Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “fixed point method”

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 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Reduced-order modeling on a near-term quantum computer

Quantum computing is an advancing area of research in which computer hardware and algorithms are developed to take advantage of quantum mechanical phenomena. In recent studies, quantum algorithms have shown promise in solving linear systems of equations as well as systems of linear ordinary differential equations (ODEs) and partial differential equations (PDEs). Reducedorder modeling (ROM) algorithms for studying fluid dynamics have shown success in identifying linear operators that can describe flowfields, where dynamic mode decomposition (DMD) is a particularly useful method in which a linear operator is identified from data. In this work, DMD is reformulated as an optimization problem to propagate the state of the linearized dynamical system on a quantum computer. This reformulation was chosen as a means of facilitating implementation on a near-term quantum computer. Quadratic unconstrained binary optimization (QUBO), a technique for optimizing quadratic polynomials in binary variables, allows for quantum annealing algorithms to be applied. A quantum circuit model (quantum approximation optimization algorithm, QAOA) is utilized to obtain predictions of the state trajectories. Results are shown for the quantum-ROM predictions for flow over a 2D cylinder at Re = 220 and flow over a NACA0009 airfoil at Re = 500 and α = 15°. The quantum-ROM predictions are found to depend on the number of bits utilized for a fixed point representation and the truncation level of the DMD model. Comparisons with DMD predictions from a classical computer algorithm are made, as well as an analysis of the computational complexity and prospects for future, more fault-tolerant quantum computers.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Clustering of red sequence galaxies in the fourth data release of the Kilo-Degree Survey

We present a sample of luminous red sequence galaxies as the basis for a study of the large-scale structure in the fourth data release of the Kilo-Degree Survey. The selected galaxies are defined by a red sequence template, in the form of a data-driven model of the colour-magnitude relation conditioned on redshift. In this work, the red sequence template was built using the broad-band optical+near infrared photometry of KiDS-VIKING and the overlapping spectroscopic data sets. The selection process involved estimating the red sequence redshifts, assessing the purity of the sample and estimating the underlying redshift distributions of redshift bins. After performing the selection, we mitigated the impact of survey properties on the observed number density of galaxies by assigning photometric weights to the galaxies. We measured the angular two-point correlation function of the red galaxies in four redshift bins and constrain the large-scale bias of our red sequence sample assuming a fixed ΛCDM cosmology. We find consistent linear biases for two luminosity-threshold samples (‘dense’ and ‘luminous’). We find that our constraints are well characterised by the passive evolution model.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Doppler Lidar Motion-Correction (DLMC) Value-Added Product Report

The U.S. Department of Energy Atmospheric Radiation Measurement (ARM) second ARM Mobile Facility (AMF2) Doppler lidar (S/N 0319-160) was deployed on the German ice breaker Polarstern during the Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC) campaign during 2019-2020 (see Figure 1). This was the first deployment of the new AMF2 Doppler lidar, as well as the first ship-based deployment of a Doppler lidar (DL) by ARM. In contrast to land-based deployments, the lidar’s heading (a.k.a. home point) and tilt are constantly changing, and lidar’s radial velocity measurements are impacted by the ship’s motion. Since the beam directions are reported relative to the instrument’s frame of reference, derivation of higher-order data products such as wind speed and direction require that the lidar’s attitude and translational velocity be properly accounted for. The Doppler Lidar Motion-Correction (DLMC) Value-Added-Product (VAP) was developed specifically for ship-based deployments of the Doppler lidar. This VAP combines raw uncorrected data from the DL and simultaneous measurements from the ARM Navigation system (NAV) (Walton 2019) to transform the beam angles from the lidar coordinate system to an Earth-fixed coordinate system. The VAP also removes the contribution of the lidar’s platform velocity from the radial velocity measurements. This report documents the methods used by the DLMC to perform these corrections.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

symfind : Addressing the Fragility of Subhalo Finders and Revealing the Durability of Subhalos

