Search NASA⌕ Search

SEARCH · Search NASA

Results for “model interpretability”

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 1,225 records · Page 68

Polarization and Compressibility of Oblique Kinetic Alfven Waves

Even though solar wind, as a collisionless plasma, is properly described by the kineticMaxwell-Vlasov description, it can be argued that much of our understanding of solar wind observational data comes from an interpretation and numerical modeling which is based on a fluid description of magnetohydrodynamics. In recent years, there has been a significant interest in better understanding the importance of kinetic effects, i.e. the differences between the kinetic and usual fluid descriptions. Here we concentrate on physical properties of oblique kinetic Alfvn waves (KAWs), which are often recognized as one of the key ingredients in the solar wind turbulence cascade. We use three different fluid models with various degrees of complexity and calculate polarization and magnetic compressibility of oblique KAWs (propagation angle q = 88), which we compare to solutions derived from linear kinetic theory. We explore a wide range of possible proton plasma b = [0.1,10.0] and a wide range of length scales krL = [0.001,10.0]. It is shown that the classical isotropic two-fluid model is very compressible in comparison with kinetic theory and that the largest discrepancy occurs at scales larger than the proton gyroscale. We also show that the two-fluid model contains a large error in the polarization of electric field, even at scales krL 1. Furthermore, to understand these discrepancies between the two-fluid model and the kinetic theory, we employ two versions of the Landau fluid model that incorporate linear low-frequency kinetic effects such as Landau damping and finite Larmor radius (FLR) corrections into the fluid description. It is shown that Landau damping significantly reduces the magnetic compressibility and that FLR corrections (i.e. nongyrotropic contributions) are required to correctly capture the polarization.We also show that, in addition to Landau damping, FLR corrections are necessary to accurately describe the damping rate of KAWs. We conclude that kinetic effects are important even at scales which are significantly larger than the proton gyroscale krL 1.

Turbulence↗

Exploring NASA OMI Level 2 Data With Visualization

Satellite data products are important for a wide variety of applications that can bring far-reaching benefits to the science community and the broader society. These benefits can best be achieved if the satellite data are well utilized and interpreted, such as model inputs from satellite, or extreme events (such as volcano eruptions, dust storms, etc.).Unfortunately, this is not always the case, despite the abundance and relative maturity of numerous satellite data products provided by NASA and other organizations. Such obstacles may be avoided by allowing users to visualize satellite data as images, with accurate pixel-level (Level-2) information, including pixel coverage area delineation and science team recommended quality screening for individual geophysical parameters. We present a prototype service from the Goddard Earth Sciences Data and Information Services Center (GES DISC) supporting Aura OMI Level-2 Data with GIS-like capabilities. Functionality includes selecting data sources (e.g., multiple parameters under the same scene, like NO2 and SO2, or the same parameter with different aggregation methods, like NO2 in OMNO2G and OMNO2D products), user-defined area-of-interest and temporal extents, zooming, panning, overlaying, sliding, and data subsetting, reformatting, and reprojection. The system will allow any user-defined portal interface (front-end) to connect to our backend server with OGC standard-compliant Web Mapping Service (WMS) and Web Coverage Service (WCS) calls. This back-end service should greatly enhance its expandability to integrate additional outside data-map sources.

