Implementation of Machine Learning in Flat Die Extrusion of Polymers.

Area of Science:

• Polymer processing
• Materials science
• Computational modeling

Background:

• Flat die extrusion of polymer sheets and films faces challenges in achieving uniform thickness and defect-free production.
• Current methods rely on trial-and-error for different polymer grades and flow rates, increasing costs and time.

Purpose of the Study:

• To investigate the application of machine learning (ML) for guiding flat die extrusion processes.
• To develop ML models that predict and help reduce thickness variations, non-uniformities, and defects in extruded polymer products.

Main Methods:

• A dataset of 200 cases was generated using computer simulations of flat die extrusion.
• Variations included die geometry, polymer properties (rheological, thermophysical), and processing conditions (throughput rate, temperatures).
• Three ML algorithms—Random Forest (RF), XGBoost, and Support Vector Regression (SVR)—were trained to predict thickness variations, pressure drops, and wall shear rate.

Main Results:

• SHapley Additive exPlanations (SHAP) analysis identified key features influencing extrusion outcomes.
• Polymer rheology, throughput rate, and the gap beneath the restrictor were critical factors affecting thickness variations and pressure drops.
• The trained ML models accurately predicted target variables, demonstrating their predictive power.

Conclusions:

• ML-based methodology offers a promising approach to optimize flat die extrusion processes.
• This method has the potential to significantly reduce or eliminate the need for extensive trial-and-error procedures.
• Implementing ML can lead to more efficient and cost-effective production of high-quality polymer sheets and films.