Experimental Investigation and Artificial Neural Network-Based Prediction of Tensile Strength in Fused Filament-Fabricated Carbon Fiber-Reinforced PETG

Abstract

Fused filament fabrication (FFF) has become an important additive manufacturing technique for producing functional polymer-composite components. The tensile performance of carbon fiber-reinforced polyethylene terephthalate glycol (PETG/CF) fabricated by FFF depends on multiple printing parameters. This study presents an integrated experimental and predictive framework for investigating the effects of extrusion temperature, printing speed, layer height, infill pattern, and infill density on the tensile strength of PETG/CF containing 15 wt.% carbon fiber. A mixed-level Taguchi L36 orthogonal array was employed, comprising 36 experimental runs with three independently printed specimens per run, resulting in 108 ASTM D638 Type V specimens. Analysis of variance showed that the printing speed had the largest contribution to tensile strength (20.51%), followed by layer height (18.29%). The highest tensile strength of 33.225 MPa was obtained using grid infill, 60% infill density, 270 °C extrusion temperature with 40 mm/s printing speed, and 0.3 mm layer height. An artificial neural network (ANN) was developed for the tensile-strength prediction, achieving R = 0.9801, R2 = 0.9569, and MAPE = 1.52% for the overall dataset. Scanning electron microscopy qualitatively revealed bead-interface defects, fiber pullout, and localized void-like features. The proposed framework provides a systematic approach for evaluating process-parameter effects and predicting tensile strength within the investigated PETG/CF parameter domain.