Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Benchmark testing”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,153 records · Page 64
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

In Search of Optimum Fresh-Cut Raw Material: Using Computer Vision Systems as a Sensory Screening Tool for Browning-Resistant Romaine Lettuce Accessions

The popularity of ready-to-eat (RTE) salads has prompted novel technology to prolong the shelf life of their ingredients. Fresh-cut romaine lettuce is widely used in RTE salads; however, its tendency to quickly discolor continues to be a challenge for the industry. Selecting the ideal lettuce accessions for use in RTE salads is essential to ensure maximum shelf life, and it is critical to have a practical way to assess and compare the quality of multiple lettuce accessions that are being considered for use in fresh-cut applications. Thus, in this work we aimed to determine whether a computer vision system (CVS) composed of image acquisition, processing, and analysis could be effective to detect visual quality differences among 16 accessions of fresh-cut romaine lettuce during postharvest storage. The CVS involved a post-capturing color correction, effective image segmentation, and calculation of a browning index, which was tested as a predictor of quality and shelf life of fresh-cut romaine lettuce. The results demonstrated that machine vision software can be implemented to replace or supplement the scoring of a trained panel and instrumental quality measurements. Overall visual quality, a key sensory parameter that determines food preferences and consumer behavior, was highly correlated with the browning index, with a Pearson correlation coefficient of −0.85. Other important sensory decision parameters were also strongly or moderately correlated with the browning index, with Pearson correlation coefficients of −0.84 for freshness, 0.79 for off odor, and 0.57 for browning. The ranking of the accessions according to quality acceptability from the sensory evaluation produced a similar pattern to those obtained with the CVS. This study revealed that multiple lettuce accessions can be effectively benchmarked for their performance as fresh-cut sources via a CVS-based method. Future opportunities and challenges in using machine vision image processing to predict consumer preferences for RTE salad greens is also discussed.

Agriculture↗
DOE OSTI2022

Evidence for Dissociation and Ionization in Shock Compressed Nitrogen to 800 GPa

Triple bonding in the nitrogen molecule (N 2 ) is among the strongest chemical bonds with a dissociation enthalpy of 9.8 eV/molecule. Nitrogen is therefore an excellent test bed for theoretical and numerical methods aimed at understanding how bonding evolves under the influence of the extreme pressures and temperatures of the warm dense matter regime. Here, we report laser-driven shock experiments on fluid molecular nitrogen up to 800 GPa and 4.0 g/cm 3 . Line-imaging velocimetry measurements and impedance matching method with a quartz reference yield shock equation of state data of initially precompressed nitrogen. Comparison with numerical simulations using path integral Monte Carlo and density functional theory molecular dynamics reveals clear signatures of chemical dissociation and the onset of L-shell ionization. Furthermore, combining data along multiple shock Hugoniot curves starting from densities between 0.76 and 1.29 g/cm 3 , our study documents how pressure and density affect these changes in chemical bonding and provides benchmarks for future theoretical developments in this regime, with applications for planetary interior modeling, high energy density science, and inertial confinement fusion research.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2024

