Detection and Correction of Silent Data Corruption for Large-Scale High-Performance Computing.
Abstract not provided.
SEARCH · Engineering Papers
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.
Abstract not provided.
Sensor data validity is extremely important in a number of applications, particularly building technologies where collected data are used to determine performance. An example of this is Oak Ridge National Laboratory s ZEBRAlliance research project, which consists of four single-family homes located in Oak Ridge, TN. The homes are outfitted with a total of 1,218 sensors to determine the performance of a variety of different technologies integrated within each home. Issues arise with such a large amount of sensors, such as missing or corrupt data. This paper aims to eliminate these problems using: (1) Kalman filtering and (2) linear prediction filtering techniques. Five types of data are the focus of this paper: (1) temperature; (2) humidity; (3) energy consumption; (4) pressure; and (5) airflow. Simulations show the Kalman filtering method performed best in predicting temperature, humidity, pressure, and airflow data, while the linear prediction filtering method performed best with energy consumption data.
Master’s Thesis: University of North Dakota
Abstract not provided.
A formula has been developed to correct laser ranging data for the effects of horizontal refractivity gradients. The formula requires the values of the horizontal pressure and temperature gradients at the laser site. The accuracy of this technique was evaluated by comparing the formula with range corrections which were computed by ray tracing.
Summary In observational epidemiological studies, the exposure that is received by an individual often cannot be precisely observed, resulting in measurement error, and a common approach to dealing with measurement error is regression calibration (RC). Use of RC, which requires assumptions about the distribution of unknown error-free (true) variables, leads to concern about the possibility of bias due to misspecification of that distribution. The simulation–extrapolation (SIMEX) method, in contrast, does not require a distributional assumption. However, analyses of large cohorts may be performed by using grouped or person-year data, and application of SIMEX to grouped data is not straightforward, particularly when there is a mixture of classical and Berkson measurement errors. We compared RC and SIMEX with grouped data analyses to assess robustness of the RC method to misspecification of the true dose distribution. We also applied SIMEX assuming mixtures of classical and Berkson errors and compared the results with those obtained by using RC for classical error only. SIMEX had less bias than RC and performed well regardless of the true dose distribution, whereas RC based on a misspecified true dose distribution showed greater bias than when based on the correctly specified true dose distribution.
Atmospheric correction of imaging spectroscopy data is required for quantitative analysis. Different models were proposed for atmospheric correction of these data. LOWTRAN-7 is a low-resolution model and computer code for predicting atmospheric transmittance and background radiance from 0 to 50,00 cm(sup -1) which was developed by the Air Force Geophysics Laboratory. The Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data used are radiometrically calibrated and include the 28 Sep. 1989 Providence Fan flight line segment 07, California. It includes a dark gravel surface defined as a calibration site by the Geologic Remote Sensing Field Experiment (GRSFE). Several ground measurements of portable spectrometer DAEDALUS AA440 Spectrafax were taken during the GRSFE, July 1989 field campaign. Comparisons of the LOWTRAN-7 corrected AVIRIS data with the ground spectrometer measurement were made.
Powered flight maneuver data used to reduce state vector uncertainties for lunar satellite orbit differential correction
A concept for performing radiometric correction of multispectral image data onboard a spacecraft at very high data rates is presented and demonstrated. This concept utilized a lookup table approach, implemented in hardware, to convert the raw sensor data into the desired corrected output data. The digital lookup table memory was interfaced to a microprocessor to allow the data correction function to be completely programmable. Sensor data was processed with this approach at rates equal to the access time of the lookup table memory. This concept offers flexible high speed data processing for a wide range of applications and will benefit from the continuing improvements in performance of digital memories.
The close environment of the central supermassive black hole of our Galaxy has been studied thoroughly for decades in order to shed light on the behavior of the central regions of galaxies in general and of active galaxies in particular. The Galactic center (GC) has shown a wealth of structures on different scales with a complicated mixture of early- and late-type stars, ionized and molecular gas, dust, and winds. Here we aim to study the distribution of water-ices and hydrocarbons in the central parsec, as well as along the line of sight. This study is made possible thanks to L-band spectroscopy. This spectral band, from 2.8 to 4.2 μm, hosts important signatures of the circumstellar medium and interstellar dense and diffuse media among which deep absorption features are attributed to water-ices and hydrocarbons. We observed the GC in the L band of the ISAAC spectrograph located on the UT1/VLT ESO telescope. By mapping the central half parsec using 27 slit positions, we were able to build the first data cube of the region in this wavelength domain. Thanks to a calibrator spectrum of the foreground extinction in the L band derived in a previous paper, we corrected our data cube for the line-of-sight extinction and validated our calibrator spectrum. The data show that a residual absorption due to water-ices and hydrocarbons is present in the corrected data cube. This suggests that the features are produced in the local environment of the GC, implying very low temperatures well below 80 K. This is in agreement with our finding of local CO ices in the central parsec described in Moultaka et al.
For over 30 years, astronauts in Earth orbit or on missions to the moon have been exposed to space radiation comprised of high-energy protons and heavy ions and secondary particles produced in collisions with spacecraft and tissue. Large uncertainties exist in the projection of risks of late effects from space radiation such as cancer and cataracts due to the paucity [corrected] of epidemiological data. Here we present epidemiological [corrected] data linking an increased risk of cataracts for astronauts with higher lens doses (>8 mSv) of space radiation relative to other astronauts with lower lens doses (<8 mSv). Our study uses historical data for cataract incidence in the 295 astronauts participating in NASA's Longitudinal Study of Astronaut Health (LSAH) and individual occupational radiation exposure data. These results, while preliminary because of the use of subjective scoring methods, suggest that relatively low doses of space radiation may predispose crew to [corrected] an increased incidence and early appearance of cataracts.
