Linking process variables to residence time distribution in a hybrid flowsheet model for continuous direct compression

Abstract Continuous manufacturing of oral-dosage pharmaceuticals is widely understood to ensure consistent product quality and higher productivity at lower costs. Here, a hybrid flowsheet model is developed in order to link the process variables to the residence time distribution (RTD) of mixed pharmaceutical powders in a continuous direct compression (CDC) process containing two separate powder blending units. An empirical equation is regressed from collected RTD data for one formulation that calculates parameters for a tanks-in-series model as a function of system throughput and blender impeller speed. The prediction power of the hybrid flowsheet was tested against the regression dataset and validated against five conditions that were not used in the regression. The hybrid flowsheet was found to perform accurately against the regression dataset and could predict the behavior of the validation dataset. The potential applications of the hybrid flowsheet were demonstrated by developing a soft sensor for predicting off-target material based on a large in-silico dataset generated by the hybrid flowsheet. The soft sensor was demonstrated to accurately predict diversion events triggered by off-target material as well as adjust the throughput and blender speed to prevent the diversion event from occurring.