sources of variability. Results: In our study, we estimate LUC GHG emissions for ethanol from four feedstocks: corn, corn stover, switchgrass, and miscanthus. We use new computable general equilibrium (CGE) results for worldwide LUC. U.S. domestic carbon emission factors are from state-level modelling with a surrogate CENTURY model and U.S. Forest Service data. This paper investigates the effect of several key domestic lands carbon content modelling parameters on LUC GHG emissions. International carbon emission factors are from the Woods Hole Research Center. LUC GHG emissions are calculated from these LUCs and carbon content data with Argonne National Laboratory’s Carbon Calculator for Land Use Change from Biofuels Production (CCLUB) model. Our results indicate that miscanthus and corn ethanol have the lowest (-10 g CO2e/MJ) and highest (7.6 g CO2e/MJ) LUC GHG emissions under base case modelling assumptions. The results for corn ethanol are lower than corresponding results from previous studies. Switchgrass ethanol base case results (2.8 g CO2e/MJ) were the most influenced by assumptions regarding converted forestlands and the fate of carbon in harvested wood products. They are greater than miscanthus LUC GHG emissions because switchgrass is a lower-yielding crop. Finally, LUC GHG emissions for corn stover are essentially negligible and insensitive to changes in model assumptions. Conclusions: This research provides new insight into the influence of key carbon content modelling variables on LUC GHG emissions associated with the four bioethanol pathways we examined. Our results indicate that LUC GHG emissions may have a smaller contribution to the overall biofuel life cycle than previously thought. Additionally, they highlight the need for future advances in LUC GHG emissions estimation including improvements to CGE models and aboveground and belowground carbon content data.

09 BIOMASS FUELS↗

Incorporating Agricultural Management Practices into the Assessment of Soil Carbon Change and Life-Cycle Greenhouse Gas Emissions of Corn Stover Ethanol Production

Land management practices such as cover crop adoption or manure application that can increase soil organic carbon (SOC) may provide a way to counter SOC loss upon removal of stover from corn fields for use as a biofuel feedstock. This report documents the data, methodology, and assumptions behind the incorporation of land management practices into corn-soybean systems that dominate U.S. grain production using varying levels of stover removal in the GREET TM (Greenhouse gases, Regulated Emissions, and Energy use in Transportation) model and its CCLUB (Carbon Calculator for Land Use change from Biofuels production) module. Tillage (i.e., conventional, reduced and no tillage), corn stover removal (i.e., at 0, 30% and 60% removal rate), and organic matter input techniques (i.e., cover crop and manure application) are included in the analysis as major land management practices. Soil carbon changes associated with land management changes were modeled with a surrogate CENTURY model. The resulting SOC changes were incorporated into CCLUB while GREET was expanded to include energy and material consumption associated with cover crop adoption and manure application. Life-cycle greenhouse gas (GHG) emissions of stover ethanol were estimated using a marginal approach (all burdens and benefits assigned to corn stover ethanol) and an energy allocation approach (burdens and benefits divided between grain and stover ethanol). In the latter case, we considered corn grain and corn stover ethanol to be produced at an integrated facility. Life-cycle GHG emissions of corn stover ethanol are dependent upon the analysis approach selected (marginal versus allocation) and the land management techniques applied. The expansion of CCLUB and GREET to accommodate land management techniques can produce a wide range of results because users can select from multiple scenario options such as choosing tillage levels, stover removal rates, and whether crop yields increase annually or remain constant. In a scenario with conventional tillage and a 30% stover removal rate, life-cycle GHG emissions for a combined gallon of corn grain and stover ethanol without cover crop adoption or manure application are 49 g CO 2 eq MJ -1 , in comparison with 91 g CO 2 eq MJ -1 for petroleum gasoline. Adopting a cover crop or applying manure reduces the former ethanol life-cycle GHG emissions by 8% and 10%, respectively. We considered two different life cycle analysis approaches to develop estimates of life-cycle GHG emissions for corn stover ethanol, marginal analysis and energy allocation. In the same scenario, this fuel has GHG emissions of 12 – 20 g CO 2 eq MJ -1 (for manure and cover crop application, respectively) and 45 – 48 g CO 2 eq MJ -1 with the marginal approach and the energy allocation approach, respectively.

09 BIOMASS FUELS↗

HPC-Driven Modeling with ML-Based Surrogates for Magnon-Photon Dynamics in Hybrid Quantum System

Here, we introduce a hybrid computational framework that merges HPC-based numerical solvers with physics-informed ML surrogates for efficient modeling of magnon-photon interactions. By running short-duration, high-fidelity Maxwell-LLG simulations and feeding their results into an ML model, we substantially cut simulation time while achieving accurate predictions across larger spatiotemporal domains.