Abstract A major question in ΛCDM is what this theory actually predicts for the properties of subhalo populations. Subhalos are difficult to accurately simulate and to find within simulations, and this propagates into uncertainty in theoretical predictions for satellite galaxies. We present Symfind , a new particle-tracking-based subhalo finder, and demonstrate that it can track subhalos to orders-of-magnitude lower masses than commonly used halo-finding tools, with a focus on Rockstar and consistent-trees . These longer survival times mean that at a fixed peak subhalo mass, we find ≈ 15%–40% more subhalos within the virial radius, R vir , and ≈35%–120% more subhalos within R vir /4 in the Symphony dark-matter-only simulation suite. More subhalos are found as the resolution is increased, in contrast to the Rockstar halo finder, which appears to be converged at smaller subhalo counts. We perform extensive numerical testing. In agreement with idealized simulations, we show that the v max , the maximum circular velocity, is systematically biased low until high resolutions ( n peak ≳ 3 × 10 4 ) are achieved, but that mass loss itself can be resolved at much more modest resolutions ( n peak ≳ 4 × 10 3 ). We show that Rockstar converges to false solutions for the mass function, radial distribution, and disruption masses of subhalos. We argue that our new method can trace resolved subhalos until the point of typical galaxy disruption without invoking post hoc orphan modeling. We outline a concrete set of steps for determining whether other subhalo finders meet the same criteria. We publicly release Symfind catalogs and particle data for the Symphony simulation suite at http://web.stanford.edu/group/gfc/symphony .

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Equivalent Fundamental-Mode Source Simulations for Spherical Uranium and Plutonium Systems [Abstract]

The equivalent fundamental-mode source refers to a “source” which is identically distributed in space, energy, and angle as that of a fundamental-mode fission source distribution. It has previously been shown that the system fixed source multiplication ($M_{fs}$) relates to the effective multiplication factor ($\kappa_{eff}$) as $M_{fs} = \frac{g^*}{1-k_{eff}}$ The term $\mathcal{g}$* therefore relates the system source distribution to the system equivalent fundamental-mode source. As previously shown, the $\mathcal{g}$* term is approximately 1 for deeply subcritical systems and then diverges as the system multiplication increases. For a system with a point source located in the center of an assembly (such a sphere), $\mathcal{g}$* will be greater than 1; this happens because there is a higher probability of induced fission when all of the starter neutrons originate from the center of the sphere. For a uniform source, $\mathcal{g}$* is less than 1, but does not diverge from 1 as dramatically as seen by a point source. One reason that $\mathcal{g}$* is an important parameter is that it is used in measurement methods to infer reactor kinetics parameters ($β_{eff}$ being one example). This work will discuss multiple ways to simulate $\mathcal{g}$* using MCNP®6.2, which include use of traditional methods versus a single input file method that was recently published. A discussion related to methods to simulate $M_{fs}$ will be presented. Spherical systems of Highly Enriched Uranium (HEU) will be compared with previous works. In addition, results of plutonium including multiple reflector materials will be investigated. Simulated systems will include both parametric studies as well as configurations used in recent experiments performed at the National Criticality Experiments Research Center (NCERC). These studies will include subtleties associated with system geometry that have not been previously shown. Last, this work will investigate the relationship between $\mathcal{g}$* and reactor kinetics parameters (such as $β_{eff}$).

