Solid State Joining of Metals by Linear Friction Welding

Imran Bhamji¹,*, Michael Preuss¹, Philip L. Threadgill², Adrian C. Addison³

¹Manchester Materials Science Centre, University of Manchester, Grosvenor Street, M1 7HS.
²Formerly with TWI Ltd, Cambridge, CB21 6AL, now retired
³TWI Ltd, Cambridge, CB21 6AL
*Corresponding author. Email: Imran.bhamji@postgrad.manchester.ac.uk

Paper published in Materials Science & Technology 2010. Vol.27. No.1. Jan 2011. pp.2-12

Abstract

Linear friction welding (LFW) is a solid state joining process in which a joint between two metals can be formed through the intimate contact of a plasticised layer at the interface of the adjoining specimens. This plasticised layer is created through a combination of frictional heating, which occurs as a result of pushing a stationary workpiece against one that is moving in a linear reciprocating manner, and applied force. The process is currently established as a niche technology for the fabrication of titanium alloy bladed disk (blisk) assemblies in aero-engines, and is being developed for nickel based superalloy assemblies. However, interest is growing in utilising the process in a wider range of applications that also employ non aero-engine metallic materials. Therefore, it is the objective of this report to provide a broad view of the capabilities of the LFW process for joining metals.

This review paper will cover relevant published work conducted to date on LFW. The basics of the process and the fundamental aspects of operating a LFW machine will firstly be described, followed by a description of the different materials that have been welded using the process. The review will then go on to describe the microstructural changes, including texture variations, and residual stresses that are produced as a result of the welding process.

1 Introduction to linear friction welding

Linear friction welding (LFW) is a solid state joining process in which a stationary part is forced against a part that is reciprocating in a linear manner in order to generate frictional heat^[1,2] (Fig.1). The heat, along with the force applied perpendicular to the weld interface, causes material at the interface to deform and plasticise. Much of this plasticised material is removed from the weld, as flash, because of the combined action of the applied force and part movement. Surface-oxides and other impurities are removed, along with the plasticised material, and this allows metal-to-metal contact between parts and allows a joint to form.

Fig.1. Schematic diagram of the linear friction welding process

A defining feature of the linear friction welding process, along with all other friction welding processes, is that it takes place in the solid state and involves no melting of the parts to be joined^[1,3]. This means the process offers advantages over fusion welding when joining metals that exhibit solidification problems (e.g. porosity, hot cracking, segregation etc). In addition, in most cases the severe deformation in the weld region during friction welding results in a refined microstructure, which can provide improved strength at the weld line relative to the parent material^[4]. Other benefits include the fact that the processes are fairly quick (in terms of total weld time; less than 10sec for Ti-6Al-4V⁴) and provide parts with good repeatability. Oxide formation is also reduced, because of the close contact between parts^[5], which means shielding gases are only rarely needed.

One of the main disadvantages of the process is the high capital cost of the equipment. The high cost of both the equipment and tooling means the LFW process can only be justifiably used for producing high value added components^[4]. This has generally confined the process to niche applications such as producing bladed disks for aero-engines. However, machines based on the principle of stored energy, rather than direct drive, have significantly lowered costs and this may mean the LFW of lower value added components can be justified^[6]. A further disadvantage of the process is that it can be very noisy.

There have been a few brief literature reviews on LFW^[7,8], but they have tended to focus on specific areas of the process, such as energy input^[7] or the use of the process for blisk manufacture^[8]. Neither of the reviews has attempted to give a wide ranging and in-depth overview of the process, which this document aims to do. Therefore, this review should be a beneficial addition to the publicly available literature on LFW.

1.1 History of linear friction welding

Linear friction welding was first patented in 1929^[9]; however, the description of the process was vague. Some discussion of the concept was then recorded in the 1960s, but it was described as 'very doubtful' because of the difficulty in generating linear reciprocation^[2]. The Caterpillar Tractor Company^[10] was the next to mention the process in a patent. However, the patent primarily focused on the machine that generates the linear reciprocation and not the actual welding process. Indeed a patent search has shown that no currently valid patents exist that protect the fundamentals of the LFW process. However many patents protect certain aspects of LFW, such as particular applications, welding methods or tooling concepts (for example^[11,12]).

To date the only major commercial use of LFW is for the joining of aero-engine compressor blades to compressor disks, to form blisks^[8] (Fig.2), although other applications are known to be in development (for example^[13]). As LFW is predominantly used for blisk manufacture, much of the published literature on the subject has focused on the joining of materials for aero-engine applications. The materials studied with regard to the process are predominantly titanium alloys, especially Ti-6Al-4V (Ti-64) (for example^[4,14]), with some work also conducted on the LFW of nickel based superalloys (for example^[15,16]). However, the process can be used to join other materials and interest in this aspect is growing.

Fig.2. A linear friction welded blisk assembly (Courtesy of MTU Aero Engines^[8])

1.2 Linear friction welding machine operation

The machine operation of the LFW process can be broken down into six separate stages. These are:

Part clamping: The parts are held using tooling designed to withstand the forces experienced during the process. Specimen and tooling preparation is critical to the process, with accurate sides and edges needed on the specimen, and a tight-fit needed between the specimen and the tooling. This generally means that the tooling is custom built to fit particular specimen geometries.

Datum and retract: The clamped parts are brought together under a small compressive force in order to determine the location of the parts and set the machine datum to zero. The parts are then retracted to leave a small separation distance between the workpieces.

Conditioning phase: Oscillation of one of the parts is increased and stabilised over a set period (usually very quickly) and the parts are brought together under a small force for a predetermined time (Fig.3).

Frictional phase: The compressive force (friction force) is increased to a set level and heat is generated at the interface. The material at the interface becomes plastic and flows out of the weld, as flash, because of the shearing motion between the two parts and the applied force. This loss of material from the weld causes the parts to shorten (or burn-off). This phase usually ends, and the next is triggered, when a predetermined loss of length, or burn-off distance, is reached. However, the next phase can also be triggered after the frictional phase has continued for a predetermined time (burn-off time) or number of oscillation cycles (burn-off cycles).

