Abstract
Friction welding of high strength aluminium alloys was considered in the current study. The mechanism of bond formation in dissimilar alloy welding was compared to similar alloy welding. The differences in flow properties between two different alloys of aluminium lead to uneven deformation behaviour while welding. Since flow behaviour directly affects bond formation and strength, it is important to understand its behaviour. Rods of aluminium alloys AA 2024 and AA 6061 were welded to themselves and to each other denoting similar and dissimilar weld systems. Burn-off length, which is one of the important parameters in friction welding was varied in steps to see how it affects the bond formation. The flash geometry of the welds was observed to interpret the flow differences. Reasons for variations in bond formation were discussed with the help of fracture surfaces.
Keywords
Introduction
Nowadays, there is an ever increasing desire to harness the benefits of different materials within a single component and, hence, the importance of dissimilar material joints. In aerospace and automotive industries, it is believed that the ability to weld different aluminium alloys improves the overall efficiency of a structure. Fusion welding of high strength aluminium alloys suffers from problems such as tendency to crack during welding, softening of heat affected zone (HAZ) and its increase in corrosion susceptibility. 1 Solid stage joining techniques like friction welding were demonstrated to have overcome these problems to a large extent.
In the current work, the quality of joints made of similar and dissimilar aluminium alloys has been investigated. The two aluminium alloys selected are Al–Cu (AA 2024) and Al–Mg–Si (AA 6061). Dissimilar materials joining can fall within metallurgically compatible and non-compatible systems. The joining of AA 2024 to AA 6061 alloys belongs to the former category. In metallurgically compatible system, although there is no problem of hard compounds formation, the major issue that still needs to be tackled is difference in parent metal strengths. This difference in strengths would have a major bearing on how the parent metals flow, which in turn determines if the rubbing surfaces deform enough to expel the oxide layers, or how the different regions evolve during the welding. For example, it has been shown that surfaces when meticulously prepared yielded better results in cases of dissimilar welds where the parent metals have wide difference in their flow stress. 2
Rotary friction welding (RFW) is one among several types of technologies available under friction group. In RFW, a full weld cycle encompasses two main stages: friction stage and upset stage. In friction stage, one of the rods to be joined is rotated at a particular angular velocity (spindle speed), while the other is held stationary in a vice. Slowly, these two rods are brought in contact under pressure (friction pressure). This causes frictional heat to develop at the interface, which makes the rubbing end of the rods to become plastic and start to extrude as flash. This proceeds until a preset loss of length in both the rods is reached (burn-off length), whereafter, the rotation is stopped within a very short duration by applying brakes. The two rods would then be held in this position for some time under a different pressure (upset pressure), usually greater than the friction pressure, for some time (upset time). This last stage where the two hot interfaces are held stationary is called upset stage.
The welding parameter settings affect the final quality of a friction joint. Friction pressure and spindle speed jointly affect the rate of friction heat generation at the interface. The burn-off length determines the total heat input to the system ensuring that the heat is sufficient enough for bonding. The upset pressure and upset time help in the final consolidation of bonding. Higher friction pressure proved beneficial while joining of AA 6061 alloy by reducing the softened regions adjacent to the weld line.3,4 Longer friction times, which technically equate to higher burn-off lengths, raised the joint strength in AA 7075 welds only up to a certain limit, after which there was stabilisation. 5 Simulation studies reported higher friction pressures lead to larger axial shortening whereas longer friction times lead to linear rate in axial shortening. 6 Both friction pressure and friction time needed to be ideal for a strong joint for particulate reinforced AA 7005 alloy. 7 Upset pressure was another factor that can affect the joint strengths. 8
It remains always a challenge to find the best set of welding parameters for different materials combination, different specimen cross-sections and sometimes even different makers of machines. When the change in parameters has direct effect on the softened zone formation, like in age hardenable aluminium alloys, the importance of these parameters becomes all the more important. The aim of the current study is to observe how burn-off length when set at different levels affects similar and dissimilar weld systems.
Experimental
Welding parameters
Fully computer controlled 200 kN continuous drive friction welding machine (ETA Technologies, Bangalore) was utilised in the present study. The machine had provision for individual spindle speed settings (up to 2500 rev min−1), friction pressure, burn-off length, upset pressure and upsetting time.
A wide range of machine settings were used in initial trials. This was to determine a suitable window within which further experiments can be conducted, a well defined flash being the criterion. In final experiments, all weld parameters were maintained constant, except burn-off length (Table 1). Three replications were made for each setting.
