Friction stir - where we are, and where we're going...

TWI Bulletin, May - June 1998

Wayne is a Principal Research Engineer in the New Technology Unit. He gained his MPhil from Brunel University (Material Technology).

Chris joined TWI 34 years ago and has wide experience in researching and developing welding technologies. His interest in friction welding dates back to the late sixties when he was closely involved in pioneering the first micro-friction welding machine.

Mike Gittos is a Principal Metallurgist in the TWI Materials Department. His main areas of research have related to aspects of joining aluminium, copper, titanium and nickel alloys. He is currently responsible for electronmicroscopy services at TWI, and in 1993, he was awarded the A F Davis silver medal by the American Welding Society.

Dick Andrews, Principal Research Metallurgist, works in TWI's Electron Beam, Friction & Forge Processes Department. He is responsible for the progress of research programmes involving friction welding, friction processing, flash, MIAB and explosive welding.

Friction stir welding (FSW) is a new solid-phase technique invented and patented at TWI ^[1] for the butt and lap welding of ferrous and non-ferrous metals and plastics. Wayne Thomas, Chris Dawes, Mike Gittos and Dick Andrews describe recent developments in FSW, with brief reference to certain process characteristics common to many of the variants within friction technology.

Friction stir welding is a continuous process that involves plunging a portion of a specially shaped rotating tool between the abutting faces of the joint. The relative motion between the tool and the substrate generates frictional heat that creates a plasticised 'third-body' region around the immersed portion of the tool. The contact of the shouldered region of the tool with the workpieces also generates significant frictional heat, as well as preventing plasticised material from being expelled. The tool is moved (relatively) along the joint line, forcing the plasticised material to coalesce behind the tool to form a solid-phase joint.

Friction stir welding has captured the attention of the fabrication industry as a solid-phase joining technique capable of good quality single-sided and double-sided butt, T, and lap joints. Friction stir welding can join a number of materials, including some, which are difficult to weld by conventional fusion processes. Aluminium and its alloys have proved particularly suited to FSW. In this connection it is noted that aluminium alloys are more widely used than any other alloy apart from steel, ranking among the most important of the common metals.

Figure 1 illustrates the salient features of the process which operates by generating frictional heat between a rotating tool (of harder material than the workpiece being welded), to plasticise the abutting weld region. Commonly, the tool is shaped with a large diameter shoulder and a small diameter, specially profiled, probe that makes contact first as it is plunged into the joint region. The components to be welded are secured to prevent the butted joint faces from being forced apart as the probe passes through and along the seam. (For thick plate welding, 25-50mm thick, usually a pilot hole of smaller diameter than the probe is drilled at the start to assist the plunging operation.) The depth of penetration is controlled by the length of the probe below the shoulder of the tool. The initial plunging friction contact heats the adjacent metal around the probe as well (as a small region of material underneath the probe), but once in contact with the top surface of the substrate, the shoulder contributes significant additional heat to the weld region. In addition, the contacting shoulder, which can be profiled to provide improved coupling, prevents highly plasticised material from being expelled from the welding region.

Once the rotating tool is in position, the thermally softened and heat-affected region take up a shape relating to that of the overall tool geometry. The heat-affected region is much wider at the top surface (in contact with the shoulder) and tapers down as the probe diameter reduces. The combined frictional heat from the probe and the shoulder creates a highly plasticised 'third-body' condition around the immersed probe and the adjacent contacting surface of the workpiece top. This highly plasticised material provides for some hydrostatic effect as the rotating tool moves along the joint, which helps the plasticised material to flow around the tool. The plasticised weld material then coalesces behind the tool, as the tool moves away.

Friction stir welding can be regarded as an autogenous keyhole joining technique in which consolidated welds are solid-phase in nature and do not show fusion welding defects. No consumable filler material or edge preparation is normally necessary. The distortion is significantly lower than that caused by fusion welding techniques.

A greater understanding of FSW can be gained from related friction technology, including rotary friction welding, friction surfacing, friction extrusion, friction hydro pillar processing, friction plunge and third-body friction joining.

Although knowledge of friction dates back to antiquity, the use of frictional heat for solid-phase joining and forming techniques has only been extant over the last century.

