Automatic control and regulation online in the fluid bed
The aim of the joint project was the realization of a measurement and control system for monitoring and controlling the pharmaceutical manufacturing step “fluid bed granulation” on a pilot granulator. The automatic monitoring and control of the granulation process was to be developed on the basis of the direct product properties, since the granulation process defines the basis for important characteristics of the subsequent pharmaceutical form (uniformity of the active ingredient content, or processability in tabletting). In the course of this, the influence of the data recorded today on the quality of the intermediate product produced was to be examined. In addition, new possible parameters for process control were to be found and analyzed online. These could be, for example, the particle size distribution in the granules or a surface characteristic of the particles.
At the end of the project, methods were to be developed that would make it possible to monitor and control critical (relevant) process parameters in fluid bed granulation. In principle, this was intended to establish improved process control of granulation processes in the fluidized bed. Improved granules are to be produced through the targeted influencing of property distributions in fluid bed granulation. The already high quality level in pharmaceutical production would be further increased with the consequence of producing fewer deviations, having to block fewer batches and thus also improving quality and drug safety for the end user as well as conserving natural resources.
Principle of artificial neural networks
Main objectives of the project:
- Identification of relevant parameters of the process or also of the granulate as control variables,
- Development of novel model-based algorithms for the prediction of process-relevant product parameters and parameter distributions from spatially and spectrally resolved product properties,
- Development of novel online sensor systems for spatially resolved quantification of process-relevant product parameters,
- Elaboration of the plant and control engineering fundamentals for the precise control of particle growth in granulation processes as well as testing on a laboratory scale,
- Validation of the new process on a fluid bed granulator at pilot scale,
- Ensuring regulatory feasibility.
Glatt Ingenieurtechnik focused on process modeling via artificial neural networks (KNN), which are then applied in the elaboration of the process control. This involved an extensive experimental test program on a laboratory apparatus in order to generate sufficient training data for modeling with KNN. Furthermore, the developed process control was scaled and validated on a pilot plant.
The subproject was divided into 12 work packages:
- Creation of a requirements profile
- Basic experiments, analytics and process analysis
- Experimental evaluation and process modeling
- Derivation of a control methodology
- Construction of a laboratory apparatus for basic experiments
- Development of adaptive, self-learning process models
- Development of model-based control of the granulation process
- Functional verification of the model-based control
- Research of technical solutions for process scaling
- Bayer Weimar GmbH und Co. KG
- Glatt Ingenieurtechnik GmbH
- Forschungszentrum für Medizintechnik und Biotechnologie (fzmb) GmbH
- Leibniz-Institut für Photonische Technologien e.V. (IPHT)
The project “ASTEROID-WS” is funded by the Free State of Thuringia within the framework of the joint project number 2016 FE 9052 and co-financed by funds of the European Union within the framework of the European Regional Development Fund (ERDF).
Funding period: 01.04.2017 – 31.03.2020
Further information on this topic and related topics can also be found in the following publications: