Scalable parallel finite volume lattice Boltzmann method for thermal incompressible flows on unstructured grids

•A parallel cell-centered finite volume thermal lattice Boltzmann method for two-dimensional unstructured grids is proposed to simulate thermal incompressible flows.•The proposed algorithm scales up to 6000 processor cores with 96.50% parallel efficiency for a case with over 5 billion unstructued gr...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:International journal of heat and mass transfer Ročník 160; s. 120156
Hlavní autoři: Xu, Lei, Chen, Rongliang
Médium: Journal Article
Jazyk:angličtina
Vydáno: Oxford Elsevier Ltd 01.10.2020
Elsevier BV
Témata:
Algorithms
Computational fluid dynamics
Computer simulation
Distribution functions
Finite volume method
Fluid flow
Fluxes
Free convection
Incompressible flow
Lattice Boltzmann equation
Mathematical analysis
Mathematical models
Microprocessors
Parallel algorithm
Thermal incompressible flows
Thermal simulation
Unstructured grids
Unstructured grids (mathematics)
Thermal incompressible flows
Parallel algorithm
Unstructured grids
68W10
Finite volume method
Lattice Boltzmann equation
35Q20
76M12
ISSN:0017-9310, 1879-2189
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:•A parallel cell-centered finite volume thermal lattice Boltzmann method for two-dimensional unstructured grids is proposed to simulate thermal incompressible flows.•The proposed algorithm scales up to 6000 processor cores with 96.50% parallel efficiency for a case with over 5 billion unstructued grid cells. A parallel cell-centered finite volume thermal lattice Boltzmann method (FV-TLBM) for two-dimensional unstructured grids is proposed to simulate thermal incompressible flows. The governing equations for the velocity and temperature fields are solved using the D2Q9 lattice model in the lattice Boltzmann method and the cell-centered finite volume method. In this paper, the advective fluxes in the governing equations are evaluated by a low-diffusion Roe scheme, and the gradients of the particle distribution functions are computed with the least-square approach. To simulate large-scale complex flows within a reasonable time, a parallel algorithm for the FV-TLBM on unstructured grids is devised. The present method is validated by numerical simulations of the natural convection in a square cavity and natural convection in a concentric annulus at different Rayleigh numbers. The results obtained by the proposed method agree well with previous studies. The parallel performance results show that the proposed algorithm has considerable scalability and the parallel efficiency is as high as 96.50% for a case with over 5 billion grid cells using 6000 processor cores.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0017-9310
1879-2189
DOI:10.1016/j.ijheatmasstransfer.2020.120156