Verification of RESRAD-BUILD Code Version 4

This report documents the verification of the RESRAD-BUILD code, Version 4.0, which was released on December 22, 2022. Two earlier reports verifying Versions 3.0 and 3.1, respectively, were published in 2001 (Kamboj, et al. 2001) and 2003 (Tetra Tech NUS 2003). Version 4.0 of the RESRAD-BUILD code has many new features and modeling enhancements over the earlier versions, including the previously released Version 3.5. Chapter 2 of this report focuses on verifying the external dose and risk modeling for point, line, area, and volume sources, as well as for floor deposition. Besides verification, the external radiation doses calculated by RESRAD-BUILD were also benchmarked with those calculated by the MCNP code (Briemeister 1993). Section J.3 of the RESRAD-BUILD User’s Manual Vol. 1 (Yu et al. 2022) documents the results of that benchmarking effort. Chapter 3 of this report focuses on verifying the ventilation modeling, from checking the remaining source inventory, releases of radionuclides to the air, air concentrations and deposited floor concentrations over time, to the radiation dose and risk associated with inhalation, ingestion, and air submersion, with and without vacuuming. The verification efforts involve designing spreadsheets to perform calculations the same as or like those performed by the RESRAD-BUILD code and then comparing the spreadsheet results with those produced by the code. When the results agree or the differences are within acceptable range, the accuracy of model implementation in the code is verified. In addition to model implementation, the implementation of key functions and features that facilitate the modeling or the use of the code were also verified during the release testing of the code. Appendix A presents the test cases developed for these verification testing, and Appendix B presents the testing results that verify proper implementation of key functions and features.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Improving ProtoDUNE pion cross-section measurements with NuGraph Michel-electron tagging

Understanding hadron-argon interactions is essential for precise neutrino energy reconstruction and final-state interaction modeling in liquid-argon time projection chamber (LArTPC) experiments such as DUNE. In particular, pion absorption and charge-exchange processes constitute significant sources of systematic uncertainty in neutrino oscillation measurements. ProtoDUNE-SP, a large-scale LArTPC prototype operated at the CERN Neutrino Platform and exposed to charged-particle test beams in the few-GeV range, enables direct measurements of these processes. This work focuses on the measurement of differential cross sections for pion absorption and charge exchange using the 2 GeV/c pion beam data from the ProtoDUNE-SP run. A key component of this analysis is the identification of Michel electrons from $\pi \rightarrow \mu \rightarrow e$ decay chains, which helps separate different interaction topologies and improves background rejection. Michel electron identification will also assist in reliably calibrating the electromagnetic response in ProtoDUNE-SP data and for the future DUNE detectors. In this analysis, we apply NuGraph to identify Michel electrons. NuGraph is a graph neural network that models detector hits as nodes connected by spatial and temporal edges for particle and topology classification in LArTPC detectors. We first benchmark NuGraph’s Michel electron classification performance using ICEBERG data, a small-scale LArTPC prototype used for DUNE electronics and reconstruction development, and then transfer the approach to ProtoDUNE-SP. This poster presents the analysis strategy, NuGraph-based classification studies, and discusses how these developments are expected to improve the pion cross-section measurement.

Razafinime, Soamasina Herilala [Cincinnati U.] (OR↗
DOE OSTI2021

Solution Irregularity Remediation for Spatial Discretization Error Estimation for S N Transport Solutions

The discrete ordinates linear Boltzmann transport equation is typically solved in its spatially discretized form, incurring spatial discretization error. Quantification of this error for purposes such as adaptive mesh refinement or error analysis requires an a posteriori estimator, which utilizes the numerical solution to the spatially discretized equation to compute an estimate. Because the quality of the numerical solution informs the error estimate, irregularities, present in the true solution for any realistic problem configuration, tend to cause the largest deviation in the error estimate vis-a-vis the true error. In this paper, an analytical partial singular characteristic tracking (pSCT) procedure for reducing the estimator’s error is implemented within our novel residual source estimator for a zeroth-order discontinuous Galerkin scheme, at the additional cost of a single inner iteration. Here, a metric-based evaluation of the pSCT scheme versus the standard residual source estimator is performed over the parameter range of a Method of Manufactured Solutions test suite. The pSCT scheme generates near-ideal accuracy in the estimate in problems where the dominant source of the estimator’s error is the solution irregularity, namely, problems where the true solution is discontinuous and problems where the true solution’s first derivative is discontinuous and the scattering ratio is low. In problems where the scattering ratio is high and the true solution is discontinuous in the first derivative, the error in the scattering source, which is not converged by the pSCT scheme, is greater than the error incurred due to the irregularity. Ultimately, a pSCT scheme is judged to be useful for error estimation in problems where the computational cost of the scheme is justified. In the presence of many irregularities, such a scheme may be intractable for general use, but in benchmarks, as an analytical tool, or in problems that have nondissipative discontinuities, the scheme may prove invaluable.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2025

