Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “network inference”

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

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

At least 1,369 records · Page 76
All sourcesNASA NTRSDOE OSTI
NASA NTRS2010

Efficient Web Services Policy Combination

Large-scale Web security systems usually involve cooperation between domains with non-identical policies. The network management and Web communication software used by the different organizations presents a stumbling block. Many of the tools used by the various divisions do not have the ability to communicate network management data with each other. At best, this means that manual human intervention into the communication protocols used at various network routers and endpoints is required. Developing practical, sound, and automated ways to compose policies to bridge these differences is a long-standing problem. One of the key subtleties is the need to deal with inconsistencies and defaults where one organization proposes a rule on a particular feature, and another has a different rule or expresses no rule. A general approach is to assign priorities to rules and observe the rules with the highest priorities when there are conflicts. The present methods have inherent inefficiency, which heavily restrict their practical applications. A new, efficient algorithm combines policies utilized for Web services. The method is based on an algorithm that allows an automatic and scalable composition of security policies between multiple organizations. It is based on defeasible policy composition, a promising approach for finding conflicts and resolving priorities between rules. In the general case, policy negotiation is an intractable problem. A promising method, suggested in the literature, is when policies are represented in defeasible logic, and composition is based on rules for non-monotonic inference. In this system, policy writers construct metapolicies describing both the policy that they wish to enforce and annotations describing their composition preferences. These annotations can indicate whether certain policy assertions are required by the policy writer or, if not, under what circumstances the policy writer is willing to compromise and allow other assertions to take precedence. Meta-policies are specified in defeasible logic, a computationally efficient non-monotonic logic developed to model human reasoning. One drawback of this method is that at one point the algorithm starts an exhaustive search of all subsets of the set of conclusions of a defeasible theory. Although the propositional defeasible logic has linear complexity, the set of conclusions here may be large, especially in real-life practical cases. This phenomenon leads to an inefficient exponential explosion of complexity. The current process of getting a Web security policy from combination of two meta-policies consists of two steps. The first is generating a new meta-policy that is a composition of the input meta-policies, and the second is mapping the meta-policy onto a security policy. The new algorithm avoids the exhaustive search in the current algorithm, and provides a security policy that matches all requirements of the involved metapolicies.

Vatan, Farrokh↗
DOE OSTI2024

Thinking Bayesian for plasma physicists

Bayesian statistics offers a powerful technique for plasma physicists to infer knowledge from the heterogeneous data types encountered. To explain this power, a simple example, Gaussian Process Regression, and the application of Bayesian statistics to inverse problems are explained. The likelihood is the key distribution because it contains the data model, or theoretic predictions, of the desired quantities. By using prior knowledge, the distribution of the inferred quantities of interest based on the data given can be inferred. Because it is a distribution of inferred quantities given the data and not a single prediction, uncertainty quantification is a natural consequence of Bayesian statistics. The benefits of machine learning in developing surrogate models for solving inverse problems are discussed, as well as progress in quantitatively understanding the errors that such a model introduces.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
NASA NTRS1994

Infrared spectroscopic measurements of the total column abundance of ethan (C2H6) above Lauder, New Zealand

Total vertical column aboundances of atmospheric ethane (C2H6) have been retrieved form more than 350 high-resolution infrared solar spectra recorded on 129 days between December 1992 and March 1994 at the Network for the Detection of Stratospheric Change (NDSC) station in Lauder, New Zealand (latitude 45.04 deg S, 169.68 deg E, 0.37 km altitude). The results are based on nonlinear least squares fits to narrow spectral intervals containing the unresolved C2H6 nu (sub 7) band (P)Q(sub 3) and (R)Q(sub 0) subbranches at 2976.8 and 2986.7/cm. The measured Lauder column abundances show a pronounced seasonal cycle with a maximum value of (8.6 +/- 0.5) x 10(exp 15) molecules/sq cm in September and a minimum value of (3.8 +/- 0.5) x 10 (exp 15) molecules/sq cm in February. Assuming a realistic vertical distribution for C2H6, these values correspond to surface level volume-mixing ratios of 0.52 and 0.23 ppbv (parts per billion by volume), respectively. The phase and relative amplitude of the Lauder C2H6 seasonal cycle are similar to corresponding values determined from infrared measurements at the Jungfraujoch (46.55 deg N, 7.98 deg E, 3.58 km altitude) and Mauna Loa (latitude 19.5 deg N, longitude 155.6 deg W, 3.5 km altitude) NDSC stations, but the inferred absolute magnitudes of the Lauder surface level mixing ratios during the same season are significantly lower. The Lauder C2H6 measurements are also compared with published southern hemisphere surface level sampling measurements and two-dimensional model calculations.

