Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Computational Science, Engineering & Technology Series
PARALLEL, DISTRIBUTED AND GRID COMPUTING FOR ENGINEERING
Edited by: B.H.V. Topping, P. Iványi
Parallel Distributed Seismic Analysis of an Assembled Nuclear Power Plant
Center for Computational Science and e-Systems, Japan Atomic Energy Agency, Tokyo, Japan
T. Yamada, "Parallel Distributed Seismic Analysis of an Assembled Nuclear Power Plant", in B.H.V. Topping, P. Iványi, (Editors), "Parallel, Distributed and Grid Computing for Engineering", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 20, pp 439-454, 2009. doi:10.4203/csets.21.20
Keywords: seismic simulation, nuclear power plant, parallel computing, balancing domain decomposition method.
An overview of the seismic simulation of nuclear power plants and reduction of parallel computation cost are discussed in this paper. The importance of nuclear power generation for ensuring national energy security is widely acknowledged. The seismic safety of nuclear power plants has attracted considerable attention after the introduction of new regulatory guidelines for the seismic design of nuclear power plants in Japan and also after several recent strong earthquakes .
A nuclear power plant is generally a huge and complex facility assembled from several tens of millions of mechanical components. The safety requirement of nuclear power plant is extremely high, enough to perform real experiments to evaluate the structural integrity of mechanical components. However, such experiments are performed on independent sets of components because the ability of experimental facilities are limited. Hence, a numerical simulation, which can treat whole assembled plant, attracts more and more attention with the growing computational technologies.
Such whole plant simulation is quite challenging because it requires much computation resources, huge data handling and rigorous modelling efforts. To simulate such complex assembled structures, a hierarchical simulation strategy was introduced and implemented as a simulation framework. In hierarchical simulation strategy, large mechanical components such as pressure vessels and small ones such as pipes are loosely coupled (one-way coupling) because the displacement transferred from the small component to the large one approaches zero as the size difference increases [2,3]. The computation cost for seismic simulation of nuclear power plant is drastically reduced with this strategy.
The balancing domain decomposition method  is also investigated to reduce the computation cost of large mechanical components. The balancing domain decomposition method is a combination of primal substructuring approach, Neumann-Neumann preconditioning and coarse grid correction and the computation cost depends greatly on the number of subdomains. A prediction curve of parallel computation cost of balancing domain decomposition method is proposed and optimal number of subdomains is estimated.
Numerical validation of an optimal number of subdomains is performed with a component of an actual nuclear power plant, High Temperature engineering Test Reactor (HTTR) , which is located at the O-arai research and development center of Japan Atomic Energy Agency. The component contains 50 millions degrees of freedom and successfully and efficiently analyzed on 128 and 512 processors on SGI Altix3700B2. The measured computation time with various numbers of subdomains shows good agreement with the estimated optimal number of subdomains.
purchase the full-text of this chapter (price £20)