Artificial neural network-based predictive models for analyzing the flexural and compressive strength of PLA/carbon parts fabricated via material extrusion-based 3D printing

This study focuses on material extrusion-based (MEX) 3D printing of PLA reinforced with milled carbon fiber composite using the fused deposition modeling technique, structured through an L 16 orthogonal array design. The effects of key 3D printing input parameters (print temperature, layer thickness, infill pattern, and print speed) on flexural and compressive properties were recorded, including flexural force and stress and compressive force and stress. Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG) algorithm was developed to predict and evaluate the accuracy of the experiment. Regression analysis showed strong correlation between predicted and actual values, with LM recoding R-squared values of 0.99,983, 0.9809, and 0.98,968 for training, validation, and testing, respectively and 0.99,446, 0.98,119, and 0.98,876 for the SCG model. Both models effectively captured the relationships between input parameters and fracture toughness, with overall R-squared values of 0.99,506 for LM and 0.98,863 for SCG, indicating robust predictive performance across all datasets.