skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Compression of magnetohydrodynamic simulation data using singular value decomposition

Abstract

Numerical calculations of magnetic and flow fields in magnetohydrodynamic (MHD) simulations can result in extensive data sets. Particle-based calculations in these MHD fields, needed to provide closure relations for the MHD equations, will require communication of this data to multiple processors and rapid interpolation at numerous particle orbit positions. To facilitate this analysis it is advantageous to compress the data using singular value decomposition (SVD, or principal orthogonal decomposition, POD) methods. As an example of the compression technique, SVD is applied to magnetic field data arising from a dynamic nonlinear MHD code. The performance of the SVD compression algorithm is analyzed by calculating Poincare plots for electron orbits in a three-dimensional magnetic field and comparing the results with uncompressed data.

Authors:
 [1];  [2];  [1];  [1]
  1. Oak Ridge National Laboratory, Oak Ridge, TN 37831 (United States)
  2. Oak Ridge National Laboratory, Oak Ridge, TN 37831 (United States). E-mail: hirshmansp@ornl.gov
Publication Date:
OSTI Identifier:
20991563
Resource Type:
Journal Article
Resource Relation:
Journal Name: Journal of Computational Physics; Journal Volume: 222; Journal Issue: 1; Other Information: DOI: 10.1016/j.jcp.2006.07.022; PII: S0021-9991(06)00355-X; Copyright (c) 2006 Elsevier Science B.V., Amsterdam, The Netherlands, All rights reserved; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; ALGORITHMS; APPROXIMATIONS; COMPUTERIZED SIMULATION; DATA TRANSMISSION; EQUATIONS; INTERPOLATION; MAGNETIC FIELDS; MAGNETOHYDRODYNAMICS; NONLINEAR PROBLEMS; ORBITS; PERFORMANCE; THREE-DIMENSIONAL CALCULATIONS

