на краю земли

2.3. Material flow and mechanism of joiningRecent experimental and computational works have provided significant insight about More

2.3. Material flow and mechanism of joiningRecent experimental and computational works have provided significant insight about severalinteresting features of materials flow in FSW and the joining process. Most of the material flow occursthrough the retreating side and the transport of the plasticized material behind the tool forms thewelded joint. Three types of flow affects the overall transport of plasticized materials during FSW.First, near the tool, a slug of plasticized material rotates around the tool. This motion is driven bythe rotation of the tool and the resulting friction between the tool and the work-piece. Second, rota-tional motion of the threaded pin tends to push material downward close to the pin which drives anupward motion of an equivalent amount of material somewhat farther away. Finally, there is a relativemotion between the tool and the work-piece. The overall motion of the plasticized material and theformation of the joint results from the simultaneous interaction of these three effects.The plastic flow models for FSW have been used to predict velocities around the tool pin. The veloc-ities have also been estimated from strain rates which, in turn, were obtained from the correlation be-tween grain-size and strain rate [13]. Comparison of the shape of the TMAZ predicted by flow modelswith macrostructural observation has shown satisfactory match [39]. Good agreement between thetorque values obtained using dynamometers and the computed values from the flow models for304L stainless steel [40], 1018 Mn steel [41] and Ti–6Al–4V alloy [72] indicates the usefulness ofthe models to understand the FSW process. Since torque is a measure of the shear stress on the tooland since the shear stress on the work-piece is responsible for both heat generation [38] and plasticflow, validation of the model predictions by experiments indicates that it is appropriate to use themodels for the estimation of several important parameters. Though numerical modelling of plasticflow can aid tool design and the optimisation of weld quality, there does not appear to have beenan application of models towards the prediction of practical processing maps [73]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023The computed stream traces on horizontal planes around the tool pin at three different elevationsare shown in Fig. 5. The stream lines indicate the presence of rotational zone, which clearly shows therecirculating flow of a plug of material around the tool pin. The thickness of the recirculating materialflow region is affected by material properties, welding parameters and rate of heat transfer into thetool. This zone occupies larger areas at higher elevation planes due to greater momentum transportfrom the rotating shoulder. The streamlines indicate that beyond the region of recirculating plasticflow, i.e. in the transition zone, material transfer occurs mainly on the retreating side. Fig. 5 also showsa flow reversal in the advancing side close to the pin, leading to a relatively stagnant zone, whichforms closer to the pin at lower elevations. An important consequence of the lack of adequate materialflow on the advancing side has been related to the formation of ‘‘wormhole” defects [74]. The stream-lines show that beyond the rotational zone the material transport occurs mainly along the retreatingside. Flow visualisation using tracers also indicates the presence of a zone where the material rotatesand advances with the tool and transitional zone where the materials move on the retreating side [75–77].The maximum velocity is attained near the edge of the shoulder at the top surface of the work-piece followed by a rapid decrease away from this region [39–41], as shown in Fig. 6. At planes nearthe bottom of the work-piece, the peaks in velocity were attained near the tool surface. The computedviscosity contours at different horizontal planes (Fig. 6) shows that the viscosity lies in the range of 105to 5  106 Pa s for FSW of aluminium alloys. It is also seen that no significant flow occurs when theFig. 5. Stream traces on different horizontal planes (a) 0.35 mm, (b) 1.59 mm and (c) 2.28 mm below the top surface for a 304stainless steel plate of thickness 3.18 mm. Welding speed 4 mm sÀ1 with tool rotation at 300 rpm [40]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Fig. 6. Spatial variation of viscosity and velocity in AA 6061-T6 [39] at planes corresponding to z = 1.27, 4.67, 8.07 and11.47 mm for a plate thickness of 12.7 mm. Distances in x- and y-direction were equivalent, but that in the z-direction wasincreased eight-fold to enhance clarity. The welding velocity was 1.59 mm sÀ1 and the rotational speed was 637 rpm.viscosity is higher than about 5  106 Pa s and that the region of plastic flow decreases with depth. Itwas observed that plastic flow ceases beyond a certain critical value of viscosity. This cut-off viscositysurface defines the geometry of the thermomechanically affected zone [39].The flow during FSW differs in character from material flow in the liquid state during conventionalfusion welding. This is evident from studies of the welding of dissimilar metals. During fusion welding,the weld pool becomes compositionally homogenous after solidification. However, during FSW of dis-similar metals, mixing does not occur in atomic scale, and it is possible to find larger concentrationdifferences in the weld metal and the region is far from homogenous. For example, Fig. 7 shows a dif-fusion couple formation between Fe and Ni at a certain location in stir region, when pure Fe and Niplates are joined together using FSW. The diffusion couple has a length scale of only 2–3 lm. Such dif-fusion couples would not exist had melting and homogenisation of weld metal occurred.During FSW, the material near the top of the work-piece is stirred under the action of the shoulderand vertical component of motion occurs mainly due to threading on the tool pin. The stirred materialfrom the top is carried down by the threads and deposited in the weld nugget. The vertical mixing be-comes prominent at low weld pitch, which is the ratio of welding speed to rotational speed. One way toFig. 