Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Variational approach”

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 379 records · Page 21
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Real-time dynamics of the Schwinger model as an open quantum system with Neural Density Operators

Ab-initio simulations of multiple heavy quarks propagating in a Quark-Gluon Plasma are computationally difficult to perform due to the large dimension of the space of density matrices. This work develops machine learning algorithms to overcome this difficulty by approximating exact quantum states with neural network parametrisations, specifically Neural Density Operators. As a proof of principle demonstration in a QCD-like theory, the approach is applied to solve the Lindblad master equation in the 1 + 1d lattice Schwinger Model as an open quantum system. Neural Density Operators enable the study of in-medium dynamics on large lattice volumes, where multiple-string interactions and their effects on string-breaking and recombination phenomena can be studied. Thermal properties of the system at equilibrium can also be probed with these methods by variationally constructing the steady state of the Lindblad master equation. Scaling of this approach with system size is studied, and numerical demonstrations on up to 32 spatial lattice sites and with up to 3 interacting strings are performed.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2021

Standardizing Performance Metrics for Building-Level Electrical Distribution Systems

Building-level electrical distribution systems comprise a myriad of current-carrying equipment, conversion devices, and protection devices that deliver power from the utility or local distributed energy resources to end-use building loads. Electric power has traditionally been generated, transmitted, and distributed in alternating current (AC). However, the last decade has seen a significant increase in the integration of native direct current (DC) equipment that has elevated the importance of DC distribution systems. Numerous studies have comparatively examined the performance of various electrical distribution systems in buildings but have failed to achieve uniform conclusions, primarily because of a lack of consistent and analogous performance evaluation methods. This paper aims to fill this gap by providing a standard set of metrics and measurement boundaries to consistently evaluate the performance of AC, DC, or hybrid AC/DC electrical distribution systems. The efficacy of the proposed approach is evaluated on a representative medium-sized commercial office building model with AC distribution and an equivalent hybrid AC/DC and DC distribution model, wherein the AC distribution model is concluded to be the most efficient. The simulation results show variation in computed metrics with different selected boundaries that verify the effectiveness of the proposed approach in ensuring consistent computation of the performance of building-level electrical distribution systems. This paper provides an initial set of guidelines for building energy system stakeholders to adopt appropriate solutions, thus leading to more efficient energy systems.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2020

Integrating Predictions for Improving Defect Classification Accuracy in NDT-based Assessment of Concrete - 20229

There is an increasing need to create predictive models for defect classification in concrete using the output of non-destructive testing (NDT) techniques. Recent advancement of machine-learning algorithms has offered several techniques for developing classification models for different types of data sets. However, the performance of these algorithms is very uncertain, mainly when applied to small and noisy datasets. For example, when human access is limited (e.g., nuclear facility), robot-based NDT is preferred. But compared to manual tests with humans present on site, the data sets are small, and more noise can exist. Therefore, it is imperative to develop new approaches to ensure a consistently high classification accuracy for inadequate data sets. This study explores the classification performance on NDT dataset using classifiers from different machine-learning algorithms, namely k-Nearest Neighbor (kNN), Decision Tree, Naive Bayes, Logistic Regression, and Support Vector Machine (SVM). The authors further integrated the predictions from these classifiers using proposed methods. The integration strategy combines the output of the classifiers based on two different measures, accuracy, and performance (ACC and PERF), using equations such as sum, average, and square-root-of-sums-of-squares (SRSS). Our results reveal varying classification accuracies across individual classifiers with different misclassifications across the test data set. The integration strategy provided significant improvement in the classification accuracy compared to the individual classifiers. Furthermore, the results indicate minimal variation across the integration methods as compared to the variation across the individual classifiers. To conclude, prediction integration offers a unique approach for combining the output of multiple classifiers to create redundancies with the potential of achieving high classification performance and improved reliability in predictive models for defect detection in concrete. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2025

RLGBS: Reinforcement Learning-Guided Beam Search for process optimization in a paper machine dryer section

