Effect of infill and density pattern on the mechanical behaviour of ABS parts manufactured by FDM using Taguchi and ANOVA approach

Purpose: The present work aims to investigate the effect of many infill patterns (rectilinear, line, grid, triangles, cubic, concentric, honeycomb, 3D honeycomb) and the infill density on the mechanical tensile strength of an Acrylonitrile Butadiene Styrene (ABS) test specimen manufactured numerically by FDM. Design/methodology/approach: Computer-Aided Design (CAD) software has been used to model the geometry and the mesostructure of the test specimens in a fully automatic manner from a G-code file by using a script. Then, a Numerical Design of Experiments (NDoE) has been carried out by using Taguchi method and the Analysis of Variance (ANOVA). The tensile behaviour of these numerical test specimens has been studied by the Finite Element Analysis (FEA). Findings: The FEA results showed that a maximal Ultimate Tensile Strength (UTS) was reached by using the ‘concentric’ infill pattern combined with an infill density of 30%. The results also show that the infill pattern and the infill density are significant factors. Research limitations/implications: The low infill densities of 20% and 30% that have already been used in many previous studies, we have also applied it in order to reduce the time of the simulations. Indeed, with high infill density, the simulations take a very excessive time. In an ongoing study, we predicted higher percentages. Practical implications: This study provided an important modelling tool for the design and manufacture of functional parts and helps the FDM practitioners and engineers to manufacture strong and lightweight FDM parts by choosing the optimal process parameters. Originality/value: This study elucidated the effect of various infill patterns on the tensile properties of the test specimens and applied for the first time a NDoE using numerical test specimens created by the mesostructured approach, which considerably minimized the cost of the experiments while obtaining an error of 6.8% between the numerical and the experimental values of the UTS.