Rinsland, Curtis P.↗
NASA NTRS1995

The sensitivity of various mode sets for probing differential rotation shear zones

The potential of future data sets from global oscillations network group (GONG) and solar oscillations investigation (SOI) for resolving long-lived azimuthal jets and shearing flows, is investigated. Various artificial data sets are constructed, containing noise resembling that of a one-year observation run. These are inverted using a two dimensional regularized least squares inversion. The ability of this method to form well localized averages of the rotation rate, as measured by the averaging kernels, is investigated using an extensive mode set and subsets. It is shown that it is possible to keep the noise in the solution down to a few nHz in much of the solar interior, while obtaining a reasonable resolution for a GONG-like data set. At low latitudes in the middle of the convection zone, an angular resolution of less than 10 deg and a radial resolution of about 0.04 solar radii, are obtained. The averaging kernels depend on the mode set, and a reduction in the number of modes tends to introduce small-scale near surface structures into the averaging kernels which would adversely affect the inferred rotation rate.

Christensen-Dalsgaard, Joergen↗
DOE OSTI2015

Large eddy simulation of a turbulent non-reacting spray jet

We performed Large Eddy Simulation (LES) of a turbulent non-reacting n-Heptane spray jet, referred to as Spray H in the Engine Combustion Network (ECN), and executed a data analysis focused on key LES metrics such as fraction of resolved turbulent kinetic energy and similarity index. In the simulation, we used the dynamic structure model for the sub-grid stress, and the Lagrangian-based spray-parcel models coupled with the blob-injection model. The finest mesh-cell size used was characterized by an Adaptive Mesh Refinement (AMR) cell size of 0.0625 mm. To obtain ensemble statistics, we performed 28 numerical realizations of the simulation. Demonstrated by the comparison with experimental data in a previous study [7], this LES has accurately predicted global quantities, such as liquid and vapor penetrations. The analysis in this work shows that 14 realizations of LES are sufficient to provide a reasonable representation of the average flow behavior that is benchmarked against the 28-realization ensemble. With the current mesh, numerical schemes, and sub-grid scale turbulence model, more than 95% of the turbulent kinetic energy is directly resolved in the flow regions of interest. The large-scale flow structures inferred from a statistical analysis reveal a region of disorganized flow around the peripheral region of the spray jet, which appears to be linked to the entrainment process.

Hu, Bing↗
NASA NTRS1999

The Role of Velocity Redistribution in Enhancing the Intensity of the He II 304 A Line in the Quiet Sun Spectrum

