Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “uncertainty estimation”

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 OSTI2021

Sensitivity and Uncertainty of the IFR-1 BISON Benchmark

The fuel performance code BISON is being used to evaluate metallic fuel for a new fast-spectrum test reactor called the Versatile Test Reactor, which is being considered by the US Department of Energy. To quantify the accuracy of BISON predictions, researchers at Oak Ridge National Laboratory have been developing a series of benchmarks based on legacy metallic fuel experiments. As part of this effort, the sensitivity of BISON predictions to variations in model inputs and the uncertainties associated with BISON predictions must be established. This report summarizes efforts to perform a comprehensive sensitivity analysis (SA) and uncertainty quantification (UQ) on a benchmark based on the IFR-1 experiment. For the SA, at least one input was chosen from every BISON model and physics module used in the benchmark. The inputs were varied individually in a series of BISON simulations. The resulting variations in benchmark predictions were normalized to calculate sensitivities. The strongest sensitivities were identified and used to inform input selections for the UQ. The UQ was performed using the Monte Carlo UQ method. A literature review was conducted to estimate uncertainty distributions for the selected inputs, and values were sampled randomly from each distribution in a series of BISON simulations. Variations in the benchmark predictions were used to estimate uncertainty distributions and confidence intervals. It was found that nearly 100% of benchmark predictions matched the corresponding legacy values within the confidence intervals. However, this is at least partially because of the wide confidence intervals associated with the benchmark predictions. The uncertainty contributions of assumptions in the benchmark, experimental uncertainties, and BISON models were quantified. Some analysis was performed to identify inputs that contributed to the uncertainties. Finally, recommendations are made for future benchmark development and future BISON development.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2025

Bayesian Optimized Deep Ensemble for Uncertainty Quantification of Deep Neural Networks: a System Safety Case Study on Sodium Fast Reactor Thermal Stratification Modeling

Deep neural networks (DNNs) are increasingly important to scientific computing and engineering system simulations. Accurate uncertainty quantification (UQ) for DNNs is critical in safety-sensitive engineering domains. Traditional Deep Ensemble (DE) methods, while easy to implement, frequently suffer from poorly calibrated uncertainty estimates and limited predictive accuracy due to reliance on fixed architectures with varied weight initializations. To address these issues, we introduce a workflow that combines Bayesian Optimization (BO) and DE. The workflow is modular, scalable, and integrates parallel BO initialized with Sobol sequences to individually optimize the hyperparameters of each ensemble member. This method enhances ensemble diversity, improves predictive accuracy, and provides reliable uncertainty estimates. We evaluate the proposed BODE approach in a sodium fast reactor thermal stratification modeling case study, where we used a densely connected convolutional neural network to predict turbulent viscosity during the reactor transient with consideration of data noise. We benchmark its performance against several optimization approaches, including baseline deep ensemble, evolutionary algorithm-optimized ensemble, ensemble formed via random search combined with greedy selection, and a BO ensemble using random initialization. Here, our results demonstrate superior performance of the developed BODE approach. In noise-free scenarios, BODE notably reduces incorrect aleatoric uncertainty and significantly enhances predictive accuracy. Under conditions of 5% and 10% Gaussian noise, BODE adaptively quantifies uncertainty proportional to data noise, achieving up to an 80% reduction in root mean square error compared to baseline methods and producing well-calibrated prediction intervals.

Bayesian optimization↗
NASA NTRS1999

Monte Carlo mixture model of lifetime cancer incidence risk from radiation exposure on shuttle and international space station

