Search NASA⌕ Search

SEARCH · Search NASA

Results for “Generalization error”

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 217 records · Page 12
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Qualitative and Quantitative Evaluation for Representative Human Reliability Analysis Methods

The Korea Institute of Nuclear Safety (KINS) is the regulatory expert organization established by the Korean government to strengthen the nation’s technical capabilities relating to nuclear safety regulation. KINS oversees the technical aspects of nuclear safety regulation, including safety reviews, inspections, education, and safety research—all conducted based on technical knowledge and accumulated regulatory experience. In 2023, KINS requested that Idaho National Laboratory (INL) validates representative human reliability analysis (HRA) methods used throughout the world, thus affording KINS with a basis for determining an HRA method adequate for its domestic regulatory purposes. The present paper mainly examines INL’s efforts in this regard. The resulting INL study covered four representative HRA methods widely used by nuclear utilities and regulatory institutes. These methods were qualitatively evaluated by applying specific evaluation criteria and determining how well each method reflected critical HRA issues. For this assessment, INL benchmarked the Halden International HRA Empirical Study. Using the Halden empirical data, along with information on human failure events (HFEs), the present study employed the selected HRA methods to estimate human error probabilities (HEPs) for the HFEs. It also performed statistical analyses to compare the HEPs predicted via the HRA methods against those from the Halden empirical data.

99 - GENERAL AND MISCELLANEOUS↗
DOE OSTI2025

Navigating the Noise: Bringing Clarity to ML Parameterization Design With O $\boldsymbol{\mathcal{O}}$(100) Ensembles

Abstract Machine‐learning (ML) parameterizations of subgrid processes (here of turbulence, convection, and radiation) may one day replace conventional parameterizations by emulating high‐resolution physics without the cost of explicit simulation. However, uncertainty about the relationship between offline and online performance (i.e., when integrated with a large‐scale general circulation model) hinders their development. Much of this uncertainty stems from limited sampling of the noisy, emergent effects of upstream ML design decisions on downstream online hybrid simulation. Our work rectifies the sampling issue via the construction of a semi‐automated, end‐to‐end pipeline for size ensembles of hybrid simulations, revealing important nuances in how systematic reductions in offline error manifest in changes to online error and online stability. For example, removing dropout and switching from a Mean Squared Error to a Mean Absolute Error loss both reduce offline error, but they have opposite effects on online error and online stability. Other design decisions, like incorporating memory, converting moisture input from specific humidity to relative humidity, using batch normalization, and training on multiple climates do not come with any such compromises. Finally, we show that ensemble sizes of may be necessary to reliably detect causally relevant differences online. By enabling rapid online experimentation at scale, we can empirically settle debates regarding subgrid ML parameterization design that would have otherwise remained unresolved in the noise.

Lin, Jerry [Department of Earth System Sciences Un↗
DOE OSTI2024

Ensemble Monte Carlo calculations with five novel moves

We introduce five novel types of Monte Carlo (MC) moves that brings the number of moves of ensemble MC calculations from three to eight. So far such calculations have relied on affine invariant stretch moves that were originally introduced by Christen (2007), walk moves by Goodman and Weare (2010) and quadratic moves by Militzer (2023). Ensemble MC methods have been very popular because they harness information about the fitness landscape from a population of walkers rather than relying on expert knowledge. Here we modified the affine method and employed a simplex of points to set the stretch direction. We adopt the simplex concept to quadratic moves. We also generalize quadratic moves to arbitrary order. Finally, we introduce directed moves that employ the values of the probability density while all other types of moves rely solely on the location of the walkers. We apply all algorithms to the Rosenbrock density in 2 and 20 dimensions and to the ring potential in 12 and 24 dimensions. We evaluate their efficiency by comparing error bars, autocorrelation time, travel time, and the level of cohesion that measures whether any walkers were left behind. Our code is open source.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Structural coherence model for predicting molten salt thermal conductivity informed by the pair distribution function

