Enhancing the Prediction of Mechanical Properties of 3D-Printed Parts Through Machine Learning

Abstract

This study investigates the application of the XGBoost machine learning algorithm to predict the mechanical properties of 3D printed polylactic acid (PLA) specimens, specifically focusing on bending elasticity (BE) and bending strength (BS). Using a face-centered central composite design (FCCCD) with 20 experiments, the specimens were subject to bending loading to determine their mechanical properties. The dataset was divided into 70% for training and 30% for testing to build and validate the predictive models. The XGBoost algorithm demonstrated exceptional performance in predicting BE, achieving R2 values of 0.999 for training data and 0.99 for test data, with significantly low mean squared error (MSE) and root mean squared error (RMSE). For BS, the model also showed strong performance, with R-squared values of 0.99 for training data and 0.93 for test data. These results indicate that XGBoost outperforms traditional Response Surface Methodology (RSM) approaches, offering higher accuracy and lower error rates. The correlation heatmap provided further insights into the relationships between process parameters and bending strength, aiding in the optimization of the 3D printing process. This study highlights the potential of advanced machine learning techniques in enhancing the predictability and optimization of 3D printing technologies.