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2024

IDAES-PSE 2.6.0 Release

The Institute for the Design of Advanced Energy Systems (IDAES) Integrated Platform is a versatile computational environment offering extensive process systems engineering (PSE) capabilities for optimizing the design and operation of complex, interacting technologies and systems. IDAES enables users to efficiently search vast, complex design spaces to discover the lowest cost solutions while supporting the full process modeling lifecycle, from conceptual design to dynamic optimization and control. The extensible, open platform empowers users to create models of novel processes and rapidly develop custom analyses, workflows, and end-user applications. IDAES-PSE 2.6.0 Release Highlights Upcoming Changes IDAES will be switching to the new Pyomo solver interface in the next release. Whilst this will hopefully be a smooth transition for most users, there are a few important changes to be aware of. The new solver interface uses a different version of the IPOPT writer (“ipopt_v2”) and thus any custom configuration options you might have set for IPOPT will not carry over and will need to be reset. By default, the new Pyomo linear presolver will be activated with ipopt_v2. Whilst are working to identify any bugs in the presolver, it is possible that some edge cases will remain. IDAES will begin deploying a new set of scaling tools and APIs over the next few releases that make use of the new solver writers. The old scaling tools and APIs will remain for backward compatibility but will begin to be deprecated. New Models, Tools and Features New Intersphinx extension automatically linking Jupyter notebook examples to project documentation New end-to-end diagnostics example demonstrated on a real problem New complementarity formulation for VLE with cubic equations of state, backward compatibility for old formulation New solver interface with presolve (ipopt_v2) in support of upcoming changes to the initialization and APIs methods, with default set to ipopt to maintain backwards compatibility; this will deprecate once all examples have been updated New forecaster and parameterized bidder methods within grid integration library Updated surrogates API and examples to support Keras 3, with backwards compatibility for older formats such as TensorFlow SavedModel (TFSM) Updated costing base dictionary to include the 2023 cost year index value Updated ProcessBlock to include information on the constructing block class Updated Flowsheet Visualizer to allow visualize() method to return value and functions Bug Fixes Fixed bug in the Modular Property Framework that would cause errors when trying to use phase-based material balances with phase equilibria. Fixed bug in Modular Properties Framework that caused errors when initializing models with non-vapor-liquid phase equilibria. Fixed typos flagged by June update to crate-ci/typos and removed DMF-related exceptions Minor corrections of units of measurement handling in power plant waste/transport costing expressions, control volume material holdup expressions, and BTX property package parameters Fixed throwing >7500 numpy deprecation warnings by replacing scalar value assignment with element extraction and item iteration calls Testing and Robustness Migrated slow tests (>10s) to integration, impacting test coverage but also yielding a nearly 30% decrease in local test runtime Pinned pint to avoid issues with older supported Python versions Pinned codecov versions to avoid tokenless upload behavior with latest version Bumped extensions to version 3.4.2 to allow pointing to non-standard install location Deprecations and Removals Python 3.8 is no longer supported. The supported Python versions are 3.9 through 3.12 The Data Management Framework (DMF) is no longer supported. Importing idaes.core.dmf will cause a deprecation warning to be displayed until the next release The SOFC Keras surrogates have been removed. The current version of the SOFC surrogate model in the examples repository is a PySMO Kriging model.

AS↗
DOE OSTI2023

Quantum mechanics based non-bonded force field functions for use in molecular dynamics simulations of materials and systems: The nitrogen and oxygen columns

Accurate Force Fields (FFs) are essential for Molecular Dynamics (MD) simulations of the dynamics of realistic materials in terms of atomic-level interactions. The FF parameters of short-range valence interactions can be derived through Quantum Mechanical (QM) calculations on model systems practical for QM (<300 atoms). Similarly, the dynamic electrostatic interactions can be described with methods such as QEq or PQEq that allow charges and polarization to adjust dynamically. However, accurately extracting long-range van der Waals (vdW) interactions from QM calculations poses challenges due to the absence of a definitive method to distinguish between the different energetic components of electrostatics, polarization, vdW, hydrogen bonding, and valence interactions. To do this we use the Perdew–Burke–Ernzerhof flavor of Density Functional Theory, including empirical D3 vdW corrections, to predict the Equation of State for each element (keeping any covalent bonds fixed), from which we obtain the two-body vdW nonbond potential. Here, we extend these calculations to include non-bonded parameters for the N and O columns of the periodic table so that we now describe columns 15 (N), 16 (O), 17 (F), and 18 (Ne) of the periodic table. For these 20 elements, we find that the two-body vdW potentials can all be mapped to a single universal two-body curve, with just three scaling parameters: Re, De, and L. We refer to this as the Universal NonBond (UNB) potential. We expect this to be useful for new MD simulations and a helpful starting point to obtain UNB parameters for the remainder of the periodic table.