To enable thermal behavior prediction and design optimization of molten salt reactors, thermal conductivity of molten salts must be characterized in terms of salt composition and temperature. Current theoretical models fail to provide consistent approximations for all halide mixtures, particularly actinide-bearing melts. This study aims to link the short-range order structure of molten salts to the mean free path of energy carriers through a simple structural coherence model informed by the partial pair distribution function. The proposed method is used to predict the thermal conductivity of 33 alkali and alkaline earth halide salts. Predictions approximate experimental measurements with a mean absolute error of 15.7% for dissociating, complexing, and actinide salts, including unary LiCl, NaCl, and MgCl 2 as well as mixtures LiF–NaF–KF (FLiNaK), LiF–BeF 2 (FLiBe), and NaCl–UCl 3 . The work provides evidence for the validity of energy carrier descriptions of molecular-level heat transfer in molten salts, with implications for improved theories of liquid energy transport in general.

Actinide mixtures↗
DOE OSTI2026

Assessing the Limitations of Self-Interaction-Corrected Functionals for Describing the Hydrated Electron

Simulating the hydrated electron using density functional theory is challenging due to the prevalence of self-interaction error in standard functionals. Hybrid functionals like PBEh(40) can reasonably describe the chemistry of an excess electron in water and partially mitigate self-interaction error by incorporating exact Hartree–Fock exchange, but they are computationally expensive making them impractical for large-scale and long-time ab initio molecular dynamics simulations. Explicit self-interaction correction schemes that are applied on an orbital-by-orbital basis offer a potential alternative when the correction is limited to the singly occupied molecular orbital obtained with a generalized gradient approximation functional. Here, we examine whether the Perdew–Zunger self-interaction correction scheme applied to the revPBE functional can provide a computationally efficient and physically sensible alternative to PBEh(40) for the hydrated electron. We find that functionals incorporating a self-interaction correction scheme should be viewed with caution when applied to the hydrated electron and its reactivity. Furthermore, we show that it is critical to consider extensive sampling and diverse chemical environments when validating their performance.

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

Beyond real: alternative unitary cluster Jastrow models for molecular electronic structure calculations on near-term quantum computers

Near-term quantum devices require wavefunction ansätze that are expressive while also of shallow circuit depth in order to both accurately and efficiently simulate molecular electronic structure. While the unitary coupled cluster ansatz (e.g., UCCSD) has become a standard, the high gate count associated with the implementation of this limits its feasibility on noisy intermediate-scale quantum (NISQ) hardware. k -Fold unitary cluster Jastrow (uCJ) ansätze mitigate this challenge by providing O( kN 2 ) circuit scaling and favorable linear depth circuit implementation. Previous work has focused on the real orbitalrotation (Re-uCJ) variant of uCJ, which allows an exact (Trotter-free) implementation. Here we extend and generalize the k -fold uCJ framework by introducing two new variants, Im-uCJ and g-uCJ, which incorporate imaginary and fully complex orbital rotation operators, respectively. Similar to Re-uCJ, both of the new variants achieve quadratic gate-count scaling. Our results focus on the simplest k = 1 model, and show that the uCJ models frequently maintain energy errors within chemical accuracy (∼1 kcal mol −1 ). Both g-uCJ and Im-uCJ are more expressive in terms of capturing electron correlation and are also more accurate than the earlier Re-uCJ ansatz. We further show that Im-uCJ and g-uCJ circuits can also be implemented exactly, without any Trotter decomposition. Numerical tests using k = 1 on H 2 , H 3 + , Be 2 , C 2 H 4 , C 2 H 6 and C 6 H 6 in various basis sets confirm the practical feasibility of these shallow Jastrow-based ansätze for applications on near-term quantum hardware.

Tkachenko, Nikolay V. [University of California, B↗
DOE OSTI2025

Accurate and efficient predictions of keyhole dynamics in laser materials processing using machine learning-aided simulations