Level 2↗

Exploring NASA OMI Level 2 Data With Visualization

Satellite data products are important for a wide variety of applications that can bring far-reaching benefits to the science community and the broader society. These benefits can best be achieved if the satellite data are well utilized and interpreted, such as model inputs from satellite, or extreme events (such as volcano eruptions, dust storms,... etc.). Unfortunately, this is not always the case, despite the abundance and relative maturity of numerous satellite data products provided by NASA and other organizations. Such obstacles may be avoided by allowing users to visualize satellite data as "images", with accurate pixel-level (Level-2) information, including pixel coverage area delineation and science team recommended quality screening for individual geophysical parameters. We present a prototype service from the Goddard Earth Sciences Data and Information Services Center (GES DISC) supporting Aura OMI Level-2 Data with GIS-like capabilities. Functionality includes selecting data sources (e.g., multiple parameters under the same scene, like NO2 and SO2, or the same parameter with different aggregation methods, like NO2 in OMNO2G and OMNO2D products), user-defined area-of-interest and temporal extents, zooming, panning, overlaying, sliding, and data subsetting, reformatting, and reprojection. The system will allow any user-defined portal interface (front-end) to connect to our backend server with OGC standard-compliant Web Mapping Service (WMS) and Web Coverage Service (WCS) calls. This back-end service should greatly enhance its expandability to integrate additional outside data/map sources.

Aura↗

Advanced Computational Modeling Approaches for Shock Response Prediction

Motivation: (1) The activation of pyroshock devices such as explosives, separation nuts, pin-pullers, etc. produces high frequency transient structural response, typically from few tens of Hz to several hundreds of kHz. (2) Lack of reliable analytical tools makes the prediction of appropriate design and qualification test levels a challenge. (3) In the past few decades, several attempts have been made to develop methodologies that predict the structural responses to shock environments. (4) Currently, there is no validated approach that is viable to predict shock environments overt the full frequency range (i.e., ~100 Hz to 10 kHz). Scope: (1) Model, analyze, and interpret space structural systems with complex interfaces and discontinuities, subjected to shock loads. (2) Assess the viability of a suite of numerical tools to simulate transient, non-linear solid mechanics and structural dynamics problems, such as shock wave propagation.

Derkevorkian, Armen↗

An Unobtrusive System to Measure, Assess, and Predict Cognitive Workload in Real-World Environments

Across many careers, individuals face alternating periods of high and low attention and cognitive workload, which can result in impaired cognitive functioning and can be detrimental to job performance. For example, some professions (e.g., fire fighters, emergency medical personnel, doctors and nurses working in an emergency room, pilots) require long periods of low workload (boredom), followed by sudden, high-tempo operations during which they may be required to respond to an emergency and perform at peak cognitive levels. Conversely, other professions (e.g., air traffic controllers, market investors in financial industries, analysts) require long periods of high workload and multitasking during which the addition of just one more task results in cognitive overload resulting in mistakes. An unobtrusive system to measure, assess, and predict cognitive workload could warn individuals, their teammates, or their supervisors when steps should be taken to augment cognitive readiness. In this talk I will describe an approach to this problem that we have found to be successful across work domains including: (1) a suite of unobtrusive, field-ready neurophysiological, physiological, and behavioral sensors that are chosen to best suit the target environment; (2) custom algorithms and statistical techniques to process and time-align raw data originating from the sensor suite; (3) probabilistic and statistical models designed to interpret the data into the human state of interest (e.g., cognitive workload, attention, fatigue); (4) and machine-learning techniques to predict upcoming performance based on the current pattern of events, and (5) display of each piece of information depending on the needs of the target user who may or may not want to drill down into the functioning of the system to determine how conclusions about human state and performance are determined. I will then focus in on our experimental results from our custom functional near-infrared spectroscopy sensor, designed to operate in real-world environments to be worn comfortably (e.g., positioned into a baseball cap or a surgeons cap) to measure changes in brain blood oxygenation without adding burden to the individual being assessed.

workload collection↗

Failure Bounding And Sensitivity Analysis Applied To Monte Carlo Entry, Descent, And Landing Simulations

In the study of entry, descent, and landing, Monte Carlo sampling methods are often employed to study the uncertainty in the designed trajectory. The large number of uncertain inputs and outputs, coupled with complicated non-linear models, can make interpretation of the results difficult. Three methods that provide statistical insights are applied to an entry, descent, and landing simulation. The advantages and disadvantages of each method are discussed in terms of the insights gained versus the computational cost. The first method investigated was failure domain bounding which aims to reduce the computational cost of assessing the failure probability. Next a variance-based sensitivity analysis was studied for the ability to identify which input variable uncertainty has the greatest impact on the uncertainty of an output. Finally, probabilistic sensitivity analysis is used to calculate certain sensitivities at a reduced computational cost. These methods produce valuable information that identifies critical mission parameters and needs for new technology, but generally at a significant computational cost.

