Melt pool width prediction with machine learning in selective laser melting

Date

2025-8

Author

Gülletutan, Umut Can

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

64
views

0
downloads

Cite This

This study presents a comprehensive investigation into the prediction of melt pool width in Laser Powder Bed Fusion (LPBF) by integrating analytical modeling and data-driven techniques. Initially, the classical Rosenthal equation was applied to estimate melt pool widths across a literature-compiled dataset of 1206 entries, covering five common alloys. However, due to its simplifying assumptions which are neglecting convection heat transfer and temperature-dependent properties the Rosenthal model consistently overestimated melt pool width, particularly under keyhole-mode conditions. To improve its predictive accuracy, the machine learning model revealed that laser power, scanning speed, and beam diameter were the influential parameters. Based on these findings, the Rosenthal equation was modified by introducing exponent-based tuning on these variables, which reduced the NMAE from 49.49% to 18.79%. To further enhance prediction capabilities, a Random Forest regression model was developed using process parameters and intrinsic material properties. The model achieved an NMAE of 8.5% on the test set, demonstrating strong generalization. External validations were conducted on SS316L and a CoFeNiCuMn high-entropy alloy (HEA), with model predictions significantly outperforming. HEA-specific material properties were derived using Thermo-Calc simulations and UV–VIS spectroscopy, enabling the model to accurately predict melt pool width even for previously untrained materials. Feature importance and SHAP analyses confirmed process parameters were the most influential predictors, while absorptivity emerged as the dominant material property. This hybrid framework offers a robust and interpretable approach for melt pool prediction, supporting improved process control and accelerated qualification of new alloys in metal additive manufacturing.

Subject Keywords

Laser Powder Bed Fusion (LPBF), Melt Pool Width, Rosenthal Equation, Machine Learning, High-Entropy Alloy (HEA)

URI

https://hdl.handle.net/11511/117361

Collections

Graduate School of Natural and Applied Sciences, Thesis