E ﬀ ect of Severe Shot Peening on the Very-High Cycle Notch Fatigue of an AW 7075 Alloy

: The severe shot peening process was applied to the notched specimens from an AW 7075 alloy with the aim to improve fatigue endurance in the very-high cycle fatigue region. To reveal the stress state in the notch vicinity, ﬁnite element analysis was performed, simulating the conditions of the used 20 kHz ultrasonic fatigue loading. Modiﬁed surface characteristics by the severe shot peening process were analyzed in terms of residual stress distribution measured by X-ray di ﬀ raction methods and near-surface microstructural observations by scanning electron microscopy. The applied severe shot peening increased the fatigue limit by 11%; however, the positive e ﬀ ect was recorded only for the loading amplitudes corresponding to the fatigue lifetimes in the range 10 7 –10 9 cycles. At higher loading amplitudes, the fatigue properties tended to decrease, most likely due to accelerated fatigue crack initiation on the surface damage features created by the peening process and also by rapid residual stress relaxation. Author Contributions: Experiment design and surface treatment, L.T.; fracture surface analysis, M.J., S.F.; residual stress evaluation, D.K.; stress analysis, J.K.; fatigue testing, F.N.; fatigue data evaluation, O.B. All authors have the


Introduction
High strength aluminum alloys based on the Al-Zn system are one of the most widely spread materials for the high strength application in the aircraft industry, combining high strength-related properties with sufficient toughness and corrosion resistance. The most serious drawback of these alloys is their low fatigue endurance, which has been reported by many authors [1,2]. The fatigue life is further reduced by the presence of notches in complexly shaped components. In the last decades, numerous methods of strengthening of the surface and subsurface layers were introduced to improve fatigue properties, especially in the high cycle fatigue area [3][4][5][6]. One of the most widely used methods is shot peening, where the treated surface is impacted by shots with high kinetic energy, resulting in the deformation of the surface layer. The positive effect of the shot peening on the fatigue properties has been proved in numerous studies [3,[7][8][9][10][11] for various materials. In the shot peening process, several factors are contributing to the final surface properties, which influence the final fatigue life of the treated material [12]. The first and most important factor is considered the residual stresses state. The severe  Table 2. Basic mechanical properties of the experimental material [7]. The microstructure of the experimental material was examined using the scanning electron microscopy (SEM): the electron backscatter diffraction method (EBSD) and electron channeling contrast imaging (ECCI). Specimens for the microstructural observation were mechanically ground and polished using diamond suspensions up to the size of 1 µm. In the final step, specimens were polished using colloidal silica. The partially recrystallized microstructure of the experimental material is shown in Figure 1. It consists of the long, partially recrystallized grains, elongated in the extrusion direction, and small intermetallic phases. The material exhibited significant fiber texture, typical for the extruded high strength aluminum alloys [24]. ECCI images revealed developed sub-grain microstructure within the elongated grains. The presence of intermetallic particles was recorded at the grain boundaries and also inside the grains. These particles were aligned in the bands, parallel with the extrusion direction. Based on the literature data, these phases are most likely Al x CuFe y , MgZn 2 , and Al x Fe y Si types [24,25]. The AW 7075 alloy in T6511 state is mostly strengthened by the coherent η' particles [26][27][28]. As in this study, no additional heat treatments were applied, and no more attention was paid to the study and characterization of the strengthening particles.

UTS (MPa) Elongation (%) Reduction of area (%) Hardness (HV10)
Specimens of the "hourglass" shape ( Figure 2) were used for the fatigue tests. Their dimensions were calculated to fulfill the resonance conditions of the fatigue testing system. More information about designing of the specimens for the ultrasonic fatigue testing can be found in [29]. In the middle of the gauge length of the specimens, a U notch with a 0.6 mm tip radius was machined, with the use of a machining tool with a negative shape. This means that during machining of the notch, the only transversal feed was used.
Metals 2020, 10, 1262 4 of 18 grain microstructure within the elongated grains. The presence of intermetallic particles was recorded at the grain boundaries and also inside the grains. These particles were aligned in the bands, parallel with the extrusion direction. Based on the literature data, these phases are most likely AlxCuFey, MgZn2, and AlxFeySi types [24,25]. The AW 7075 alloy in T6511 state is mostly strengthened by the coherent η' particles [26][27][28]. As in this study, no additional heat treatments were applied, and no more attention was paid to the study and characterization of the strengthening particles. Specimens of the "hourglass" shape ( Figure 2) were used for the fatigue tests. Their dimensions were calculated to fulfill the resonance conditions of the fatigue testing system. More information about designing of the specimens for the ultrasonic fatigue testing can be found in [29]. In the middle of the gauge length of the specimens, a U notch with a 0.6 mm tip radius was machined, with the use of a machining tool with a negative shape. This means that during machining of the notch, the only transversal feed was used.