Estimating uncertainty in lifetime cancer risk for human exposure to space radiation is a unique challenge. Conventional risk assessment with low-linear-energy-transfer (LET)-based risk from Japanese atomic bomb survivor studies may be inappropriate for relativistic protons and nuclei in space due to track structure effects. This paper develops a Monte Carlo mixture model (MCMM) for transferring additive, National Institutes of Health multiplicative, and multiplicative excess cancer incidence risks based on Japanese atomic bomb survivor data to determine excess incidence risk for various US astronaut exposure profiles. The MCMM serves as an anchor point for future risk projection methods involving biophysical models of DNA damage from space radiation. Lifetime incidence risks of radiation-induced cancer for the MCMM based on low-LET Japanese data for nonleukemia (all cancers except leukemia) were 2.77 (90% confidence limit, 0.75-11.34) for males exposed to 1 Sv at age 45 and 2.20 (90% confidence limit, 0.59-10.12) for males exposed at age 55. For females, mixture model risks for nonleukemia exposed separately to 1 Sv at ages of 45 and 55 were 2.98 (90% confidence limit, 0.90-11.70) and 2.44 (90% confidence limit, 0.70-10.30), respectively. Risks for high-LET 200 MeV protons (LET=0.45 keV/micrometer), 1 MeV alpha-particles (LET=100 keV/micrometer), and 600 MeV iron particles (LET=180 keV/micrometer) were scored on a per particle basis by determining the particle fluence required for an average of one particle per cell nucleus of area 100 micrometer(2). Lifetime risk per proton was 2.68x10(-2)% (90% confidence limit, 0.79x10(-3)%-0. 514x10(-2)%). For alpha-particles, lifetime risk was 14.2% (90% confidence limit, 2.5%-31.2%). Conversely, lifetime risk per iron particle was 23.7% (90% confidence limit, 4.5%-53.0%). Uncertainty in the DDREF for high-LET particles may be less than that for low-LET radiation because typically there is very little dose-rate dependence. Probability density functions for high-LET radiation quality and dose-rate may be preferable to conventional risk assessment approaches. Nuclear reactions and track structure effects in tissue may not be properly estimated by existing data using in vitro models for estimating RBEs. The method used here is being extended to estimate uncertainty in spacecraft shielding effectiveness in various space radiation environments.

NASA Center LaRC↗
NASA NTRS1999

SeaWiFS Postlaunch Technical Report Series: The SeaWiFS Solar Radiation-Based Calibration and the Transfer-to-Orbit Experiment - Volume 5

The solar radiation-based calibration (SRBC) of the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) was performed on 1 November 1993. Measurements were made outdoors in the courtyard of the instrument manufacturer. SeaWiFS viewed the solar irradiance reflected from the sensor's diffuser in the same manner as viewed on orbit. The calibration included measurements using a solar radiometer designed to determine the transmittances of principal atmospheric constituents. The primary uncertainties in the outdoor measurements are the transmission of the atmosphere and the reflectance of the diffuser. Their combined uncertainty is about 5 or 6%. The SRBC also requires knowledge of the extraterrestrial solar spectrum. Four solar models are used. When averaged over the responses of the SeaWiFS bands, the irradiance models agree at the 3.6% level, with the greatest difference for SeaWiFS band 8. The calibration coefficients from the SRBC are lower than those from the laboratory calibration of the instrument in 1997. For a representative solar model, the ratios of the SRBC coefficients to laboratory values average 0.962 with a standard deviation of 0.012. The greatest relative difference is 0.946 for band 8. These values are within the estimated uncertainties of the calibration measurements. For the transfer-to-orbit experiment, the measurements in the manufacturer's courtyard are used to predict the digital counts from the instrument on its first day on orbit (August 1, 1997). This experiment requires an estimate of the relative change in the diffuser response for the period between the launch of the instrument and its first solar measurements on orbit (September 9, 1997). In relative terms, the counts from the instrument on its first day on orbit averaged 1.3% higher than predicted, with a standard deviation of 1.2% and a greatest difference of 2.4% or band 7. The estimated uncertainty for the transfer-to-orbit experiment is about 3 or 4%.

Hooker, Stanford B.↗
NASA NTRS2020

Chrono-Validation of Near-Real-Time Landslide Susceptibility Models via Plugin Statistical Simulations

The idea behind any validation scheme in landslide susceptibility studies is to test whether a model calibrated on a certain data can predict an unknown dataset of the same nature (landslide presences/absences and covariates). Almost the entirety of landslide susceptibility studies are validated by subsetting a single dataset into a training and test sets. This dataset usually corresponds either to event-specific or to historical inventories. Very rarely, a multi-temporal inventory is available and, in the few cases where this condition is met, the validation practices involve training a model on a specific landslide inventory, deriving a single predictive equation and validating it on a subsequent landslide inventory. This commonly leads landslide predictive studies, even those with a strong statistical rigor, to neglect the uncertainty estimation in their modeling scheme. In statistics, validation can also be performed via statistical simulations. This means that after fitting a given model, one can generate any number of predictive functions and test their predictive skills on any type and number of unknown datasets. In this work, we take a similar direction and we apply it to model and validate three separate co-seismic inventories, including an uncertainty estimation phase. We mapped these inventories within the same area in Indonesia, for three earthquakes occurred in 2012, 2017 and 2018. Specifically, we build three event-specific Bayesian Generalize Additive Models of the binomial family. From each model we then simulate 1000 predictive realizations over the remaining two inventories, by using a plug-in scheme where all the morphometric covariates are kept fixed and only the ground motion is replaced according to the prediction target. By doing so, we introduce a new analytical tool for near-real-time landslide predictive purposes, which is able to produce a probabilistic model which stands in between the definitions of susceptibility and hazard. In fact, our model is able to accurately estimate “where” and “when” - although not “how frequently” - landslide have occurred by featuring the multitemporal information of the trigger. In our findings, the simulations are quite similar to the fitted models; and the nine combinations we analyse produce excellent performance. This result confirms the assumption that “the past is the key to the future”, as we show that the relative contribution of each variable and their interactions in each probabilistic model remains practically the same across temporal replicates. This information is not trivial because it supports the routines implemented in global near-real-time applications.

