*Article* **Nugget Formation and Mechanical Behaviour of Friction Stir Welds of Three Dissimilar Aluminum Alloys**

#### **Neves Manuel 1,2,\*, Ivan Galvão 1,3, Rui M. Leal 1,4, José D. Costa <sup>1</sup> and Altino Loureiro <sup>1</sup>**


Received: 14 May 2020; Accepted: 9 June 2020; Published: 11 June 2020

**Abstract:** The aim of this research was to investigate the influence of the properties of the base materials and welding speed on the morphology and mechanical behavior of the friction stir welds of three dissimilar aluminum alloys in a T-joint configuration. The base materials were the AA2017-T4, AA5083-H111, and AA6082-T6 alloys in 3 mm-thick sheets. The AA6082-T6 alloy was the stringer, and the other alloys were located either on the advancing or retreating sides of the skin. All the T-joint welds were produced with a constant tool rotation speed but with different welding speeds. The microstructures of the welds were analyzed using optical microscopy, scanning electron microscopy with energy dispersive spectroscopy, and the electron backscatter diffraction technique. The mechanical properties were assessed according to micro-hardness, tensile, and fatigue testing. Good quality welds of the three dissimilar aluminum alloys could be achieved with friction stir welding, but a high ratio between the tool's rotational and traverse speeds was required. The welding speed influenced the weld morphology and fatigue strength. The positioning of the skin materials influenced the nugget morphology and the mechanical behavior of the joints. The joints in which the AA2017 alloy was positioned on the advancing side presented the best tensile properties and fatigue strength.

**Keywords:** friction stir welding; three dissimilar aluminum alloys; welding speed; T-joints; microstructure; mechanical properties

#### **1. Introduction**

Aluminum alloys of the 5xxx, 6xxx, and 2xxx series are widely used in various industrial sectors, such as shipbuilding, aerospace, building structures, bridge structures, and even military vehicles, due to their lightness, mechanical strength, and resistance to corrosion [1,2]. T-joints are currently used in these industries to increase the stiffness of thin plates.

Friction stir welding (FSW) is a solid-state welding process that prevents porosity and cracking, but the formation of defects is influenced by the flow of materials around a tool. The flow of material in this zone depends on the tool's geometry, the tool's rotation speed and welding speed, the tool's axial force or displacement, and even the tool's tilt angle, whether in similar or dissimilar material welds [3,4]. Fratini et al. [5], when comparing similar welds for the AA6082-T6 and AA2024-T4 alloys

in a T-joint configuration, concluded that the temperature fields and strain rate influence the flow of a material in the stir zone and, hence, the integrity of welds.

In welds between dissimilar materials of different families, the properties of the base materials have to be considered because of the formation of intermetallic compounds that have very different physical and mechanical properties from the base materials and influence the flow of materials in the stir zone [6,7]. However, even for dissimilar welds of materials within the same family, the influence of the properties of the materials on the formation of the weld can be significant. Silva et al. [8] stated that in the dissimilar butt welds of AA7075-T6 to AA2024-T3, the mixing of the two materials in the stir zone was greatly influenced by the geometry and rotational speed of the tool, only obtaining satisfactory mixing at high rotational speeds.

Dinaharan et al. [9] stated that the material that is on the advancing side occupies most of the stir zone and influences the strength of butt weld in aluminum 6061 in rolled and cast plates. Barbini et al. [10] showed that, for butt welds between AA2024-T3 and AA7050-T7651, a better material flow in the stir zone is obtained when the AA2024-T3 is located on the advancing side. Better material flow in the stir zone was also observed for butt welds between AA2024-T6 and AA6061-T6, when the latter material was placed on the advancing side [11]. The results mentioned above suggest that the flow of materials in the stir zone depends on the mechanical properties of the materials on the advancing and retreating sides. Manuel et al. [12] also stated that for T-joints between AA6082 and AA5083, material flow and defect formation are greatly influenced by the joint type, tool geometry, and process parameters, as well as the materials' properties. The investigations found in the literature about three dissimilar aluminum alloys only concern lap joints [13,14].

The FSW of three dissimilar aluminum alloys has scientific and industrial interest, since there is a need for new combinations of materials and there is no clear understanding of how material properties influence weld formation. The aim of this study was to analyze the influence of the properties of the base materials and the variation of the welding speed on the morphology and mechanical properties of welds of three dissimilar aluminum alloys arranged in a T-joint configuration.

#### **2. Materials and Methods**

The experimental tests were performed on 3 mm-thick sheets of AA2017-T4, AA5083-H111, and AA6082-T6 aluminum alloys. Alloys that require plastic deformation during construction are used without much hardening, as in the case the alloys 5083-H111 and 2017-T4, while reinforcements are made of hardened alloys, as is the case for 6082-T6. The chemical composition and mechanical properties of these alloys are listed in Tables 1 and 2, respectively.


**Table 1.** Chemical composition of the base materials (wt.%).

**Table 2.** Mechanical properties of the tested aluminum alloys.


A tool of H13 quenched and tempered steel, composed of a progressive (cylindrical and conical) threaded pin of 5.2 mm in length and a shoulder of 18 mm in diameter with a concavity of 5◦

(see Figure 1a) was used. Previous tests have shown that this tool geometry has obtained defect-free welds in dissimilar aluminum alloys [12].

**Figure 1.** (**a**) Geometry of the tool; (**b**) welding setup and starting point to measure tool penetration.

The setup used to perform the welds is shown in Figure 1b. The stringer protruded 1.4 mm in order to provide enough material to fill the empty volumes between plates and dies in the fillets. The AA6082 alloy was the stringer for all the weld series. The AA5083 and AA2017 alloys were positioned on the advancing and retreating sides of the skin, respectively, in some series—they were designated as 562 and the opposite in another series, designated as 265. The dimensions of the stringer and skin plates were 330 × 37.4 × 3 and 330 × 80 × 3 mm, respectively. The oxides were removed from the interfaces by sanding, and the plates were cleaned with alcohol just before the welding.

The welds were performed in position control using a Cincinnati Milacron 207 Mk milling machine, and the tool's rotational speed (w-500 rpm), plunge depth (7.1 mm), and tilt angle (3◦) were maintained constant for all series. The tool's welding speed (v) was changed according to Table 3. The weld series designation consisted of the designation of the sequence of the materials, followed by the welding speed. The parameters were chosen based on previous experience. The (*w*/*v*) ratio is also indicated in Table 3 due to its relationship with the heat input, which increases with the ratio [15].


**Table 3.** Welding parameters used to make the three dissimilar T-joints.

During welding, the thermal cycles were measured by k-type thermocouples embedded in small holes close to the shoulder's trajectory, on the advancing and retreating sides; see Figure 1b. A data translation device with an acquisition rate of 75 Hz and cold junction compensation was used to record the thermal cycles.

