Development of an Intelligent Tablet Press Machine for the In-Line Detection of Defective Tablets Using Machine...

Area of Science:

• Pharmaceutical Manufacturing
• Process Analytical Technology (PAT)
• Machine Learning in Pharmaceuticals

Background:

• Tablet quality control is crucial for drug efficacy and safety.
• Traditional methods for detecting tablet defects are often time-consuming and labor-intensive.
• Integrating Process Analytical Technology (PAT) with machine learning (ML) offers a path towards real-time quality monitoring.

Purpose of the Study:

• To develop a tablet press machine (TPM) integrated with ML and deep learning (DL) for in-line detection of tablet defects.
• To predict tablet defects, including capping and tablet breaking force (TBF), using real-time processing data.
• To establish a TPM as a PAT tool for non-destructive quality assessment.

Main Methods:

• Metformin HCl granules were compressed using a TPM at a commercial scale.
• Random Forest (RF) and Artificial Neural Network (ANN) models were trained on in-line data (compression force, ejection force, speed).
• A PAT-enabled TPM was designed and manufactured, incorporating an RF model for defect detection and sorting.

Main Results:

• The RF model achieved 93.7% predictive accuracy (AUC 0.895), outperforming the ANN model (92.6% accuracy, AUC 0.878).
• The integrated TPM successfully sorted defective tablets in real-time with 99.43% sorting accuracy.
• Defective tablet detection accuracy reached 93.71% using the developed ML-based TPM.

Conclusions:

• ML-based TPMs can effectively detect tablet defects non-destructively during wet granulation scale-up.
• This approach serves as a foundational TPM model for in-line PAT in multi-product, small-batch manufacturing.
• Implementation reduces labor, time, API consumption, and environmental impact.