Data-driven prediction of tensile strength and energy consumption in MEX using machine learning models

Machine Learning (ML) techniques were employed to predict Tensile Strength (TS) and Energy Consumption (EC) in the Material EXtrusion (MEX) process. Four key process parameters were considered: Infill Percentage (IP), Layer Height (LH), Number of Walls (N w ), and Bed Temperature (BT). Several supervised ML models were trained and evaluated, including Linear Regression (LR), Regression Trees (RTs), Ensemble of RTs (ERTs), Support Vector Machine (SVM), Kernel Regression (KR), and Artificial Neural Network (ANN). Each model was assessed using five-fold cross-validation, and a separate test set was used to determine the model’s ability to capture the relationship between input parameters and output responses. Among the models investigated, the ERTs consistently achieved the highest predictive performance, with R² values exceeding 99% for EC and 85% for TS, respectively. The obtained results highlight the potential of ML to enhance prediction of mechanical properties and energy consumption in the MEX process using data obtained through Design Of Experiments (DOE). • Machine Learning (ML) predicts tensile strength and energy use in FDM processes. • Four FDM parameters studied: IP, LH, Nw, and BT. • Six ML models trained with Cross-Validation and test set evaluation. • Ensemble of Regression Trees gave best results: R 2 > 99% (EC), > 85% (TS). • ML shows strong potential for DOE-based FDM prediction.