Simulation of distributed product structures in combined discrete and continuous production processes for solid, particulate products

• Applicants: Professor Dr.-Ing. Christoph Herrmann; Professor Dr.-Ing. Arno Kwade
• Subject Area: Mechanical Process Engineering ; Production Automation and Assembly Technology
• Term: since 2018
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 413141366

In the production of solid dosage forms, the ultimate goal is consistently high quality defined by narrow ranges for critical quality attributes. The overall goals of this project are therefore to build robust process-structure models and, based on these, an agent-based simulation for ring-layer granulation. Agent-based simulations offer the possibility to overcome conflicts between level of detail and computation time, which often occur in other simulation approaches for particle-based processes such as flow sheet and discrete element methods. Furthermore, they can predict process outcomes in near real-time when linked to live sensor data. Accordingly, the process-structure models, the agent-based simulation approaches and the live sensor data are to be bundled in a cyber-physical production system (CPPS). This also incorporates the previous investigations on the mixing and tabletting process and connects them with the novel ring-layer granulation. The basis for the model developments is the in-depth experimental investigation of the continuous ring-layer granulation. Online process analytics will be established to measure the product properties and soft sensors will be developed to describe the meaningful, but not yet directly measurable intermediate properties of hold-up and granulate density. In addition, the energy requirements during the (sub-)processes are measured in order to record these as a target variable in addition to the product quality to be achieved and to take them into account in the prediction of favourable process settings. Mechanistic and grey-box (using machine learning methods) process and property models are developed to determine the product structure and property distribution based on process parameters and sensor values for each process step. In order to take into account the interdependencies and particle transport along the process chain, the developed models will be embedded in an agent-based process chain simulation, which will enable a near-real-time mapping of the process. The implementation of the project is based on the partners' intensive joint research in the areas of mechanistic models for particulate processes, agent-based simulation and implementation of cyber-physical production systems.

• DFG Programme: Research Grants
• Co-Investigators: Dr. Jan Henrik Finke; Dr.-Ing. Max Juraschek