Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Model Inference”

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 91 records · Page 5
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

A GPU‐Accelerated Generative Adversarial Model for Causal Inference

We develop a GPU-accelerated machine learning generative adversarial model designed to facilitate causal inferences from observational data. Our model's theoretical framework is conceptualized in a manner that is amenable to being operable and scalable for high-performance computing platforms. We leverage GPU acceleration to develop a parallel evolutionary algorithm to achieve large-scale parallel computation of the model within a now widely accessible computing platform. This capability both enhances computational speedup and efficiency and also extends the use of the model to a broader range of substantive research domains while maintaining the underlying theoretical properties of the model.

GPU↗
DOE OSTI2025

Bayesian calibration of bubble size dynamics applied to CO 2 gas fermenters

To accelerate the scale-up of gaseous CO 2 fermentation reactors, computational models need to predict gas-to-liquid mass transfer which requires capturing the bubble size dynamics, i.e. bubble breakup and coalescence. However, the applicability of existing models beyond air–water mixtures remains to be established. Here, an inverse modeling approach, accelerated with a neural network surrogate, calibrates the breakup and coalescence closure models, that are used in class methods for population balance modeling (PBM). The calibration is performed based on experimental results obtained in a CO 2 -air–water-coflowing bubble column reactor. Bayesian inference is used to account for noise in the experimental dataset and bias in the simulation results. To accurately capture gas holdup and interphase mass transfer, the results show that the breakage rate needs to be increased by one order of magnitude. In conclusion, the inferred model parameters are then used on a separate configuration and shown to also improve bubble size distribution predictions.

09 BIOMASS FUELS↗
DOE OSTI2019

DiaMonD: An Integrated Multifaceted Approach to Mathematics at the Interfaces of Data, Models, and Decisions

For many problems in computational science and engineering, the so-called forward problem—solution of the underlying mathematical model to yield output quantities of interest, given input parameters—is difficult enough for frontier complex models, which are often characterized by coupled multiphysics and possibly stochastic behavior over a wide range of length and time scales. However, we face the need to move beyond the forward problem, to address what is often the ultimate goal of CS&E: decision making. This requires us to tackle a spectrum of mathematical problems that subsume, and are thus even more difficult than, the forward problem. First, given experimental data, we wish to estimate unknown parameters characterizing a model of a natural or engineered system by solving an inverse problem. Second, we seek the optimal configuration of the system (or experiments) by solving an optimal design problem. Third, the optimal operation of the system must be determined by solving an optimal control problem. And fourth, we must quantify uncertainties as they propagate through all of the preceding problems, from data to model inference to prediction and finally to optimal design and control. Optimal mathematical methods and algorithms for such end-to-end, data-to-decisions modeling and simulation of complex problems—requiring integrated solution of forward, inverse, optimization, and UQ problems for large, multiphysics, multiscale models—entails challenges of the highest order, while also presenting great opportunities for applied mathematics research. Unfortunately, research on forward, inverse, optimization, and UQ problems has all-too-often progressed in isolation. This has led to mathematical methods that perform well when considering the forward simulation in isolation, but are prohibitive or suboptimal or unstable when combined with other methods within the framework of inversion, optimization, or UQ. The reverse is also true: general-purpose methods developed within the inversion, optimization, and UQ fields often become prohibitive when applied to complex CS&E problems since they do not exploit their structure.

58 GEOSCIENCES↗
DOE OSTI2019

Scalable optimal Bayesian classification of single-cell trajectories under regulatory model uncertainty

Single-cell gene expression measurements offer opportunities in deriving mechanistic understanding of complex diseases, including cancer. However, due to the complex regulatory machinery of the cell, gene regulatory network (GRN) model inference based on such data still manifests significant uncertainty. The goal of this paper is to develop optimal classification of single-cell trajectories accounting for potential model uncertainty. Partially-observed Boolean dynamical systems (POBDS) are used for modeling gene regulatory networks observed through noisy gene-expression data. We derive the exact optimal Bayesian classifier (OBC) for binary classification of single-cell trajectories. The application of the OBC becomes impractical for large GRNs, due to computational and memory requirements. To address this, we introduce a particle-based single-cell classification method that is highly scalable for large GRNs with much lower complexity than the optimal solution. The performance of the proposed particle-based method is demonstrated through numerical experiments using a POBDS model of the well-known T-cell large granular lymphocyte (T-LGL) leukemia network with noisy time-series gene-expression data.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2019