Paper drying is responsible for over two-thirds of energy consumption in the U.S. pulp and paper industry, presenting significant potential for energy savings through optimization of process parameters. Current approaches often assume fixed operating conditions, neglecting dynamic ambient and process variations that limit achievable savings and real-world applicability. To this end, we develop a physics-based simulation environment for a paper machine dryer section and propose a reinforcement learning (RL) framework to minimize overall energy consumption by optimizing drying process parameters under diverse operating conditions. To mitigate overdrying and numerical instabilities caused by suboptimal local RL actions, we introduce Reinforcement Learning-Guided Beam Search (RLGBS), which explores multiple action sequences in parallel using beam search. Instead of making step-by-step decisions, RLGBS prioritizes solutions based on cumulative probability, reducing the impact of individual suboptimal actions. Experiments demonstrate that RLGBS achieves consistent energy savings under unseen operating conditions not encountered during training, outperforming conventional RL methods. While validated in drying optimization, this framework is broadly applicable to other RL-based industrial process control problems.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2021

A weighted Shifted Boundary Method for free surface flow problems

The Shifted Boundary Method (SBM) belongs to the class of unfitted (or immersed, or embedded) finite element methods and was recently introduced for the Poisson, linear advection/diffusion, Stokes, Navier-Stokes, acoustics, and shallow-water equations. By reformulating the original boundary value problem over a surrogate (approximate) computational domain, the SBM avoids integration over cut cells and the associated problematic issues regarding numerical stability and matrix conditioning. Accuracy is maintained by modifying the original boundary conditions using Taylor expansions. Hence the name of the method, that shifts the location and values of the boundary conditions. In this article, we extend the SBM to the simulation of incompressible Navier-Stokes flows with moving free-surfaces, by appropriately weighting its variational form with the elemental volume fraction of active fluid. This approach prevents spurious pressure oscillations in time, which would otherwise be produced if the total active fluid volume were to change abruptly over a time step. In fact, the proposed weighted SBM method induces small mass (i.e., volume) conservation errors, which converge quadratically in the case of piecewise-linear finite element interpolations, as the grid is refined. Finally, we present an extensive set of two- and three-dimensional tests to demonstrate the robustness and accuracy of the method.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2024

A weighted shifted boundary method for immersed moving boundary simulations of Stokes' flow

The Shifted Boundary Method (SBM) belongs to the class of unfitted (or immersed, or embedded) finite element methods, and relies on reformulating the original boundary value problem over a surrogate (approximate) computational domain. The surrogate domain is constructed so as to avoid cut cells and the associated problematic implementation and numerical integration issues. Accuracy is maintained by modifying the original boundary conditions using Taylor expansions: hence the name of the method, that shifts the location and values of the boundary conditions. Here, in this article, we extend the SBM to the simulation of incompressible Stokes flow, by appropriately weighting its variational form with the elemental volume fraction of active fluid. This approach allows to drastically reduce spurious pressure oscillations in time, which are produced if the total volume of active fluid were to change abruptly over a time step. The proposed Weighted SBM (W-SBM) exactly preserves states of hydrostatic equilibrium, and induces small mass and momentum conservation errors, which converge as the grid is refined. This is in analogy to cutFEMs and related unfitted approaches, which rely on an affine representation of cut boundaries. We demonstrate the robustness and accuracy of the proposed method with an extensive suite of two-dimensional tests.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Ensemble variational Fokker-Planck methods for data assimilation

Particle flow filters solve Bayesian inference problems by smoothly transforming a set of particles into samples from the posterior distribution. Particles move in state space under the flow of an McKean-Vlasov-Itˆo process. This work introduces the Variational Fokker-Planck (VFP) framework for data assimilation, a general approach that includes previously known particle flow filters as special cases. The McKean-Vlasov-Itˆo process that transforms particles is defined via an optimal drift that depends on the selected diffusion term. It is established that the underlying probability density - sampled by the ensemble of particles - converges to the Bayesian posterior probability density. For a finite number of particles the optimal drift contains a regularization term that nudges particles toward becoming independent random variables. Based on this analysis, we derive computationally-feasible approximate regularization approaches that penalize the mutual information between pairs of particles, and avoid particle collapse. Moreover, the diffusion plays a role akin to a particle rejuvenation approach that aims to alleviate particle collapse. The VFP framework is very flexible. Different assumptions on prior and intermediate probability distributions can be used to implement the optimal drift, and localization and covariance shrinkage can be applied to alleviate the curse of dimensionality. A robust implicit-explicit method is discussed for the efficient integration of stiff McKean- Vlasov-Itˆo processes. Here, the effectiveness of the VFP framework is demonstrated on three progressively more challenging test problems, namely the Lorenz ’63, Lorenz ’96 and the quasi-geostrophic equations.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