BACKFLIP: A Comparison of Market-Benchmark Backsheet Technologies to Novel Non-Fluoro-Based Coextruded Materials and Their Correlation and Impact on PV Module Degradation Rates: Final Results of the Study at 4000 Hours or 2 Years

As the photovoltaic (PV) industry is rapidly expanding around the world, there has been an increasing interest in extending the lifespan of PV modules. Concern has also emerged regarding the recyclability of modules and their component materials, including fluoropolymer-based backsheets. Laminated polyethylene-terephthalate (PET) core backsheets have traditionally been used in the PV industry, but new, co-extruded polyolefin (PO) backsheets show promise as an improved alternative. Mini-module and coupon samples of seven different backsheets (made of layers including contemporary PET and fluoropolymers, novel PO, and polyamide (PA) materials) were run through hygrometric- or UV photolytic-accelerated aging to identify and better understand each material's degradation modes and the backsheets' field reliability. In addition to the artificial aging, the natural weathering methods used in this study are described. The comprehensive set of chemical, mechanical, and structural characterizations at intermittent read points in this study is presented, including: visual appearance and color; gloss; mechanical tensile testing; I-V performance; electroluminescence (EL) imaging; dielectric breakdown; FTIR-chemical structure; X-ray-polymer structure (WAXS); and DSC-crystalline content. After 4000 h of accelerated aging or 2y of outdoor aging, a strong correlation occurs between initial physical characteristics (mechanical tensile test) and operating performance (EL and I-V characteristics).

14 SOLAR ENERGY↗
DOE OSTI2021

A Poisson equation method for prescribing fully developed non-Newtonian inlet conditions for computational fluid dynamics simulations in models of arbitrary cross-section

Prescribing inlet boundary conditions for computational fluid dynamics (CFD) simulations of internal flow in complex geometries such as anatomical vascular models is challenging. In the absence of patient-specific inlet velocity data, a common approach for long blood vessels is to assume that the inlet flow is fully developed. In vessels of irregular cross section, however, prescribing fully developed conditions is complicated due to the lack of a general closed-form analytical solution. In this study, we develop a simple Poisson equation method for prescribing fully developed inlet conditions for the flow of either Newtonian or non-Newtonian fluids in CFD models of arbitrary cross-section. We first derive the generalized Poisson equation for fully developed flow of a non-Newtonian fluid and we then develop and verify a methodology for numerically computing the solution on any planar boundary domain. In addition, we develop a simple extension of the method for prescribing a non-orthogonal inlet velocity that represents fully developed flow from an upstream tube that is connected to the CFD inlet at a non-orthogonal angle. This may be used to investigate a common source of uncertainty in CFD simulations of internal flow that is due to a lack of information concerning the exact streamwise flow direction at the inlets. Comparison to several Newtonian and non-Newtonian benchmark verification solutions shows the method to be extremely accurate. As a practical demonstration case, we use the method to prescribe fully developed conditions on multiple non-circular inlets for the non-Newtonian flow of blood in a patient-specific model of the inferior vena cava (IVC). Finally, we further demonstrate the utility of the method by performing a sensitivity study using the patient-specific IVC model, wherein we investigate the influence of inlet velocity flow direction on the non-Newtonian IVC hemodynamics. Given its simplicity and computational efficiency, the method is shown to be far superior to alternative approaches for prescribing fully developed inlet conditions in such complicated geometries. In conclusion, to facilitate the adoption of our Poisson equation method, we have distributed our OpenFOAM source code and the associated test cases from this study as open-source software.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Insights into the Surface Electronic Structure and Catalytic Activity of InO x /Au(111) Inverse Catalysts for CO 2 Hydrogenation to Methanol

