Predictive machine learning and printing parameter optimization for enhanced impact performance of 3D-printed Onyx-Kevlar composites

Additive manufacturing enables the design of lightweight, high-performance composites customised for specific mechanical applications. This study aims to enhance the impact resistance of 3D-printed Onyx-Kevlar fiber-reinforced composites by optimizing printing parameters and predictive machine learning (ML) modeling. A Taguchi L27 orthogonal array was used to systematically examine the influence of fiber orientation, volume fraction, infill density, and pattern on impact performance. Statistical analysis identified fiber volume fraction and infill density as the most significant factors. To reduce experimental load and enable accurate property estimation, supervised ML models, Support Vector Regression (SVR), and Linear Regression, were developed and trained on experimental data. Both models demonstrated high predictive performance, with R² values of 0.9166 and 0.9747, respectively. Model accuracy was further validated through learning curves, residual analysis, and performance metrics. Scanning Electron Microscopy (SEM) provided microstructural insights into fracture behavior, correlating with model predictions. he combined approach provides a reliable and efficient method to optimize composite performance with less trial-and-error experimentation. This study demonstrates the utility of predictive ML in guiding material design and highlights its applicability in developing impact-resistant composites for protective and structural engineering applications.