Search NASA⌕ Search

SEARCH · Search NASA

Results for “processing parameters”

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 901 records · Page 50
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Five percent measurement of the gravitational constant in the Large Magellanic Cloud

We perform a novel test of General Relativity by measuring the gravitational constant in the Large Magellanic Cloud (LMC). The LMC contains six well-studied Cepheid variable stars in detached eclipsing binaries. Radial velocity and photometric observations enable a complete orbital solution, and precise measurements of the Cepheids’ periods permit detailed stellar modelling. Both are sensitive to the strength of gravity, the former via Kepler’s third law and the latter through the gravitational free-fall time. We jointly fit the observables for stellar parameters and the gravitational constant. Performing a full Markov Chain Monte Carlo analysis of the parameter space including all relevant nuisance parameters, we constrain the gravitational constant in the Large Magellanic Cloud relative to the Solar System to be G LMC /G SS = $0.93{^{+0.05}_{-0.04}}$. We discuss the implications of this 5% measurement of Newton’s constant in another galaxy for dark energy and modified gravity theories. This result excludes one Cepheid, CEP-1812, which is an outlier and needs further study: it is either a highly unusual system to which our model does not apply, or it prefers G LMC < G SS at 2.6σ. We also obtain new bounds on critical parameters that appear in semi-analytic descriptions of stellar processes. In particular, we measure the mixing length parameter to be $α = 0.90{^{+0.36}_{-0.26}}$ (when assumed to be constant across our sample), and obtain constraints on the parameters describing turbulent dissipation and convective flux.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

The VMM3a ASIC

The VMM3a is a custom low-noise Application Specific Integrated Circuit. It is intended to be used in the front end readout electronics of both the Micromegas and sTGC detectors of the ATLAS New Small Wheels upgrade project at CERN. It is fabricated in the 130 nm GlobalFoundries 8RF-DM process. Here, the 64 channels with highly configurable parameters, although designed for a specific project, can meet the processing needs of signals from various detector types in several other applications. In this note, the VMM3a, which is the production version of the VMM family, will be presented along with the features incorporated and its performance.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2020

Optimization of a Cyclone Using Multiphase Flow Computational Fluid Dynamics

The U.S. Department of Energy National Energy Technology Laboratory's (NETL) 50 kW th chemical looping reactor (CLR) has an underperforming cyclone, which was designed using empirical correlations. To improve the performance of this cyclone using computational fluid dynamics (CFD)-based modeling simulations, four critical design parameters including the vortex tube radius and length, barrel radius, and the inlet width and height were optimized. Here, NETL's open source multiphase flow with interphase exchange (MFiX) CFD code has been used to model a series of cyclones by systematically varying the geometric design parameters. To perform the optimization process, the surrogate modeling and sensitivity analysis followed by the optimization capability in nodeworks was used. The basic methodology for the process is to employ a statistical design of experiments (DOE) method to generate sampling simulations that fill the design space. Corresponding CFD models are then created, executed, and postprocessed. A response surface is created to characterize the relationship between input parameters and the quantities of interest (QoI). Finally, the CFD-surrogate is used by an optimization method to find the optimal design condition based on the objective and constraints prescribed. The resulting optimal cyclone has a larger diameter and longer vortex tube, a larger diameter barrel, and a taller and narrower solids inlet. The improved design has a predicted pressure drop 11 times lower than the original design while reducing the mass loss by a factor of 2.3.

42 ENGINEERING↗
DOE OSTI2021

Assessing the representation of the Australian carbon cycle in global vegetation models