Surface Treatment
The used surface treatment was severe shot peening with parameters chosen according to works [3,7], where optimization for the best fatigue performance of smooth specimens was performed. The gauge lengths of the specimens were severely shot peened with the following parameters: Almen intensity 9.6 N, coverage 650%. Ceramic beads CEZ 100 were used for the treatment, and they were shot at the incidence angle close to 90 • with respect to the specimen axis. During the severe shot peening process, specimens were rotated at the constant speed of 180 rpm to ensure homogeneous coverage around the whole gauge length.
The effect of the severe shot peening on the surface layer character was analyzed by SEM (Tescan LYRA 3, Tescan, Brno, Czech Republic), and the residual stress state was evaluated by the X-ray diffraction measurements . The Proto iXRD device (PROTO Manufacturing Ltd., Oldcastle, ON, Canada) was used for the measurements, using the Cr K α radiation with an irradiated area of 0.5 mm 2 . The diffraction signal from {222} α planes was collected at 2θ = 156.9 • . The measurements were carried out using the sin 2 ψ method, with nine inclinations between ±39 • . The measurements were carried out in axial (φ = 0 • ) and tangential (φ = 90 • ) directions. To obtain the depth profile of the residual stress distribution, the surface was gradually removed by electrolytic polishing.

Fatigue Life Analysis
The fatigue testing was carried out using an ultrasonic fatigue testing system (Lasur Sarl, France) operating at the loading frequency close to 20 kHz. The specimens were subjected to the symmetrical tension-compression loading (R = −1). The setup of the ultrasonic fatigue testing device is described in more detail in works [3,7,29,30]. This loading type was considered as one of the most damaging because, during the compression part of the cycle, the superposition of the compression residual stresses with the compression part of the loading cycle can overrun local compression yield strength, resulting in the micro-plastic deformation, relieving the residual stresses. The specimens were loaded at stress levels corresponding to the fatigue lifetimes in the range from 10 6 to 10 9 cycles. After reaching 10 9 cycles without fracture, the tests were terminated. During the tests, the specimens were cooled by the stream of dry air.
As the fatigue tests carried out on the ultrasonic fatigue testing device were performed in the strain control regime, the actual stress level in the gauge length had to be calculated. For the calculation of the stress level in the conventional (smooth) specimens used for ultrasonic fatigue testing, an analytical solution by Bathias [31] was already developed and is well accepted among other researchers. In the case of the notched specimens, the actual distribution of the stress in the notch vicinity had to be treated using finite element (FE) simulation. The calculations were performed in the ANSYS system (ANSYS, Inc., Canonsburg, PA, USA). The 3D model of the specimen was meshed by the 20-nodes elements SOLID 186. In the regions of the stress concentration, the mesh was refined in order to describe the stress distribution more precisely (see Figure 3). The basic input parameters to the calculations were the dynamic modulus of elasticity E d = 71.7 GPa, Poisson's ratio ν = 0.33, and mass density ρ = 2800 kg/m 3 . The 3D model was prepared for modal and harmonic analysis. Modal analysis was used in order to design the specimen so that its intrinsic frequency of longitudinal vibrations is as close as possible to 20 kHz. The 3D model was important, especially in this step (modal analysis), because it allowed revealing other spatial shapes (not only the longitudinal one) and frequencies of intrinsic vibrations. It was necessary to ensure that other intrinsic vibration modes had their frequencies outside the interval 19.5-20.5 kHz. In the harmonic analysis, cyclic displacement with the amplitude of 20 µm and with the intrinsic frequency 19,995 Hz was applied at one of the ends of the specimen. Note that the actual intrinsic frequency was calculated in the modal analysis. The full method of the harmonic analysis was chosen.
this step (modal analysis), because it allowed revealing other spatial shapes (not only the longitudinal one) and frequencies of intrinsic vibrations. It was necessary to ensure that other intrinsic vibration modes had their frequencies outside the interval 19.5-20.5 kHz. In the harmonic analysis, cyclic displacement with the amplitude of 20 μm and with the intrinsic frequency 19,995 Hz was applied at one of the ends of the specimen. Note that the actual intrinsic frequency was calculated in the modal analysis. The full method of the harmonic analysis was chosen.

