Evaluation of Microstructure and Mechanical Properties of Al-Zn-Mg-Cu Alloy Repaired via Additive Friction Stir Deposition

As the need for maintenance and repairs of aging aircraft continue to rise [1,2], there is a growing necessity for cost efficient and effective methods of repair for flight-critical components. Additive manufacturing (AM) has become an attractive approach for fabricating and repairing aerospace components [3], in part, due to the benefit in reducing production costs associated with build time and waste material of traditional manufacturing methods [4]. However, for fusion-based AM processes (e.g., selective laser melting and electron-beam melting), certain alloys suffer from poor weldability [5–8], which inhibit the additive fabrication or repair. One such alloy system is Al-Zn-Mg-Cu aluminum alloys, in particular AA7075, which exhibit hot cracking and porosity defects via fusion-based AM methods [9–12].

The AA7075 is a precipitate strengthened aluminum alloy with the addition of zinc, magnesium, and copper alloying elements. AA7075-T651 is strengthened from the formation of two primary phases, ŋ' (MgZn₂) and ŋ (MgZn₂), which are formed from Guinier-Preston zones during heat treatment [13–15]. After peak aging, with continual input of thermal energy, additional secondary phases can form along grain boundaries such as S-phase (A_l2CuMg), E-phase (Al₁₈Mg₃Cr₂), and the T-phase (Al₃₂(Mg,Zn)₄₉) [16–18]. These phases typically are brittle stress risers that can lead to primary crack nucleation during cyclic and static loading. For wrought AA7075-T6, the primary damage mechanisms for both static and cyclic loading are derived from iron-rich and silicon-based intermetallics; Al₆(Fe,Mn), Al₃Fe, Al(Fe,Mn,Si), Al₂₃Fe₄Cu and Al₇Cu₂Fe, and Mg₂Si [19–21]. These constituent particles are brittle and incoherent to the aluminum matrix that either crack or debond from the matrix, nucleating cracks that facilitate fracture.

It is well established that AA7075 is traditionally considered unweldable alloy, and when subjected to high thermal gradients, hot cracking occurs in the microstructure. Therefore, fusion-based AM, in which high thermal gradients are introduced into the microstructure, typically results in hot cracking and material anisotropy when fabricating or repairing AA7075. These deleterious defects on microstructure reduce the mechanical response of the material. Additionally, due to these challenges in AM of AA7075, there are only a few studies on beam-based manufacturing of AA7075 [10–12,22–24]. However, a new solid-state AM process for fabrication and repair, additive friction stir deposition (AFS-D), mitigates these thermal defects by facilitating material deposition below the melting temperature of the alloy.

The AFS-D process is a solid-state AM process that relies on similar physics of friction stir welding (FSW) to additively deposit material to a substrate and subsequent additive layers. Illustrated in Fig. 1(a), a hollow rotating tool comes into contact with the substrate to generate frictional heat that softens the work pieces as well as the feedstock. The feedstock is then deposited through the center of the hollow rotating tool via an actuator that forces material onto the softened substrate. The combination of pressure, frictional heat, and severe plastic deformation of both substrate and feedstock creates a metallurgical bond, which allows for deposition of consecutive layers upon a substrate. As such, the AFS-D process provides a new pathway for rapid repair of aerospace components without intrinsic microstructural defects typically exhibited in fusion-based AM technologies.

There are limited published studies dedicated to the AFS-D process [25–35], and fewer focusing on its validity for repair applications. Joey Griffiths et al. [36] evaluated the hardness and microstructure of a repaired AA7075 plate in the as-deposited configuration. The repaired plate originally had a singular hole that was filled in using the AFS-D process. Additionally, Avery et al. [30] investigated the fatigue behavior of an as-deposited monolithic AFS-D AA7075 component, and its resulting microstructure. Lastly, Yoder et al. [37] conducted a heat-treatment study of AFS-D produced AA7075. The authors demonstrated that the heat-treated AA7075 achieved 95% of yield stress (YS), 94% of ultimate tensile strength, and 78% elongation to failure in comparison with wrought AA7075-T6. To the best of the authors' knowledge, these are the only studies on AFS-D processing of AA7075. Therefore, this paper will focus on elucidating the process-structure-property-performance (PSPP) relationships of AA7075 repaired via AFS-D. In this work, a thorough microstructural characterization of the AFS-D repair of damaged AA7075 plate is carried out via a combination of optical microscope, scanning electron microscope (SEM), and transmission electron microscope (TEM) techniques. To quantify the mechanical response of heat-treated AA7075 repairs, monotonic and fully reversed strain-controlled fatigue experiments were conducted. Lastly, post-mortem analysis was conducted to further quantify the process-structure-property relationships of heat-treated AFS-D repair and wrought AA7075-T651.

