1. Introduction
AA 2xxx and AA 7xxx aluminum alloys are widely used in aircraft structures due to their low density, high strength, and high stiffness [
1]. Specifically, the 2024-T3 alloy is primarily employed in the fuselage for the rivets, skin, wings, skeleton, etc. The 7075-T6 alloy serves as the essential material in wing spars, baffle frames, and wing panels. However, traditional welding methods cause several defects in the 2024-T3 and 7075-T6 alloys [
2]. On the other hand, friction stir welding (FSW) stands out as a typical solid-state joining technology, which is capable of fabricating high-quality welded joints with minimal welding deformation [
3,
4]. Therefore, FSW is being increasingly adopted in industries such as aerospace, shipping, railway, electronics, and others [
5,
6].
During the welding process, changes in local temperature and deformation induce plastic deformation, which occurs in the welded zone (WZ) and thereby leads to the formation of residual stress, which is a self-balancing system that is not dependent on external loads [
3,
7]. The residual stress and the external loads can result in secondary deformations, as well as the redistribution of residual stress, which significantly influences the integrity and safety of welded structures [
8,
9,
10]. Previous studies have shown that the maximum residual stress in the longitudinal direction (LD) is generated in the heat-affected zone (HAZ), while the minimum compressive residual stress originates on the advancing side just beyond the WZ [
11,
12]. Furthermore, another study indicated that in the 2024/7075 dissimilar welded joints, the longitudinal residual stress on the 7075 side was higher than that on the 2024 side [
13]. This is related to the position within the welded joint, as the heat input on the advancing side is relatively higher than that on the retreating side. Additionally, compared to the 2024 alloy, 7075 aluminum exhibits higher mechanical properties, including yield strength and hardness.
Due to the strong local thermal coupling effect in the FSW process, the microstructure of the welded joint becomes diversified and non-uniform [
14,
15]. The WZ, which is composed of fine equiaxed grains experiences the highest thermal cycle and the most intense plastic deformation during the welding process, causing dynamic recrystallization in this region [
16]. In the thermo-mechanically affected zone (TMAZ), plastic deformation induces the formation of several dislocations within the crystal structure. Additionally, the increase in temperature leads to dynamic recovery, thereby resulting in the formation of a large number of sub-grains [
17]. A previous study demonstrated that the HAZ exclusively underwent the thermal cycle that results in a partial coarsening of the grains, which typically leads to a morphology that closely resembles that of the base metal [
18].
Previous studies have also demonstrated that the fatigue crack growth rate in FSWed joints is influenced by the interaction between residual stress distribution and the microstructure [
19,
20]. The high level of compressive residual stress near the welded zone can induce crack closure effects, while tensile stress increases the effective stress intensity factor range [
21,
22]. Fratini et al. [
23,
24] noted that stress relief did not alter the hardness and microstructural properties of FSWed structures, but the closure phenomena resulting from residual stress did affect the growth rates. Ilman et al. [
25] stated that the effect of the microstructure on the fatigue crack growth rate was not significant when compared to residual stress. However, in the study by Anderson-Wedge et al. [
26], the changes in crack growth were attributed to the depletion of strengthening precipitates in thermally affected zones rather than to the residual stresses.
Considering the varying conclusions from previous studies, this paper aims to investigate the coupling effect between residual stress and microstructure on the fatigue crack growth behavior of 2024 and 7075 aluminum alloys. The residual stresses were measured with the X-ray diffraction method, and the microstructure was observed with TEM. The finite element model was built using ANSYS and FRANC3D to calculate the fatigue crack growth rate, and the results were compared with those of the subsequent experiments.
4. Fatigue Crack Growth Experiment
The fatigue crack growth experiments in this study were conducted using an INSTRON 810 testing machine at Northwestern Polytechnical University, Xi’an, China, with a maximum loading force of 26 kN (
Figure 3a). This setup ensured symmetrical load distribution with a load error of less than ±1% and minimal variation in the indicating value. The testing machine was equipped with an accurate counting device to meet the testing requirements. The compact tension (C(T)) specimens were machined according to the ASTM-647 standard [
27]; the size of the specimen is shown in
Figure 3b. It can be seen that the cracks in the C(T) specimens were parallel to the welded joints. To facilitate fixture fixation and crack growth observation, the surfaces of the specimen were polished before the experiment (
Figure 3c). Fatigue tests were performed at R = 0.1 for all the C(T) specimens, with a fatigue load frequency of 20 Hz. Load ranges of 6.21 kN for the 2024 and 7075 homogeneous specimens, 8.04 kN for the 2024/7075 dissimilar specimens, and 3.26 kN for the base metal were applied.
