Tensile performance prediction of multilayer FDM-printed polymers using Taguchi design and neural networks

Abstract

The effects of process parameters on monolithic fused deposition modeling-printed polymers have been extensively investigated in the literature, while the optimization of tensile performance and development of predictive models of multilayer laminated polymer structures have received limited attention. Thus, the current research investigated the synergistic effects of infill density, printing speed, and wall thickness on the tensile behavior of multilayer samples made of polylactic acid/acrylonitrile butadiene styrene/polyethylene terephthalate glycol by fused deposition modeling. The Taguchi L9 orthogonal design, statistical analyses, and artificial neural network-based prediction models were used. Three main process parameters were studied: infill density (30%–90%), printing speed (50–150 mm/second), and wall thickness (2–6 mm). The maximum tensile stress of 48.74 MPa was achieved at an infill density of 90%, a printing speed of 50 mm/second, and a wall thickness of 6 mm. Analysis of variance and signal-to-noise ratio analyses revealed that wall thickness had the most significant effect (38.74%), followed by infill density (34.09%) and printing speed (26.13%). The results of regression and stress–strain analyses also confirmed the effects of the selected parameters on mechanical performance. The artificial neural network model demonstrated superior prediction capability accuracy ( R 2  = 0.9956) as compared to the linear regression and Taguchi estimation. Scanning electron microscopy observations showed better interlayer bonding and reduced internal void formation for specimens with better tensile performance. The proposed experimental–statistical approach is an effective approach for the optimization of multilayer fused deposition modeling-printed polymer structures, and may facilitate the design of lightweight and mechanically reliable components for engineering applications.