Multi-Formalism Modeling for Disaster Resilience, Forecasting, and Response

Disaster response preparedness, risk mitigation, and effective post-event response are critical for national security, public health, and economic integrity. For areas impacted by extreme events, timely and accurate impact assessments of lifeline infrastructures are critical. Traditional disaster management and situational awareness platforms are not capable of handling the massive influx of available information from traditional and non-traditional sources. Next-generation analytic assessment approaches, where use of dynamically sourced, heterogeneous data with near real-time availability are integrated into a continually-updated analytics framework. The objective of this research was to develop a foundational multi-formalism modeling platform to integrate novel data sources for use in improving situational awareness after a disaster event. Specifically, to 1) provide an architecture for machine-learning based data fusion various multi-sensor datasets; 2) classify, geolocate, and temporally bin novel data sources for use in spatially-enabled machine-learning based data fusion; and 3) derive confidence-based observation data with machine-learning based prediction to fill spatial and temporal observation gaps. This project developed a reusable cloud-based pipeline architecture for ingesting, storing, and analyzing relevant data, executing a range of deep-learning models, and visualize the information geographically. The pipeline development was informed by previous PNNL investments (StreamKit) and built upon leading-edge technologies consisting of: 1) a raw datastore; 2) a data enrichment pipeline with custom and third-party analytics; 3) enriched data storage; and 4) a custom UI to display the enriched data. Machine learning/deep learning models were selected and tested to classify social media images for relevancy and presence of flood waters. Our custom trained relevance models were run on a validation dataset and the overall accuracy of the relevance classifiers ranged from 65% to 90%. We found that convolutional neural network (CNN) models were both faster and more accurate than more traditional machine learning models. We applied the relevance classifier to remove non-relevant images and then undertook a limited training of CNN models to detect flood on ~1,800 labeled images. The testing accuracy of the flood detection models were ~88%. To improve the geographic location of both social media and oblique aerial images we tested a multi-step workflow built on a pre-trained deep-learning model. To identify similarity features from test and reference scenes, we applied the Xception and Faiss models to a set of test images with inaccurate spatial locations, as well and precisely-located reference scenes from the general area. Matching reference scenes to the test scenes with the closest similarity index had limited success in improving the precision of test scene location and additional methods have been identified for testing. Data fusion between authoritative geographic data (satellite imaging) and non-authoritative crowdsourced data is an underlying theme throughout this research and methods are tested bring these data to a common geographic form through feature matching and transformations. The automated extraction of supporting information for disaster response from oblique aerial images or ground-level social media images is accomplished through semantic image segmentation and transformation. The non-authoritative data are used to validate the authoritative data collections and processing and optionally, t+0 flood models. In addition, the common form geographic data are combined with other static landscape features to develop spatially-explicit training vectors used in deep generative statistical inference models to develop complex non-linear relationships in static and dynamic data and ultimately derive a spatially continuous and probabilistic assessment of flood damage with the event domain.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2026

Machine learning at the Spallation Neutron Source accelerator and target

We describe the ongoing efforts to apply Machine Learning techniques to improve the performance of our accelerator and target. Specially, we are looking to minimize halo beam losses in the absence of a proper physics model, automatically detect and log anomalies in the target support systems such as cooling, and detect and prevent errant beam pulses in the linac. We also describe the infrastructure we use to acquire and stream data to the GPU cluster for training, our code development cycle, and edge computing for model inference. To minimize halo beam losses, we use a Reinforcement Learning technique tested on a virtual accelerator. The target anomaly detection is trained on archived data using incomplete physics models and is made part of the existing target reporting system. The errant beam prevention analyzes beam current and beam phase waveforms as well as accelerator configuration data to predict errant pulses. We also develop continual learning to adapt to changes in the accelerator.

Accelerator Physics↗
DOE OSTI2025

Theoretical and methodological challenges in hierarchical Bayesian inference for model-form uncertainty

This report describes challenges associated with the hierarchical Bayesian approach to inform model-form uncertainty (MFU) representations, which are parameterized modifications to a mathematical models’ governing equations to express uncertainty in form of the equations. To inform model-form uncertainties, hierarchical Bayesian inference is often employed. Here, the MFU parameters are distributed parametrically, and the hyperparameters of the parametric distribution are informed through Bayesian inference, with the aim of determining the MFU parameter distribution that best agrees with calibration data. In practice, however, we have found the hierarchical Bayesian approach falls short of this aim. We discuss theoretical and methodological challenges of the approach, and we present several numerical demonstrations of these challenges. To conclude, we suggest promising alternative approaches for future investigation.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1993

The cosmic-ray He-3/He-4 ratio from 100 to 1600 MeV/amu

The Superconducting Magnet Instrument for Light Isotopes (SMILI) flew for 19 hours on September 1, 1989, with a residual overburden of 5 g/sq cm. It measured the charge, rigidity, and velocity of 30,000 cosmic-ray helium nuclei, with velocity determined by time-of-flight and Cerenkov techniques. Using these data, the flux and isotopic composition of helium as a function of energy were determined. The observed isotopic composition is consistent with that expected from interstellar propagation models inferred from the secondaries of CNO, in contrast to earlier observations which indicated an overabundance of He-3. We discuss constraints that this result places on cosmic-ray transport and solar modulation models.

