A data-driven process control strategy aligned with quality by design using a local linear modelling method and fault...

Area of Science:

• Pharmaceutical Engineering
• Process Control
• Chemical Engineering

Background:

• Continuous granulation requires intelligent control systems adhering to Quality by Design (QbD) principles.
• Existing control systems struggle to identify specific disturbance sources, hindering regulatory compliance.
• Automated root cause identification is crucial for robust pharmaceutical manufacturing processes.

Purpose of the Study:

• To propose a novel Fault Detection and Diagnosis (FDD) method for continuous granulation.
• To develop a control framework element that recognizes and corrects process disturbances based on their source.
• To demonstrate the FDD method's capability in isolating root causes within a non-linear design space.

Main Methods:

• Modeling the twin-screw granulation design space using linear Partial Least Squares (PLS).
• Utilizing the Prediction Reliability Enhancing Parameter (PREP) method for comprehensive data collection.
• Dynamically generating local models around operational points to reduce FDD algorithm error.
• Employing an optimization framework for fault diagnosis by comparing observed and predicted particle size distributions (PSD).

Main Results:

• The FDD method successfully identified the root cause of simulated process disturbances.
• Local modeling and optimization framework effectively addressed the complex, interdependent nature of granulation inputs.
• The approach demonstrated systematic isolation of faults, overcoming input interdependencies.

Conclusions:

• The proposed FDD method is a significant step towards realizing a QbD-compliant control system for continuous granulation.
• This approach enables intelligent control by diagnosing disturbances, not just compensating for them.
• The study highlights the potential of FDD in enhancing process understanding and control in pharmaceutical manufacturing.