Page 99 - MSAM-2-3

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

properties. Due to the layer-by-layer processing, as EB-PBF, L-PBF, and DED fabricated Ti-6Al-4V will
microstructure and mechanical properties change due produce highly varied microstructure, which will also lead
to different cooling rates and thermal gradients within a to different machining behavior.
single AM part. The previous research observed that the In this study, L-PBF-, EB-PBF-, and DED-processed
grain size increases along the build direction in Ti-4Al-4V Ti-6Al-4V are studied due to their different AM processing
AM samples, and mechanical properties such as hardness, conditions and resulting grain morphologies. The authors
yield strength, and elongation gradient also vary along the build on their previous work on PBF ML study for process-
build direction [24,72-74] . In L-PBF and DED metal parts, heat structure-property linkages in the machining behavior of
treatment is also necessary to achieve the final geometrical Ti-6Al-4V . In this study, scanning electron microscopy
[7]
shape and desired mechanical properties. Heat treatment (SEM), electron backscatter diffraction (EBSD), and X-ray
directly affects the microstructure of the final metal AM diffraction (XRD) analysis methods are applied. Machining
parts. The previous research observed that there exists behavior, that is, specific cutting energy, is computed from a
a preferred grain growth orientation in the Ti-6Al-4V cutting force model using recorded data through a Taguchi
sample before and after heat treatment, and grain growth experiment. Gradient tree-based boosting machine
and transformation are evident after heat treatment [32,75,76] . learning models and benchmark linear regression models
Since post-processing is necessary for most metal AM parts are developed for a thorough validation study.
to achieve desired tolerance and remove support structures,
machining behavior such as tool wear, cutting force, and 2. Methodology
specific cutting energy should be investigated due to large
variations in material properties among Ti-6Al-4V alloys 2.1. Sample preparation
fabricated through different AM processes. In this study, EB-PBF Ti-6Al-4V specimens were fabricated

To reduce tool wear and cutting force, cryogenic using an Arcam A2 EB-PBFchine with 50 μm layer
machining was used in machining Ti-6Al-4V alloys. Bordin thickness. L-PBF specimens were manufactured using
et al. investigated the tool wear and surface finish on EB-PBF the EOSINT M280 system with 200W laser power, 80 μm
2
Ti-6Al-4V samples and found that adhesive wear was the spot size, and approximately 40 kW/mm power density.
primary tool wear mechanism in both dry and cryogenic Half of the L-PBF specimens were heat-treated to relieve
turning. When compared with dry machining, cryogenic residual stress. DED Ti-6Al-4V specimens were fabricated
machining led to a decrease in adhered layer thickness on using a MERKE IV machine with rapid plasma deposition
the rake face, while crater wear was eliminated by cryogenic technology. The layer thickness in the Z direction is around
cooling. Cryogenic cooling improves surface integrity and 3 – 4 mm. All DED specimens were heat-treated to relieve
leads to a better surface finish . Polishetty et al. compared residual stresses.
[77]
the machining force and surface finish of SLM Ti-6Al-4V 2.2. Material structure data extraction
samples with conventional produced wrought Ti-6Al-4V
samples and found that cutting forces in SLM samples 2.2.1. SEM data extraction
were higher than the machining of wrought parts. Another According to our previous research , the 2-point correlation
[7]
significant finding was that when the cutting speed functions [81-83] , chord length distributions (CLDs) , and
[84]
increased, cutting forces increased on machining SLM principal component analysis (PCA) have been proven to be
Ti-6Al-4V samples while cutting force decreased during valid low-order SEM microstructure feature extraction tools.
machining of wrought Ti-6Al-4V samples. SLM samples
had better surface roughness than wrought parts [78,79] . Wu The 2-point statistics denote the Ti-6Al-4V
et al. observed extra strain hardening caused by strain microstructure conditional probability of finding the same
gradient near the grain boundaries in microstructure phase (α/β) in spatial bins whose centers are separated by
gradient material, which leads to different properties a vector set.
compared with homogeneous material . The difference The CLDs describe the probability of finding a specific
[80]
in mechanical properties between heterogeneous and length chord within the microstructure. Considering the
homogenous materials affects the machining mechanisms anisotropy of AM materials, CLDs were resolved along and
and will result in a different machining behavior. directions of the microstructure to keep function analyses
Based on prior research, it is evident that the machinability consistent.
and machining behavior for AM Ti-6Al-4V parts are clearly
different from traditionally produced Ti-6Al-4V parts due to 2.2.2. XRD data extraction
their difference in material characterization and mechanical XRD was used to directly measure the strain resulting from
properties. In addition, different AM technologies such the distortion of the crystalline lattice structure, along

Volume 2 Issue 3 (2023) 7 https://doi.org/10.36922/msam.0999