(137f) Reduced Order-Discrete Element Method Modeling of Comilling for Efficient Integration into Continuous Process

Continuous pharmaceutical manufacturing is known for the advantages it offers for cost, efficiency and reduced ‘time to market’, specifically in the time of rising manufacturing, research and development costs, and competition from generics. The gradual shift from batch to continuous process along with introduction of US food and drug administration’s (FDA) Quality by Design (QbD) initiative has provided the necessary impetus for research efforts to build quality into the product through a thorough understanding of the effect of variability in material properties and process conditions on product properties. The role of process modeling in this paradigm is apparent as effect of a disturbance in continuous process on the product quality needs to be well predicted [1].

Comilling unit in the dry and wet granulation routes in pharmaceutical manufacturing is used to break granules into a required size distribution. Particle size reduction in comill through the impact of a rotating impeller on granules and their eventual exit through a screen is a complex phenomenon. Past work on predicting the breakage has used Population balance models (PBM) where the kernel that represents the probability of granule breakage is represented semi-empirically through their dependence on impeller speed, granule size etc. A mechanistic understanding of the breakage phenomenon can be achieved through discrete element method (DEM) modeling of breakage in comill where particle-particle, particle-wall interactions are captured through underlying system of equations based on Newton’s laws of motion. The DEM model proposed in [2] takes into account a threshold impact energy above which the granules break. A multi scale framework is established where cumulative energy of granules from DEM simulations are utilized in an energy based PBM kernel and an iterative algorithm is used for estimating material specific parameters in the kernel. However, the high computational expense of DEM simulations and the required cumbersome post processing limits its use in any predictive capacity for continuous manufacturing such as integration into flowsheet model. Reduced order modeling can bridge this gap [3].

Current work proposes a reduced order-discrete element method modeling to represent the multi- dimensional and time dependent energy data from DEM simulations. The methodology reduces the computational expense associated with the high fidelity DEM simulations while retaining accuracy of the mechanistic data [4]. DEM simulations of comilling are run at various flow rates and impeller speeds based on an experimental design and the energy distribution data from these simulations are used to build a reduced order model. An analytical relationship that captures dependence of energy of granules on size, time, flow rate and impeller speed is developed. This enables accurate representation of energy based kernel in PBM model. A comparison of various surrogate modeling techniques such as kriging, radial basis functions is also presented. The improvement achieved through this framework is drastic as the DEM simulations take approximately five days where as the reduced order DEM model incorporated into PBM takes few seconds. Through this work, a framework to represent the mechanistic information from DEM efficiently into PBM of a comill operation is established, thus enabling its incorporation into flowsheet modeling of DG and WG routes of continuous manufacturing. The incorporation of reduced order model into a continuous WG route flowsheet model is also demonstrated.

References

[1] A. Rogers, A. Hashemi, M. Ierapetritou, Modeling of Particulate Processes for the Continuous Manufacture of Solid-Based Pharmaceutical Dosage Forms, Processes, 1 (2013) 67.

[2] N. Metta, Ierapetritou, M., Ramachandran, R., A combined experimental and computational approach using discrete element method for the development of a mechanistically motivated breakage kernel, AIChE annual meeting, San Francisco, CA, USA, 13-18 November 2016.

[3] D. Barrasso, A. Tamrakar, R. Ramachandran, Model Order Reduction of a Multi-scale PBM-DEM Description of a Wet Granulation Process via ANN, Procedia Engineering, 102 (2015) 1295-1304.

[4] A. Rogers, M.G. Ierapetritou, Discrete Element Reduced-Order Modeling of Dynamic Particulate Systems, Aiche Journal, 60 (2014) 3184-3194.