A combined experimental and crystal plasticity study of grain size effects in magnesium alloys

This work presents a method to incorporate the micro Hall-Petch equation into the crystal plasticity finite element (CPFE) framework accounting for the microstructural features to understand the coupling between grain size, texture, and loading direction in magnesium alloys. The effect of grain size and texture is accounted for by modifying the slip resistances of individual basal and prismatic systems based on the micro Hall-Petch equation. The modification based on the micro Hall-Petch equation endows every slip system at each microstructural point with a slip system-level grain size and maximum compatibility factor, which are in turn used to modify the slip resistance. While the slip-system level grain size is a measure of the grain size, the maximum compatibility factor encodes the effect of the grain boundary on the slip system resistance modification and is computed based on the Luster-Morris factor. The model is calibrated using experimental stress-strain curves of Mg-4Al samples with three different grain sizes from which the Hall-Petch coefficients are extracted and compared with Hall-Petch coefficients predicted using original parameters from previous work. The predictability of the model is then evaluated for a Mg-4Al sample with different texture and three grain sizes subjected to loading in different directions. The calibrated parameters are then used for some parametric studies to investigate the variation of Hall-Petch slope for different degrees of simulated spread in basal texture, variation of Hall-Petch slope with loading direction relative to basal poles for a microstructure with strong basal texture, and variation of yield strength with change in grain morphology. The proposed approach to incorporate the micro Hall-Petch equation into the CPFE framework provides a foundation to quantitatively model more complicated scenarios of coupling between grain size, texture and loading direction in the plasticity of Mg alloys.

Crystal plasticity↗
DOE OSTI2023

Mechanisms Controlling the Energy Barrier for Ion Hopping in Polymer Electrolytes

Here, the present work studies the mechanisms controlling the energy barrier for ion hopping in conducting polymers. Polymer electrolytes usually show Arrhenius-like temperature dependence of the conductivity relaxation time (characteristic time of local ion rearrangements) at temperatures below their glass transition T g . However, our analysis reveals that the Arrhenius fit of this regime leads to unphysically small prefactors, τ 0 $\ll$ 10 –13 s. Imposing a value of 10 –13 s for this parameter renders the fairly unexpected result that the energy barrier for charge transport in these polymers has strong temperature dependence even below T g . Our study also reveals significant temperature variations of the dielectric permittivity and the instantaneous shear modulus in the glassy state of these polymers. Using the Anderson and Stuart model, we demonstrate that these variations provide strong justifications for the temperature variation of energy barrier for ion hopping. Most importantly, the proposed approach reveals that the energy barrier controlling ion hopping in polymer electrolytes is significantly (~30–40%) lower than that estimated using traditional Arrhenius fit. These new insights call for revisions of many earlier results based on apparent Arrhenius fits, and the newly proposed approach can provide more accurate guidance for the design of solid-state electrolytes with enhanced ionic conductivity.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Subnanometer Visualization of Spatially Varying Local Field Resonances that Drive Tip-Enhanced Optical Spectroscopy

Our knowledge of the electromagnetic fields that power modern nanoscale optical measurements, including (non)linear tip-enhanced Raman and photoluminescence, chiefly stems from numerical simulations. Aside from idealized in silico vs heterogeneous (nano)structures in the laboratory, challenges in quantitative descriptions of nanoscale light–matter interactions more generally stem from the very nature of the problem, which lies at the interface of classical and quantum theories. This is particularly the case in ultrahigh spatial resolution measurements that are sensitive to local optical field variations that take place on subnanometer length scales. Here this work approaches this challenge through extinction-based spectral nanoimaging experiments. We demonstrate <1 nm spatial resolution in hyperspectral extinction measurements that track spatially varying plasmon resonances. We describe the principles behind our experiments and highlight more general implications of our observations.