The LFW process is always carried out under load control, but other parameters also play a role in controlling the welding process. For example when using a burn-off distance the load is controlled throughout the welding process, however the burn-off is also monitored (although not controlled) and at a set burn-off distance the next phase (forge phase) is triggered. Similarly with burn-off time or cycles the load is controlled throughout welding and the amount of time or cycles determines the transition to the next phase.

Forge phase: The amplitude is decayed to zero over a predetermined time to ensure good alignment (usually very quickly), and a forge force is rapidly applied and held for a set time to consolidate the joint. The forge force can either be the same as or higher than (more common) the friction force.

Release phase: The welded parts are released from the clamps and removed from the machine.

Fig.3. Schematic diagram of the parameter traces that are obtained during the linear friction welding process. A number of input variables are defined in the diagram. Burn-off is defined as the loss of length occurring as the process continues, whilst upset is the total loss of length measured after the weld has been produced.

The frictional phase can be further broken down into three separate phases that are determined by material behaviour (Fig.4)^[17].

Fig.4. Phases of the linear friction welding process, determined from consideration of material behaviour (redrawn from^[17]).

Initial phase: The parts are forced together and heat is generated through friction. The rubbing together of parts causes asperities to wear down and flatten so that the true area of contact increases towards 100%. A small amount of part shortening occurs due to wear particle expulsion. It is critical for the rest of the process that enough frictional heat is generated in this phase of the process.

Transition phase: The temperature at the interface rises and therefore the strength of the material at the interface decreases. The applied load can then cause this low strength material to soften and plastically deform.

Equilibrium phase: The hot plasticised material at the interface is expelled, as flash, with the help of the oscillatory motion and applied pressure.

The main input variables during the process are:

• Frequency: number of oscillatory cycles per second.
• Ramp-up time: time taken to increase the welding parameters to the required steady state level (Fig.3).
• Amplitude: maximum displacement of the oscillating sample from its equilibrium position.
• Friction pressure: pressure applied during the frictional phase of the process. Pressure is calculated by using the nominal area of contact at zero amplitude.
• Burn off distance, time or cycles: possible factors that trigger the start of the forge phase.
• Decay time: time taken to reduce the amplitude to zero at the beginning of the forge phase.
• Forge pressure: pressure applied during the forge phase of the process.
• Forge time: the amount of time that the forge pressure is applied.

There are also other variables that are of importance, but which are a consequence of the main variables and cannot be easily and precisely changed through the variation of the main variables. These consequent parameters are:

• Upset: the total loss of length (shortening) due to the process (the upset will always exceed the burn-off distance mainly because of the loss of length occurring during the forge phase).
• Shear (or in-plane) force: force parallel to the oscillatory movement.
• Burn-off rate: rate of shortening (i.e. gradient of the burn-off curve, Fig.3).
• Welding time: total time taken to weld a specimen.

1.3 Quality control systems for linear friction welding

Although no extensive quality control systems have been reported in the public domain for producing and testing linear friction welds, procedures have been disclosed for rotary friction welds^[18]. As RFW and LFW are similar processes the procedures for RFW should also be applicable to LFW.

It has been stated that destructive and non-destructive tests can be used to control the quality of friction welds^[18]. However, because of the costs of these tests, visual examinations, loss of length (upset) checks and control of weld parameters is sufficient to ensure weld quality in most non-critical situations^[18]. Statistical process control is usually the primary method of quality control for non-critical applications, and is also an important quality control method in critical applications. This process relies on keeping all significant welding parameters within set boundaries well away from values that may cause poor weld properties. The LFW process has been shown to produce welds of consistent quality when welding parameters are controlled, which would suggest that statistical process control is a good method of ensuring quality. It is also recommended^[18] that destructive tests are conducted at appropriate intervals to ensure weld quality. These can be in the form of tensile and/or other mechanical tests, or metallographic examination.

For welds in critical applications, such as those in blisks, more stringent quality control systems will be required. The use of non-destructive testing, along with the other control systems mentioned, may be needed to ensure weld quality in these situations^[18].

Table 1. Guidelines for quality assurance in rotary friction welds (BS EN ISO 15620:200018). Indicated is the percentage of samples that should be inspected for each inspection technique.

Category	Visual examination	Check of total length loss (upset)	Parameter monitoringa)	Recording of parameters	Destructive testing mechanical and micrographic	Non-destructive testing
A	100%	100%	-	100%	b)	c)
	-	-	100%
B	100%	10%	-	to be defined	b)	c)
	-	-	50%
C	50%	5%	-	periodic, at least once every 6 months	b)	c)
	-	-	20%
a) At least monitoring of friction pressure, forge pressure, burn-off distance, total length loss (upset) and welding time. b) Frequency to be defined. c) The application of non-destructive testing depends on conditions of use and further machining. When developing the welding procedure specification, weldments which are subjected to dynamic loads in service without removal of the upset material should be tested non-destructively after removal of the upset material.
Category A: Where failure of welded components is dangerous for the product and the environment. Category B: Where failure of welded components will cause considerable damage. Category C: Where failure of welded components will cause limited damage.

2 General materials aspects of linear friction welding

LFW is particularly suited to joining materials that have good high temperature properties, especially compressive yield and shear strength, and a low thermal conductivity^[1]. The good high temperature mechanical properties allow a high level of frictional heat generation, whilst the low thermal conductivity helps confine heat to the interface. These properties, therefore, allow welds to be made using low welding parameters as a high level of heat can be confined close to the interface^[1]. This makes titanium and nickel alloys, with their good high temperature mechanical properties and low thermal conductivities, particularly suitable for the process.

Regardless of this, very good linear friction welds have been made in materials with quite different properties from those of titanium and nickel. Various alloys of aluminium^[1,19] and aluminium metal matrix composites^[20,21] have been adequately welded using the process, even though these materials have relatively poor high temperature mechanical properties and very high thermal conductivities. However, high parameters (high applied forces relative to the material strength, and high amplitudes and frequencies) are generally needed to get the heat concentrations required to produce sound welds, and the parameter window in which good welds are achievable is reduced, relative to that in titanium^[1,4,19]. The parameter window for adequate welds in some advanced nickel based superalloys may also be small because of the sudden loss of strength of the material at high temperatures^[22], which is related to the rapid dissolution of strengthening phases as the temperature exceeds 800 - 900&C (see section 3.2.2).