Australia plays an important role in the global terrestrial carbon cycle on inter-annual timescales. While the Australian continent is included in global assessments of the carbon cycle such as the global carbon budget, the performance of dynamic global vegetation models (DGVMs) over Australia has rarely been evaluated. We assessed simulations of net biome production (NBP) and the carbon stored in vegetation between 1901 to 2018 from 13 DGVMs (TRENDY v8 ensemble). We focused our analysis on Australia's short-term (inter-annual) and long-term (decadal to centennial) terrestrial carbon dynamics. The TRENDY models simulated differing magnitudes of NBP on inter-annual timescales, and these differences resulted in significant differences in long-term vegetation carbon accumulation (–4.7 to 9.5 Pg C). We compared the TRENDY ensemble to several satellite-derived datasets and showed that the spread in the models' simulated carbon storage resulted from varying changes in carbon residence time rather than differences in net carbon uptake. Differences in simulated long-term accumulated NBP between models were mostly due to model responses to land-use change. The DGVMs also simulated different sensitivities to atmospheric carbon dioxide (CO 2 ) concentration, although notably, the models with nutrient cycles did not simulate the smallest NBP response to CO 2 . Our results suggest that a change in the climate forcing did not have a large impact on the carbon cycle on long timescales. However, the inter-annual variability in precipitation drives the year-to-year variability in NBP. We analysed the impact of key modes of climate variability, including the El Niño–Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD), on NBP. While the DGVMs agreed on sign of the response of NBP to El Niño and La Niña and to positive and negative IOD events, the magnitude of inter-annual variability in NBP differed strongly between models. In addition, we find that differences in the timing of simulated phenology and fire dynamics are associated with differences in simulated or prescribed vegetation cover and process representation. We further find model disagreement in simulated vegetation carbon, phenology, and apparent carbon residence time, indicating that the models have different types and coverage of vegetation across Australia (whether prescribed or emergent). Our study highlights the need to evaluate parameter assumptions and the key processes that drive vegetation dynamics, such as phenology, mortality, and fire, in an Australian context to reduce uncertainty across models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Advances in imaging of chemically reacting flows

Many important chemically reacting systems are inherently multi-dimensional with spatial and temporal variations in the thermochemical state, which can be strongly coupled to interactions with transport processes. Fundamental insights into these systems require multi-dimensional measurements of the thermochemical state as well as fluid dynamics quantities. Laser-based imaging diagnostics provide spatially and temporally resolved measurements that help address this need. The state of the art in imaging diagnostics is continually progressing with the goal of attaining simultaneous multi-parameter measurements that capture transient processes, particularly those that lead to stochastic events, such as localized extinction in turbulent combustion. Development efforts in imaging diagnostics benefit from advances in laser and detector technology. This article provides a perspective on the progression of increasing dimensionality of laser-based imaging diagnostics and highlights the evolution from single-point measurements to 1D and 2D multi-parameter imaging and 3D high-speed imaging. This evolution is demonstrated using highlights of laser-based imaging techniques in combustion science research as an exemplar of a complex multi-dimensional chemically reacting system with chemistry–transport coupling. Imaging diagnostics impact basic research in other chemically reacting systems as well, such as measurements of near-surface gases in heterogeneous catalysis. The expanding dimensionality of imaging diagnostics leads to larger and more complex datasets that require increasingly demanding approaches to data analysis and provide opportunities for increased collaboration between experimental and computational researchers in tackling these challenges.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Development of a high fidelity CFD model for solvent evaporation and transport in porous structure during battery electrode drying

An efficient battery manufacturing process is the key to the mass production of Electric Vehicles (EV), in which drying is one of the most energy-intensive steps significantly influencing the battery cell performance. An accurate 3D CFD model for drying is essential for predicting the drying mechanism and optimizing its parameters. By optimizing the drying process, it is possible to reduce energy consumption and cost during battery manufacturing, minimize binder loading and maximize active material loading to achieve superior electrochemical performances and facilitate wider and faster public adoption of EV. This project aims to optimize the drying process during electrode manufacturing by leveraging high-fidelity, porous electrode simulations for solvent evaporation. By optimizing this process, we seek to reduce energy consumption during battery manufacturing, while minimizing binder loading and maximizing active material loading, with the overall goal of enhancing electrical vehicle performance.

Horner, Jeffrey Scott [Sandia National Laboratorie↗
DOE OSTI2026

Active learning of a crystal plasticity flow rule from discrete dislocation dynamics simulations

