The outcome of this project is a higher understanding of multiphase flows with a focus on dense granular and bubbly flows. This is achieved by implementing an innovative numerical algorithm for particle tracking on recently developed, exremely powerful graphic processing units (GPUs). The core idea is to combine the flexible open-source computational fluid dynamics (CFD) solver "OpenFOAM" on conventional central processing units (CPUs) and a particle tracking module running simultaneously on GPUs. This will result in an extremely accelerated multiphase flow calculation, since GPUs are designed for massively parallel computations of particle trajectories. Such a combined CPU-GPU simulator has not been used for multiphase flow calculations before. The expected speed-up, i.e., the reduction in simulation time compared to a conventional particle-based simulation code, is expected to be in the order of 20 to 100 (!).
The resulting particle-based simulation code will be used for four-way-coupled simulations. Thus, the simulations will take into account fluid-particle, particle-fluid, particle-particle as well as particle-wall interactions. Currently, these simulations are very time-consuming, or even impossible, hindering the progress in the understanding of dense multiphase flows. This strong need for advancing multiphase flow simulation technology is reflected by various efforts by, e.g., the U.S. Dept. of Energy (DOE) or the pharmaceutical industry. The proposed project exactly adresses the scientific challenges connected to this need by combining novel algorithms and massively parallel computing on GPUs. During the project the applicant will (i) explore the virtues, merits and limitations of the developed CPU-GPU simulator, (ii) demonstrate its performance and (iii) will use the knowledge gained to understand and model dense granular as well as bubbly flows. For the latter the code will be extended during the implementation phase in Austria. This will enable the applicant to study also complex three-phase systems like slurry bubble columns or coalescence and breakage dominated flows in, e.g., bioreactors.
Prof. Sundaresan's group at the host institution in Princeton will provide the necessary knowledge for the core design of the algorithm. The group will also deliver reference results for the benchmark simulations and will help in validating the developed simulator. The continuation of the cooperations with Dr. Charles Radeke as well as Prof. Glasser's group at the Rutgers University will also contribute to the success of the project.