Integrated statistical analysis and machine learning-based predictive modelling of mechanical properties in 3D printed acrylonitrile butadiene styrene via material extrusion for unmanned aerial vehicle applications

Abstract

Additive manufacturing (AM) through fused filament fabrication (FFF) offers great potential for producing lightweight and customized components for unmanned aerial vehicles (UAVs). However, achieving consistent mechanical performance remains challenging due to nonlinear interactions among control parameters (CPs). This study investigates the influence of deposited raster height (DRH), extrusion nozzle temperature (ENT), and raster deposition rate (RDR) on the mechanical properties of acrylonitrile butadiene styrene (ABS). Twenty-seven ASTM-compliant specimens (D638, D695, D790) were fabricated and tested for tensile strength (TS), compressive strength (CS), and flexural strength (FS). The measured values ranged from 22.08–33.81 MPa (TS), 38.29–185.71 MPa (CS), and 45.32–90.04 MPa (FS). Taguchi's L27 orthogonal array and analysis of variance (ANOVA) identified DRH as the dominant factor for TS (35.43%, F = 7.44, P = 0.015) and CS (61.75%, F = 85.86, P = 0), whereas ENT primarily influenced FS (70.52%, F = 323.8, P = 0). Eleven machine learning (ML) algorithms were developed, among which extreme gradient boosting regression (XGBR) achieved the highest accuracy for TS (coefficient of determination (R²) = 0.9381) and CS (R² = 0.9597), while gradient boosting regression (GBR) performed best for FS (R² = 0.9826). The optimized parameter combination (0.1 mm DRH, 20 mm/s RDR, 240 °C ENT) was applied to fabricate UAV arms. Finite element analysis (FEA) indicated a maximum von Mises stress of 19.49 MPa and deformation of 10.632 mm, confirming structural reliability under service loads. This integrated Taguchi–ANOVA–ML–FEA framework provides a robust pathway for optimizing polymer-based UAV components produced via FFF.