Friction was first studied around 1500 AD by Leonardo da Vinci, and more specifically by Amontons (1699) who postulated that, the coefficient of dry friction was independent of the contact area, the speed of motion, and the applied load. However, beyond normal limits, Coulomb (1779) found that with increased load the coefficient of friction increased and that it was also dependent on the speed of motion.

However, with respect to friction-welding technology, the coefficient of friction is not in itself relevant as the conventional theory only applies to comparatively light loads with undeformed faces.

In friction joining and forming, the process has more relation to a fluid layer of high viscosity between solid components in relative motion and under significant compressive loading. The thixotropic properties and the fluid flow features that occur in conventional friction welding have been reported.^[2] In some respects, the science of friction processes is probably more closely allied to that of rheology.

For abutting components of similar geometry (cross-sectional area and mass) and like material, an equal contribution of material from each component forms a common plasticised layer or transient 'third body'. This third body, in a series of infinitely thin laminae, averages 50% of the relative component velocity, effectively as a quasi-hydrodynamic intermediate zone. However material is preferentially drawn from components that are dissimilar in geometry (and/or physical material properties) as in friction surfacing which is further augmented by the relative traverse movement. Similar differences can occur in FSW within the transient third-body plasticised region.

Background to FSW

Friction stir welding was first patented by TWI in 1991, and since that time there has been wide spread interest in the technology. A growing number of reports has been published which generally confirm that FSW offers a number of useful technical advantages for an ever increasing range of materials.^[3-34]

The technology surrounding the overall tool geometry, (the probe, and shoulder profiles), and the tool attitude have been reported to be the heart of the FSW technique and a number of tool features have been disclosed.^{[4,8,22,24-26,30]} For example, rotating tools with whisk type probes that allow plasticised material, especially from comparatively soft metals, to flow through the probe and paddle type probes, where plasticised material can flow more easily around the probe, have been investigated.^[22] Reference has also been made to probe tip features, in the form of a V-cut into probe tip, similar in effect to a mini whisk.^[30] The introduction of shoulder profiles such as spiral grooves and shoulder texturing such as pock marking are reported to improve the coupling, between the shoulder and substrate material.

There is asymmetry associated with rotation, as illustrated by the^[5,17,22] advancing and retreating sides shown in Fig.1, which are similar to deposition characteristics of friction surfacing. When process parameters are unsatisfactory, a sub-surface void or even a surface-breaking defect can occur running parallel with the joint^[17] on the advancing side of the tool probe. Probe size and shape, in relation to rubbing velocity, and welding speed, also have a major influence on weld quality.^[7]

Distortion is found to be almost non-existent in FSW which offers a major advantage for the manufacture of most weldments.^[11-15] Mechanical properties claimed include 180° rolled, three-point, and hammer bends without failure and for 2014A T6, 5083 and 7075-T7351 tensile strengths equal to that of the overaged parent metal. Fatigue endurance tests have produced values approximately twice those normally attained for fusion welds.^[11,13,15]

An estimated maximum temperature, which exceeded 400°C and was probably close to 480°C, was also noted for 7075 material.^[20] The effects of FSW on microstructure have been studied ^[6,20,21,23] and for 7075 aluminium were found to be less extreme than those occurring in fusion welding. Also a reduction in the grain size by dynamic recrystallisation, as well as a change from slightly elongated grains found in wrought bar, to that of more equiaxed grains is found in the weld region.^[19] The grain size in the weld nugget area is typically very low; for example, between 2-10 (microns in diameter) for a range of aluminium alloys.^[20,23] In addition, the dislocation density can be significantly reduced in the weld nugget area, in contrast to that in the parent metal.^[20,23] The microstructure developed in the FSW process has similarities with certain features associated with superplastic deformation. The small grain size, a nominally comparable strain rate, and in some cases a similar operating temperature suggest that a greater understanding of FSW can be gained by consideration of superplastic deformation related technology.^[36] Corrosion tests on parent material and FSW welded specimens indicated that FSW as-welded 2024-T3 exhibited less than 6% reduction in strength after exposure whereas for 7075-T6 the reduction was about 1%.^[29]

