Hybrid Parallelization of Population Balance Models for Massively Accelerated Modeling of Particulate Processes

Hybrid Parallelization of Population Balance Models for Massively Accelerated Modeling of Particulate Processes

Franklin Bettencourt, Anik Chaturbedi, Rohit Ramachandran

Particulate processes are common in industry and are used to produce commercially important products such as fertilizers, detergents, polymers, pharmaceuticals, food products, catalysts, and more. Particulate processes in general are complex to understand and manufacture efficiently and companies spend millions making a particulate process work before they can sell a high quality product. Recently population balance models (PBM) have been effectively used to model these particulate systems with a great deal of accuracy however to tune these models to replicate the real process behavior, even with the fastest computers they can take days or weeks to solve. In this work, a simpler way of parallelizing these simulations by using Message Passing Interface (MPI) and a more advanced way using a hybrid MPI-OpenMP method have been applied to PBM. We study the speed up and the scale up of these parallelization techniques for different system sizes and different computer architecture to come up with the fastest way to solve a PBM to date. Successful implementation of these parallelization techniques would eventually lead to the use of PBMs for real time control of critical quality attributes (CQA) of particulate processes, leading to faster time-to-market products.