Baseline model

A simplified baseline model is created to represent what has been observed in two fatigue experiments on AA7075-T651 double edge-notched (DEN) specimens [59]. In these experiments, observations of MSFC incubation, nucleation, and

propagation were made in a 1.50 mm x 0.50 mm region within one of the two notch roots [32]. The loading condition in the notch root is approximately R=0.1 (min/max strain ratio) with 1% maximum strain in the RD. Figure 1.3 illustrates the DEN notch root, the three-dimensional microstructure, and the notch root loading conditions simulated here and in previous studies [7,31,32]. The simulated boundary conditions in this study are the same as those detailed by Bozek et al. [7]:

constrained, axial tension with R and maximum strain as given above.

The DEN experiments revealed MSFC nucleation occurs within O(10¹) to O(10³) load cycles, and subsequent MSFC propagation through the first grain requires another O(10²) to O(10³), or more, load cycles. The upper bound on the number of cycles to propagate a nucleated MSFC through the first grain could not be determined, because the DEN specimens failed prior to this occurring for some of the observed cracks. Figure 1.1 is a set of microstructural images from one of the observed MSFC’s that propagated the fastest. A comparison of the in-verse pole figure, Figure 1.1(d), to the scanning electron micrographs, Figure 1.1(a) through Figure 1.1(c), shows the crack nucleated between 100 and 300 cycles, and propagated through the first grain by approximately 1000 cycles.

Due to the computational intractability of modeling all load cycles through in-cubation, nucleation, and early MSFC propagation, this baseline study simulates an MSFC as stationary during two cycles of loading. From the mesh convergence study presented in Section 1.2, it was determined that the mesh size required for convergence of near crack front fields consumes approximately 10 thousand CPU hours of run-time per load cycle. This is computationally tractable to thousands of load cycles for one model simulation, assuming access to a large computational cluster with approximately 10 million CPU hours of available run-time, but

repe-tition of this task for multiple model simulations is currently intractable. Since the cracks are instead modeled as stationary, the plastic wake of a propagating MSFC is not captured. This wake is likely to reduce CT D at any load and can theoret-ically cause crack closure at low applied tensile loads. However, for the loading considered here, it is assumed (observations of closure were not made) the crack does not close at minimum load. Therefore, ∆CT D computed for a stationary MSFC model should be a sufficient approximation of a propagating MSFC. Since MSFC propagation is not simulated here, the actual stress and plastic slip states near the cracks are not accurately produced. However, the intent of this study - to compare results from a parametric study and begin to understand the relative in-fluence of variations among microstructural heterogeneities on MSFC propagation metrics - can be satisfied by modeling a stationary crack.

The baseline model geometry, Figure 1.4, represents four grains of typical ND dimension for AA7075-T651 and a surface particle likely to crack under the afore-mentioned loads. The ND grain dimension, L^{N D}_g , is 20 µm because this is a typical dimension observed in AA7075-T651 microstructure [76]. Observed RD and TD grain dimensions are not used or necessary in the baseline model, because the largest MSFC size modeled is much smaller than either dimension. RD and TD dimensions of L^RD_g =60 µm and L^{T D}_g =40 µm, respectively, are used to eliminate boundary effects on the crack. The particle radii in the ND and TD, L^{N D}_p and L^{T D}_p , respectively, are both 2 µm, because these are common radii in both dimen-sions for Al₇Cu₂Fe particles in this material, based on statistical data recorded by Harlow et al. [29]. The RD particle radius, L^RD_p , is 6 µm, because this is also a common radius and the resulting particle size and aspect ratios frequently cause particle cracking, as predicted by the incubation filter from Bozek et al. [7].

More importantly, this particle is predicted to crack for all of the following grain

grain C

Figure 1.4: Baseline model geometry. (a) Perspective view of entire model with grain identifier and dimension labels. (b) Magnified view of particle and crack with particle and crack dimension labels.

orientations modeled here. The particle-grain interface and all grain boundaries are simulated as perfectly bonded, because no debonding was observed at these material interfaces during experiments.

Three grain orientations are modeled with the crystal plasticity formulation from Matous and Maniatty [48] and the AA7075-T651 constitutive parameters given by Bozek et al. [7] and in Appendix A. Each texture, Table 1.1, is common in AA7075-T651, and together they represent three different plastic slip character-istics, revealed by the Schmid factors of their 12 primary slip systems, Table 1.2.

The three orientations are referred here as the ‘weak’, ‘rotated’, and ‘strong’ ori-entations. The weak orientation has one slip system with Schmid factor close to 0.50, which will result in early onset of plastic slip in comparison to an orientation with no high Schmid factors. The rotated orientation can also be categorized as weak since it actually has two high Schmid factors, but the delineation refers to the rotated (or twisted) cube texture family to which this orientation belongs [58, 76].

Table 1.1: Three face-centered cubic (FCC) grain orientations investigated in the baseline model.

Bunge Euler angles [rads]

orientation ID ϕ1 Φ ϕ2

weak 3.988913 0.730795 2.806933 rotated 3.803586 -0.661472 2.590874 strong 4.390221 0.721284 1.114852

Table 1.2: Schmid factors for the FCC grain orientations in Table 1.1 and loading along RD.