77 NANOSCIENCE AND NANOTECHNOLOGY↗
DOE OSTI2020

Theoretical H + O 3 rate coefficients from ring polymer molecular dynamics on an accurate global potential energy surface: assessing experimental uncertainties

Thermal rate coefficients and kinetic isotope effects have been calculated for an important atmospheric reaction H/D + O 3 → OH/OD + O2 based on an accurate permutation invariant polynomial-neural network potential energy surface, using ring polymer molecular dynamics (RPMD), quasi-classical trajectory (QCT) and variational transition-state theory (VTST) with multidimensional tunneling. The RPMD approach yielded results that are generally in better agreement with experimental rate coefficients than the VTST and QCT ones, especially at low temperatures, attributable to its capacity to capture quantum effects such as tunneling and zero-point energy. Additionally, the theoretical results support one group of existing experiments over the other. In addition, rate coefficients for the D + O 3 → OD + O 2 reaction are also reported using the same methods, which will allow a stringent assessment of future experimental measurements, thus helping to reduce the uncertainty in the recommended rate coefficients of this reaction.

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

Software for the frontiers of quantum chemistry: An overview of developments in the Q-Chem 5 package

This article summarizes technical advances contained in the fifth major release of the Q-Chem quantum chemistry program package, covering developments since 2015. A comprehensive library of exchange-correlation functionals, along with a suite of correlated many-body methods, continues to be a hallmark of the Q-Chem software. The many-body methods include novel variants of both coupled-cluster and configuration-interaction approaches along with methods based on the algebraic diagrammatic construction and variational reduced density-matrix methods. Methods highlighted in Q-Chem 5 include a suite of tools for modeling core-level spectroscopy, methods for describing metastable resonances, methods for computing vibronic spectra, the nuclear-electronic orbital method, and several different energy decomposition analysis techniques. High-performance capabilities including multithreaded parallelism and support for calculations on graphics processing units are described. Q-Chem boasts a community of well over 100 active academic developers, and the continuing evolution of the software is supported by an "open teamware" model and an increasingly modular design.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2026

Single-shot ultrafast dynamics of nanosecond pulsed plasmas: Transition from ps–ns nonequilibrium to near-full ionization

The ultrafast multi-stage evolution of state-defining properties in atmospheric-pressure nanosecond pulsed plasmas is quantified using single-discharge, jitter-free, continuous streak-sweep spectroscopy of N 2 (C → B) molecular and N + /O + ionic species with a single-shot time resolution as short as ∼160 ps. An early period of extreme nonequilibrium is observed with vibrational temperatures [T V (C)] dropping from ∼8000 K at <200 ps after plasma breakdown to <4000 K within ∼1 ns, with near-ambient rotational temperatures [T R (C)] ∼ 300 K due to limited collisional energy transfer. This early-time T V (C) trend is representative of a direct and unquenched look into the high-energy tail of the electron energy distribution function; thus, it tracks real-time changes in the mean electron energy via the N2(C) emission signatures. This is followed by the rapid onset of N + /O + ionic emission after a distinct time delay of ∼14.9 ns due to multistep chemical kinetics. The ionic emission enables determination of electron densities (n e ) >2 × 10 19 cm −3 and electron temperatures (T e ) >36 000 K, indicating the transition to a nearly fully ionized regime. The ps–ns temporal dynamics are also compared between air and N 2 plasmas to assess the influence of collisional partners, as well as across the anode, cathode, and central gap regions to identify spatial variations in plasma behavior. Finally, this work demonstrates versatile continuously probing approach capable of spectro-temporal and spatially resolved single-shot measurements of key state variables in the ps–ns evolution of atmospheric-pressure nanosecond pulsed plasmas.

Electric discharges↗
DOE OSTI2026

Reliability Assessment of Cooling Fans for PV Inverters: Testing, Modeling, and Case Studies

The reliability of photovoltaic (PV) inverters is critical for long-term solar system performance, with cooling fan failures frequently leading to costly downtime. While much research exists on general cooling fan reliability, little attention has been given to fans operating within PV inverters and their unique environmental challenges. Here, this article proposes a comprehensive methodology to address this gap. First, a failure mode and effects analysis is performed on fans to identify the key failure mechanisms in PV applications, their corresponding stressors, and the models necessary for lifetime prediction. Second, an accelerated life test is designed and conducted to collect valuable experimental data for PV inverter fans in a reasonable amount of time. Third, a mathematical conversion of dynamic mission profiles into effective constant stress levels is derived. Fourth, case studies are given, showcasing lifetime estimates that account for geographic variations in mission profile data. The results demonstrate that this integrated approach leads to an accurate reliability assessment for PV inverter cooling fans.

