464607 Model-Based Process Analysis in Continuous Manufacturing of Pharmaceuticals: Calibration of a Fluid Bed Dryer Model

Monday, November 14, 2016
Grand Ballroom B (Hilton San Francisco Union Square)
Michael Ghijs1,2, Fien De Leersnyder2, Daan Van Hauwermeiren1,2, Valérie Vanhoorne3, Jurgen Vercruysse4, Philippe Cappuyns5, Séverine T.F.C. Mortier1,2, Thomas De Beer2, Krist V. Gernaey6 and Ingmar Nopens7, (1)BIOMATH, Ghent University, Ghent, Belgium, (2)Laboratory of Pharmaceutical Process Analytical Technology, Ghent University, Ghent, Belgium, (3)Department of Pharmaceutics, Ghent University, Ghent, Belgium, (4)Laboratory of Pharmaceutical Technology, Ghent University, Ghent, Belgium, (5)Janssen Pharmaceutica, Beerse, Belgium, (6)CAPEC-PROCESS Research Center, Dept. of Chemical and Biochemical Engineering, Technical University of Denmark, Kgs. Lyngby, Denmark, (7)BIOMATH, Dept. of Mathematical Modeling, Statistics and Bioinformatics, Ghent University, Ghent, Belgium

The goal of this work is to investigate the dominating phenomena in the fluid bed dryer in a continuous pharmaceutical wet granulation line, which are acting on the Particle Size Distribution (PSD) and moisture content of the product. Knowledge about these phenomena is built with model-based analysis, using a semi-mechanistic model. A thorough and novel experimental data collection campaign is done to create sufficient data for the calibration and validation of this model. Along with that, this work provides clarity into the application of a model for researching the particle property dynamics in a fluid bed dryer, and to produce concrete criteria for the success of every step in the application of the model. In this way, model-based process analysis for the transition to continuous manufacturing in the pharmaceutical industry, becomes more concrete.

The interaction and evolution of PSD and moisture content is investigated in a ConsiGma‑25™ semi-continuous fluid bed dryer, which is fed with wet granules from a twin-screw granulator through pneumatic transport in a tube. In the dryer unit, the continuous flow of granules is divided over its six separate drying segments, in which the granules are batch dried. This partitioning into parallel and sequential batch drying processes approaches a continuous mode over the entire dryer.

In total, 49 different drying experiments were performed with wet granules that were produced under constant granulation settings. For the dryer, the following process settings were varied: the dryer inlet air flow rate was varied between 360 and 440 m³/h and the inlet air temperature between 40 and 60°C. The filling time of a dryer cell, governing the cell load, was varied between 60 and 180 seconds. For each of the combinations of the previous settings and one center point of the settings, experiments were conducted for, at least, five different drying times, ranging from the filling time plus 20 seconds to maximally six times the filling time. As such, the drying times in the experiments span from 80 to 1080 seconds. The high-resolution PSD of the dried granules was collected using a particle size analyser, QICPIC (Sympatec), with a dispensing system, Gradis (Sympatec). The sample is collected after drying and directly sieved into five fractions. The moisture content of each sieve fraction, as well as the total sample, is determined by means of the Loss-On-Drying method (Mettler LP16 moisture analyzer).

An obvious observation can be made that the moisture content of the granules of the total sample decreases when drying time increases. Two phases can be distinguished during drying: (1) a fast phase, where evaporation of the surface water occurs, and (2) a slower drying phase in which water from the core needs to migrate to the surface to get evaporated (Mortier et al., 2012). It can be noticed that smaller granules have a lower moisture content compared to the larger granules, caused by the larger specific surface of the small granules. This phenomenon can only be noticed when shorter drying times are applied. When the drying time is increased, the moisture content of the different size fractions does not deviate significantly, indicating that drying is finished. The experimental data also indicates that most of the breakage occurs during the transport of the granules to the dryer and the loading into the dryer segments. The extent of this breakage was significantly different between the two formulations used in the experiments.

Through the application of a 2D Population Balance Model (PBM), the evolution of PSD and moisture content is simulated. The PBM describes the granules in the system according to a two-dimensional grid of discrete classes of particle size and moisture content, where the last one is given as the wet radius of the particle. Breakage and loss of moisture content of the granules are solved sequentially within each time step, approaching a simultaneous simulation of these phenomena (Mortier, 2014). Breakage is solved using the fixed-pivot discretization scheme. The change in wet radius is modelled according to a drying curve with two regimes, separated by the point where the wet radius equals the particle radius. When the wet radius is smaller, it is assumed that the moisture in the particle resides in its pores, resulting in a lower drying rate. The evolution of the moisture content is solved using the high resolution finite volume scheme (Mortier et al., 2013).

The three steps in the application of the model are model selection, calibration and validation. Model selection denotes selecting the breakage function, which describes the breakage regime in the system. Once a suitable breakage function is found, its empirical components, as well as the drying rates, are calibrated to the system for some of the different process conditions, using the experimental data. The other part of the process conditions are reserved for validation of the model. The methodology of these three steps, including the success criteria, will be presented. The study is done based on criteria from the statistics of comparison of distributions, applying the moments of a PSD to assess calibration and validation success. Each moment contains different information on the PSD. Therefore, incorporating various moments into a success criterion results in an all-round assessment of the PBM prediction. This way, the concrete and statistically sound comparison of PSD is brought into the calibration and validation steps of a PBM, important steps that are often overlooked in practice (Mortier, 2014).


Mortier, S. T. F. C., De Beer, T., Gernaey, K. V., Vercruysse, J., Fonteyne, M., Remon, J. P., Vervaet, C., Nopens, I. (2012). Mechanistic modelling of the drying behaviour of single pharmaceutical granules. European Journal of Pharmaceutics and Biopharmaceutics, Vol. 80(3), 682–9.

Mortier, S. T. F. C., Gernaey, K. V., De Beer, T., Nopens, I. (2013). Development of a Population Balance Model of a pharmaceutical drying process and testing of solution methods. Computers & Chemical Engineering, Vol. 50, 39–53.

Mortier, S.T.F.C. (2014). Modelling the drying behaviour of wet granules in the context of fully continuous pharmaceutical tablet manufacturing. Ghent University. Faculty of Bioscience Engineering, Ghent, Belgium.

Extended Abstract: File Not Uploaded
See more of this Session: Poster Session: Pharmaceutical
See more of this Group/Topical: Pharmaceutical Discovery, Development and Manufacturing Forum