Authors: N. Nicolaï, I. Nopens, K.V. Gernaey and T. De Beer

EUPAT 8

Title: Analysis of inline spatial filter velocimetry for granule size feedback control of a continuous pharmaceutical twin-screw granulator.

Introduction

In continuous pharmaceutical twin-screw wet granulation, granule productivity depends largely on the control of liquid and solid feed rates during production. However, even if operational variables such as pump speed, screw speed, temperatures etc. are precisely controlled, the rate mechanisms that account for granule formation are sensitive to changes in feed material properties (e.g. powder rheometry), equipment status (e.g. barrel fouling) and environmental conditions (e.g. humidity). Therefore, real-time control of granulation processes is a requirement for industrial scale production.

Consider for instance the ammonia phosphate granulation cycles used to produce up to a kilo ton of granular fertilizer a day. The continuous stream of granules coming out of these mega plants is within strict size and shape specifications ranging over just a few mm. The trick lies in the cascade of sieves optimised to mechanically sort out all off-size granular material which is subsequently reused as a feed at the start of the process. Here, tight monitoring and control on the recycle flows rates is of paramount importance to maintain on-spec product. Yet such an approach is not applicable for continuous pharmaceutical granulation because of product traceability issues. This indicates an urgent need in the pharmaceutical industry for alternative methods based on novel PAT and QbD principles which can assure granule quality at the source.

Precise inline measurement of both process conditions along with granule growth is seen as a valuable PAT solution for the real-time monitoring of continuous twin-screw production lines. To counteract unintended product variability in e.g. granule distribution mean, median and width, dedicated automatic control strategies also need to be implemented here. This said, the current work tries to identify the need and applicability of feedback control loops for the continuous control of granule size characteristics measured inline during twin-screw granulation using spatial filter velocimetry.

Materials and Methods

α-Lactose monohydrate filler (Pharmatose 200M, Caldic Belgium NV, Hemiksem, Belgium) was granulated with distilled water as granulation liquid. All granulation experiments were performed with the twin screw granulation unit of the turnkey ConsiGmaTM-25 from-powder-to-tablet line (GEA Pharma Systems NV, Collete). To obtain inline granule size measurements using the Parsum IPP 70 instrument, a custom process-instrument interfacing device was made. Both process and instrument data were collected in real-time with SIMATIC SIPAT 4.0 (Siemens AG, Brussels, Belgium) configured with OPC collectors. As a mathematical assistant MATLAB 7.12 (The MathWorks, Inc., Natick MA, USA) was used as a COM automation server to execute function calls from the SIPAT method configured by the user.

Results and discussion

Initial experiments indicate that the software settings of the Parsum IPP 70 are of utmost importance when trying to reproduce accurate inline measurements. Therefore, substantial work was devoted to the optimisation of the particle buffer size, i.e. the number of valid particles used in computing granule size distributions, and the maximal particle loading, which is a measurement threshold to compensate for high granule concentrations in the sensing volume. The aim is to find reproducible signals, i.e. same process conditions lead to the same measurement signal, and signals that are sensitive for dynamic changes in the process conditions, as will be explained. Satisfactory results were found for the Dv10, Dv50 and Dn90 metrics. Next the identification of different signal interactions, i.e. dynamic correlations, as well as their controllability, i.e. the ability to keep acceptable control performance on the outputs. From measurement it is clear that the size metrics such as the percentiles derived from the same size population show strong interaction, this clearly hampers there individual controllability as process outputs. With this information at hand, new possibilities for feedback control of twin-screw granulation line will be further investigated.