3D Data for Fatigue in Superalloys
9
not exist. Analysis of the data may be either based on voxelized or meshed data
formats, depending on whether access to microstructural descriptors is desired, or
direct property simulation. However, some microstructural descriptors may require
meshed data formats as well, such as grain boundary inclinations.
For TriBeam EBSD datasets, DREAM.3D [75] is used to perform all reconstruction steps except for distortion corrections, which are performed using the methods
described for strontium titanate [50] and a nickel superalloy [76]. The reconstruction
steps can be clustered into four major groups: slice alignment, data cleanup, grain
or feature segmentation, and artifact removal.
Although generating a preliminary dataset reconstruction is trivial with modern
software tools, creating a high-quality reconstruction is still a significant challenge
and often requires more time than dataset collection. Procedures that reduce
noise or improve data quality greatly enhance the ability to extract high fidelity
information from the dataset for modeling. Alignment and segmentation are by far
the most difficult tasks. Alignment can be particularly challenging for small datasets
where the morphology of a few dominant features dictates shifts computed during
registration. Creating sample pedestals like the one fabricated using wire EDM
shown in Fig. 5 that are small enough to collect data from the entire sample surface
makes alignment significantly easier, and recovering the original sample shape
provides a simple validation of alignment quality [20]. The pedestal fabrication
procedure is a coarser scaled equivalency to the FIB procedures pioneered by Uchic.
In many instances the pedestals used to collect data shown here were of the order
of 1 × 1 mm in cross section by several mm in height. Orientation gradients and
systemic misindexing due to pseudosymmetry are the most serious challenge for
segmentation, when present.
Dataset volumes can become many terabytes in size, mostly due to the collection
of raw EBSD patterns (EBSPs) or full spectrum EDS maps. The approximate scale
Fig. 5 Wire EDM is often
used to create custom
mm-scaled sample pedestals
for targeted and untargeted
TriBeam sectioning. The
roughly 10 μm EDM
heat-affected zone is
mechanically polished away
before TriBeam experiments
or is located adjacent to a
region where data will not be
collected. The pedestal
geometry is used in order to
reduce material redeposition
during laser ablation and to
prevent shadowing of the
EBSD signal at high sample
tilt angles
9
not exist. Analysis of the data may be either based on voxelized or meshed data
formats, depending on whether access to microstructural descriptors is desired, or
direct property simulation. However, some microstructural descriptors may require
meshed data formats as well, such as grain boundary inclinations.
For TriBeam EBSD datasets, DREAM.3D [75] is used to perform all reconstruction steps except for distortion corrections, which are performed using the methods
described for strontium titanate [50] and a nickel superalloy [76]. The reconstruction
steps can be clustered into four major groups: slice alignment, data cleanup, grain
or feature segmentation, and artifact removal.
Although generating a preliminary dataset reconstruction is trivial with modern
software tools, creating a high-quality reconstruction is still a significant challenge
and often requires more time than dataset collection. Procedures that reduce
noise or improve data quality greatly enhance the ability to extract high fidelity
information from the dataset for modeling. Alignment and segmentation are by far
the most difficult tasks. Alignment can be particularly challenging for small datasets
where the morphology of a few dominant features dictates shifts computed during
registration. Creating sample pedestals like the one fabricated using wire EDM
shown in Fig. 5 that are small enough to collect data from the entire sample surface
makes alignment significantly easier, and recovering the original sample shape
provides a simple validation of alignment quality [20]. The pedestal fabrication
procedure is a coarser scaled equivalency to the FIB procedures pioneered by Uchic.
In many instances the pedestals used to collect data shown here were of the order
of 1 × 1 mm in cross section by several mm in height. Orientation gradients and
systemic misindexing due to pseudosymmetry are the most serious challenge for
segmentation, when present.
Dataset volumes can become many terabytes in size, mostly due to the collection
of raw EBSD patterns (EBSPs) or full spectrum EDS maps. The approximate scale
Fig. 5 Wire EDM is often
used to create custom
mm-scaled sample pedestals
for targeted and untargeted
TriBeam sectioning. The
roughly 10 μm EDM
heat-affected zone is
mechanically polished away
before TriBeam experiments
or is located adjacent to a
region where data will not be
collected. The pedestal
geometry is used in order to
reduce material redeposition
during laser ablation and to
prevent shadowing of the
EBSD signal at high sample
tilt angles