Multi-objective optimization of surface roughness, dimensional errors and density in FFF 3D-printed glass fiber-reinforced PP parts via adaptive neuro-fuzzy inference modeling

Purpose This paper aims to investigate the effect of fused filament fabrication (FFF) 3D printing parameters on surface roughness, dimensional error and density of glass fiber-filled polypropylene (GF/PP) parts. It is a promising material to be used to obtain surgical models for bones, due to its high thermal and mechanical properties. Design/methodology/approach The experimental approach focuses on the manufacture of cuboid parts, measurement of surface roughness, dimensions and mass, and the use of adaptive neuro-fuzzy inference system (ANFIS) models to analyze the influence of 3D printing parameters such as printing temperature, print speed (PS), nozzle diameter (ND) and layer height (LH) on the responses. Multi-objective optimization using the desirability function is then applied to determine the optimal 3D printing parameters. In addition, the results of the ANFIS models were compared to other machine learning models such as genetic algorithm, support vector machine for regression and random forest for regression. Findings The research identifies the optimal 3D printing parameters recommended to simultaneously minimize surface roughness and dimensional error, and to maximize density. Specifically, the optimal parameters were identified as a low temperature of 230°C, a moderate print speed of 20 mm/s, a large nozzle diameter of 0.8 mm and a low layer height of 0.1 mm. Originality/value This study will help select appropriate 3D printing parameters for fabricating surgical models using GF/PP, a material that has only recently become commercially available for FFF processes, and has therefore been scarcely explored in the literature for such applications.