Friction welding parameters
Parent metals
The chemical composition and certain other properties of parent metals are given in Table 2. The rod dimensions are shown in Fig. 1. Rubbing surfaces were ground and cleaned with acetone just before welding.
Dimensions of parent metal rods and regions of interest
Chemical composition (wt-), mechanical and thermal properties of parent materials
Characterisation techniques
Tensile tests were conducted according to ASTM E8 standard on a 40 ton FIE UTM machine. Vickers microhardness was measured across the interface at a load of 100 g.
To investigate the different microstructures in the welding zone, samples were cut in half across the interface to reveal regions of interest: half and quarter positions (Fig. 1). Modified Keller's reagent was used as etchant, composition being 5 mL HNO3, 5 mL HCl, 5 mL HF and 185 mL of water. The etching time was 20 s for AA 2024 and 30 s for AA 6061. In case of dissimilar welds, AA 2024 was etched first, repolished, and then AA 6061 was etched to reveal the microstructure. Leica-500 was used to observe optical microstructures and Hitachi S-3000 H model for SEM studies.
Results and discussions
Deformation behaviour
With increasing burn-off length, the amount of material extruding as flash was also higher. This was observed commonly in all welds (Fig. 2). The geometry or shape of the flash, however, was different. In AA 6061 welds, the flash is seen curling more than AA 2024. The flash of AA 6061 also had more corrugations. These characteristic features of flash repeated in dissimilar welds.
Cross-section of welds showing amount of flash for different burn-off lengths (Top row: 1 mm; Bottom row: 5 mm)
Higher amount of flash for larger values of burn-off length is because of high heat input into the system. 10 The volume and shape of flash depend mainly on flow stress property of a material with respect to temperature. Thermal cycle during friction stage and subsequent intensity of pressure in forge stage determine the flash. Out of the two alloys, AA 2024 is stronger than AA 6061 at room temperature (Table 1). The fact that AA 2024 extruded less as seen in Fig. 2 confirms that AA 2024 maintained its strength throughout the weld cycle.
Thermal properties play a role in deciding which metal deforms preferentially in dissimilar alloy weld. For example, if the thermal diffusivity of one alloy is significantly lesser than the other, then heat would pile up with in a shorter length of that side of the joint. In such a scenario, the metal with lower thermal diffusivity will deform more, even if its strength were higher to start with. Whereas in other cases, even though the parent material properties are almost same at room temperature, one material may begin to lose the strength faster with raise in temperature, thus giving rise to higher amount of flash. 11 In the present case, it can be reasonably assumed that the temperature distribution within both the parent metals is more or less similar. This is because thermal properties of both the parent metals are close to each other (Table 2). Thus, with almost similar thermal history, AA 6061 curled more as flash than AA 2024.
Corrugations were seen in AA 6061 flash on the outer surface of the flash. The absence of such a feature in AA 2024 suggests that the temperature regime experienced by AA 6061 was more than sufficient to make AA 6061 plastically free flowing.
Another notable observation is that the amount of flash from AA 2024 in dissimilar weld was relatively smaller than in a similar weld (Fig. 2). With thermal properties being almost the same for AA 2024 and AA 6061 (Table 2), the difference in the flash amounts in AA 2024 may have to do with the differences in its flow behaviour in similar and dissimilar systems.
From Fig. 3, it can be observed that the total loss of length has been always higher for AA 6061 welds than AA 2024 welds for the same burn-off length. This is because of the softer nature of AA 6061. The total loss in length for dissimilar welds fell in between similar welds in all cases except for 5 mm burn-off length. As would be shown in the section on ‘Fractography’, the bonding across the whole cross-section was almost complete at 5 mm burn-off length. At lower burn-offs, the softer metal (AA 6061) would have continuously lost its length in both friction and upset stages even as the bonding area slowly progressed. After full cross-section bonding was attained, any extra burn-off length setting would only result in ‘empty’ deformation without any further benefit of bonding. This may be perceived as the unstable point after which the loss in length is no longer linear to the magnitude of burn-off length. In addition, when there is no more rubbing and bonding taking place at the interface, the region contributing to the flash could have shifted towards the interior, away from the centre. This in turn can make the whole flash forming process more sluggish. Since no such abrupt changes happened in similar welds, the loss in length more or less kept up with burn-off length.
