ABSTRACT
This study intends to contribute to the state of the art of Fused-Filament Fabrication (FFF) of short-fiber-reinforced polyamides by optimizing process parameters to improve the performance of printed parts under uniaxial tensile loading. This was performed using two different approaches: a more traditional 2k full factorial design of experiments (DoE) and multiple polynomial regression using an algorithm implementing machine learning (ML) principles such as train-test split and cross-validation. Evaluated parameters included extrusion and printing bed temperatures, layer height and printing speed. It was concluded that when exposed to new observations, the ML-based model predicted the response with higher accuracy. However, the DoE fared slightly better at predicting observations where higher response values were expected, including the optimal solution, which reached an UTS of 117.1 ± 5.7 MPa. Moreover, there was an important correlation between process parameters and the response. Layer height and printing bed temperatures were considered the most influential parameters, while extrusion temperature and printing speed had a lower influence on the outcome. The general influence of parameters on the response was correlated with the degree of interlayer cohesion, which in turn affected the mechanical performance of the 3D-printed specimens.