After welding, 63 × 25 × 25 mm specimens were transversely removed to the welding direction of all the series, ground down with sandpaper P2500, and then polished using 3 and 1 μm diamond suspensions. Modified Keller's reagent was effective in revealing the grain boundary of AA2017 and Weck's reagent was effective for such for AA6082, but no effective reagent was found for AA5083. The grain size was determined by the Heyn intercept method. A Leica DM4000M LED optical

microscope was used to analyze the samples. In order to better identify which of the alloys was present in each zone of the nugget, semi-quantitative chemical analyses were performed by scanning electron microscopy/energy dispersive spectroscopy (Zeiss, MERLIN, Field Emission Scanning Electron Microscope-Gemini II/Oxford Instruments, X-MAXN) at 10 kV, bearing in mind that the AA2017 alloy was Cu-rich, the AA5083 alloy was Mg-rich and the AA6082 alloy had intermediate contents of Mg and Si. The samples for electron backscatter diffraction (EBSD) analysis were grinded and polished with 3 and 1 μm diamond suspensions and finished with colloidal silica. Finally, an electrolytic polishing was performed by using a solution of nitric acid and methanol at −10 ◦C and 15–20 V for 30 s. EBSD was performed with an FEI QUANTA 400F scanning electron microscope provided with an EBSD detector system using a scan step size of 0.65 μm.

The Vickers microhardness profiles of the welds were determined using an HMV-G SHIMADZU tester on the weld's cross-section, using 200 g for 15 s. The distance between the test points was 0.5 mm in the nugget zone and 1 mm in the heat-affected zone (HAZ) and base materials.

The base materials were tensile tested at room temperature and at high temperatures (320 and 450 ◦C) using loading speeds of 2 and 72 mm/min, respectively on an Instron 4206 machine provided with a three-stage oven. Three specimens were tested for each base material.

The welded specimens for the tensile and fatigue tests were cut transversely to the direction of the weld with the dimensions of 180 × 20 mm (length × width). The shape and size of the fatigue specimens are shown in Figure 2. The edges of the specimens were rounded and polished to avoid a concentration of surface stresses and the initiation of cracks. The tensile tests were performed in accordance with the ASTM E8 standard for testing metallic materials [16], and the loads were applied in the direction of the skin. Three tensile specimens were tested for each weld series. The local strain fields were recorded with an ARAMIS 3D 5 M optical extensometer from GOM GmbH with digital image correlation (DIC). This is a real time system of measurement of 3D surface strains, based on triangulation, that employs image registration and the tracking of changes in images over time. The specimens were prepared by applying a random black speckle pattern over the previously mat white-painted side surface.

**Figure 2.** Fatigue specimen—shape and size.

The fatigue tests were carried out using an Instron servo-hydraulic machine coupled to an Instron Fast Track 8800 acquisition and control system. The range of stresses varied between 150 and 200 MPa with a frequency of 15–25 Hz; the frequency decreased as the maximum load applied increased, and the stress ratio was set to 0.02. Two test pieces were used for each test condition. In cases where there was greater dispersion, a third trial was carried out. The fracture surface of the fatigue specimens was studied using a Zeiss, MERLIN, field emission scanning electron microscope.

#### **3. Results and Discussion**

#### *3.1. Thermal Cycles in the Welds*

The formation of defects is currently attributed to insufficient heat generation in the weld, inadequate material flow around the pin, and the insufficient consolidation of the deformed material

at the back of the pin, all of which are factors controlled by welding parameters [17]. For this reason, the influence of some process parameters on the thermal cycles induced in the welds was analyzed.

Figure 3 shows the thermal cycles measured on the advancing side for the 562 series. As the welding speed of the series increased, the peak temperature and the cooling cycle time decreased due to the lower ratios (*w*/*v*); see Table 3.

**Figure 3.** The thermal cycles measured in the 562 series, as performed using different welding speeds.

Though these thermal cycles were recorded far from the center of the nugget (11 mm), the curves illustrate the influence of the welding speed on the heat input during the process. On the other hand, the maximum temperature reached in the welds was higher on the advancing side than on the retreating side by about 27 ◦C, as illustrated in Figure 4a. This was due to the asymmetry of heat generation around the tool, as has been shown by other researchers [18,19]. The induced thermal cycle was also greatly influenced by the materials that were located on the advancing and retreating sides. Figure 4b shows that the measured peak temperature was higher when AA2017 was on the advancing side than when it was on the retreating side. This means that there was a larger energy consumption to plastically deform this material than AA5083, so the material position had a more significant effect on the heat generated than the asymmetry of the process.

**Figure 4.** Peak temperature measured (**a**) on the advancing and retreating sides of the weld series 265 and (**b**) on the advancing side of the 562 and 265 weld series.

Figure 4b also shows that the peak temperature observed for the 265-230 series was higher than that obtained in the 265-120 series; this was due to a slight offset of the tool closer to the thermocouple on the advancing side in the 265-230 series.

#### *3.2. Morphology of the Welds*

All welds showed good surface appearance; however, the differences mentioned in the thermal cycles should have influenced the formation of the weld nugget. Figure 5 illustrates the cross-sectional macrographs of the 562 and 265 weld series. This image shows that all the welds had well-defined radii of the fillets, which indicates the effectiveness of the welding parameters and the adopted T-joint configuration. The points where chemical analysis (EDS) was performed are marked with numbers in these macrographs, e.g., Z1, Z2, and Z3, to better aid the interpretation of the flow of the three materials in the nugget. The chemical composition of the different zones is indicated in Table 4.

**Figure 5.** Cross-section macrographs of the three dissimilar materials weld series: (**a**) 562-30, (**b**) 265-30, (**c**) 562-120, (**d**) 265-120, (**e**) 562-280, and (**f**) 265-230.


**Table 4.** Chemical composition in the various zones of both the 562-30 and 265-30 series (wt.%).

An analysis of the macrographs revealed differences in nugget morphologies that varied with increasing welding speeds, as well as with the position of the base materials on the advancing or retreating side. It was possible to observe a great asymmetry in the flow of materials in all the macrographs in relation to the welding center line in both the 562 and 265 series.

The 562-30 series featured three onion ring structures—two located in the skin and one in the stringer fillet zone (marked with ellipses 1, 2, and 3, respectively)—and other zones with different colors; see Figure 5a. Zone 1 essentially consisted of AA2017 positioned on the rear side, which was dragged by the shoulder and remained at the top of the nugget, as seen in Table 4. This table shows only some of the measurements in the 562-30 and 265-30 series in order to illustrate the main differences and reduce the size of the article. The onion ring structure just below ellipse 1—zone 2—was essentially composed of interspersed layers of the three alloys; the dark layers consisted of AA2017. The second onion ring structure indicated by ellipse 2—zone 3—had interleaved layers composed of the three alloys, varying in composition with the location, but with less contribution from AA2017; see Table 4. In zone 4, there was a really good contribution from AA5083. Zone 5 consisted of AA5083 coming from the advancing side. While the material flow on top was a shoulder-driven flow, the last flows were, in fact, pin-driven flows. The onion ring structure was composed of the three materials on the advancing side, as illustrated by zone 7. This showed a good mix of the three materials in the stir zone, contrary to what has been suggested for dissimilar welds [20].

Zone 6 corresponded to the stringer region next to the advancing side fillet, influenced by the action of the pin tip. There was the formation of an onion ring structure, resulting from the contribution of only AA6082 and AA2017. This showed that only one material of the skin, the AA2017, flowed downward into the fillet zone between the tool and dies, which is contrary to what was suggested by Manuel et al. [12] for dissimilar welds.