7. Concentration profiles of Fe and Ni at a location in stir region for FSW of pure Fe and Ni. Points represent data obtainedby atomic emission spectroscopy and the solid lines indicate computed results [78]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023understand material flow experimentally is to use inert markers before starting the weld [77], andthen determine their final positions using serial sectioning parallel to the top surface. The tracer tech-nique does not reveal the actual flow path of the material but it shows the final position of the tracer.Schmidt et al. [75] estimated the average velocity of material flow through shear layers during FSW ofaluminium alloy based on experimental investigation of tracer flow. They estimated the averagevelocity to be approximately 0.1–0.3 times the shoulder rotational speed. The order of magnitudeof the material velocity is comparable with the reported values from computational flow models [39].An interesting experiment involved the introduction of copper markers parallel and perpendicularto the weld centreline, in both real experiments and computational models during FSW of AA 7075-T6[79]. The positions of the markers after welding were then compared for the two cases, enabling somequalitative conclusions to be made about the material flow as a function of position, and on the natureof the material bonding that takes place in the advancing side. It was concluded that the weld pitch(welding velocity/rotational speed) and the tool pin shape determined whether the flow was simpleor complex in nature. When pitch was small, markers were continuous rather than dispersed, indicat-ing better welds [80]. A conical tool pin led to more effective material flow with fewer defects. This isanalogous to chip formation in the machining processes – just as chip formation morphology changesfrom discontinuous to segmented to wavy to continuous with increase in cutting speed, the flow pat-tern in FSW also changes with weld pitch [79] as shown in Fig. 8.Steel balls have been used to trace flow in butt welds of 6061 and 7075 aluminium alloys [82], theirpositions after welding being determined using radiography. Several initial configurations of ballswere investigated and the ‘‘stop action technique” was used by suddenly interrupting the forward mo-tion of the pin which is then quickly unscrewed from the work-piece, leaving the threading in the key-hole intact. When the trailing edge of the keyhole was examined using metallography, it showedmicrostructural banding of the extruded material. The bottom portion indicated vertical striations,whereas horizontal striations were observed close to the shoulder indicating downward and rotationalflows, respectively.Fig. 8. Analogy between chip morphology in machining and material flow in FSW [79]. Mid-plane horizontal sections of AA7075-T6 show the re-distribution of copper about the center-line after welding when copper foil is placed between the plates tobe welded. Tool was made of C40 steel, with shoulder diameter of 12 mm. The cylindrical pin had a dimeter of 3.0 mm while theconical pin had a major diameter equal to 5.00 mm and minor diameter of 1.90 mm and both were 2.70 mm in height (for theconical pin, this corresponds to cone angle 30). R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Some idea about the flow of materials can be obtained when microscopic observations are com-bined with knowledge of the crystallographic texture and microtexture [83]. Far from the tool, theundeformed grains have low-angle boundaries while near the tool, elongated subgrains form becauseof the rotation induced by the tool. These subgrains eventually develop high-angle (P15) grainboundaries [84]. Orientation imaging microscopy can provide shear texture maps which, in turn, givesan idea of the strength of flow.Macrostructural observations of transverse cross-sections of FS-welded aluminium alloy speci-mens reveals ‘‘onion-ring” shaped structures of the type illustrated in Fig. 9. These have been linkedto the nature of material flow during FSW [85,86,81], but the detailed mechanism of pattern formationis not completely understood. The pattern has been attributed to the geometry of the extrusion ofcylindrical sheets of material during each rotation of the tool [86]. Measurement of the spacing ofthe markings has been found to be equal to the weld pitch, i.e., the length the tool moves forwardin one rotation. Electron back scattered diffraction studies indicate that the bands have different den-sities of second-phase particles, rather than any significant difference in the local grain structure ortexture [87,88]. This is confirmed by studies of crack propagation in FS-welds [89], where it is ob-served that cracks deflect around the onion-ring structure, consistent with local variations in hard-ness. In AA 2024-T351 rolled sheet, segregated banded microstructure consisting of alternatinghard particle-rich and hard particle-poor regions form with the crack path tending to follow regionsof high particle density [90].In FSW of dissimilar aluminium alloys of cast A356 and wrought 6061, onion-ring patterns consistof lamellar mixture of the two alloys of equal width [91]. When AA 2024 and AA 6061 are welded, theintercalated flow pattern can be visualized by differential etching of the AA 2024 producing contrastrelative to AA 6061 [92–94]. Apart from welding parameters, the base metal microstructure also af-fects the rings. For example, direct-chill casting of AA 5182 produces a dense and uniformly distrib-uted intermetallic particle structure while during strip casting intermetallics are broken-up duringhot rolling and distributed along the primary grain boundaries [95]. Strip-cast base material, withintermetallic particles present in bands, show a definite onion-ring structure in the weld. In contrast,the diffuse rings appear in the weld when the starting material is homogeneous [96].As a result of recent research, several important features of material flow during FSW are now wellunderstood. The flow path of the plasticized materials across both the advancing and the retreatingsides of the tool and the process of joining is now better understood. However, recent research has alsouncovered several complexities and special features of the material flow in FSW such as the lack ofmixing of the plasticized materials in atomic scale and the formation of certain interesting patternsthat still require further work.Fig. 9. Optical image showing the macroscopic features [81] in a transverse section of a friction-stir welded 2195-T81 Al–Li–Cualloy. Note the onion-ring and the adjacent large upward movement of material. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–10232.4. Temperature fields and cooling ratesThe welding variables and the material properties affect the temperature profiles, cooling rates,microstructure and the resulting properties of the welded joint. The temperature fields in FSW exhibitcertain special features. The peak temperatures are significantly lower than those attained in conven-tional fusion welding processes. Furthermore, diffuse heat source in the FSW process and relativelylow welding speed often results in low cooling rate in FSW.Thermal cycles in FSW can be measured using thermocouples (Fig. 10) [97] and in situ neutron-dif-fraction (Fig. 11) [98] where measured lattice-distortion is related to change in temperature. Whenexperimental data are not available, a recourse is to estimate the temperature profiles and the coolingrates using a computational model after the model has been properly validated with experiments.Fig. 10. 3D plot of thermal cycles at the bottom surface along the weld centerline of AA 6061-T6 [97]. The thermal cycle at thefront is for the location where tool pin is inserted into the work-piece while other thermal cycles correspond to locations alongthe weld centreline in the direction of tool movement.Fig. 11. Instantaneous temperature distribution, measured using neutron-diffraction, on the top surface along the weldcenterline behind the the tool pin, for FSW of AA 6061-T6 [98]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Asymmetry in the temperature field is also a special feature of FSW. For example, in welds per-formed on 304L stainless steel, the experimental and the computed peak temperatures were about100 K higher on the advancing side than the retreating side, as shown in Fig. 12 [4]. This large temper-ature difference in the two sides is higher than the values typically observed in the FSW of other mate-rials under most welding conditions. However, the asymmetry in temperature can be appreciatedowing to very low thermal conductivity of stainless steel compared to most other alloys.Some of the early work on heat transfer during FSW was based on simple analytical or numericalheat conduction models that neglected convective heat transfer due to motion of the plasticised mate-rial. Models which neglect convective heat transfer are limited to materials with high thermal conduc-tivity (aluminum alloys) or low Péclet number, Pe, the latter defined as Pe ¼ qC pucLk ð19Þ where uc is the characteristic velocity, and L is the characteristic length. The Péclet number indicatesthe relative importance of heat transfer by convection to that by conduction. When Pe is much lowerthan one, heat is transported mainly by conduction. Under these conditions, the errors in the calcula-tion of temperature fields would not be significant if convection were ignored. Let us consider a typicalFSW of an aluminium alloy. Considering approximate values of q, C p and k as 2700 kg mÀ3,3:77 J kg KÀ1 and 7:02 J mÀ1 sÀ1 KÀ1, respectively, and assuming L to be the average length of stirzone (6 mm) and u ’ 150 mm sÀ1 , the Péclet number is 12. Therefore, even for a high thermal conduc-tivity material such as aluminium, convective heat transfer is an important mechanism for heat trans-fer near the tool during FSW. The typical order of magnitude of the Péclet number for steel is 100. Forsuch high values of Pe, the temperature profiles cannot be accurately calculated ignoring convectiveheat transfer. When convective heat transfer is ignored, the peak temperature and the size of thedeformation zone are overestimated and the flow stress and the tool force are underestimated[65,43]. Furthermore, since the two-dimensional models [74,99] do not consider vertical mixing dur-ing FSW, their application in high Péclet systems can lead to overprediction of local temperature.Plastic flow in FSW is quite complex due to an interplay of variation in strain rates and flow stressleading to variation in viscosity which affects the flow. However, comprehensive three-dimensionalheat transfer and material flow calculations considering spatially variable thermophysical propertiesand non-Newtonian viscosity have been reported for the FSW of aluminium alloys, steels and Ti–6Al–4V [39–41]. It was found that the typical peak strain rate during FSW reaches about 100 sÀ1 atFig. 12. Peak temperature distribution along the intersection of horizontal midplane and transverse plane through the tool-axis,for the FSW of 304L stainless steel using a tool rotational speed of 600 rpm and a welding speed of 10.2 mm minÀ1. The toolshoulder and tool pin diameters were 25.4 and 5.6 mm, respectively [4]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023locations where the velocity gradient is highest, such as near the shoulder edge at the work-piece sur-face and at the pin surface at lower elevations. The strain rate drops sharply to about 30 sÀ1 a few mmbelow the top surface. The strain rate decreases rapidly with depth due to a significant decrease invelocities through viscous dissipation. The computed strain rates were comparable with the 20 sÀ1 va-lue estimated based on measured grain-size and calculated peak temperature from a thermal modelreported by Frigaard et al. [9].The computed temperature profiles along the longitudinal and transverse sections through tool-axis and at the top surface of Ti–6Al–4V work-piece are shown in Fig. 13a–c, respectively. The temper-ature profiles on the longitudinal mid-section (Fig. 13a) and on the top surface of the work-piece (Fig.13c) are compressed in front of the tool and expanded behind it. The computed results are consistentwith the fact that heat is supplied rapidly to the cold region of the work-piece ahead of the tool whileheat is transported at a slower rate to material already preheated behind the tool.Peak temperatures in the work-piece are attained close to the edge of the tool shoulder and signif-icant spatial gradients of temperature exist in the vicinity of the tool surfaces. Measurements of tem-peratures close to a rotating tool are difficult for two reasons. The material transport caused by themotion of the tool makes it difficult to focus on a single location; any embedded thermocouple maybe displaced due to plastic flow. The strong temperature gradient near the tool means that a small er-ror in thermocouple location can lead to a large error in temperature.Thermal cycles affect the structure and properties of welded materials. For example, in precipita-tion-strengthened aluminum alloys, the hardness profile depends greatly on the precipitate distribu-tion and only slightly on the grain-size [100,101]. If a peak temperature greater than 675 K is attainedFig. 