Orientation-dependent residual stress governs anisotropic warpage in laser powder bed fusion stainless steel: Experimental analysis and predictive modelling

Page view(s)
0
Checked on
Orientation-dependent residual stress governs anisotropic warpage in laser powder bed fusion stainless steel: Experimental analysis and predictive modelling
Title:
Orientation-dependent residual stress governs anisotropic warpage in laser powder bed fusion stainless steel: Experimental analysis and predictive modelling
Journal Title:
Additive Manufacturing
Keywords:
Publication Date:
11 March 2026
Citation:
Wang, C., Sun, C.-N., Chang, T., Liang, X., Ma, N., & Wang, P. (2026). Orientation-dependent residual stress governs anisotropic warpage in laser powder bed fusion stainless steel: Experimental analysis and predictive modelling. Additive Manufacturing, 121, 105158. https://doi.org/10.1016/j.addma.2026.105158
Abstract:
Anisotropic warpage is one of the major challenges in laser powder bed fusion (LPBF), especially for overhanging and thin-walled structures. This study demonstrates that orientation-dependent residual stress, arising from the coupled process environment, is the key driver of warpage in LPBF stainless steel cantilever beams. X-ray diffraction and contour mapping revealed that different build orientations relative to the fixed recoater–gas-flow configuration lead to distinct residual-stress states and deformation modes. Builds aligned with the recoater direction exhibited higher dislocation densities and shear-induced stresses, resulting in thickness-sensitive warpage while builds perpendicular to recoater direction showed more uniform cooling and bending-dominated deformation. Machine learning (ML) models trained on experimental data captured nonlinear geometry-warpage relationships. The tree models (Random Forest and XGBoost) achieved the highest accuracy in average with k-fold validations when predicting maximum deflection, while the Radial Basis Function was the most reliable model to predict polynomial fitting parameters. These supervised learning models demonstrated their effectiveness for nonlinear and small-size dataset cases with consistent R2 values and Mean Absolute Error values with low standard deviations. The integrated experimental-ML approach offers a predictive tool for optimizing build orientation and geometry in LPBF, highlighting the importance of residual stress anisotropy in distortion control.
License type:
Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
Funding Info:
This research / project is supported by the Agency for Science, Technology and Research - Manufacturing, Trade, and Connectivity Programmatic Fund - Advanced Modelling Models for Additive Manufacturing
Grant Reference no. : M22L2b0111

This research / project is supported by the Agency for Science, Technology and Research - Career Development Fund
Grant Reference no. : C210112043
Description:
ISSN:
2214-8604
Files uploaded:

File Size Format Action
r1-manuscript.pdf 2.89 MB PDF Request a copy