120
D. W. Eastman et al.
6 Chapter Summary
Microtensile testing protocols and unique characterization routes have been developed and employed to obtain multiscale, microstructurally dependent benchmarks
for CPFEM models of commercial polycrystalline alloys, such as René 88DT. Initial
microtensile testing was performed on samples with gage cross sections on the
order of a few hundred microns machined via Wire EDM and on much smaller
samples with gage cross sections of 20 × 20 μm machined using a FIB. The
samples machined with wire EDM exhibited a mechanical response that was very
similar to the bulk material. The much smaller FIB-machined samples were more
tractable for CPFEM simulations but were found to be too small to capture a true
polycrystalline response. A femtosecond laser machining protocol was developed
and used to machine microtensile samples with cross-sectional dimensions of 50–
100 microns at a rate that was an order of magnitude faster than FIB milling;
machining of samples 25 times larger than could be achieved with the FIB was
accomplished in a matter of minutes.
Testing samples of different sizes revealed an inherent effect of sample size
on yield strength. It was seen that the sample strength decreased with decreasing
sample size and that the variability of the measured strengths increased as the
sample size decreased. These variations are attributed to the influence of geometric
averaging of oligocrystalline specimens. Combining the microtensile tests with
local strain and orientation mapping provided robust microstructural benchmarks
for direct comparisons with CPFEM simulations. More than just acquiring the bulk
stress-strain response of a sample, investigating local response of the material and
the corresponding microstructural features that lead to this behavior provides a new
level of benchmarking for ever-improving CPFEM modeling capabilities. Local
strain maps can be used to compare directly with simulations instantiated with the
microstructural data of these samples. This explicit benchmark is both quantitative
in terms of the local magnitudes of strain observed as well as qualitative in terms
of where in the microstructure these concentrations are located. Moreover, the
correlation of microstructural information with strain maps enhances mechanistic
understanding and provides detailed microstructural benchmarks. Ensuring that this
behavior is captured in CPFEM simulations should be a key goal of future model
development.
The long-term goal of this work was to be able to explicitly characterize tested
samples in 3D to be able to instantiate simulations and collect attendant 3D
benchmarks of mechanical behavior. The development and availability of advanced
characterization and mechanical testing techniques, like the ones outlined in this
chapter, make it possible to marry experimental and modeling like never before.
While challenges of benchmarking with 3D data within the ICME paradigm still
exist, such as error propagation from experimental results and the handling of large
datasets, the need and opportunity for making this level of analysis more routine and
readily available for model development is clear.
D. W. Eastman et al.
6 Chapter Summary
Microtensile testing protocols and unique characterization routes have been developed and employed to obtain multiscale, microstructurally dependent benchmarks
for CPFEM models of commercial polycrystalline alloys, such as René 88DT. Initial
microtensile testing was performed on samples with gage cross sections on the
order of a few hundred microns machined via Wire EDM and on much smaller
samples with gage cross sections of 20 × 20 μm machined using a FIB. The
samples machined with wire EDM exhibited a mechanical response that was very
similar to the bulk material. The much smaller FIB-machined samples were more
tractable for CPFEM simulations but were found to be too small to capture a true
polycrystalline response. A femtosecond laser machining protocol was developed
and used to machine microtensile samples with cross-sectional dimensions of 50–
100 microns at a rate that was an order of magnitude faster than FIB milling;
machining of samples 25 times larger than could be achieved with the FIB was
accomplished in a matter of minutes.
Testing samples of different sizes revealed an inherent effect of sample size
on yield strength. It was seen that the sample strength decreased with decreasing
sample size and that the variability of the measured strengths increased as the
sample size decreased. These variations are attributed to the influence of geometric
averaging of oligocrystalline specimens. Combining the microtensile tests with
local strain and orientation mapping provided robust microstructural benchmarks
for direct comparisons with CPFEM simulations. More than just acquiring the bulk
stress-strain response of a sample, investigating local response of the material and
the corresponding microstructural features that lead to this behavior provides a new
level of benchmarking for ever-improving CPFEM modeling capabilities. Local
strain maps can be used to compare directly with simulations instantiated with the
microstructural data of these samples. This explicit benchmark is both quantitative
in terms of the local magnitudes of strain observed as well as qualitative in terms
of where in the microstructure these concentrations are located. Moreover, the
correlation of microstructural information with strain maps enhances mechanistic
understanding and provides detailed microstructural benchmarks. Ensuring that this
behavior is captured in CPFEM simulations should be a key goal of future model
development.
The long-term goal of this work was to be able to explicitly characterize tested
samples in 3D to be able to instantiate simulations and collect attendant 3D
benchmarks of mechanical behavior. The development and availability of advanced
characterization and mechanical testing techniques, like the ones outlined in this
chapter, make it possible to marry experimental and modeling like never before.
While challenges of benchmarking with 3D data within the ICME paradigm still
exist, such as error propagation from experimental results and the handling of large
datasets, the need and opportunity for making this level of analysis more routine and
readily available for model development is clear.