Temporal validation↗
DOE OSTI2024

Bridging the time scale in exascale computing of chemical systems (Final Technical Report)

This report summarizes the work carried out with support of the United States Department of Energy under Award DE-SC0019441. The theme of this project was to develop and apply methods that allowed for the acceleration of atomistic calculations, particularly in challenging areas such as multiphase systems, electrified interfaces, uncertainty estimation, and applications requiring chemical accuracy, which tend to be applications where simulation time is severely bottlenecked by the computational time requirements. Much of the focus was on the application of emerging machine-learning methodologies, although a wide range of methodologies were employed. This report has two major sections. The first focuses on the methodological advances themselves. Within this part, we report a number of major advances, a few examples of which are described here. We report the first machine-learning scheme for the acceleration of electronically grand-canonical calculations (that is, those applicable to electrochemistry). We report new methods of performing transfer learning, in which physics-based priors can be used to provide predictions, often with uncertainty estimates, of images well outside of training sets; we also offer ways to fine-tune these transfer-learning models. We provide a new systematic means to generate and apply minimal training data sets to very large (10,000’s of atoms) systems, with only small training sets appropriate for electronic structure. We developed new methodologies to integrate surface vibrations into surface adsorption calculations. We made advances to the applicability of diffusion Monte Carlo methods to allow (learned) force prediction, finite-size error correction, and force-free means of searching for transition states. We integrated machine-learned atomistic predictions into mechanism generation codes. Additionally, we released new software including AmpTorch, a modernized version of our original atomistic machine-learning code Amp. The second part of this report focuses on the scientific applications that accompanied, and were often enabled by, the methodological advances described earlier. A few examples follow, but full details are in the individual chapters of the report. For example, we developed a general theory of phonon-induced friction on molecular adsorbates. We showed fundamentally how solvent influences the adsorption and desorption process and how it differs from the processes typically involved at the solid–gas interface, making aqueous-phase and electrocatalysis different from traditional thermocatalysis. We examined how metal–insulator and magnetic transitions can be probed, and accelerated exciton dynamics via Frenkel Hamiltonian parameters. We showed that the nearsighted force-training approach, developed within this project, can predict both the stability and reactivity of large nanoparticles, and can also lead to insights on catalyst coverage on binding energies and entropies. These applied studies, which generally integrated with our method development, allowed us to push forward the theoretical understanding of several reaction classes.

08 HYDROGEN↗
DOE OSTI2022

Uncertainties in estimating global potential yields and their impacts for long-term modeling

Estimating realistic potential yields by crop type and region is challenging; such yields depend on both biophysical characteristics (e.g., soil characteristics, climate, etc.), and the crop management practices available in any site or region (e.g., mechanization, irrigation, crop cultivars). A broad body of literature has assessed potential yields for selected crops and regions, using several strategies. In this study we first analyze future potential yields of major crop types globally by two different estimation methods, one of which is based on historical observed yields (“Empirical”), while the other is based on biophysical conditions (“Simulated”). Potential yields by major crop and region are quite different between the two methods; in particular, Simulated potential yields are typically 200% higher than Empirical potential yields in tropical regions for major crops. Applying both of these potential yields in yield gap closure scenarios in a global agro-economic model, GCAM, the two estimates of future potential yields lead to very different outcomes for the agricultural sector globally. In the Simulated potential yield closure scenario, Africa, Asia, and South America see comparatively favorable outcomes for agricultural sustainability over time: low land use change emissions, low crop prices, and high levels of self-sufficiency. In contrast, the Empirical potential yield scenario is characterized by a heavy reliance on production and exports in temperate regions that currently practice industrial agriculture. At the global level, this scenario has comparatively high crop commodity prices, and more land allocated to crop production (and associated land use change emissions) than either the baseline or Simulated potential yield scenarios. This study highlights the importance of the choice of methods of estimating potential yields for agro-economic modeling.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Simulation of hardened cement degradation and estimation of uncertainty in predicted failure times with peridynamics

