Comparative Analysis of Machine Learning Methods for Predicting Tensile Strength of Dual-Extruded FDM 3D- Printed ABS/PC Composite

The mechanical behaviour of FDM composite materials is largely influenced by process variables, material composition, & inter-layer bonding. In order to predict the tensile strength of dual-extruded FDM composites fabricated from alternating layers of Acrylonitrile Butadiene Styrene (ABS) and Polycarbonate (PC) using a dual-extrusion printer, this study compares various Machine Learning (ML) algorithms. Such kind of systematic comparison of ML modelling on an alternating-layer ABS/PC composite along with the use of permutation importance analysis has never been performed before. The following process parameters are taken into account: layer height (0.20, 0.25, 0.30 mm), nozzle temperature (250, 255, 260 °C), print speed (50, 55, 60 mm/s), shell/wall layers (2, 3, 4), and bed temperature (90, 95, 100 °C). The experimental design is based on Taguchi L27 orthogonal array and ASTM D638 is used for preparing the tensile specimens. The tensile results range between 22.98 MPa and 49.38 MPa, indicating a strong influence of process parameters. Four ML regression models-Linear Regression (LR), Support Vector Regression (SVR), Gradient Boosting (GB), and AdaBoost are used, with the best model selected based on the highest Coefficient of Determination (R²) and the lowest RMSE value. LR achieves the highest R² value of 0.4595, followed by AdaBoost (0.4478), SVR (0.4211), and GB (0.3947). Learning-curve analysis reveals that ensemble models are overfitting. Permutation importance, which is model-independent, predicts layer thickness as the most important parameter. The results show that data-driven modelling can provide reasonable predictions of the tensile strength of dual-extrusion FDM composites.