Fast and Scalable Turbulent Flow Simulation with Two-Way Coupling

摘要

Despite their cinematic appeal, turbulent flows involving fluid-solid couplingremain a computational challenge in animation. At the root of thiscurrent limitation is the numerical dispersion from which most accurateNavier-Stokes solvers suffer: proper coupling between fluid and solid oftengenerates artificial dispersion in the form of local, parasitic trains of velocityoscillations, eventually leading to numerical instability. While successiveimprovements over the years have led to conservative and detail-preservingfluid integrators, the dispersive nature of these solvers is rarely discusseddespite its dramatic impact on fluid-structure interaction. In this paper, weintroduce a novel low-dissipation and low-dispersion fluid solver that cansimulate two-way coupling in an efficient and scalable manner, even forturbulent flows. In sharp contrast with most current CG approaches, weconstruct our solver from a kinetic formulation of the flow derived from statisticalmechanics. Unlike existing lattice Boltzmann solvers, our approachleverages high-order moment relaxations as a key to controlling both dissipationand dispersion of the resulting scheme. Moreover, we combine our newfluid solver with the immersed boundary method to easily handle fluid-solid coupling through time adaptive simulations. Our kinetic solver is highlyparallelizable by nature, making it ideally suited for implementation onsingle- or multi-GPU computing platforms. Extensive comparisons withexisting solvers on synthetic tests and real-life experiments are used to highlightthe multiple advantages of our work over traditional and more recentapproaches, in terms of accuracy, scalability, and efficiency.