Page 100 - MSAM-2-3

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

with the associated residual stresses . According to the The near-surface residual stress from eight materials
[85]
previous study, the presence of shear stresses is indicated is shown in Table 4. Along the build direction, the DED
by splitting in the d-spacing-sin Ψ plot . Therefore, a sample shows larger compressive and shear stresses in the
2
[86]
regression analysis was used to calculate the principal Z direction. L-PBF-AM sample shows large compressive
stresses of each specimen. stress in the Y direction. On the top layers, both the
EB-PBF and DED samples show large compressive stress
2.2.3. EBSD data extraction in the Y direction but minimal stress in the Z direction.
For constructing the valid PSP linkages among AM L-PBF heat-treated sample shows massive tensile stress on
Ti-6Al-4V, the Schmid and Taylor models were utilized top layers in the Y direction and compressive stress in the
in this study. For feature extraction from EBSD data, Z direction, while the as-AM L-PBF sample shows a small
the Schmid Factors (SFs) were calculated from different stress value on the top region.
specimens based on three main slip systems, and Taylor The ratio of critical resolved shear stress (CRSS) applied
Factors (TFs) were calculated based on three different to the Schmid model on basal, prismatic, and first-order
strain rate conditions. pyramidal slip systems is 1:0.7:3 [88,89] . In the Taylor model,
considering the milling process, the deformation strain
2.3. Machining experiment and specific cutting energy (DS) was expressed as a pure shear strain component and a
The machining experiment was performed on a Haas rotation component in Equation I:
VF2SS vertical CNC center. The cutting tools selected
for the experiments were 6.35 mm (1/4 inch) nominal   0 0 ε   0 0 ε    
diameter six flute carbide end mills with KC635M TiAlN  2   2 
coating (Model HPFT250S6075). DS =  0 0 0  +  0 0 0  (I)
 ε   ε 
Three levels of machining parameters selected for the  0 0  − 0 0 
machining experiments are shown in Table 3. As described  2   2 
above, two feed directions, XY and XZ, which are vertical For milling machining conditions, a generalized
and parallel to the build orientation, were utilized in the deformation strain form was applied to the Taylor model.
experiment for EB-PBF, as-AM L-PBF, heat-treated L-PBF, For each surface, three replicate EBSD samples were
and DED specimens. collected as the representative data, as shown in Table 5,
Specific cutting energy for each path was calculated which presents the average values of the SFs and TFs on
based on our cutting force model. each surface under different strain rates.
3. Results 3.2. Ti-6Al-4V machining behavior

3.1. Material microstructure representation The results from the Taguchi experiment, the specific
cutting energy analyses, are shown in Figures 7 and 8.
In this study, 200 microstructure statistical volume elements The ANOVA results in Table 6 indicate the P-values
(SVEs) are used to represent each AM Ti-6Al-4V sample for material, feed, and speed are <0.05, which shows a
level (L-PBF XY HT, L-PBF XZ HT, L-PBF XY NHT, L-PBF significant impact on specific cutting energy. The depth
XZ NHT, EB-PBF XY, EB-PBF, DED XY, and DED XZ), of cut and material interaction shows a significant effect
as shown in Figure 6. In SVEs, the influence coefficient on specific cutting energy. As previous research shows, the
decayed to around zero within a ~210 μm region . After specific cutting energy has a significant difference among
[87]
SEM feature extraction and PCA, the number of dimensions L-PBF Ti-6Al-4V, and the DED Ti-6Al-4V also proved that
was reduced from over 3000 initially to 238 as the PCA the AM processes significantly affect machining behavior.
cumulative variance approaches >99%. This is critical for From the interaction plot, the DED-specific cutting
developing an SP linkage model that is both computationally energy decreases as the depth of cut increases, which show
feasible and accurate in predicting material response.
that the thermal softening phenomenon dominates the
machining response in DED parts, which has been observed
Table 3. Machining parameters in the machining of titanium with coarser grains [93-95] .

Levels 1 2 3 3.3. Process-structure-property linkages testing and
Depth of cut (mm) 0.6350 1.270 1.905 validation
Feed (mm per tooth) 0.0127 0.0254 0.0381 For each data sample, features include 238 principal
Speed (m per min) 24.384 36.576 48.768 components, three residual stress, 9 SFs, 9 TFs, and three

Volume 2 Issue 3 (2023) 8 https://doi.org/10.36922/msam.0999