Comparison between Compartment and Computational Fluid Dynamics Models for Simulating Reactive Crystallization Processes

This work compares two different computational approaches aimed at describing the reactive crystallization or precipitation process in stirred tanks. The first approach is a full computational fluid dynamics (CFD) model coupled with population balance modeling, which is accelerated by the operator-splitting method and hybrid MPI-OpenMP parallelization. Here, emphasis is given to the hybrid MPI-OpenMP parallelization that improves parallel scalability, when the operator-splitting method is used to take relatively large time steps, despite the large separation of time-scales in such processes. The second approach is a compartment model (CM) enhanced by an automatic tool for the generation of compartments based on some relevant features of the system. The two models are compared for a case study of particular interest: the reactive coprecipitation of Ni-Mn-Co hydroxide in a continuous stirred tank, main precursor to produce cathode active materials of lithium-ion batteries. The obtained results demonstrate the effectiveness of hybrid parallelization in improving the parallel scalability of the CFD model. In addition, it is shown that the CM can produce less accurate but still relevant predictions with relatively small computational cost.