For the 265-30 series, only one onion ring was formed on the skin and another was formed on the stringer fillet zone, as indicated by ellipses 1 and 2 in Figure 5b. The top layer in this weld consisted of AA6082 (see zone 1), as opposed to the 562-30 series, which had AA2017 on the top; see Table 4. Zone 2 consisted of interspersed layers of the skin materials AA2017 and AA5083. The AA6082 alloy also probably coexisted in this zone, but it was difficult to distinguish via the chemical composition analysis. The peripheral zone of the nugget on the retreating side consisted of layers of only AA5083 and AA2017, according to zone 3. The onion ring structure in the fillet (zone 4) consisted of interleaved layers composed of AA6082 and AA2017, with no AA5083.

As the welding speed increased, in the 562-120 series, the onion ring structure on the fillet tended to disappear, although the material flow remained in this zone; see Figure 5c. In addition, the two onion ring structures in the skin tended to separate, with material from the retreating side entering between the onion rings. This is more visible in Figure 6, where the morphology of the material flow from the 562-30 and 562-120 weld series is compared at a higher magnification. Figure 6a,c compares the advancing sides and the retreating sides (Figure 6b,d).

For the 265-120 series (Figure 5d), the onion ring structure tended to fade, revealing a lack of time or an inability for an orderly flow of layered materials but with increased participation of the AA2017 alloy. This suggested that this alloy had a greater capacity for hot plastic flow and a greater need for time and/or temperature for the AA5083 alloy to flow in layers when it was located on the retreating side.

The onion ring structures disappeared in the 562-280 and 265-230 series, and the stir zone presented a more chaotic appearance with the formation of small internal cavities, either in the center or in the advancing side, as marked with arrows in Figure 5e,f. This was because the heat input decreased as the *w*/*v* ratio decreased, so less material was dragged by the tool and there was less time and temperature for the stable and periodic flow of materials (as suggested by Yoon [21]), thus causing the formation of defects.

Furthermore, it was found that the AA5083 alloy never went downwards to the stringer, regardless of its advancing or retreating side position. This showed that the formation of the nugget depended not only on the position of the materials and process parameters but also on their intrinsic ability to deform at a high temperature.

Figure 7a shows the engineering tensile stress–strain curves of the three base materials obtained at temperatures of 320 and 450 ◦C, as well as the variation of the yield stress with temperature; see Figure 7b. The alloys still had very different yield stresses and tensile strengths at 320 ◦C, and AA2017 had the highest values. The AA2017 and AA6082 alloys showed significant work softening soon after reaching yield stress, unlike the AA5083 alloy that hardened until a strain of 5% before starting to soften. In turn, the AA5083 alloy exhibited much greater plastic strain at fracture than the AA2017 and AA6082 alloys, probably due to dynamic recovery, as suggested by Shi et al. [22]. At 450 ◦C, a noticeable loss of mechanical strength was observed for all the alloys, as was an obvious increase in strain at fracture. Some dissolution and coarsening of strengthening precipitates and partial recrystallization may explain the loss of strength and the increased ductility of the heat-treatable alloys (AA6082 and AA2017) [22,23]. In addition to the loss of strength, the AA5083 alloy had a large steady state flow due to dynamic recrystallization [24]. For this temperature, the yield stress of this alloy was higher than that of AA6082; see Figure 7b.

**Figure 6.** Comparison of material flow between the advancing and retreating sides of 562 welds series: (**a**) Advancing Side-562-30, (**b**) Retreating Side-562-30, (**c**) Advancing Side-562-120, and (**d**) Retreating Side-562-120.

**Figure 7.** Variation of the mechanical properties of base materials with temperature: (**a**) Engineering tensile stress–strain curves and (**b**) yield stress curves.

The heated material was mainly exposed to compression and shear stresses in the stir zone due to its complex interaction with the tool and the colder surrounding material [25]. Though the tensile behavior of the materials at high temperatures did not match what happens in welding, it helped to understand the material flow and was easier to obtain. In welding, temperature decreases rapidly as the distance to the tool increases, so, for the AA5083 alloy, it should be difficult for the tool to drag the material in this zone because some work hardening occurs at a low temperature. Therefore, a little volume of material is dragged around the pin, which results in the poor weldability of the alloy. This is compatible with the difficulty for the AA5083 alloy to flow downwards from the skin to the stringer, which was observed in the welds presented above. The AA6082 and AA2017 alloys experienced significant temperature softening, which allowed the AA2017 to flow downwards from the skin to the stringer, as illustrated above. Similar behavior was observed by Leitão et al. [26] for the AA6082 and AA5083 alloys used in butt welds. The results presented above suggest that, for this welding process, the AA2017 and AA6082 alloys show better weldability than the AA5083.

#### *3.3. Microstructure*

The microstructures in the nugget of the welds between the three dissimilar Al alloys were very complex and difficult to uncover, especially when taking the definition of the grain boundary into consideration. The etchants for each material were different and did not always work when the materials were together. Figure 8a shows a micrograph of the AA6082 base material etched with Weck's reagent. The material was composed of grains elongated in a rolling direction with an average grain size of 59 × 26.3 μm. The distribution of each material in the nugget was very complex, as is illustrated in Figure 8b, for a 562-120 weld. This image shows, at a smaller magnification, a refined grain structure in the areas where AA6082 was present. In the surrounding areas, which were made up of the other materials, the grain boundary was not revealed with this etchant. Figure 8c,d illustrates the microstructure of the same alloy in the nuggets of the 562-30 and 265-30 weld series, respectively. When comparing these last two figures with Figure 8a, a marked refinement of the grain in the nugget is visible. The difference in the average grain size in the nugget between the 562-30 and 265-30 series was marginal, as the grain sizes were about 5.1 ± 1.5 and 5.2 ± 1.6 μm, respectively. A large standard deviation was observed, because the grain size varied with the location. According to the thermal cycles illustrated in Section 3.1, a larger difference in grain size from both weld series would be expected. A possible cause of this similarity was that most of the grain sizes were measured in the central nugget or even on the retreating side, where the grains were most visible.

Increasing the welding speed—from 30 to 280 mm/min, for instance—had a very small effect on the nugget grain size, with it remaining within the range of 4–5 μm. For AA2017, although its grain size (20.2 × 9.5 μm) was smaller than that of AA6082, the nugget grain size was also in the 4–5 μm range for all the welding series. The significant plastic deformation suffered by the materials in the stir zone and the recrystallization occurring in small and very confined areas, as seen in Figure 8b, could penalize grain growth with the ephemeral increase in heat input. The authors believe that the coexistence of the three alloys in the nugget cancelled the effect that the increase in welding speed had on the decrease in grain size, contrary to that suggested by Kalemba-Rec et al. [27]. A study by Ahmed et al. [28] found that grain size decreases with the increase in welding speed for similar welds, but the same effect does not occur for dissimilar welds.

The nuggets of both weld series were analyzed by EBSD to study the weld microstructure further, as shown in Figures 9 and 10 for the 562 and 265 welds, respectively. The regions under analysis are indicated by dots in Figures 9a and 10a to make them more discernible in the weld macrographs, although they are small vertical or horizontal scanning lines. It can be observed from the figures that the onion-ring regions of both welds tended to present a refined microstructure, which agreed well with the metallographic study, but the grain size was not uniform in the nugget. A gradient in grain structure is visible in these maps, with the grains composing the onion-ring regions presenting a smaller grain size and being more equiaxed than the grains in neighboring areas; see Figure 9b,c

and Figure 10b–d. Plastically deformed grains are visible both in the onion-ring regions and the neighboring areas (see Figures 9b and 10c), which agrees well with the flow features characterized above. A further refined microstructure (2 μm) is also observed in Figure 9d, and this corresponds to a region in the stringer that was under the action of the pin tip where little heat was generated. The great microstructural heterogeneity of the weld nugget was due not only to the differences in local heat generated but also to the very complex flow features of the different alloys that made it up.