We present observational evidence of the effect of small scale ("microturbulent") velocities in enhancing the intensity of the He II lambda304 line with respect to other transition region emission lines, a process we call "velocity redistribution". We first show results from the 1991 and 1993 flights of SERTS (Solar EUV Rocket Telescope and Spectrograph). The spectral resolution of the SERTS instrument was sufficient to infer that, at the spatial resolution of 5", the line profile is nearly gaussian both in the quiet Sun and in active regions. We were then able to determine, for the quiet Sun, a lower limit for the amplitude of non-thermal motions in the region of formation of the 304 A line of the order of 10 km/s. We estimated that, in the presence of the steep temperature gradients of the solar Transition Region (TR), velocities of this magnitude can significantly enhance the intensity of that line, thus at least helping to bridge the gap between calculated and observed values. We also estimated the functional dependence of such an enhancement on the relevant parameters (non-thermal velocities, temperature gradient, and pressure). We then present results from a coordinated campaign, using SOHO/CDS and H-alpha spectroheliograms from Coimbra Observatory, aimed at determining the relationship between regions of enhanced helium emission and chromospheric velocity fields and transition region emission in the quiescent atmosphere. Using these data, we examined the behavior of the He II lambda304 line in the quiet Sun supergranular network and compared it with other TR lines, in particular with O III lambda600. We also examined the association of 304 A emission with the so-called "coarse dark mottle", chromospheric structures seen in H-alpha red wing images and associated with spicules. We found that all these observations are consistent with the velocity redistribution picture.

Andretta, Vincenzo↗
DOE OSTI2024

Embedding Neural Thermal Scattering (NeTS) Modules in SERPENT for Higher Fidelity Advanced Reactor Analysis

When a neutron born in fission thermalizes to the order of $k$ $B$ $T$, it’s de-Broglie wavelength and energy approach the order of inter-atomic spacing and elementary lattice oscillations, respectively. $S$($a,β,t$) or the scattering law, uuantify these temperature-dependent crystallographic contributions to total cross section (or reaction rate). In a Monte Carlo analysis, cumulative distribution functions (CDFs) of $S$($a,β,t$) are loaded to memory from “A Compact ENDF” (ACE) files for stochastically selecting thermal scattered neutron trajectories. In this work, novel neural thermal scattering (NeTS) modules for $S$($a,β,t$) CDFs are designed, trained, serialized and embedded within SERPENT using Python’s limited C-API for on-the-fly deployment of crystalline graphite $S$($a,β,t$) sampling. Torchscript tracing and Numba just-in-time (JIT) compilation streamline neural inference on NVIDIA GPUs with CUDA libraries. Demonstrations of bare sphere thermalization of fast and thermal sources show excellent agreement between embedded NeTS in SERPENT and MCNP. With an explicit model of the reactor, NeTS can predict on-the-fly changes in TREAT neutron spectra as a function of local temperature, which can serve to improve transient and accident predictions in a multiphysics analysis framework. This framework can be further extended to account on-the-fly for changes in local graphitic microstructure to scattering cross sections, and outlines a novel coupling of modern machine learning with state-of-the-art reactor physics methods.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Low Temperature CO 2 Hydrogenation on Unsupported Mo 2 C Catalysts

CO 2 hydrogenation to methanol, a key reaction for decarbonizing the fuel and chemical industries, requires catalyst formulations that hydrogenate CO 2 selectively to methanol at temperatures where methanol conversion is not significantly equilibrium limited (<423 K). Herein we report continuous CO 2 hydrogenation at low temperatures (348-408 K, H 2 /CO 2 = 0.1-50, 5-35 bar) with high selectivity to methanol (up to ca. 80%) over unsupported β-Mo 2 C catalysts. Active site density quantification via titration with trifluoroacetic acid at reaction temperatures enables an assessment of site-specific rates. Methanation and reverse water gas shift (RWGS) occur concurrently with methanol synthesis during CO 2 hydrogenation over Mo 2 C. Reaction pathway analysis, product cofeeds, and reversibility formalisms show that all products form through primary reaction pathways from CO 2 , but secondary reactions of CO contribute significantly to rates of methanation. Dependences of forward rates on reactant and product concentration determined by independently varying the CO 2 , H 2 , CO, H 2 O, CH 3 OH, and CH 4 pressure in conjunction with reversibility formalisms reveal that all products form through H-assisted CO 2 activation and involve partially hydrogenated CO 2 -derived intermediates. Here, these inferences were verified by quantitative agreement between measured site-time yields and site-time yields predicted by closed form kinetic rate expressions in an integral reactor model over widely varying conditions (85-2000 kPa H 2 , 80-1500 kPa CO 2 , 0-45 kPa H 2 O, 0-21 kPa CO, 0-25 kPa CH 3 OH, 0-75 kPa CH 4 , 5-87 mol Mo s s mol CO 2 -1 ). Coverages calculated based on the kinetic model reveal that the Mo 2 C surface is covered with bidentate CO- and CO 2 -derived intermediates of the stoichiometry H 2 CO 2 and H 2 CO, indicating that H 2 and CO x do not compete for surface occupancy but instead adsorb cooperatively to form partially hydrogenated intermediates. Hydrogenation of the CO-derived H 2 CO** intermediate favors methanation, while hydrogenation of CO 2 -derived H 2 CO 2 ** favors methanol synthesis. Together, these findings demonstrate the ability of unsupported Mo 2 C to catalyze the hydrogenation of CO 2 to methanol at low temperatures and provide insight into the reaction network and mechanisms involved in its formation.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