In this study, a commercially available Meld Manufacturing B8 AFS-D machine was used to deposit square cross-section of 9.53 mm solid AA7075-T651 feedstock rods, with nominal composition (wt%) Al–5.6Zn–2.6Mg–1.6Cu, which were machined using a wire EDM from AA7075-T651 rolled plate. The same rolled plate that the feedstock was machined from was also used as the substrate for the AFS-D repair. Prior to the deposition process, feedstock rods were coated with a dry graphite lubricant as specified by the machine manufacturer, Meld Manufacturing, in order to minimize friction through the hollow tool during deposition. Without this lubricant, the deposition process becomes unattainable due to jamming via thermal expansion of feedstock within the hollow tool. The feedstock rods were deposited onto a 6.35 mm thick AA7075-T651 plate with a milled groove that simulated a damaged plate, as shown in Fig. 1(a). The repaired plate, Fig. 1(b), was 304 mm in length, 177 mm in width, and 6.35 mm in thickness and was constructed with a single depositional layer. The depositional parameters were 225 rpm tool rotational speed, with feedstock feed rate of 50.8 mm/min, and traversing velocity of 50.8 mm/min.

The simulated-damaged groove was machined from a 6.35 mm thick AA7075-T651 plate, using a 12.7 mm diameter ball end mill. The groove was machined into the surface of the plate to a depth of 3.175 mm. After repairing the plate with the AFS-D process, the plate was milled flat in order to create a uniform gage and to remove unwanted flash. It is important to note that during deposition, the tool face/shoulder was in contact with the substrate for the entire length of the deposition, which is illustrated in Fig. 2. Therefore, the feedstock was directly deposited within the milled groove. Additionally, there are four protrusions on the tool face that extend 2.23 mm into the substrate. Figure 2 displays how the protrusions are positioned in pairs and are staggered in location from the center of the tool. Thus, the inner protrusions have a significant influence on the stirred material deposited within the damaged groove. However, due to the depth of the damage grooved (3.175 mm), there is a separation of a direct protrusion stir zone (DPSZ) and nonprotrusion stir zone (NPSZ) as illustrated in Fig. 2.

In the work by Avery et al. [30] on the characterization of as-deposited AA7075 processed via AFS-D, a reduction in mechanical properties due to loss of strengthening precipitates was quantified. Therefore, this investigation of AFS-D AA7075 repair considered the effect of heat treatment on the as-deposited samples in order to achieve a T6 temper to regain wrought-like mechanical properties. Post-processing heat treatment of samples was performed according to ASM standards to recover the T6 temper [38].

Due to the natural aging properties of AA7075, a sample of the as-deposited material was naturally aged for a minimum of 4400 h. The as-deposited sample was used for microstructural comparative analysis to the heat-treated AFS-D condition. The as-deposited and heat-treated AFS-D repair was ground and polished to 3 µm diamond suspension for Vickers microhardness measurements across the entire cross section of the repair to elucidate any mechanical differences throughout the repair. Microhardness measurements were conducted with a Clemex Automated Microhardness Tester using parameters: 300 gf, 10 s dwell, 500 µm indent spacing, and grid spacing of 6 × 24 mm.

Metallographic samples of the as-deposited and heat-treated AFS-D repair cross section were ground and polished to 3 µm diamond suspension followed by vibratory polish using 0.03 µm colloidal silica. An Apreo SEM and Tescan Lyra FIB-FESEM, coupled with an EDAX Hikari Super Electron Backscatter Diffraction (EBSD) camera and Octane Elite Silicon Drift Detector, was used to perform Euler EBSD grain size, electron dispersive X-ray spectroscopy, and post-mortem analysis. The EBSD scans were performed at the same magnification in all three build locations and orthogonal directions over an area of approximately 160 × 160 µm, with a step size of 0.5 µm. The scans were performed at 15 kV and the EBSD data was post-processed using the grain dilation algorithm in the OIM software with settings of a minimum grain size of 3 pixels and requiring that a grain must contain multiple rows of pixels. The grain size analysis was produced in compliance with ASTM 2627-13.

