Additive Manufacturing

This study develops a data-driven framework to predict melt pool geometry in directed energy deposition (DED), linking process parameters (laser power, scan speed, feed rate) to geometrical outcomes (width, depth, height). Through experimental datasets and comparative modeling, the authors show that Gaussian Process Regression (GPR) outperforms other ML models in low-data regimes, offering both high accuracy and uncertainty-aware predictions. The work highlights ML’s role as a practical decision-support tool for process optimization, bridging experimental observations with predictive modeling in additive manufacturing where physics-based models are complex and computationally expensive.