GPU Accelerated Multiple-Relaxation-Time Lattice Boltzmann Simulation of Convective Flows in a Porous Media

• Main Authors: Md Mamun Molla, Md Jahidul Haque, Md Amirul Islam Khan, Suvash C. Saha
• Format: Article
• Language: English
• Published: Frontiers Media S.A. 2018-11-01
• Series: Frontiers in Mechanical Engineering
• Subjects: GPU parallel computing; CUDA C; porous media; Brinkman–Forchheimer model; MRT-LBM; natural and mixed convection
• Online Access: https://www.frontiersin.org/article/10.3389/fmech.2018.00015/full

A two-dimensional (2D) multiple-relaxation-time (MRT)-lattice Boltzmann method (LBM) is used for porous media with the Brinkman–Forchheimer extended Darcy model to investigate the natural and mixed convection flows in a square cavity. This Brinkman–Forchheimer model is directly applied by using the forcing moments as a source term. A Tesla K40 NVIDIA graphics card has been used for the present graphics processing unit (GPU) parallel computing via compute unified device architecture (CUDA) C platform. The numerical results are presented in terms of velocity, temperature, streamlines, isotherms, and local and average Nusselt numbers. For the wide range of Rayleigh numbers, (Ra = 103 to 1010), Reynolds numbers, Darcy numbers, and porosities, the average Nusselt number is compared with the available results computed by finite element method (FEM) and single-relaxation-time (SRT) lattice Boltzmann method-LBM and, showing great compliance. The results are also validated with the previous experimental results. The simulations speed up to a maximum of 144x using CUDA C in GPU compared with the time of FORTRAN 90 code using a single core CPU simulation.