Chemistry↗
DOE OSTI2023

Airfoil Computational Fluid Dynamics - 9k shapes, 2 AoA's

This dataset contains aerodynamic quantities - including flow field values (momentum, energy, and vorticity) and summary values (coefficients of lift, drag, and momentum) - for 8,996 airfoil shapes, computed using the HAM2D CFD (computational fluid dynamics) model. The airfoil shapes were designed using the separable shape tensor parameterization that encodes two-dimensional shapes as elements of the Grassmann manifold. This data-driven approach learns two independent spaces of parameter from a collection of sample airfoils. The first captures large-scale, linear perturbations, and the second defines small-scale, higher-order perturbations. For this data, we used the G2Aero database of over 19,000 airfoil shapes to learn a parameter space that captured a wide array of shape characteristics. We fixed the linear deformations to be the mean over the database and sampled new shapes over a four-dimensional parameter space of higher-order perturbation. This sampling approaches allows for isolated analysis of non-linear airfoil shape deformations while holding other aspects (e.g., airfoil thickness) approximately constant. The aerodynamic quantities for the generated airfoil were obtained using the HAM2D code, which is a finite-volume Reynolds-averaged Navier-Stokes (RANS) flow solver. We employ a fifth-order WENO scheme for spatial reconstruction with Roe's flux difference scheme for inviscid flux and second-order central differencing for viscous flux. A preconditioned GMRES method is applied for implicit integration. The Spalart-Allmaras 1-eq turbulence model is used for the turbulence closure, and the Medida-Baeder 2-eq transition model is applied to account for the effects of laminar turbulent transition. The airfoil grid is generated with a total of 400 points on the airfoil surface, the initial wall-normal spacing of y+ = 1, and an outer boundary located at 300 chord lengths away from the wall. The CFD simulations are performed at a freestream Mach number of 0.1, Reynolds number of 9M, and at two angles of attack, 4 deg. and 12 deg. The simulations were performed using the Bridges-2 system at the Pittsburgh Supercomputing Center in February 2023 as part of the INTEGRATE project funded by the Advanced Research Projects Agency - Energy in the U.S. Department of Energy. The data was collected, reformatted, and preprocessed for this OEDI submission in July 2023 under the Foundational AI for Wind Energy project funded by the U.S. Department of Energy Wind Energy Technologies Office. This dataset is intended to serve as a benchmark against which new artificial intelligence (AI) or machine learning (ML) tools may be tested. Baseline AI/ML methods for analyzing this dataset have been implemented, and a link to their repository containing those models has been provided. The .h5 data file structure can be found in the GitHub Repository resource under explore_airfoil_9k_data.ipynb.

9k↗
DOE OSTI2024

Defining Detection Limits for Continuous Monitoring Systems for Methane Emissions at Oil and Gas Facilities

Networks of fixed-point continuous monitoring systems are becoming widely used in the detection and quantification of methane emissions from oil and gas facilities in the United States. Regulatory agencies and operators are developing performance metrics for these systems, such as minimum detection limits. Performance characteristics, such as minimum detection limits, would ideally be expressed in emission rate units; however, performance parameters such as detection limits for a continuous monitoring system (CMS) will depend on meteorological conditions, the characteristics of emissions at the site where the CMS is deployed, the positioning of CMS devices in relation to the emission sources, and the amount of time allowed for the CMS to detect an emission source. This means that certifying the performance of a CMS will require test protocols with well-defined emission rates and durations; initial protocols are now being used in field tests. Field testing results will vary, however, depending on meteorological conditions and the time allowed for detection. This work demonstrates methods for evaluating CMS performance characteristics using dispersion modeling and defines an approach for normalizing test results to standard meteorological conditions using dispersion modeling.

Chen, Qining (ORCID:0000000316908091)↗
DOE OSTI2021

Ocean Energy Systems Wave Energy Modeling Task 10.4: Numerical Modeling of a Fixed Oscillating Water Column