Cross-sectional samples for the TEM analysis were prepared using FEI Quanta 3D dual beam SEM-focused ion beam (SEM-FIB) equipment. In FIB-SEM an acceleration voltage of 30 kV was used to produce the electron transparent samples. To remove the Ga ion implantation during the FIB milling process, an Ar-ion milling in a Gatan precision ion polishing system was conducted using an acceleration voltage of 0.5 kV, incident angle of 5 deg, and milling time of 45 s. Bright field TEM images and selected area electron diffraction patterns were acquired using a FEI Tecnai F-20 TEM operated at 200 keV beam energy.

For monotonic and fatigue testing, a modified ASTM E606/E606M-12 specimen design was used with a 25.4 mm gauge extensometer as shown in Fig. 1(c). The quasi-static monotonic and fatigue specimens were machined using a wire EDM in the transverse direction for heat-treated AFS-D repair specimens (Fig. 1(b)), where the wrought specimens were oriented in the rolling direction. A MTS 60 kN servo hydraulic load frame was used for the monotonic tests in strain control at a rate of 0.001/s until fracture at ambient temperature. For the fatigue specimens, the surfaces were low pressure ground using 600 grit sandpaper prior to testing. Strain-control fatigue experiments were conducted using a MTS servo hydraulic load frame equipped with a 60 kN load cell. Fatigue specimens were tested at R = −1 in ambient temperature and relative humidity (50%), until fracture. Upon fracture, the specimen fracture surfaces were mounted for SEM fractography analysis to determine crack initiation and crack propagation mechanisms.

The microstructure-sensitive fatigue (MSF) model implemented in this study was first proposed by McDowell [39,40] for aluminum alloy A356-T6 [41]. The model was further expanded to additional aluminum alloys [42–45], other materials [46–52], and AM processes [30,53,54]. To elucidate the microstructural effect of AFS-D on fatigue resistance, the MSF was calibrated for the AFS-D using initial calibration parameters determined by Avery et al. [30]. For the purpose of this paper, modeling constants from Avery et al. [30] were used as a basis for the MSF model calibrations for wrought and heat-treated AFS-D repair. For brevity's sake, details on the MSF model and the application to the AFS-D process can be found elsewhere [30].

Figure 3(a) displays a macroscopic view of a cross section of the heat-treated AFS-D repaired plate etched with a 20% sodium hydroxide solution. A hardness profile across the repaired region is displayed in Fig. 3(b), which revealed an average hardness across the sample of approximately 210 HV. Figures 3(c) and 3(d) are optical images of the refined microstructure within the repair and interface of the repair, respectively. In Fig. 3, the original milled groove that was repaired is outlined in black, whereas the region of the repair that is not directly stirred by the protrusions on the face of the AFS-D tool is outlined in red (NPSZ). From the optical micrograph, Fig. 3(a), the two regions DPSZ and NPSZ exhibit separate morphological differences in microstructure. The DPSZ shows a uniform microstructure of refined grains (slightly darker region in micrograph) within the repaired grooved (outlined in black) that extends outward past the original damaged area into the surrounding microstructure, due to the second set of outer protrusions displayed in Fig. 2. The NPSZ outlined in red of the original simulated damage region, exhibits unique flow lines of feedstock and substrate material mixing. This nonuniform mixing is seen only in the NPSZ and the interface between NPSZ and DPSZ. Furthermore, as seen in Fig. 3(b), the microhardness measurements demonstrated a 210 HV in the heat-treated material, which is above the nominal wrought Vickers hardness of 175 HV. It is hypothesized that the refinement of the microstructure, due to Hall-Petch effect, increased mechanical performance, which led to the overall increase in hardness measurement. The estimated contribution of the reduction in grain size corresponds with a 14 HV increase in hardness. However, the estimated Hall-Petch strengthening contribution does not elucidate why the average hardness of the heat-treated specimen is 210 HV. Therefore, other strengthening contribution has likely increased the mechanical performance. Note that the outlined middle region in Fig. 3(b) exhibits a 5% reduction in average hardness across the entire repaired cross section. This reduction in hardness is postulated to be the heat affected zone of the repair.