13. Computed temperature profiles (K) in (a) xz-plane, (b) yz-plane and (c) z-plane, i.e. top surface of 7.2 mm thick work-piece of Ti–6Al–4V alloy welded with translational speed of 1.6 mm sÀ1 and rotational speed of 275 rpm [72]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023at any location, complete dissolution and re-precipitation takes place there. The dissolution andgrowth of the precipitates during the welding thermal cycle may lead to softening. Even when thepeak temperature is lower than 675 K, the density of the strengthening precipitate reduces as the peaktemperature gets closer to 675 K [100].Dimensional analysis, a useful tool for understanding a complex situation, can be used to estimatepeak temperature in the work-piece using available numerically computed and experimentally mea-sured thermal cycles for different alloys [102]. The non-dimensional peak temperature, T Ã, and heatinput, Q à , are given by T à ¼ T P À T aT S À T a and Q à ¼ f r8AxC pkU ð20Þ where T a , and T S are the peak, ambient and solidus temperatures of the work-piece, respectively, frepresents the fraction of heat that is transported into the work-piece, r8 is the yield strength valueat a temperature which is 0.8 times the solidus temperature and A is the surface area of the shoulder.The thermal properties at the mean temperature, defined as ðT S þ T aÞ=2, are used in the calculations.Note that the non-dimensional heat input, Q à , is similar to a recovery factor, or the ratio of actual totheoretical temperature recovery [103] for surface of a body undergoing adiabatic viscous heating: Q ¼ f r8 Ax|fflfflffl{zfflfflffl}actual U 2C p|{z}theoretical ð21Þ A correlation between dimensionless peak temperature, T Ã, and dimensionless heat input, Q à , valid fora range of Q à between 5  103 and 5  105, is obtained from experimentally measured and computedresults (Fig. 14): T à ¼ 0:131 logðQ Ã Þ þ 0:196 ð22Þ In principle, the yield stress of the material drops dramatically as the solidus temperature isapproached. If the yield stress decreases, so does the rate of heat generation, reaching zero in the lim-iting case of melting. The decrease in temperature allows the material to recover it strength, thus per-mitting a steady temperature to be established below the solidus [104]. However, in practice it ispossible that localised melting can occur under non-steady conditions. The unsteady condition is10.80.6T* = 0.131 log (Q*) + 0.19610 0.40.2 standard deviation = 0.055linear fit6061 aluminum alloy1018 C-Mn steel304 stainless steelreported-aluminum alloyreported-1018 steelreported-304 steel 0 4 4.5 5 5.5 log (Q*)10Fig. 14. Variation of dimensionless peak temperature [102] with non-dimensional heat input. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023possible because heat generation is instantaneous while heat transfer takes time. Incipient meltinghas been reported in the nugget zone at high travel speeds in an aluminium alloy (AA 7100) [105],and locally melted films were observed during the FSW of cast magnesium alloy AZ91 [106].Cooling rate determines the precipitation sequence in age hardenable alloys. It also affects thegrain-size. In fact, friction-stir processing, combined with large cooling rates obtained at higher rota-tional speed, has been found to be useful for grain refinement and has been suggested as a techniquefor producing bulk nano-crystalline materials [107].In high carbon steels, for example S70C (0.72 wt.% C), FSW can result in martensite formation [108].To keep the volume fraction of martensite low, either the peak temperature should be lower than A1critical temperature or the cooling rate should be slower than the critical cooling rate. The peak tem-perature decreases with increase in welding speed or decrease in rotational speed. The cooling ratedecreases with decrease in welding speed and with decrease in peak temperature. Low welding speedcombined with low rotational speed produce slow cooling rate and prevent martensite formation[109].2.5. Welding variablesThe welding speed, the tool rotational speed, the vertical pressure on the tool, the tilt angle of thetool and the tool design are the main independent variables that are used to control the FSW process.The heat generation rate, temperature field, cooling rate, x-direction force, torque, and the power de-pend on these variables. The effects of several of the independent variables on the peak temperaturehave been discussed in the previous section. In short, peak temperature increases with increasingrotational speed and decreases slightly with welding speed. Peak temperature also increases with in-crease in the axial pressure. Fig. 15 shows significant increase in peak temperature with increase inrotational speed.During FSW, the torque depends on several variables such as the applied vertical pressure, tool de-sign, the tilt angle, local shear stress at the tool material interface, the friction coefficient and the ex-tent of slip between the tool and the material. Measured torque values can provide some idea aboutthe average flow stress near the tool and the extent of slip between the tool and the work-piece forcertain conditions of welding, when other variables are kept constant.The torque decreases with increase in the tool rotation speed due to increase in the heat generationrate and temperature when other variables are kept constant. It becomes easier for the material toflow at high temperatures and strain rates. However, torque is not significantly affected by the changein welding speed as shown in Fig. 16. The relative velocity between the tool and the material is influ-enced mainly by the rotational speed. Therefore, the heat generation rate is not significantly affectedby the welding speed. High traverse speeds tend to reduce heat input and temperatures. The torqueFig. 