Prediction of the lack-of-fusion defect of laser powder bed fusion based on deep learning

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 situations. This paper studies the creation of a DL model on a small experimental dataset with unbalanced data and the prediction of the LOF defect of LPBF utilizing the created DL model. Data analytics is mainly conducted based on four DL methods, including Elman neural networks, Jordan neural networks, deep neural networks (DNN) with weights initialized by the deep belief network, and the regular DNN based on four algorithms: “rprop+”, “rprop−”, “sag,” and “slr.” It is shown that the regular DNN after the z-score standardization of the small dataset helps create a more accurate DL model and achieve better analytics and prediction results than the three other DL methods in this paper. The three other DL methods do not work well in the prediction of LOF based on the small dataset (with unbalanced data).

Keywords

Additive manufacturing / Laser powder bed fusion / Deep learning / Deep neural network / Defect prediction

Acknowledgments

The author thanks the support from Mississippi State University, Mississippi, USA.

Funding

None.

Conflict of interest

The author declares no conflicts of interest.

Author contributions

This is a single-authored article.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data

The dataset in this paper is available from the author of reference¹⁰ upon appropriate request.

References

[1] Raihan AS, Harper A, Era IZ, et al. A data-efficient sequential learning framework for melt pool defect classification in laser powder bed fusion. J Manuf Processes. 2025; 145:201-210. doi: 10.1016/j.jmapro.2025.03.118

[2] Ni C, Zhu J, Zhang B, et al. Recent advance in laser powder bed fusion of Ti-6Al-4V alloys: Microstructure, mechanical properties and machinability. Virtual Phys Prototyp. 2025; 20(1):e2446952. doi: 10.1080/17452759.2024.2446952

[3] Nabavi SF, Dalir H, Farshidianfar A. A comprehensive review of recent advances in laser powder bed fusion characteristics modeling: Metallurgical and defects. Int J Adv Manuf Technol. 2024; 132(5):2233-2269. doi: 10.1007/s00170-024-13491-1

[4] Ero O, Taherkhani K, Hemmati Y, Toyserkani E. An integrated fuzzy logic and machine learning platform for porosity detection using optical tomography imaging during laser powder bed fusion. Int J Extrem Manuf. 2024; 6(6):065601. doi: 10.1088/2631-7990/ad65cd

[5] Gu Z, Mani Krishna KV, Parsazadeh M, et al. Deep learning-based melt pool and porosity detection in components fabricated by laser powder bed fusion. Prog Addit Manuf. 2025; 10(1):53-70. doi: 10.1007/s40964-024-00603-2

[6] Zhao J, Yang Z, Chen Q, et al. Real-time detection of powder bed defects in laser powder bed fusion using deep learning on 3D point clouds. Virtual Phys Prototyp. 202531;20(1):e2449171. doi: 10.1080/17452759.2024.2449171

[7] Pouyanfar S, Sadiq S, Yan Y, et al. A survey on deep learning: Algorithms, techniques, and applications. ACM Comput Surv (CSUR). 2018; 51(5):1-36. doi: 10.1145/3234150

[8] Narasimharaju SR, Zeng W, See TL, Zhu Z, Scott P, Jiang X, Lou S. A comprehensive review on laser powder bed fusion of steels: Processing, microstructure, defects and control methods, mechanical properties, current challenges and future trends. J Manuf Processes. 2022; 75:375-414. doi: 10.1016/j.jmapro.2021.12.033

[9] Ganta MG, Kurek M. Influence of post-processing methods on the fatigue performance of materials produced by selective laser melting (SLM). Int J Adv Manuf Technol. 2025; 9:1-32. doi: 10.1007/s00170-024-14920-x

[10] Wang GW. Microstructure and Mechanical Properties of Oxide Dispersion Strengthened Nickel-Based Superalloys by Laser Additive Manufacturing. [Dissertation, Zhongnan University, China]; 2023. Available from: https://www.cnki.net [Last accessed on 2025 May 28].

[11] Bramer M. Data for Data Mining. Principles of Data Mining. London: Springer; 2016. p. 9-19. doi: 10.1007/978-1-4471-7307-6

[12] Alabdulwahab S, Moon B. Feature selection methods simultaneously improve the detection accuracy and model building time of machine learning classifiers. Symmetry. 2020; 12(9):1424. doi: 10.3390/sym12091424

[13] Maseer ZK, Yusof R, Bahaman N, Mostafa SA, Foozy CF. Benchmarking of machine learning for anomaly based intrusion detection systems in the CICIDS2017 dataset. IEEE Access. 2021; 9:22351-22370. doi: 10.1109/ACCESS.2021.3056614

[14] Mohammed JZ, Wagner M. Data Mining and Analysis: Fundamental Concepts and Algorithms. Cambridge: Cambridge University Press; 2014.

[15] Lewis ND. Deep Learning Made Easy with R. A Gentle Introduction for Data Science. South Carolina: CreateSpace Independent Publishing Platform; 2016.

[16] Elman JL. Finding structure in time. Cogn Scil. 1990; 14(2):179-211. doi: 10.1207/s15516709cog1402_1

[17] Jordan MI. Serial order: A parallel distributed processing approach. Adv Psychol. 1997;121:471-495.

[18] Jang H, Plis SM, Calhoun VD, Lee JH. Task-specific feature extraction and classification of fMRI volumes using a deep neural network initialized with a deep belief network: Evaluation using sensorimotor tasks. Neuroimage. 2017; 145:314-328. doi: 10.1016/j.neuroimage.2016.04.003

[19] Ghasemi F, Mehridehnavi A, Fassihi A, Pérez-Sánchez H. Deep neural network in QSAR studies using deep belief network. Appl Soft Comput. 2018; 62:251-258. doi: 10.1016/j.asoc.2017.09.040

[20] Hinton GE. Training products of experts by minimizing contrastive divergence. Neural Comput. 2002;14:1771-800. doi: 10.1162/089976602760128018

[21] Hinton G. A practical guide to training restricted Boltzmann machines. Momentum. 2010; 9(1):926.

[22] Fischer A, Igel C. Training restricted Boltzmann machines: An introduction. Pattern Recognit. 2014; 47(1):25-39. doi: 10.1016/j.patcog.2013.05.025

[23] Hann J, Pei J, Kamber M. Data Mining: Concepts and Techniques. Netherlands: Elsevier; 2011.

[24] Chung H, Lee SJ, Park JG. Deep Neural Network using Trainable Activation Functions. In: 2016 International Joint Conference on Neural Networks (IJCNN). IEEE; 2016. p. 348-352. doi: 10.1109/IJCNN.2016.7727219