Stress Analysis in the Notch
Harmonic analysis at the intrinsic frequency fn = 19,995 Hz (calculated within the modal analysis) was performed to describe stress concentration in the notch vicinity. Firstly, the basic approximation of the stress concentration at the notch was performed by the equivalent ellipse concept introduced by Murakami [32]. The equivalent ellipse concept estimates the stress concentration factor Kt in the form of the following equation:

Stress Analysis in the Notch
Harmonic analysis at the intrinsic frequency f n = 19,995 Hz (calculated within the modal analysis) was performed to describe stress concentration in the notch vicinity. Firstly, the basic approximation of the stress concentration at the notch was performed by the equivalent ellipse concept introduced by Murakami [32]. The equivalent ellipse concept estimates the stress concentration factor K t in the form of the following equation: where t is the depth of the notch, and r is the radius in the notch root. For our specimen, t = 0.5 mm and r = 0.6 mm. The stress concentration factor calculated according to the equivalent ellipse concept is K t = 3.2.
The FE analysis enables us to describe the stress concentration more precisely and in detail. The results ascertained for the harmonic longitudinal tension-compression vibrations with the amplitude of vibrations A = 20 µm of the σ xx stress component (direction of the specimen's axis) are shown in Figure 4. In Figure 4a, the stress concentration in the gauge length of the specimen is apparent. Detail of the stress concentration in the notch is shown in the section in Figure 4b-the highest stress is in the notch root, as expected. The stress distribution across the diameter in the middle of the specimen is shown in Figure 4c. The figure shows that the highest σ xx stress in the notch root is 457.32 MPa (for the displacement amplitude of 20 µm). The stress factor S f expresses the level of stress σ xx in the central part of the specimen caused by the displacement amplitude of 1 µm. This means that the maximum stress factor is S f (max) = 22.866 MPa/µm. The average stress across the diameter is calculated according to equation (2): The result of the calculation is σ xx (avg) = 186.308 MPa (again for the displacement amplitude of 20 µm). Consequently, the average stress factor is S f (avg) = 9.315 MPa/µm. Finally, the stress concentration factor following from FE analysis is: K t = σ xx (max)/σ xx (avg) = 2.455. and from FE analysis (Kt = 2.455), it is obvious that the analytical solution is more conservative and predicts higher stresses in the notch root. For determination of the stress level in the tested specimens, the stress factor Sf(avg) = 9.315 MPa/μm was used. The Sf(avg) multiplied by the amplitude of loading displacement corresponds to the nominal stress in the cross-section, where failure occurs. In this study, the nominal stress was used, so it is comparable with other works where smooth specimens were tested.

Effect of Shot Peening on the Surface State
The surface state of the fatigue specimens was examined in terms of three critical factors related to the shot peening process, which are affecting the resulting fatigue properties-surface state, strain hardening, and the residual stresses. The micrographs comparing the surface state of the notched specimens after machining and shot peening are shown in Figure 5. The surface of the not peened (NP) specimens has a smooth character, but the detailed view of the notch root ( Figure 5b) reveals the presence of the traces perpendicular to the loading axis, caused by the machining process. Due to their orientation, it can be assumed that they act as preferential crack initiation locations. The surface of the specimens after the severe shot peening (SSP) is shown in Figure 5c,d. After the severe shot peening process, uniform deformation of the surface is achieved, as can be seen from the detailed view of the notch root in Figure 5d. When comparing the factors of stress concentration following from analytical solution (K t = 3.2) and from FE analysis (K t = 2.455), it is obvious that the analytical solution is more conservative and predicts higher stresses in the notch root. For determination of the stress level in the tested specimens, the stress factor S f (avg) = 9.315 MPa/µm was used. The S f (avg) multiplied by the amplitude of loading displacement corresponds to the nominal stress in the cross-section, where failure occurs. In this study, the nominal stress was used, so it is comparable with other works where smooth specimens were tested.