The keyhole phenomenon has been widely observed in laser materials processing, including laser welding, remelting, cladding, drilling, and additive manufacturing. Keyhole-induced defects, primarily pores, dramatically affect the performance of final products, impeding the broad use of these laser-based technologies. The formation of these pores is typically associated with the dynamic behavior of the keyhole. So far, the accurate characterization and prediction of keyhole features, particularly keyhole depth, as a function of time, has been a challenging task. In situ characterization of keyhole dynamic behavior using the synchrotron X-ray technique is informative but complicated and expensive. Current simulations are generally hindered by their poor accuracy and generalization abilities in predicting keyhole depths due to the lack of accurate laser absorptance data. In this study, we develop a machine learning-aided simulation method that accurately predicts keyhole dynamics, especially in keyhole depth fluctuations, over a wide range of processing parameters. In two case studies involving titanium and aluminum alloys, we achieve keyhole depth prediction with a mean absolute percentage error of 10 %, surpassing those simulated using the ray-tracing method with an error margin of 30 %, while also reducing computational time. This exceptional fidelity and efficiency empower our model to serve as a cost-effective alternative to synchrotron experiments. Our machine learning-aided simulation method is affordable and readily deployable for a large variety of materials, opening new doors to eliminate or reduce defects for a wide range of laser materials processing techniques.

Computational fluid dynamics↗
DOE OSTI2024

Implicit-explicit Runge-Kutta for radiation hydrodynamics I: Gray diffusion

Radiation hydrodynamics are a challenging multiscale and multiphysics set of equations. To capture the relevant physics of interest, one typically must time step on the hydrodynamics timescale, making explicit integration the obvious choice. On the other hand, the coupled radiation equations have a scaling such that implicit integration is effectively necessary in non-relativistic regimes. A first-order Lie-Trotter-like operator split is the most common time integration scheme used in practice, alternating between an explicit hydrodynamics step and an implicit radiation solve and energy deposition step. However, such a scheme is limited to first-order accuracy, and nonlinear coupling between the radiation and hydrodynamics equations makes a more general additive partitioning of the equations non-trivial. Here, we develop a new formulation and partitioning of radiation hydrodynamics with gray diffusion that allows us to apply (linearly) implicit-explicit Runge-Kutta time integration schemes. In conclusion, we prove conservation of total energy in the new framework, and demonstrate 2nd-order convergence in time on multiple radiative shock problems, achieving error 3–5 orders of magnitude smaller than the first-order Lie-Trotter operator split at the hydrodynamic CFL, even when Lie-Trotter applies a 3rd-order TVD Runge-Kutta scheme to the hydrodynamics equations.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2025

Forecast for a growth-rate measurement using peculiar velocities from LSST supernovae

We investigate whether the cosmic growth-rate parameter fσ 8 can be measured using peculiar velocities (PVs) derived from type Ia supernovae (SNe Ia) in the Vera C. Rubin Observatory’s Legacy Survey of Space and Time (LSST). We produced simulations of different SN types using a realistic LSST observing strategy that incorporated noise, a photometric detection from the difference-image analysis (DIA) pipeline, and a PV field modeled from the Uchuu universe machine simulations. We tested three different observational scenarios that ranged from ideal conditions with spectroscopic host galaxy redshifts and spectroscopic SN typing to realistic photometric typing that resulted in a contamination with non-Ia SNe. Using a maximum likelihood technique, we showed that the LSST can measure fσ 8 with a precision of 10% in the redshift range 0.02 < z < 0.14 for our most realistic scenario. In three tomographic bins, the LSST will be able to constrain the growth-rate parameter with errors below 18% up to redshift z = 0.14. We also tested the contamination effect on the maximum likelihood method and found that for a contamination fraction below ∼2%, we recovered unbiased measurements. The results of this analysis highlight that the LSST SN sample is expected to complement traditional redshift-space distortion measurements at high redshift. This will provide a novel avenue for testing general relativity and different dark energy models.

Rosselli, D↗
DOE OSTI2025

Melting point of group IV and V transition metal diborides

The environmental conical nozzle levitator system enables cooling trace experiments above 4000 K. These cooling trace experiments were used to determine the melting points through observation of thermal arrest of group IV and V transition metal diborides (TiB 2 , T m = 3313 ± 33 K; ZrB 2 , T m = 3404 ± 38 K; HfB 2 , T m = 3529 ± 33 K; and TaB 2 , T m = 3284 ± 10 K; note: mean ± 2 standard error). Temperature measurements were conducted utilizing dual single-color pyrometers at wavelengths of 0.9 and 0.65 µm. This system utilizes aerodynamic levitation and dual laser heating to achieve temperatures approaching 4000 K. The samples were synthesized from commercial powders using ceramic gel-casting to produce high-density spherical components. The measurements obtained were compared to previously reported melting temperature values and were generally found to be in agreement with most of the published values.

