Sampat, Chaitanya. Parallel solution to multi-scale, multi-dimensional coupled DEM-PBM model for high shear granulation using high performance computing. Retrieved from https://doi.org/doi:10.7282/t3-jssh-ss48
DescriptionParticulate processes are prevalent in the pharmaceutical industry but, the physics underlying these processes are complicated. The particle-level models used to describe these systems require large amounts of computation to solve which makes simulations slow. Another approach to model these systems is less accurate bulk model which does not capture all the particle level data but is more efficient to simulate and requires lesser computational power. A quicker and a more accurate way to model such systems is to use a multi-scale model i.e. use the particle-scale data into a bulk model. Using a coupled model does not decrease the time taken by each individual component of the simulation, thus there is also a need to increase the speed of the separate simulations.
In this work, a unidirectional multi-scale model was used to model the high shear wet granulation process. A multi-dimensional population balance model (PBM) was developed with a mechanistic kernel, which obtained collision data from the discrete element modeling (DEM) simulation. The PBM was run in parallel using MPI + OMP hybrid technique. The DEM simulations were performed on LIGGGHTS, which runs in parallel using MPI. Speedup of about 14 times was obtained for the PBM simulations and around 12 for the DEM simulations. This coupling was performed using the radical pilot for scaling studies from 1 to 128 cores for the PBM and up to 256 cores for the DEM. A further improvement to the PBM code was also done by developing a code in CUDA C++ such that it could utilize up to 1024 cores on the NVIDIA graphical processor units (GPU). Using this developed framework, the granulation process has been modeled accurately much faster than existing approaches in literature