(124a) Multi-Scale Modeling and Validation of Twin Screw Granulation Processes

Process robustness and controllability is of particular importance in the highly-regulated pharmaceutical industry. While batch processing is typically used in pharmaceutical manufacturing operations, continuous processing has numerous potential advantages in cost, efficiency, scalability, and controllability. In wet granulation, a liquid binder is sprayed onto a powder mixture of active pharmaceutical ingredient (API) and excipient particles, which aggregate to form larger granules. Granulation processes are often inefficient because granule size and composition can be widely distributed and difficult to control. Continuous granulators that have been developed or investigated for use in pharmaceutical processing include fluidized bed granulators, twin screw granulators, and high shear granulators. In this study, a multi-scale model for twin screw granulation was developed and validated with experimental data.

Wet granulation is governed by three rate processes: wetting and nucleation, consolidation and aggregation, and attrition and breakage. Population balance models are used track the number of particles with a given set of characteristics, such as size and composition, taking these rate processes into account. A three-dimensional population balance model was developed to simulate concurrent distributions in size, composition, and liquid binder content. It has been demonstrated that these properties strongly influence the aggregation rate. In order to represent the continuous granulator, spatial coordinates were added to the population balance model, along with an inflow and outflow of particles. This model was coupled with discrete element modeling (DEM). Axial and radial velocities within the twin screw granulator were determined from DEM simulations, and these particle fluxes were input into the population balance model. Experimental data was used to select values for adjustable parameters in the aggregation and breakage rate kernels. Using these rate constants, simulations were performed under varying process parameters, such as liquid binder spray rate. These results were compared to experimental data, demonstrating the predictive capability of the model.