Effect of Shot Peening on the Surface State
The surface state of the fatigue specimens was examined in terms of three critical factors related to the shot peening process, which are affecting the resulting fatigue properties-surface state, strain hardening, and the residual stresses. The micrographs comparing the surface state of the notched specimens after machining and shot peening are shown in Figure 5. The surface of the not peened (NP) specimens has a smooth character, but the detailed view of the notch root ( Figure 5b) reveals the presence of the traces perpendicular to the loading axis, caused by the machining process. Due to their orientation, it can be assumed that they act as preferential crack initiation locations. The surface of the specimens after the severe shot peening (SSP) is shown in Figure 5c,d. After the severe shot peening process, uniform deformation of the surface is achieved, as can be seen from the detailed view of the notch root in Figure 5d. The degree of the strain hardening of the surface layer was examined using SEM and the ECCI technique. Figure 6 shows a comparison of the microstructure in the vicinity of the notch root on the NP and SSP specimens. In the case of the NP specimen, a shallow deformed layer with the maximal thickness of several μm can be seen. This layer is formed due to the machining process, and the level of the deformation is not severe. NP specimen exhibits quite a smooth surface, with occasional occurrence of shallow discontinuities. Significant changes are observed in the specimens after the SSP process (Figure 6c,d). Based on the level of the deformation and on the approximate grain size, the strengthened layer can be divided into two sub-layers. In the near-surface region, a severely deformed layer up to the depth of approximately 20-30 μm is observed, within which the grains are refined below 1 μm. With increasing depth, the degree of grain refinement is decreasing, and in the range of 30-45 μm under the surface, the area with modest grain refinement is observed. In the depths of 50 μm and more, the microstructure is visually unchanged, but it can be assumed that at least a small increase of the dislocation density in this area is created. The surface of the SSP specimens contains shallow discontinuities in higher occurrence than in the case of the NP specimens. A large number of small cracks in the heavily deformed layer were also recorded. These are most likely formed on the grain boundaries of the former elongated grains. Their presence was recorded also in the unbroken (run-out) specimens . The degree of the strain hardening of the surface layer was examined using SEM and the ECCI technique. Figure 6 shows a comparison of the microstructure in the vicinity of the notch root on the NP and SSP specimens. In the case of the NP specimen, a shallow deformed layer with the maximal thickness of several µm can be seen. This layer is formed due to the machining process, and the level of the deformation is not severe. NP specimen exhibits quite a smooth surface, with occasional occurrence of shallow discontinuities. Significant changes are observed in the specimens after the SSP process (Figure 6c,d). Based on the level of the deformation and on the approximate grain size, the strengthened layer can be divided into two sub-layers. In the near-surface region, a severely deformed layer up to the depth of approximately 20-30 µm is observed, within which the grains are refined below 1 µm. With increasing depth, the degree of grain refinement is decreasing, and in the range of 30-45 µm under the surface, the area with modest grain refinement is observed. In the depths of 50 µm and more, the microstructure is visually unchanged, but it can be assumed that at least a small increase of the dislocation density in this area is created. The surface of the SSP specimens contains shallow discontinuities in higher occurrence than in the case of the NP specimens. A large number of small cracks in the heavily deformed layer were also recorded. These are most likely formed on the grain boundaries of the former elongated grains. Their presence was recorded also in the unbroken (run-out) specimens . The distribution of the residual stresses in the not peened and severely shot peened specimens is shown in Figure 7. In the case of the NP specimens, the machining process creates a shallow compressive residual stress field, with a peak at 0.02 mm under the surface and maximal value stated between −100 and −150 MPa. With increasing depth, these residual stresses quickly disappear. After severe shot peening, the residual stresses are introduced deeply to the material. The peak of the compressive residual stresses is achieved in depth between 0.03 and 0.04 mm, with values between −300 and −350 MPa, followed by the plateau up to the depth of 0.08 mm, after which a steep decrease of the residual stress level occurs. The residual stress levels in the axial direction (ϕ = 0°) are approximately 50 MPa higher than those in the circumferential direction (ϕ = 90°). The distribution of the residual stresses in the not peened and severely shot peened specimens is shown in Figure 7. In the case of the NP specimens, the machining process creates a shallow compressive residual stress field, with a peak at 0.02 mm under the surface and maximal value stated between −100 and −150 MPa. With increasing depth, these residual stresses quickly disappear. After severe shot peening, the residual stresses are introduced deeply to the material. The peak of the compressive residual stresses is achieved in depth between 0.03 and 0.04 mm, with values between −300 and −350 MPa, followed by the plateau up to the depth of 0.08 mm, after which a steep decrease of the residual stress level occurs. The residual stress levels in the axial direction (φ = 0 • ) are approximately 50 MPa higher than those in the circumferential direction (φ = 90 • ). The distribution of the residual stresses in the not peened and severely shot peened specimens is shown in Figure 7. In the case of the NP specimens, the machining process creates a shallow compressive residual stress field, with a peak at 0.02 mm under the surface and maximal value stated between −100 and −150 MPa. With increasing depth, these residual stresses quickly disappear. After severe shot peening, the residual stresses are introduced deeply to the material. The peak of the compressive residual stresses is achieved in depth between 0.03 and 0.04 mm, with values between −300 and −350 MPa, followed by the plateau up to the depth of 0.08 mm, after which a steep decrease of the residual stress level occurs. The residual stress levels in the axial direction (ϕ = 0°) are approximately 50 MPa higher than those in the circumferential direction (ϕ = 90°).