accelerated life testing (ALT)↗
DOE OSTI2021

Regional Relative Risk, a Physics-Based Metric for Characterizing Airborne Infectious Disease Transmission

Airborne infectious disease transmission events occur over a wide range of spatial scales and can be an important means of disease transmission. Physics- and biology-based models can assist in predicting airborne transmission events, overall disease incidence, and disease control strategy efficacy. We describe a new theory that extends current approaches for the case in which an individual is infected by a single airborne particle, including the scenario in which numerous infectious particles are present in the air but only one causes infection. A single infectious particle can contain more than one pathogenic microorganism and be physically larger than the pathogen itself. This approach allows robust relative risk estimates even when there is wide variation in (i) individual exposures and (ii) the individual response to that exposure (the pathogen dose-response function can take any mathematical form and vary by individual). Based on this theory, we propose the regional relative risk—a new metric, distinct from the traditional relative risk metric, that compares the risk between two regions. In theory, these regions can range from individual rooms to large geographic areas. In this paper, we apply the regional relative risk metric to outdoor disease transmission events over spatial scales ranging from 50 m to 20 km, demonstrating that in many common cases minimal input information is required to use the metric. Also, we demonstrate that the model predictions are consistent with data from prior outbreaks. Future efforts could apply and validate this theory for other spatial scales, such as transmission within indoor environments. This work provides context for (i) the initial stages of an airborne disease outbreak and (ii) larger-scale disease spread, including unexpected low-probability disease “sparks” that potentially affect remote populations, a key practical issue in controlling airborne disease outbreaks.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Analysis of Space-Conditioning Loads in Commercial Buildings

Space conditioning end-uses, which include heating, cooling, and ventilation, represent a significant fraction of commercial building energy use, with a wide variety of heating and cooling technology options available in the market. In the interest of improving the overall efficiency of heating, ventilation and air-conditioning (HVAC) technologies, governments, utilities and private sector entities have implemented a variety of market transformation policies that aim to influence consumer purchase decisions. To evaluate the costs and benefits of such programs, analysts typically postulate a hypothetical default equipment choice, and compare it to one that provides comparable service with lower energy and/or power use. The corresponding reduced operating cost provides a benefit that offsets the potential higher cost of improved efficiency. Typically, life-cycle cost or cash-flow analyses are used to quantify the net economic benefit. These analyses require the capability to assess how a given equipment design would perform across a broad range of characteristics, both of the building and of the local weather. While these assessments can be performed using customized building simulations, it is generally not practical to develop and validate detailed building simulation code to cover all the potential variations of equipment design and installation. An alternative, and somewhat simpler, approach is to solely use detailed building simulations to generate time series of heating and cooling loads in commercial buildings. These loads can then be used as input to more detailed, stand-alone engineering models that simulate HVAC system performance under different equipment designs. This approach was used to evaluate a range of high-efficiency commercial packaged air conditioner design options for the Department of Energy’s Appliance and Equipment Standards Program (DOE-EERE 2015). While there may be some loss of precision relative to full simulation, the accuracy of this approach is sufficient for practical applications of cost-benefit analysis. This report describes the development of a database of commercial building heating and cooling loads, generated using the EnergyPlus software package, a whole building energy use model supported by the Department of Energy (DOE-EERE 2020a). EnergyPlus takes as input a set of configuration files that describe the building itself (size, zoning, envelope characteristics, etc.) and the various systems within it (HVAC, lighting, water heating, etc.). This analysis uses a publicly available collection of commercial reference buildings (CRB), comprised of sixteen building types and three vintages (DOE-EERE 2020b; Deru et al 2011). Each building is simulated in eighteen different locations, covering a wide range of climatic conditions. The prototype building description files assign the type of HVAC equipment used, and capacities across climate zones.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2021

