Page 43 - MSAM-4-3

Materials Science in

Additive Manufacturing

ORIGINAL RESEARCH ARTICLE
Artificial intelligence-driven defect detection

and localization in metal 3D printing using
convolutional neural networks

1
Xinyi Yin * , Jan Akmal 1,2 , and Mika Salmi 1
1 Materials to Products Research Group, Department of Energy and Mechanical Engineering, Aalto
University, Espoo, Finland
2 EOS Metal Materials, Electro Optical Systems Finland Oy, Turku, Finland
(This article belongs to the Special Issue: Smart Additive Manufacturing: Product and Process
Qualification through Innovation in Design, Modeling, Monitoring, Machine Learning, Metrology, and
Materials Science)

Abstract

h
Metal additive manufacturing (AM) has attracted significant i nterest i n igh-value
industries due to its ability to produce complex parts flexibly, but the reliance on costly
manual monitoring remains a major burden for quality control. Artificial intelligence
(AI)-driven models for automated defect detection are emerging as promising
solutions. This study contributes a new annotated dataset for AI research in AM and
evaluates the performance of four widely used convolutional neural network (CNN)
*Corresponding author: models in detecting powder bed morphology defects, based on layer-wise images
Xinyi Yin
(xinyi.yin@aalto.fi) acquired by the EOSTATE PowderBed system during the metal laser-based powder
bed fusion process. The models were trained through transfer learning methods with
Citation: Yin X, Akmal J, Salmi M. manually labeled and pre-processed data. Results demonstrated that ResNet50 and
Artificial intelligence-driven defect
detection and localization in metal EfficientNetV2B0 achieved over 99% accuracy in defect classification, while YOLOv5
3D printing using convolutional outperformed Faster region-based-CNN in defect detection and localization. However,
neural networks. Mater Sci Add lower average precision values in object detection tasks were attributed to variability
Manuf. 2025;4(3):025150022.
doi: 10.36922/MSAM025150022 in defect scales and annotation quality. This study confirms the potential of AI-based
models for defect identification in AM, with YOLOv5 demonstrating clear advantages in
Received: April 7, 2025
managing complex, multi-scale defects. Future improvements will focus on expanding
1st revised: May 7, 2025 the dataset and refining annotation strategies to further enhance model robustness.
2nd revised: May 20, 2025
Accepted: May 20, 2025 Keywords: Machine learning; Defects detection; Quality control; AI-driven models; Metal
Published online: June 25, 2025 additive manufacturing; Image classification; Object detection
Copyright: © 2025 Author(s).
This is an Open-Access article
distributed under the terms of the
Creative Commons Attribution 1. Introduction
License, permitting distribution,
and reproduction in any medium, Metal additive manufacturing (AM) has attracted significant attention in high-value
provided the original work is industries, such as aerospace, automotive, healthcare, and nuclear energy, due to its
properly cited. ability to produce customized, complex parts with excellent mechanical properties while
1-3
Publisher’s Note: AccScience minimizing waste. However, a significant challenge in AM is the frequent occurrence
Publishing remains neutral with of inevitable defects. Even in metal laser-based powder bed fusion (PBF-LB) alone,
regard to jurisdictional claims in
published maps and institutional around 10 typical high-frequency defects may occur, including lack of fusion, keyhole
affiliations. porosity, gas porosity, solidification cracking, solid-state cracking, surface-connected

Volume 4 Issue 3 (2025) 1 doi: 10.36922/MSAM025150022