Page 105 - MSAM-2-3

Materials Science in Additive Manufacturing Validation of a novel ML model for AM-PSP

Table 9. Comparison of P-values for AM Ti‑6Al‑4V 2‑point coefficients
2‑Point coefficient EB‑PBF EB‑PBF LPBFHT LPBFHT LPBFNHT LPBFNHT DED DED
XY XZ XY XZ XY XZ XY XZ
EB-PBF XY \ 0.021 0.046 0.001 0.063 0.005 0.025 0.078
EB-PBF XZ \ \ 0.591 0.019 0.170 0.005 0.088 0.378
LPBFHT XY \ \ \ 0.056 0.130 0.009 0.030 0.826
LPBFHT XZ \ \ \ \ 0.777 0.241 0.208 0.060
LPBFNHT XY \ \ \ \ \ 0.377 0.381 0.046
LPBFNHT XZ \ \ \ \ \ \ 0.020 0.013
DED XY \ \ \ \ \ \ \ 0.019
DED XZ \ \ \ \ \ \ \ \
Abbreviations: DED: Directed energy deposition; EB-PBF: Electron beam powder bed fusion; LPBF: Laser powder bed fusion.

A B C

D E

Figure 10. F-score for different training conditions: (A) MP, (B) MP + EBSD, (C) MP + XRD, (D) MP + SEM, and (E) all feature above.
Abbreviations: EBSD: Electron backscatter diffraction; MP: Machining parameters; SEM: Scanning electron microscope; XRD: X-ray diffraction.

5. Conclusion methods were applied to this study to save computational
cost and time. Finally, the new PSP linkages, which include
A valid Ti-6Al-4V AM PSP linkage is presented herein. the most employed Ti-6Al-4V AM technologies (PBF and
The PSP linkage focuses on connecting the AM Ti-6Al-4V DED), were established through this study.
structure information with post-processing machining
behavior. This PSP linkage covers L-PBF, EB-PBF, and DED The key findings of this research are as follows:
Ti-6AL-4V data, which expands the previous research (i) Established a validated statistical methodology to
version and overcomes the application limitation of the distinguish the microstructure difference among
previous PBF Ti-6Al-4V PSP linkages. different AM processes fabricated Ti-6Al-4V alloys
To validate the Ti-6Al-4V PSP linkages established (L-PBF with/without heat treatment, EB-PBF, and
in the previous research based on the PBF dataset, wire DED).(ii) It was shown that due to inherently
feed plasma-based directed energy deposition Ti-6Al-4V different grain morphologies across PBF and DED
materials were applied to evaluate the efficacy of the processes, the developed ML-based PSP linkages could
ML-based PSP linkage accuracy. Similar to a prior study , be limited if one of the AM process categories (e.g.,
[7]
statistical methods were employed to define the properties of PBF) is used as a training set to predict the material
the material, and revised feature dimensionality reduction response, that is, machining behavior of other AM

Volume 2 Issue 3 (2023) 13 https://doi.org/10.36922/msam.0999