Transparent predictive modelling of the twin screw granulation process using a compensated interval type-2 fuzzy...

Area of Science:

• Pharmaceutical Sciences
• Chemical Engineering
• Data Science

Background:

• Granulation is a critical unit operation in pharmaceutical manufacturing.
• Accurate prediction of granule properties is essential for process optimization and product quality.
• Existing modeling approaches may not fully capture the complexities and uncertainties in granulation processes.

Purpose of the Study:

• To develop a novel systematic modeling framework using machine learning for the granulation process.
• To predict the properties of granules produced by twin screw granulation (TSG).
• To refine predictive models by accounting for uncertainties and unmodeled behaviors.

Main Methods:

• Elicitation of an interval type-2 fuzzy model to predict granule properties.
• Integration of a Gaussian mixture model (GMM) to characterize fuzzy model error residuals.
• Validation of the proposed modeling algorithms through laboratory-scale experiments.

Main Results:

• Successful prediction of the size of granules produced by TSG.
• Most predictions fell within a 95% confidence interval, demonstrating high accuracy.
• The integrated framework effectively refined the predictive model by addressing uncertainties.

Conclusions:

• The developed machine learning framework provides a robust approach for modeling the TSG process.
• The study demonstrates the successful application of interval type-2 fuzzy models and GMM for pharmaceutical granulation.
• This approach enhances the prediction accuracy of granule properties, aiding in process control and development.