Optimisation of an in-line Raman spectroscopic method for continuous API quantification during twin-screw wet granulation and its application for process characterisation.

In a previous publication, the development of an in-line Raman spectroscopic method for continuous API quantification during twin-screw wet granulation was presented. An in-line method was developed successfully and the developed method showed an acceptable prediction error. A disadvantage of the developed method was that a measurement was only possible in the dark since light influenced the Raman spectra and made a data interpretation impossible. Therefore, the measurement setup for the implementation of the Raman probe was optimised in the present study to allow a measurement in interior light and to further improve the predictive performance. With the optimised setup, two different calibration models were developed and compared. For the first calibration model, spectra were collected in the dark as before and for the second in interior light. The dark calibration model was able to predict the API content with an RMSEP of 0.31% and the light model with an RMSEP of 0.29%. Thus, both PLS models showed prediction errors in the same order. Consequently, it was possible to evaluate Raman spectra which were collected in interior light. Further, the previous prediction error of 0.60% could be clearly decreased. The optimised Raman method was applicable to evaluate the mixing efficiency of the twin-screw granulator during a split feeding process. The quality of the mixture was monitored behind different barrel sections by Raman spectroscopy and the corresponding API concentrations were predicted by the developed calibration model. For a screw length of 40 D and a screw configuration with two kneading blocks a good mixing ability was observed. For a screw length of 20 D and one kneading block the mixing efficiency was largely acceptable whereas a broad scattering of the API content was observed when no kneading blocks were used. In a second part, an experimental design was performed for each screw configuration to evaluate the influence of the barrel-fill level and screw speed on the mixing efficiency. The quality of the mixture using the entire barrel length was minimally influenced by the fill-level. For the other two positions, the screw speed influenced the quality of the mixture slightly. Thus, for an appropriate mixing, a certain barrel length and a screw configuration with two kneading blocks were necessary.