15. Relationship between rotational speed and peak temperature in FS-welds of AA 6063 [110]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Fig. 16. Variation of torque on the tool with weld pitch [111].increases only slightly with the increase in traverse speed because material flow becomes somewhatmore difficult at slightly lower temperatures. The torque on the tool can be used to calculate thepower required from P ¼ xM, where M is the total torque on the tool.Excessive x-direction force can be an important indicator of potential for tool erosion and, in ex-treme cases, tool breakage. Axial pressure also affects the quality of the weld. Very high pressures leadto overheating and thinning of the joint while very low pressures lead to insufficient heating andvoids. Power requirement also increases with the increase in axial pressure.2.6. Tool designTool design influences heat generation, plastic flow, the power required, and the uniformity of thewelded joint. The shoulder generates most of the heat and prevents the plasticized material fromescaping from the work-piece, while both the shoulder and the tool pin affect the material flow. In re-cent years several new features have been introduced in the design of tools. Several tools designed atTWI are shown in Table 2. The Whorl and MX-Triflute have smaller pin volumes than the tools withcylindrical pins [112,113]. The tapered threads in the whorl design induce a vertical component ofvelocity that facilitates plastic flow. The flute in the MX-Triflute also increases the interfacial area be-tween tool and the work-piece, leading to increased heat generation rates, softening and flow of mate-rial. Consequently, more intense stirring reduces both the traversing force for the forward tool motionand the welding torque [112,113].Although cylindrical, Whorl and Triflute designs are suitable for butt welding, they are not useful forlap welding, where excessive thinning of the upper plate can occur together with the trapping ofadherent oxide between the overlapping surfaces. Flared-Triflute and A-skew tools were developedto ensure fragmentation of the interfacial oxide layer and a wider weld than is usual for butt welding[114]. The Flared-Triflute tool is similar to MX-Triflute with an expanded flute, while A-skewTM tool is athreaded tapered tool with its axis inclined to that of the machine spindle. Both of these tools increasethe swept volume relative to that of the pin, thus expanding the stir region and resulting in a widerweld and successful lap joints.Motion due to rotation and translation of the tool induces asymmetry in the material flow andheating across the tool pin. It has been demonstrated that during FSW, material flows primarily onthe retreating side [40,41,74,49,77]. To overcome this problem, TWI devised a new tool, Re-stir, whichapplies periodic reversal of tool rotation. This reversal of rotation eliminates most problems associatedwith inherent asymmetry of conventional FSW.With the exception of FSW with Re-stir tool, material flow is essentially asymmetric about jointinterface. Understanding the asymmetry in material flow is important for optimal tool design. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Table 2A selection of tools designed at TWI [112,16] Tool Cylindrical WhorlTM MX trifluteTM Flared trifluteTM A-skewTM Re-stirTM Schematics Tool pin shape Cylindricalwith threads Tapered withthreads Threaded,tapered with Tri-flute withflute ends Inclinedcylindrical with Tapered withthreads three flutes flared out threads Ratio of pin 1 0.4 0.3 0.3 1 0.4 volume tocylindricalpin volume Swept volume 1.1 1.8 2.6 2.6 Depends on pin 1.8 to pinvolumeratio angle Rotary reversalApplication NoButt welding;fails in lapwelding NoButt weldingwith lowerweldingtorque NoButt weldingwith furtherlower weldingtorque NoLap weldingwith lowerthinning ofupper plate NoLap weldingwith lowerthinning ofupper plate YesWhen minimumasymmetry inweld property isdesired Colegrove and Shercliff [44,47] calculated material flow for the FSW of AA 7075 using computationalfluid dynamics. They compared the mechanical efficiency of a triangular tool with convex surfaces(Trivex) with that of a conventional tool (Triflute) by examining the streamlines around these tools.They suggested that the Triflute tool produced a strong auguring action, thus increasing the downwardforce (Fig. 17).Zhao et al. [115] studied the effect of tool pin design on the weldability and mechanical propertiesof welded 2014 Al plates. Cylindrical and tapered tool pins did not ensure effective mixing in the ver-tical direction leading to wormholes at the base of the TMAZ. However, when tapered tools withthreads were used, defect free welds were obtained. Other studies have also confirmed that tools withscrew threads generate more heat and improve flow of the softer material by exerting a downwardforce. Since the material flows mainly on the retreating side, insufficient plasticity and material flowresults in a ‘‘wormhole” on the advancing side [36,76]. This effect becomes more prominent at lowtemperatures due to sluggish flow of materials. The choice of pin angle, which is the angle betweenthe conical surface of the pin and its axis, is another important parameter which influences theFSW process; increasing the angle leads to a more uniform temperature distribution along the verticaldirection, which helps in reducing distortion [68]. An angle of ’40 is thought to be optimum for 7xxxaluminium alloys [68].Design of tools based on a quantitative understanding of material flow is just beginning. Cole-grove and Shercliff [44,47] used a thermal model based on FLUENT to design a tool to minimisethe traversing force during FSW of aluminium alloy. They examined several tool geometries includ-ing Trivex, which is triangular in shape with convex surfaces, and MX-Trivex which had similar shapewith threads. They observed that the traversing force and downward forces were considerably lowerfor the Trivex tools relative to those for Triflute, especially at lower applied shear stress where con-siderable slip occurs between the tool and the work-piece. They suggested that Trivex with its con-vex surfaces avoids sticking to the material that reduces the shear force at the tool–metal interfaceand consequently reduces the traversing force. Triflute, on the other hand, has features that impedeflow and the tool sticks to the material even at low applied shear stress. The entrapped material inthe tool leads to a large shearing effect causing correspondingly greater tool forces. Interestingly, R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Fig. 17. Streamlines for isothermal model [47] that used a limiting shear stress of 40 MPa: (a) Triflute tool, (b) single streamlinefor Triflute tool showing vertical movement and (c) Trivex tool. Tool rotation 457 rpm and translation at 457 mm minÀ1.they [44,47] observed no change in terms of heat input or power requirement by different tooldesign.The quantitative prediction of the forces that the tool experiences is particularly useful in designingtools for hard materials such as steel. Such predictions should ideally include a consideration of plasticflow in