12 slip systems

1 2 3 4 5 6 7 8 9 10 11 12

weak 0.00 0.00 0.00 0.20 0.33 0.13 0.33 0.20 0.13 0.46 0.13 0.33 rotated 0.49 0.29 0.20 0.29 0.21 0.08 0.29 0.20 0.49 0.21 0.29 0.08 strong 0.02 0.22 0.24 0.03 0.24 0.26 0.33 0.01 0.32 0.07 0.09 0.02

The strong orientation has no high Schmid factors and a higher yield strain than the other two orientations. Three unique types of plasticity are expected to oc-cur in the weak, rotated, and strong orienations, respectively: early yielding with initial slip localization in one direction, early yielding with initial slip localization in two directions, and late yielding with slip localization in three or more direc-tions. The reader must be aware that the word choices of ’weak’ and ’strong’ are referring to the relative yield strengths, and not the fracture toughnesses, of these orientations.

Three crack lengths, i.e. radii, a, of a semi-circular surface crack, are also mod-eled to simulate three sizes of an MSFC inside the first grain(s) encountered after nucleation. The first crack length is a=a₁=3µm to simulate an MSFC immediately subsequent to nucleation, when the crack is likely still within the micro-notch root influence of the particle, i.e. the region of elevated stresses due to the high stiff-ness of the particle relative to the grain stiffstiff-ness. The second crack length is a=a₂=11µm to simulate an intragranular MSFC that is equi-distant from both the particle-grain interface and the next grain boundary. The third crack length is a=a₃=19µm to simulate a transgranular MSFC as it is 1µm from reaching grain C on the RD-ND surface shown in Figure 1.4(a). The crystal plasticity model em-ployed in this study has no inherent length scale, but the crack varying in size and location with respect to the fixed sizes and locations of microstructural features will influence the computed fields among modeled crack sizes.

Table 1.3 is a summary of all orientations and crack lengths studied for the base-line model in Figure 1.4. For each orientation and at each crack length, a model is simulated where grains A and B both have the same orientation to resemble a crack nucleating inside a grain. Conversely, there are other models where grains A and B have differing orientations to resemble a crack nucleating on a grain bound-ary. In all but three models, the particle is modeled to best represent Al₇Cu₂Fe:

linear elastic, isotropic with a Young’s Modulus of 166 GPa and a Poisson’s Ratio of 0.3. However, for the strong orientation and an intragranularly nucleated crack, alternate models are made for all three crack lengths where the particle is assigned the strong FCC orientation. These alternate models are compared to the models where the elastic particle is present to determine the influence of the particle on MSFC metrics.

Multiple microstructural fatigue metrics are followed in the neighborhood of the crack. For the previously motivated reasons, ∆CT D is utilized to measure crack blunting. Since blunting is caused by permanent deformation in the neighborhood of the crack, the mechanism of such deformation for FCC materials - accumulated irreversible slip on the 12 primary slip planes - is to be calculated and compared with CT D. The same five slip-based metrics studied by Hochhalter et al. [31, 32]

for nucleation, Equation 1.3 through Equation 1.7, are used here.

D₁ = max of all slips accumulated on the 12 slip systems. D₄ is a measure of the maximum of energies dissipated due to plastic slip on the 4 slip planes, where energy is the product of the slip rate, ˙γ_p^α, and resolved shear stress, τ_p^α, on a plane. D5 is also a measure of the maximum of energies dissipated on the 4 slip planes, but this last metric includes stress normal to the plane, σ_n^p. Since Hochhalter et al. [32]

found the increase rate of irreversible slip per cycle, ∆D_i for i = 1...5, to be related to the number of cycles required for nucleation, this metric is also analyzed here. Maximum stress, both the maximum principal stress, σ₁, and the maximum tangential stress, σ^max_θθ , in the neighborhood of the crack and angular location, θ, of

Table 1.3: All permutations of crack sizes and region properties simulated for the baseline model shown in Figure 1.4. Region properties are identified as ‘elastic’ for linear elastic, isotropic or one of the three texture orientation identifiers from Table 1.1 for crystal plastic.

region properties

model ID crack size particle grain A grain B grain C

M1 a1 elastic strong strong rotated

M2 a₁ strong strong strong rotated

M3 a₁ elastic rotated rotated rotated

M4 a1 elastic weak weak rotated

M5 a₁ elastic strong weak rotated

M6 a₁ elastic strong rotated weak

M7 a1 elastic rotated weak strong

M8 a₂ elastic strong strong rotated

M9 a₂ strong strong strong rotated

M10 a2 elastic rotated rotated rotated

M11 a₂ elastic weak weak rotated

M12 a₂ elastic strong weak rotated

M13 a2 elastic strong rotated weak

M14 a₂ elastic rotated weak strong

M15 a₃ elastic strong strong rotated

M16 a3 strong strong strong rotated

M17 a₃ elastic rotated rotated rotated

M18 a₃ elastic weak weak rotated

M19 a3 elastic strong weak rotated

M20 a₃ elastic strong rotated weak

M21 a₃ elastic rotated weak strong

σ^max_θθ along the non-local arc are also investigated, because Hochhalter et al. found maximum tangential stress to also be related to the number of cycles to nucleation and θ to align with the direction of nucleation.