Mechanical characterization and ML-based prediction of 3D-printed PLA+: effects of infill pattern, infill density, and orientation

Abstract Additive manufacturing (AM) enables the fabrication of lightweight, customizable components. Among its technologies, Fused Filament Fabrication (FFF) stands out for simplicity, low cost, and versatility. This work investigates how infill pattern, density, and raster angle affect PLA+ specimens using a combined experimental–computational approach that integrates mechanical testing, statistical analysis (ANOVA), and machine learning (ML). Specimens were printed with a Prusa i3 MK3 + under FFF conditions, evaluating four infill patterns (Honeycomb, Grid, Triangles, Gyroid), four density levels (20–80%), and three raster angles (0°, 45°, 90°) in a full factorial design. Tensile, impact, and stiffness tests showed infill density as the dominant factor, accounting for 61.7% of the variance in tensile strength and 57.1% in impact strength, followed by infill pattern and raster angle. Significant interactions ( p < 0.05) confirmed that mechanical behaviour depends on parameter combinations rather than individual factors. The highest tensile strength (52.3 MPa) and elongation at break (9.8%) were obtained with the Honeycomb pattern at 80% density and 0° raster angle, while impact strength peaked at 45° raster angle. The main contribution is the integration of a systematic full-factorial experimental design with ML models (CatBoost and XGBoost) to analyze and predict the mechanical response of FFF-printed structures. The results establish a hierarchy of parameter influence: infill density governs mechanical performance, followed by infill pattern and raster angle. This framework identifies parameter combinations that balance strength, impact strength, and manufacturability, while predicting the mechanical properties of untested printing configurations.