Spectral bisection of adaptive finite element meshes for parallel processing

摘要

In the present paper the spectral bisection method is used for partitioning finite element (FE) meshes. For adaptive FE meshes it is beneficial to operate on the coarse basic mesh rather than bisecting the refined model. This requires the bisection of the weighted graph corresponding to the coarse meshes. In this paper, the spectral bisection method is improved to solve the weighted graph partitioning problem. The adaptive FE mesh is mapped into the bisection of the associated weighted graph. The partitioning is performed on the initial coarse mesh, and then mesh refinements is carried out. Since the number of elements in the refined mesh is several times that of the initial mesh, the bisection is performed with far less effort. Examples are included to illustrate performance of the proposed method. The results are compared to those of the subdomain approach.