Figure 4(a) displays the macroscopic etched cross section of the AFS-D repair, with areas of interest highlighted. Figures 4(b)–4(d) display the grain morphology and distribution at the top of the repair, middle of the repair, and the bottom repair, respectively. Figure 4(e) exhibits the interface of the bottom of the repaired groove to the substrate, wherein the interface displays a thin line of further refined grains. The microstructure outside of the repair exhibits a similar equiaxed grain refinement found in the repaired region, as shown in Fig. 4(f). For comparative purposes, an EBSD map of the substrate material away from the influence of the repaired plate region exhibited a grain size of 100 + µm, as seen in Fig. 4(g). From Fig. 4, it is clear that the repaired AA7075 plate exhibited a refinement in the microstructure from the top to the bottom of the repair within the simulated-damaged groove. This refinement is notable in comparison with the initial microstructure of the substrate seen in Fig. 4(g) where large grains dominate the microstructure. The refinement of the microstructure is attributed to the continuous dynamic recrystallization (CDRX) facilitated from the high-shear strain and hot deformation of the AFS-D process. Additionally, further refinement of the microstructure is seen at the interface at the bottom of the repaired groove and substrate, Fig. 4(e). It is postulated that due to being in the NPSZ, deposited material flow within the region is discontinuously stirred within the bottom of the repair, which leads to material flow lines of substrate and deposited material seen in Fig. 4(a). Therefore, due to the inadequate flow of material, as more feedstock is deposited within a region, it builds up pressure at the boundary of the bottom of the repair and substrate, further refining the grains. Notably, the heat-treatment process did not cause any observable abnormal grain growth observed.

Figure 5 displays the inverse pole figure EBSD maps of an as-deposited repair specimen (Figs. 5(a)–5(c)) and heat-treated repair specimen (Figs. 5(d)–5(f)) for the top, middle, and bottom sections of the repair, as shown previously in Fig. 4, for comparative purposes of grain size and morphology, respectively. Quantitative grain size analysis of the as-deposited repair is observed in Fig. 5(g). Cumulative grain size plot analysis of the heat-treated repair is displayed in Fig. 5(h). From the cumulative grain size plot in Fig. 5(g), the as-deposited repair exhibited an average grain diameter of 5 µm in the top and middle regions, whereas the bottom of the repair has an average grain diameter of 6 µm. This difference in grain size is postulated to be a result of the direct protrusion stir zone (Fig. 2), where the protrusions on the tool face have direct influence and mechanical stirring of the microstructure. The protrusions act as sources of deformation with higher strain rates that allows for CDRX to occur more readily. Thus, the protrusions allow for grain nucleation to occur at a higher rate and more evenly in comparison with the bottom of the repair (NPSZ), where deformation is dominated by the direct pressure of the depositing feedstock and far-field strain of the protrusions and tool shoulder into the substrate, as illustrated in Fig. 2. The heat-treated AFS-D specimen exhibited similar trends of grain size where the grain size at the top of the repair was 3.5 µm, the middle of the repair was 4.5 µm, and the bottom of the repair was 6 µm, as seen in Fig. 5(h).

Figure 6(a) shows the macroscopic optical image of the heat-treated AFS-D repair cross section, with designated area of interest at the interface of the DPSZ and NPSZ. Figure 6(b) displays the EBSD map highlighting flow lines of material mixing of the substrate and feedstock. The distribution of the high-angle (>10 deg) grain boundaries are displayed in black and the low-angle (<9.5 deg) grain boundaries are shown in red as displayed in Fig. 6(c). Figure 6(d) shows the grain reference orientation deviation (GROD) map of the refined microstructure at the DPSZ/NPSZ interface. The inverse pole figure map (Fig. 6(b)) of the material flow lines exhibits a boundary layer of substrate material to feedstock that produces an interface of finely equiaxed grains. This is further illustrated by the grain boundary map, seen in Fig. 6(c), wherein an abundance of high angle to low angle grain boundaries were observed at the mixing interface, which is evidence for CDRX. This interface was observed to contain nucleated defect-free crystallites (blue grains) via CDRX validated by the low amount of average intergranular misorientation, as shown in Fig. 6(d). Thus, the interface of mixing between substrate and feedstock is a site for grain nucleation from CDRX. It is important to note that the material flow lines directly correspond to the NPSZ where the tool face protrusions do not directly stir the material into the bottom of the repair. Therefore, we postulate that the NPSZ experienced a shear-strain gradient from the DPSZ/NPSZ interface to the bottom of the repair. Consequently, the nonuniform shear strain produces inadequate mixing of substrate to feedstock, revealed by material flow lines.

