(219e) Modelling and Analysis of Roller Compaction Using the Discrete Element Method

Roller compaction is commonly adopted in dry granulation processes due to its reliability when dealing with cohesive materials and its suitability for both batch and continuous manufacturing. Despite its widespread use, the mechanics of this process is not fully understood, leading to high reliance on physical trial-and-error in process design and sub-optimal performance in operations. Simulations based on the Discrete Element Method (DEM) can provide unprecedented insight into the mechanics of the roller compaction process and enable rapid and cost-effective virtual trials that inform the relationship between operational parameters and process performance.

In this work we present an efficient DEM modelling methodology for the roller compaction process and a parametric analysis of process mechanics based on DEM simulation data.

The proposed methodology combines meso-scopic modelling with inertial scaling to achieve three orders of magnitude increase in computational efficiency relative to the traditional DEM modelling approach, making the simulation of the full-scale compaction process practical. Material model calibration based on shear cell and uniaxial compression measurements is employed to achieve an excellent prediction of the experimentally observed roller compaction force and mass flow rate at different screw speeds and roller gaps. The model calibration is performed using an automated approach that combines Design of Experiments (DoE), machine learning and genetic algorithm optimization in the interest of efficiency.

A virtual parametric analysis is performed, focusing on the effect of screw speed and roll gap on the specific roll force and the relative density distribution in the compact. The analysis reveals the determining role of the mass flow rate on the roll force as well as a complex interplay between the process parameters and the compact relative density. A spatially non-uniform and temporally variable compact density distribution is observed in all cases. This is in contrast with the results for a compact made by slugging and may partially explain the difference in performance of the two processes. The simulations reveal a complex stress state in the nip region which has a cyclic behavior that can be linked to the motion of the screw and may explain the observed spatial and temporal variance of the relative density distribution in the compact.