Operating FSW under water provides increased cooling rate, leading to a smaller heat affected zone and consequent improvement in the properties with some materials. ^[30] The use of fluid and/or air coolants, applied externally, or through internal spaces within the tool, has been claimed to improve traverse rates by 20-100% over that achieved without cooling.^[31]

At the end of a run when the tool is withdrawn a hole is left and the possible need for run-off plates has been noted,^[11] as well as the use of other friction methods to fill in the end. Programmable or adjustable FSW tools to provide retractable or variable length probes have been described.^[25,26] These adjustable tools are designed to allow the probe to penetrate the workpiece, from zero, extending to the depth required as the workpiece is traversed and is gradually withdrawn to zero penetration after the weld is completed. One version changes the shoulder diameter in conjunction with probe depth.^[25]

To ensure complete penetration, or weld root closure for FSW, the bottom corners of the plates to be welded are chamfered, and subsequently filled with plasticised material from the remaining plate material.^[32} In another method a recessed backing plate is used so that plasticised material is extruded to form a small bead on the rear side of the workpiece.^[33]

Although preliminary, a simple heat flow model indicates that the heating during FSW occurs largely at the periphery of the tool shoulder and conducts into the plate for bonding.^[18] A more comprehensive model also agrees that only the shoulder provides significant heat input, the probe's contribution being small.^[34] (The contribution from the shoulder is likely to be thickness related, reducing in relative effect as the plate thickness increases).

Thick Plate Welding

Friction welding equipment

All the friction stir welding trials described here were carried out at TWI on a modified milling machine, designated FW14, that incorporated a fixed 3° tilt away from the direction of travel. Transmission power was provided by a two speed reversing motor, which developed 22kW at 1430 rev/min and 15kW at 960 rev/min, which was found ample for a range of spindle speeds between 200 and 1270 rev/min. The available traverse rate ranged between 0.5 and 15 mm/sec (30 and 900 mm/min). Rotational speed was checked by a hand held optical digital tachometer and traverse speed set using the machine controls and checked with a hand held stopwatch.

Workpiece dimensions were 50mm wide x 500mm long (50mm plate), and 150mm wide x 500mm long (75mm plate), of 6082 aluminium alloy in the T6 condition, and chemical compositions shown in Table 1.

Table 1: Chemical composition of substrate materials, wt% (TWI analysis ref. no. S/97/339)

Tool geometry

Trials were carried out with the Whorl^TM FSW tool shown in Fig.2, which included a scoop shouldered profile with a frustum-shaped probe that incorporated a helical ridge with side flats designed to augur plasticised material downwards. The tool inclination included a combined 3° tilt with the acute angle towards the start of the weld in line with the longitudinal direction of the weld seam, and a 1° tilt towards the retreating side of the weld by tilting the workpiece.

Surface appearance

Typically the surface appearance of FSW is a regular series of part circular ripples, which point towards the start of the weld. These ripples are essentially cycloidal and are produced by the final sweep of the trailing circumferential edge of the shoulder, during traverse. The pitch between the ripples is determined by the rotational speed of the tool and the traverse rate of the workpiece, increasing with the latter. The combined relative motion is by definition a superior trochoid, ie a cycloid with a high degree of overlap in successive revolutions. For aluminium alloys under optimised conditions, the surface colour is normally silvery-white.

Metallographic examination

Figures 3 and 4 show sections of the welds in 6082 T6 condition aluminium alloy, for 50 and 75mm thick plate, where these welds are characterised by well-defined weld nuggets and flow contours, almost spherical in shape. In detail, the contours depend on tool design and process parameters used. For heat treatable materials, a well-defined heat affected zone surrounds the weld nugget region, and extends to the shoulder diameter at the plate surface. For the weld nugget itself, full dynamic re-crystallisation occurs giving a fine equiaxed grain structure of two to four microns in diameter. Typically the parent metal composition is unchanged in the metal region and there is no apparent segregation of alloying elements. A hardness traverse taken from a 50mm thick test weld recorded the following mean values:

Parent metal 100HV_2.5
Weld nugget 65HV_2.5
HAZ region 52HV_2.5.