Adequate welds have also been demonstrated in various steels^[23,24,25] as well as tungsten based alloys and cobalt based superalloys^[8]. In addition, intermetallics such as titanium aluminides^[26,27] and nickel aluminide^[28] have been welded successfully using LFW.

Although LFW could be a promising process for producing highly dissimilar welds (welds between different material classes), this aspect has been unexplored in the publicly available literature. Any literature that does exist is confined to welding different alloys of the same base material^[15,29]. The only mention of the LFW of dissimilar materials is in a patent by the Northwestern Polytechnical University in Xian, China, which protects welding parameters for the LFW of aluminium to steel^[30]. Although the LFW of dissimilar materials has been almost totally unexplored, there has been a large amount of work on the RFW of dissimilar materials, with some good results produced^[18,31]. As the two processes are fundamentally similar (i.e. in both processes heat is generated by contact and relative movement between parts, and plasticised material is ejected by a combination of part movement and applied force), this provides hope that dissimilar welds can also be produced with the LFW process.

This review of different materials that have been linear friction welded shows that the process can be applied to more than just titanium and nickel alloys, and could be applied to applications away from the aero-engine industry. This suggests that the limited amount of information on the LFW of non aero-engine materials is related to a lack of research and not feasibility issues. A large number of different materials have been joined using RFW^[18], and as it is a very similar process to LFW there is no reason to suppose that these materials cannot be welded, with equal success, using this process.

3 Detailed characteristics of linear friction welds

3.1 Mechanical properties

The mechanical properties of linear friction welds in a number of different materials have been investigated, and the results from these studies have shown the joint strengths either surpassing or being just less than those of the parent material during tensile testing. The yield and ultimate tensile strengths of defect free Ti-64 welds (i.e. welds without voids or oxides) exceeded those of the parent material^[4,29] (tested according to ASTM E8M-01, with strain rate 1x10^-4 sec^-1; displacement velocity 2mm min^{-1 [29]}). Similarly linear friction welds in a ferritic steel^[25] gave improvements in both strength and ductility relative to the parent material (strain rate 1x10^-4 sec^-1), and welds in an aluminium alloy^[19] demonstrated a strength just less than that of the parent material (failed at 85% of the parent strength; tested according to ASTM E8-87) during tensile testing. Adequate welds have also been produced between different nickel based superalloys^[15], and it has been demonstrated that welds between 720Li and IN718 (nickel based superalloys) have a resistance to fatigue crack propagation at least comparable to the parent IN718^[32].

The effects of producing successive linear friction welds (i.e. one weld on top of the other) have been reported^[33]. The aim was to simulate the effects of blisk repair, and involved the removal of a weld (welds were produced in Ti-6Al-2Sn-4Zr-6Mo) by cutting parallel to the weld plane, 3mm from the weld line, and welding onto this cut surface. The results from this study, which simulated up to two repairs (i.e. up to three successive welds), showed that tensile failure always occurred away from the weld lines, in the parent material. High cycle fatigue tests also produced failure in the parent material, but not enough samples were examined to determine if the presence of the welds had a definite impact on fatigue properties.

3.2 Microstructure

Although the interface temperature during LFW is not expected to exceed the melting temperature of the material being welded, the peak interface temperature is, nevertheless, very high, and could be close to the solidus temperature of the material being welded^[16]. In all reported cases, this high temperature, along with the applied pressure, causes significant microstructural variation close to the weld interface.

No widely accepted nomenclature exists for microstructural regions in linear friction welds, but there would be a clear advantage in establishing such a system as it would avoid much confusion in the literature. The following proposal is based on a nomenclature which has received widespread acceptance for friction stir welding^[34]. It is proposed that the weld is divided into four regions, each defined as follows:

• Parent material (PM): This is material, some distance from the weld line, where no change in microstructure, mechanical properties or other properties can be detected.
• Heat affected zone (HAZ): In this region the microstructure and/or other properties have been changed by heat from the weld, but there is no optically visible plastic deformation. Changes could for example include one or more of grain growth, change in precipitate morphology, changes in mechanical or physical properties.
• Thermo-mechanically affected zone (TMAZ): In this region the material has been subjected to more heat than in the HAZ and shows clear evidence of plastic deformation (see Fig.5d for a typical TMAZ microstructure in Ti-64). Phase transformations could also take place, in some materials, in this zone. Changes in this region would be expected to be more apparent than in the heat affected zone.
• Weld zone (WZ): In many materials there will be a region, close to the weld line, where the microstructure is very different to that in other parts of the weld. This is usually because of recrystallisation, which produces a region consisting of very fine equiaxed grains (see Fig.5c for a typical WZ microstructure in Ti-64), and/or phase transformations. Since this area has been subjected to heat and plastic flow, it is a sub-group of the TMAZ.

An identical approach to this could also be used for rotary and inertia friction welds.

3.2.1 Microstructure of titanium linear friction welds

The microstructure of linear friction welds in bimodal (α and β structure, Fig.5a and b) titanium alloys have been studied by a number of researchers^[4,35,36]. From this work it is evident that during LFW the material close to the weld line exceeds the β-transus temperature. A very fine grained structure is typical of microstructures in this region, which is related to the material being exposed to high temperature and strain resulting in dynamic recrystallisation. The fully β transformed microstructure is rapidly cooled, after the frictional phase of the process, which avoids β grain coarsening and leaves a Widmanstatten microstructure of α and β plates delineated by prior β grain boundaries (Fig.5c). It has also been suggested that the cooling rates experienced by the weld could be high enough to produce martensite (α) (with some retained metastable β)^[36], suggesting that cooling rates greater than 410°Cs^{-1 [4,35]} can be achieved in titanium LFW.

Slightly further away from the weld line the non-recrystallised region of the TMAZ is present. In this region grains are heavily deformed and are reoriented (in terms of morphology) so that their long dimension is perpendicular to the applied force (Fig.5d)^[4]. Titanium welds do not display a prominent HAZ (although subtle HAZ's may exist that are very hard to detect), and there is possibly a direct transition from the TMAZ to the parent material^[4].