Fatigue Test Results
Fatigue test results in the form of S-N curves are shown in Figure 8, and the regression curves were plotted according to the Basquin function with fitting coefficients shown in Table 3. From the results, the low inclination character of the fatigue curves is clearly seen, as in the case of the not peened specimens, the difference in the fatigue strengths for 10 6 and 10 9 of loading cycles is less than 40 MPa. This number is even lower when considered severely shot peened specimens (∆σ a less than 20 MPa). The fatigue limit (for 10 9 cycles) for NP specimens was recorded to be 93 MPa. The applied shot peening process is found to be beneficial in terms of the fatigue limit, as, in the case of the SSP specimens, the fatigue limit was recorded to be 107 MPa (an increase of 15%). The regression curves of the NP and SSP specimens are crossing at approximately 10 7 cycles, revealing that the positive effect of the shot peening process on the fatigue properties of notched specimens is limited only for the very-high cycle region, while for the lower lifetimes (and higher loading amplitudes), the severe shot peening reduces the fatigue strength.
Metals 2020, 10, x FOR PEER REVIEW 10 of 18 Figure 7. The residual stress profile measured in the notch of not peened (NP) and severe shot peened (SSP) specimens.

Fatigue Test Results
Fatigue test results in the form of S-N curves are shown in Figure 8, and the regression curves were plotted according to the Basquin function with fitting coefficients shown in Table 3. From the results, the low inclination character of the fatigue curves is clearly seen, as in the case of the not peened specimens, the difference in the fatigue strengths for 10 6 and 10 9 of loading cycles is less than 40 MPa. This number is even lower when considered severely shot peened specimens (Δσa less than 20 MPa). The fatigue limit (for 10 9 cycles) for NP specimens was recorded to be 93 MPa. The applied shot peening process is found to be beneficial in terms of the fatigue limit, as, in the case of the SSP specimens, the fatigue limit was recorded to be 107 MPa (an increase of 15%). The regression curves of the NP and SSP specimens are crossing at approximately 10 7 cycles, revealing that the positive effect of the shot peening process on the fatigue properties of notched specimens is limited only for the very-high cycle region, while for the lower lifetimes (and higher loading amplitudes), the severe shot peening reduces the fatigue strength.  The fracture surfaces of the specimens failed during the fatigue tests were subjected to the fractography analysis. In the NP series, the fracture surfaces of all specimens have a similar character The fracture surfaces of the specimens failed during the fatigue tests were subjected to the fractography analysis. In the NP series, the fracture surfaces of all specimens have a similar character (Figures 9 and 10). Multiple initiation locations (indicated by white arrows in Figures 9a and 10a) are observed on all tested specimens, suggesting strong notch sensitivity of the tested aluminum alloy. The cracks initiate on imperfections on the notch surface, created by the machining process, and in the first stage, they propagate in the perpendicular direction with respect to the loading axis (Figures 9b and 10c). After this stage, the cracks propagate through the transition region, reaching the stable stage of the fatigue crack propagation (Figures 9c and 10d). The fracture surfaces, especially in the regions of the first and second stage of the fatigue crack propagation, exhibit a significant degree of surface wear and oxidation, which prevents the clear distinguishing of individual regions. The area of the stable crack propagation (Figures 9c and 10d) exhibits crystallographic features (facets), which is typical for the wrought precipitation strengthened aluminum alloys. The area of the final fracture shows a transcrystalline character with the dimple morphology (Figure 9d).  (Figures 9 and 10). Multiple initiation locations (indicated by white arrows in Figure 9a and Figure  10a) are observed on all tested specimens, suggesting strong notch sensitivity of the tested aluminum alloy. The cracks initiate on imperfections on the notch surface, created by the machining process, and in the first stage, they propagate in the perpendicular direction with respect to the loading axis ( Figure 9b and Figure 10c). After this stage, the cracks propagate through the transition region, reaching the stable stage of the fatigue crack propagation (Figure 9c and Figure 10d). The fracture surfaces, especially in the regions of the first and second stage of the fatigue crack propagation, exhibit a significant degree of surface wear and oxidation, which prevents the clear distinguishing of individual regions. The area of the stable crack propagation (Figure 9c and Figure 10d) exhibits crystallographic features (facets), which is typical for the wrought precipitation strengthened aluminum alloys. The area of the final fracture shows a transcrystalline character with the dimple morphology ( Figure 9d).  The fracture surfaces of the SSP specimens failed during cyclic loading are shown in Figures 11  and 12. The multiple fatigue crack initiation was recorded on all fractured specimens, including those on the lowest level of the loading stress amplitude. All of the fatigue cracks start to propagate from the surface/near-surface region. Various discontinuities in the strengthened layer, introduced during the severe shot peening process, act as initiation locations for fatigue cracks. The cracks propagate perpendicularly to the loading axis through the strengthened layer. After overcoming of the strengthened layer (1st stage of fatigue crack propagation-FCP, according to a Figure 11b), a short transition region of the fatigue crack propagation was recorded, where the crack propagation starts to deviate from the initial crack plane (2nd stage of fatigue crack propagation according to Figure  11b). This area is characterized by a transcrystalline character and a very rough surface (Figure 11b and Figure 12b). With further crack propagation, the propagation mechanism has changed again, and the cracks continue to propagate by transcrystalline mechanism with a crystallographic manner, producing the characteristic appearance of the fracture surface, consisting of a large number of facets (Figure 11d and Figure 12d). The same appearance of the fracture surfaces was recorded for all severely shot peened specimens, regardless of the loading amplitude, but the number of the initiation sites is reduced with decreasing loading amplitude. The fracture surfaces of the SSP specimens failed during cyclic loading are shown in Figures 11 and 12. The multiple fatigue crack initiation was recorded on all fractured specimens, including those on the lowest level of the loading stress amplitude. All of the fatigue cracks start to propagate from the surface/near-surface region. Various discontinuities in the strengthened layer, introduced during the severe shot peening process, act as initiation locations for fatigue cracks. The cracks propagate perpendicularly to the loading axis through the strengthened layer. After overcoming of the strengthened layer (1st stage of fatigue crack propagation-FCP, according to a Figure 11b), a short transition region of the fatigue crack propagation was recorded, where the crack propagation starts to deviate from the initial crack plane (2nd stage of fatigue crack propagation according to Figure 11b). This area is characterized by a transcrystalline character and a very rough surface (Figures 11b and 12b). With further crack propagation, the propagation mechanism has changed again, and the cracks continue to propagate by transcrystalline mechanism with a crystallographic manner, producing the characteristic appearance of the fracture surface, consisting of a large number of facets (Figures 11d and 12d). The same appearance of the fracture surfaces was recorded for all severely shot peened specimens, regardless of the loading amplitude, but the number of the initiation sites is reduced with decreasing loading amplitude.