This paper reports on an ongoing international effort to establish guidelines for numerical modeling of wave energy converters, initiated by the International Energy Agency Technology Collaboration Program for Ocean Energy Systems. Initial results for point absorbers were presented in previous work, and here we present results for a breakwater-mounted Oscillating Water Column (OWC) device. The experimental model is at scale 1:4 relative to a full-scale installation in a water depth of 12.8 m. The power-extracting air turbine is modeled by an orifice plate of 1–2% of the internal chamber surface area. Measurements of chamber surface elevation, air flow through the orifice, and pressure difference across the orifice are compared with numerical calculations using both weakly-nonlinear potential flow theory and computational fluid dynamics. Both compressible- and incompressible-flow models are considered, and the effects of air compressibility are found to have a significant influence on the motion of the internal chamber surface. Recommendations are made for reducing uncertainties in future experimental campaigns, which are critical to enable firm conclusions to be drawn about the relative accuracy of the numerical models. It is well-known that boundary element method solutions of the linear potential flow problem (e.g., WAMIT) are singular at infinite frequency when panels are placed directly on the free surface. This is problematic for time-domain solutions where the value of the added mass matrix at infinite frequency is critical, especially for OWC chambers, which are modeled by zero-mass elements on the free surface. A straightforward rational procedure is described to replace ad-hoc solutions to this problem that have been proposed in the literature.

16 TIDAL AND WAVE POWER↗
DOE OSTI2022

Background-Aware 3-D Point Cloud Segmentation With Dynamic Point Feature Aggregation

With the proliferation of LiDAR sensors and 3-D vision cameras, 3-D point cloud analysis has attracted significant attention in recent years. In this article, we propose a novel 3-D point cloud learning network, referred to as dynamic point feature aggregation network (DPFA-Net), by selectively performing the neighborhood feature aggregation (FA) with dynamic pooling and an attention mechanism. DPFA-Net has two variants for semantic segmentation and classification of 3-D point clouds. As the core module of the DPFA-Net, we propose an FA layer, in which features of the dynamic neighborhood of each point are aggregated via a self-attention mechanism. In contrast to other segmentation models, which aggregate features from fixed neighborhoods, our approach can aggregate features from different neighbors in different layers providing a more selective and broader view to the query points and focusing more on the relevant features in a local neighborhood. In addition, to further improve the performance of semantic segmentation, we exploit the background–foreground (BF) information and present two novel approaches, namely, two-stage BF-Net and BF regularization. Experimental results show that the proposed DPFA-Net achieves the state-of-the-art overall accuracy score of 89.22% for semantic segmentation on the Stanford large-scale 3-D Indoor Spaces (S3DIS) dataset and provides consistently satisfactory performance across different tasks of semantic segmentation, part segmentation, and 3-D object classification. Furthermore, our model achieves 93.1% accuracy on the ModelNet40 dataset and provides a mean shape intersection-over-union (IoU) value of 85.5% for part segmentation on the ShapeNet-Part dataset. It is a also computationally more efficient compared to other methods.

3-D↗
DOE OSTI2022

Compressed basis GMRES on high-performance graphics processing units

Krylov methods provide a fast and highly parallel numerical tool for the iterative solution of many large-scale sparse linear systems. To a large extent, the performance of practical realizations of these methods is constrained by the communication bandwidth in current computer architectures, motivating the investigation of sophisticated techniques to avoid, reduce, and/or hide the message-passing costs (in distributed platforms) and the memory accesses (in all architectures). This article leverages Ginkgo’s memory accessor in order to integrate a communication-reduction strategy into the (Krylov) GMRES solver that decouples the storage format (i.e., the data representation in memory) of the orthogonal basis from the arithmetic precision that is employed during the operations with that basis. Given that the execution time of the GMRES solver is largely determined by the memory accesses, the cost of the datatype transforms can be mostly hidden, resulting in the acceleration of the iterative step via a decrease in the volume of bits being retrieved from memory. Together with the special properties of the orthonormal basis (whose elements are all bounded by 1), this paves the road toward the aggressive customization of the storage format, which includes some floating-point as well as fixed-point formats with mild impact on the convergence of the iterative process. We develop a high-performance implementation of the “compressed basis GMRES” solver in the Ginkgo sparse linear algebra library using a large set of test problems from the SuiteSparse Matrix Collection. We demonstrate robustness and performance advantages on a modern NVIDIA V100 graphics processing unit (GPU) of up to 50% over the standard GMRES solver that stores all data in IEEE double-precision.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Formation and detection of metastable formic acid in a supersonic expansion: High resolution infrared spectroscopy of the jet-cooled cis -HCOOH conformer