Fig.5. Microstructure of a Ti-64 linear friction weld^[4]. a&b: Parent material (consisted of alternating layers of the two types of microstructure), c: Widmanstatten microstructure in the weld zone (which had a width of ~180µm), d: Non-recrystallised region of the thermo-mechanically affected zone (which had a width of ~290µm on one side of the weld), showing deformation of the parent microstructure^[4]. (with kind permission from Springer Science+Business Media. © The Minerals, Metals & Materials Society and ASM International 2005)

3.2.2 Microstructures of non-titanium linear friction welds

There has not been an extensive amount of research conducted into the LFW of materials other than titanium, and much of the research that is available does not investigate the microstructures of the welds in depth. However, it is worthwhile to briefly mention some of the research that is publicly available.

LFW of nickel based superalloys: The welding of IN718 produced a significant loss of strength at the weld line due to dissolution of the γ' and γ'' phases that strengthen the material^[16]. Strengthening precipitates were also seen to dissolve at the weld line of linear friction welds in other nickel alloys^[15]. Oxides have been reported at the weld line of some nickel based superalloys^[15,37], but welding these types of material in a vacuum has not yet been proposed. Work on the LFW of a single crystal to a polycrystalline nickel based superalloy^[15] has shown a relationship between weldability and the orientation of the single crystal. The materials were easiest to weld when the primary slip system of the single crystal was favourably orientated to give a high Schmid factor. When the single crystal material was not orientated for easy activation of primary slip, the welding process was unsuccessful.

LFW of an aluminium alloy: Linear friction welds in Al-Fe-V-Si 8009^[19] aluminium alloy showed a decrease in strength at the weld line, relative to the parent material. The overaging or dissolution of hardening particles, during the welding process, may have been significant in reducing strength at the weld line in this material.

LFW of titanium aluminides: There has been some degree of success in LFW α₂ ²⁷ and γ³⁸ titanium aluminides. When welding α₂ titanium aluminides it is important that cooling rates are kept to a minimum to avoid undesirable microstructures²⁷. At high cooling rates the relatively slow β-to-α transformation causes a microstructure of retained β, which has low notch toughness, or α₂ martensite, which is very brittle. In order to produce a more desirable microstructure of a mixture of α₂ and β, very slow cooling rates are needed. It has been suggested that this could be achieved by optimising welding parameters, with a regime of low forces and amplitudes and high frequencies proposed^[38]. Despite these difficulties crack free welds have been produced in α₂ titanium aluminide.

Crack-free welds were also produced in a γ titanium aluminide^[38]. A fine lamellar α₂/γ microstructure was formed in this material by using welding parameters that gave long welding times and therefore relatively slow heating and cooling rates.

3.3 Weldability

Although most of the public work on weldability and welding parameters has been conducted on Ti-64, the work should be equally applicable to other materials. Vairis and Frost^[14] showed that for sound linear friction welds to be formed in Ti-64, a specific power input parameter must be exceeded. It was shown that frequency, amplitude and pressure have an effect on this parameter, which was defined as:

with α being the amplitude, f the frequency, P the pressure and A the weld area. From this equation it can be seen that the power input can be increased by increasing the frequency, amplitude or pressure.

Wanjara and Jahazi^[4] showed that another parameter, the upset, was also important in forming sound welds and demonstrated that a minimum level of upset was necessary to consolidate the weld. Therefore the power input parameter devised by Vairis and Frost^[14] cannot be used as an exclusive criterion for obtaining sound welds.

3.4 Effects of welding parameters

The influences of various welding parameters on the size of recrystallised β grains in the WZ of Ti-64 welds have been reported^[4]. This work gives a good insight as to how the interface temperature varies as welding parameters are changed, as the β grain growth will be dependent on it. It was shown that an increase in frequency or pressure increases the size of the prior β grains (Fig.6). This is thought to be because an increase in power input, associated with an increase in frequency or pressure, causes the temperature at the interface to be greater. However, the increased prior β grain size could also be related to slow cooling rates when high parameters were used^[4].

Fig.6. Effects of frequency, amplitude and pressure on average prior β grain size in the weld zone of Ti-64 linear friction welds^[4]. a: amplitude, P: pressure, f: frequency, s: burn-off distance. Open markers indicate poor welds. (with kind permission from Springer Science+Business Media. © The Minerals, Metals & Materials Society and ASM International 2005)

More recent work, however, has shown that the prior β grain size actually decreases at high pressures, which has been interpreted as a reduction in the peak welding temperature^[39]. This decrease in temperature was attributed to a large amount of flash expulsion at high pressures, which caused large heat rejection, and short welding times. The conclusion that the welding temperature decreases when using higher pressures is further supported by ^[40]. The results from ^[40] will be discussed in more detail in sections 3.5 and 3.6.

The reasons for the discrepancies between studies are not clear. However, there are differences in the materials being welded (Ti-64^[4,40], Ti-6Al-2Sn-4Zr-6Mo^[39]), and the machine, parameters and specimen geometry being used, which could have caused the differing findings. Given these differences none of the studies should be dismissed, and the effects of different variables on welding temperature should be further investigated.

It was also demonstrated that an increase in oscillation amplitude results in a reduction of the prior β grain size^[4] even though the power input (according to equation 1) increases (Fig.6). It was reasoned that large oscillation amplitudes expose a considerable amount of material to the surrounding atmosphere, resulting in increased convective heat dissipation. Therefore, an increase in power input would not create the expected temperature increase.

These parameter studies clearly demonstrate that the welding temperature can be controlled, to some extent, through the control of welding parameters. Minimising the peak welding temperature is an important objective, particularly when attempting to join highly dissimilar materials (e.g. aluminium to copper) by LFW. In such cases, the formation of intermetallics is usually a major obstacle. Therefore, minimising the welding temperature aims at avoiding the formation of detrimental intermetallic phases at the weld line. No publicly available work has investigated whether a similar control of welding temperature can be achieved when welding materials other than titanium. However, there is no reason to assume that the relationships observed for the LFW of titanium would not occur in other materials because similar changes in power input and heat dissipation should take place with parameter changes.