Continuum-scale material deformation models, such as crystal plasticity (CP), can significantly enhance their predictive accuracy by incorporating input from lower-scale (i.e. mesoscale) models. The procedure to generate and extract the relevant information is however typically complex and ad hoc, involving decision and intervention by domain experts, leading to long development times. In this study, we develop a principled approach for calibration of continuum-scale models using lower scale information by representing a CP flow rule as a Gaussian process model. This representation allows for efficient parameter space exploration, guided by the uncertainty embedded in the model through a process known as Bayesian optimization (BO). We demonstrate a semi-autonomous BO loop which instantiates discrete dislocation dynamics simulations whose initial conditions are automatically chosen to optimize the uncertainty of a model CP flow rule. Our self-guided computational pipeline efficiently generated a dataset and corresponding model whose error, uncertainty, and physical feature sensitivities were validated with comparison to an independent dataset four times larger, demonstrating a valuable and efficient active learning implementation readily transferable to similar material systems.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Effect of heating/cooling dynamics in the hysteresis loop and tunable IR emissivity of VO 2 thin films

We experimentally investigate the semiconductor-to-metal transition (SMT) in vanadium dioxide thin films using an infrared thermographic technique. During the semiconductor to metal phase change process, VO 2 optical properties dynamically change and infrared emission undergoes a hysteresis loop due to differences between heating and cooling stages. The shape of the hysteresis loop was accurately monitored under different dynamic heating/cooling rates. In order to quantify and understand the effects of different rates, we used a numerical modelling approach in which a VO 2 thin layer was modeled as metamaterial. The main experimental findings are interpreted assuming that both the rate of formation and shape of metallic inclusions are tuned with the heating/cooling rate. The structural transition from monoclinic to tetragonal phases is the main mechanism for controlling the global properties of the phase transition. However, our experimental results reveal that the dynamics of the heating/cooling process can become a useful parameter for further tuning options and lays out a macroscopic optical sensing scheme for the microscopic phase change dynamics of VO 2 . Our study sheds light on phase-transition dynamics and their effect on the infrared emission spectra of VO 2 thin films, therefore enabling the heating/cooling rate to be an additional parameter to control infrared emission characteristics of thermal emitters. The hysteresis loop represents the phase coexistence region, thus being of fundamental importance for several applications, such as the operation of radiative thermal logic elements based on phase transition materials. For such applications, the phase transition region is shifted for heating and cooling processes. We also show that, depending on the way the phase change elements are heated, the temperature operation range will be slightly modified.

42 ENGINEERING↗
DOE OSTI2024

Sensitivity of modelled passive margin stratigraphy to variations in sea level, sediment supply and subsidence

Abstract We produced a 10 Myr synthetic stratigraphic section using a forward stratigraphic model that generates marine deltaic stratigraphy over geological timescales. We recursively fit the model using a Bayesian inversion algorithm to test: (1) if it could be accurately reconstructed; (2) if the parameters used to create it could be recovered; and (3) the sensitivity of the model output to given model parameters and the attendant physical processes. The original synthetic stratigraphic section was produced with cyclical sea‐level variations of 40 and 30 m with 2.4 and 10 Myr periods respectively. Sediment was also supplied cyclically, in 2.4 and 10 Myr cycles with amplitudes of 30 and 80 tons/100 kyr, respectively, varying from a mean of 232 tons/100 kyr. Parameter values were sampled to fit the model using a Markov chain Monte Carlo algorithm, resulting in a ±5 m (1σ) variation between the experimental output and the original. Sea level varied by ±7 m (1σ) within the posterior distribution of parameters. As a result, both the 10 Myr and 2.4 Myr sea‐level cycles could be extracted from the original output. The variation in sediment supply was approximately ±38 tons/100 kyr (1σ) and, as a result, only the larger long‐term supply variations could be accurately recovered in refitting the model. The variation in thermal, flexural and total subsidence across those parameter sets is less than ±10 m (1σ). The original section experienced 150 m of total subsidence at the depocentre. Our results demonstrate the distinct and interpretable imprint of sea level and subsidence on continental margin stratigraphy can be quantified. Moreover, we conclude that sea‐level change produces a defined effect on the geometries of stratigraphic architecture, and that techniques applied for the purpose of delineating sea‐level variation from continental margin strata have a well‐founded conceptual basis.