High-resolution direct absorption infrared spectra of metastable cis-formic acid (HCOOH) trapped in a cis-well resonance behind a 15 kcal/mol barrier are reported for the first time, with the energetically unstable conformer produced in a supersonic slit plasma expansion of trans-formic acid/H 2 mixtures. We present a detailed high-resolution rovibrational analysis for cis-formic acid species in the OH stretch (ν 1 ) fundamental, providing first precision vibrational band origin, rotational constants, and term values, which in conjunction with ab initio calculations at the couple-cluster with single, double, and perturbative triple [CCSD(T)]/ANOn (n = 0, 1, 2) level support the experimental assignments and establish critical points on the potential energy surface for internal rotor trans-to-cis isomerization. Relative intensities for a- and b-type transitions observed in the spectra permit the transition dipole moment components to be determined in the body fixed frame and prove to be in good agreement with ab initio CCSD(T) theoretical estimates but in poor agreement with simple bond-dipole predictions. As a result, the observed signal dependence on H 2 in the discharge suggests the presence of a novel H atom radical chemical mechanism for strongly endothermic “up-hill” internal rotor isomerization between trans- and cis-formic acid conformers.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Extremely Overdoped Superconducting Cuprates via High Pressure Oxygenation Methods

Within the cuprate constellation, one fixed star has been the superconducting dome in the quantum phase diagram of transition temperature vs. the excess charge on the Cu in the CuO2-planes, p, resulting from O-doping or cation substitution. However, a more extensive search of the literature shows that the loss of the superconductivity in favor of a normal Fermi liquid on the overdoped side should not be assumed. Many experimental results from cuprates prepared by high-pressure oxygenation show Tc converging to a fixed value or continuing to slowly increase past the upper limit of the dome of p = 0.26–0.27, up to the maximum amounts of excess oxygen corresponding to p values of 0.3 to > 0.6. These reports have been met with disinterest or disregard. Our review shows that dome-breaking trends for Tc are, in fact, the result of careful, accurate experimental work on a large number of compounds. This behavior most likely mandates a revision of the theoretical basis for high-temperature superconductivity. That excess O atoms located in specific, metastable sites in the crystal, attainable only with extreme O chemical activity under HPO conditions, cause such a radical extension of the superconductivity points to a much more substantial role for the lattice in terms of internal chemistry and bonding.

Sederholm, Linda (ORCID:0000000172090509)↗
DOE OSTI2021

Method for boosting dispersive spectrograph stability 1000× using interferometry with crossfaded pairs of delays

We demonstrate a key step along a technical route to achieving cm/s scale accuracy for astronomical spectrographs over long (multi-year) time scales, which is critical for the Doppler characterization of Earth sized exoplanets, and measurement of small cosmic redshift drift over many years. This same technique also enables searching exoplanet atmospheres for biosignificant molecules in direct planet imaging using, otherwise, insufficiently low resolution and drift prone dispersive (grating or prism) spectrographs. Using a method called crossfading for externally dispersed interferometers (EDIs) to get highly robust spectra, we recently demonstrated a factor of 1000 x reduction in the net shift of an EDI measured ThAr line to a deliberate simulated wavelength translation of the detector. This 1000 x gain in disperser stability can be combined with conventional stability gains afforded by fiber scramblers, vacuum tanks, and thermal control, to provide an additional 1 to 3 orders of magnitude reduction in the net point spread function shift drift. Crossfading combines high- and low-delay fringing signals that react oppositely in phase to cancel their net reaction to a detector wavelength drift. This can be implemented by an interferometer addition to a facility spectrograph.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Methane, ethane, and propane production in Greenland ice core samples and a first isotopic characterization of excess methane