The AFS-D process deposited material with the combination of heat and pressure on the feedstock and the substrate material. Thus, an important factor in repairing any damaged part is determining the influence of the repair process on the base material (substrate). Therefore, Fig. 7 exhibits the microstructural evolution of the substrate material directly under the repaired groove, and the effect AFS-D has on nonrepaired base material (substrate). Figure 7(a) displays an optical micrograph of the repaired section with areas of interest corresponding to base material below the repaired groove. Figures 7(b), 7(e), and 7(h) correspond to the EBSD maps ranging from just under the repaired milled groove to the bottom of the substrate. Figures 7(c), 7(f), and 7(i) display the grain boundary maps for the substrate locations, and Figs. 7(d), 7(g), and 7(j) show the GROD maps which elucidate the average intergranular misorientation. From Fig. 7, it is clearly depicted that the microstructure transitions from a nondeformed state at the bottom of the repaired plate (Figs. 7(h)–7(j)), to a highly deformed state illustrated by the abundance of low angle grain boundaries and sub-grains directly underneath the repaired milled grooved, as seen in Figs. 7(b)–7(d). The high forces and thermal input of the AFS-D process has changed the substrate microstructure into a gradient of deformation, seen in Figs. 7(d), 7(g), and 7(j) as the accumulation of geometrically necessary dislocations. The deformation is highlighted by the concentration of low angle boundaries, indicating crystallographic misorientation, due to the force and heat exerted onto the work piece. As pressure and heat are applied, sub-grains are formed within the larger grains, seen in Fig. 7. These sub-grains and high-angle grain boundaries become nucleation sites for recrystallized grains. The GROD maps, Figs. 7(d), 7(g), and 7(j), depict a narrow band of defect-free grains located between larger highly deformed grains that are formed within the microstructure directly under the repaired groove. Thus, the substrate undergoes a microstructural evolution as discontinuous dynamic recrystallization nucleates new grains from the deformed original microstructure.

Figure 8 shows the backscattered SEM images of the polished as-deposited (A) and heat-treated (B) surfaces. The representative SEM micrograph of the as-deposited repair, (Fig. 8(a)), reveals Zn-Mg-Cu rich precipitates at the grain boundaries. Rutherford et al. [35] reported that these precipitates were due to the repeated thermal cycles and overaging of the AA7075 microstructure. While the repair process in the present study involved only a single deposition pass, the thermal input was sufficient to dissolute and precipitate out additional phases. Conversely, Fig. 8(b) shows the microstructure of the heat-treated sample that underwent solutionizing treatment for dissolution of the Zn-Mg-Cu rich precipitates and reestablishment of η' (MgZn₂) and η (MgZn₂) strengthening phases. From the backscatter SEM images, it is clear that the heat-treatment process successfully solutionized the Zn-Mg-Cu rich particles. Two sets of nanoscale particles were visible in these microstructures. Figure 8(c) displays a representative TEM bright field image from the as-deposited AFS-D repair sample. The TEM image confirms the presence of 100–200 nm diameter Zn-Mg-Cu rich particles in the grain interiors. The slightly larger particles are visible at the grain boundaries. Figure 8(d) shows a representative TEM bright field image of a heat-treated AFS-D repair sample. The same distribution of 100–200 nm diameter intermetallic particles were observed inside the grains. The grain boundary regions exhibited numerous ∼15–30 nm sized precipitates but lack the larger phases along the grain boundaries.

