Page 101 - MSAM-2-3

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

A B C D

E F G H

Figure 6. Representative SEM images of AM Ti-6Al-4V samples. (A) EB-PBF XY plane, (B) LPBFHT XY plane, (C) LPBFNHT XY plane, (D) DED XY
plane, (E) EB-PBF XZ plane, (F) LPBFHT XZ plane, (G) LPBFNHT XZ plane, and (H) DED XZ plane.
Abbreviations: DED: Directed energy deposition; EB-PBF: Electron beam-powder bed fusion; SEM: Scanning electron microscopy.
Table 4. Crystal strain and residual stress

Material ε22 ε23 Ε33 σ22 σ23 σ23 (MPa)
(MPa) (MPa)
L-PBF XY HT 0.002297 −0.00062 −0.00246 273.343 −74.137 −292.612
L-PBF XZ HT −0.00096 0.000059 −0.00189 −114.478 7.021 −224.865
L-PBF XY NHT −0.00134 −0.00018 −0.00047 −159.341 −20.825 −55.981
L-PBF XZ NHT −0.00179 0.000861 −0.00099 −213.367 102.459 −118.336
EB-PBF XY −0.00411 0.000716 0.000364 −488.495 85.204 43.275
EB-PBF XZ 0.000951 −0.00105 −0.00083 113.169 −124.95 −99.31
DED XY −0.0027 −0.00119 0.000658 −321.300 −141.610 78.264
DED XZ 0.00057 0.00364 −0.0055 67.830 433.160 −654.321
Abbreviations: DED: Directed energy deposition; EB-PBF: Electron beam powder bed fusion; LPBF: Laser powder bed fusion.
Table 5. Schmid factor and Taylor factor

Material Basal (SD) Prismatic Pyramidal ε = 0.5 ε = 1 ε = 1.5
(Schmid) (Schmid) (Schmid) (Taylor) (Taylor) (Taylor)
LPBF XY HT 0.2902 0.3928 0.1151 2.1288 3.6471 5.1626
LPBF XZ HT 0.2873 0.3982 0.1170 2.1139 3.7023 5.1586
LPBF XY NHT 0.3005 0.3825 0.1148 2.1923 3.7147 5.1753
LPBF XZ NHT 0.2795 0.4110 0.1194 2.1130 3.6349 5.0466
EB-PBF XY 0.2994 0.3963 0.1167 2.0285 3.5247 4.8956
EB-PBF XZ 0.2935 0.3851 0.1171 2.0923 3.5557 5.1285
DED XY 0.2761 0.3491 0.1118 2.1296 3.9260 5.5082
DED XZ 0.2800 0.4377 0.1334 2.0096 3.2098 4.5987
Abbreviations: DED: Directed energy deposition; EB-PBF: Electron beam powder bed fusion; LPBF: Laser powder bed fusion.
machining parameters. The prediction response is the specific SEM microstructure functions (MP + SEM), and all features
cutting energy. Root mean square error (RMSE) is used as the (All). The rationale for this approach is to understand the
evaluation metric to show the model’s accuracy. The XGBoost individual and cumulative interaction effects of machining
model and linear regression model are applied in this study. conditions, grain size, grain density, grain orientation, and
residual stress on machining behavior.
Five different feature combinations were designed: The first condition in this study uses 14400 L-PBF
Machining parameters only (MP), machining parameters and data points as a training set, and the testing set is DED
EBSD features (MP + EBSD), machining parameters and data. During the training process in the XGBoost model,
residual stress (MP + XRD), machining parameters and the grid-search method was used for digging hyper-

Volume 2 Issue 3 (2023) 9 https://doi.org/10.36922/msam.0999