Machine learning driven optimisation of mechanical and microstructural behaviour in FFF-printed Onyx–HSHT composites

Additive Manufacturing (AM) is a transformative technology that allows the construction of lightweight, high-strength composite structures with customised geometries, making it mostly suitable for aerospace, defence, and structural applications. This study examines the mechanical behaviour of Onyx–High-Strength High-Toughness (HSHT) fibre composites fabricated using Fused Filament Fabrication (FFF). It establishes an integrated optimisation outline combining experimental testing, statistical analysis, machine learning (ML), and scanning electron microscopy (SEM). Tensile, flexural, and impact tests revealed significant adaptability in performance, with tensile strength of 177.50 MPa, flexural strength of 68.9 MPa, and impact energy of 7.487 J. Meanwhile, Analysis of Variance (ANOVA) established substantial sensitivity of mechanical properties to processing parameters, with F-values of 21.54, 70.39, and 294.15, respectively. Predictive ML models, predominantly the Random Forest Regressor, revealed high accuracy (R2 up to 0.996), and interactive linear regression effectively modelled flexural strength. SEM analysis explained critical microstructural factors, such as void morphology and fibre–matrix interface quality, that influence mechanical performance. The combined DOE–ANOVA–ML–SEM approach provides a robust, data-driven methodology for the practical prototyping and optimisation of continuous fibre-reinforced AM composites, enabling accelerated design and improved performance prediction of advanced composite materials for practical engineering applications.