Optimization and predictive modelling of mechanical characteristics of 3D printed olefin block copolymer polyethylene: RSM and machine learning methods

Abstract

The mechanical behavior of FDM-printed Olefin Block Copolymer (OBC) polyethylene is examined and optimized in this work, using an integrated experimental, statistical, and machine learning framework. The effects of infill density (20–80%), nozzle temperature (200–240°C), and printing speed (40–60 mm/s) on tensile and flexural strength were investigated using a Central Composite Design within the Response Surface Methodology (RSM) framework. The results indicate that infill density is the most dominant parameter, followed by printing speed, while nozzle temperature shows a comparatively minor effect. Optimal conditions were achieved with a maximum tensile strength of 37.21 MPa and flexural strength of 58.4 MPa. Coefficients of prediction exceeding 0.998 for both tensile and flexural responses were obtained from the RSM regression model. Four supervised machine learning models, including Linear Regression, Polynomial Regression, Random Forest, and Support Vector Regression, were implemented and evaluated to further improve predictions. Nonlinear models, particularly Polynomial Regression and Random Forest, achieved near-perfect predictive accuracy (R² = 0.99) with low error metrics. The dominance of infill density was quantitatively confirmed by explainable artificial intelligence using SHAP values, which provided a clear physical interpretation consistent with the ANOVA results and experimental observations. The suggested framework offers a valid approach to optimizing the process parameters and developing mechanically efficient FDM-printed OBC parts for advanced engineering applications.