Multivariate feed forward process control and optimization of an industrial, granulation based tablet manufacturing...

Area of Science:

• Pharmaceutical Technology
• Process Engineering
• Data Science

Background:

• High drug load tablets (>60%) present challenges in achieving consistent disintegration time and tableting yield.
• Traditional batch granulation processes often lack robust control mechanisms for complex formulations.

Purpose of the Study:

• To understand and control variability in disintegration time and tableting yield for high drug load tablets.
• To implement multivariate methods for enhanced process control in high shear wet granulation.
• To develop predictive models for key process parameters and output variables.

Main Methods:

• Utilized historical data from 95 industrial batches for model development.
• Applied Batch Evolution Models (BEMs) and Batch Level Models (BLMs) for process analysis.
• Employed Partial Least Squares (PLS) and Artificial Neural Network (ANN) for predictive modeling.

Main Results:

• Achieved reliable prediction for granulation water amount (±2 kg), tableting speed (±5000 tablets/h), and core disintegration time (±100 s).
• Identified time-dependent process variables and raw material properties influencing output variability.
• Correlated process deviations with observed differences in output variables, identifying improvement opportunities for 80% of batches with high disintegration time.

Conclusions:

• Multivariate methods, including BEMs and BLMs, effectively enhance process control in high shear wet granulation.
• Predictive models built from historical data enable feed forward control for improved tablet quality.
• Systematic process improvement is achievable through the application of trained predictive models.