Özet:

We developed a lattice Boltzmann Method code that is able to solve conjugate heat transfer problems on graphical processing unit. The solver uses Cartesian meshes of D2Q9 lattices and solves transient heat equation on both fluid and solid regions at once after obtaining the macroscopic quantities from the flow field. The newly developed thermal solver was constructed on the previously developed and verified flow solver. Both solvers were coded using parallel computing platform and programing model of NVIDIA, Compute Unified Device Architecture (CUDA). The thermal solver was tested first for the problems of conduction in solid, then conjugate heat transfer problems were solved using the fluid and thermal solvers in conjunction. The thermal solver was designed to run on both solid and fluid regions. Comparison of the obtained results with the analytical and the computational results available in the literature show that the developed solver is able to model conjugate heat transfer problem realistically. After verification of the solver, a conjugate heat transfer benchmark problem, heat and mass transfer through a suddenly expanded channel with finite thickness of wall was revisited. Solid-fluid interface temperature distributions were obtained for different values of parameters such as Prandtl and Reynolds numbers. These results hopefully contribute to clarify the mismatch between the results available in the literature. The main drawback of the LBM solver is its relatively longer computational time compared to its rivals because of its considerably dense mesh needs. In this study, this drawback is resolved by designing the solver to be able to run on GPU.

Anahtar Kelimeler: