Predictive modeling and experimental validation of mechanical-microstructural relationships in 3D-printed Onyx-fibre...

Additive Manufacturing (AM) has emerged as a revolutionary fabrication technology for producing high-performance polymer-based composite materials. Fused Filament Fabrication (FFF), a leading AM technology, facilitates the fabrication of continuous fibres in a thermoplastic matrix to provide a customised mechanical response. In this study, the tensile and flexural properties of Onyx matrix composites reinforced with continuous carbon fibre (CF) and glass fibre (GF) are investigated. Specimens were fabricated using a Markforged 3D printer and tested in accordance with ASTM standards. A Taguchi L27 design was employed to examine the effect of infill density, fibre volume fraction, fibre orientation, and roof and floor layers. The principal results indicated maximum tensile strengths of 286.8 MPa (CF) and 183.2 MPa (GF), and a flexural strength of 208.4 MPa (CF). ANOVA indicated the significance of the variables (high F-values, low residuals), especially in the CF flexural and GF tensile models. The regression analysis provided high predictive accuracy for all models, with predicted R-squared values of 98.73% (CF flexural), 97.69% (GF tensile), 91.04% (GF flexural), and 84.22% (CF tensile). Separate machine learning models trained on experimental data further predicted mechanical properties: a linear regression model provided R² = 99.49% for flexural data, while a multi-output random forest model predicted tensile strength and elongation with R² = 99.38% and 96.95%, respectively. SEM analysis of the fractured specimens validated failure modes, such as fibre pullout, matrix cracking, and delamination in CF (brittle) compared to ductile debonding in GF. The integrated Taguchi-ANOVA, machine learning, and microstructural analysis offers a robust platform for optimising the performance of FFF-fabricated fibre-reinforced Onyx composites.