Page 68 - ESAM-1-1

Engineering Science in

Additive Manufacturing

PERSPECTIVE ARTICLE
Advancing machine learning in additive

manufacturing: Perspectives on data challenges,
model development, and industrial adoption

†
†
Mutahar Safdar , Jiarui Xie , and Yaoyao Fiona Zhao*
Department of Mechanical Engineering, Additive Design and Manufacturing Laboratory,
McGill University, Montreal, Canada

Abstract

Over the past decade, 100 of scientific studies have harnessed statistical artificial
intelligence, specifically machine learning (ML), to address the existing challenges
to process repeatability and part quality in additive manufacturing (AM). ML has also
been applied to support structure and material design for AM. This rapidly expanding
field at the intersection of two growing disciplines provides opportunities to mature
AM technology in the industry. There have been numerous review articles and survey
reports to summarize ML applications at design, processes, structure, and property
phases. While ML models, data handling, and learning techniques for AM concerns
have been summarized in these articles, no work has specifically focused on the core
† These authors contributed equally
to this work. ML challenges of data scarcity and model complexity – two important aspects to
develop ML models for the AM industry. This perspective highlights existing data
*Corresponding author:
Yaoyao Fiona Zhao and model challenges and discusses the opportunities to address them through
(yaoyao.zhao@mcgill.ca) advanced techniques in data science and ML. By enhancing the usefulness of AM
Citation: Safdar M, Xie J, datasets and by leveraging the strengths of cutting-edge ML models, the research
Zhao YF. Advancing machine progress at the intersection of ML and AM can be effectively translated into real-
learning in additive manufacturing: world industrial applications making the deployment of ML models in the industry
Perspectives on data challenges,
model development, and industrial much easier.
adoption. Eng Sci Add Manuf.
2025;1(1):025040004.
doi: 10.36922/ESAM025040004 Keywords: Additive manufacturing; Artificial intelligence; Machine learning; Data
scarcity; Model complexity; Perspective; Industrial integration
Received: January 24, 2025
Revised: March 3, 2025
Accepted: March 6, 2025
1. Introduction
Published online: March 19, 2025
Additive manufacturing (AM) enables the direct conversion of digital designs into physical
Copyright: © 2025 Author(s).
This is an Open-Access article objects by depositing, joining, or solidifying materials layer-by-layer, without requiring
1
distributed under the terms of the molds, tools, or extensive manual intervention. This approach marks a transformative
Creative Commons Attribution shift in product design and manufacturing. Unlike traditional manufacturing methods,
License, permitting distribution,
and reproduction in any medium, AM offers distinct advantages, such as the capability to produce intricate, complex,
provided the original work is and highly customized geometries at a lower cost, significantly shortening production
properly cited. cycles and time-to-market. Furthermore, it facilitates the development of innovative
2
Publisher’s Note: AccScience material compositions and designs with tailored functionalities. AM also holds great
Publishing remains neutral with promise for reducing environmental impacts by minimizing material waste, enhancing
regard to jurisdictional claims in
published maps and institutional material efficiency, and eliminating the need for assemblies, thereby streamlining the
3-5
affiliations. manufacturing process. However, despite its advantages, AM processes often face

Volume 1 Issue 1 (2025) 1 doi: 10.36922/ESAM025040004