**Figure 8.** Microstructure of the: (**a**) base material AA 6082, (**b**) nugget of the 562-120 series, (**c**) nugget of the 562-30 series, and (**d**) nugget of the 265-30 series.

**Figure 9.** Electron backscatter electron backscatter diffraction (EBSD) analysis conducted for the 562-120 weld: (**a**) zones analyzed and EBSD maps registered in zones 1 (**b**), 2 (**c**), and 3 (**d**).

**Figure 10.** EBSD analysis conducted for the 265-30 weld: (**a**) zones analyzed and EBSD maps registered in zones 1 (**b**), 2 (**c**), and 3 (**d**).

#### *3.4. Hardness and Tensile Behavior*

The skin hardness profiles of the 562 and 265 weld series are shown in Figure 11a for defect-free welds. The hardness of the base materials is also shown in the same figure using dashed (AA2017) and dotted (AA5083) lines. It appears that the increase in welding speed did not significantly change the hardness in the HAZ, either on the AA5083 or AA2017 sides. However, there was a slight loss in hardness in the HAZ close to the tool path, mainly when AA2017 was located on the advancing side for the lowest welding speed. This could be attributed to the dissolution of hardening precipitates, as stated by Dong et al. [29]. The stir zone had an irregular hardness pattern with some peaks in the current welds as a result of the non-homogeneous mixing of the three different base materials during the process, as illustrated in Section 3.2.

The stringer hardness profiles for the 562 and 265 series are shown in Figure 11b. In all the series, there was a significant reduction in hardness in the thermomechanically-affected zone (TMAZ) and HAZ, which was generally attributed to the dissolution and coarsening, respectively, of the hardening precipitates [29]. This reduction in hardness was higher for the lowest welding speed series due to the higher and longer thermal cycles, as per Section 3.1.

Figure 12 shows the tensile curves of the skin of defect-free specimens of the 562 and 265 series, as well as of the three base materials for comparison. Table 5 summarizes the mean values of tensile strength and strain at failure of each series, as well as the zones where the failure occurred. The effect of welding speed on weld strength is well-illustrated in Figure 12, where the welds performed at the highest speed were slightly above the others but just below the AA5083 alloy.

Table 5 shows that specimens from both series (562 and 265) made at the lowest welding speed (30 mm/min) had the lowest efficiency values (losses of about 20%) and broke in the HAZ, close to the stir zone, on the AA5083 side. The efficiency is defined as the ratio between the maximum tensile strength of each weld series and the tensile strength of the least resistant base material, in this case being AA5083. Specimens from both series performed at 120 mm/min had the highest efficiency values, as suggested by the hardness results, although the difference was small; see Figure 11a. In addition, these test specimens broke in the HAZ on the AA5083 side but far from the stir zone. This is illustrated in Figure 13, which represents the distribution of the strain fields, obtained by an optical extensometer

in two specimens produced with different welding speeds, and at an instant close to failure. Figure 13a illustrates a specimen from the 562-30 series whose highest plastic deformation, the red zone, and failure occurred in the HAZ next to the stir zone, while Figure 13b illustrates a specimen form the 562-120 series that also broke in the HAZ but far from the stir zone.

(**b**)

**Figure 11.** Effect of variation of the welding speed on the hardness profile in the (**a**) skin and (**b**) stringer.



**Figure 12.** Tensile stress–strain curves of the 562 and 265 weld series and the base materials.

**Figure 13.** Strain distribution in tensile test specimens close to failure from the (**a**) 562-30 and (**b**) 562-120 series.

Figure 14 illustrates the influence of the welding speed on the fracture surface morphology of test pieces from series 562-30 and 562-280 performed at speeds of 30 and 280 mm/min, respectively. Figure 14a shows a ductile fracture surface of a 562-30 specimen with thin dimples and some larger dimples, as shown in the enlargement (5000x) of a zone in the lower right corner of the image. The fractures of the remaining series had similar morphologies because they occurred in the HAZ of the same alloy. Figure 14b shows a ductile fracture surface of a 562-280 series specimen, but the fracture was caused by a defect here, already mentioned in Section 3.2 and indicated by an arrow in the image.

#### *3.5. Fatigue Strength*

Figure 15 shows the S–N (stress range/number of cycles to failure) curves for the 562 and 265 series made at 30 and 120 mm/min, as well as the curve for the AA5083 alloy (the least resistant of the base materials). The fatigue test specimens that did not break for more than one and a half million cycles are shown with a horizontal arrow.

(**a**) (**b**)

**Figure 14.** Fracture surface morphology of a specimen in the (**a**) 562-30 and (**b**) 562-280 series.

**Figure 15.** Influence of welding speed on the fatigue strength of the 562 and 265 series.

This figure shows that all the weld series presented a lower fatigue strength than the base material, which indicates that the welding process reduced fatigue strength regardless of the welding parameters used. However, the 562-120 series, produced at the advancing speed of 120 mm/min, presented better resistance to fatigue compared to the 562-30 series, performed at a lower speed of 30 mm/min but with an increase in the fatigue strength of only 4.5% at 5 <sup>×</sup> 105 cycles.

A similar increase in the fatigue strength with the increase in the welding speed could be observed in the 265 series for identical welding speeds of about 5.6%. Ericsson and Sandstrom [30] observed that for low welding speeds, their welds using the AA6082 alloy showed a slight increase in mechanical strength and fatigue compared to the higher welding speed due to the increased amount of heat generated in welding per unit length. In the current study, the opposite behavior was observed, which can be explained by the increase in mechanical strength with the increase in the welding speed, as shown in Table 5.

Figure 15 further shows that the 265 series had a slightly higher fatigue strength than the 562 series, while the 265-120 series had the highest strength. This may also have been related to the tensile properties of the welds, as specified in Table 5.

One factor that influenced the fatigue strength had to do with the surface finish of the samples tested, as the presence of small surface defects made crack initiation and propagation more probable, thus leading to premature failures [31]. The specimens were polished before testing in the current work; however, the welding crown was more difficult to polish without significantly reducing the

thickness of the skin, which introduced some variability in the results. The specimens broke mostly at the TMAZ or close to the HAZ in the AA5083 alloy, either in the higher or lower stress ranges.

The fracture surface of the fatigue specimens had identical morphology. Figure 16 illustrates the fracture surface of the specimen from the 562-30 series subjected to a stress range of 180 MPa that broke after 193,700 cycles. Figure 16a shows a general view of the fracture surface of the specimen, where the fracture zones analyzed in more detail are indicated with rectangles. The crack started on the welding surface marked with an arrow and was propagated by fatigue through the thickness of the skin. The orientation of the characteristic fatigue striations was perpendicular to the growth direction of the crack, as shown in Figure 16b,c in more detail. Figure 16c shows the detail marked with a rectangle in Figure 16b in greater magnification. In the final part of the crack propagation phase, a ductile fracture occurred, as evidenced by the presence of small dimples and some larger dimples in the transition zone, as seen in Figure 16d. The fracture of this test specimen occurred in the TMAZ on the advancing side of the AA5083 alloy.