Figure 9 displays the tensile stress–strain experimental results of the wrought AA7075-T651, AFS-D as-deposited, and AFS-D heat-treated repaired material. All tensile curves shown in this plot are representative of three samples tested in each condition. The wrought AA7075-T651 exhibited a higher average yield stress (YS) at 512 MPa, ultimate tensile stress (UTS) at 558 MPa, and elongation to failure of 17%, when compared with as-deposited and heat-treated samples. The as-deposited specimens exhibited a YS of 285 MPa, UTS of 421 MPa, and elongation to failure of 6.5%. The as-deposited specimens exhibited the lowest mechanical performance in comparison with the heat-treated and wrought experimental data, but outperformed monotonic AFS-D results from Avery et al. [30] that showed YS of 150 MPa and UTS of 295 MPa. The reduction in mechanical properties reported by Avery et al. [30] is attributed to the increased amount of thermal cycles for the multilayered larger build in that study. However, in this present study, the repaired AFS-D AA7075 was only subjected to a single thermal cycle to complete the repair. Therefore, the strengthening phases η' and η had less thermal activation and diffusivity to grow in size, leading to a reduction in the ability to impede statistically stored dislocations before crack nucleation and fracture. It is important to note that both the as-deposited and as-received wrought samples had very tight clusters with regard to monotonic tensile results, and as such are plotted as averaged lines. In contrast, heat-treated samples had significant scatter in strain to failure, which can be seen in Fig. 9(b). However, mechanical properties where calculated with only the YS being comparable between all samples, which had an average stress of 490 MPa YS, while UTS and elongation to failure varied from 530 to 555 MPa and 2–9%, respectively. Thus, the heat treatment allowed for direct recovery of strengthening precipitates to achieve wrought-like strength properties and to recover Zn-Mg-Cu particles as seen in Fig. 8(a). The monotonic tensile properties are summarized in Table 1.

Figure 10 shows the Kocks-Mecking plot of the strain-hardening rate (θ = dσ/dε) against the net flow stress (σ−σ_y) at a strain rate of 0.001/s for the wrought, as-deposited repair, and heat-treated repair. From Fig. 10, it is clear that there is no stage I or stage II hardening behavior observed. Stage III was observed at the onset of yield for all three sample conditions followed by stage IV. The wrought AA7075-T651 had an initial θ of about 12,500 MPa and then decreased rapidly exhibiting stage III hardening behavior until saturation at 20 MPa net flow stress, wherein stage IV hardening dominated. The as-deposited repair exhibited an initial θ of about 11,000 MPa, then decreased rapidly until saturation at 40 MPa net flow stress, where stage IV hardening then dominated. Lastly, the heat-treated condition had an initial θ of about 16,500 MPa and then decreased rapidly exhibiting stage III hardening behavior until saturation at 40 MPa net flow stress, wherein stage IV hardening dominated. Interestingly, the wrought material experienced the largest linear decrease in stage II hardening with the largest slope in comparison with the AFS-D repair samples. The as-deposited repair results are similar to FSWed AA7075 results wherein the base material experienced the highest degree of dynamic recovery in its strain hardening [55,56]. Thus, for the as-deposited and heat-treated repair, the decreased stage III hardening corresponds with the refined microstructure of the equiaxed grains. Due to CDRX, defect-free grains with low dislocation density were present within the microstructure and during loading, resulted in a higher density of dislocations being generated than recovered. Thus, an increase in the hardening capability of the material was observed, where Table 1 displays the hardening exponents for all three conditions with the as-deposited displaying the highest hardening exponent at n = 0.1971 and wrought with the lowest n = 0.0409.

Regarding the cyclic behavior, Fig. 11 depicts the strain-controlled fatigue experimental results for wrought AA7075 and the heat-treated AFS-D repair. The fatigue results show a clear trend where the heat-treated specimens exhibited a reduction in the average cycles to failure compared with wrought. While the refinement of the microstructure, along with reducing stress risers in the material by refining intermetallic particles in combination with reestablishing strengthening phases, was expected to produce fatigue behavior of the AFS-D repair similar to the wrought control, a clear decrease in the fatigue resistance was observed. The likely cause of the observed reduction in fatigue resistance is discussed later.

For comparison, Fig. 12(a) shows an SEM image of wrought specimen WR6 that was fatigue tested at 0.3% strain amplitude and failed at 664,742 cycles. Figure 12(b) displays a magnified SEM image of the striations that progressed from the crack initiation site near the free surface. Figure 12(c) is a magnified SEM image of the crack initiation site found near the free surface and with river marks that progressed radially outward away from the nucleation site. Figure 12(d) shows a backscatter SEM image of the initiation site, where brighter iron-rich intermetallics can be seen at the surface (white arrows), as well as fractured particles (black arrows). These iron-rich intermetallics likely facilitated crack nucleation, where iron-rich intermetallics have been reported to be the primary source of fatigue crack nucleation in wrought AA7075-T651 [57].