Schmelz, William J.↗
DOE OSTI2021

Analysis of energetics and economics of sub-ambient hybrid post-combustion carbon dioxide capture

Adsorption of CO 2 from post-combustion flue gas is one of the leading candidates for globally impactful carbon capture systems. In this report, work focused on understanding the opportunities and limitations of sub-ambient CO 2 capture processes utilizing a multistage separation process. A hybrid process design using a combination of pressure-driven separation of CO 2 from flue gas (e.g., adsorption- or membrane-based separation) followed by CO 2 -rich product liquefaction to produce high-purity (>99%) CO 2 at pipeline conditions is considered. The operating pressure of the separation unit is a key cost parameter and also an important process variable that regulates the available heat removal necessary to reach the sub-ambient operating conditions. The economic viability of applying pressure swing adsorption (PSA) processes using fiber sorbent contactors with internal heat management was found to be most influenced by the productivity of the adsorption system, with productivities as high as 0.015 mol CO2 /kg sorb -1 sec -1 being required to reduce costs of capture below $60/ton CO 2 captured. This analysis was carried out using a simplified two-bed process, and thus there is opportunity for further cost reduction with exploration of more complex cycle designs. Three exemplar fiber sorbents (MIL-101(Cr), UiO-66, and zeolite 13X) were considered for application in the sub-ambient process of PSA unit. Among the considered sorbents, zeolite 13X fiber composites were found to perform better at ambient temperatures as compared to sub-ambient. MIL-101(Cr) and UiO-66 fiber composites had improved purity, recovery, and productivity at colder temperatures reducing costs of capture as low as $61/ton CO 2 . Future economic improvement could be achieved by reducing the required operating pressure of the PSA unit and pushing the Pareto frontier closer to the final pipeline requirement via a combination of PSA cycle design and material selection.

42 ENGINEERING↗
DOE OSTI2025

Impact of U-10Mo HALEU fuel element tolerances on the Massachusetts Institute of Technology reactor safety and operational performance – Thermal hydraulics

The U.S. is coordinating efforts for the conversion of six U.S. High Performance Research Reactors (USHPRR) including one critical facility from highly enriched uranium (HEU) to low-enriched uranium (LEU). In order to continue the mission of these reactors, including the Massachusetts of Institute of Technology Reactor (MITR), and achieve similar performance, high-assay low-enriched uranium (HALEU) with a high-density metallic alloy of uranium with 10 wt% molybdenum (U-10Mo) is being evaluated. The impact of the fabrication specification and tolerances was assessed following the preliminary design of the MITR LEU fuel elements using the U-10Mo monolithic alloy. This research focuses on the analysis of fabrication specification impact on thermal hydraulics (TH) characteristic of the MITR LEU core as a function of the variation of the relevant fuel specification parameters (e.g., coolant channel gap thickness, fuel plate thickness, etc.). The analyses are performed based on an all-fresh LEU fuel conversion plan identified in a preliminary safety analysis report submitted to the Nuclear Regulatory Commission. The reactor power margin to the onset of nucleate boiling (ONB) is assessed under the limiting safety system settings (LSSS), where a scram occurs, to ensure there is sufficient margin to the reactor safety limit, which is defined by the onset of flow instability that occurs after the ONB. The best estimate plus uncertainty approach is employed to analyze this TH characteristic, which yields realistic results while maintaining adequate conservatism, utilizing a statistical uncertainty propagation method with the STAT7 code. The TH characteristic is analyzed as a function of the variability of the specification parameters resulting from the fabrication process. The main findings of this study show that the MITR core can meet the TH safety and operational requirements at the all-LEU initial core startup (cycle 1), selected transition cycles (most reactive cycle and most limiting cycle: cycle 3 and 5, respectively) and equilibrium (cycle 14) cores under all limiting fabrication parameter combinations considered. In addition, the analyses show that the dependency of the core power margin to ONB on those specification parameters that have the most direct impact on TH performance is non-linear but monotonically decreasing within the specification tolerances. The third order polynomial fit curves are reported in detail for selected limiting cases and can serve as a powerful tool for future MITR fuel management in cases such as when HALEU supply is established that may allow additional cycle length or other operational benefits.

Conversion↗
DOE OSTI2023

Physics-based Models, Machine Learning, and Experiment: Towards Understanding Complex Electrode Degradation

Degradation phenomena in Li-ion batteries are highly complex, coupled, and sensitive to use history and operating conditions. In this study, we show how tracking model parameters in continuum-level physics-based models, expedited by machine learning, can be useful in testing hypotheses for degradation mechanisms. An exemplary analysis using this approach is presented for a set of lithium trivanadate ( L i x V 3 O 8 ) cathodes cycled over a range of current rates. A simple cell revival process is combined with the parameter estimates over the course of cycling to extract valuable insights into cathode evolution and eliminate hypothesized degradation mechanisms for these cathodes. The presented approach is expected to be broadly applicable for degradation analysis of other electrodes.