Gaebler, John A.↗

Internal Variability and Disequilibrium Confound Estimates of Climate Sensitivity from Observations

An emerging literature suggests that estimates of equilibrium climate sensitivity (ECS) derived from recent observations and energy balance models are biased low because models project more positive climate feedback in the far future. Here we use simulations from the Coupled Model Intercomparison Project Phase 5 (CMIP5) to show that across models, ECS inferred from the recent historical period (1979-2005) is indeed almost uniformly lower than that inferred from simulations subject to abrupt increases in CO2-radiative forcing. However, ECS inferred from simulations in which sea surface temperatures are prescribed according to observations is lower still. ECS inferred from simulations with prescribed sea surface temperatures is strongly linked to changes to tropical marine low clouds. However, feedbacks from these clouds are a weak constraint on long-term model ECS. One interpretation is that observations of recent climate changes constitute a poor direct proxy for long-term sensitivity.

Marvel, Kate↗

Assessing impacts of canopy 3D structure on chlorophyll fluorescence radiance and radiative budget of deciduous forest stands using DART

Although remote sensing (RS) of solar-induced chlorophyll fluorescence (SIF) is increasingly used as a valuable source of information about vegetation photosynthetic activity, the RS SIF observations are significantly influenced by canopy-specific structural features (i.e., canopy architecture including leaf area index and presence of woody components), atmospheric conditions during their acquisition (e.g., proportion of direct and diffuse irradiance) and observational geometric configurations (e.g., sun and viewing directions). Radiative transfer (RT) models have the potential to provide a better understanding of the canopy structural effects on the SIF emission and RS signals. Here, we used the DART model to assess the daily influence, from morning to evening, of forest 3D architecture on SIF nadir radiance, emission, escape factor and nadir yield of eight 100 m × 100 m forest study plots established in a temperate deciduous forest of the Smithsonian Environmental Research Center (Edgewater, MD, USA). The 3D architecture of each plot was derived from airborne LiDAR. DART simulations of these 3D forest plots and their 1D (i.e., vertical profile of sun-adapted and shade-adapted leaves) and 0D (i.e., homogeneous layer of sun-adapted leaves above an homogeneous layer of shade-adapted leaves) abstractions were compared to assess the relative errors (ε1D−3D and ε0D−3D) associated with horizontal and vertical structural heterogeneity, respectively. Forest 3D structure, especially horizontal heterogeneity, had a great influence on forest nadir SIF radiance, resulting in ε1D−3D up to 55% at 8:00 and 18:00 (i.e., for oblique sun directions). The key indicators of this impact, in the descending order of importance, were the SIF escape factor (ε1D−3D up to 40%), the attenuation of incident photosynthetically active radiation (ε1D−3D less than 5%), and the SIF emission yield (ε1D−3D less than 2%). The influence of forest architecture on the nadir SIF escape factor and SIF yield (ε1D−3D up to 40%) varied over time, with differences in forest stand structure, and per spectral domain, being always larger between 640 and 700 nm than between 700 and 850 nm. In addition, woody elements demonstrated a large influence on forest SIF radiance due to their “shading” effect (ε up to 17%) and their “blocking” effect (ε ≈ 10%), both of them higher for far-red than for red SIF. These results underline the importance of 3D forest canopy architecture, especially 2D heterogeneity, and inclusion of woody elements in RT modeling used for interpretation of the RS SIF signal, and subsequently for the estimation of gross primary production and detection of vegetation stress.

