Anisotropy and internal flaws effects on fatigue response of notched 3D-printed PLA parts

In this study, the fatigue strength of notched 3D printed polylactic acid (PLA) samples were examined experimentally, considering the notch shapes, raster patterns, and internal voids/flaws between the filaments. Three different raster patterns were considered, which were 0, 45, and 90, and three different notch shapes were considered. The samples were subjected to five different cyclic loads ranging from 40 % to 80 % of the samples' ultimate tensile strength with a load ratio of 0.1. Results revealed that filaments could act as sharp cracks in the case of 90 & DEG; raster orientation. However, this effect in the cases of 0 & DEG; and 45 & DEG; was slightly smaller. Machine learning techniques including linear, polynomial and random forest (RF) regression models have been utilized for fatigue life predictions of the samples. R-squared values were utilized to compare the accuracy of the regression models. Among the regression models, polynomial regression best predicted fatigue lives of the notched samples with the R-squared value equal to 92 %. Predicted results with machine learning techniques were then compared with those obtained through an analytical approach by employing critical distance theory and using stress field and stress gradient distributions obtained from numerical simulations. It was found that all machine learning models except linear regression at high load levels predicted better than an approach based on analytical and numerical methods.