Computational & Technology Resources
an online resource for computational,
engineering & technology publications |
|
Civil-Comp Proceedings
ISSN 1759-3433 CCP: 94
PROCEEDINGS OF THE SEVENTH INTERNATIONAL CONFERENCE ON ENGINEERING COMPUTATIONAL TECHNOLOGY Edited by:
Paper 32
Preconditioning Techniques for Matrices arising in the Discretization of the Neutron Diffusion Equation in Hexagonal Geometry S. González-Pintor1, D. Ginestar2 and G. Verdú1
1Department of Chemical and Nuclear Engineering,
, "Preconditioning Techniques for Matrices arising in the Discretization of the Neutron Diffusion Equation in Hexagonal Geometry", in , (Editors), "Proceedings of the Seventh International Conference on Engineering Computational Technology", Civil-Comp Press, Stirlingshire, UK, Paper 32, 2010. doi:10.4203/ccp.94.32
Keywords: time dependent neutron diffusion equation, block iterative methods, block preconditioners.
Summary
To improve the safety of nuclear power
reactors it is necessary to develop fast and
accurate plant simulators. Under general assumptions, the
neutronic population inside a nuclear power reactor
can be modelled by the time dependent
neutron diffusion equation
in the approximation of two energy groups [1].
To analyze the behaviour of Vodo-Vodyanoi Energetichesky Reactor (VVER) nuclear power reactors it is
necessary to discretize this equation in a hexagonal mesh.
The spatial discretization selected consists of
a high order finite element method based on a triangular mesh
which assumes that the neutronic flux can be expanded
in terms of the modified Dubiner's polynomials [3,2].
Once this discretization
has been selected, the semidiscrete version
of the time dependent neutron diffusion equation is solved.
Since the ordinary differential equations resulting of the
discretization of diffusion equations are, in general, stiff,
implicit methods are necessary. We have used
a finite differences method, that needs to solve a
large system of linear equations for each time step.
Since the energy groups structure defines a natural
partition of the matrix of the system into different
blocks with good properties different methods
to solve the linears systems are studied that
use this block structure. In this way
we have studied the performance of block iterative methods
for the solutions of these systems of equations such as the
block Jacobi and the block Gauss-Seidel methods combined
with different variational acceleration techniques.
Also we have proposed an inexpensive preconditioner for the system.
The methods have been tested for the matrices obtained for
a two-dimensional transient benchmark problem. For this case, we have
observed that the most efficient method is the
preconditioned Gauss-Seidel method. The
performance of the method does not have a strong dependence
on the variational acceleration technique used.
References
purchase the full-text of this paper (price £20)
go to the previous paper |
|