Fractography

Samples of welded 50mm thick 6082 T6 plate were notched in the parent material and the weld nugget region. After fracturing by bending, the surfaces were examined by scanning electron microscopy. As shown in Fig.5, both the weld nugget and the parent material failed in a ductile manner by microvoid coalescence. However, there was an absence of relatively large microvoids in the weld nugget sample and this may be a consequence of the break up of the primary constituent particles during stir welding.

Mechanical integrity

For 50mm plate, transverse sections were hammer bend-tested to 180°, and for 75mm thick plate transverse sections were three-point bend-tested, see Fig.6. A number of tensile tests were carried out with failure at 175 N/mm 2 typically occurring in the HAZ region. Necking occurs where the hardness is lowest, as shown in Fig. 7. 

Discussion

Research work at TWI aims at evaluating a range of materials and developing a number of tool profiles. These tools are designed to improve the flow of plasticised material around the probe itself, to enable substantially thicker plates to be joined and allow relatively high traverse rates to be achieved.

Tool design

The main probe development variants of the Whorl^TM tool are shown in Fig.2. Essentially the probe is shaped as a frustum that displaces less material than a cylindrical tool of the same diameter as the probe shoulder change in section. Providing a certain minimum probe tip diameter is maintained the frustum shape means less effort is required to traverse through the plasticised material. The frustum also is more uniformly stressed and the shape ensures that the lower surface of the helical ridge provides a clear downward augering force, with less interference from the next ridge below. The core of the probe need not run parallel with the helical ridge nor does the ridge have to be of uniform pitch. In this respect the helical ridge is not a simple external thread which has to engage with an internal thread, but is essentially an auger which is immersed in the super-plasticised medium which the tool creates.

To enable more effective flow of the plasticised material, it is preferred that the distance between each ridge is greater than the thickness of the ridge itself. The Whorl^TM concept provides for probe cross-sections that, apart from circular, can be nominally oval, flattened or re-entrant. In this way the probe displacement volume is less than its volume of rotation, again to enable the easier flow of plasticised material. In addition, the inclination of the continuous spiral ridge can range from nominally 45° to as low as 5° or as high as 85° to the axis. This range allows adjustment in the degree of stirring versus downward drive of the plasticised material. For this downward thrust the rotation, as viewed from underneath the shoulder, must be in a clockwise direction for a right hand spiral, and vice versa. (A whorl type sea shell, with a spiral of progressively changing pitch is shown in Fig.8).

Inherent to all rotating systems is the natural dynamic orbit associated with rotating machinery. Together with the shape of the probe, it is this inherent or even nominally superimposed eccentricity that helps the plasticised material to be transported around the probe. In this respect non-circular flat-sided (paddle type) probes, enable the plasticised material to pass rearwards more easily with each half turn of rotation.

Shoulder development

For tools positioned perpendicular to the workpiece the leading edge of the shoulder in effect provides some preheat and hence thermal softening of the workpiece in front of the probe, which can be of advantage when dealing with harder or difficult to weld materials. The greater the area of the shouldered region making contact with the work surface the greater the preheat available, but this becomes less effective for substantially thick plate. Increasing the diameter of the shouldered region, however, has practical limitations and tends to produce side flash on the weld surface.

Inclined tools are also used, where the trailing edge of the shoulder is set slightly below the workpiece top surface, which helps to consolidate the weld. Essentially hydrostatic pressure within the third-body region leads to subsequent restoration of the full joint thickness as the FSW tool moves away as illustrated in Fig.9. For a given tool shape and tool inclination, the process only involves two parameters, the rotational and travel speed. This enables the FSW to be readily optimised.

Process asymmetry of FSW

Under satisfactory operating conditions the effect of rotation, with respect to the traverse, is not immediately obvious. However, there is always an inherent asymmetry, where the sum of the velocities of angular rotation and workpiece traverse is slightly greater on the outward (advancing) side compared with the inward (retreating) side see Fig.1. This nominal difference would not be expected to lead to any significant asymmetry itself. However, there is a major differential pressure with a tendency to void formation on the advancing side which has to be filled by the plasticised material from the retreating side under the available hydrostatic pressure. When the process parameters are incorrect, a void occurs which may be either surface or even surface-breaking. It can run parallel with the joint along its entire length on the advancing side of the tool probe. This feature is analogous to friction surfacing deposits made with non-optimised process conditions, where a distinct surface feature develops on the advancing side.

