دانشگاه: State University of New York at Stony Brook
: 4.88 MB
مدل سازی پدیده های طبیعی با روش بولتزمن شبکه
Abstract: We have adopted a numerical method in computational fluid dynamics, the Lattice Boltzmann Method (LBM), for simulation and visualization of natural phenomena. The LBM is an alternative and promising numerical scheme for modeling the fluid dynamics from a microscopic perspective, and recovering the Navier-Stokers equations. The LBM has the following obvious advantages: (1) it uses local and simple operators to model complex and nonlinear fluid behaviors; (2) it discretizes the micro-physics of local interactions between fluids and objects, and thus can handle very complex boundary conditions, such as deep urban canyons, curved walls, arbitrarily-shaped objects and dynamic boundaries of moving objects; (3) due to its discrete nature, the LBM lends itself to multi-resolution approaches, and its computational pattern, is easily parallelizable. In addition to the traditional Single-relaxation-time LBM, we introduce a more general version: the Multiple-relaxation-time LBM, which is more appropriate for coupling physical properties (temperature, body forces, etc.) to fluid dynamics and it provides more stable computation. We have applied this method to model light objects floating in the wind, heat shimmering of air, mirage, solid melting and flowing, and contaminant plume dispersion in urban environment. We have accelerated the LBM on commodity graphics processing units (GPUs), achieving interactive performance for our applications. Moreover, we have modelled the front spreading phenomena of fire and flows on 3D surfaces. Our method provides fast and simple simulations and allows users to conveniently control the propagation behaviors. Our LBM-based approaches provide a physically based framework for modeling and simulating natural phenomena that enables the development of computer graphics, movies, games, and scientific prediction simulations.