Figure 13(a) displays the macroscopic SEM image of the AFS-D heat-treated repair fracture surface of specimen HT11 cycled at a strain amplitude of 0.2% until failure at 535,508 cycles. Figure 13(b) is a magnified SEM image of the crack initiation point at the center of the specimen, where a boundary line at the repair/substrate interface exhibits large defects. Figure 13(c) is a further magnified SEM image of the defect area, where a tumultuous area of aggregate particles was observed to reside on the grain boundaries. Figure 13(d) displays a representative area of the defect region wherein intergranular cracking was observed to occur around small aggregate particles. From the macroscopic SEM image in Fig. 13(a), the fractured heat-treated sample displays two distinct regions in the fracture surface that will be discussed subsequently. The regions correspond to the AFS-D repair and substrate side of the repaired specimen. The AFS-D repair region is characterized with a very flat region of crack growth that protruded through a majority of the region, with river marks that flowed from the center of the specimen and grew radially outward. Conversely, the substrate side of the repair exhibited a typical tumultuous morphology of crack growth with micro-cliffs of crack coalescence present and abundant within the region. However, the dominating feature of the fracture surface is the fish-eye crack initiation point in the center of the sample. These fracture surface phenomena are typically observed due to metallurgical defects in material processing, typically with insufficient mixing and metallurgical bonding of the material. Yet, recalling Fig. 4, the AFS-D heat-treated repair was fully dense at the bottom of the repair/substrate interface where the fish-eye defect has originated. Furthermore, from Figs. 13(c) and 13(d), a regional band can be seen at the repair/substrate interface. The defect band is characterized with intergranular cracking between grain boundaries and small aggregate particles residing on the grain boundaries as well. These particles are seen randomly dispersed around grain boundaries and are on the order of magnitude of a nanometer length scale in diameter. The narrow band of the defect zone runs the length of most of the sample indicating that the distribution is not entirely localized to a few microns but on the order of magnitude of millimeters in scale, suggesting that the defect was a product of the AFS-D process. However, the presence of the fish-eye morphology and the defect zone in the fracture surface are likely the cause for the scatter in the monotonic and fatigue results.

Figure 13(d) displays the magnified SEM micrograph of the defect zone. Energy dispersive spectroscopy (EDS) was performed, at 5 kV to encapsulate the aluminum excitation spectrum and reduce beam spot size, on two regions within the defect zone, including a control with EDS chemical analysis on an area that exhibited no aggregate particles, and an area abundant of aggregate particles. The matrix EDS results seen in Fig. 13(e) display typical chemical analysis of the aluminum matrix of some Zn and Mg alloying elements present. However, the EDS spectrum of the aggregate particles, Fig. 13(f), suggests the presence of Carbon and Oxygen, signifying an oxide film that has formed at or around carbon particles. These carbon particles are hypothesized to either come from the wear of the steel tool face/protrusions or from the graphite lubricant that is applied to the surface of the feedstock before deposition. The oxygen is postulated to be from the reaction of the aluminum in open air forming an oxide film from the fracture surface. The fractography results indicate that the presence of carbon is due to the graphite lubricant used to minimize jamming of feedstock during processing because of the large distribution of finely sized particulates in the matrix observed in the micrographs. The use of lubricant has been observed to be necessary for depositions to prevent jamming within the hollow tool due to thermal expansion of feedstock that can occur when the lubricant is not used. Interestingly, examination of the location of the carbon defect zone in Fig. 13(d) at the boundary of the repair/substrate interface revealed that this region is at the bottom of the nonprotrusion stirred zone at the largest depth of the milled groove. As noted previously, this region experienced a strain gradient during deposition, where the actuator force and frictional heat dominated the stirring of the material. Thus, the graphite lubricant from the surface of the feedstock, due to lower shear strain at the bottom of the repair, likely did not evenly mix with the deposited material, but accumulated at the interface. However, the graphite was not seen to cause crack initiation or damage in monotonic or fatigue specimens in the as-deposited condition as reported by Avery et al. [30]. Thus, it is likely that during heat treatment of the AFS-D repaired AA7075, the increase in temperature allowed for thermodynamic diffusivity of carbon elements to diffuse into the grain boundaries and collectively create a defect network that ultimately reduced the mechanical properties of the material and created significant scatter in both tensile and fatigue behavior.