In this article, the direct conversion of carbon dioxide (CO 2 ) into methanol (CH 3 OH) via low-temperature hydrogenation is crucial for recycling anthropogenic CO 2 emissions and producing fuels or high value chemicals. Nevertheless, it continues to be a great challenge due to the trade-off between selectivity and catalytic activity. For CO 2 hydrogenation, In 2 O 3 catalysts are known for their high CH 3 OH selectivity. Subsequent studies explored depositing metals on In 2 O 3 to enhance CO 2 conversion. Despite extensive research on metal (M) supported In 2 O 3 catalysts, the role of In-M alloys and M/In 2 O 3 interfaces in CO 2 activation and CH 3 OH selectivity remains unclear. In this work, we have examined the behavior of In/Au(111) alloys and InO x /Au(111) inverse systems during CO 2 hydrogenation using synchrotron-based ambient-pressure X-ray photoelectron spectroscopy (AP-XPS) and catalytic tests in a batch reactor. Indium forms alloys with Au(111) after deposition. The In-Au(111) alloys display high reactivity towards CO 2 and can dissociate the molecule at room temperature to generate InO x nanostructures. At very low coverages of In (≤ 0.05 ML), the InO x nanostructures are not stable under CO 2 hydrogenation conditions and the active In-Au(111) alloys produces mainly CO and little methanol. An increase in indium coverage to 0.3 ML led to stable InOx nanostructures under CO 2 hydrogenation conditions. These InO x /Au(111) catalysts displayed a high selectivity (~ 80 %) towards CH 3 OH production and an activity for CO 2 conversion that was at least 10 times larger than that of plain In 2 O 3 or Cu(111) and Cu/ZnO(000$\overline{1)}$ benchmark catalysts. The results of AP-XPS show that InO x /Au(111) produces methanol via methoxy intermediates. Inverse oxide/metal catalysts containing InOx open up a possibility for improving CO 2 → CH 3 OH conversion in processes associated with the control of environmental pollution and the production of high value chemicals.

36 MATERIALS SCIENCE↗
DOE OSTI2025

SEFOR Core I-E

This repository contains the VTB model developed for the Southwest Experimental Fast Oxide Reactor (SEFOR) core configuration I-E using the NEAMS tools. The MOOSE Reactor Module was employed to set up the geometry mesh, MC2-3 was applied to generate multigroup cross sections at various temperatures, and Griffin was used to calculate the k-eff values of Core I-E across these temperature conditions. To validate the deterministic results from Griffin, reference solutions were obtained with the Monte Carlo code Shift. The reactivity feedback was then derived and compared against SEFOR experimental data from isothermal tests in order to assess the performance of Griffin. Reproducing these results requires licensed access to several code systems, including Shift, Griffin, and MC2-3. Instructions for obtaining and accessing these codes can be found at the following locations: https://www.ornl.gov/scale/releases for SCALE/Shift, https://mooseframework.inl.gov/ncrc/applications/ncrc_root_griffin.html for Griffin, and https://www.anl.gov/nse/mc23 for MC2-3.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2025

A comparative analysis of YOLOv8 and U-Net image segmentation approaches for transmission electron micrographs of polycrystalline thin films

