PREON® technology is the result of many years of research with one goal: Finding a new approach to fluid simulation that produces physically accurate simulations in unprecedented resolutions.
We have developed a stable and highly functional implicit solver, which combines the numerical robustness of grid-based methods with the intrinsic advantages of mesh-free approaches. The superior performance of PREON® is defined by
Schematic formula for estimating the computational effort for a single simulation timestep
PREON® employs the SPH framework to formulate innovative implicit methods which handle time-steps at CFL 1 without compromising on simulation quality. This formulation can handle time steps many times larger than explicit SPH methods.
Comparison of the computational cost between an explicit and an implicit formulation.
For any particle-based method, the costs per time step are defined by the computation time for updating the neighbors and forces per particle. PREON® employs highly efficient, spatially adaptive data structures which update the neighborhood of many million particles. A state-of-the-art hybrid parallel programming implementation ensures great scalability on the desktop as well as on high performance clusters. The implicit formulation requires the solution of a linear system of equation.
We are continuously optimizing this core innovation, significantly reducing computational costs and evolving the performance of the PREON® core further and further.
Particles carry physical values of the partial volume they are representing. Like the color of a pixel, these quantities are constant within the particle itself. Thus, finer particles capture more details. An adaptive algorithm can reduce computational effort significantly by using finer particles in regions of interest and coarser ones where less detail is sufficient. PREON® allows you to simulate at three different levels of resolution, including automatic coarsening and refinement of particles in multiple user-defined regions. The different levels are fully coupled, and the computational overhead for the refinement algorithm is small which yields tremendous speed-up in various applications.