Graphics processing unit accelerated high-order moment-encoded lattice Boltzmann method for liquid–vapor phase-change simulations

Although visually appealing, the simulation of the liquid–vapor phase-change flows faces numerous challenges in accurately reproducing complex bubble dynamics (nucleation, growth, coalescence, departure and breakup, etc.). Recently, a cascaded lattice Boltzmann method (CLBM) was proposed to solve the liquid–vapor phase-change process, which is based on the three-dimensional nineteen-velocity (D3Q19) lattice. Although this method has faithfully reproduced complete boiling regimes, it systematically comes at the price of a substantial memory requirement: double D3Q19 discretizations for the LB equation requires more variables per grid node and makes it hard to run higher resolution simulations on a laptop or workstation. To open up kinetic-based liquid-vapor phase-change simulation to industrial applications, we employ the recent high-order moment-encoded LBM (HOME-LBM) framework for solving the pseudopotential multiphase model with a memory footprint reduction by a factor of two while coupling with the temperature solved by a three-times faster lightweight finite difference method through a non-ideal equation of state. We show that our HOME-LBM phase-change solver can get similar accuracy and lower computing costs on numerical simulations of several canonical problems than CLBM. It is verified that the resulting solver can simulate both the isothermal multiphase flow and the liquid–vapor phase-change process. In addition, we present more challenging examples to demonstrate our solver's visual complexity, accuracy, and realism.