The MSF model correlations of total-life estimations for the wrought and heat-treated AFS-D repair of AA7075 are displayed in Fig. 14(a). The MSF model parameters for the wrought AA7075 were taken from [42], whereas the MSF model parameters for the AFS-D repair were based on calibration results by Avery et al. [30]. However in this present work, the specific microstructural input parameters, including grain size (5.25 µm) and inclusion size (1.7 µm) as it pertains to the AFS-D repair, were determined via metallographic examination.

Figures 14(b) and 14(c) display the MSF model correlations for the upper, mean, and lower bounds of the wrought and heat-treated repair. By varying the minimum and maximum microstructural features including the intermetallic particle diameter and grain size of the material, the MSF model was able to bound the experimental fatigue results, thus capturing the experimental fatigue scatter. From Avery et al. [30], the MSF model for the wrought material relied on the intermetallic particle diameter and grain size of 5.2 µm and 50 µm for the upper bounds, respectively. For the lower bounds, an intermetallic particle diameter and grain size, of 25.6 µm and 150 µm, respectively, was used. The upper and lower bounds for the heat-treated samples were determined using intermetallic particle diameter and grain size of 0.7 µm and 1.5 µm, respectively for upper bounds, and 7 µm and 14.5 µm, respectively for lower bounds. The upper and lower bounds of the MSF model for the wrought AA7075 correlated well and bounded the scatter of the experimental fatigue results. However, the MSF lower bounds of the heat-treated repair did not completely capture the scatter of the experimental results likely due to the presence of the nonuniform localized planar defects caused by the graphite lubricant.

The objective of this work was to quantify the process-structure-property-performance (PSPP) relationships of the solid-state AFS-D process for the repair of AA7075-T651. In order to understand the potential applicability of this solid-state additive repair technique, in-depth microstructural characterization and mechanical testing were carried out. Based on the results of this study, the following conclusions were drawn:

• The AFS-D approach was used for repair of a machined groove without the observation of hot cracking or other deleterious microstructural defects.

• Microstructural characterization revealed that the deposited repair exhibited a refined microstructure with an equiaxed grain morphology.

• The combination of optical and EBSD analysis revealed the variation of the microstructural features of the repaired plate. In particular, this characterization revealed a difference in microstructure in two regions: the direct protrusions stir zone (DPSZ) and the nonprotrusion stir zone (NPSZ).

• The DPSZ exhibited a uniform microstructure of equiaxed grains throughout the first 2 mm of AFS-D material, and no material flow lines were observed in this region.

• The NPSZ was characterized with material flow lines of mixing between the substrate and deposited feedstock. The flow line boundaries exhibited refined defect-free nucleated grains at the interface, which was also exhibited at the bottom of the repair/substrate interface.

• The heat-treatment process for the AFS-D samples was not observed to produce any abnormal grain growth, and the difference in hardness for the as-deposited state was attributed to the loss of the strengthening precipitates.

• The heat treatment of the repaired plate successfully restored the mechanical performance of the material to wrought-like static properties; however, significant scatter in the strain to failure was observed in the tensile results.

• Experimental fatigue results exhibited significant scatter for the heat-treated AFS-D repaired plate. Additionally, the average overall fatigue results of the heat-treated AFS-D repaired plate displayed a decrease in fatigue performance when compared with the wrought control samples.

• Post-mortem analysis of the heat-treated specimens revealed a fish-eye fracture feature, where the fatigue crack nucleation site was located at the center of the specimen toward the repair/substrate interface. It was also revealed that a planar defect of aggregate particles facilitated intergranular fracture of the aluminum matrix.

• EDS chemical analysis elucidated that the aggregate particulates found at the nucleation site are carbon and oxygen rich. The carbon is postulated to come from the dry graphite lubricant that was applied to the feedstock before deposition.

• In general, the MSF model was found to bound the scatter of the experimental fatigue results for both the wrought and AFS-D repair AA7075. However, for the AFS-D repair lower bounds prediction, the carbon impurities likely accelerated fatigue initiation mechanisms that were not fully captured by the MSF model. Future MSF modeling work will need to address fatigue crack initiation mechanisms from carbon impurities.

The authors are grateful for the support of the Alabama Transportation Institute (ATI), the Analytical Research Center (AARC), and the US Department of Defense Strategic Environmental Research and Development Program under contract WP18-C4-1323.

There are no conflicts of interest.