Beatty, J. J.↗
DOE OSTI2025

Protein-ligand binding affinity prediction using multi-instance learning with docking structures

Recent advances in 3D structure-based deep learning approaches demonstrate improved accuracy in predicting protein-ligand binding affinity in drug discovery. These methods complement physics-based computational modeling such as molecular docking for virtual high-throughput screening. Despite recent advances and improved predictive performance, most methods in this category primarily rely on utilizing co-crystal complex structures and experimentally measured binding affinities as both input and output data for model training. Nevertheless, co-crystal complex structures are not readily available and the inaccurate predicted structures from molecular docking can degrade the accuracy of the machine learning methods. We introduce a novel structure-based inference method utilizing multiple molecular docking poses for each complex entity. Our proposed method employs multi-instance learning with an attention network to predict binding affinity from a collection of docking poses. We validate our method using multiple datasets, including PDBbind and compounds targeting the main protease of SARS-CoV-2. The results demonstrate that our method leveraging docking poses is competitive with other state-of-the-art inference models that depend on co-crystal structures. This method offers binding affinity prediction without requiring co-crystal structures, thereby increasing its applicability to protein targets lacking such data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Single Episode Policy Transfer in Reinforcement Learning

Transfer and adaptation to new unknown environmental dynamics is a key challenge for reinforcement learning (RL). An even greater challenge is performing near-optimally in a single attempt at test time, possibly without access to dense rewards, which is not addressed by current methods that require multiple experience rollouts for adaptation. To achieve single episode transfer in a family of environments with related dynamics, we propose a general algorithm that optimizes a probe and an inference model to rapidly estimate underlying latent variables of test dynamics, which are then immediately used as input to a universal control policy. This modular approach enables integration of state-of-the-art algorithms for variational inference or RL. Moreover, our approach does not require access to rewards at test time, allowing it to perform in settings where existing adaptive approaches cannot. In diverse experimental domains with a single episode test constraint, we report our method significantly outperforms existing adaptive approaches and shows favorable performance against baselines for robust transfer.

42 ENGINEERING↗
DOE OSTI2021

PRISM3D: a 3-D reference seismic model for Iberia and adjacent areas

SUMMARY We present PRISM3D, a 3-D reference seismic model of P- and S-wave velocities for Iberia and adjacent areas. PRISM3D results from the combination of the most up-to-date earth models available for the region. It extends horizontally from 15°W to 5°E in longitude, 34°N to 46°N in latitude and vertically from 3.5 km above to 200 km below sea level, and is modelled on a regular grid with 10 and 0.5 km of grid node spacing in the horizontal and vertical directions, respectively. It was designed using models inferred from local and teleseismic body-wave tomography, earthquake and ambient noise surface wave tomography, receiver function analysis and active source experiments. It includes two interfaces, namely the topography/bathymetry and the Mohorovičić (Moho) discontinuity. The Moho was modelled from previously published receiver function analysis and deep seismic sounding results. To that end we used a probabilistic surface reconstruction algorithm that allowed to extract the mean of the Moho depth surface along with its associated standard deviation, which provides a depth uncertainty estimate. The Moho depth model is in good agreement with previously published models, although it presents slightly sharper gardients in orogenic areas such as the Pyrenees or the Betic-Rif system. Crustal and mantle P- and S-wave wave speed grids were built separately on each side of the Moho depth surface by weighted average of existing models, thus allowing to realistically render the speed gradients across that interface. The associated weighted standard deviation was also calculated, which provides an uncertainty estimation on the average wave speed values at any point of the grid. At shallow depths (<10 km), low P and S wave speeds and high VP/VS are observed in offshore basins, while the Iberian Massif, which covers a large part of western Iberia, appears characterized by a rather flat Moho, higher than average VP and VS and low VP/VS. Conversely, the Betic-Rif system seems to be associated with low VP and VS, combined with high VP/VS in comparison to the rest of the study area. The most prominent feature of the mantle is the well known high wave speed anomaly related to the Alboran slab imaged in various mantle tomography studies. The consistency of PRISM3D with previous work is verified by comparing it with two recent studies, with which it shows a good general agreement.The impact of the new 3-D model is illustrated through a simple synthetic experiment, which shows that the lateral variations of the wave speed can produce traveltime differences ranging from –1.5 and 1.5 s for P waves and from –2.5 and 2.5 s for S waves at local to regional distances. Such values are far larger than phase picking uncertainties and would likely affect earthquake hypocentral parameter estimations. The new 3-D model thus provides a basis for regional studies including earthquake source studies, Earth structure investigations and geodynamic modelling of Iberia and its surroundings.