Total loss in length of rods after welding for various settings of burn-off length
Interface microstructures
Microstructures observed at the interface revealed very fine grains. Fine grains at the interface are a result of dynamic recrystallisation. During friction welding, combined effects of high temperatures and stress result in dynamic recrystallisation. 12 This fine grained region is referred to as dynamically recrystallised zone (DRZ).
In AA 2024 welds (Fig. 4), the width of DRZ varied: it is thinner at quarter position than at half position. Generally, quarter position experiences higher amounts of heat and strain due to higher relative velocity compared to the centre. The centre, where the relative velocity is almost zero, can receive heat from adjacent regions by conduction only. Thus, it is expected that DRZ is wider at the quarter than at the half position, which was true in the case of AA 6061 welds (Fig. 5).
Dynamically recrystallised zone width variations at interface in AA 2024 welds
Dynamically recrystallised zone width variations at interface in AA 6061 welds
For 1 mm burn-off length, a double cone shape of the DRZ is seen at half position (Fig. 4), similar to what was reported in another study. 13 The ‘pinched’ feature of DRZ at the centre suggests that the DRZ formation was initiated away from the centre and only afterwards started to advance to the centre. This observation thus far is in agreement with the theory, as discussed in the preceding paragraph, that the DRZ forms at the quarter position first. However, as the burn-off length was raised, the greater amounts of extruding flash would have started to modify the DRZ width. The modification may not be uniform at quarter and half positions, though, since the flow patterns are different. The extruding flash has a relatively higher chance of shearing off more metal at the quarter than from the deeply embedded half region. In other words, the hotter metal at the periphery would have clearer route to extrude out as a flash when compared to the ‘locked in’ metal at the centre. Even though the DRZ was thicker at the quarter position, to begin with, the extruding metal would have easily sheared it away in to flash. This could be the same reason why, overall, the DRZ became thinner at all the positions when burn-off length increased, even as the rate of thinning of quarter position was greater. This conjecture is reinforced by the fact that there was heavy bending of the grains at the quarter position compared to half position.
In dissimilar welds, the DRZ is thicker at quarter position (Fig. 6). Hence, it can be said that AA 2024 displayed a contrasting behaviour when welded to itself and when it was welded to AA 6061. This suggests that the resulting microstructure after welding differs in a similar and dissimilar weld system. That AA 6061 behaved in a similar fashion in both similar and dissimilar welds may indicate that dual behaviour is possible only by the stronger of the alloys.
Dynamically recrystallised zone width variations at interface in AA 2024–AA 6061 dissimilar welds
It is to be noted that the amount of flash was always greater in case of AA 6061 than AA 2024 in all conditions (Fig. 2). Hence, the phenomenon of thinner DRZ at the quarter due to shearing of extruding metal should have been seen in AA 6061 rather than in AA 2024. The reason could well have to do from which part of the rod is the flash fed.
Figure 7 shows a typical spread of yield strength (YS) of material from the interface. Yield strength is lowest exactly at the interface as the temperature is highest here during welding. Gradually, the YS increases and reaches the strength of the parent metal unaffected by the temperature. The strength of AA 2024 is higher than that of AA 6061 at room temperature, and it is expected that it remains so, even at higher temperatures. Because of this reason, the AA 2024 curve was shown always above AA 6061. Since, the applied friction pressure used in the present study is constant for all the weld systems, the friction pressure curve cuts AA 2024 closer to the interface (point A) than AA 6061 (point B). Thus, it can be seen that the volume of softened metal that would extrude is lesser in AA 2024. This explains the smaller amount of flash seen in AA 2024 welds.
Variation of yield stress with distance from interface at particular point of time during friction stage
Hardness studies, which would be discussed in following section, revealed that it takes a longer distance from the interface to achieve the parent metal strength in AA 6061. Hence, the AA 6061 slope in Fig. 7 was drawn to be less steep than AA 2024. The soft metal flowing at such great distances makes it less probable to shear away DRZ, i.e. the flow path is more ‘streamlined’ in AA 6061. This phenomenon would explain the untouched wider DRZ at quarter position in AA 6061.