Advancing Diagnostic Model Evaluation to Better Understand Water Shortage Mechanisms in Institutionally Complex River Basins

Abstract Water resources systems models enable valuable inferences on consequential system stressors by representing both the geophysical processes determining the movement of water and the human elements distributing it to its various competing uses. This study contributes a diagnostic evaluation framework that pairs exploratory modeling with global sensitivity analysis to enhance our ability to make inferences on water scarcity vulnerabilities in institutionally complex river basins. Diagnostic evaluation of models representing institutionally complex river basins with many stakeholders poses significant challenges. First, it needs to exploit a large and diverse suite of simulations to capture important human‐natural system interactions as well as institutionally aware behavioral mechanisms. Second, it needs to have performance metrics that are consequential and draw on decision‐relevant model outputs that adequately capture the multisector concerns that emerge from diverse basin stakeholders. We demonstrate the proposed model diagnostic framework by evaluating how potential interactions between changing hydrologic conditions and human demands influence the frequencies and durations of water shortages of varying magnitudes experienced by hundreds of users in a subbasin of the Colorado River. We show that the dominant factors shaping these effects vary both across users and, for an individual user, across percentiles of shortage magnitude. These differences hold even for users sharing diversion locations, demand levels or water right seniority. Our findings underline the importance of detailed institutional representation for such basins, as institutions strongly shape how dominant factors of stakeholder vulnerabilities propagate through the complex network of users.

Hadjimichael, Antonia↗
NASA NTRS2019

GWTC-1: A Gravitational-Wave Transient Catalog of Compact Binary Mergers Observedby LIGO and Virgo During the First and Second Observing Runs

We present the results from three gravitational-wave searches for coalescing compact binaries with component masses above1M during the first and second observing runs of the advanced gravitational-wave detector network. During the first observing run (O1), from September 12, 2015 to January 19, 2016, gravitational waves from three binary black hole mergers were detected. The second observing run (O2), which ran from November 30, 2016 to August 25, 2017, saw the first detection of gravitational waves from a binary neutron star in spiral, in addition to the observation of gravitational waves from a total of seven binary black hole mergers, four of which we report here for the first time: GW170729, GW170809, GW170818, and GW170823. For all significant gravitational-wave events, we provide estimates of the source properties. The detected binary black holes have total masses between18.6þ3.2−0.7M and 84.4þ15.8−11.1M and range in distance between 320þ120−110 and 2840þ1400−1360Mpc. No neutron star–black hole mergers were detected. In addition to highly significant gravitational-wave events, we also provide a list of marginal event candidates with an estimated false-alarm rate less than 1 per 30 days. From these results over the first two observing runs, which include approximately one gravitational-wave detection per 15 days of data searched, we infer merger rates at the 90% confidence intervals of 110−3840Gpc−3y−1 for binary neutron stars and 9.7−101Gpc−3y−1 for binary black holes assuming fixed population distributions and determine a neutron star–black hole merger rate 90% upper limit of 610Gpc−3y−1.

