Page 50 - IJAMD-1-1

International Journal of AI

for Material and Design

ORIGINAL RESEARCH ARTICLE
Integration of physics-based data in deep

learning model training for predicting the
effect of sulfur content in the directed energy

deposition process

1,2
Stanley Jian Liang Wong , Chengxi Chen , Eddie Zhi’En Tan , and Hua Li *
1
2
1,2
1 Department of Singapore Centre for 3D Printing, School of Mechanical and Aerospace Engineering,
Nanyang Technological University, Singapore, Republic of Singapore
2 Makino Asia Pte Ltd, Singapore, Republic of Singapore

Abstract

The training of a machine learning model solely on experimental data, encompassing
both pre- and post-process information, can reveal the general relationship of the
directed energy deposition process. However, models trained in this manner encounter
limitations in capturing critical in-process information occurring during deposition.
This paper details the training of a deep learning model through the integration
of in-process physics-based simulation information and a pre-process experiment
dataset. The sulfur content of stainless steel 316L was selected as critical in-process
information affecting the final track geometry and was captured using computational
fluid dynamics simulation of a single-track deposition process, which cannot be
captured accurately through experimentation. The physics-based simulation dataset
*Corresponding author: was generated by obtaining the contour of deposition and dilution of the solidified
Hua Li track cross-section. The experiment was conducted using central composite design,
(lihua@ntu.edu.sg) and data augmentation was achieved through curve fitting using a response surface
Citation: Wong SJL, Chen C, methodology regression model. Statistical analysis assessing the quality of simulation
Tan EZ, Li H. Integration of physics- and experiment data was conducted. Among six baseline models, a deep learning
based data in deep learning model
training for predicting the effect of model with a specified training sequence of experiment and simulation data, denoted
2
sulfur content in the directed energy as DL-AugExp-Sim-Exp, exhibited the best-performing R and root mean square error
deposition process. Int J AI Mater prediction accuracy for cross-section track shape. Notably, deep learning models
Design. 2024;1(1):2355.
2
https://doi.org/10.36922/ijamd.2355 trained with both experiment and simulation information demonstrated a lower R
value compared to models trained solely with experiment data, revealing a tradeoff
Received: December 1, 2023 2
Accepted: January 3, 2024 between R value and additional prediction capability. In summary, in this study,
Published Online: January 17, 2024 the integration of a physics-based simulation dataset demonstrated the additional
prediction capability concerning the effect of sulfur content on track geometry.
Copyright: © 2024 Author(s).
This is an Open-Access article
distributed under the terms of the
Creative Commons Attribution Keywords: Additive manufacturing; Directed energy deposition; Machine learning
License, permitting distribution,
and reproduction in any medium,
provided the original work is
properly cited. 1. Introduction
Publisher’s Note: AccScience
publishing remains neutral with Additive manufacturing (AM) is rapidly gaining popularity within the manufacturing
regard to jurisdictional claims in
published maps and institutional industry, resulting in an increase in the number of developed AM machines. However,
affiliations. the quest for optimal parameters remains an exhaustive process, often relying on a

Volume 1 Issue 1 (2024) 44 https://doi.org/10.36922/ijamd.2355