Omar Regaieg↗

Coevolution of Machine Learning and Process-Based Modelling to Revolutionize Earth and Environmental Sciences: A Perspective

Machine learning (ML) applications in Earth and environmental sciences (EES) have gained incredible momentum in recent years. However, these ML applications have largely evolved in ‘isolation’ from the mechanistic, process-based modelling (PBM) paradigms, which have historically been the cornerstone of scientific discovery and policy support. In this perspective, we assert that the cultural barriers between the ML and PBM communities limit the potential of ML, and even its ‘hybridization’ with PBM, for EES applications. Fundamental, but often ignored, differences between ML and PBM are discussed as well as their strengths and weaknesses in light of three overarching modelling objectives in EES, (1) nowcasting and prediction, (2) scenario analysis, and (3) diagnostic learning. The paper ponders over a ‘coevolutionary’ approach to model building, shifting away from a borrowing to a co-creation culture, to develop a generation of models that leverage the unique strengths of ML such as scalability to big data and high-dimensional mapping, while remaining faithful to process-based knowledge base and principles of model explainability and interpretability, and therefore, falsifiability.

Saman Razavi↗

New Insights into the First Two PSP Solar Encounters Enabled by Modeling Analysis with ADAPT-WSA

Parker Solar Probe's (PSP's) unique orbital path allows us to observe the solar wind closer to the Sun than ever before. Essential to advancing our knowledge of solar wind and energetic particle formation is identifying the sources of PSP observations. We report on results for the first two PSP solar encounters derived using the Wang–Sheeley–Arge (WSA) model driven by Air Force Data Assimilative Photospheric Flux Transport (ADAPT) model maps. We derive the coronal magnetic field and the 1 R⊙ source regions of the PSP-observed solar wind. We validate our results with the solar wind speed and magnetic polarity observed at PSP. When modeling results are very reliable, we derive time series of model-derived spacecraft separation from the heliospheric current sheet, magnetic expansion factor, coronal hole boundary distance, and photospheric field strength along the field lines estimated to be connected to the spacecraft. We present new results for Encounter 1, which show time evolution of the far-side mid-latitude coronal hole that PSP corotates with. We discuss how this evolution coincides with solar wind speed, density, and temperature observed at the spacecraft. During Encounter 2, a new active region emerges on the solar far side, making it difficult to model. We show that ADAPT-WSA output agrees well with PSP observations once this active region rotates onto the near side, allowing us to reliably estimate the solar wind sources retrospectively for most of the encounter. We close with ways in which coronal modeling enables scientific interpretation of these encounters that would otherwise not have been possible.

Solar wind↗

Stochastic Verification by Analysis for Autonomous Systems Management Architecture (ASMA)

The Gateway Vehicle Systems Manager (VSM) is the top-level of a distributed, hierarchical software control system. VSM is data-driven and will make decisions related to mission, fault, resource management and vehicle control. These attributes combined with a high degree of autonomy make it susceptible to emergent behavior. In order to achieve the high level of confidence needed in this critical system, the VSM team has developed a multifaceted verification strategy employing traditional verification techniques, simulation, model checking, and runtime verification. Individual algorithms are verified using conventional testing and model checking using assume-guarantee contracts. A discrete event-based simulation approach is being developed to verify timelines. This presentation describes an enhancement to the verification approach using analysis to enhance system robustness by detecting and resolving the potential for emergent behavior. The verification by analysis employs a Software in the Loop (SITL) environment with real flight software executing on emulated processors, simulations of vehicle subsystems, flight dynamics, and human inputs. Since the possible input space and configuration data set are too large for exhaustive testing, a Monte Carlo approach is used to cover feasible scenarios, augmented with corner cases and known higher-risk scenarios. A key problem in using Monte Carlo-based system verification is evaluating test results to ensure that system behavior is correct. The presentation describes the approach the VSM team uses to monitor behavior for compliance with predetermined boundaries and to identify anomalous behavior for further analysis. This presentation describes the multi-level systems approach to verification, and the simulation-based layer that covers the feasible state space: 1. Overview of the Gateway VSM 2. Special challenges due to heterogeneous, hierarchical architecture 3. Modeling and simulation environment using flight software and system simulations 4. Developing input sets to ensure state-space coverage 5. Developing model and data configuration sets to ensure model coverage 6. Interpreting results without predetermined outcomes 7. Lessons learned and future work