Discussion
Based on the recorded fatigue behavior of the NP and SSP specimens, the beneficial effect of the severe shot peening treatment on the notched fatigue properties in the very-high cycle fatigue region is clearly visible. The fatigue limit at 10 9 cycles is increased from 93 MPa (for NP) to 107 MPa (SSP), which represents an increase of 15%. The analysis of regression curves shows that the beneficial effect of the SSP is recorded for the region beyond 10 7 loading cycles. On the other hand, for the shorter lifetimes and higher loading amplitudes, the SSP causes worsening of the fatigue properties. According to the regression curve coefficients, the fatigue strength for the 10 6 cycles is, in the case of the SSP specimens, 7% lower than for the NP specimens. Similar behavior is also observed in other works, in which ultrasonic fatigue testing of NP and SP specimens has been performed [3,33]. In these works, the phenomenon is attributed to a higher residual stress relaxation rate due to higher applied loading amplitudes at a lower number of cycles of interest. Since the elastic residual stresses are mostly resulting from increased dislocation density, the application of dynamic loading causes further slip of the bonded dislocations. In this case, the fatigue properties improvement by the residual stress effect is lost in the first few thousand loading cycles, and only the effect of the grain refined layer takes place [34]. However, the possible positive effect of the grain refined layer is strongly opposed by the negative effect of the increased surface roughness, which causes acceleration of the fatigue crack initiation. Under higher loading, also the fatigue crack propagation increments