**Figure 16.** Fracture surfaces of a sample from the 562-30 series: (**a**) general fracture surface, (**b**) detail from location 1, (**c**) detail from location 3, and (**d**) detail from location 2.

#### **4. Conclusions**

The carried out research led to the following conclusions:


• Placing the more resistant alloy (AA2017) on the advancing side rather than on the retreating side generates higher local weld temperature and provides stronger joints and with better fatigue behavior.

**Author Contributions:** The conceptualization and revision of the article, A.L. and I.G.; methodology, data treatment, and writing—review and editing of the manuscript, N.M. and R.M.L.; supervision and formal analysis of the results of the fatigue tests, J.D.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by FEDER and FCT, as indicated in the acknowledgement.

**Acknowledgments:** This research is sponsored by FEDER funds through the program COMPETE–Programa Operacional Factores de Competitividade–and by national funds through FCT–Fundação para a Ciência e a Tecnologia –, under the project UIDB/00285/2020. The author N. Manuel is supported by Ministério das Pescas from Angola through a fellowship. All support is gratefully acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **E**ff**ect of FSW Traverse Speed on Mechanical Properties of Copper Plate Joints**

#### **Tomasz Machniewicz 1, Przemysław Nosal 1, Adam Korbel <sup>1</sup> and Marek Hebda 2,\***


Received: 22 January 2020; Accepted: 14 April 2020; Published: 20 April 2020

**Abstract:** The paper describes the influence of the friction stir welding travel speed on the mechanical properties of the butt joints of copper plates. The results of static and fatigue tests of the base material (Cu-ETP R220) and welded specimens produced at various travel speeds were compared, considering a loading applied both parallel and perpendicularly to the rolling direction of the plates. The mechanical properties of the FSW joints were evaluated with respect to parameters of plates' material in the delivery state and after recrystallisation annealing. The strength parameters of friction stir welding joints were compared with the data on tungsten inert gas welded joints of copper plates available in the literature. The results of microhardness tests and fractographic analysis of tested joints are also presented. Based on the above test results, it was shown that although in the whole range of considered traverse speeds (from 40 to 80 mm/min), comparable properties were obtained for FSW copper joints in terms of their visual and microstructural evaluation, their static and especially fatigue parameters were different, most apparent in the nine-fold greater observed average fatigue life. The fatigue tests turned out to be more sensitive criteria for evaluation of the FSW joints' qualities.

**Keywords:** FSW; copper; butt joint; mechanical properties; fatigue performance; traverse and rotation speed

#### **1. Introduction**

Friction stir welding (FSW) is a relatively novel technique for joining materials, and thanks to the wide possibilities of application, it has been gaining popularity rapidly recent years [1]. In particular, this technique can be used as one of few methods for joining metals that are hard to weld or have physicochemical properties significantly different from each other, i.e., in cases when the use of conventional fusion welding is strongly limited and braze welding does not yield sufficient joint strength [2]. For this reason, FSW is commonly used not only for welding aluminium alloys [3,4] but also for many other metallic materials such as magnesium alloys [1,5–7], titanium alloys [1,8], steel [9,10], copper [11–16], and different combinations of dissimilar metals [1].

Friction stir welding, as a solid-state process, is based on the plastic deformation of joined materials, elicited by using a special tool consisting in general of a shoulder and a mixing pin (Figure 1). The basic function of this tool is to generate heat, which enables the plasticisation of the materials, and then their mixing. Around the FSW joint, as shown in Figure 1, three characteristic zones within which the material properties have changed in relation to the base material (BM) are generally specified. These zones are: (i) the heat affected zone (HAZ), (ii) the thermo-mechanically affected zone (TMAZ) and (iii) the weld nugget (WN), where the material is fully recrystallized [1]. In addition, due to the complex motion of the working tool, there are two distinctive sides of the formed joint: the advancing

side—where the tangential vector of the rotational speed of the tool is compatible with its travel speed vector, and the retreating side—where the senses of mentioned vectors are opposite (Figure 1).

**Figure 1.** Schematic drawing of the friction stir welding (FSW) process.

The kinematics of the FSW process consist of two main motions, rotational and progressive, along the interface of joined materials. These motions correspond to the two basic parameters of the technological process which are the rotation speed (ω) and the traverse speed of the welding (*V*). These parameters affect the amount of heat generated during the process and thus the quality and properties of the wrought weld. Thus, for a given tool geometry, the mechanical properties of FSW joints are greatly influenced by the abovementioned technological parameters applied to the welding process as confirmed by numerous examples in literature [1,3,12,15,16]. A poor choice of these parameters may lead to various kinds of imperfections of the joints, significantly reducing their strength properties [17–22]. Although extensive research on the influence of FSW process parameters on the weld quality has already been conducted, existing publications most often concern aluminium alloys and static joint properties. The database on non-aluminium materials is more modest, and only rudimentary information can be found on the fatigue properties of produced joints. In particular, one of the materials that is still poorly known in this respect is copper, despite the fact that, due to the fact of its physical and chemical properties, copper combined with FSW technology is increasingly used in many areas of industry including the nuclear [22], energy [23] and automotive sectors [24]. Elements made in this way must have the required strength, not only in static but also in fatigue loading conditions, which can only be achieved through the appropriate choice of FSW process parameters.

A review of the literature provides divergent data on FSW parameters optimal for copper sheets. There are examples of significantly different combinations of rotation speed/traverse speed/plate thickness recommended for welding copper, e.g., 1000 rpm/30 mm·min−1/2 mm [11], 400 rpm/100 mm·min−1/3 mm [12], 250 rpm/61 mm·min−1/4 mm [13] and 1300 rpm/170 mm·min−1/6 mm [14]. Xue et al. [15], while welding copper plates with a thickness of 5 mm, applied various rotary speeds ω ≥ 400 rpm for the traverse speed of 50 mm/min and various traverse speeds *V* ≥ 50 mm/min for the rotary speed of 800 rpm. In the entire range of the considered parameters, they observed the systematic effect of joint strength increasing with increasing traverse speed for constant ω = 800 rpm, and with decreasing rotation speed at constant *V* = 50 mm/min. For the same sheet thickness, Khodaverdizadeh et al. [16] used both higher (*V* = 75 mm/min, ω = 600–900 rpm) and lower (*V* = 25 mm/min, ω = 600 rpm) process parameters relative to the research conducted by Xue et al. [15]. The obtained test results confirmed that the highest mechanical parameters of the joint were received for the combination of the lowest of the considered rotation speeds (600 rpm) and the highest traverse speed (75 mm/min). These results, however, are contrary to the generally expressed view that, due to the high thermal conductivity and high melting temperature of copper, a moderate welding speed is recommended, so that the right amount of heat energy can be generated [13]. On the

other hand, excessive heat generation during welding of copper sheets hardened beforehand by cold rolling causes a more intensive reduction of strength parameters of the joint within the weld, due to the recrystallisation process. Because of the high degree of complexity of the processes occurring during FSW, it is important to determine the optimal technological parameters in order to obtain the highest possible strength properties of joints. For this purpose, the results of mechanical tests of butt joint specimens of electrolytic copper (99.998% Cu) with a thickness of *t* = 5 mm made by FSW with different traverse speeds are presented in this paper. Based on these, the influence of the applied process parameters on structural changes, microhardness and mechanical properties of the fabricated joints were analysed. When selecting optimal FSW traverse speed, static tests results are usually taken into account during research [11–15], but this work also considers the fatigue properties of FSW joints (i.e., fatigue lives). This is an approach rarely found in the literature but is considered by the authors as more sensitive criteria for evaluation of the FSW joints' qualities. Moreover, data on the impact of FSW welding parameters on the properties of copper joints are usually limited to one specimen orientation. The results in this paper demonstrate that the variations of FSW parameters may have a qualitatively different effect on the properties of friction stir welds oriented longitudinally and transversely to the rolling direction of the jointed plates. After considering these additional criteria, the optimal combination of transverse and rotation speed selected in respect to the best strength parameters of FSW copper joints proved to be different from the proposals recommended so far in the literature.

