Page 28 - MSAM-1-4
P. 28
Materials Science in Additive Manufacturing Process optimization of SEBM IN718 via ML
A B
C
D
Figure 4. Classification of surface morphology. (A) Processing parameters window of surface morphology. (B–D) Typical surface morphology and
corresponding cross sections of samples: (B) 15 mA and 2 m/s (top), 25mA and 3m/s (bottom); (C) 15 mA and 4 m/s; and (D) 7.5 mA and 7 m/s.
A B
Figure 5. Surface morphology analysis. (A) Effect of energy density and beam current on surface morphology. (B) Relationship between energy density,
relative density and surface morphology.
characteristics of surface morphology. Low-energy 3.2. Prediction of relative density by machine
density results in porous surfaces, while high-energy learning
density results in uneven surfaces. Despite the same In this study, SVR and GPR machine learning algorithms
energy density, it is still easy to get uneven or unformed were trained to predict the relative density. There were 65
surfaces when the beam current is too high. The surface groups of basic data (including six repeated low relative
morphology is also related with the relative density, as density data), in which 52 data were for training, while 13
shown in Figure 5B. The porous surface has a low relative data were for testing, as shown in Figure 3. Training the
density due to the lack-of-fusion defects mentioned machine learning model by the method is described in
above. Samples with even surfaces usually own higher section 2.2. The appropriate hyper-parameter was selected
relative density, compared to those with uneven surfaces. by the Grid-Search method. Hyper-parameter is critical to
Different from SLM which uses laser as energy source, machine learning’s performance. SVR has the radial basis
there was no reduction in relative density due to keyhole function kernel and hyper-parameters C = 300, γ = 3.5, and
when the energy density is too high . GPR has squared exponential kernel and hyper-parameter
[46]
Volume 1 Issue 4 (2022) 5 https://doi.org/10.18063/msam.v1i4.23