Verification and Validation↗

Observations of the Temperature Dependent Response of Ozone to NOx Reductions in an Urban Plume

Observations of NO{sub x} in the Sacramento, CA region show that mixing ratios decreased by 30% between 2001 and 2008. Here we use an observation-based method to quantify net ozone production rates in the outflow from the Sacramento metropolitan region and examine the O{sub 3} decrease resulting from reductions in NO{sub x} emissions. This observational method does not rely on assumptions about detailed chemistry of ozone production, rather it is an independent means to verify and test these assumptions. We use an instantaneous steady-state model as well as a detailed 1-D plume model to aid in interpretation of the ozone production inferred from observations. In agreement with the models, the observations show that early in the plume, the NO{sub x} dependence for O{sub x} (O{sub x} = O{sub 3}+NO{sub 2}) production is strongly coupled with temperature, suggesting that temperature dependent biogenic VOC emissions can drive O{sub x} production between NO{sub x}-limited and NO{sub x}-suppressed regimes. As a result, NO{sub x} reductions were found to be most effective at higher temperatures over the 7 year period. We show that violations of the California 1-hour O{sub 3} standard (90 ppb) in the region have been decreasing linearly with decreases in NO{sub x} (at a given temperature) and predict that reductions of NO{sub x} concentrations (and presumably emissions) by an additional 30% (relative to 2007 levels) will eliminate violations of the state 1 hour standard in the region. If current trends continue, a 30% decrease in NO{sub x} is expected by 2012, and an end to violations of the 1 hour standard in the Sacramento region appears to be imminent.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Deep Learning for Unsupervised Insider Threat Detection in Structured Cybersecurity Data Streams

Situational awareness data is a prototypical example of structured heterogeneous streaming data with asynchronous timing that can quickly scale beyond the cognitive power of a human analyst. We present deep learning methods that can train in real time from unlabeled streams and provide probabilistic anomaly scores. We compare our novel deep and recurrent neural network models to three baseline systems using threat detection recall on the CERT Insider Threat Dataset v6.2. In our experiments, our deep neural network and our recurrent neural network systems outperformed all baselines, and we found that using aggregated activity count features yielded the best performance. Our best model produces on average an anomaly score of 95.53 percentile for the day, suggesting it can facilitate a great reduction in analyst workload. Our models are also interpretable, in that they break down the anomaly score into individual features of user behavior, which could further aid analysts in reviewing potential cases of insider threat.

Tuor, Aaron R.↗

Dynamic Measurements of the Structural Evolution of Material Defects at the Mesoscale

This is a technical report to be published as part of the FY2021 annual overview package. We have designed experiments and developed analytical methods using images from x-ray and electron microscopy in order to quantify the behavior of crystal dislocation defects at their native nanometer-length and micro-length scales. The results of these analyses are used to directly inform theoretical physics models and guide interpretation of dark field x-ray microscopy (DFXM) images that offer spatial resolution surpassing the resolution of beamline optics. Additionally, we developed new analytical methods to automate the alignment of optics hardware used in both synchrotron and x-ray free electron laser beamline facilities. The tools and workflows developed for this project are now being deployed as proof-of-concept in reaction history analysis (RHA) for future programmatic integration.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Robotic Measurement of Holdup Deposit Volume in Gaseous Diffusion Piping to Quantify U-235 Content - 18375