In dissimilar welds (Fig. 6), the interface is almost straight at the half position, whereas it is uneven at the quarter position. The unevenness of interface indicates intermixing of parent metals. Since, the temperatures and strain rates at quarter position were higher, they would have contributed to intensive mixing, which is an indicator of strong bonding. 14
Weld strength
The tensile strengths of AA 2024 and AA 6061 similar welds were ∼80 and 68 of the parent metals respectively (Fig. 8). Weld strength was not significantly affected by the burn-off length. Thus, there is no advantage of increasing burn-off length within the range studied in this present work. In fact, higher burn-off length leads to wastage of parent metal as flash. This result is similar to what was reported in another study, 1 that beyond a certain limit, there is no effect of increasing burn-off length on the joint strength in high strength aluminium alloys.
weld strength for different configurations and burn-off lengths
Nevertheless, the tensile strength of dissimilar welds (AA 2024–AA 6061) increased with burn-off length. Thus, higher burn-off length proved to be beneficial in dissimilar welds than in similar welds.
Fractography
Three kinds of failures are reported in friction welds of high strength aluminium alloys: HAZ, cup and cone, and fracture by shear across the weld plane. 1 In the current study, AA 2024 welds showed HAZ type of failure, AA 6061 welds showed fracture by shear and dissimilar welds showed cup and cone fracture (Fig. 9).
Fracture modes of different welds when burn-off length was 3 mm
The fracture surfaces of AA 2024 looked to be not much different from each other with various burn-off lengths (Fig. 10). This is in line with the observation that there is no difference in tensile strength even. On the contrary, although the strength again did not alter much in AA 6061 welds, the fracture surfaces do not look the same (Fig. 11). With increase in burn-off length, in AA 6061 welds, the tapered area of the fracture was also seen spreading, at the expense of the flat region at the centre. The fracture location photos (Fig. 9) indicate how the crack initiated in HAZ on one side of the joint, travelled parallel to the bond line for some distance and finally ended up in HAZ on the other side. In addition, since it is mostly a HAZ failure in all cases, the strength did not alter much.
Fracture surfaces of AA 2024 welds with different burn-off lengths
Fracture surfaces of AA 6061 welds with different burn-off lengths
Dissimilar welds also displayed almost same fracture behaviour as AA 6061 welds: taper feature on the periphery and flat region at the centre, with the taper region spreading with increasing burn-off length (Fig. 12). However, unlike AA 6601 welds, higher burn-off length enhanced the strength in dissimilar welds. This seemingly contradictory behaviour can be explained by noting how the flat regions appearing in the centre region of the fracture surfaces are not equivalent. The centre region in AA 6061 welds is not exactly flat but has abrasion marks on it, which means some portions of the parent metal from that particular side have been transferred to the other side when the fracture occurred. This hints at the proper rubbing of surfaces and bond formation between the two parent metals. In contrast, the flat region in dissimilar welds shows no marks of rubbing. Thus, most of the strength of the dissimilar welds was contribution from the periphery regions. The centre section just simply debonded.
Fracture surfaces of AA 2024 side in AA 2024–AA 6064 welds with different burn-off lengths
As has been discussed in the previous section, there is more heat generation as well as higher strain rates near to the periphery than at the centre. Hence, the bonding is quicker in the periphery region. When the burn-off length was set low, the heat generated and its subsequent conduction towards the centre of the weld would not have been sufficient enough for the centre region to form a good bond. However, when the burn-off length was increased, the higher heat input in to the system helped spread the bonding area towards the central region.
In the current study, the bonding started from the periphery and then progressed towards the centre in dissimilar welds. In another study involving friction welding of AA 5052, i.e. low carbon steel, 8 the transfer of softer metal AA 5052 initiated at the centre and advanced to the periphery, which is an opposite trend to what is seen in the present study. The starting location of bonding, whether at periphery or centre, depends on friction pressure and spindle speed. 5 At low friction pressures and high spindle speeds, bonding starts at the centre. However, when friction pressure rises, the trend reverses. In the present study, the latter condition, i.e. higher friction pressure, is applicable; hence, the bonding initiated at the periphery. These observations suggest how important it is to have optimum parameter settings when welding dissimilar materials.
The bonding in friction welding is directly influenced by heat flux during the friction stage. The conditions that affect heat flux are friction coefficient, pressure distribution, relative velocity of bonding surfaces and conduction, convection and radiation. 15 It is safe to assume that these were maintained more or less the same in all the weld systems in the current work, except the pressure distribution. This is because the above properties of a metal do not alter much with small alloy additions. Since pressure distribution is the only parameter that could have varied between different weld systems, it can be reasonably argued that this was a consequence of flow properties of the parent metals. The different amount of flash shapes as discussed in the previous section substantiates this argument.
