Page 75 - IJAMD-2-2

International Journal of AI for
Materials and Design

ORIGINAL RESEARCH ARTICLE
Prediction of the lack-of-fusion defect of laser

powder bed fusion based on deep learning

Lidong Wang*
Institute for Systems Engineering Research, Mississippi State University, Mississippi, United States
of America
(This article belongs to the Special Issue: Applications of Deep Learning in Advanced Materials
Processing)

Abstract
Laser powder bed fusion (LPBF) is one of the additive manufacturing (AM) techniques
and the most studied laser-based AM process for metals and alloys. The optimization
of the laser process parameters of LPBF and the prediction of defects, for example,
keyholes, cracks, and lack of fusion (LOF), are important for improving the quality
of products made with LPBF. Deep learning (DL) is powerful in analyzing complex
processes and predicting anomalies; however, much data is generally required for
training a DL model. Experimental studies on AM (e.g., LPBF) habitually employ the
design of experiments to decrease the number of experiments and save time and
costs. Hence, the experimental data are not prepared for DL model creation in most
*Corresponding author: situations. This paper studies the creation of a DL model on a small experimental
Lidong Wang dataset with unbalanced data and the prediction of the LOF defect of LPBF utilizing
(lidong@iser.msstate.edu)
the created DL model. Data analytics is mainly conducted based on four DL methods,
Citation: Wang L. Prediction of including Elman neural networks, Jordan neural networks, deep neural networks
the lack-of-fusion defect of laser
powder bed fusion based on deep (DNN) with weights initialized by the deep belief network, and the regular DNN based
learning. Int J AI Mater Design. on four algorithms: “rprop+”, “rprop−”, “sag,” and “slr.” It is shown that the regular DNN
2025;2(2):69-78. after the z-score standardization of the small dataset helps create a more accurate
doi: 10.36922/IJAMD025060005 DL model and achieve better analytics and prediction results than the three other DL
Received: February 5, 2025 methods in this paper. The three other DL methods do not work well in the prediction
1st revised: March 27, 2025 of LOF based on the small dataset (with unbalanced data).
2nd revised: May 15, 2025
Keywords: Additive manufacturing; Laser powder bed fusion; Deep learning; Deep
3rd revised: May 29, 2025
neural network; Defect prediction
Accepted: June 3, 2025
Published online: June 16, 2025
Copyright: © 2025 Author(s). 1. Introduction
This is an Open-Access article
distributed under the terms of the
Creative Commons Attribution Additive manufacturing (AM) is one of the key elements in Industry 4.0. The reliability
License, permitting distribution, and quality of metal AM parts or components are critical, especially for a laser powder
and reproduction in any medium, bed fusion (LPBF) process. Melt pool defects of LPBF, for example, balling, keyholes,
provided the original work is
1
properly cited. and lack of fusion (LOF) can compromise structural integrity. LPBF has advantages,
Publisher’s Note: AccScience such as complex and precise parts, excellent material densification, etc. The laser
Publishing remains neutral with scanning strategy is one of the significant factors affecting the quality of LPBF parts,
regard to jurisdictional claims in
published maps and institutional where printing defects often occur at the path’s endpoints and between the paths. The
affiliations. optimization of process parameters and the improvement of forming environment are

Volume 2 Issue 2 (2025) 69 doi: 10.36922/IJAMD025060005