Artificial Neural Network-Based Structural Analysis of 3D-Printed Polyethylene Terephthalate Glycol Tensile Specimens

Abstract

Materials are a mainstay of both industry and everyday life. The manufacturing and processing of materials is a very important sector as it affects both the mechanical properties and the usage of the final products. In recent years, the increased use of 3D printing and, by extension, its materials have caused the creation of gaps in terms of strength that require further scientific study. In this study, the influence of various printing parameters on 3D-printed specimens made of polyethylene terephthalate glycol (PETG) polymer was tested. More specifically, three printing parameters were selected—infill, speed, and type—with three different values each (50%, 70%, and 90%), (5 mm/s, 20 mm/s, and 35 mm/s) and (Grid, Rectilinear, and Wiggle). From the combinations of the three parameters and the three values, 27 different specimens were obtained and thus, 27 equivalent experiments were designed. The measurements were evaluated, and the process was modeled with the Artificial Neural Network (ANN) method, revealing a strong and robust prediction model for the tensile test, with the relative error being below 10%. Both infill density and infill pattern were identified as the most influential parameters, with the Wiggle type being the strongest pattern of all. Additionally, it was found that the infill density acts increasingly on the strength, whereas the printing speed acts decreasingly.