Implementation of Machine Learning in Flat Die Extrusion of Polymers DOI Creative Commons
Nickolas D. Polychronopoulos, Ioannis E. Sarris, J. Vlachopoulos

et al.

Molecules, Journal Year: 2025, Volume and Issue: 30(9), P. 1879 - 1879

Published: April 23, 2025

Achieving a uniform thickness and defect-free production in the flat die extrusion of polymer sheets films is major challenge. Dies are designed for one scenario, grade with specified rheological behavior, given throughput rate. The different grades at flow rates requires trial-and-error procedures. This study investigated application machine learning (ML) to provide guidance reduced thickness, non-uniformities, without defects. A dataset 200 cases was generated using computer simulation software extrusion. encompassed variations geometry by varying gap under restrictor, thermophysical properties, processing conditions, including rate temperatures. used train evaluate following three powerful algorithms: Random Forest (RF), XGBoost, Support Vector Regression (SVR). ML models were trained predict variations, pressure drops, lowest wall shear (targets). Using SHapley Additive exPlanations (SHAP) analysis provided valuable insights into influence input features, highlighting critical roles rheology, rate, beneath restrictor determining targets. ML-based methodology has potential reduce or even eliminate use trial error

Language: Английский

Implementation of Machine Learning in Flat Die Extrusion of Polymers DOI Creative Commons
Nickolas D. Polychronopoulos, Ioannis E. Sarris, J. Vlachopoulos

et al.

Molecules, Journal Year: 2025, Volume and Issue: 30(9), P. 1879 - 1879

Published: April 23, 2025

Achieving a uniform thickness and defect-free production in the flat die extrusion of polymer sheets films is major challenge. Dies are designed for one scenario, grade with specified rheological behavior, given throughput rate. The different grades at flow rates requires trial-and-error procedures. This study investigated application machine learning (ML) to provide guidance reduced thickness, non-uniformities, without defects. A dataset 200 cases was generated using computer simulation software extrusion. encompassed variations geometry by varying gap under restrictor, thermophysical properties, processing conditions, including rate temperatures. used train evaluate following three powerful algorithms: Random Forest (RF), XGBoost, Support Vector Regression (SVR). ML models were trained predict variations, pressure drops, lowest wall shear (targets). Using SHapley Additive exPlanations (SHAP) analysis provided valuable insights into influence input features, highlighting critical roles rheology, rate, beneath restrictor determining targets. ML-based methodology has potential reduce or even eliminate use trial error

Language: Английский

Citations

0