ORGS5-1491
Machine Learning-Based Optimization and Experimental Validation of Maleic Anhydride Compatibilizer Composition in PP/HDPE PCR Blends
Topic
GS5. Graduate Student Oral Session V: Sustainable Polymers and Circular Materials
When and Where
Sep 28, 2026
15:12 - 15:24
Room 105
Session Chairs
Taehoo CHANG
Taejun EOM
Chae Bin KIM
Presenter(s)
Sungwoong Choi (KITECH)
Co-Author(s)
Abstract
The heterogeneity of raw materials and the non-linear changes in physical properties that occur during the upcycling of Post-Consumer Recycled (PCR) plastics are major factors that reduce the efficiency of conventional experimental approaches and threaten the reliability of data-driven predictive models. In this study, we propose an integrated workflow combining statistical data purification with machine learning to maximize the tensile strength of PP/HDPE blends and improve prediction accuracy. Outliers in the experimental data were first removed to secure data reliability, and a regression model was then constructed to predict tensile strength from the blend composition and the contents of the compatibilizer (MA) and initiator (DCP). In addition, model-interpretation techniques were applied to examine the contribution of each factor to the mechanical properties, and an optimization using the predictive model as the objective function was employed to reverse-engineer high-strength compositions, which are to be verified through actual synthesis. The integrated machine learning framework proposed in this study is expected to provide a practical methodology for overcoming the property variability of recycled waste plastics and accelerating the development of high-performance PCR blends.