Arroucau, P.↗
DOE OSTI2024

Increasing the power of weak lensing data with multipole-based intrinsic alignment estimators

ABSTRACT It has long been known that galaxy shapes align coherently with the large-scale density field. Characterizing this effect is essential to interpreting measurements of weak gravitational lensing, the deflection of light from distant galaxies by matter overdensities along the line of sight, as it also produces coherent galaxy alignments that we wish to interpret in terms of a cosmological model. Existing direct measurements of intrinsic alignments using galaxy samples with high-quality shape and redshift measurements typically use well-understood but sub-optimal projected estimators, which do not make good use of the information in the data when comparing those estimators to theoretical models. We demonstrate a more optimal estimator, based on a multipole expansion of the correlation functions or power spectra, for direct measurements of galaxy intrinsic alignments. We show that even using the lowest order multipole alone increases the significance of inferred model parameters using simulated and real data, without any additional modelling complexity. We apply this estimator to measurements of parameters of the non-linear alignment model using data from the Sloan Digital Sky survey, demonstrating consistent results with a factor of ∼2 greater precision in parameter fits to intrinsic alignments models. This result is functionally equivalent to quadrupling the survey area, but without the attendant costs – thereby demonstrating the value in using this new estimator in current and future intrinsic alignments measurements using spectroscopic galaxy samples.

Astronomy & Astrophysics↗
DOE OSTI2022

Nuclear masses learned from a probabilistic neural network

Machine learning methods and uncertainty quantification have been gaining interest throughout the last several years in low-energy nuclear physics. In particular, Gaussian processes and Bayesian neural networks have increasingly been applied to improve mass model predictions while providing well-quantified uncertainties. In this work, we use the probabilistic Mixture Density Network (MDN) to directly predict the mass excess of the 2016 Atomic Mass Evaluation within the range of measured data, and we extrapolate the inferred models beyond available experimental data. The MDN provides not only mean values but also full posterior distributions both within the training set and extrapolated testing set. We show that the addition of physical information to the feature space increases the accuracy of the match to the training data as well as provides for more physically meaningful extrapolations beyond the the limits of experimental data.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
NASA NTRS1994

Location of the emitting regions in the mode-switching radio pulsar PSR 1926 + 18

We present an analysis of average single-frequency profiles of the radio pulsar PSR 1926 + 18 in two different modes. Assuming a purely dipolar field and taking into account retardation, aberration, magnetic line curvature, and ignoring the magnetic field line sweepback, we come to the following conclusions: (1) mode switching in PSR 1926 + 18 at 430 MHz may be understood in terms of different emission heights of the two modes; (2) the weaker mode appears when regions emitting the first two components move away from the star by an additional distance of not more than several kilometers from their original locations; (3) the trailing conal component is emitted essentially from the same layer of the magnetosphere in both modes, in contrast to the first conal component, which appears to change its location in the magnetosphere; (4) core and conal components are emitted from distinctly different emission heights: core components not more than 75 km, and conal components not less than 215 km from the surface of the star. Our model infers three-dimensional information about the emitting region from single-frequency observations. It has been tested previously with a mode switching pulsar that had one component in the average profile in two modes. This paper shows that the same model can explain positions and widths of the components in a more complex profile during mode switching, provided that the components can be traced from one mode to another.

