Development of Calibration Models using Process Analytical Technology for Advanced Crystallization Process Control

Abstract

With the technological advancement in the last decade, process analytical technology (PAT) is intensively developed with the purpose of monitoring key process variables in real-time. PAT tools have opened up new possibilities for improving process performance. The benefits of this approach are products of predefined properties, higher batch production reproducibility and high quality in a series of batches using the concept of Quality by Design (QbD) and Quality by Control (QbC). The crystallization process is of the greatest importance in pharmaceutical production. The purpose of this research is to develop a system that enables continuous monitoring and optimal crystallization process control. Based on a calibration model for monitoring the concentration of solute, continuous process monitoring and an advanced process control strategy maintain optimal conditions and products of desired critical quality attributes. Neural network-based calibration models were developed to model the dependence of concentration of an active pharmaceutical ingredient on temperature and spectral data obtained by UV-Vis measurements. The best-performing model was developed for a reduced data-set (800–200 nm) with 20 neurons and tangent-hyperbolic transfer function. Developed models will be used for continuously monitoring of the active pharmaceutical ingredient (API) in the crystallization system.