DataDriven Modeling of Bead Morphology in Wire Arc Additive Manufacturing: A Machine Learning Approach

The geometric morphology of individual weld beads is a fundamental determinant of structural precision and surface integrity in Wire Arc Additive Manufacturing (WAAM). This study evaluates the predictive capabilities of Random Forest Regression (RFR) and Multiple Linear Regression (MLR) for forecasting bead width and height using a robotic GMAW experimental dataset. Comparative analysis demonstrates that the RFR model significantly outperforms the linear approach, achieving a higher R2 score of 0.84 and a lower Mean Squared Error (MSE) of 0.127. This superior performance is attributed to the ensemble model's ability to capture the non-linear thermal dynamics and stochastic melt pool behavior inherent in the process. These results establish the RFR framework as a precise computational tool for adaptive slicing strategies, enabling enhanced dimensional control in the fabrication of complex metallic components.