The results of the fatigue crack growth rate experiment of the 2024 homogeneous FSWed specimens with different rotational speeds are presented in
Figure 4. It can be seen that the crack growth rate of the 2024 base metal is higher than that of the welded specimen. Under the same welding speed, the fatigue crack growth rate of the 400 rpm specimen is higher than that the of 200 rpm specimen. As seen in
Figure 4b, the crack path of 200 rpm was deflected towards the loading hole. With further crack propagation, the crack crossed through the WZ into the thermo-mechanically affected zone (TMAZ).
Figure 4c shows that the crack growth path of the 400 rpm specimen remained within the WZ. According to previous studies on the dynamic tensile properties of the WZ, the increase in rational speed reduces the size of the particles in the WZ and makes them more uniform [
28]. This enhances the mechanical properties of the WZ and reduces the fatigue crack growth rate in this region. Therefore, it can be concluded that tensile longitudinal residual stress promotes the crack growth and that it offsets the decrease in fatigue crack growth rate by grain refinement.
Figure 5a displays the fatigue crack growth results of the 7075 homogeneous FSWed specimens with different welding speeds. It is evident that welding speed exerts a significant influence on the crack growth rate. At the initial stage, the crack growth rate of the 7075 base metal was much higher than that of the FSWed specimen. Notably, the fatigue crack growth rate of the specimen with a welding speed of 100 mm/min was higher than that of the specimen with a speed of 150 mm/min. In a previous study, it was demonstrated that the mechanical properties of the 100 mm/min and 150 mm/min specimens were similar [
29]; however, the longitudinal residual stress of the 100 mm/min specimen was higher than that of the 150 mm/min specimen. Consequently, the reduction in longitudinal residual stress contributed to the lower fatigue crack growth rate.
Figure 6a shows the fatigue crack growth rate curve of the 2024/7075 dissimilar FSWed specimen with a welding speed of 150 mm/min and rational speed of 400 rpm. It was compared with the 2024 and 7075 homogeneous FSWed specimens under the same welding parameters. The results indicated that, during the initial stage of crack growth, the fatigue crack growth rate of the 2024/7075 dissimilar specimens was between that of the 2024 and 7075 homogenous specimens. With further crack propagation, the dissimilar crack growth rate gradually approached that of the 2024 homogeneous specimen. This could be attributed to the lower hardness of the 2024 material, which facilitates crack propagation and ultimately allows the crack to extend into the 2024 side.
6. Microstructure Analysis
The fracture morphologies of the 2024 (400 rpm-150 mm/min) specimens in the fatigue propagation zone are presented in
Figure 15. These specimens exhibited fatigue striations. Notably, the crack at the front exhibited an inclination rather than expanding vertically. This phenomenon occurred as the residual stress in the specimen gradually decreased from the top to the bottom, and because the grain size at the bottom of the WZ was smaller than that at the top [
13]. As a result, the rate of fatigue crack propagation at the top was faster than that at the bottom.
In
Figure 16, the TEM diagrams of the 2024 welded joints with different rotational speeds are shown using the Talos F200X at Xi’an University of Architecture and Technology in Xi’an, China. The TEM images of the WZ confirm the presence of needle-shaped S-phase (Al
2CuMg) precipitates [
31]. In the WZ (
Figure 16a,d), most of the dislocations appeared in the form of distortion and entanglement, which resulted from the significant distortion caused by the rotation of the stirring probe. In the TMAZ, the S-phase was also present, with more dislocation entanglement observed around the grain boundaries (
Figure 16b,e). This corresponds to the residual stress measurements shown in
Figure 2. In the HAZ of the 400 rpm-150 mm/min specimen (
Figure 16c), the precipitated S-phase increased due to the influence of the welding thermal cycling, while the extent of dislocation entanglement decreased relatively. In
Figure 16f, fewer dislocation cell substructures and AlCu
3 precipitates (remnants from the aging treatment process) were observed in the base material.
A comparison between the 200 rpm and 400 rpm specimens revealed that the 200 rpm specimen exhibited a higher dislocation density and a greater presence of precipitated phases in the WZ and TMAZ. During the process of fatigue crack propagation, these precipitated phases not only pinned dislocations but also facilitated the generation of additional dislocations, thereby leading to dislocation entanglement. As a result, the dislocation density and sliding resistance increased within the welding joint, which thus resulted in a decreased rate of fatigue crack propagation. This explains why the fatigue crack growth rate of the 200 rpm specimen is lower than that of the 400 rpm from a microstructural perspective, as shown in
Figure 5a. Additionally, when compared to the WZ, the S-phase precipitates in the TMAZ and HAZ were coarser; this is a phenomenon that is well-documented in the literature [
32,
33]. Therefore, the crack predominantly propagated toward the TMAZ, as shown in
Figure 4 and
Figure 5.