25 ENERGY STORAGE↗
DOE OSTI2024

False vacuum decay and nucleation dynamics in neutral atom systems

Metastable states of quantum many-body systems with confinement offer a means to simulate false vacuum phenomenology, including nonequilibrium dynamical processes like decay by nucleation, in truncated limits. Recent work has examined the decay process in one-dimensional (1D) ferromagnetic Ising spins and superfluids. Here, in this paper, we study nucleation dynamics in 1D antiferromagnetic neutral atom chains with Rydberg interactions, using both numerical simulations and analytic modeling. We apply a staggered local detuning field to generate the metastable and ground states. Our efforts focus on two dynamical regimes: decay and annealing. In the first, we corroborate the phenomenological decay rate scaling and determine the associated parameter range for the decay process; in the second, we uncover and elucidate a procedure to anneal the metastable state from the initial to the final system, with intermediate nucleation events. We further propose experimental protocols to prepare the required states and perform quenches on near-term neutral atom quantum simulators, examining the experimental feasibility of our proposed setup and parameter regime.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

An all optical approach for comprehensive in-operando analysis of radiative and nonradiative recombination processes in GaAs double heterostructures

We demonstrate an all optical approach that can surprisingly offer the possibility of yielding much more information than one would expect, pertinent to the carrier recombination dynamics via both radiative and nonradiative processes when only one dominant deep defect level is present in a semiconductor material. By applying a band-defect state coupling model that explicitly treats the inter-band radiative recombination and Shockley–Read–Hall (SRH) recombination via the deep defect states on an equal footing for any defect center occupation fraction, and analyzing photoluminescence (PL) as a function of excitation density over a wide range of the excitation density (e.g., 5–6 orders in magnitude), in conjunction with Raman measurements of the LO-phonon plasmon (LOPP) coupled mode, nearly all of the key parameters relevant to the recombination processes can be obtained. They include internal quantum efficiency (IQE), minority and majority carrier density, inter-band radiative recombination rate (W r ), minority carrier nonradiative recombination rate (W nr ), defect center occupation fraction (f), defect center density (N t ), and minority and majority carrier capture cross-sections (σ t and σ tM ). While some of this information is thought to be obtainable optically, such as IQE and the W r /W nr ratio, most of the other parameters are generally considered to be attainable only through electrical techniques, such as current-voltage (I-V) characteristics and deep level transient spectroscopy (DLTS). Following a procedure developed herein, this approach has been successfully applied to three GaAs double-heterostructures that exhibit two distinctly different nonradiative recombination characteristics. The method greatly enhances the usefulness of the simple PL technique to an unprecedented level, facilitating comprehensive material and device characterization without the need for any device processing.

14 SOLAR ENERGY↗
DOE OSTI2023

Accelerating Nature: Induced Atomic Order in Equiatomic FeNi

Abstract The production of locally atomically ordered FeNi (known by its meteoric mineral name, tetrataenite) is confirmed in bulk samples by simultaneous conversion X‐ray and backscattered γ‐ray 57 Fe Mössbauer spectroscopy. Up to 22 volume percent of the tetragonal tetrataenite phase is quantified in samples thermally treated under simultaneous magnetic‐ and stress‐field conditions for a period of 6 weeks, with the remainder identified as the cubic FeNi alloy. In contrast, all precursor samples consist only of the cubic FeNi alloy. Data from the processed alloys are validated using Mössbauer parameters derived from natural meteoritic tetrataenite. The meteoritic tetrataenite exhibits a substantially higher degree of atomic order than do the processed samples, consistent with their low uniaxial magnetocrystalline anisotropy energy of ≈1 kJ·m −3 . These results suggest that targeted refinements to the processing conditions of FeNi will foster greater atomic order and increased magnetocrystalline anisotropy, leading to an enhanced magnetic energy product. These outcomes also suggest that deductions concerning paleomagnetic conditions of the solar system, as derived from meteoritic data, may warrant re‐examination and re‐evaluation. Additionally, this work strengthens the argument that tetrataenite may indeed become a member of the advanced permanent magnet portfolio, helping to meet rapidly escalating green energy imperatives.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Results of an interlaboratory study on the working curve in vat photopolymerization