3.5 Residual stresses in linear friction welds

Measurements of residual stresses, deep inside the weldments, have been carried out on linear friction welds in a number of different titanium alloys^{[40,41,42,43,44]} and in an aluminium metal matrix composite (AA2124 reinforced with 25% silicon carbide)^[21]. Each of these studies have been carried out by using either high energy synchrotron X-ray or neutron diffraction, and each rely on calculating strains and stresses from changes in lattice parameter^[45]. The studies are in broad agreement, and most of them show tensile stresses in all three directions at central locations (i.e. mid-width and mid-thickness) near the weld line. The stresses were highest in the transverse direction (in the weld plane but perpendicular to the direction of oscillation, which is the reciprocating direction) and lowest in the axial direction (direction of force application).

The value of the highest stress differs significantly between studies and ranges from about 20% of the welded materials yield strength in ^[40] to about 90% of the yield strength in ^[43] (both of these studies were on Ti-64, but welding parameters were not disclosed in either study). All of the studies show a sharp drop in stresses on either side of the weld line, becoming compressive in the adjacent region before finally approaching zero (Fig.7).

Fig.7. Stress in the three spatial directions as a function of position along the axial direction (y-axis)^[41]. a: reciprocating direction, b: axial direction, c: transverse direction (study conducted on Ti-64).

A particular issue with residual stress calculations is the accurate determination of the strain-free lattice spacing (d₀), which is used as a reference value when determining residual strains and stresses. The d₀ value will change across the weld line because of microstructural changes and changes in the phase specific chemical composition. When these changes in d₀ are taken into account much lower stress values are reported (as in ^[40]), relative to without d₀ correction (as in ^[43]). As well as this, the stresses in the axial direction become negligible when d₀ variations are taken into account, suggesting that the residual stress state is predominantly biaxial in the plane of weld interface^[40]. Whilst the d₀ variation is clearly important for an accurate understanding of residual stresses in linear friction welds, it is important to note that some uncertainty in d₀ values occurs at locations very close to the weld line (<0.5mm)^[42]. This also means that there is some uncertainty associated with the residual stress results directly at the weld line and more work is needed to accurately determine the stresses in this region.

Although the magnitude of residual stress in weldments can vary with direction, as has been discussed in the preceding paragraphs, it is important to note that this variation is related to sample geometry rather than the oscillation and applied force directions defined by the welding process. Residual stresses in linear friction welds are a result of the thermal mismatch created by different cooling rates at various positions across the weld (material at the surface of the weld will cool more quickly than that in the centre). As the weld geometry will affect the thermal profile within a weld, it is clear that it will also have a substantial effect on residual stresses, for a given material.

In almost all of the welds studied for residual stress the sample was reciprocated in the small dimension, with a larger dimension in the transverse direction. As the distance between the weld centre and the surface is longer in the transverse direction, relative to that in the reciprocating, it is likely that there will be a larger thermal gradient, and hence thermal mismatch, in this direction. This explains why generally larger tensile stresses are observed in the transverse direction compared to the reciprocating direction.

The low stresses in the axial direction, generally observed when appropriate d₀ corrections were applied across the weld line^[40], demonstrate that only small thermal mismatches are generated in this direction. The small thermal mismatches may be related to the high weld pressure that is applied in the axial direction, which results in sufficient plasticity to compensate for any mismatch.

The results described in this section show that very significant stresses develop during linear friction welding, which can have a detrimental effect on the long term performance of the welded component. Consequently, the development of an appropriate post weld heat treatment (PWHT) that relieves the stresses sufficiently without compromising microstructure and mechanical properties is of major importance. A few studies have investigated the influence of PWHT on residual stresses, and reductions in stresses ranging between 75 and 90% have been reported for various titanium alloys^[42,44]. An important dependence of residual stress relief on specimen size has been reported in Ti-64^[43]. In a small, lab scale, linear friction welded sample negligible stress levels existed after PWHT, whilst in a full scale blisk significant tensile stresses still existed after an identical PWHT. Both welds displayed similar residual stress profiles in the as welded condition^[43].

It has also been suggested that residual stress development can be minimised by optimising welding parameters, with low stress levels being observed when high applied pressures were used during LFW of Ti-64^[40]. These low stresses are thought to result from a low peak welding temperature at high pressures, which was also suggested in ^[39] (see section 3.4). This result clearly demonstrates that stresses can be minimised by the choice of appropriate welding parameters, which is particularly important if sufficient residual stress mitigation by a subsequent PWHT is difficult to obtain (for instance dissimilar welds).

3.6 Texture in linear friction welds

Textures in Ti-64 linear friction welds of different sizes (lab scale and full scale specimens) have been investigated^[36]. With both specimen geometries a strong transverse texture, , in the α phase was seen close to the weld line. In simple terms, the c-axis of the hexagonal close packed a crystallites are predominantly orientated parallel to the transverse direction (Fig.8). This strong texture was present across a region of approximately ±100µm from the weld line in the lab scale specimen, and ±50µm from the weld line in the full scale specimen^[36]. (The TMAZ in the lab scale specimen spanned a region of ±250µm from the weld line, with a WZ of ±80µm, whilst the TMAZ in the full scale specimen spanned ±500µm with a WZ of ±150µm36). It has been suggested that this strong α texture is the result of a β deformation texture generated during LFW.

On subsequent cooling this β texture, in combination with strong variant selection during β to α phase transformation (12 different α variants can form from a single β grain but here only a single variant was activated), resulted in the strong transverse texture observed at the weld line ^[36]. Although these strong textures have been reported in Ti-64 linear friction welds, more research is needed to determine the significance of these textures on the mechanical and physical properties of the weld.

Fig.8. Crystallite orientation at the weld line of lab and full scale Ti-64 linear friction welds^[36]. The region of strong transverse texture spans ±100µm from the weld line in the lab scale specimens and ±50µm in the full scale specimens.

