NIR-Based Intelligent Sensing of Product Yield Stress for High-Value Bioresorbable Polymer Processing

Area of Science:

• Materials Science
• Polymer Engineering
• Biomaterials

Background:

• Polylactide (PLA) is a crucial bioresorbable polymer for medical implants and drug delivery systems.
• Careful processing is essential to prevent degradation and maintain the integrity of PLA-based medical devices.
• Real-time quality control is vital in the medical device industry, necessitating robust analytical methods.

Purpose of the Study:

• To evaluate the potential of fused in-process sensor data and multivariate regression for predicting the mechanical strength of extruded PLA.
• To assess the suitability of these methods as intelligent sensors for real-time quality analysis in the medical device industry.
• To compare the performance of linear and nonlinear regression methods for predicting PLA mechanical properties.

Main Methods:

• Combined in-process measurements of temperature, pressure, and near-infrared (NIR) spectroscopy.
• Applied multivariate regression techniques, including Partial Least Squares (PLS), Random Forest (RF), and Support Vector Regression (SVR).
• Utilized Principal Component (PC) dimension reduction prior to applying nonlinear methods.

Main Results:

• Fusion of NIR and conventional process sensor data is required for robust predictions across varying processing conditions.
• While PLS showed moderate performance, RF and SVR, after PC dimension reduction, demonstrated excellent predictive performance.
• Nonlinear methods proved more reliable than linear methods over the full processing range for mechanical property prediction.

Conclusions:

• Nonlinear regression methods (RF, SVR) show significant potential to outperform traditional linear methods for soft sensing of mechanical properties in PLA processing.
• These nonlinear approaches can meet industrial robustness standards, even with limited training data typical in high-value material processing.
• The study highlights the capability of advanced data fusion and machine learning for real-time quality assurance in regulated industries like medical devices.