Digital Library / STP / STP1398-EB / STP13638S



Accuracy and Parallel Performance of PENTRAN™ Using the VENUS-3 Benchmark Experiment

Haghighat, A
Professor of Nuclear Engineering,Penn State University,PA,

Abderrahim, HA
Manager of Nuclear Fuel,SCK•CEN,

Sjoden, GE
Associate Professor,USAF AcademyUSAF Academy,CO,


Pages: 8    Published: Jan 2001


Download this paper for $25 PDF (296K)          View License Agreement
        Click here to download the complete source publication for $273 PDF (17M)


Source: STP1398-EB


Abstract

We use the VENUS-3 experimental benchmark to examine the accuracy and parallel performance of PENTRAN™ (Parallel Environment Neutral-particle TRANsport), a three- dimensional parallel Sn code. VENUS-3 with its partial length assemblies is a 3-D benchmark with 370 measurements (58Ni(n,p), 115In(n,n'), and 27Al(n,α) dosimeters). PENTRAN results are very close to experimental results; more than 95% of the Calculated-to-experimental (C/E) values are within ±10%. The remaining 5% of the values are within ±10% and ±15%. The PENTRAN calculation (with 84748 spatial meshes, 80 directions, and 26 energy groups) is performed on an IBM SP2 parallel computer using different numbers of processors and different parallel algorithms. It is demonstrated that PENTRAN requires less than 1.4 hrs wall-clock time on32 processors, and it exhibits an excellent parallel performance with a scalability of ̃90%.


Keywords:
VENUS-3, PENTRAN™, parallel processing, benchmark, Sn transport method

Paper ID: STP13638S
Committee/Subcommittee: E10.08
DOI: 10.1520/STP13638S
CrossRef ASTM International is a member of CrossRef.