Soft sensor for content prediction in an integrated continuous pharmaceutical formulation line based on the residence time distribution of unit operations

In this study, a concentration predicting soft sensor was achieved based on the Residence Time Distribution (RTD) of an integrated, three-step pharmaceutical formulation line. The RTD was investigated with color-based tracer experiments using image analysis. Twin-screw wet granulation (TSWG) was directly coupled with a horizontal fluid bed dryer and an oscillating mill. Based on integrated measurement, we proved that it is also possible to couple the unit operations in silico. Three surrogate tracers were produced with a coloring agent to investigate the separated unit operations and the solid and liquid inputs of the TSWG. The soft sensor's prediction was compared to validating experiments of a 0.05 mg/g (15% of the nominal) concentration change with High -Performance Liquid Chromatography (HPLC) reference measurements of the active ingredient proving the ad-equacy of the soft sensor (RMSE < 4%).

Automated summary

In this study, a concentration predicting soft sensor was developed based on the Residence Time Distribution (RTD) of a pharmaceutical formulation line. The RTD was investigated using color-based tracer experiments with image analysis. The study also demonstrated the feasibility of coupling unit operations in silico. The soft sensor showed accurate predictions compared to validation experiments, proving its adequacy.