Astrophysics↗
DOE OSTI2021

A Reconfigurable Neural Network ASIC for Detector Front-End Data Compression at the HL-LHC

Despite advances in the programmable logic capabilities of modern trigger systems, a significant bottleneck remains in the amount of data to be transported from the detector to off-detector logic where trigger decisions are made. We demonstrate that a neural network (NN) autoencoder model can be implemented in a radiation-tolerant application-specific integrated circuit (ASIC) to perform lossy data compression alleviating the data transmission problem while preserving critical information of the detector energy profile. For our application, we consider the high-granularity calorimeter from the Compact Muon Solenoid (CMS) experiment at the CERN Large Hadron Collider. The advantage of the machine learning approach is in the flexibility and configurability of the algorithm. By changing the NN weights, a unique data compression algorithm can be deployed for each sensor in different detector regions and changing detector or collider conditions. To meet area, performance, and power constraints, we perform quantization-aware training to create an optimized NN hardware implementation. The design is achieved through the use of high-level synthesis tools and the hls4ml framework and was processed through synthesis and physical layout flows based on a low-power (LP)-CMOS 65-nm technology node. The flow anticipates 200 Mrad of ionizing radiation to select gates and reports a total area of 3.6 mm 2 and consumes 95 mW of power. The simulated energy consumption per inference is 2.4 nJ. Furthermore, this is the first radiation-tolerant on-detector ASIC implementation of an NN that has been designed for particle physics applications.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2020

Turbulence model reduction by deep learning

A defining problem of turbulence theory is to produce a predictive model for turbulent fluxes. These have profound implications for virtually all aspects of the turbulence dynamics. In magnetic confinement devices, drift-wave turbulence produces anomalous fluxes via cross-correlations between fluctuations. In this work, we introduce an alternative, data-driven method for parametrizing these fluxes. The method uses deep supervised learning to infer a reduced mean-field model from a set of numerical simulations. We apply the method to a simple drift-wave turbulence system and find a significant new effect which couples the particle flux to the local gradient of vorticity. Notably, here, this effect is much stronger than the oft-invoked shear suppression effect. We also recover the result via a simple calculation. The vorticity gradient effect tends to modulate the density profile. In addition, our method recovers a model for spontaneous zonal flow generation by negative viscosity, stabilized by nonlinear and hyperviscous terms. We highlight the important role of symmetry to implementation of the new method.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2024

Latent Stochastic Differential Equations for Modeling Quasar Variability and Inferring Black Hole Properties

Quasars are bright and unobscured active galactic nuclei (AGN) thought to be powered by the accretion of matter around supermassive black holes at the centers of galaxies. The temporal variability of a quasar’s brightness contains valuable information about its physical properties. The UV/optical variability is thought to be a stochastic process, often represented as a damped random walk described by a stochastic differential equation (SDE). Upcoming wide-field telescopes such as the Rubin Observatory Legacy Survey of Space and Time (LSST) are expected to observe tens of millions of AGN in multiple filters over a ten year period, so there is a need for efficient and automated modeling techniques that can handle the large volume of data. Latent SDEs are machine learning models well suited for modeling quasar variability, as they can explicitly capture the underlying stochastic dynamics. In this work, we adapt latent SDEs to jointly reconstruct multivariate quasar light curves and infer their physical properties such as the black hole mass, inclination angle, and temperature slope. Our model is trained on realistic simulations of LSST ten year quasar light curves, and we demonstrate its ability to reconstruct quasar light curves even in the presence of long seasonal gaps and irregular sampling across different bands, outperforming a multioutput Gaussian process regression baseline. Our method has the potential to provide a deeper understanding of the physical properties of quasars and is applicable to a wide range of other multivariate time series with missing data and irregular sampling.