Materials science↗
DOE OSTI2025

High-resolution national mapping of natural gas composition substantially updates methane leakage impacts

Methane is emitted from oil and gas operations alongside heavier hydrocarbons and non-hydrocarbon gases, shaping emissions management decision-making, including air quality impacts. Yet, most assessments assume fixed gas composition, overlooking significant spatial and temporal variations. Here, we generate a high-resolution, data-driven map of natural gas composition across the United States, reconstructing methane, heavier hydrocarbons, and non-hydrocarbon species using spatio-temporal interpolation and oil-and-gas production patterns. Our approach is able to reduce composition prediction errors by 39% in terms of Mean Absolute Error (MAE) compared to standard techniques and reveals that methane loss rates have been underestimated by more than 50% in some regions. Beyond methane, we uncover substantial variability in co-emitted gases, exposing blind spots in current emissions inventories and emissions management frameworks. Our work enables more accurate emissions assessments, guides targeted measurement strategies, and informs emissions management decision-making. It also provides a general framework for prediction in environmental applications that integrate sparse measurements with auxiliary variables.

03 NATURAL GAS↗
DOE OSTI2026

Modeling the behavior of concentrated aqueous HNO 3 using machine learning interatomic potentials

We develop two multi-defect machine learning interatomic potentials (MLIPs) trained at the BLYP-D2 and PBE-D3 density functional theories using the DeepMD-kit, allowing for the investigation of structural and thermodynamic properties of nitric acid over a wide range of concentrations via molecular dynamics (MD) simulations. We directly compute the degree of dissociation, α, and pK a from MD simulations, revealing that HNO 3 behaves as a weaker acid at higher concentrations, noting that our standard-state pK a value is in excellent agreement with the experimental one. In general, good agreement is observed with experimental results such as α and density outside the training dataset, with only modest deviations at low-to-medium concentrations. We benchmark our custom multi-defect DeepMD MLIPs against foundational models MACE-MP0 and MACE-OFF23. The foundation models capture some aspects of HNO 3 /NO 3 − solvation in concentrated nitric acid but show noticeable density errors and miss subtle structural features relevant to spectroscopy, whereas the bespoke DeepMD MLIPs yield more compact solvation shells, reproduce density-concentration trends, and run ∼12–15× faster than MACE-MP0. Although classical FFs are still more efficient and match experimental densities better, they lack chemical reactivity and thus cannot predict α or pK a , underscoring the need for system-specific reactive MLIPs beyond universal MLIPs.

Dinpajooh, Mohammadhasan [Pacific Northwest Nation↗
DOE OSTI2024

Understanding Mixed Precision GEMM with MPGemmFI: Insights into Fault Resilience

Emerging deep learning workloads urgently need fast general matrix multiplication (GEMM). Thus, one of the critical features of machine-learning-specific accelerators such as NVIDIA Tensor Cores, AMD Matrix Cores, and Google TPUs is the support of mixed-precision enabled GEMM. For DNN models, lower-precision FP data formats and computation offer acceptable correctness but significant performance, area, and memory footprint improvement. While promising, the mixed-precision computation on error resilience remains unexplored. To this end, we develop a fault injection framework that systematically injects fault into the mixed-precision computation results. We investigate how the faults affect the accuracy of machine learning applications. Based on the characteristics of error resilience, we offer lightweight error detection and correction solutions that significantly improve the overall model accuracy by 75% if the models experience hardware faults. The solutions can be efficiently integrated into the accelerator's pipelines.

Fang, Bo↗
DOE OSTI2024

AI-Driven Crack Detection for Remanufacturing Cylinder Heads Using Deep Learning and Engineering-Informed Data Augmentation

Detecting cracks in cylinder heads traditionally relies on manual inspection, which is time-consuming and susceptible to human error. As an alternative, automated object detection utilizing computer vision and machine learning models has been explored. However, these methods often face challenges due to a lack of sufficiently annotated training data, limited image diversity, and the inherently small size of cracks. Addressing these constraints, this paper introduces a novel automated crack-detection method that enhances data availability through a synthetic data generation technique. Unlike general data augmentation practices, our method involves copying cracks from one location to another, guided by both random and informed engineering decisions about likely crack formations due to cyclic thermomechanical loads. The innovative aspect of our approach lies in the integration of domain-specific engineering knowledge into the synthetic generation process, which substantially improves detection accuracy. We evaluate our method’s effectiveness using two metrics: the F2 score, which emphasizes recall to prioritize detecting all potential cracks, and mean average precision (MAP), a standard measure in object detection. Experimental results demonstrate that, without engineering insights, our method increases the F2 score from 0.40 to 0.65, while maintaining a stable MAP. Incorporating detailed engineering knowledge further enhances the F2 score to 0.70 and improves MAP to 0.57, representing increases of 63% and 43%, respectively. These results confirm that our approach not only mitigates the limitations of traditional data augmentation but also significantly advances the reliability and precision of crack detection in industrial settings.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

Problem-tailored Simulation of Energy Transport on Noisy Quantum Computers

The transport of conserved quantities like spin and charge is fundamental to characterizing the behavior of quantum many-body systems. Numerically simulating such dynamics is generically challenging, which motivates the consideration of quantum computing strategies. However, the relatively high gate errors and limited coherence times of today's quantum computers pose their own challenge, highlighting the need to be frugal with quantum resources. In this work we report simulations on quantum hardware of infinite-temperature energy transport in the mixed-field Ising chain, a paradigmatic many-body system that can exhibit a range of transport behaviors at intermediate times. We consider a chain with L = 12 sites and find results broadly consistent with those from ideal circuit simulators over 90 Trotter steps, containing up to 990 entangling gates. To obtain these results, we use two key problem-tailored insights. First, we identify a convenient basis &#x2013; the Pauli Y basis &#x2013; in which to sample the infinite-temperature trace and provide theoretical and numerical justifications for its efficiency relative to, e.g., the computational basis. Second, in addition to a variety of problem-agnostic error mitigation strategies, we employ a renormalization strategy that compensates for global nonconservation of energy due to device noise. We discuss the applicability of the proposed sampling approach beyond the mixed-field Ising chain and formulate a variational method to search for a sampling basis with small sample-to-sample fluctuations for an arbitrary Hamiltonian. This opens the door to applying these techniques in more general models.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2025

FedCSpc: A Cross-Silo Federated Learning System With Error-Bounded Lossy Parameter Compression

Cross-Silo federated learning is widely used for scaling deep neural network (DNN) training over data silos from different locations worldwide while guaranteeing data privacy. Communication has been identified as the main bottleneck when training large-scale models due to large-volume model parameters and gradient transmission across public networks with limited bandwidth. Most previous works focus on gradient compression, while limited work tries to compress parameters that can not be ignored and extremely affect communication performance during the training. Here, to bridge this gap, we propose FedCSpc: an efficient cross-silo federated learning system with an XAI-driven adaptive parameter compression strategy for large-scale model training. Our work substantially differs from existing gradient compression techniques due to the distinct data features of gradient and parameter. The key contributions of this paper are fourfold. (1) Our designed FedCSpc proposes to compress the parameter during the training using the state-of-the-art error-bounded lossy compressor – SZ3. (2) We develop an adaptive compression error bound adjustment algorithm to guarantee the model accuracy effectively. (3) We exploit an efficient approach to utilize the idle CPU resources of clients to compress the parameters. (4) We perform a comprehensive evaluation with a wide range of models and benchmarks on a GPU cluster with 65 GPUs. Results show that FedCSpc can achieve the same model accuracy as FedAvg while reducing the data volume of parameters and gradients in communication by up to 7.39× and 288×, respectively. With 32 clients on a 4 Gb size model, FedCSpc significantly outperforms FedAvg in wall-clock time in the emulated WAN environment (at the bandwidth of 1 Gbps or lower without loss of generality).

