Multi-target machine learning for predicting mechanical properties of FDM-printed polymer components.

Additive manufacturing using Fused Deposition Modelling (FDM) is increasingly adopted for producing polymer components; however, achieving consistent and repeatable mechanical performance remains a challenge due to process-induced variability and material behavior. Existing studies have predominantly focused on predicting individual mechanical properties using single-output machine learning (ML) models, which limits their ability to capture interdependencies among multiple responses. The present study addresses this gap by investigating whether a Multi-Target Machine Learning (MTML) framework can effectively predict multiple mechanical properties of FDM-fabricated polymer components simultaneously. Experimental datasets were generated from tensile, hardness, and impact tests conducted on specimens fabricated using Polylactic Acid (PLA), Acrylonitrile Butadiene Styrene (ABS), and Polyethylene Terephthalate Glycol-modified (PETG) materials. Several regression-based ML models, including K-Nearest Neighbour (KNN), Decision Tree (DT), Random Forest (RF), Extra Trees (ET), Gradient Boosting (GB), and Extreme Gradient Boosting (EGB), were implemented within both single-target and multi-target learning paradigms. The predictive performance of the models was evaluated using standard statistical metrics, including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R2-score. The results demonstrate strong agreement between predicted and experimental values, with R2 values ranging from 0.74 to 0.998, indicating the effectiveness of the MTML framework in capturing complex, non-linear relationships among mechanical responses. The study confirms that the proposed MTML approach improves predictive reliability and modeling efficiency compared to conventional single-output strategies. The findings contribute to advancing data-driven predictive modeling in FDM-based additive manufacturing and provide a robust foundation for future applications in process optimization, quality assessment, and intelligent manufacturing systems. The results demonstrate strong predictive capability, with R2 values ranging from approximately 0.74 to 0.998, depending on the material, mechanical property, and regression model. Among the evaluated algorithms, Gradient Boosting Regression (GBR) consistently achieved the highest accuracy for ductility-related properties such as reduction in area and elongation, while Extra Trees Regression (ETR) and GBR showed robust performance for strength-related properties across multiple materials. Overall, the study confirms that the proposed MTML framework improves prediction reliability and modelling efficiency compared to conventional single-output approaches. The findings provide a data-driven foundation for mechanical property estimation in FDM and highlight GBR as the most efficient and reliable regression algorithm within the investigated experimental scope.