Citation Formats

Castillo Negrete, D. del, Hirshman, S.P., Spong, D.A., and D'Azevedo, E.F. Compression of magnetohydrodynamic simulation data using singular value decomposition. United States: N. p., 2007. Web. doi:10.1016/j.jcp.2006.07.022.
Castillo Negrete, D. del, Hirshman, S.P., Spong, D.A., & D'Azevedo, E.F. Compression of magnetohydrodynamic simulation data using singular value decomposition. United States. doi:10.1016/j.jcp.2006.07.022.
Castillo Negrete, D. del, Hirshman, S.P., Spong, D.A., and D'Azevedo, E.F. Thu . "Compression of magnetohydrodynamic simulation data using singular value decomposition". United States. doi:10.1016/j.jcp.2006.07.022.
@article{osti_20991563,
title = {Compression of magnetohydrodynamic simulation data using singular value decomposition},
author = {Castillo Negrete, D. del and Hirshman, S.P. and Spong, D.A. and D'Azevedo, E.F.},
abstractNote = {Numerical calculations of magnetic and flow fields in magnetohydrodynamic (MHD) simulations can result in extensive data sets. Particle-based calculations in these MHD fields, needed to provide closure relations for the MHD equations, will require communication of this data to multiple processors and rapid interpolation at numerous particle orbit positions. To facilitate this analysis it is advantageous to compress the data using singular value decomposition (SVD, or principal orthogonal decomposition, POD) methods. As an example of the compression technique, SVD is applied to magnetic field data arising from a dynamic nonlinear MHD code. The performance of the SVD compression algorithm is analyzed by calculating Poincare plots for electron orbits in a three-dimensional magnetic field and comparing the results with uncompressed data.},
doi = {10.1016/j.jcp.2006.07.022},
journal = {Journal of Computational Physics},
number = 1,
volume = 222,
place = {United States},
year = {Thu Mar 01 00:00:00 EST 2007},
month = {Thu Mar 01 00:00:00 EST 2007}
}
  • Numerical calculations of magnetic and flow fields in magnetohydrodynamic (MHD) simulations can result in extensive data sets. Particle-based calculations in these MHD fields, needed to provide closure relations for the MHD equations, will require communication of this data to multiple processors and rapid interpolation at numerous particle orbit positions. To facilitate this analysis it is advantageous to compress the data using singular value decomposition (SVD, or principal orthogonal decomposition, POD) methods. As an example of the compression technique, SVD is applied to magnetic field data arising from a dynamic nonlinear MHD code. The performance of the SVD compression algorithm ismore » analyzed by calculating Poincare plots for electron orbits in a three-dimensional magnetic field and comparing the results with uncompressed data.« less
  • Higher order singular value decomposition (HOSVD) is explored as a tool for analyzing and compressing gyrokinetic data. An efficient numerical implementation of an HOSVD algorithm is described. HOSVD is used to analyze the full six-dimensional (three spatial, two velocity space, and time dimensions) gyrocenter distribution function from gyrokinetic simulations of ion temperature gradient, electron temperature gradient, and trapped electron mode driven turbulence. The HOSVD eigenvalues for the velocity space coordinates decay very rapidly, indicating that only a few structures in velocity space can capture the most important dynamics. In almost all of the cases studied, HOSVD extracts parallel velocity spacemore » structures which are very similar to orthogonal polynomials. HOSVD is also used to compress gyrokinetic datasets, an application in which it is shown to significantly outperform the more commonly used singular value decomposition. It is shown that the effectiveness of the HOSVD compression improves as the dimensionality of the dataset increases. (C) 2012 Elsevier Inc. All rights reserved.« less
  • A modified truncated singular value decomposition (MTSVD) is employed to unfold proton recoil pulse-height spectra into neutron energy spectra, using experimentally measured response functions. To illustrate the method, spectra from [sup 252]Cf and [sup 239]PuBe sources are unfolded. The relative error, defined in terms of the 1-norm, using the MTSVD method is found to be approximately half that of the truncated singular value decomposition for the [sup 252]Cf spectra. Relative errors for the [sup 239]PuBe spectra were approximately equal for the two methods. The method is limited by the precision of the measurement of the response functions and the pulse-heightmore » spectra. More precise measurements would allow the use of larger truncation parameters and are likely to result in more accurate reconstructed neutron spectra. The MTSVD method is particularly suited to real-time on-line unfolding of spectra.« less
  • Error estimates of time activity curves are necessary to obtain efficient estimates of dynamic of dynamic cardiac SPECT kinetic parameters which are determined using weighted least squares fitting that incorporates these error estimates. In cardiac SPECT, iterative algorithms are used to obtain attenuation corrected reconstructions, and the use of an iterative algorithm makes it difficult to estimate the errors of the estimated reconstruction. An alternate approach is to estimate the reconstruction by solving the system of normal equations using singular value decomposition. This method was applied to dynamic data acquired from a canine study. A canine was injected with 25more » mCi of Tc-99m-teboroxime and was imaged using a three-detector SPECT system (Picker PRISM 3000). Sequential 5 sec tomographic acquisitions were acquired for 15 min, allowing both the wash-in and wash-out of teboroxime to be measured. The projection data were reconstructed into 64x64 transaxial slices for each 5 sec acquisition using singular value decomposition to calculate the reconstructed estimate, the variance of the estimate, and the covariance between tissue and blood regions-of-interest. One 4096x4096 singular value decomposition was obtained in 71 hours using a 40 mHz Supper SPARC processor. Tissue and blood time-activity curves were generated from the attenuation corrected transaxial reconstructions. The blood activity curve was generated from a region drawn inside the left ventricle. A two-compartment model was fit to the blood and tissue activity curves to give weighted least squares estimates of blood volume fraction and wash-in and wash-out rate constants specifying teboroxime kinetics for regions of the left ventricular myocardium. As expected the weighted least squares estimates of the kinetic parameters had smaller variances than the unweighted estimates, thus demonstrating more efficient parameter estimation.« less
  • The authors describe a new attenuation correction method for fluorescent X-ray computed tomography (FXCT) applied to image nonradioactive contrast materials in vivo. The principle of the FXCT imaging is that of computed tomography of the first generation. Using monochromatized synchrotron radiation from the BLNE-5A bending-magnet beam line of Tristan Accumulation Ring in KEK, Japan, the authors studied phantoms with the FXCT method, and they succeeded in delineating a 4-mm-diameter channel filled with a 500 {micro}g I/ml iodine solution in a 20-mm-diameter acrylic cylindrical phantom. However, to detect smaller iodine concentrations, attenuation correction is needed. They present a correction method basedmore » on the equation representing the measurement process. The discretized equation system is solved by the least-squares method using the singular value decomposition. The attenuation correction method is applied to the projections by the Monte Carlo simulation and the experiment to confirm its effectiveness.« less