Page 79 - MSAM-4-3

Materials Science in Additive Manufacturing Interpretable GP melt track prediction

in the melt pool can lead to defects in the trajectories. Five melt pool is disturbed by the expansion, the P-R instability
types of trajectory characteristics—regularity, distortion, makes the surface of the melt pool protrude, which leads to
hump, necking, and protrusion—were identified in the the splitting of the melt and the splashing of the melt. The
experiments, and their morphologies are displayed in combined effect of these factors on the melt pool leads to
Figure 5. drastic changes in the morphology of the melt pool, thus

Figure 5A displays a regularity-shaped melt track, increasing the risk of defects. 28,29
characterized by continuous edges that do not deviate To fully characterize the melt pool geometry,
from the preset scanning path and are symmetrical quantitative analysis was performed by extracting
along the centerline of the scanning trajectory. Figure 5B geometrical parameters, such as aspect ratio, roundness,
features a distortion-shaped melt track, which usually has and convex defects of the melt pool. During the LPBF
22
continuous edges, but the center of the melt track deviates process, images captured by the high-speed camera in the
from the pre-defined scanning path, probably due to the side-axis monitoring system are subject to interference
Marangoni effect that distorts the trajectory. Figure 5C from intense metal powder splatter, plasma plumes,
presents a hump-shaped melt track, which usually has and substrate surface reflections. To address this, the
23
different thicknesses due to bulging. This is associated with ROI method was first applied to identify the maximum
the Plateau-Rayleigh (P-R) instability generated by the connected region (128 × 128 pixels) with grayscale values
pressure difference acting on the surface of the melt pool.
24
Figure 5D illustrates a melt track with necking-shaped exceeding 200. This approach effectively eliminates most
morphology, characterized by a contraction in width splatter interference while preserving the integrity of the
25
or even the appearance of breakpoints – a phenomenon melt pool morphology. Second, the k-means segmentation
that is also believed to be caused by cylindrical fluid jets. method (segmentation category: 7; i.e., the category with
Figure 5E features a protrusion-shaped melt track, often the largest sum of pixel values) was used to remove the
26
accompanied by a sudden increase in width, forming plume, proximity powder splash, and metal material arc
an angular lateral bulge – a phenomenon that is usually light, among other parameters. The large particle droplet
attributed to high surface tension melt pools and the splash and the melt pool were then separated by the kernel
Marangoni effect. 27 operation (kernel size: 5). Connectivity domain analysis
was then used to select the largest area to obtain the main
2.3. Data processing body of the melt pool. According to the main body, the
High surface tension makes the melt pool surface unstable, area s m, perimeter p m, convexity defect t m, length l m, width
which can easily lead to melt rupture. The Marangoni effect w m, and center of mass coordinates x,y of the melt pool
induces the melt to flow from the high-temperature region were calculated, and the flow is displayed in Figure 6. The
to the low-temperature region, leading to metal splashing input features x m = [s m, ar m, c m, t m, y] of the melt pool were
from the surface of the melt pool. When the surface of the obtained based on Equations II and III.

A B C

D E

Figure 5. Five different types of melt track morphology in laser powder bed fusion: (A) regularity, (B) distortion, (C) hump, (D) necking, and
(E) protrusion. Scale bar: 200 μm

Volume 4 Issue 3 (2025) 5 doi: 10.36922/MSAM025200030