Nowakowski, L. A.↗
DOE OSTI2022

Inference, Prediction, & Entropy-Rate Estimation of Continuous-Time, Discrete-Event Processes

Inferring models, predicting the future, and estimating the entropy rate of discrete-time, discrete-event processes is well-worn ground. However, a much broader class of discrete-event processes operates in continuous-time. Here, we provide new methods for inferring, predicting, and estimating them. The methods rely on an extension of Bayesian structural inference that takes advantage of neural network’s universal approximation power. Based on experiments with complex synthetic data, the methods are competitive with the state-of-the-art for prediction and entropy-rate estimation.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
NASA NTRS2014

A Corrected Formulation of the Multilayer Model (MLM) for Inferring Gaseous Dry Deposition to Vegetated Surfaces

The Multilayer Model (MLM) has been used for many years to infer dry deposition fluxes from measured trace species concentrations and standard meteorological measurements for national networks in the U.S., including the U.S. Environmental Protection Agency's Clean Air Status and Trends Network (CASTNet). MLM utilizes a resistance analogy to calculate deposition velocities appropriate for whole vegetative canopies, while employing a multilayer integration to account for vertically varying meteorology, canopy morphology and radiative transfer within the canopy. However, the MLM formulation, as it was originally presented and as it has been subsequently employed, contains a non-physical representation related to the leaf-level quasi-laminar boundary layer resistance that affects the calculation of the total canopy resistance. In this note, the non-physical representation of the canopy resistance as originally formulated in MLM is discussed and a revised, physically consistent, formulation is suggested as a replacement. The revised canopy resistance formulation reduces estimates of HNO3 deposition velocities by as much as 38% during mid-day as compared to values generated by the original formulation. Inferred deposition velocities for SO2 and O3 are not significantly altered by the change in formulation (less than 3%). Inferred deposition loadings of oxidized and total nitrogen from CASTNet data may be reduced by 10-20% and 5-10%, respectively, for the Eastern U. S. when employing the revised formulation of MLM as compared to the original formulation.

vegetated surfaces↗
NASA NTRS1981

Inference of dust opacities for the 1977 Martian great dust storms from Viking Lander 1 pressure data

The tidal heating components for the dusty Martian atmosphere are computed based on dust optical parameters estimated from Viking Lander imaging data, and used to compute the variation of the tidal surface pressure components at the Viking Lander sites as a function of season and the total vertical extinction optical depth of the atmosphere. An atmospheric tidal model is used which is based on the inviscid, hydrostatic primitive equations linearized about a motionless basic state the temperature of which varies only with height, and the profiles of the tidal forcing components are computed using a delta-Eddington approximation to the radiative transfer equations. Comparison of the model results with the observed variations of surface pressure and overhead dust opacity at the Viking Lander 1 site reveal that the dust opacities and optical parameters derived from imaging data are roughly representative of the global dust haze necessary to reproduce the observed surface pressure amplitudes, with the exception of the model-inferred asymmetry parameter, which is smaller during the onset of a great storm. The observed preferential enhancement of the semidiurnal tide with respect to the diurnal tide during dust storm onset is shown to be due primarily to the elevation of the tidal heating source in a very dusty atmosphere.

Zurek, R. W.↗
DOE OSTI2021

Inferring pesticide toxicity to honey bees from a field‐based feeding study using a colony model and Bayesian inference

Abstract Honey bees are crucial pollinators for agricultural crops but are threatened by a multitude of stressors including exposure to pesticides. Linking our understanding of how pesticides affect individual bees to colony‐level responses is challenging because colonies show emergent properties based on complex internal processes and interactions among individual bees. Agent‐based models that simulate honey bee colony dynamics may be a tool for scaling between individual and colony effects of a pesticide. The U.S. Environmental Protection Agency (USEPA) and U.S. Department of Agriculture (USDA) are developing the VarroaPop + Pesticide model, which simulates the dynamics of honey bee colonies and how they respond to multiple stressors, including weather, Varroa mites, and pesticides. To evaluate this model, we used Approximate Bayesian Computation to fit field data from an empirical study where honey bee colonies were fed the insecticide clothianidin. This allowed us to reproduce colony feeding study data by simulating colony demography and mortality from ingestion of contaminated food. We found that VarroaPop + Pesticide was able to fit general trends in colony population size and structure and reproduce colony declines from increasing clothianidin exposure. The model underestimated adverse effects at low exposure (36 µg/kg), however, and overestimated recovery at the highest exposure level (140 µg/kg), for the adult and pupa endpoints, suggesting that mechanisms besides oral toxicity‐induced mortality may have played a role in colony declines. The VarroaPop + Pesticide model estimates an adult oral LD 50 of 18.9 ng/bee (95% CI 10.1–32.6) based on the simulated feeding study data, which falls just above the 95% confidence intervals of values observed in laboratory toxicology studies on individual bees. Overall, our results demonstrate a novel method for analyzing colony‐level data on pesticide effects on bees and making inferences on pesticide toxicity to individual bees.

59 BASIC BIOLOGICAL SCIENCES↗