Probabilistic design space exploration and optimization via bayesian approach for a fluid bed drying process.

Area of Science:

• Pharmaceutical Manufacturing
• Process Engineering
• Bayesian Statistics

Background:

• The International Conference on Harmonisation (ICH) Q8 introduced Design Space (DS) as a framework for pharmaceutical development to ensure product quality.
• Exploring a reliable and robust DS is challenging due to process complexity and uncertainties in pharmaceutical manufacturing.

Purpose of the Study:

• To investigate the probabilistic Design Space (DS) for a fluid bed drying process.
• To explain process operability and performance reliability using a Bayesian approach.
• To identify a probability DS that guarantees product quality at a desired reliability level, considering material and process uncertainty.

Main Methods:

• Development of a Bayesian model integrating a surrogate-based predictive model with uncertainty quantification for material variability.
• Discretization of the operational variable domain using a grid search technique to explore the probabilistic DS.
• Application of optimization techniques to maximize the DS region and improve operability.

Main Results:

• The Bayesian approach effectively identifies a probabilistic Design Space (DS).
• The method ensures product quality at a desired reliability level.
• It successfully accounts for both material and process uncertainties in fluid bed drying.

Conclusions:

• The Bayesian approach is a powerful tool for defining a probabilistic Design Space (DS) in pharmaceutical manufacturing.
• This methodology enhances the reliability and robustness of pharmaceutical processes.
• It provides a framework for guaranteeing product quality under uncertainty.