Contrast of RELAP5/MOD3.2 Results From Different Computing Platforms (NUREG/IA-0157)
On this page:
Download complete document
This page includes links to files in non-HTML format. See Plugins, Viewers, and Other Tools for more information.
Date Published: April 1999
A. López, CNA
J. M. Sierra, CDI
|Central Nuclear de Almaraz
C/Claudio Coello, 123
|Control Data Ibérica, SA
Paseo de La Castellana, 93
Prepared as part of:
The Agreement on Research Participation and Technical Exchange
under the International Code Application and Maintenance Program (CAMP)
Office of Nuclear Regulatory Research
U.S. Nuclear Regulatory Commission
Washington, DC 20555-0001
This work has focused in detecting (and, where possible, correcting) problems ocurred during the installation of Relap5/Mod 3.2 on different computing environments that encompass all major workstations (namely: DEC, HP, IBM, SGI and SUN), besides that of CONVEX, all of which run under the UNIX operating system. Once that the production version - that obtained with the supplied installation procedure - of Relap5 was obtained for each platform, it was run a common sample case and the results produced were analysed. As a result of that analysis, substantial differences were observed in the magnitude of the selected components (i.e., pressure, mass flow and temperature in the intact and broken loop, etc.), establishing as a possible cause for the discrepancies: the level of optimization adopted for compilation, the chosen maximum time step size, and the inherent precision of the platforms under consideration (32 or 64 bit architectures, compliance with IEEE-754, etc.). Those discrepancies, however, lay within acceptable limits when the results of the different machines are produced by non-optimised versions of Relap5 (only for some machines) and using a maximum time step size of about 0.125 s.. Hence, the recommendations are two-fold: to users, about the need to run the test (both, with and without optimisation) whenever a new version of Relap5, operating system and/or compiler level is installed, and to urge their system software providers to supply robust and reliable products, and to developers, about the need to deliver a code which is compliant with ANSI standards for Fortran, to establish a more rigid criteria to evaluate the maximum time step size and the convenience to put in place a test, or set of them, to be used as an acceptance/rejection criteria of the Relap5 version produced.
This work has been performed by Central Nuclear de Almaraz (CNA), a member of Unidad Eléctrica, S.A. (UNESA) that participates in the Code Application and Maintenance Project (CAMP), and is part of the "in kind" contribution of Spain to the CAMP Project.
Being the CAMP project of an international and multidisciplinary scope, this work reinforzes that nature for it is made out of contribution from the following groups: the Spanish PWR Plant Operators, the Spanish Nuclear Safety Council (CSN), the Turkish Atomic Energy Authority (TAEA) and the High Technical School of Mines of Madrid (ETSIMM). Their effort and suggestions are gathered here along with our deep gratitude.