Air trapped in polar ice provides unique records of the past atmospheric composition ranging from key greenhouse gases such as methane (CH 4 ) to short-lived trace gases like ethane (C 2 H 6 ) and propane (C 3 H 8 ). Recently, the comparison of CH 4 records obtained using different extraction methods revealed disagreements in the CH 4 concentration for the last glacial in Greenland ice. Elevated methane levels were detected in dust-rich ice core sections measured discretely, pointing to a process sensitive to the melt extraction technique. To shed light on the underlying mechanism, we performed targeted experiments and analyzed samples for methane and the short-chain alkanes ethane and propane covering the time interval from 12 to 42 kyr. Here, we report our findings of these elevated alkane concentrations, which scale linearly with the amount of mineral dust within the ice samples. The alkane production happens during the melt extraction step of the classic wet-extraction technique and reaches 14 to 91 ppb of CH 4 excess in dusty ice samples. We document for the first time a co-production of excess methane, ethane, and propane, with the observed concentrations for ethane and propane exceeding their past atmospheric background at least by a factor of 10. Independent of the produced amounts, excess alkanes were produced in a fixed molar ratio of approximately 14 : 2 : 1, indicating a shared origin. The measured carbon isotopic signature of excess methane is (-47.0 ± 2.9) ‰ and its deuterium isotopic signature is (-326 ± 57) ‰. With the co-production ratios of excess alkanes and the isotopic composition of excess methane we established a fingerprint that allows us to constrain potential formation processes. This fingerprint is not in line with a microbial origin. Moreover, an adsorption–desorption process of thermogenic gas on dust particles transported to Greenland does not appear very likely. Instead, the alkane pattern appears to be indicative of abiotic decomposition of organic matter as found in soils and plant leaves.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

Architecture-Preserving Provable Repair of Deep Neural Networks

Deep neural networks (DNNs) are becoming increasingly important components of software, and are considered the state-of-the-art solution for a number of problems, such as image recognition. However, DNNs are far from infallible, and incorrect behavior of DNNs can have disastrous real-world consequences. This paper addresses the problem of architecture-preserving V-polytope provable repair of DNNs. A V-polytope defines a convex bounded polytope using its vertex representation. V-polytope provable repair guarantees that the repaired DNN satisfies the given specification on the infinite set of points in the given V-polytope. An architecture-preserving repair only modifies the parameters of the DNN, without modifying its architecture. The repair has the flexibility to modify multiple layers of the DNN, and runs in polynomial time. It supports DNNs with activation functions that have some linear pieces, as well as fully-connected, convolutional, pooling and residual layers. To the best our knowledge, this is the first provable repair approach that has all of these features. We implement our approach in a tool called APRNN. Using MNIST, ImageNet, and ACAS Xu DNNs, we show that it has better efficiency, scalability, and generalization compared to PRDNN and REASSURE, prior provable repair methods that are not architecture preserving.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Projective Integral Updates for High-Dimensional Variational Inference

Variational inference is an approximation framework for Bayesian inference that seeks to improve quantified uncertainty in predictions by optimizing a simplified distribution over parameters to stand in for the full posterior. Capturing model variations that remain consistent with training data enables more robust predictions by reducing parameter sensitivity. This work introduces a fixed-point optimization for variational inference that is applicable when every feasible log density can be expressed as a linear combination of functions from a given basis. In such cases, the optimizer becomes a fixed-point of projective integral updates. When the basis spans univariate quadratics in each parameter, the feasible distributions are Gaussian mean-fields and the projective integral updates yield quasi-Newton variational Bayes (QNVB). Other bases and updates are also possible. Since these updates require high-dimensional integration, this work begins by proposing an efficient quasirandom sequence of quadratures for mean-field distributions. Each iterate of the sequence contains two evaluation points that combine to correctly integrate all univariate quadratic functions and, if the mean-field factors are symmetric, all univariate cubics. More importantly, averaging results over short subsequences achieves periodic exactness on a much larger space of multivariate polynomials of quadratic total degree. The corresponding variational updates require four loss evaluations with standard (not second-order) backpropagation to eliminate error terms from over half of all multivariate quadratic basis functions. Furthermore, this integration technique is motivated by first proposing stochastic blocked mean-field quadratures, which may be useful in other contexts. A PyTorch implementation of QNVB allows for better control over model uncertainty during training than competing methods. Experiments demonstrate superior generalizability for multiple learning problems and architectures.

Gaussian mean-field↗