79 ASTRONOMY AND ASTROPHYSICS↗
NASA NTRS2014

Oceanic Storm Characteristics off the Kennedy Space Center Coast

Natural cloud-to-ground lightning may behave differently depending on the characteristics of the attachment mediums, including the peak current (inferred from radiation fields) and the number of ground strike locations per flash. Existing literature has raised questions over the years on these characteristics of lightning over oceans, and the behaviors are not yet well understood. To investigate this we will obtain identical electric field observations over adjacent land and ocean regions during both clear air and thunderstorm periods. Oceanic observations will be obtained using a 3-meter NOAA buoy that has been instrumented with a Campbell Scientific electric field mill and New Mexico Techs slow antenna, to measure the electric fields aloft. We are currently obtaining measurements from this system on-shore at the Florida coast, to calibrate and better understand the behavior of the system in elevated-field environments. Sometime during winter 2013, this system will be moored 20NM off the coast of the Kennedy Space Center. Measurements from this system will be compared to the existing on-shore electric field mill suite of 31 sensors and a coastal slow antenna. Supporting observations will be provided by New Mexico Techs Lightning Mapping Array, the Eastern Range Cloud to Ground Lightning Surveillance System, and the National Lightning Detection Network. An existing network of high-speed cameras will be used to capture cloud-to-ground lightning strikes over the terrain regions to identify a valid data set for analysis. This on-going project will demonstrate the value of off-shore electric field measurements for safety-related decision making at KSC, and may improve our understanding of relative lightning risk to objects on the ground vs. ocean. This presentation will provide an overview of this new instrumentation, and a summary of our progress to date.

Ocean↗
DOE OSTI2022

Mass Detection for Heavy-Duty Vehicles using Gaussian Belief Propagation

Predicting vehicle mass is critical to accurately estimate energy use and emissions of commercial trucks. However, data from vehicle telematics is often not at sufficient temporal resolution or accuracy for use in model-based detection methods. In this work, a new statistical mass prediction technique is described for heavy-duty vehicles that incorporates the use Gaussian Belief Propagation (GBP) for probabilistic inference. Similar to Bayesian inference models, the GBP model typically requires less labeled training data than other contemporary machine learning techniques. First, a factor graph is constructed, and a set of Gaussian belief nodes with associated means and variances are fitted to the training data. To better handle noisy input data, the GBP mass prediction model utilizes a k-nearest factors (kNF) algorithm for probabilistic inference on unseen testing data. The proposed method is compared with a classical weighted k-nearest neighbors (kNN) regressor. This statistical kNF-GBP model works even with low-quantity, low-quality initial training data, while being capable of realtime mass estimation. Unlike the kNN regressor, the GBP model produces a measure of uncertainty with its predictions. The proposed method is validated using curve-sampled driving data collected from multiple cloud-connected Class 8 regional haul diesel trucks. Both the kNN regressor and the kNF-GBP mass prediction model were able to predict payload mass with coefficients of determination above 0.97 with minimal data preprocessing.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
NASA NTRS2012

Observations of GEO Debris with the Magellan 6.5-m Telescopes

Optical observations of geosynchronous orbit (GEO) debris are important to address two questions: 1. What is the distribution function of objects at GEO as a function of brightness? With some assumptions, this can be used to infer a size distribution. 2. Can we determine what the likely composition of individual GEO debris pieces is from studies of the spectral reflectance of these objects? In this paper we report on optical observations with the 6.5-m Magellan telescopes at Las Campanas Observatory in Chile that attempt to answer both questions. Imaging observations over a 0.5 degree diameter field-of-view have detected a significant population of optically faint debris candidates with R > 19th magnitude, corresponding to a size smaller than 20 cm assuming an albedo of 0.175. Many of these objects show brightness variations larger than a factor of 2, suggesting either irregular shapes or albedo variations or both. The object detection rate (per square degree per hour) shows an increase over the rate measured in the 0.6-m MODEST observations, implying an increase in the population at optically fainter levels. Assuming that the albedo distribution is the same for both samples, this corresponds to an increase in the population of smaller size debris. To study the second issue, calibrated reflectance spectroscopy has been obtained of a sample of GEO and near GEO objects with orbits in the public U.S. Space Surveillance Network catalog. With a 6.5-m telescope, the exposures times are short (30 seconds or less), and provide simultaneous wavelength coverage from 4500 to 8000 Angstroms. If the observed objects are tumbling, then simultaneous coverage and short exposure times are essential for a realistic assessment of the object fs spectral signature. We will compare the calibrated spectra with lab-based measurements of simple spacecraft surfaces composed of a single material.