During D and D of old gaseous diffusion uranium enrichment facilities, radiometric assay of U-235 holdup in pipes is a costly, time consuming, and labor-intensive process. Subject to human interpretation using approximate modeling, radiometric assay introduces significant challenges. Taking advantage of routine demolition activities in which D and D cuts pipes open, robotic in-pipe assay is explored. The novel method introduced here generates models of the internal pipe and deposit surface geometries that are used to derive volumetric quantities. The pipe surface is sensed using a non-contact inductive proximity sensor and the deposit surface is sensed using an optical laser triangulation sensor. These sensors are mounted on a spinning disk and driven down the pipe to construct a helical point cloud of discrete measurements. Surfaces are fit to each of the two point clouds to create a watertight volume that represents the holdup. This provides the location and volume per foot of pipe that is used to compare against criticality incredible (CI) thresholds during evaluation. The robotic system collects data autonomously, deploying and returning to the same pipe opening from which it is launched. This provides redundant measurements of the pipe and deposit surfaces as well as odometry used in localization. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Dense Plasma X-ray Scattering: Methods and Applications

We have developed accurate x-ray scattering techniques to measure the physical properties of dense plasmas. Temperature and density are inferred from inelastic x-ray scattering data whose interpretation is model-independent for low to moderately coupled systems. Specifically, the spectral shape of the non-collective Compton scattering spectrum directly reflects the electron velocity distribution. In partially Fermi degenerate systems that have been investigated experimentally in laser shock-compressed beryllium, the Compton scattering spectrum provides the Fermi energy and hence the electron density. We show that forward scattering spectra that observe collective plasmon oscillations yield densities in agreement with Compton scattering. In addition, electron temperatures inferred from the dispersion of the plasmon feature are consistent with the ion temperature sensitive elastic scattering feature. Hence, theoretical models of the static ion-ion structure factor and consequently the equation of state of dense matter can be directly tested.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Contaminant point source localization error estimates as functions of data quantity and model quality

We develop empirically-grounded error envelopes for localization of a point contamination release event in the saturated zone of a previously uncharacterized heterogeneous aquifer into which a number of plume-intercepting wells have been drilled. We assume that flow direction in the aquifer is known exactly and velocity is known to within a factor of two of our best guess from well observations prior to source identification. Other aquifer and source parameters must be estimated by interpretation of well breakthrough data via the advection-dispersion equation. We employ high performance computing to generate numerous random realizations of aquifer parameters and well locations, simulate well breakthrough data, and then employ unsupervised machine optimization techniques to estimate the most likely spatial (or space-time) location of the source. Tabulating the accuracy of these estimates from the multiple realizations, we relate the size of 90% and 95% confidence envelopes to the data quantity (number of wells) and model quality (fidelity of ADE interpretation model to actual concentrations in a heterogeneous aquifer with channelized flow). We find that for purely spatial localization of the contaminant source, increased data quantities can make up for reduced model quality. For space-time localization, we find similar qualitative behavior, but significantly degraded spatial localization reliability and less improvement from extra data collection. Since the space-time source localization problem is much more challenging, we also tried a multiple-initial-guess optimization strategy. Furthermore, this greatly enhanced performance, but gains from additional data collection remained limited.

54 ENVIRONMENTAL SCIENCES↗

Can a disk model explain Beta Lyrae?

The Beta Lyrae eclipsing binary system is interpreted in terms of a model where the primary, i.e., spectroscopically observable, B star transfers matter to a 'canonical' accretion disk surrounding a 'gainer' which may be a rather ordinary B0.5 V star. Disk models are calculated using the TLUSDISK program. Most of the optical radiation from the secondary object is found to come from the disk rim, provided that the rate of mass transfer is not much lower than about 0.0001 solar mass per year. The rather high disk rim then introduces severe constraints on the orbital inclination; these constraints are discussed in detail. It is concluded that the disk plays an essentially passive role, and the observable radiation from its face is insignificant. A model in which most ultraviolet radiation would be coming from the face of the disk is found unacceptable both because of these constraints and because it would require an unrealistically small radius for the accreting star. It is proposed that the bulk of the ultraviolet radiation comes from a small unocculted segment of the gainer.

Hubeny, Ivan↗