Page 11 - ESAM-1-1

Engineering Science in
Additive Manufacturing ML in MAM monitoring and control through images

sample under investigation. Then, the camera is connected real-time thermography to detect anomalies with a
to a computer system equipped with image processing broader observation range and increased installation
software for real-time analysis and recording of the images flexibility. For instance, Marshall et al. installed an IR
53
54
captured. Furthermore, adequate lighting conditions are camera to monitor the changing temperature dynamics
maintained to ensure optimal visibility of the sample during of bulk part and a pyrometer for measuring the melt
imaging. Before the experiment, the camera is calibrated pool temperature, as depicted in Figure 4. Similarly, a
to ensure accurate color representation and focus. Next, coaxial thermal monitoring device was used by Zheng
during the monitoring process, the images are analyzed to et al. to track the melt pool’s temperature distribution
55
track any changes in the sample morphology. For instance, during the LPBF process. Yan et al. extracted thermal
56
Cannizzaro et al. placed an off-axis, low-cost camera to
47
automatically capture images of the powder bed during
the layering process. They discovered that defects like
holes and spattering were easily detectable when visualized
in images. Similarly, Iravani-Tabrizipour et al. installed
48
off-axis CCD cameras to observe the deposition layer
area during the DED process. They then employed image
processing algorithms to assess the deposition layer height,
ensuring the stability of the printing process.
Researchers have identified melt pool characteristics
as pivotal factors influencing printing stability and sample
quality, prompting a shift in focus toward monitoring
the melt pool itself. For example, Pandiyan et al. set up
49
a simple CCD camera to coaxially monitor melt pool
morphology during the DED process. They utilized
U-net methods to classify printing quality based on the
observed melt pool characteristics. Other researchers have
also monitored melt pool morphology attributes such as
width, length, and area, making adjustments to assess if the
corresponding features can be controlled. Nevertheless, Figure 3. Experimental setup with an off-axis high-speed camera in the
50
39
the rapid movements of the melt pool present challenges direct energy deposition process. Reproduced with permission from
Elsevier. Copyright © 2024 The Author(s). Distributed under a CC-BY
in capturing clear and distinct melt pool images, leading 4.0 license (https://creativecommons.org/licenses/by/4.0/).
to the adoption of high-speed cameras, as demonstrated
in Figure 3. Li et al. installed a coaxial high-speed camera
46
to capture melt pool images during the multi-track and
multilayer LPBF process. In addition, Li et al. successfully
51
extracted melt pool boundaries captured with a high-
speed camera, significantly enhancing the fine-tuning of
laser parameters for melt pool stabilization. Furthermore,
Gaikwad et al. utilized a high-speed camera to capture
52
melt pool droplets during the DED process and effectively
measured their size and velocity from the images. These
advancements underscore the critical role of high-speed
cameras in capturing dynamic melt pool behavior, enabling
precise monitoring and control of the MAM process to
enhance printing quality and stability.
Monitoring melt pool morphology can indeed require
sophisticated equipment to ensure stability and reduce
image interference, potentially leading to longer acquisition
and processing times. Furthermore, challenges arise in
capturing the sub-surface conditions of the melt pool, such Figure 4. Experimental setup with an off-axis infrared camera in the
direct energy deposition process. Reproduced with permission from
54
as bubbles, inclusions, and other defects. Temperature Elsevier. Copyright © 2016 The Authors. Distributed under a CC-BY 4.0
monitoring has emerged as a valuable alternative, offering license (https://creativecommons.org/licenses/by/4.0/).

Volume 1 Issue 1 (2025) 5 doi: 10.36922/esam.8548