Modeling the degradation of cement-based infrastructure due to aqueous environmental conditions continues to be a challenge. In order to develop a capability to predict concrete infrastructure failure due to chemical degradation, here we created a chemomechanical model of the effects of long-term water exposure on cement paste. The model couples the mechanical static equilibrium balance with reactive–diffusive transport and incorporates fracture and failure via peridynamics (a meshless simulation method). The model includes fundamental aspects of degradation of ordinary Portland cement (OPC) paste, including the observed softening, reduced toughness, and shrinkage of the cement paste, and increased reactivity and transport with water induced degradation. This version of the model focuses on the first stage of cement paste decalcification, the dissolution of portlandite. Given unknowns in the cement paste degradation process and the cost of uncertainty quantification (UQ), we adopt a minimally complex model in two dimensions (2D) in order to perform sensitivity analysis and UQ. We calibrate the model to existing experimental data using simulations of common tests such as flexure, compression and diffusion. Then we calculate the global sensitivity and uncertainty of predicted failure times based on variation of eleven unique and fundamental material properties. We observed particularly strong sensitivities to the diffusion coefficient, the reaction rate, and the shrinkage with degradation. Also, the predicted time of first fracture is highly correlated with the time to total failure in compression, which implies fracture can indicate impending degradation induced failure; however, the distributions of the two events overlap so the lead time may be minimal. Extension of the model to include the multiple reactions that describe complete degradation, viscous relaxation, post-peak load mechanisms, and to three dimensions to explore the interactions of complex fracture patterns evoked by more realistic geometry is straightforward and ongoing.

36 MATERIALS SCIENCE↗
DOE OSTI2021

A Systems Approach to Estimating the Uncertainty Limits of X-Ray Radiographic Metrology

Micro- and nanomanufacturing capabilities have rapidly expanded over the past decade to include complex three-dimensional (3D) structure fabrication; however, the metrology required to accurately assess these processes via part inspection and characterization has struggled to keep pace. X-ray computed tomography (CT) is considered an ideal candidate for providing the critically needed metrology on the smallest scales, especially internal features, or inaccessible regions. X-ray CT supporting micro- and nanomanufacturing often push against the poorly understood resolution and variation limits inherent to the machines, which can distort or hide fine structures. In this study, we have developed and experimentally verify a comprehensive analytical uncertainty propagation signal variation flow graph (SVFG) model for X-ray radiography in this work to better understand resolution and image variability limits on the small scale. The SVFG approach captures, quantifies, and predicts variations occurring in the system that limit metrology capabilities, particularly in the micro/nanodomain. This work is the first step to achieving full uncertainty modeling of CT reconstructions and provides insight into improving X-ray attenuation imaging systems. The SVFG methodology framework is applied to generate a complete basis set of functions describing the major sources of variation in radiographs. Five models are identified, covering variation in energy, intensity, length, blur, and position. Radiographic system experiments are defined to measure the parameters required by the SVFGs. Best practices are identified for these measurements. The SVFG models are confirmed via direct measurement of variation to predict variation within 30% on average.

47 OTHER INSTRUMENTATION↗
DOE OSTI2022

Estimating epistemic uncertainty in soil parameters for nonlinear site response analyses: Introducing the Latin Hypercube Sampling technique

This study quantifies the effects of epistemic uncertainty in soil parameters on nonlinear (NL) site response analysis (SRA) results, validated against the data recorded at a well-instrumented geotechnical downhole array located in Japan. To this end, a one-dimensional soil column model of the Service Hall Array (SHA) near the Kashiwazaki-Kariwa Nuclear Power Plant (KKNPP) is developed using the finite element (FE) program LS-DYNA. The dynamic stress–strain relationship is characterized by a modified two-stage hyperbolic (MTH) NL backbone curve formulation capable of capturing soil behavior at both small- and large-shear strains. The model is then validated against the ground motion recordings to capture the model bias. The uncertainties associated with the shear-wave velocity profile (a small-strain soil property) and soil shear strength (a large-strain soil property) are incorporated in NL SRA to quantify their separate and joint randomization effects on the results. This study proposes using the Latin Hypercube Sampling (LHS) method as an efficient alternative to commonly used methods, such as Standard Monte Carlo (SMC), to account for uncertainty propagation in such reliability analysis. Both low-intensity and design-level records from the recordings at the SHA are applied to study the contribution of the small- and large-strain NL dynamic soil properties. Results from 46,200 NL FE analyses (23,100 per input motion) are presented. Measured and predicted site response, using recorded ground motions at this downhole array, is compared to assess the significance of soil parameter uncertainty on the observed ground motion dispersions. It is demonstrated that increasing the number of soft realizations and implementing higher level earthquake intensity lead to higher ground motion dispersion. Unlike past studies in randomization of Vs profiles with the SMC method, the LHS method is shown to have no significant effect on the predicted median surface response spectra and amplification factors (AFs) for this case study.

