3D printing using fused deposition modeling (FDM) includes a multitude of control parameters. It is difficult to predict a priori what surface finish will be achieved when certain values are set for these parameters. The objective of this work is to compare the models generated by decision tree algorithms (C4.5, random forest, and random tree) and to analyze which makes the best prediction of the surface roughness in polyethylene terephthalate glycol (PETG) parts printed in 3D using the FDM technique. The models have been created using a dataset of 27 instances with the following attributes: layer height, extrusion temperature, print speed, print acceleration, and flow rate. In addition, a dataset has been created to evaluate the models, consisting of 15 additional instances. The models generated by the random tree algorithm achieve the best results for predicting the surface roughness in FDM parts.

The fused deposition modeling (FDM) technique is used today by companies engaged in the fabrication of traffic signs for the manufacture of light-emitting diode LED spotlights. In this sector, the surface properties of the elements used (surface finish, hydrophobic features) are decisive because surfaces that retain little dirt and favor self–cleaning behavior are needed. A design of experiments (L27) with five factors and three levels has been carried out. The factors studied were: Layer height (LH), print temperature (T), print speed (PS), print acceleration (PA), and flow rate (F). Polyethylene terephthalate glycol (PETG) specimens of 25.0 × 25.0 × 2.4 mm have been printed and, in each of them, the surface roughness (R_a,0, R_a,90), sliding angle (SA₀, SA₉₀), and contact angle (CA₀, CA₉₀) in both perpendicular directions have been measured. Taguchi and ANOVA analysis shows that the most influential variables in this case are printing acceleration for R_{a, 0} (p–value = 0.052) and for SA₀ (p–value = 0.051) and flow rate for R_{a, 90} (p–value = 0.001) and for SA₉₀ (p–value = 0.012). Although the ANOVA results for the contact angle are not significant, specimen 8 (PA = 1500 mm/s² and flow rate F = 110%) and specimen 10 (PA =1500 mm/s² and F = 100%) have reached contact angle values above or near the limit value for hydrophobia, respectively.

In the present work, ten data mining algorithms have been used to generate models capable of predicting the surface roughness of parts printed on polylactic acid (PLA) by using fused deposition modeling (FDM). The models have been trained using experimental data measured on 27 horizontal (XY) and 27 vertical (XZ) specimens, printed using different values for the parameters studied (layer height, extrusion temperature, print speed, print acceleration and flow). The models generated by multilayer perceptron (MLP) and logistic model trees (LMT) have obtained the best results in a cross-validation. Although it does not obtain such optimal results, the J48 algorithm (C4.5) allows the generation of models in the form of a decision tree. These trees permit to determine which print parameters have an influence on the surface roughness. For XY specimens, the surface roughness measured in the direction parallel to the extrusion path (R_a,0,XY ) depends on the flow, the print temperature and the layer height; in the direction perpendicular to the extrusion path, the surface roughness (R_a,90,XY) depends only on the flow. For XZ specimens, the surface roughness measured in the direction parallel to the extrusion path (R_a,0,XZ) depends only on the print speed; in the direction perpendicular to the extrusion path (R_a,90,XZ), it depends on the layer height and the extrusion temperature. According to the study carried out, the most suitable set up provides values of R_a,0,XY, R_a,90,XY, R_a,0,XZ and R_a,90,XZ equal to 0.46, 1.18, 0.45 and 11.54, respectively. A practical application of this work is the manufacture of PLA frame glasses using FDM.