the thermal model [39–41]. The errors resulting from ignoring plastic flow for high thermalconductivity alloys are smaller than for low thermal conductivity materials such as steel. The errorsare greatest near the tool where an accurate knowledge of temperature is important for the smoothoperation of the FSW process.Tool wear is an important concern for the FSW process, particularly for high temperature harderalloys. Titanium, which has a melting point of 1941 K, has been successfully friction-stir weldedusing sintered TiC welding tools, which require a water cooling arrangement to extract excessiveheat from the tool. Mandal et al. [116] proposed a technique to reduce tool wear. Heat sources wereintroduced in front of the tool to preheat the work-piece, creating a hot channel for the tool to movein, thus reducing the tool wear. The model was tested using a modified Rosenthal equation for acombination of three heat sources, one corresponding to the tool and two for the preheat aheadof the tool. Though the concept is novel, the computational modelling ignored material flow, thewear on the tool was not quantified and the conclusion about reduction in wear was intuitive. Thereis a need for the development of reliable models for the wear of the FSW tools. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–10232.7. DefectsCommon defects in friction-stir welds include porosity and surface defects. At a constant rotationalspeed, an increase in the travel speed leads to wormhole initiation near the bottom of the weld. ***-thermore, the size of the wormholes increases with the travel speed [117] because of inadequatematerial flow towards the bottom of the weld. There are indications that the travel speed to rotationalspeed ratio is an important variable in the formation of the wormhole defect [118]. For the same mate-rial and tool geometry, a high ratio tends to favour the formation of wormhole defects [119].Most of the heat generation occurs at the interface between the tool shoulder and the work-piece.Significant heterogeneity in heat generation at that interface can lead to defect formation in the formof excess flash due to surface overheating [117,120].The propensity for cracks or voids increases with the welding speed although there is an alloy-dependence [121]. For example, defects dominated in AA 5083-O and AA 2024-T3 but not in AA6063-T6 in which there is a significant drop in hardness within the TMAZ. The defects tend to occuron the advancing side where an abrupt microstructural transition occurs from the highly refined nug-get zone to the TMAZ while the transition was gradual on the relatively defect free retreating side.Cast Al–12 wt.% Si alloys contain coarse particles of silicon which can be used to reveal the flow anddefect evolution during FSW [122]. Insufficient heating causes the brittle Si particles to crumble sothat the observation of fine Si particles after FSW is indicative of defect formation due to limited mate-rial flow. Therefore, it is not surprising that finer particles were detected on the advancing side nearthe bottom of the tool pin where inadequate flow is associated with void formation.It has been suggested that the force on the tool in the x-direction, which can be measured using adynamometer, could be used to predict defect formation in FSW on the assumption that large forcesindicate sluggish flow [123]. Maps showing the fraction of stirred material which is at a given flowstress and temperature, with superimposed strain rate contours (Fig. 18) can also indicate the ten-dency for defect generation [124]. Referring to Fig. 18, at 80 rpm a greater amount of material is beingdeformed at a higher strain rate with both the Triflat and Trivex tools than with the cylindrical tool. At200 rpm, greater amount of material was deformed at moderate temperatures (725–735 K) and strainrates (1–50 sÀ1) with the Trivex tool which shows largest deformation domain. It was concluded [124]that optimum welds correspond to maps with large regions of moderate to high strain rates and tem-peratures 30–50 K below the solidus temperature. Higher temperatures will lead to softening of thematerial and reduce the size of the deformation region.Tool design and welding variables affect materials flow patterns. However, no specific character ofthe material flow has been related with the porosity formation and no unified mechanism of porosityformation exists.Elangovan et al. [125] examined the effects of rotational speed and tool pin design on defect for-mation in friction-stir processing of AA 2219. Five pin profiles (straight cylindrical, tapered cylindrical,threaded cylindrical, triangular and square) were used to fabricate joints at various tool rotationalspeeds. The square tool pin profile produced the least defect content in the weld as the flat faces pro-duced a pulsating action which led to more effective stirring. Also, a square tool has higher eccentricitywhich is defined as the ratio of the dynamic volume swept by the tool to the static volume of the tool.For example, a square tool has eccentricity of pd =4 : d =2 ¼ p=2 % 1:57 where d is the diagonal of thesquare.In FSW lap joints of AA5083 and SS400, the size of the voids formed along the weld interface in-creases with increase in diameter of the tool pin and tool tilt angle [126]. Large diameter pin produces(>5 mm) more heat and forms intermetallic compound FeAl3 and Fe2Al5 instead of FeAl formed at low-er temperature. Aluminum rich FeAl3 and Fe2Al5 are harder and more brittle than FeAl and they reducejoint shear strength. Increasing the tool tilt angle (>1) also increases the heat generation rate, formsaluminum rich precipitates and decreases joint strength [126].One advantage of the FEM model [36] is its ability to predict void formation because the arbitrary-Langrangian–Eulerian formulation allows for large material deformation and for the grid to track thematerial so that separation can occur between the work-piece and tool. Fig. 19 shows void formationat the lower advancing side, near the trailing edge of the pin/work-piece interface, due to incompletedeposition of plastic material. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Fig. 18. Material condition maps for (a) Triflat, (b) Trivex and (c) cylindrical tool pin at 80 and 200 rpm [124].Fig. 19. Void formation at the lower advancing side due to incomplete filling [36] modeled using ALE formulation of FEM.Temperature contours are shown in C. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–10232.8. Residual stressThe presence of residual stress in a weld plate affects its distortion behaviour and ability to sustainapplied loads while maintaining structural integrity [127–129]. While compressive stresses can insome circumstances be beneficial [130], tensile stresses can cause crack initiation and aid its propa-gation leading to catastrophic failure.Residual stresses in the weld can be measured by using: (1) diffraction studies using X-ray or neu-tron sources without destroying the weld and (2) destructive hole-drilling methods [127,128,131].Since diffraction is expensive and conventional hole-drilling is only suited to uniform plane stressaround the hole, Ya et al. [132] used Moiré interferometry incremental-hole-drilling method to assessresidual stress in a friction-stir weld. Synchrotron diffraction has also been used to measure residualstress in such welds in the context of fatigue behaviour [133].Using neutron-diffraction, Peel et al. [134] showed that longitudinal stress increases as traversespeed increases due to steeper thermal gradients and reduced time for stress-relaxation. Fig. 20 showsthat transverse stresses do not display a direct dependence on the rotational speed. It also shows thatthe weld zone is under tension and the work-piece material is under compression.As in ordinary welds, residual stresses develop in constrained assemblies during FSW due to expan-sion during heating and contraction during cooling; a feature unique to FSW is the additional stresscaused by the rotational and translational components of the tool so that the welding parametersof FSW must affect the final state of stress [135]. The stirring action of the tool is believed to relievesome of the stresses within the thermomechanically affected zone [136]. A finite element analysis ofclamped FS-welded AA 6061 samples has shown that lateral and longitudinal residual stressesdecrease with an increase in the rotational speed [135]. A higher welding speed enhances theFig. 20. Measured (a) longitudinal x-direction and (b) the lateral y-direction stresses in AA5083 welds at different weldingspeeds across a transverse section [134]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023longitudinal residual stress but reduces it along the lateral direction. The analysis also showed that themaximum temperature gradients in the sample are located just beyond the edge of the tool shoulder.As might be expected, the residual stress distribution is dramatically altered on unclamping the sam-ples after friction-stir welding and this must be taken into account in any modeling effort.The role of plasticity during the friction-stir welding process is known to be important in the cal-culation of residual stress; a large overestimation of the magnitudes of the residual stress [136] canresult when this effect is ignored. Nevertheless, the outcome that the longitudinal stress along theweld centreline is tensile and larger than all the other component stresses is believed to be correct[138]. Fig. 21 shows some detail at a point where the FSW tool is half way along the length [137];the longitudinal stress is lower in front of the tool (marked by A) as it has not been affected by thermalstress or by structural loading. Behind the tool, marked by D, compressive stresses exist. Transversestress shows similar trend though it is spread over a very larger area behind the tool compared to thatin longitudinal stress as the end of work-piece right behind the tool is not constrained leading to freethermal expansion. The stress in the vertical direction is negligible along the edges of the work-piece(region B) while it is compressive in the remaining regions.Fig. 21. Normal stresses in MPa in (a) the longitudinal x-direction, (b) the lateral y-direction and (c) the vertical z-directionwhen the tool is mid-way between the two ends of AA 6061 weld plates [137]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Tensile stresses present in FS-welded samples lead to poor mechanical properties. Applying exter-nal tensioning during welding [139] induces compressive stresses which have the benefit of inhibitingcrack propagation. Models predict that increasing tensioning levels to values higher than 50% of theroom temperature yield stress in AA2024 aluminium alloys, leads to tension in the weld being re-placed by desirable compressive stresses [140]. This is an important finding and can be used by weld-ers as a guiding principle to produce high-quality welds. Not surprisingly, crack propagation rates areknown to correlate strongly with the state of stress [141].There has been systematic work on the resistance to fatigue of friction-stir welds in aluminium andtitanium alloys [142], which demonstrates the key role of residual stresses in controlling crack growthwithin the HAZ along the welding direction [143]. Although the levels of residual stress can be smallerthan in conventional welding scenarios, they have a large effect on near-threshold crack propagationrates. In AA 7050-T7451, fatigue crack propagation in the weld nugget is inferior, while it is superior tothat of base alloy in the HAZ [143]. In HAZ, compressive residual stresses dominate crack growth[144]. For the weld nugget region, microstructure and intergranular failure mechanism dominatecrack growth. In precipitation hardened aluminium alloys the particles in the HAZ are much coarserand less coherent than the base material. This should in general reduce the fatigue crack growth ratein the HAZ relative to the base material, but only if tensile residual stresses are not present within theHAZ [142]. Post-weld solution heat-treatment and aging also restores the strengths of the nugget tothe same level as the base material [83]. In AA 7050 grain boundaries are sensitized during FSW[145], leading to intergranular fracture when fatigued in a 3.5% NaCl solution.3. Microstructure and properties of friction-stir welded alloys3.1. Aluminium alloysMany aluminium alloys are strong by virtue of precipitation hardening through natural or artificialageing from the solution-treated condition. The heat associated with welding changes the microstruc-ture of the material. In Fig. 22, HVmin and HVmax represent the hardness in the solution-treated andprecipitation hardened states. The effect of welding is to cause a drop in hardness from HVmax towardsHVmin as the peak temperature experienced increases, curve (a), Fig. 22. This is because precipitateswill coarsen and reduce in number density in regions remote from the heat source, and will re-entersolution when the peak temperature is sufficiently high [100]. Some re-precipitation may occur duringthe cooling part of the thermal cycle, resulting in a hardness value beyond HVmin, curve (b), Fig. 22. TheFig. 