Accuracy↗

Temperature, humidity and air flow in the emplacement drifts using convection and dispersion transport models

A coupled thermal-hydrologic-airflow model is developed, solving for the transport processes within a waste emplacement drift and the surrounding rockmass together at the proposed nuclear waste repository at Yucca Mountain. Natural, convective air flow as well as heat and mass transport in a representative emplacement drift during post-closure are explicitly simulated, using the MULTIFLUX model. The conjugate, thermal-hydrologic transport processes in the rockmass are solved with the TOUGH2 porous-media simulator in a coupled way to the in-drift processes. The new simulation results show that large-eddy turbulent flow, as opposed to small-eddy flow, dominate the drift air space for at least 5000 years following waste emplacement. The size of the largest, longitudinal eddy is equal to half of the drift length, providing a strong axial heat and moisture transport mechanism from the hot to the cold drift sections. The in-drift results are compared to those from simplified models using a surrogate, dispersive model with an equivalent dispersion coefficient for heat and moisture transport. Results from the explicit, convective velocity simulation model provide higher axial heat and moisture fluxes than those estimated from the previously published, simpler, equivalent-dispersion models, in addition to showing differences in temperature, humidity and condensation rate distributions along the drift length. A new dispersive model is also formulated, giving a time- and location-variable function that runs generally about ten times higher in value than the highest dispersion coefficient currently used in the Yucca Mountain Project as an estimate for the equivalent dispersion coefficient in the emplacement drift. The new dispersion coefficient variation, back-calculated from the convective model, can adequately describe the heat and mass transport processes in the emplacement drift example.

54 ENVIRONMENTAL SCIENCES↗

Multi-objective surrogate-assisted calibration of CPFEM models using macroscopic response and in situ EBSD measurements of grain reorientation trajectories

Crystal plasticity finite element method (CPFEM) models are widely used to simulate the deformation behaviour of polycrystalline materials, but their calibration is often limited by their high computational cost and the non-convexity of the optimisation landscape. Here, this study develops a multi-objective surrogate-assisted calibration workflow that couples a multi-objective genetic algorithm (MOGA) with an adaptively trained deep neural network (DNN) surrogate model to efficiently identify CPFEM parameters from experimental data. The workflow is demonstrated on three crystal plasticity (CP) formulations of increasing complexity — Voce hardening (VH), two-coefficient latent hardening (LH2), and six-coefficient latent hardening (LH6) — using in situ electron backscatter diffraction (EBSD) measurements of Alloy 617 under uniaxial tensile loading. The CPFEM models are calibrated against the experimentally observed stress–strain response and reorientation trajectories of eight grains, then validated against eight additional trajectories and overall texture evolution. Across the CP formulations, the macroscopic response was reproduced reliably, while differences emerged in the robustness and accuracy of the grain-scale predictions. Including grain reorientation trajectories in the multi-objective calibration improved texture evolution predictions and filtered out physically inconsistent parameter sets that can arise from calibrating against only the stress–strain data. The workflow also demonstrates good transferability of calibrated parameters from a low- to a high-fidelity microstructural model. These results provide practical guidance for integrating in situ microstructural data into CPFEM through efficient, repeatable, and physically meaningful multi-objective calibration.

Crystal plasticity finite element method↗

PINN surrogate of Li-ion battery models for parameter inference, Part I: Implementation and multi-fidelity hierarchies for the single-particle model

To plan and optimize energy storage demands that account for Li-ion battery aging dynamics, techniques need to be developed to diagnose battery internal states accurately and rapidly. Here, this study seeks to reduce the computational resources needed to determine a battery's internal states by replacing physics-based Li-ion battery models - such as the single-particle model (SPM) and the pseudo-2D (P2D) model - with a physics-informed neural network (PINN) surrogate. The surrogate model makes high-throughput techniques, such as Bayesian calibration, tractable to determine battery internal parameters from voltage responses. This manuscript is the first of a two-part series that introduces PINN surrogates of Li-ion battery models for parameter inference (i.e., state-of-health diagnostics). In this first part, a method is presented for constructing a PINN surrogate of the SPM. A multi-fidelity hierarchical training, where several neural nets are trained with multiple physics-loss fidelities is shown to significantly improve the surrogate accuracy when only training on the governing equation residuals. The implementation is made available in a companion repository (https://github.com/NREL/PINNSTRIPES). The techniques used to develop a PINN surrogate of the SPM are extended in Part II for the PINN surrogate for the P2D battery model, and explore the Bayesian calibration capabilities of both surrogates.

25 ENERGY STORAGE↗

Persistence and decontamination of surrogate radioisotopes in a model drinking water distribution system

Contamination of a model drinking water system with surrogate radioisotopes was examined with respect to persistence on and decontamination of infrastructure surfaces. Cesium and cobalt chloride salts were used as surrogates for cesium-137 and cobalt-60. Studies were conducted in biofilm annular reactors containing heavily corroded iron surfaces formed under shear and constantly submerged in drinking water. Cesium was not detected on the corroded iron surface after equilibration with 10 and 100 mg L{sup -1} solutions of cesium chloride, but cobalt was detected on corroded iron coupons at both initial concentrations. The amount of adhered cobalt decreased over the next six weeks, but was still present when monitoring stopped. X-ray absorption near-edge spectroscopy (XANES) showed that adhered cobalt was in the III oxidation state. The adsorbed cobalt was strongly resistant to decontamination by various physicochemical methods. Simulated flushing, use of free chlorine and dilute ammonia were found to be ineffective whereas use of aggressive methods like 14.5 M ammonia and 0.36 M sulfuric acid removed 37 and 92% of the sorbed cobalt, respectively.