Discussion
Based on the recorded fatigue behavior of the NP and SSP specimens, the beneficial effect of the severe shot peening treatment on the notched fatigue properties in the very-high cycle fatigue region is clearly visible. The fatigue limit at 10 9 cycles is increased from 93 MPa (for NP) to 107 MPa (SSP), which represents an increase of 15%. The analysis of regression curves shows that the beneficial effect of the SSP is recorded for the region beyond 10 7 loading cycles. On the other hand, for the shorter lifetimes and higher loading amplitudes, the SSP causes worsening of the fatigue properties. According to the regression curve coefficients, the fatigue strength for the 10 6 cycles is, in the case of the SSP specimens, 7% lower than for the NP specimens. Similar behavior is also observed in other works, in which ultrasonic fatigue testing of NP and SP specimens has been performed [3,33]. In these works, the phenomenon is attributed to a higher residual stress relaxation rate due to higher applied loading amplitudes at a lower number of cycles of interest. Since the elastic residual stresses are mostly resulting from increased dislocation density, the application of dynamic loading causes further slip of the bonded dislocations. In this case, the fatigue properties improvement by the residual stress effect is lost in the first few thousand loading cycles, and only the effect of the grain refined layer takes place [34]. However, the possible positive effect of the grain refined layer is strongly opposed by the negative effect of the increased surface roughness, which causes acceleration of the fatigue crack initiation. Under higher loading, also the fatigue crack propagation increments are higher, so the crack relatively quickly grows through the grain refined layer, and its effect becomes negligible. The fatigue curves of both states exhibit low steep character. Benedetti et al. [12] tested fatigue properties of the same alloy after the shot peening process, applied on the specimens with various notches. They reported a similar low steep character of the fatigue curves. On the other hand, they recorded "knee" behavior, with run-out specimens after approximately 10 6 cycles for the specimens with a notch with similar and higher stress concentration factor. After the application of the shot peening process, they observed a change in the behavior, and only specimens with the sharpest notch exhibited such behavior. It should be mentioned that completely different specimen shapes and loading modes (four-point bending with R = 0.05) and also different shot peening parameters were used, resulting in the different residual stress profiles and surface texture. This can explain the differences in the observed behavior from the results presented in this study.
A comparison of the S-N curves of the NP and SSP specimens obtained on the smooth [3] and notched specimens using the same experimental conditions is shown in Figure 13. The results show a similar trend for the NP series, recorded for both notched and smooth specimens. The obvious shift of the S-N curves to lower stress amplitudes due to stress concentration is observed for the notched specimens. A slightly different trend is recorded when the SSP series are compared. In the case of the smooth specimens, the crossing of the Basquin's fits of the S-N curves is expected at approximately 10 5 cycles. On the contrary, for the notched specimens, the crossing of the fits of S-N curves is recorded at approximately 10 7 cycles. This can be attributed to the strong notch sensitivity of the AW 7075 alloy due to its high strength in the artificially aged condition.
Metals 2020, 10, x FOR PEER REVIEW 15 of 18 are higher, so the crack relatively quickly grows through the grain refined layer, and its effect becomes negligible. The fatigue curves of both states exhibit low steep character. Benedetti et al. [12] tested fatigue properties of the same alloy after the shot peening process, applied on the specimens with various notches. They reported a similar low steep character of the fatigue curves. On the other hand, they recorded "knee" behavior, with run-out specimens after approximately 10 6 cycles for the specimens with a notch with similar and higher stress concentration factor. After the application of the shot peening process, they observed a change in the behavior, and only specimens with the sharpest notch exhibited such behavior. It should be mentioned that completely different specimen shapes and loading modes (four-point bending with R = 0.05) and also different shot peening parameters were used, resulting in the different residual stress profiles and surface texture. This can explain the differences in the observed behavior from the results presented in this study. A comparison of the S-N curves of the NP and SSP specimens obtained on the smooth [3] and notched specimens using the same experimental conditions is shown in Figure 13. The results show a similar trend for the NP series, recorded for both notched and smooth specimens. The obvious shift of the S-N curves to lower stress amplitudes due to stress concentration is observed for the notched specimens. A slightly different trend is recorded when the SSP series are compared. In the case of the smooth specimens, the crossing of the Basquin's fits of the S-N curves is expected at approximately 10 5 cycles. On the contrary, for the notched specimens, the crossing of the fits of S-N curves is recorded at approximately 10 7 cycles. This can be attributed to the strong notch sensitivity of the AW 7075 alloy due to its high strength in the artificially aged condition. According to [1], the surface fatigue crack initiation is usually observed for the not peened specimens from high strength Al alloys in the whole range of the tested fatigue lifetime. In the case of the shot peened specimens, the surface initiation is observed in the medium lifetimes, while for the higher numbers of cycles, the sub-surface fatigue crack initiation is observed. In the presented study, the surface crack initiation dominates for both tested states and in the whole range of the evaluated fatigue lifetime. In NP specimens, the crack initiates from the small local imperfection in the notch root area, caused by the machining process. Multiple fatigue crack initiation implies strong notch sensitivity of an AW 7075 alloy [35]. In the SSP specimens, the crack always initiates from the surface or in the near-surface region (the initiation can occur on the subsurface discontinuities; however, it is impossible to define the exact initiation location). The transversal cross-sections of the run-out specimen in the notch area ( Figure 14) show numerous discontinuities in the strain hardened layer. During cyclic loading, these discontinuities can serve as the initiation points in the final fatigue failure.
Parameters determining the fatigue lifetime of the severe shot peened components are loading amplitude, grain size, depth, and amplitude of the residual stress field. In the case of the lower applied loading amplitudes, the compressive residual stresses slow down fatigue crack initiation and According to [1], the surface fatigue crack initiation is usually observed for the not peened specimens from high strength Al alloys in the whole range of the tested fatigue lifetime. In the case of the shot peened specimens, the surface initiation is observed in the medium lifetimes, while for the higher numbers of cycles, the sub-surface fatigue crack initiation is observed. In the presented study, the surface crack initiation dominates for both tested states and in the whole range of the evaluated fatigue lifetime. In NP specimens, the crack initiates from the small local imperfection in the notch root area, caused by the machining process. Multiple fatigue crack initiation implies strong notch sensitivity of an AW 7075 alloy [35]. In the SSP specimens, the crack always initiates from the surface or in the near-surface region (the initiation can occur on the subsurface discontinuities; however, it is impossible to define the exact initiation location). The transversal cross-sections of the run-out specimen in the notch area ( Figure 14) show numerous discontinuities in the strain hardened layer. During cyclic loading, these discontinuities can serve as the initiation points in the final fatigue failure.
(below the peak of compressive residual stresses)-subsurface fatigue crack initiation. The second type of crack initiation occurs when the continuous cyclic loading causes relaxation of the residual stresses to a level when they no longer restrain the initiation or propagation of a crack from preexisting discontinuities. The occurrence of this mechanism is supported by the stress concentration at the surface caused by the notch, see Figure 4d. As no internal fatigue crack initiation is observed, the residual stresses and their possible relaxation seem to be the critical factors affecting the resulting fatigue properties. This statement also meets the conclusion of other authors [36], who have reported that the induced compressive residual stresses on the surface have only a small effect on the crack nucleation but significantly slow down the crack propagation in the first stages.

