A Parallel Free Surface Lattice Boltzmann Method for Large-Scale Applications

In recent years the lattice Boltzmann method (LBM) has been established as a popular alternative to conventional computational fluid dynamics. With a free surface extension to the method, simulation of bubbly flows and even metal foaming processes is possible. The extension is based on a volume of fluid approach and an explicit tracking of the interface, including a reconstruction of the curvature to model surface tension. In order to accomplish realistic engineering applications, large domain sizes are required, and thus efficient parallelization for several thousand processes is inevitable. Our previous implementation of the parallel free surface algorithm used all-to-all communication schemes resulting in only moderate parallel efficiency when using more than hundred processes. Therefore, the algorithm has been adapted to communicate updates only locally in a restricted neighborhood, which complicates data exchange between processes, in particular when bubbles extend across several subdomains and in case topological changes occur through the coalescence of bubbles. The novel algorithm increases parallel efficiency and enables usage of several thousand processors, rendering large-scale engineering applications like simulation of liquid water in a fuel cell possible. It has been integrated into the waLBerla LBM framework, which features basic tools for communication and data management, designed for massively parallel flow simulations. With this implementation, free surface simulations exhibit parallel efficiency of 90% on up to 9 152 cores.