SAMPSON Parallel Computation for Sensitivity Analysis of TEPCO's Fukushima Daiichi Nuclear Power Plant Accident

¹ The Institute of Applied Energy, Shimbashi SY Bldg. 1-14-2 Nishi-Shimbashi, Minato-ku, Tokyo, 105-0003 Japan
² Tokyo Institute of Technology, 2-12-1-W8-33 Oookayama, Meguro-ku, Tokyo, 152-8550 Japan
³ AICS, RIKEN, 7-1-26 Minatojima-minami-machi, Chuo-ku, Kobe, Hyogo, 650-0047 Japan
⁴ INRIA-UIUC Join Laboratory for Petascale Computing, 1205 W. Clark St., Urbana, Illinois, 61801 USA

^* E-mail: mpellegrini@iae.or.jp, leobago@matsulab.is.titech.ac.jp

Abstract

On March 11^th 2011 a high magnitude earthquake and consequent tsunami struck the east coast of Japan, resulting in a nuclear accident unprecedented in time and extents. After scram started at all power stations affected by the earthquake, diesel generators began operation as designed until tsunami waves reached the power plants located on the east coast. This had a catastrophic impact on the availability of plant safety systems at TEPCO's Fukushima Daiichi, leading to the condition of station black-out from unit 1 to 3. In this article the accident scenario is studied with the SAMPSON code. SAMPSON is a severe accident computer code composed of hierarchical modules to account for the diverse physics involved in the various phases of the accident evolution. A preliminary parallelization analysis of the code was performed using state-of-the-art tools and we demonstrate how this work can be beneficial to the nuclear safety analysis. This paper shows that inter-module parallelization can reduce the time to solution by more than 20%. Furthermore, the parallel code was applied to a sensitivity study for the alternative water injection into TEPCO's Fukushima Daiichi unit 3. Results show that the core melting progression is extremely sensitive to the amount and timing of water injection, resulting in a high probability of partial core melting for unit 3.