Although textures close to the weld line in lab and full scale specimens were similar, they were quite different slightly further away from it^[36]. In the full scale specimens alternating bands of transverse and type textures (c-axis of the α crystallites aligned almost parallel to the reciprocating direction) were observed a small distance away from the weld line (Fig.9). It appears, therefore, that two (of the 12 possible) types of variant were selected on β→α transformation in full scale welds. At the moment, this difference in variant selection between lab scale and full scale welds has not been explained.

Fig.9. α crystallite orientation at and near the weld line in a full scale Ti-64 linear friction welded blisk^[36].

Results from ^[40] (also on Ti-64) have suggested an influence of welding parameters on texture development in linear friction welds (along with an influence on residual stress, as described in section 3.5). When welding with a low applied pressure the before mentioned strong transverse α texture is observed at the weld line. However, an almost random α texture developed at the weld line when high pressures were used. To date, the mechanisms by which welding pressure can minimise the α texture at the weld line is not entirely clear, but it seems obvious that variant selection is less active when using high applied pressures to weld Ti-64.

4 Equations of power input and modelling of linear friction welding

There have been some attempts at modelling heat generation in linear friction welds, and a heat input model for the equilibrium phase (see section 1.2) of the process has been developed^[17]. The heat input power per unit area in this model was defined as:

where Τ is the shear stress, ν the sliding velocity, F_S the shear force, µ the co-efficient of friction, FN the force normal to the interface, α the amplitude, Ω the angular frequency (2Π*oscillation frequency), t the time, D the thickness, and W the width of the specimen (weld area = DW). A is the contact area, which is not constant and will change during the oscillatory cycle.

Equation 2 is correct for the most part, but the contact area term D(W - αsin(Ωt)) needs to be adjusted slightly. In this term D(α sin(Ωt)) is the area not in contact at any time during the oscillatory cycle, and DW is the nominal area of contact at zero amplitude. The modulus of the sine term needs to be taken in this part of the equation (i.e. values in the αsin (Ωt) term need to be turned positive when the sine term turns negative in the second half of the oscillatory cycle) and it would appear that this was not done in the original paper.

Although the equation of power input into the weld (equation (2)) is likely to be sufficient for a retrospective analysis of heat input, it may not satisfy requirements for a forward, predictive, analysis from a given set of initial welding parameters (i.e. pressure, force, frequency, amplitude and burn-off distance). This is largely because of the difficulty in predicting the shear force (force in the reciprocating direction) that results from the initial parameters. The shear force will be dependent on both sliding velocity and the real (or true) contact area (area of contact of microscopic asperities)^[46], and this is not brought out in equation (2). Although equation (2) does have a shear stress that is contact area dependent, this is solely a result of dividing a constant shear force by an area that is changing due to the oscillation. Therefore, the model is not an accurate depiction of the welding process as the shear force also oscillates with both contact area and velocity14. This unsatisfactory representation of the shear force may also mean that there is an inadequate representation of the power generation, and better models may be needed to accurately describe the power generation and temperature profiles produced by the LFW process.

A further limitation of the model is that it does not take into account the energy and heat losses that occur because of flash formation. As a result, it is impossible to relate the outcome of the model to the microstructural properties of the weld. A coupled thermal and mechanical model that can predict both temperature and burn-off rate, and therefore predict the energy in the weld and flash at a particular time, may be more useful.

There have been only a few studies^[47,48] on using finite element modelling for developing a process model of LFW. The studies focused on the LFW of Ti-64 blocks and the thermal profiles were validated by undertaking thermocouple measurements^[48] or undertaking thermocouple measurements together with monitoring the axial shortening (upset)^[47]. In ^[47] the discrepancy between the predicted and measured temperature profiles were smaller than 12% while the predicted and measured upset differed by less than 16%. In ^[48] temperature measurements corresponded very well to the model in the initial stages of the welding process, but there were large discrepancies between predicted and measured temperatures during the later stages of LFW.

Although these results show some promise, it would appear that each of the studies were only validated with one weld. Therefore, there is no indication that the models will still be applicable if the welding parameters are changed significantly from the values used in these particular publications. Moreover, in ^[47] an atypical weld pressure for linear friction welding of Ti-64, of just 20MPa, was applied, while a pressure that is at least two times greater is usually considered to be necessary for structurally sound welds^[4].

Despite the criticism of the LFW process models currently available in the open literature, it is important to remember that the development of any reliable process model is difficult. It will largely depend on the accurate knowledge of friction coefficients and forces, and will rely on an accurate material data base over a wide temperature range and for exceptionally high strain rates. Much of this information is not available in the public domain. An additional complication, when considering the generation of a material data base, comes from the fact that the material that is pushed into the WZ experiences extremely high heating rates and consequently non-equilibrium conditions that are very difficult to simulate.

5 Summary

A review of published information on LFW has been conducted, and from this the following general conclusions can be drawn.

• The bulk of literature on LFW concerns the welding of titanium alloys, with some work also done on welding nickel alloys. However, a few publications focusing on other materials have confirmed that the LFW process allows sound joints to be made in a number of different materials. These include, amongst others, welds in various steel and aluminium alloys, as well as welds in titanium and nickel aluminides. This suggests that the process could be used for applications away from the aero-engine industry. To date, the production of aero-engine blisks is still the only major commercial application of the process.
• The peak interface temperature during LFW is very high, although it is not expected to exceed the melting points of the materials being welded. As a result of such high temperatures near the weld line, along with the applied force and plastic deformation taking place during LFW, significant microstructural changes have been reported in linear friction welds.
• The microstructure of a LFW is usually characterised by a thermo-mechanically affected zone (TMAZ) close to the weld line, a heat affected zone (HAZ) further from the weld line and the unaffected parent material microstructure further still. The TMAZ usually displays a region, at the weld line, that is recrystallised and contains fine grains (the weld zone), and a region, further away from the weld line, that is heavily deformed, but not recrystallised. The HAZ is affected by the transferred heat during welding, but is not plastically deformed (this region is not prominent in some materials, e.g. Ti-64).
• Welding parameters appear to have a significant influence on the welding temperature, and this property of the process may be important when welding challenging materials (e.g. when welding dissimilar materials). However, more work is needed to better understand how the welding parameters affect welding temperature (as different studies have produced contradictory results) and the range of temperatures that can be achieved by changing welding parameters.
• Residual stresses have been studied in a small number of linear friction welded systems, and in all cases significant stresses have been reported in the material close to the weld interface. In the case of titanium it has been demonstrated that these stresses can be relieved by more than 75% when applying alloy-typical annealing procedures during PWHT. In addition, a significant reduction in stress generation was found when using high weld pressures.
• To date, very little research seems to have been conducted to develop a widely applicable process model for LFW. Even though such a model might not be fully predictive at the first stage, due to the complexity of LFW, it is clear that it could help to significantly improve our understanding of the process, and in this way provide strategies to further improve welding parameters for LFW.

