Incompressible flow simulation based on Smoothed Particle Hydrodynamics (SPH) has been gaining interest in the computer graphics community for more than a decade. The most prominent reason is; its potential for handling breaking waves and small scale phenomena such as droplets and interaction with fine detailed concave objects. However, because of its particle based nature, it has some inherent problems that not only hinder performance, but also plausibility of simulations and renderings.
The dominant force for incompressible fluids is the pressure force. Approaches that are relying on a state equation (SESPH) have been commonly used for pressure determination until a recent approach, namely predictive-corrective incompressible SPH (PCISPH) emerged. The most important advantage of PCISPH over existing methods is allowing much larger time steps.
In this thesis, the standard SPH model and the PCISPH model are discussed and then two novel approaches are presented; which are targeted at improving the performance by means of allowing larger time steps. The first improvement is a boundary handling scheme for SPH fluids. In combination with PCISPH, it allows even larger time steps compared to existing boundary handling methods. Additionally, it enhances the plausibility of simulations by allowing more realistic small scale particle-boundary interactions. The second improvement is an adaptive time-stepping scheme for PCISPH, which automatically estimates appropriate time steps independent of the simulation scenario without any extra parameter setting. Due to its adaptivity, the overall computation time of dynamic scenarios, especially with highly turbulent fluids is significantly reduced compared to simulations with constant time steps. When these two methods are combined, a speed-up of up to an order of magnitude can be achieved.
You may download it in PDF format from here.