Images (SEM) reveal particles on the fracture surfaces of AA 2024 and AA 6061 welds (Fig. 13). These could be precipitates, a characteristic feature of precipitate strengthened aluminium alloys, that would have acted as initiation sites for the fracture. The shape of these particles in AA 2024 was not round so much as in AA 6061. The particles in AA 6061 were seen deeply embedded inside dimples, whereas in AA 2024, the particles were found in shallower pits. This observation agrees with fracture surfaces seen in Fig. 10, wherein the AA 2024 welds failed in almost a brittle fashion with no visible elongation.
Fractographs (SEM) of AA 2024 (left) and AA 6061 welds; arrows point to some of precipitate locations
The visibility of precipitates on the fracture surface indicates that these have coarsened from the actual finer precipitates. Coarsened precipitates are generally found in HAZ of a weld. Hence, it can be deduced that fracture has taken place in HAZs of both AA 2024 and AA 6061 similar welds.
Figure 14 shows fracture surface of AA 2024 rod at the centre zone after tensile testing of AA 2024–AA 6061 dissimilar welds. When the burn-off length was 1 mm, the features are indicative of a shear failure without any strong bonding between AA 2024 and AA 6061. For higher burn-off length of 5 mm, few areas of AA 6061 are seen adhering on to the underlying AA 2024. That these regions are AA 6061 can be identified from the ductile features they exhibit, a characteristic of AA 6061 fracture.
Fractographs (SEM) showing half position of AA 2024 rods obtained after tensile testing of dissimilar welds
Microhardness
Microhardness carried out at quarter position (Figs. 15–17) in different weld systems suggest a softened zone adjacent to the weld interface, indicative of HAZ. Similar trend in hardness is reported in friction welds of high strength aluminium alloys in few other studies.6,10,16 Since the alloys AA 2024 and AA 6061 are precipitation hardenable, the precipitates that render strength to this alloy would have gone in to solid solution in HAZ due to the heat or sometimes even could have become coarser.
Microhardness across interface at quarter position in AA 2024 welds
Microhardness across interface at quarter position in AA 6061 welds
Microhardness across interface at quarter position in AA 2024–AA 6061 dissimilar welds
The hardness valley in AA 6061 is deeper and wider than in AA 2024.The reason could be disparity in the reprecipitation rates in each of these alloys.
There is a sudden rise in hardness at the interface, appreciable in AA 2024 and limited in AA 6061 sides. The increase in hardness could be due to finer grains or reprecipitation. 12 Another condition that may have contributed to this spike could be accumulation of residual stresses at the interface. 17 The differences in flow properties of parent metals can result in residual stresses.
Higher settings of burn-off length were not seen having a significant effect either on the width or depth of the softened zone. This could be the reason why the strength of similar welds did not change much with different burn-off lengths. It also means that the increase in weld strength with higher burn-off lengths in dissimilar welds was mainly due to increase in bonding area rather than any changes in the softened zone.
In another study, it was concluded that the softened zone played a critical role in determining joint strength of dissimilar welds, 3 whereas in the current study, the softened zone attributes did not change much within the parameter settings. So, it can be said that in dissimilar welds involving aluminium alloys, friction parameters need to be optimised, keeping in view both area of bonding and softened zone formation.
Conclusions
The influence of burn-off length on weld strength of similar and dissimilar welds was investigated. High strength aluminium alloys, i.e. AA 2024 and AA 6061, were considered as base materials. The strength of the similar welds was not affected significantly by burn-off length, and the failure was mainly due to presence of a softened zone beside the interface. However, the burn-off length had a proportional impact on strength in case of dissimilar welds. This was due to the simultaneous increase in bonding area when the burn-off length was raised. The bonded region has grown from periphery to the core. Hence, it can be concluded that in dissimilar welds, the burn-off length should be sufficiently high enough to spread the bonding area up to the axial centre of the rods. Only after this critical burn-off length condition is met, the dissimilar welds would start to be influenced by the softened zone symptoms.
The current study points toward a maximum limit for burn-off length in dissimilar welds, beyond which the strength of joints would remain almost same. This is important from a technological point of view, as such a limit would save costly parent materials from being wasted merely as flash. Although the conclusions of this study can be extended to other metallurgically compatible materials, caution must be exercised for incompatible pairs. Higher burn-off length may in fact lead to formation of hard intermetallics. The current study also indicates the importance of burn-off length parameter would only grow further as the strength mismatch between parent metals widens.