Prediction of residual stress and hybrid layer location for additively manufactured samples using ultrasonic spectroscopy and machine learning

Authors

• Jazmin Ley, The University of Texas Rio Grande ValleyFollow
• John Greenhall, Los Alamos National Laboratory
• Milo Prisbrey, Los Alamos National Laboratory
• Cristian Pantea, Los Alamos National Laboratory
• Joseph A. Turner, University of Nebraska - Lincoln

Document Type

Article

Publication Date

10-2026

Abstract

Current methods for measuring residual stress in additively manufactured components face limitations in sensitivity or are destructive. Resonant ultrasound spectroscopy (RUS) offers a promising approach for assessing residual stresses in hybrid additively manufactured components. Hybrid additive manufacturing (AM) involves using secondary energy sources or processes to create functionally graded components at specific locations. Each hybrid layer or step refines grain size, increases dislocation density, and consequently alters residual stresses. This study employs finite element models to simulate residual stress in hybrid AM components. Using machine learning (ML), we predict both stress levels and the location of hybrid layers. A total of 1000 simulations generated a sufficiently large dataset for ML models. Partial least squares regression (PLSR) predicts residual stress with a mean absolute error (MAE) of approximately 3 MPa, which is less than 1% of residual stresses found in hybrid AM layers. Simultaneously, the location of the hybrid layer exhibits degeneracies in the resonant modes, necessitating a more robust algorithm. The ML models are able to predict the location of a hybrid layer within 0.14 mm using 80% of the data as a training set. This work lays the groundwork for future noninvasive and nondestructive measurements of residual stresses through experimental RUS combined with finite element modeling and machine learning.

Comments

Original published version available at https://doi.org/10.1016/j.ndteint.2026.103820

Recommended Citation

Ley, Jazmin, John Greenhall, Milo Prisbrey, Cristian Pantea, and Joseph A. Turner. "Prediction of residual stress and hybrid layer location for additively manufactured samples using ultrasonic spectroscopy and machine learning." NDT & E International (2026): 103820. https://doi.org/10.1016/j.ndteint.2026.103820

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.

Publication Title

NDT & E International

DOI

10.1016/j.ndteint.2026.103820