Acceleration of particle based fluid simulation with adhesion boundary conditions using GPU

Abstract

We present adhesion boundary conditions for smoothed particle hydrodynamics (SPH) with implicit surfaces. An existing method called ghost SPH addresses adhesion boundary conditions and produces plausible liquid animations using ghost particles. The generation of ghost particles, however, takes considerable computation time when it is implemented on graphics processing units (GPUs). The purpose of this paper is to accelerate ghost SPH using GPUs. In order to accelerate the processing of adhesion boundary conditions, we propose a new boundary model that can skip the ghost particle generation process in air and solid objects. The proposed technique is not just efficient but also inherits other advantages of implicit surfaces such as smoothness. Our test results show that the proposed method efficiently produces natural fluid adhesion motion without air or solid particles and achieves more than a hundredfold speed up compared to ghost SPH.