A flight test evaluation conducted at the NASA Ames Research Center of precision-code differential GPS for terminal approach and landing is described. Differential corrections to satellite range and range-rate measurements are computed in a stationary ground reference system, uplinked to the test aircraft, and employed to correct corresponding airborne satellite data. Differentially corrected range and range-rate data, combined with inertial data, are then utilized with satellite position and velocity to compute aircraft position and velocity employing an eight-state extended Kalman filter.
Several 3.8 cm radar frames were calibrated empirically by histogram fitting, because no instrument background data is available. Then data were corrected for geometric distortion by: (1) redetermining position of individual frames using most accurate recent lunar ephemerides; (2) reprojecting frames into simple cylindrical map projection; (3) using most recent catalog of lunar craters to determine the exact positions of features identifiable on radar frames; and (4) correcting for apparent distortion (misplacement of features in frames) by resampling using a different bilinear interpolation derived for each of the parallelopideds of the set defined for each frame. A hardcopy set of corrected frames was produced. Attempts to produce a mosaic of such corrected frames continue. The resulting mosaic can be used to show the systematic relationship between photographic thermal IR and radar data at different wavelengths in a region dominated by both mare and highland terrain.
With an eye to future experiments at the Large Hadron Collider (LHC), O(aa{sub s}) QCD corrections to the Drell-Yan process are calculated for vertex functions and gluon bremsstrahlung. Use is made of fully differential cross sections, so that the result obtained in the present study can readily be applied in analyzing experimental data (in correcting data from future experiments at LHC). It is shown both analytically and numerically that the results are independent of unphysical parameters, including the parameter separating the regions of soft and hard gluons and the quark mass. A numerical analysis of radiative effects was performed by means of the FORTRAN code READY with allowance for the experimental cuts used at the Compact Muon Solenoid (CMS) detector
LANDSAT frames were geometrically corrected and data sets from six different dates were overlaid to produce a 24 channel (six dates and four wavelength bands) data tape. Changes in the extent of the snowpack could be accurately and easily determined using a change detection technique on data which had previously been classified by the LARSYS software system. A second phase of the analysis involved determination of the relationship between spatial resolution or data sampling frequency and accuracy of measuring the area of the snowpack.
Solar heating of the relative humidity (RH) probe on Vaisala RS92 radiosondes results in a large dry bias in the upper troposphere. Two different algorithms (Miloshevich et al., 2009, MILO hereafter; and Wang et al., 2013, WANG hereafter) have been designed to account for this solar radiative dry bias (SRDB). These corrections are markedly different with MILO adding up to 40 % more moisture to the original radiosonde profile than WANG; however, the impact of the two algorithms varies with height. The accuracy of these two algorithms is evaluated using three different approaches: a comparison of precipitable water vapor (PWV), downwelling radiative closure with a surface-based microwave radiometer at a high-altitude site (5.3 km m.s.l.), and upwelling radiative closure with the space-based Atmospheric Infrared Sounder (AIRS). The PWV computed from the uncorrected and corrected RH data is compared against PWV retrieved from ground-based microwave radiometers at tropical, midlatitude, and arctic sites. Although MILO generally adds more moisture to the original radiosonde profile in the upper troposphere compared to WANG, both corrections yield similar changes to the PWV, and the corrected data agree well with the ground-based retrievals. The two closure activities – done for clear-sky scenes – use the radiative transfer models MonoRTM and LBLRTM to compute radiance from the radiosonde profiles to compare against spectral observations. Both WANG- and MILO-corrected RHs are statistically better than original RH in all cases except for the driest 30 % of cases in the downwelling experiment, where both algorithms add too much water vapor to the original profile. In the upwelling experiment, the RH correction applied by the WANG vs. MILO algorithm is statistically different above 10 km for the driest 30 % of cases and above 8 km for the moistest 30 % of cases, suggesting that the MILO correction performs better than the WANG in clear-sky scenes. Lastly, the cause of this statistical significance is likely explained by the fact the WANG correction also accounts for cloud cover – a condition not accounted for in the radiance closure experiments.
A description is given of the structure and use of a computer program that simulates the operation of a parallel processor simulation system. The program is part of an investigation to determine algorithms that are suitable for simulating continous systems on a parallel processor configuration. The simulator is designed to accurately simulate the problem-solving phase of a simulation study. Care has been taken to ensure the integrity and correctness of data exchanges and to correctly sequence periods of computation and periods of data exchange. It is pointed out that the functions performed during a problem-setup phase or a reset phase are not simulated. In particular, there is no attempt to simulate the downloading process that loads object code into the local, transfer, and mapping memories of processing elements or the memories of the run control processor and the system control processor. The main program of the simulator carries out some problem-setup functions of the system control processor in that it requests the user to enter values for simulation system parameters and problem parameters. The method by which these values are transferred to the other processors, however, is not simulated.
We correct the mistaken claim made in~\cite{Guzey:2005ec,Guzey:2006xi} that the minimal model of the dual parameterization of nucleon generalized parton distributions (GPDs) gives a good, essentially model-independent description of high-energy data on deeply virtual Compton scattering (DVCS). In the implementation of the dual parameterization in~\cite{Guzey:2005ec,Guzey:2006xi}, the numerical prefactor of two in front of the DVCS amplitude was missing. We show that the corrected minimal model of the dual parameterization significantly overestimates the HERA data (H1 and ZEUS) on the DVCS cross section.