#### **2. Materials and Methods**

Base material samples and FSW butt joints were fabricated from 5 mm thick plates of high purity electrolytic copper (Cu-ETP R220) with a chemical composition containing more than 99.9% Cu, according to the EN 573-1 standard. The FSW welds were performed on a conventional Jafo CNC milling machine at a constant rotation speed ω = 580 rpm and at three different traverse speeds: *V* = 40, 60 and 80 mm/min. The other FSW parameters were zero tilt angle, plunged depth of 0.3 mm, tool plunging speed of 2.5 mm/min and dwelling time of 5 s. The dimensions of the plate welded in such a way were approximately 300 mm × 300 mm. A view of the tool geometry used for welding and a representative example of the weld are presented in Figure 2. Samples from base material and welded plates for static and fatigue tests, with geometry shown in Figure 3, were cut out using the water-jet technique. The good quality (defect-free) of the FSW joints produced in accordance with the above parameters was confirmed by visual observation and ultrasonic evaluation.

**Figure 2.** The FSW joint made at ω = 580 rpm, *V* = 60 mm/min (**a**) and geometry of the tool used for welding (**b**) (dimensions in mm).

**Figure 3.** Specimen geometry used in static tests (**a**) and fatigue tests (**b**) (dimensions in mm).

For FSW joints, the welds were located in the centre of the samples as presented in Figure 3. Specimens, both from the base material and from welded plates, were made for two sheet configurations, i.e., longitudinally (orientation L) and transversally (orientation T) to the direction of their rolling process, as presented schematically in Figure 4. The test plan was additionally extended by tensile tests of specimens subjected to earlier annealing at 600 ◦C. These results were considered as a reference point for the assessment of weld strength properties in the heat affected zone (HAZ). The annealing temperature was adopted based on temperature maps, as presented in Figure 5, registered by the thermal imaging camera on the surface of sheets during the welding process in the tool working area. Although the temporary weld temperature (*T*) just behind the rotating shoulder was about 800 ◦C, it stabilised a short distance from the tool at 600 ◦C.

**Figure 4.** Schematic presentation of the longitudinal (L) and transversal (T) specimen orientation.

Static and fatigue tests were carried out using the MTS 810 universal testing machine, with a load limit of 100 kN. Tensile tests were conducted according to the EN ISO 6892-1:2016 standard under displacement control at a rate of 0.5 mm/min. The strains were measured using axial extensometer (Epsilon 3542-025M-025-ST) with a gauge base of 25 mm and a measuring range of ±6.25 mm. The extensometer was mounted in the centre of the samples, covering the width of the weld when FSW joints were tested.

**Figure 5.** The temperature field (*T*) registered during FSW process by a thermal imaging camera on the copper plate surface behind the tool.

The stress-controlled constant-amplitude fatigue tests were performed with the stress ratio *R* = 0 using sinusoidal waveform loading at frequency *f* = 20 Hz. For the solid samples of the base material, four stress ranges were considered: Δ*S* = 210, 220, 230, 240 MPa. In the case of samples with an FSW joint, comparative fatigue tests were carried out only in the stress range of Δ*S* = 160 MPa due to the fact of their lower strength and large scatter of the results. The number of specimens subjected to monotonic tensile tests and fatigue tests is summarised in Table 1.


**Table 1.** Tensile and fatigue test matrix.

The microhardness measurements were performed for base material and three FSW joint specimens (i.e., one specimen per considered travel speed) in accordance with the ISO 6507-1 standard on Nexus 423A hardness equipment with a Vickers indenter, applying a loading force of 25 g and a measuring time of 10 s. The metallographic investigation was conducted using an Eclipse ME600 light optical microscope.

#### **3. Results and Discussion**

#### *3.1. Microhardness Tests and Metallographic Analysis*

The microhardness distribution determined for the various traverse speeds on the cross-section of the joints along the centre of the specimens' thickness is presented in Figure 6.

**Figure 6.** Microhardness profiles of the FSW joints corresponding to various traverse speeds and comparison of representative microstructure images in selected zones of the FSW copper joints produced with a traverse speed of 80 mm/min.

This figure also presents the representative microstructure images of the weld nugget, heat affected zone and base material for a joint welded at a speed of *V* = 80 mm/min. For this *V*-speed value, the profile of microhardness had a characteristic "W" shape (Figure 6) which is typical for most FSW joints [13,15–18,25]. For each of the traverse speeds used, a sudden decrease of microhardness in the HAZ was found. The minimum value of the microhardness (76.2 HV) was measured on the border of the HAZ and WN. The above trends corresponded well to the average grain sizes observed in the respective zones of the joint. Thus, the higher hardness in the WN zone than in HAZ was related to smaller grain size which was 10 μm for WN and 90 μm for HAZ. The base material, despite the larger grain size (approximately 70 μm), displayed a hardness and strength properties higher than those measured in the welded zone which is the effect of strain hardening arising from the sheet rolling process. A lower traverse speed value increases the amount of heat generated during the FSW process which has a direct impact on the lower hardness in the TMAZ and is manifested by a flatter microhardness profile for speeds of 60 and 40 mm/min (Figure 6). These trends, although widely shown in the literature [13,15,16,25], are contrary to other available data [26–28] showing that microhardness of the FSW copper joints increasing over the level related to the base material. For example, in the work of Berenji [26], for pure copper plates with the same thickness as those described in this paper (5 mm) and with very similar tool geometry and nearly the same rotation speed (600 rpm), a consistent increase of microhardness in the weld zone, over 75 HV for base material, was observed only when the traverse speed was above 25 mm/min. As in other works [27,28], this effect increased with the traverse speed. The highest reported ratios of such obtained microhardness of weld zone over the microhardness of the base material were (in HV): 90/75 [26], 114/77 [27] and 105/60 [28]. It may be noted that increased microhardness in the region of the FSW joints was observed in the case of soft copper conditions (microhardness below 80 HV) [26–28], while base material under higher strain with microhardness of over 80 HV results in a drop in hardness in the welded zone [13,15,16,25] which was also confirmed by the results of this paper.