Seitzer, Patrick↗
DOE OSTI2023

KEBLM: Knowledge-Enhanced Biomedical Language Models

Pretrained language models (PLMs) have demonstrated strong performance on many natural language processing (NLP) tasks. Despite their great success, these PLMs are typically pretrained only on unstructured free texts without leveraging existing structured knowledge bases that are readily available for many domains, especially scientific domains. As a result, these PLMs may not achieve satisfactory performance on knowledge-intensive tasks such as biomedical NLP. Comprehending a complex biomedical document without domain-specific knowledge is challenging, even for humans. Inspired by this observation, we propose a general framework for incorporating various types of domain knowledge from multiple sources into biomedical PLMs. We encode domain knowledge using lightweight adapter modules, bottleneck feed-forward networks that are inserted into different locations of a backbone PLM. For each knowledge source of interest, we pretrain an adapter module to capture the knowledge in a self-supervised way. We design a wide range of self-supervised objectives to accommodate diverse types of knowledge, ranging from entity relations to description sentences. Once a set of pretrained adapters is available, we employ fusion layers to combine the knowledge encoded within these adapters for downstream tasks. Each fusion layer is a parameterized mixer of the available trained adapters that can identify and activate the most useful adapters for a given input. Our method diverges from prior work by including a knowledge consolidation phase, during which we teach the fusion layers to effectively combine knowledge from both the original PLM and newly-acquired external knowledge using a large collection of unannotated texts. After the consolidation phase, the complete knowledge-enhanced model can be fine-tuned for any downstream task of interest to achieve optimal performance. Extensive experiments on many biomedical NLP datasets show that our proposed framework consistently improves the performance of the underlying PLMs on various downstream tasks such as natural language inference, question answering, and entity linking. These results demonstrate the benefits of using multiple sources of external knowledge to enhance PLMs and the effectiveness of the framework for incorporating knowledge into PLMs. Finally, while primarily focused on the biomedical domain in this work, our framework is highly adaptable and can be easily applied to other domains, such as the bioenergy sector.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2021

Dynamic Earthquake Triggering in Southern California in High Resolution: Intensity, Time Decay, and Regional Variability

Abstract Earthquake triggering by seismic waves has been recognized as a phenomenon for nearly 30 years. However, our ability to study dynamic triggering has been limited by our ability to capture the triggering stresses accurately and record the resultant earthquakes. Here we use full waveforms from a dense seismic network and a modern, high‐resolution seismic catalog to measure triggering in Southern California from 2008 to 2017 based on interevent time ratios. We find that the fractional seismicity rate change, which we term triggering intensity or triggerability, as a function of peak strain change for the period of ∼20 s due to distant earthquakes is monotonically increasing and compatible with earlier measurements made with a disjoint data set from 1984 to 2008. A triggering strain of 1 microstrain is equivalent to the local productivity generated by an M 1.8 earthquakes. This result implies that a prediction of seismicity rate changes can be made based on recorded ground shaking using the same formalism as currently used for aftershock prediction. For a teleseismic event, this small level of triggering occurs throughout the region and thus aggregates to a regional effect. We find that the triggering rate decays after the triggerer follows an Omori‐Utsu law, but at a much slower rate than a typical aftershock sequence. The slow decay rate suggests that an ancillary process such as creep or fluid flow must be part of dynamic triggering. The prevalence of triggering in areas of creep or fluid involvement reinforces this inference. A triggering cascade of secondary earthquakes is insufficient to explain the data.

Miyazawa, Masatoshi↗