Metallic thin films offer a platform to experimentally study the dynamics of microstructural evolution, but the required transmission electron microscopy (TEM)-based imaging generates complex images that are challenging to segment and quantify. This work provides a comparative analysis of a new YOLOv8 model and an established U-Net model for bright-field TEM images of polycrystals, employing a framework leveraging physical observables to evaluate performance against two hand-traced benchmark datasets. This methodology obviates the comparison of large, diversely structured, and manually labeled datasets that are required to assess performance on a per-image/per-pixel basis. It is found that the YOLOv8 model, adapted for real-time instance segmentation, has up to 43× faster inferencing (NVIDIA GeForce RTX 4090) compared to U-Net and reconstructs hand-traced grain size distributions (GSDs) with excellent fidelity, finding mean diameter within 3% for grains near an optimal magnification; for grains that deviate from the optimal pixel-diameter, the size of small- (large)-diameter grains is systematically over- (under)-estimated. This is partially mitigated by including scale-aware augmentations during training. Moreover, when the bias is corrected post-inference by a rigid shift in distribution, the YOLOv8 model reproduces ground truth GSDs with exceptional fidelity, with statistical tests indicating <5% probability that the distributions are distinct. Based on ground truth data, calibration curves pertaining to this shift can be constructed for a given model. This issue is not present in the U-Net model’s results, indicating that for quantitative measurements where the true size of objects is of interest, special procedures must be implemented for YOLO-based models.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Density Measurements of Molten LiF–BeF 2 and LiF–BeF 2 –LaF 3 Salt Mixtures by Neutron Radiography

The densities of eutectic (LiF) 2 –BeF 2 and mixtures of this salt (FLiBe) with LaF 3 were measured by dilatometry and by neutron attenuation from 673 K to 1,073 K. Because LaF 3 has a limited solubility in FLiBe, it was necessary to determine the amount of LaF 3 in solution before the density could be determined. The FLiBe density determination was favorably benchmarked against the literature data. A simple comparison was not available for the LaF 3 –FLiBe mixtures, so extrapolation of published data was necessary based on analysis using the Molten Salt Thermal Properties Database-Thermochemistry, or MSTDB-TC, developed by the US Department of Energy. Solubilities for LaF 3 in FLiBe ranged from 1 to 4 mol % over 673 to 1,073 K. The salt system was heated and cooled over 24 h to evaluate potential changes in composition and hysteresis during the measurement. Changes in the meniscus were observed, and these were included in the correction for density determinations. Salt surface tension may have led to supersaturation of LaF 3 in the salt because the solubility curve was nonlinear with respect to the inverse temperature, as would be expected for an ideal system. Surface tension measurements are currently underway to test this hypothesis.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Comprehensive Basis-Set Testing of Extended Symmetry-Adapted Perturbation Theory and Assessment of Mixed-Basis Combinations to Reduce Cost

Hybrid or “extended” symmetry-adapted perturbation theory (XSAPT) replaces traditional SAPT’s treatment of dispersion with better performing alternatives while at the same time extending two-body (dimer) SAPT to a many-body treatment of polarization using a self-consistent charge embedding procedure. The present work presents a systematic study of how XSAPT interaction energies and energy components converge with respect to the choice of Gaussian basis set. Errors can be reduced in a systematic way using correlation-consistent basis sets, with aug-cc-pVTZ results converged within <0.1 kcal/mol. Similar (if slightly less systematic) behavior is obtained using Karlsruhe basis sets at much lower cost, and we introduce new versions with limited augmentation that are even more efficient. Pople-style basis sets, which are more efficient still, often afford good results if a large number of polarization functions are included. The dispersion models used in XSAPT afford much faster basis-set convergence as compared to the perturbative description of dispersion in conventional SAPT, meaning that “compromise” basis sets (such as jun-cc-pVDZ) are no longer required and benchmark-quality results can be obtained using triple-ζ basis sets. The use of diffuse functions proves to be essential, especially for the description of hydrogen bonds. As a result, the “δ(Hartree–Fock)” correction for high-order induction can be performed in double-ζ basis sets without significant loss of accuracy, leading to a mixed-basis approach that offers 4× speedup over the existing (cubic scaling) XSAPT approach.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2021