22. A schematic diagram showing dissolution and re-precipitation in age-hardenable aluminium alloys. HV denotes theVickers hardness number. Adapted from [146]. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023ultimate result is the continuous line with a minimum in hardness somewhere in the heat-affectedzone, due to the competing effects of dissolution and re-precipitation [146]. But in contrast to agehardenable AA 6082, where a minimum hardness occurs in the HAZ, FSW of non-hardenable AA5082 results in uniform hardness across the weld [147].This general scenario may be complicated by the effects of deformation in FSW as described for thespecific example of AA 2219. AA 2219 is a copper precipitation-strengthened alloy containing about6.3 wt.% Cu, which because of its strength and toughness at low temperatures, is used for containingliquefied gases for rockets of various kinds. It is frequently supplied in the T87 condition, meaning thatit has been solution-treated, cold-worked (10% reduction in rolling) and artificially aged. It can bewelded using arc processes but this results in a reduction in the cross-weld strength because the proofstrength of the fusion zone decreases to about 140 MPa compared with the 370 MPa of the plate [148].The former can be increased to between 220 and 275 MPa using pulsed GTAW or pulsed electronbeam welding techniques because this promotes finer grains in the fusion zone [148].Friction-stir welding does not seem to have an advantage over arc welding with respect to thestrength of the fusion zone, as illustrated in Fig. 23. Notice that the TMAZ is somewhat softer thanthe fusion zone because the latter dynamically recrystallises into a fine grain structure. It is the coars-ening of the Al2Cu precipitates in the TMAZ that is partly responsible for its softening [150]. Sometransmission electron micrographs across the weld are illustrated in Fig. 24; these show clearly thehuge changes due to the heat from the process. The formation of coarse precipitates at the grainboundaries, and their associated precipitate-free zones, are common detrimental features in themicrostructure.The fact that FSW welds in precipitation hardened alloys lead to a weak zone is not surprising giventhat the majority of strengthening in most strong alloys comes from precipitates [151]. There is someevidence that manipulation of pin profiles and FSW parameters may help improve slightly, the hard-ness in the central region [125] or indeed, in the HAZ [152]. Theoretical work has also been done to seeif cryogenic cooling after the tool pass can help retain alloying elements in solution after the peak tem-perature is experienced, so that the alloy can then naturally age and not develop the coarsened micro-structures typical of slow cooling from the peak temperature [153]. However, computations indicatethat the advantage in so doing is likely to be minimal [153].Fig. 25a shows that the grain-size increases with increase in peak temperature caused by increasein rotational speed. Here grain-size is related to peak temperature by assuming static grain-growth ofdynamically recrystallised grains, during the cooling of the thermal cycle [110]: D2 ¼ D2 þ At eðÀQ=RT p Þ ð23Þ where D0 and D are the initial and recrystallized grain-size, A is a constant, t is the time to cool to448 K, Q is the activation energy for grain-growth, T p is the peak temperature. Assumption of isother-mal condition leads to over-prediction of grain-sizes [110]. Precipitate free zones form near the grainboundaries because grain boundaries act as sinks for nearby dislocations, reducing nucleation sites forFig. 23. Open circles are data from friction-stir weld, plotted relative to the centreline of the FSW for alloy AA2219 [149];similar data have been reported by Paglia and Buchheit [150]. Filled circles from a variety of arc and electron-beam welds [148].In the latter case all values are plotted at zero on the horizontal axis to represent the fusion zone. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Fig. 24. Microstructure variations across aluminium alloy AA2219 joined using friction-stir welding. (a) Unaffected parentmaterial in the precipitation hardened condition; (b) the nugget; (c) the heat-affected zone. Micrographs courtesy of Dr.Christian Paglia.Fig. 25. Relationship between (a) peak temperature and grain-size in AA 6063 [110] and (b) grain-size and volume fraction ofPFZs.precipitates and also as precipitation sites, effectively reducing the solute content around them. Asgrain-size increase, assuming constant width of PFZs, their volume fraction decreases with increasepeak temperature as shown in Fig. 25b [110].Not all alloys of aluminium are precipitation hardened. In the 2000 series alloys, the strength de-pends more on grain-size (d), which has been expressed in terms of the Zener–Holloman parameter[154]: log d ¼ a þ b log Z ð24Þ R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023where a and b are empirical constants based on data from extrusion experiments [155] and the hard-ness is then related to d using a form typical of the Hall–Petch type equation: HV ¼ HV0 þ c= d ð25Þ where c is a constant.In the cast Al–Si alloys, friction-stir welding breaks up the large silicon particles in the nugget andthe TMAZ, Fig. 26; as a consequence, the fracture is located in the base plate during cross-weld tensiletests because in this case, it is the coarse silicon particles which control failure [156,5]. FSW can alsoheal casting defects such as porosity [157].Fig. 26. (a) A section through a friction-stir weld made in an Al–Si casting alloy. There are pores indicated in the base metal(BM). HAZ represents the heat-affected zone, TMAZ the thermomechanically affected zone, and SN the stir nugget. Opticalmicrographs showing the microstructure in (b) the base metal; (c) heat-affected zone; (d) the thermomechanically affectedzone, where considerable refinement of the silicon has occurred. Optical micrographs of regions (e–g) of the stir nugget asidentified in macroscopic section presented above. Micrographs courtesy of Professor H. Fujii. R. Nandan et al. / Progress in Materials Science 53 (2008) 980–1023Corrosion studies indicate that the weld zones produced by friction-stir welding have comparableenvironmentally assisted cracking susceptibility as the unaffected parent [150].
Hide