Adaptive Particle Splitting Based on Turbulence Energy for Fluid Simulations on GPUs

Abstract: We present a simulation method for particle-based fluids to capture small-scale features such as splashes in higher resolution than the underlying base simulation. For this purpose, we split particles in visually important regions of the fluid into smaller, high-resolution particles. To reduce the computational overhead introduced by the additional simulation scale, we propose splitting only those particles found within turbulent surface regions that are visible to the camera. We extract these particles in screen space and decide to split them based on their turbulent energy. Further, to compensate for irregularities in the quantity field that are introduced by transitioning particles, we use a simple blend-ing approach to maintain stability. We fully implemented our method for graphics processing units to further accelerate the computational speed. In early experiments we could achieve speed increases of up to two over a high-resolution simulation while preserving similar visual qualities. 1