#### *3.2. Monotonic Tests Results*

The tensile curves (σ—engineering stress; ε—engineering strain) for solid samples, including the base material specimens (specimens TBM), loaded parallel and perpendicularly to the rolling direction, and annealed specimens are presented in Figure 7. Determined from the static tensile tests, mechanical parameters of the solid and welded specimens, represented by yield stress (YS) ultimate stress (UTS) and reduction of area (AR), are summarised in Table 2.

**Figure 7.** Tensile curves of solid samples: (**a**) base material—orientation L; (**b**) base material—orientation T, (**c**) base material—comparison of the representative curves for L and T orientation, (**d**) base material after annealing at 600 ◦C.


**Table 2.** Mechanical properties of tested samples.

As presented in Figure 7, the strength properties of solid sheets for the given orientation (L and T) were characterised by high repeatability, apart from a few percent spread of the strain at failure. Samples loaded along the rolling direction (TBM-L) above the elastic strain range showed almost perfectly plastic characteristics (Figure 7a). Comparison of the tensile curves for both orientations (Figure 7c) demonstrate that the strain at failure was slightly higher in the case of loading in accordance with the rolling direction of the sheet. In the transversal direction (specimen TBM-T, Figure 7c), where a slight strengthening effect was observed, the yield stress was lower by about 4% and the ultimate tensile strength was higher by about 3%, compared to longitudinal orientation (Table 2). The recrystalising annealing at 600 ◦C (specimens TRM, Figure 7d) caused a seven-fold decrease in yield stress and a decrease of about 17% in ultimate strength, with a simultaneous increase in strain at failure by close to 60%, and an increase in the value of reduction of area at failure by about 2%–4% (Table 2).

Monotonic tests of FSW joints comprised three tensile tests for each of the three traverse speeds considered. The cracking of the joints under static loading always occurred on the retreating side of the welds in the area of the TMAZ as shown in Figure 8. The exemplary tensile curves presented in Figure 9, corresponding to the highest traverse speed (i.e., *V* = 80 mm/min—FSW80 specimens), were characterised by the largest observed scatter, which was particularly pronounced for the longitudinal orientation (specimens FSW80-L, Figure 9a). Two samples from this series fractured before reaching the extreme point on the static tensile curve (Figure 9a), which was also manifested by lower values of strain at failure, as well as lower ultimate strengths (Table 2). The tensile curves representative of the joints welded at different traverse speeds are compared in Figure 10.

**Figure 8.** Fracture location of the tensile FSW joint samples.

**Figure 9.** Tensile curves of FSW joint specimens produced with rotary speed ω = 580 rpm and traverse speed *V* = 80 mm/min: (**a**) orientation L; (**b**) orientation T.

**Figure 10.** Comparison of the representative tensile curves of the specimens joined by FSW with various traverse speeds: (**a**) orientation L; (**b**) orientation T.

As may be observed, for longitudinal orientation the tensile curves regardless of *V*-speed are almost overlapped, except for the aforementioned effect of the lower ductility associated with the traverse speed of 80 mm/min (Figure 10a). In the transverse orientation (Figure 10b), a consistent trend of decrease of YS and UTS values with increasing traverse speed was found, which is qualitatively different to the available literature data [15,16,29]. Assuming that reduction of the joint strength parameters is induced by a copper recrystallisation process in the HAZ, this effect should be more

intensive for a lower *V*-value, which corresponds to the higher amount of thermal energy generated during the FSW process. Besides the aforementioned experimental data, this mechanism is also confirmed by numerical simulations of the FSW welding process [30]. Macroscopic analysis of tensile fracture surfaces of FSW joints did not show distinct defects in the weld structure. The only dark streaks were observed on the fracture of the FSW specimen with the lowest strength parameters (sample FSW80-L1) as presented in Figure 11. Energy-dispersive spectroscopy (EDS) analysis of the areas marked on Figure 11 revealed the presence of iron, probably from the FSW tool, with content ranging from 1% to 4%.

**Figure 11.** The fracture surface of the FSW80-L1 sample, showing areas of EDS analysis and iron content.

Reduction of the strength properties of FSW joints in relation to solid material samples, as the effect of the copper recrystallisation in the heat affected zone, is clearly visible in Figure 12.

**Figure 12.** Comparison of representative tensile curves of base material (TBM), annealed material (TRM) and FSW specimens welded with *V* = 60 mm/min (FSW60): (**a**) orientation L; (**b**) orientation T.

The ratios of the mechanical parameters observed for FSW joints at different traverse speeds to the parameters specific to the solid material (YS/YSBM and UTS/UTSBM) for both sample orientations are shown in Figure 13. Resulting from the FSW process, the YS decreased on average by more than 60% (Figure 13), i.e., from 230–240 MPa to 80–100 MPa (Table 2). This effect was almost independent of the rolling direction of the sheet. A smaller but still significant decrease, amounting to 15%, was also observed for UTS. Figure 13 reveals a consistent trend of decreasing strength parameters of FSW joints with increasing traverse speed. Although (with the exception of the YS changes for transversal orientation (Figure 13b)) the variations presented in Figure 13 are very moderate, these trends are, in qualitative terms, opposite to those described in the literature [15,16,27]. In general, with increasing traverse speed, the amount of thermal energy generated during FSW decreases [4]. This contributes to maintaining higher strength parameters due to the lower intensity of the copper recrystallisation process. The excessive scatter of the strength parameters observed in Figure 13a (longitudinal orientation) for the speed *V* = 80 mm/min confirms the earlier mentioned objections regarding this series of joints.

**Figure 13.** The mechanical parameters of FSW copper joints (YS and UTS) at different traverse speeds, normalised by base material properties (YSBM and UTSBM): (**a**) longitudinal orientation, (**b**) transversal orientation.

Nevertheless, the strength parameters of FSW welds, particularly the YS, remain higher in relation to the properties of the copper subjected to recrystalising annealing (Figure 12, Table 2). In addition, these FSW joint parameters were also significantly higher—UTS by close to 30%, and YS by about 67%—than the literature values for properties of butt joints welded from pure copper sheets using the tungsten inert gas (TIG) method [31]. This may be due to the strong plastic deformation of the material caused by the tool acting on the plates being joined by the FSW process, which does not occur in fusion welding (including TIG welding). The result is a refined microstructure formed in the region of the weld nugget and in the zone of thermo-mechanical impact (see Section 3.1.) which improves the mechanical properties of the joints. In addition, the temperature of the FSW process is lower than at the fusion welding, thus avoiding problems of porosity, cracking, sheet distortion and large size of heat affected zone. The higher strength of FSW joints compared to solid copper samples after annealing indicates that the reduction of mechanical properties of these joints is predominantly related to the loss of the cold-work hardening effect as a result of the recrystallisation process in the heat affected zone of the joint. Therefore, the improvement of strength parameters of the joints made using the FSW technique from strain-hardened copper sheets is possible by limiting the material's heating during welding. This, in addition to selecting the optimal FSW parameters, may be implemented using a cooling medium during the FSW process [29].

#### *3.3. Fatigue Test Results*

The fatigue test results for base material specimens for both orientations are compared in the form of *S*–*N* curves in Figure 14. As may be observed, the specimens loaded perpendicularly to the rolling direction of the sheet repeatedly showed slightly lower fatigue strength (of about 4%) than the samples of longitudinal orientation. This is opposite to the trend of ultimate strength changing for both orientations under monotonic loading (Table 2).

