(407e) Use of Process Data for Granule Moisture Estimation in the Consigma 25TM Semi-Continuous Fluid Bed Dryer

Continuous pharmaceutical manufacturing lines containing a wet granulation unit operation - like twin screw granulation - require the implementation of a subsequent drying step in order to achieve the desired granule moisture. Several continuously and semi-continuously operated dryers are available to the pharmaceutical industry, e.g., [1–3]. The semi-continuously operated ones [1,3] have in common, that the continuous material stream is split into separated mini-portions of material, each of them being dried in an individual cell of the dryer. In this study, the ConsiGma 25^TM dryer will be investigated in detail. It consists of six drying cells that are consecutively filled and – after drying has finished – emptied.

The most important material property during the drying process is the moisture content of the granules. If this quantity would be available for all the dryer cells, sophisticated feedback control strategies could be developed and implemented, ensuring that the granule moisture after drying is the same for all the processed granules. However, the systems that are commonly installed in industry only allow for the measurement of granule moisture in one cell and not in all of the cells. The moisture is mainly affected by the granules initial moisture content, the inlet air temperature, the air volume flow and the drying time. Therefore, a mathematical model that uses this data for predicting the moisture content in the individual cells is of interest. Further, the model should not be computationally demanding in order to allow its real-time execution in parallel to the process. This work presents such a model.

Method

A mathematical model has been developed, which uses available process data in a ConsiGma 25^TM drier for estimating the granule moisture content. The proposed model is based on mass- and energy balances. Some parameters of the model are known (e.g., material properties like specific heat capacity of water), whereas others need to be identified (e.g., heat and mass transfer coefficients). Since up to six cells need to be simulated simultaneously, the modeling approach was kept as simple as possible by proposing appropriate simplifications (e.g., lumped parameter model, perfect mixing). A set of experiments was planned and executed and the generated process data was used to identify the model parameters. A second set of experiments - at different process settings - was conducted for validating the proposed model.

Results

The proposed model could predict the granule moisture of the validation dataset by an error of approximately 1wt.% (wet basis) of moisture, which is considered to be sufficiently accurate for the intended purpose. Further, it was demonstrated that the simulation could be executed approximately 400 times faster than real-time on an Intel XEON E3-1245V2, 3.4Ghz, desktop computer, suggesting that its execution is also feasible in real-time on a less powerful PLC.

Conclusion

A model for granule moisture prediction in a ConsiGma25^TM drier has been successfully developed and implemented in simulation. By means of that model, in a next step advanced process control concepts can be realized in order to keep the granule moisture close to the desired value.

References

1. Glatt MODCOS line Available online: https://www.glatt.com/en/products/continuous-technologies-pharma/ (accessed on Apr 30, 2019).
2. Bohle Qbcon 1 Available online: https://www.continuous-production.com/continuous-granulation-continuous-... (accessed on Apr 29, 2020).
3. GEA ConsiGma line Available online: https://www.gea.com/en/products/consigma-ctl.jsp (accessed on Apr 30, 2019).