Engineering↗
DOE OSTI2021

Local-Distance Seismic Event Relocation and Relative Magnitude Estimation, Applications to Mining Related Seismicity in the Powder River Basin, Wyoming

Recent efforts to characterize small ( M w < 3 ) seismic events at local distances have become more important because of the increased observation of human-triggered and induced seismicity and the need to advance nuclear explosion monitoring capabilities. The signals generated by low-magnitude seismic sources necessitate the use of nearby short-period observations, which are sensitive to local geological heterogeneity. Local to near-regional distance ( < 300 km ) surface and shear waves can dominate short-period observations from small, shallow seismic sources. In this work, we utilize these observations to estimate precise, relative locations and magnitudes of ~ 700 industrial mining events in Wyoming, using nearly 360,000 observations. The precise, relative location estimates (with formal location uncertainty estimates of less than 1 km) collapse a diffuse collection of mining events into discrete clusters associated with individual blasting operations. We also invert the cross-correlation amplitudes to estimate precise, relative moment magnitude estimates, which help validate and identify disparities in the event sizes reported by regional network catalogs. Joint use of multiple phases allows for the inclusion of more seismic events due to the increase in the number of observations. In some cases, using a single phase allowed us to relocate only 50% of the original reported seismic events within a cluster. Combining shear- and surface-wave phases increased the number of events to above 90% of the original events, allowing us to characterize a broader range of event sizes, source to station distances, and event distributions. This analysis takes a step toward making a fuller characterization of small industrial seismic events observed at local distances.

58 GEOSCIENCES↗
DOE OSTI2021

YAHFC: A Code Framework to Model Nuclear Reactions and Estimate Correlated Uncertainties

Reaction modeling is a key ingredient in designing experiments and interpreting their results, and is an essential component in the process of evaluating nuclear data and assembling nuclear data libraries used in nuclear technology applications. Typically, experimental data are available only for a handful of reaction channels and theory models are used to fill in the gaps. In addition, theory is often called upon as the arbitrator between discrepant data. Most importantly, theory and modeling are required for an accurate determination of uncertainties in the evaluated data and the correlations between the multiple channels. A fast, accurate, and flexible modeling capability has been developed at LLNL with the code system YAHFC (Yet Another Hauser-Feshbach Code). YAHFC is a Monte Carlo, Hauser-Feshbach code framework, making full use of dynamic memory allocation, derived types, and parallel computing. YAHFC can generate events to simulate experiments and is guiding experiments designed to measure inelastic neutron scattering from actinide targets. YAHFC is also being used to analyze decays from surrogate experiments, thereby enabling the inference of reaction cross sections inaccessible by direct measurement. Finally, by modeling nuclear reactions with constraints from experimental data, YAHFC can deliver complete nuclear data libraries, with evaluated uncertainties, using the modernized Generalized Nuclear Data Structure (GNDS).

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2021

Neural Density Estimation and Uncertainty Quantification for ChemCam Spectra [Slides]

The ChemCam instrument of Curiosity uses laser-induced breakdown spectroscopy (LIBS). It fires a laser at target and vaporizes rock surfaces, creating a plasma. Three spectrographs divide the plasma light into wavelengths for chemical analysis: ultraviolet, violet, and visible near-infrared. Regression methods (SVR, PCR, CNN) have been employed for calibration (prediction of the elemental composition of samples); however, labeled ChemCam samples are limited. Here, we focus on unsupervised learning and employ generative models from ChemCam analysis. Further, we use labels (supervised) in combination to the generative model to compute uncertainties related to predictions. We report generative modeling can be successfully applied to model real-world data. Normalizing flow models can be efficiently constructed on latent spaces for fast downstream inference. Unsupervised and supervised learning can be combined to form an uncertainty quantification framework.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