High-Performance Computing, Parallel Methods and Algorithms in Geophysical Problems
To improve the accuracy of numerical solution of problems without increasing the number of nodes, one can concentrate mesh nodes around the areas with large solution variation. A discrete stochastic method for moving structured adaptive mesh generation in 3D space based on self-organizing maps is proposed. The method is attractive, first of all, from the point of parallelization and convenience of the use: the underlying algorithms have an inherent parallelism and use simple data structures for storing mesh nodes. Unlike classical methods, which are highly complex and, thus, are not studied well, the proposed method allows us to generate meshes of any dimensionality and does not impose constrains on initial and boundary data.
Note. Abstracts are published in author's edition
Mail to Webmaster |
[Home] [Conferences] |
© 1996-2008, Siberian Branch of Russian Academy of Sciences, Novosibirsk
Last update: 06-Jul-2012 (11:52:47)