Acknowledgements

The authors would like to thank the EPSRC and TWI Ltd for financial support, and the judging panel of the 2009 Materials Literature Review Prize for their valuable comments on this report. The authors would also like to acknowledge the help and support of other staff within the friction department of TWI and the assistance provided by Dr Moataz Attallah (University of Manchester).

References

1. A. W. E. Nentwig: 'Untersuchungen zum linear-reibsscheissen von metallen', Schweissen und Schneiden, 1995, 47(8), 648-653.
2. V. I. Vill: 'Friction welding of metals'; 1962, New York, American Welding Society.
3. B. Crossland: 'Friction Welding ', Contemporary Physics, 1971, 6(12), 559-574.
4. P. Wanjara and M. Jahazi: 'Linear friction welding of Ti-6Al-4V: Processing, microstructure, and mechanical-property inter-relationships', Metall. Mater. Trans. A, 2005, 36A(8), 2149-2164.
5. W. M. Thomas: 'Friction stir welding and related friction process characteristics', Proc. 7th Int. Conf. on 'Joints in Aluminium - INALCO '98' (ed: M. H. Ogle, P. L. Threadgill, and S. J. Maddox), TWI Ltd, Cambridge, UK, 15-17 April 1998, 1999, Woodhead Publishing, 157-174.
6. S. Kallee: 'Development of low-cost linear friction machine', Assembly Automation, 2001, 21(2).
7. M. Maalekian: 'Friction welding - critical assessment of literature', Sci. Technol. Weld. Join., 2007, 12(8), 738-759.
8. M. E. Nunn: 'Aero engine improvements through linear friction welding', Proc. 1st Int. Conf. on 'Innovation and Integration in Aerospace Sciences', Paper no 2005-0040, Queen's University Belfast, 4-5 August 2005, 2005, CEIAT, Queen's University Belfast.
9. W. Richter: 'Herbeifuehrung einer haftverbindung zwischen plaettchen aus werkzeugstahl und deren traegern nach art einer schweissung oder loetung', Patent: DE477084, 1929.
10. R. Maurya and J. Kauzlarich: 'Bonding apparatus - friction welding by reciprocal motion', Patent: Caterpillar Tractor Co, US3420428-A; DE1552871-A; CA844858-A, 1969.
11. K. T. Slattery: 'Structural assembly's e.g. wing, preform for e.g. aircraft, has set of structural members connected by linear friction weld joint to base member, where each structural member defines connection surfaces', Patent: Boeing Co, US2007186507-A1, 2007.
12. R. D. Trask, S. H. Goetschius, and S. A. Hilton: 'Process for linear friction welding - comprises providing two members to be joined to one another at an interface by linear friction welding, where one of the two members is stub, etc', Patent: United Technologies Corp, EP850718-A; EP850718-A1; JP10193141-A; US5865364-A; EP850718-B1; DE69734232-E; DE69734232-T2, 1998.
13. K. T. Slattery: 'Preform forming method for structural assembly, involves moving pair of structural units relative to two base units to form plasticized region of material, and urging structural units against surface of base units', Patent: Boeing Co, US2005127140-A1, US7225967-B2, 2005.
14. A. Vairis and M. Frost: 'High frequency linear friction welding of a titanium alloy', Wear, 1998, 217(1), 117-131.
15. M. Karadge, M. Preuss, P. J. Withers, and S. Bray: 'Importance of crystal orientation in linear friction joining of single crystal to polycrystalline nickel-based superalloys', Mater. Sci. Eng., A, 2008, 491(1-2), 446-453.
16. C. Mary and M. Jahazi: 'Linear friction welding of IN-718: process optimization and microstructure evolution', Advanced Materials Research, 2007, 15-17, 357-362.
17. A. Vairis and M. Frost: 'On the extrusion stage of linear friction welding of Ti6Al4V', Mater. Sci. Eng. A-Struct. Mater. Prop. Microstruct. Process., 1999, 271(1-2), 477-484.
18. 'BS EN 15620:2000: Welding - Friction welding of metallic materials', British Standards Institution, 2000.
19. H. H. Koo and W. A. Baeslack: 'Structure, properties, and fracture of linear friction welded Al-Fe-V-Si alloy 8009', Materials Characterisation, 1992, 28(2), 157-164.
20. R. J. Harvey, M. Strangwood, and M. B. D. Ellis: 'Bond-line structures in friction welded Al2O3 particle reinforced aluminium alloy metal matrix composites (MMC's) ', Proc. 4th Int. Conf. on 'Trends in Welding Research' (ed: H. B. Smartt, J. A. Johnson, and S. A. David), Gatlingburg, Tennessee, 5-8 June 1995, 1995, ASM, 803-808.
21. T. S. Jun, F. Rotundo, L. Ceschini, and A. M. Korsunsky: 'A study of residual stresses in Al/SiCp linear friction weldment by energy-dispersive neutron diffraction', Advances in Fracture and Damage Mechanics Vii, 2008, 385-387, 517-520.
22. G. T. Murray: 'Introduction to engineering materials: behaviour, properties and selection'; 1993, New York, CRC Press.
23. B. J. Ginn and T. G. Gooch: 'Toughness of 12%Cr ferritic/martensitic steel welds produced by non-arc welding processes', Welding Journal, 1998, 77(8), 341S-349S.
