Parallel algorithms for moving boundary problems by local remeshing
| Autor: | Xuan Zhou, Jianjun Chen, Lisha He, Yao Zheng, Jianjing Zheng, Zhoufang Xiao |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
business.industry
Computer science Computation General Engineering Parallel algorithm Boundary (topology) Domain decomposition methods Parallel computing Computational fluid dynamics Solver Rigid body 01 natural sciences Bottleneck 010305 fluids & plasmas Computer Science Applications 010101 applied mathematics Computational Theory and Mathematics 0103 physical sciences 0101 mathematics business Software |
| Zdroj: | Engineering Computations. 36:2887-2910 |
| ISSN: | 0264-4401 |
| DOI: | 10.1108/ec-11-2018-0545 |
| Popis: | Purpose The purpose of this paper is to develop parallel algorithms for moving boundary simulations by local remeshing and compose them to a fully parallel simulation cycle for the solution of problems with engineering interests. Design/methodology/approach The moving boundary problems are solved by unsteady flow computations coupled with six-degrees-of-freedom equations of rigid body motion. Parallel algorithms are developed for both computational fluid dynamics (CFD) solution and grid deformation steps. Meanwhile, a novel approach is developed for the parallelization of the local remeshing step. It inputs a distributed mesh after deformation, then marks low-quality elements to be deleted on the respective processors. After that, a parallel domain decomposition approach is used to repartition the hole mesh and then to redistribute the resulting sub-meshes onto all available processors. Then remesh individual sub-holes in parallel. Finally, the element redistribution is rebalanced. Findings If the CFD solver is parallelized while the remaining steps are executed in sequential, the performance bottleneck of such a simulation cycle is observed when the simulation of large-scale problem is executed. The developed parallel simulation cycle, in which all of time-consuming steps have been efficiently parallelized, could overcome these bottlenecks, in terms of both memory consumption and computing efficiency. Originality/value A fully parallel approach for moving boundary simulations by local remeshing is developed to solve large-scale problems. In the algorithm level, a novel parallel local remeshing algorithm is present. It repartitions distributed hole elements evenly onto all available processors and ensures the generation of a well-shaped inter-hole boundary always. Therefore, the subsequent remeshing step can fix the inter-hole boundary involves no communications. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|