54 ENVIRONMENTAL SCIENCES↗

Sensitivity Study of Multiscale and Phenomenological Elasto-Viscoplastic Grade 91 Material Models for Component-Scale Response

Many advanced nuclear reactor concepts currently being developed are targeting higher operating temperatures relative to the current fleet of light water nuclear reactors, for efficiency gains and other operational considerations. The design of high temperature structural components with reliable long-term operational performance will depend on material models that accurately capture the inelastic deformation mechanisms active in these environments. In this work, we perform a detailed parameter sensitivity analysis of two unified elasto-viscoplastic Grade 91 material models capable of capturing long term high temperature creep deformation. The first model is a phenomelogical material model from the Nuclear Engineering Material Library (NEML) developed at Argonne National Lab. The NEML model parameters and their uncertainty were fit to a range of Grade 91 experimental data using Bayesian Markov Chain Monte Carlo analysis. The second model is a LAROMance data-driven surrogate material model developed at Los Alamos National Lab. The LAROMance model is fit to a large database of responses produced by a mechanistic crystal plasticity based polycrystal model. Parameters for the LAROMance surrogate material model reflect the pedigree of the Grade 91 microstructure. Both material models have been integrated into the Grizzly code, based on the open-source MOOSE multiphysics simulation framework, to simulate both the progression of aging mechanisms and the effects of that aging on nuclear power plant structures. Grizzly is used analyze a three-dimensional Grade 91 piping system to compare the long-term inelastic response predicted by these two fundamentally different models and assess the sensitivity of the material model input parameters on this quantity of interest.

42 ENGINEERING↗

Uncertainty Quantification using Deep Ensembles for Decision Making in Cyber-Physical-Human Systems

In this paper and its companion, Differential Equation Approximation Using Gradient-Boosted Quantile Regression, Robison et al., we examine an approach to quantifying model uncertainty with the aim of increasing the trustworthiness of computational models in human-machine interactions. In Differential Equation Approximation Using Gradient-Boosted Quantile Regression, we focus on gradient-boosted decision trees, while in this one, we give more details about deep ensembles. Uncertainty quantification is crucial for building trustworthy autonomous decision-making agents in human-machine teams. There are two types of uncertainties: aleatoric and epistemic. The former is related to the inherent stochasticity (noise) of the process, whereas the latter is associated with the lack of knowledge or representation capability of models, such as neural networks. By lack of knowledge, we mean the model’s inability to accurately predict outputs for all possible inputs. The aleatory uncertainty can be estimated fairly easily with, for example, filters, whereas epistemic uncertainty is challenging to compute. This paper uses deep ensembles to quantify both aleatory and epistemic uncertainty. It can act as an uncertainty-aware surrogate transition model for decision-making frameworks. "Uncertainty-aware" means that the surrogate transition model should make predictions along with confidence in those predictions. In the context of decision-making, the transition models are ordinary differential equations (ODEs). Since ODEs can be simulated to make one-step or multi-step predictions, a good surrogate model for them should perform reasonably well in both modes. In a multi-step approach, the trajectory sampling method TS∞ was used to propagate uncertainty over multiple steps. The cartpole dynamical system was selected to demonstrate the ability of deep ensembles as good surrogate transition models for decision-making frameworks. The deep ensembles modeled the dynamics of cartpole ODEs and made uncertainty-aware predictions in single-step and multi-step transition modes.

CPH systems↗

Stress intensity factor models using mechanics-guided decomposition and symbolic regression

The finite element method can be used to compute accurate stress intensity factors (SIFs) for cracks with complex geometries and boundary conditions. In contrast, handbook solutions act as surrogate SIF models that provide significantly faster evaluation times. However, the development of conventional surrogate SIF models relies on manual development based on low-order parameterizations. This limits surrogate model accuracy and generalizability. Here, in this paper, we develop a framework for the automated development of mechanics-guided handbook SIF solutions by using interpretable machine learning via genetic programming for symbolic regression (GPSR). Formalizing the mechanics-based approach of Raju and Newman, SIF training data is decomposed into multiple subsets. This decomposition enables parallel GPSR model development of subfunctions, each of which accounts for specific geometrical corrections with respect to a known analytical model. Using this mechanics-based approach with GPSR allows for equations to be learned with improved accuracy and reduced complexity relative to the Raju Newman equations while maintaining the inherent interpretability of mathematical expressions. In this paper, we present equations that match the complexity of the Raju Newman equations while having reduced error, as well as equations with similar errors and reduced complexity.

42 ENGINEERING↗