Machine Learning for Identifying Damage and Predicting Properties in 3D-Printed PLA/Lygeum spartum Biocomposites

This paper offers an experimental approach to use machine learning (ML) models and Acoustic Emission (AE) data for the identification of damage mechanisms and predict the mechanical properties of 3D printed bio-composite. The specimens were produced using a bio-filament made of a PLA matrix reinforced with 10% wt. of Lygeum spartum fibers. AE signals were gathered during tensile and flexural tests in order to monitor the progression of damage under mechanical loading. subsequently, using Random Forest Regression (RFR), Support Vector Regression (SVR), Artificial Neural Network (ANN), and Decision Tree (DT) models, the stress levels of the specimens under both test conditions were predicted. These algorithms were trained with 80% of the data, in a Python environment. 20% remained, to be used for testing. The models' accuracy was assessed using R-squared (R²) and Mean Squared Error (MSE) metrics. While the other models also demonstrated outstanding prediction capabilities for both tensile and flexural stresses, the RFR model outperformed the others. In addition, 5-fold cross-validation yielded results consistent with the hold-out test, further validating the models' accuracy. This research demonstrates how well these machine learning algorithms analyze AE data for material property evaluation, paving the way for data-driven approaches in material testing and health monitoring.