The working curve informs resin properties and print parameters for stereolithography, digital light processing, and other photopolymer additive manufacturing (PAM) technologies. First demonstrated in 1992, the working curve measurement of cure depth vs radiant exposure of light is now a foundational measurement in the field of PAM. Despite its widespread use in industry and academia, there is no formal method or procedure for performing the working curve measurement, raising questions about the utility of reported working curve parameters. Here, an interlaboratory study (ILS) is described in which 24 individual laboratories performed a working curve measurement on an aliquot from a single batch of PAM resin. The ILS reveals that there is enormous scatter in the working curve data and the key fit parameters derived from it. Further, the measured depth of light penetration D p varied by as much as 7x between participants, while the critical radiant exposure for gelation E c varied by as much as 70x. This significant scatter is attributed to a lack of common procedure, variation in light engines, epistemic uncertainties from the Jacobs equation, and the use of measurement tools with insufficient precision. The ILS findings highlight an urgent need for procedural standardization and better hardware characterization in this rapidly growing field.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Modeling powder spreadability in powder-based processes using the discrete element method

Powder-bed fusion (PBF) processes refer to a subset of Additive Manufacturing (AM) techniques where powder is spread on the build-plate before melting (by a laser or electron beam). While PBF processes are attractive due to their ability for realizing complex structures that are either difficult or impossible to create through conventional means, the parts fabricated with these techniques can exhibit defects such as pores, inclusions, and excessive surface roughness. To minimize these defects, much research has been dedicated towards process maturation by optimizing laser or electron beam parameters. However, these developmental efforts typically do not address the recoating process where achieving dense and uniform layers of powder is a necessity for ensuring process repeatability and part quality. While the recoating process can be studied through experimentation, the dynamics of particle movement are difficult to analyze experimentally. Therefore, here, in this study, powder spreading in PBF was simulated through the Discrete Element Method (DEM) to elucidate the mechanisms that control powder-bed quality. Utilizing the Buckingham Pi theorem, a dimensionless metric referred to as the spreading index is developed that combines powder-bed density, roughness, and particle size to assess the quality of powder layers. The formulated spreading index is then related to several dimensionless quantities that provide insight into the mechanisms dominating powder spreading in PBF. The DEM simulations conducted in this work focused on the scenario where powder is spread onto an existing powder bed and revealed that a reduction in the recoating velocity causes an increase in the spreading index while little to no impact on the spreading index was observed when varying layer thickness from 30 μm to 75 μm.Particle size effects on the powder-bed quality were also investigated.

36 MATERIALS SCIENCE↗
DOE OSTI2020

On Transfer Learning Of Neural Networks Using Bi-Fidelity Data For Uncertainty Propagation

Due to their high degree of expressiveness, neural networks have recently been used as surrogate models for mapping inputs of an engineering system to outputs of interest. Once trained, neural networks are computationally inexpensive to evaluate and remove the need for repeated evaluations of computationally expensive models in uncertainty quantification applications. However, given the highly parameterized construction of neural networks, especially deep neural networks, accurate training often requires large amounts of simulation data that may not be available in the case of computationally expensive systems. In this paper, to alleviate this issue for uncertainty propagation, we explore the application of transfer learning techniques using training data generated from both high- and low-fidelity models. We explore two strategies for coupling these two datasets during the training procedure, namely, the standard transfer learning and the bi-fidelity-weighted learning. In the former approach, a neural network model mapping the inputs to the outputs of interest is trained based on the low-fidelity data. The high-fidelity data are then used to adapt the parameters of the upper layer(s) of the low-fidelity network, or train a simpler neural network to map the output of the low-fidelity network to that of the high-fidelity model. In the latter approach, the entire low-fidelity network parameters are updated using data generated via a Gaussian process model trained with a small high-fidelity dataset. The parameter updates are performed via a variant of stochastic gradient descent with learning rates given by the Gaussian process model. Using three numerical examples, we illustrate the utility of these bi-fidelity transfer learning methods where we focus on accuracy improvement achieved by transfer learning over standard training approaches.

97 MATHEMATICS AND COMPUTING↗