Large-scale SPH simulation for three-dimensional flow

Abstract

Computational load and numerical instability are two of the most challenging problems in enlarging the scale of smoothed particle hydrodynamics (SPH) simulation and applying SPH to a 3D problem. In this work, a study on advanced algorithm is conducted for large-scale SPH simulation in 3D. To reduce computational cost, the concept of a grid-based, linked-list algorithm was adopted and tested by being applied to solve a rotating cylinder problem. The result shows that the size of the search grid should be above or equal to the size of the support domain to obtain a stable and accurate solution. A zero order filtering algorithm is applied for reduction of noise signal in density and pressure for every 50 steps. Consequently, noise signal in density and pressure is apparently decreased and potential instability of computation has decreased. The improved SPH algorithm was applied to simulate 3D water discharge and the result shows that the algorithm satisfies Bernulli's theorem with acceptable degree of errors.