A Python-based Phase-field MRT Lattice Boltzmann Model for Three-component Flows

This study presents the development of a Python-based phase-field MRT-LBM model capable of handling high-density ratio flows up to O(۱۰۰۰) while ensuring computational stability. The model is implemented using PyTorch, leveraging GPU parallel computing to enhance performance. A comparative analysis between single-relaxation-time (SRT) and multi-relaxation-time (MRT) formulations demonstrates that the MRT model more effectively recovers the governing equations and maintains numerical stability, making it a robust choice for multiphase flow simulations. To assess computational efficiency, performance benchmarking was conducted using RTX ۳۰۷۰ and RTX ۴۰۷۰ GPUs, achieving ۷.۳۲× and ۱۶.۳۰× process speed, respectively, compared to CPU-based implementations. These results underscore the potential of PyTorch as a viable alternative to low-level programming languages (e.g., FORTRAN, C++) for computational fluid dynamics (CFD) simulations, particularly in lattice Boltzmann methods (LBM). The findings highlight the feasibility of high-performance, GPU-accelerated numerical solvers in advancing CFD applications while maintaining ease of implementation through high-level programming frameworks.

کلیدواژه ها:

Lattice Boltzmann Method ، Multi-Component flows ، PyTorch