In all specimens with T-orientation, and in specimens with L-orientation welded at a speed of *V* = 60 mm/min, fatigue cracks occurred on the retreating side in the plane of the notch formed along the edge of the rotating tool plunge as illustrated in Figure 15a. In the samples with longitudinal orientation welded at speeds of *V* = 40 mm/min and 80 mm/min, fatigue cracks were also located on the retreating side but, similar to static tests, at some distance from the weld edge (Figure 15b). The plane of cracks developed at the welds' edges was nearly perpendicular to the face of the sample (Figure 15a), while in the TMAZ (Figure 15b) the cracks propagated on a plane inclined to the face of sample at an angle of about 45–60 degrees. In general, the cracks initiated in the corner of the sample on the face side of weld as illustrated in Figure 15c.

**Figure 14.** *S*–*N* curves of Cu-ETP R220 base material specimens.

**Figure 15.** Location of the fatigue cracks observed in FSW joints: (**a**) crack at the edge of the weld, (**b**) crack at some distance from the edge of the weld, (**c**) typical mode of fatigue crack initiation.

Fatigue strength of the specimens containing FSW joints was significantly lower compared to base material specimens. In the tests of the FSW joint specimens with longitudinal orientation (Figure 16a), the highest fatigue lives (*N*f), and simultaneously the most repetitive results were obtained for samples welded at a speed of *V* = 60 mm/min. For this configuration, other considered traverse speeds, i.e., the lower (*V* = 40 mm/min) and the higher (*V* = 80 mm/min), led to significant lower fatigue lives and much greater scatter of the test results. One of the specimens welded with a traverse speed of 80 mm/min cracked during the first loading sequence. For all samples from the two series, of the lowest fatigue life was characterised by the distinct location and morphology of the cracks. FSW joint specimens oriented transversally to the rolling direction (Figure 16b) were characterised by more repeatable results, regardless of the traverse speed used. For this orientation, the difference between the average fatigue lives for specimen series associated with all considered *V*-values did not exceed 22%. In contrast to samples with L orientation, the highest fatigue lives were observed for the joint specimens welded at a speed of *V* = 80 mm/min. Concurrently, the speed of *V* = 60 mm/min, which had given the highest durability for the samples with L-orientation, provided the lowest fatigue lives in the case of transversally oriented specimens.

**Figure 16.** Influence of traverse welding speed on fatigue life of FSW joint specimens (*R* = 0, Δ*S* = 160 MPa): (**a**) orientation L; (**b**) orientation T.

Comparing the fatigue test results obtained for specimens cut parallel (L) and perpendicularly (T) to the rolling direction, it may be pointed out that the traverse speed *V* = 60 mm/min was the most advantageous of the parameters used, due to the highest level of repeatability of observed fatigue lives. For this *V*-value, the difference in average fatigue lives for both specimen orientations was only 3%, while for the speed of 40 mm/min the average fatigue life for the transversal orientation was more than three times higher, and for a speed of 80 mm/min it was more than ten times higher, compared to the longitudinal orientation of the joints. Even for the most favourable traverse speed, the fatigue strength of specimens containing FSW joints, at fatigue life of 130,000 cycles, was lower by about 30% compared to the base material specimens.

Available research papers concerning the impact of FSW parameters on the mechanical properties of copper FSW joints [11–16,25–29] usually focus on the microstructure, hardness profiles and static properties of joints. To the best knowledge of the authors, there are no publications on FSW copper joint properties which reference various joint orientation and/or fatigue properties. However, as the results of this paper demonstrate, accounting for the last factors may qualitatively change the conclusions about the strength properties of considered joints. As follows from Section 3.2. (Table 2 and Figure 13), the traverse speed had a rather moderate effect on the static strength parameters of the tested FSW joints. For tested FSW copper joints, contrary to available literature data [15,16,27], YS- and UTS-values decreased as traverse speed increased. However, this trend was significant only for YS-variation in the case of joints with T-orientations, for which the YS decreased by almost 19% with an increase of *V*-speed from 40 to 80 mm/min. For the other cases, the changes in YS and UTS values did not exceed 6%. Taking into consideration only the tensile and microhardness properties of joints, the traverse speed of 40 mm/min could be selected as the optimal one. Fatigue tests, however, proved to be more sensitive criteria for evaluation of the FSW joints' qualities. These tests clearly confirmed the weakness of the joints made at a speed of 80 mm/min with longitudinal orientation (which was earlier indicated only by the scatter of tensile test results (Figure 13)) and revealed the poor quality of the joints produced at a speed of 40 mm/min for the same orientation which had not been demonstrated earlier by other studies.

#### **4. Conclusions**

In this work, the properties of pure copper FSW joints were studied by means of microhardness, tensile and fatigue tests as well as microstructure and fractography analysis. As the main variables in conducted tests, three values of traverse speed and two different plate orientations in terms of rolling direction were considered. The constant rotation speed value (580 rpm) and different traverse speeds (40, 60 and 80 mm/min) were preliminarily selected for producing defect-free joints in terms of their visual and ultrasonic evaluations. The most relevant conclusions can be summarised as follows:


of the base material after annealing. Regardless of traverse speed, the tensile strength parameters of FSW joints were higher for transversal than for longitudinal orientation. Tensile tests revealed a consistent trend of decreasing strength parameters of FSW joints (i.e., yield stress and ultimate stress) with increased traverse speed. Although this effect is moderate, it is qualitatively different from the data reported so far in the literature.

3. The fatigue tests turned out to be more sensitive criteria for evaluation of the FSW joints' qualities compared to other kind of examinations applied in this work. These tests clearly confirmed the weakness of the FSW joints produced at a speed of 80 mm/min in longitudinal orientation, which was earlier indicated only by the scatter of tensile test results, and revealed the poor quality of the joints made at a speed of 40 mm/min for the same orientation, which had not been predicted earlier by other studies. All samples from the two series of the lowest fatigue life were characterised by distinct location and morphology of the cracks. Considering all static and fatigue tests—for given plates, adopted tool geometry and specified rotary speed—the traverse speed of 60 mm/min proved to be the most advantageous. It should be noted that conclusions on the quality of the FSW joints resulting from fatigue tests would be different if they were referred separately to each of the considered joint orientations. This means that the variations of FSW parameters may have a qualitatively different effect on the properties of friction stir welds oriented longitudinally and transversely to the rolling direction of the joined plates.

**Author Contributions:** Conceptualization, T.M. and P.N.; methodology, T.M., M.H. and P.N.; static and fatigue tests, T.M., P.N. and A.K.; microhardness tests and metallographic analysis, M.H. and P.N.; visualization, P.N. and T.M.; data curation, P.N., T.M. and M.H.; writing—original draft preparation, T.M.; writing—editing and revision, T.M., P.N., A.K. and M.H.; correspondence, M.H.; funding acquisition, T.M., P.N. and M.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** The financial support from subsidy granted by the Polish Ministry of Science and Higher Education within project no. 16.16.130.942 is acknowledged. This work has been supported within the framework of the project: The innovative system for coke oven wastewater treatment and water recovery with the use of clean technologies - INNOWATREAT that has received funding from the Research Fund for Coal and Steel under grant agreement no. 710078 and from The Ministry of Science and Higher Education Poland from financial resources on science in 2017–2019.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