Precision test of statistical dynamics with state-to-state ultracold chemistry

Chemical reactions represent a class of quantum problems that challenge both the current theoretical understanding and computational capabilities1. Reactions that occur at ultralow temperatures provide an ideal testing ground for quantum chemistry and scattering theories, because they can be experimentally studied with unprecedented control, yet display dynamics that are highly complex. Here we report the full product state distribution for the reaction 2KRb → K 2 + Rb 2 . Ultracold preparation of the reactants allows us complete control over their initial quantum degrees of freedom, whereas state-resolved, coincident detection of both products enables the probability of scattering into each of the 57 allowed rotational state-pairs to be measured. Our results show an overall agreement with a state-counting model based on statistical theory, but also reveal several deviating state-pairs. In particular, we observe a strong suppression of population in the state-pair closest to the exoergicity limit as a result of the long-range potential inhibiting the escape of products. Finally, the completeness of our measurements provides a benchmark for quantum dynamics calculations beyond the current state of the art.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

LSTM-Based Data Integration to Improve Snow Water Equivalent Prediction and Diagnose Error Sources

Accurate prediction of snow water equivalent (SWE) can be valuable for water resource managers. Recently, deep learning methods such as long short-term memory (LSTM) have exhibited high accuracy in simulating hydrologic variables and can integrate lagged observations to improve prediction, but their benefits were not clear for SWE simulations. Here we tested an LSTM network with data integration (DI) for SWE in the western United States to integrate 30-day-lagged or 7-day-lagged observations of either SWE or satellite-observed snow cover fraction (SCF) to improve future predictions. SCF proved beneficial only for shallow-snow sites during snowmelt, while lagged SWE integration significantly improved prediction accuracy for both shallow- and deep-snow sites. The median Nash–Sutcliffe model efficiency coefficient (NSE) in temporal testing improved from 0.92 to 0.97 with 30-day-lagged SWE integration, and root-mean-square error (RMSE) and the difference between estimated and observed peak SWE values d max were reduced by 41% and 57%, respectively. DI effectively mitigated accumulated model and forcing errors that would otherwise be persistent. Moreover, by applying DI to different observations (30-day-lagged, 7-day-lagged), we revealed the spatial distribution of errors with different persistent lengths. For example, integrating 30-day-lagged SWE was ineffective for ephemeral snow sites in the southwestern United States, but significantly reduced monthly-scale biases for regions with stable seasonal snowpack such as high-elevation sites in California. These biases are likely attributable to large interannual variability in snowfall or site-specific snow redistribution patterns that can accumulate to impactful levels over time for nonephemeral sites. These results set up benchmark levels and provide guidance for future model improvement strategies.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Efficient high-fidelity TRISO statistical failure analysis using Bison: Applications to AGR-2 irradiation testing

The ability of tri-structural isotropic (TRISO) fuel to contain fission products is largely dictated by the quality of the manufacturing process, since most of the fission product release is expected to occur due to coating layer failure in a small number of particles containing defects. The Bison fuel performance code has capabilities to predict failure in individual particles, accounting for the presence of defects, and to apply statistical analysis methods to compute the probability of failure in a set of fuel particles. Bison has recently undergone significant development both to improve its physical representations of fuel particle behavior and to improve the efficiency of its statistical failure calculations. Physical model improvements include new capabilities to account for the pressure generated by fission gases on inner pyrolytic carbon (IPyC) crack surfaces and to use local material coordinate orientation to accurately incorporate the anisotropy in the material properties in aspherical particles. To improve statistical modeling efficiency, a direct integration approach which involves directly integrating the failure probability function associated with statistically varying parameters has been developed. The direct integration approach is much more efficient than the Monte Carlo (MC) schemes commonly employed, and allows Bison to directly run high-dimensional fuel performance models, which improves the accuracy of failure probability calculations. Finally, a set of benchmark problems is considered here to compare the MC and direct integration approaches, and a statistical failure analysis of compacts in the Advanced Gas Reactor (AGR)-2 experiments is performed using the direct integration approach.

