Layer-by-Layer Partitioning of Finite Element Meshes for Multicore Architectures

Novikov, A.K.; Piminova, N.K.; Kopysov, S.P.; Sagdeeva, Yulia

doi:10.1007/978-3-319-55669-7_9

Cited by 1 publication

(1 citation statement)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A substructural tree diagram and substructural sets were established in such a way that the omitted substructures were sequentially condensed into the retained substructure to construct the reduced model. A layer-by-layer partitioning of finite element meshes for multicore architecture was presented by Novikov et al (2017) [15] using a neighborhood criterion to partition the mesh into layers and combining them into blocks and assigning them into different parallel processors. Badia and Verdugo (2018) [1] investigated the use of domain decomposition preconditioners for unfitted finite element methods such as extended finite element method defining the coarse degrees of freedom in the definition of the preconditioner.…”

Section: Introductionmentioning

confidence: 99%

Reducing Computational and Memory Cost of Substructuring Technique in Finite Element Models

Kadkhodaee

Moslemi

2018

NMCE

View full text Add to dashboard Cite

Substructuring in the finite element method is a technique that reduces computational cost and memory usage for analysis of complex structures. The efficiency of this technique depends on the number of substructures in different problems. Some subdivisions increase computational cost, but require little memory usage and vice versa. In the present study, the cost functions of computations and memory usage are extracted in terms of number of subdivisions and optimized mathematically. The results are presented in the form of tables which recommend the proper substructuring for different number of elements. A combined case is also considered which investigates balanced reduction of computational and memory cost for 2D problems. Several numerical examples are analyzed numerically to demonstrate the abilities and efficiency of the proposed computational algorithm for structured and unstructured mesh.

show abstract

Section: Introductionmentioning

confidence: 99%