On the Prediction of Mechanical and Aesthetical Behavior of AM Specimens Through Machine Learning: A Preliminary Study

3D printing, including the widely used technique of Fused Filament Fabrication (FFF), offers several benefits, such as freedom in design, faster production times, customization, and lightweight. In fact, it has the potential to serve a variety of sectors, from the everyday products market to the industry. However, the properties of FFF parts, in terms of mechanical strength and geometrical accuracy, are strongly influenced by printing process parameters, necessitating thorough investigations. This preliminary study proposes a comparison between data-driven Machine Learning (ML) models, i.e. Quadratic Regression (QR), Regression Tree (RT), and Neural Network (NN), to predict the impact of printing process parameters on both mechanical (tensile properties) and aesthetical (bending angle, thickness, and flatness tolerance) behavior of additively manufactured tough-Polylactic Acid (PLA) specimens. The investigated models were trained by using experimental data obtained from Design of Experiment (DoE) where the building orientation on plate, the infill percentage and the pattern were varied. An additional dataset was used to test the accuracy of the prediction. The results suggest that ML models perform well in predicting the tensile properties, particularly the Ultimate Tensile Stress (UTS). At the same time, further investigation is necessary to improve the prediction for aesthetical features.