Conclusions
It was shown that the application of the severe shot peening increased fatigue life in the case of the notched specimens from AW 7075 aluminum alloy.
The positive effect of severe shot peening was recorded for the stress amplitudes corresponding to the fatigue lifetimes in the range 10 7 -10 9 of loading cycles. The severe shot peening process introduced the compressive residual stress field, which, together with the grain refinement, successfully increased the number of cycles necessary for fatigue crack initiation and also slowed down the propagation of the short cracks.
In the lower ranges of fatigue lifetimes, the application of the severe shot peening resulted in a decrease in fatigue strength. When the higher loading amplitudes were applied, the cyclic loading most likely resulted in early relaxation of the compressive residual stresses, the effect of axial stress concentration was stronger, and the positive effects of severe shot peening were outbalanced by the high roughness and creation of the subsurface discontinuities.
Surface roughness features created by the impact of the shot peening media caused further stress concentration points, which again accelerated fatigue crack initiation in the region of higher loading stresses.
Considering the experimental results, it can be stated that the severe shot peening process can be successfully applied for the increase of fatigue life in case of complexly shaped parts manufactured from the strong notch sensitive AW 7075 aluminum alloy and operated in the very-high cycle region.  Parameters determining the fatigue lifetime of the severe shot peened components are loading amplitude, grain size, depth, and amplitude of the residual stress field. In the case of the lower applied loading amplitudes, the compressive residual stresses slow down fatigue crack initiation and short crack propagation from the existing discontinuities in the hardened layer. When the strengthened surface layer is created, the fatigue crack initiation can occur by two basic mechanisms. First, the crack has to be nucleated in the specimen interior under the strengthened surface layer (below the peak of compressive residual stresses)-subsurface fatigue crack initiation. The second type of crack initiation occurs when the continuous cyclic loading causes relaxation of the residual stresses to a level when they no longer restrain the initiation or propagation of a crack from preexisting discontinuities. The occurrence of this mechanism is supported by the stress concentration at the surface caused by the notch, see Figure 4d. As no internal fatigue crack initiation is observed, the residual stresses and their possible relaxation seem to be the critical factors affecting the resulting fatigue properties. This statement also meets the conclusion of other authors [36], who have reported that the induced compressive residual stresses on the surface have only a small effect on the crack nucleation but significantly slow down the crack propagation in the first stages.

Conclusions
It was shown that the application of the severe shot peening increased fatigue life in the case of the notched specimens from AW 7075 aluminum alloy.
The positive effect of severe shot peening was recorded for the stress amplitudes corresponding to the fatigue lifetimes in the range 10 7 -10 9 of loading cycles. The severe shot peening process introduced the compressive residual stress field, which, together with the grain refinement, successfully increased the number of cycles necessary for fatigue crack initiation and also slowed down the propagation of the short cracks.
In the lower ranges of fatigue lifetimes, the application of the severe shot peening resulted in a decrease in fatigue strength. When the higher loading amplitudes were applied, the cyclic loading most likely resulted in early relaxation of the compressive residual stresses, the effect of axial stress concentration was stronger, and the positive effects of severe shot peening were outbalanced by the high roughness and creation of the subsurface discontinuities.
Surface roughness features created by the impact of the shot peening media caused further stress concentration points, which again accelerated fatigue crack initiation in the region of higher loading stresses.
Considering the experimental results, it can be stated that the severe shot peening process can be successfully applied for the increase of fatigue life in case of complexly shaped parts manufactured from the strong notch sensitive AW 7075 aluminum alloy and operated in the very-high cycle region.

Conflicts of Interest:
The authors declare that there is no conflict of interest.