36 MATERIALS SCIENCE↗
DOE OSTI2026

Multithreaded copy ('cp')

This is a modification to 'cp' and 'mv' commands to make them multi-threaded. Simple benchmarks showed that multi-threading could reduce the time to copy a large Linux source directory by over 2x. The 'cp' and 'mv' utilities are part of the existing Coreutils (https://www.gnu.org/software/coreutils/) software package that get installed on all Linux distros. Changes: * Add '-j|--parallel ' flags to 'cp' and 'mv'. This allows the utilities to recursively copy regular files in directories in parallel. This does NOT parallelize multiple single file copies to a destination (like 'cp file2 file2 file3 dst/'). Along with this, add in new 'CP_NUM_THREADS' and 'MV_NUM_THREADS' environment variables to set the number of threads. This can be useful when you want to enable parallelism by default in /etc/profile. The maximum number of threads is internally capped to the number of CPUs. * Add a '-j' flag to 'sort' to complement its existing '--parallel' flag. This is only done for consistency with 'cp' and 'mv'. * Add test cases for the new flags. Also, run each 'cp' and 'mv' test both in single-threaded and multithreaded modes for extra coverage.

Hutter, AnthonyJ [Lawrence Livermore National Labo↗
DOE OSTI2023

AI-accelerated protein-ligand docking for SARS-CoV-2 is 100-fold faster with no significant change in detection

Protein-ligand docking is a computational method for identifying drug leads. The method is capable of narrowing a vast library of compounds down to a tractable size for downstream simulation or experimental testing and is widely used in drug discovery. While there has been progress in accelerating scoring of compounds with artificial intelligence, few works have bridged these successes back to the virtual screening community in terms of utility and forward-looking development. We demonstrate the power of high-speed ML models by scoring 1 billion molecules in under a day (50 k predictions per GPU seconds). We showcase a workflow for docking utilizing surrogate AI-based models as a pre-filter to a standard docking workflow. Our workflow is ten times faster at screening a library of compounds than the standard technique, with an error rate less than 0.01% of detecting the underlying best scoring 0.1% of compounds. Our analysis of the speedup explains that another order of magnitude speedup must come from model accuracy rather than computing speed. In order to drive another order of magnitude of acceleration, we share a benchmark dataset consisting of 200 million 3D complex structures and 2D structure scores across a consistent set of 13 million “in-stock” molecules over 15 receptors, or binding sites, across the SARS-CoV-2 proteome. We believe this is strong evidence for the community to begin focusing on improving the accuracy of surrogate models to improve the ability to screen massive compound libraries 100 × or even 1000 × faster than current techniques and reduce missing top hits. The technique outlined aims to be a fast drop-in replacement for docking for screening billion-scale molecular libraries.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

CO 2 storage site characterization using ensemble-based approaches with deep generative models

Estimating spatially distributed properties such as permeability from available sparse measurements is a great challenge in efficient subsurface CO 2 storage operations. In this paper, a deep generative model that can accurately capture complex subsurface structure is tested with an ensemble-based inversion method for accurate and accelerated characterization of CO 2 storage sites. We chose Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) for its realistic reservoir property representation and Ensemble Smoother with Multiple Data Assimilation (ES-MDA) for its robust data fitting and uncertainty quantification capability. WGAN-GP are trained to generate high-dimensional permeability fields from a low-dimensional latent space and ES-MDA then updates the latent variables by assimilating available measurements. Several subsurface site characterization examples including Gaussian, channelized, and fractured reservoirs are used to evaluate the accuracy and computational efficiency of the proposed method and the main features of the unknown permeability fields are characterized accurately with reliable uncertainty quantification. Furthermore, the estimation performance is compared with a widely-used variational, i.e., optimization-based, inversion approach, and the proposed approach outperforms the variational inversion method in several benchmark cases. We explain such superior performance by visualizing the objective function in the latent space: because of nonlinear and aggressive dimension reduction via generative modeling, the objective function surface becomes extremely complex while the ensemble approximation can smooth out the multi-modal surface during the minimization. This suggests that the ensemble-based approach works well over the variational approach when combined with deep generative models at the cost of forward model runs unless convergence-ensuring modifications are implemented in the variational inversion.

42 ENGINEERING↗