SZ3↗
DOE OSTI2025

VARI3D & PERSENT: Perturbation and Sensitivity Analysis

The nodal diffusion method is one of the most widely used approaches in modern reactor analysis. In the nodal diffusion method, a coarse multi-group set of “homogenized” parameters is constructed such that the complex geometry of a reactor core along with the energy dependence of neutron and gamma ray cross sections in a nuclear reactor are conserved in the simpler geometry. The homogenization is typically done on a fuel assembly level as is the case in the DIF3D code developed at Argonne National Laboratory. The nodal methodology is used primarily to predict fuel cycle behavior of nuclear systems of which there is a substantial amount of validation in the literature. Another use of the nodal method is to obtain reactivity coefficients and kinetics parameters for use in a safety analysis of a given nuclear reactor. While there are many ways to obtain reactivity worth and kinetics parameters, the work presented in this manuscript is unique as it provides the user with the ability to compute reactivity worths, kinetics parameters, and cross section sensitivities with a Cartesian and hexagonal geometry based transport code. This manuscript serves as a single manual for two separate codes: VARI3D and PERSENT. The VARI3D code (VARIational 3D) is based upon the classic finite difference diffusion theory solver available in DIF3D. The PERSENT code (PERturbation and SENitivity for Transport) is based upon the variational nodal method employed in DIF3D termed VARIANT. The VARIANT solver was added to DIF3D in 1995 and has seen continued development and use for the last 18 years. Because VARI3D primarily uses deprecated coding practices, rather than incorporating the perturbation and sensitivity treatments for transport within VARI3D, a new coding development was built using modern Fortran coding. The primary purpose of this manual is to describe the theory behind PERSENT (and by convenience, that of VARI3D) and discuss the input and output of PERSENT along with giving potential users an idea of how to use it. While this manuscript does describe the input and output of VARI3D, the PERSENT code is intended to be the replacement capability of VARI3D as PERSENT can generate nearly identical (if not superior) diffusion theory results. In this manuscript, the relevant aspects of generalized perturbation theory and exact perturbation theory that apply to both VARI3D and PERSENT are covered. The input and output of VARI3D is displayed by excerpting several of the example problems. Similarly, the input and output of PERSENT is displayed along with tips on how best to use the code. Note that the input and output of the inhomogeneous solver wrapped around DIF3D (DIF3D_IFS) is also discussed as it is needed to carry out some of the sensitivities in PERSENT such as reaction rate ratios. This manuscript describes several perturbation and sensitivity problems, and the results computed using PERSENT. From these sections, potential users should find that PERSENT provides not only the typical tables of numbers desired in perturbation and sensitivity analysis work, but also can visually plot the result for a more thorough understanding of the space and energy distribution (Section 5). Overall, PERSENT is observed to produce accurate reactivity worths and sensitivities for the displayed set of test problems and clearly demonstrates the need to have a transport-based sensitivity capability as evident from the thousands of percent errors observed in the 21-group hexagonal fast reactor problem (covered in Section 7). The uncertainty calculation capability is described in Section 3 and demonstrated in Section 7.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2026

Optimization of Random Phase Approximation Calculations for Improved Energies of Molecules, Solids, and Surfaces

We present an optimized random phase approximation method (optRPA26) that significantly improves upon conventional RPA through an optimized choice of reference orbitals and energy components, rather than a modification of the RPA correlation functional itself. The method employs an empirically constructed hybrid functional to generate DFT orbitals to evaluate the RPA correlation energy, which is then scaled by a constant. Comprehensive benchmarks across molecules, bulk solids, and surface systems demonstrate that optRPA26 consistently achieves high accuracy, with mean absolute errors of 0.05 eV for W4-11-RE reaction energies, 0.07 eV for cohesive energies, 0.09 eV for metal oxide formation energies, 0.11–0.12 eV for adsorption of small molecules on metals, and 0.06 eV for adsorption on oxides. In addition, optRPA26 correctly captures phase stability in metal oxides and magnetic metals. The optRPA26 approach can be run using standard RPA implementations, highlighting its potential as a general-purpose reference method that can accurately capture covalent, ionic, metallic, and van der Waals bonding in molecules, solids, and interfaces.

Adsorption↗