24. W. Y. Li, T. J. Ma, S. Q. Yang, Q. Z. Xu, Y. Zhang, J. L. Li, and H. L. Liao: 'Effect of friction time on flash shape and axial shortening of linear friction welded 45 steel', Mater. Lett., 2008, 62(2), 293-296.
25. T. J. Ma, W. Y. Li, Q. Z. Xu, Y. Zhang, J. L. Li, S. Q. Yang, and H. L. Liao: 'Microstructure evolution and mechanical properties of linear friction welded 45 steel joint', Adv. Eng. Mater., 2007, 9(8), 703-707.
26. W. A. Baeslack, T. F. Broderick, M. Juhas, and H. L. Fraser: 'Characterization of solid-phase welds between TI-6Al-2Sn-4Zr-2Mo-0.1Si and Ti-13.5Al-21.5Nb titanium aluminide', Mater. Charact., 1994, 33(4), 357-367.
27. P. L. Threadgill: 'The prospects for joining titanium aluminides', Mater. Sci. Eng. A-Struct. Mater. Prop. Microstruct. Process., 1995, 193, 640-646.
28. P. L. Threadgill: 'Solid state joining of nickel aluminides: Preliminary study', Proc. 3rd Int. Conf. on 'Synthesis, Processing and Modelling of Advanced Materials' (ed: F. H. Froes, T. Khan, and C. M. Ward-Close), Paris, France, 25-27 June1997, 1997, ASM, 153-158.
29. M. Corzo, O. Casals, J. Alcala, A. Mateo, and M. Anglada: 'Mechanical evaluation of linear friction welds in titanium alloys through indentation experiments', Revista De Metalurgia, 2005, 41(6), 403-409.
30. T. Ma, S. Yang, Y. Zhang, and D. Xi: 'linear rubbing jointing method of aluminium piece and steel piece', Patent: Northwestern Polytechnical University, CN101020275-A, 2007.
31. G. M. Workman and E. D. Nicholas: 'Friction welding aluminium and its alloys to different materials', Metals and Materials, 1986, 2(3), 138-140.
32. H. Y. Li, Z. W. Huang, S. Bray, G. Baxter, and P. Bowen: 'High temperature fatigue of friction welded joints in dissimilar nickel based superalloys', Mater. Sci. Technol., 2007, 23(12), 1408-1418.
33. A. Mateo, M. Corzo, M. Angada, J. Mendez, P. Villechaise, J. P. Ferte, and O. Roder: 'Welding repair by linear friction in titanium alloys', Mater. Sci. Technol., 2009, 25(7), 905-913.
34. P. L. Threadgill: 'Terminology in friction stir welding', Sci. Technol. Weld. Join., 2007, 12, 357-360.
35. D. Helm and G. Lutjering: 'Microstructure and properties of friction-welds in titanium alloys', Proc. 9th Int. Conf. 'Titanium '99' (ed: I. V. Gorynin and S. S. Ushkov), St Petersburg, Russia, 7-11 June 1999, 1999, CRISM, 1726-1733.
36. M. Karadge, M. Preuss, C. Lovell, P. J. Withers, and S. Bray: 'Texture development in Ti-6Al-4V linear friction welds', Mater. Sci. Eng., A, 2007, 459(1-2), 182-191.
37. C. Mary and M. Jahazi: 'Multi-scale analysis of IN-718 microstructure evolution during Linear Friction Welding', Adv. Eng. Mater., 2008, 10(6), 573-578.
38. W. A. Baeslack, P. L. Threadgill, E. D. Nicholas, and T. F. Broderick: 'Linear friction welding of Ti-48Al-2Cr-2Nb (at. pct) titanium aluminide', Proc. of 'Titanium '95 - Science and Technology' (ed: P. A. Blenkinsop, W. J. Evans, and H. M. Flower), Birmingham, UK, 22-26 October 1995, 1996, Institute of Materials, 424-431.
39. M. M. Attallah, M. Preuss, and S. Bray: 'Microstructural development during linear friction welding of titanium alloys', Proc. 8th Int. Conf. on 'Trends in Welding Research' (ed: S. A. David), Pine Mountain, Georgia, 1-6 June 2008, 2008, ASM.
40. J. Romero, M. M. Attallah, M. Preuss, M. Karadge, and S. E. Bray: 'Effect of the forging pressure on the microstructure and residual stress development in Ti-6Al-4V linear friction welds', Acta Mater., 2009, 57(18), 5582-5592.
41. M. R. Daymond and N. W. Bonner: 'Measurement of strain in a titanium linear friction weld by neutron diffraction', Physica B-Condensed Matter, 2003, 325(1-4), 130-137.
42. P. Frankel, M. Preuss, A. Steuwer, P. J. Withers, and S. Bray: 'Comparison of residual stresses in Ti-6Al-4V and Ti-6Al-2Sn-4Zr-2Mo linear friction welds', Mater. Sci. Technol., 2009, 25(5), 640-650.
43. M. Karadge, P. Frankel, A. Steuwer, C. Lovell, S. Bray, P. J. Withers, and M. Preuss: 'Microstructure, texture, local tensile properties and residual stress relief in Ti-6Al-4V linear friction welds', Proc. of 'Materials Science & Technology' Cincinnati, Ohio, 2006, Association for Iron and Steel Technology, 35-46.
44. M. Preuss, J. Quinta da Fonseca, A. Steuwer, L. Wang, P. J. Withers, and S. Bray: 'Residual stresses in linear friction welded IMI550', Journal of Neutron Research, 2004, 12(1), 165-173.
45. P. J. Withers: 'Residual stress and its role in failure', Rep Prog Phys, 2007, 70(12), 2211-2264.
46. F. P. Bowden and D. Tabor: 'The friction and lubrication of solids'; 2001, Oxford classic texts.
47. T. Jun, Z. Tiancang, L. Pengtao, L. Jing, and M. Yu: 'Numerical computation of a linear friction welding process', Physical and Numerical Simulation of Materials Processing, Pts 1 and 2, 2008, 575-578, 811-815.
48. A. Vairis and M. Frost: 'Modelling the linear friction welding of titanium blocks', Mater. Sci. Eng., A, 2000, 292(1), 8-17.