Root Influences on Mobilization and Export of Mineral-bound Soil Organic Matter

Biogeochemical cycles within mountainous watersheds are key regulators of ecosystem carbon storage and downstream nutrient loadings, and they have shown to be particularly vulnerable to climate change impacts. Increasing temperature and persistent droughts have already dramatically changed vegetation cover across the mountainous western US, with unknown consequences for soil carbon and nutrient cycles belowground. What remains elusive is to what extent associated changes in root-soil interactions may mobilize the vast pool of organic matter (OM) that has been stabilized by associations with minerals for centuries or millennia. Although plant root-driven OM mobilization from minerals may be a central control on carbon loss and nutrient export, such mechanisms are currently missing from conceptual and numerical models. The overall objective of this Exploratory Project is to identify the biogeochemical mechanisms by which roots destabilize mineral-associated organic matter (MAOM) and the cumulative impact on carbon and nutrient fate. To accomplish this goal, we integrated well-controlled laboratory experiments with in-field measurements and a scalable modeling approach. First, we conducted model system experiments to assess the vulnerability of MAOM to exudate-mediated mechanisms. Our results show that common root exudates effectively destabilize MAOM not only through direct, ligand-driven mobilization mechanisms, but also indirect, microbially-mediated mechanisms relying on secondary metabolites and enzymes. We further found that OM bound to poorly crystalline Fe and Al (hydr)oxides is more vulnerable to exudate-induced destabilization than OM bound to more crystalline phases, particularly in response to direct, ligand-promoted mechanisms. These findings demonstrate that the stability of MAOM is not just a function of their inherent properties, but also will depend in large parts on the ability of plant roots and microbes to produce exudates capable of triggering suitable mobilization mechanisms. We further employed a well-controlled rhizobox approach, combining advanced microsensor and mass spectrometry techniques, to resolve spatiotemporal variations in the composition and availability of exudates along single growing roots of grasses. Our results show that the composition of functionally relevant exudate compounds varies at extremely short time scales, seemingly shifting from ligands such as aromatic acids around root tips to less reactive metabolites such as amino acids around mature root segments. These results suggest a prevalence of direct MAOM mobilization mechanisms around the root tip, while indirect MAOM mobilization strategies may dominate around more mature root segments. Additionally, we utilized microsensor measurements in rhizoboxes to parameterize a rhizosphere (hydro)biogeochemistry reactive transport model (eSTOMP-ROOTS). The resulting model was used to assess how root exudations affects the stability of MAOM. Model simulations show that diel pulses of root exudation are strong enough to cause oscillations in biogeochemical conditions—particularly in pH, oxygen concentrations, and microbial activity— that repeatedly disrupt MAOM. Finally, we assessed field-scale impacts of root-driven MAOM destabilization over short and long time scales. A combination of in-field incubations and field-based characterizations showed that MAOM may not only be vulnerable to seasonal variations in root activity, but also to root-driven weathering over pedogenic time scales. In sum, results from our Exploratory Award highlight the strong control plant roots exert on MAOM mobilization and, thus, on the potential for carbon and nutrient export from watersheds.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Machine Learning Inference of Random Medium Properties

Earth materials are heterogeneous across a range of spatial scales, but the resolvability of small structures is limited by sparse data coverage, noise, bandlimitedness, and other difficulties. In practice, heterogeneities below a certain size cannot be recovered from seismic data except through statistical medium descriptions, which even then can be difficult to uniquely determine. To improve the characterization of such heterogeneities, we develop a novel supervised machine learning (ML) model that provides insight about the recoverability of statistical medium properties from elastic waveform data and succeeds despite cycle-skipping and other challenges well known from elastic waveform inversion. We demonstrate the approach using random media generated by superimposing self-affine random variations on homogeneous and layered background structures. After training on sparsely-recorded, high-frequency waveforms from hundreds of different random medium realizations, we show the ability of our ML model to recover correlation lengths and other statistical properties of interest to near-surface and crustal seismology, among other fields. For frequency passbands and spatial offsets encountered in seismology, Gaussian correlation lengths and the amplitude of the random variations relative to the background model are recovered even in challenging scenarios involving unknown medium parameters, complex crustal structures, and low signal-to-noise ratio. In comparison, von Kármán correlation lengths, which are related to larger-wavelength variations of the medium than Gaussian correlation lengths, are not as well recovered. These results provide one of the first and most systematic investigations of the recoverability of statistical properties of heterogeneities below the resolution limit of deterministic seismic tomography, and suggest practical ML strategies for high-frequency waveform seismology.

58 GEOSCIENCES↗