Weld joints

It should be noted that, for critical applications, double-sided FSW joints are preferred to single-sided, as they are essentially more process tolerant. Given the freedom of using double-sided welding, traverse speeds could be significantly increased. Conversely, to achieve optimum results from full penetration one-sided welding, traverse speed is relatively slow. For a given thickness, a double-sided weld involves smaller tools that displace less material than an equivalent single-sided tool, and would produce a smaller HAZ.

However, the FSW double-sided operation with simultaneous combined passes above and below the plates has further advantages. Moreover, the probes need not touch but be positioned sufficiently close so that the transient third-body encompasses both upper and lower weld regions, as illustrated in Fig.10. To avoid any problems associated with a zero velocity zone in mid-thickness, the probes can slightly be displaced along the joint, the rear probe benefiting from heat developed by the forward probe. This also allows overlapping weld regions, if desired. This simultaneous double-sided welding enables higher overall welding rates to be attained, provides a more balanced heat input giving even less distortion , and reduces the reactive forces on the work holding fixtures. Figure 11 illustrates the concept for a contra-rotating double-sided FSW machine for the fabrication of extruded sections.

In a butt joint the weld interface is essentially in line with the axis of the FSW tool and this is the most common arrangement, where penetration through the thickness is the main requirement. In contrast for a lap joint, the weld interface is transverse to the tool and the breadth of the weld interface is crucial to weld strength. Comparisons between butt, lap and T joints with corresponding effective weld sections are shown in Fig.12.

The interface of the lap joint is under compression while welding, unlike the butt joint where the probe tends to part the components to be welded. Because of this, the butt is more at risk from atmospheric contamination than the lap joint, especially for reactive materials.

Gas shielding

In practice most metal surfaces have some surface oxide, together with water vapour, hydrocarbons, and other adsorbed impurities. The oxide is the principal concern in solid phase bonding for aluminium and its alloys, where the oxide films formed are very tenacious, and the thinnest of surface films can inhibit adhesion and associated metal transfer.

Normally the FSW process, because of the stirring across the interfaces and close coupling of the shoulder, does not need a gas shield, but for the more reactive materials or more demanding requirements this may not always be the case. In operation plasticised material may become exposed to the atmosphere, through any gaps between the abutting plates and from underneath the leading edge of the rotating tool, as illustrated in Fig.13. Unlike conventional friction welding where the surface oxide is disrupted and subsequently dispersed, via the radial movement of plasticised material to the outer flash, the FSW technique can only disperse the oxide and absorbent layer within the weld itself.

The transfer of any exposed plasticised material, from the leading to the trailing parts of the weld zone or across the joint, may reduce the overall weld integrity. In these circumstances a gas shield should be considered, not only to minimise contamination, but in some cases to improve the weld surface finish. On the other hand, the very action of FSW avoids the formation of 'flat' spots as found in say flash butt welding, and surface films are positively broken up. Thus in practice FSW has been found satisfactory without further resort to some gas shielding or other protective measures such as chemical fluxes for a wide range of alloys. It is noted that reactive metals may require protection outside the interface region.

Brand identification and witness marks

An example of identification and witness marks is shown in Fig.14. These impressed marks were produced during the FSW operation and came from inverse embossing of the support plate. Typically these identification marks would not be positioned on the joint centre line but be offset slightly. Also, as an aid to quality inspection, witness marks of various depths can be designed to provide visible evidence that the desired process conditions have been achieved. Moreover, suitably embossed rollers can be used which repeatedly identify the company, date of welding, grade of material, as well as establishing quality control witness marks.

Conclusion

Friction stir welding provides a novel, relatively hazard-free, solid phase joining process, which produces sound joints, especially butt, in a range of materials and thicknesses. The FSW process is already in commercial use and has been found to be a robust, parameter tolerant, technique that has much to offer in the welding aluminium alloys.

References