Predicting Pharmaceutical Particle Size Distributions Using Kernel Mean Embedding.

Area of Science:

• Pharmaceutical Manufacturing
• Chemical Engineering
• Data Science

Background:

• Continuous manufacturing is increasingly adopted in the pharmaceutical industry for solid dosage forms.
• High-quality process models are crucial for optimizing continuous pharmaceutical manufacturing processes.
• Understanding particle size distribution changes in wet granulation is essential for downstream processing, like fluid bed drying.

Purpose of the Study:

• To develop a data-driven modeling framework linking wet granulation machine settings to granule output distributions.
• To enable prediction of particle size distributions without assumptions about their nature.
• To provide a simulation tool for granule size distribution as a soft sensor.

Main Methods:

• A two-step data-driven approach was employed.
• Measured particle size distributions were transformed into a high-dimensional feature space.
• Machine settings and output distributions were related in this feature space, followed by inverse transformation for interpretation.

Main Results:

• A validated data-driven framework for predicting pharmaceutical particle size distributions was established.
• The model effectively links wet granulation process settings to granule size output distributions.
• The approach allows for interpretation in the original measurement space, facilitating understanding.

Conclusions:

• The developed data-driven framework reliably predicts pharmaceutical particle size distributions.
• This predictive capability is vital for quality assurance and troubleshooting in pharmaceutical production lines.
• The framework supports applications like model-based experimental design and real-time model predictive control.