Static shoulder friction stir welding of Ti-6Al-4V; process and evaluation

M J Russell ¹ , P L Threadgill ¹ , M J Thomas ² and B P Wynne ^2
1 TWI Ltd, Granta Park, Great Abington Cambridge CB21 6AL, United Kingdom
² IMMPETUS (Institute for Microstructural and Mechanical Process Engineering: The University of Sheffield), Mappin Street, Sheffield S1 3JD, UK.

Paper presented at 11th World Conference on titanium (Ti-2007), (JIMIC - 5), Kyoto, Japan, 3-7 June, 2007.

Friction Stir Welding (FSW) is now an established production technology for aluminium components, and work is continuing on the development of FSW for high temperature materials including titanium alloys. Titanium is one of the more challenging materials to weld by this process. Recent developments at TWI have led to significantly improved results in the FSW of titanium alloys, particularly in the areas of process control and stability. In particular, the development of static shoulder friction stir welding tools, (which give more uniform heating through the thickness) have had a profound effect on weld quality and ease of weld production. This paper describes the principles of the process, and gives examples of the weld quality that can be obtained. In particular, one development weld has been subjected to a very detailed metallurgical assessment, and this has given very valuable information on the nature of the microstructure and textures, and how they vary in different parts of the weld. This valuable information will provide a standard against which future improvements to the welding process can be compared.

1. Introduction

Friction stir welding is now an established process for aluminium alloys, but its application to titanium alloys has been hampered by the physical and mechanical properties of the material. The poor thermal conductivity of alloys such as Ti-6Al-4V means that heat distribution in the weld is not uniform, and since heat is generated mostly at the upper surface when conventional tools are used there is a significant temperature gradient through the thickness. Friction stir welding is primarily a hot working process, and unlike aluminium alloys, the hot working temperature range of titanium is limited. Therefore, an even temperature distribution is essential for the best results. Tool technology has also presented a significant challenge. Titanium alloys retain significant strength at the hot working temperature. Tool materials must not only be able to withstand the high welding forces and torque levels but must also be inert to titanium at high temperature (typically 1100-1200°C).

This work describes an alternative approach to the friction stir welding of titanium, in which the method of heat generation has been radically altered. The paper describes some preliminary experiences with the new approach, and gives details of the microstructures of a weld made in Ti-6Al-4V.

2. Conceptual design

It was apparent that obtaining a more uniform heat generation profile would require radical changes to tool design. Since the rotating shoulder on conventional tools generates the majority of the heat at the upper surface, experiments were carried out with some alternative approaches, one of which was the static shoulder approach. The welding mechanism consists of a rotating pin located in a non-rotating shoulder component, which slides over the surface of the material during welding. An inert gas shroud is provided by a specially made chamber fixed to the welding head which protects the welding region and trailing edges during welding. Other components for the static shoulder welding head are illustrated by the cross sectional view in Figure 1.

Fig.1. Stationary shoulder FSW mechanism 1) Rotating spindle 2) Draw bar 3) ISO 50 Tool holder 4) Water cooling jackets 5) Argon input 6) Support bearing 7) Stationary tool head 8) Ti workpiece 9) Backing plate 10) Sliding shoe 11) Rotating pin 12) Sliding seal 13) Argon supply 14) Gas chamber 15) Inert gas input

During Static Shoulder FSW (SSFSW), a tool probe (component 11) rotates through a non-rotating shoulder, which is held within a sliding shoe (components 7 and 10). The shoulder components do not directly contribute to the heat generated during welding. This approach enables the process to produce focussed heat input around the tool pin, and eliminates the problem of surface overheating. SSFSW generates a consistent linear heat input throughout the weld cross-section, which is particularly suitable for the welding of low conductivity materials. The overall view of the SSFSW head is shown in Figure 2. A very smooth 'polished' weld surface can be reliably produced for over 1m long weld runs, and an example is shown in Figure 3.

Fig.2. The SSFSW welding head
a) external view	b) underneath, showing gas ports

Fig.3. Surface appearance of a typical weld

3. Results

Initial trials on 6.35mm Ti-6Al-4V showed very encouraging results. The process was remarkably stable, and the weld surface appearance was very smooth, and much better than found in conventional friction stir welds.

Welds were made under a variety of conditions. Good results were obtained with speeds of 60-80mm/min, using rotation speeds of 400-500rev/min. Force control was used to maintain a constant pressure on the welding tool. Clearly the non-rotating shoulder will experience very high stresses and temperatures, and research on the optimum material for this component is continuing. Materials tested have included metallic and ceramic parts, the latter seeming to offer more promise.

Mechanical properties, as measured by room temperature cross weld tensile tests are as expected. 100% joint efficiency can be obtained, although there is a reduction in ductility when failure occurs close to the weld zone. In many samples, the failure was remote from the weld, in which case elongation equalled parent plate values. The status at the moment is that of a successful technology demonstration, and further studies are in hand to optimise a number of aspects of the process.

A macro section of a typical weld is shown in Figure 4. It can be seen that the weld is defect free, but that the thermal profile was not quite parallel. This was due partly to a small taper on the probe, but mainly to the loss of heat through the tool bed. Nevertheless, this profile is considered satisfactory, and a significant improvement on what can be obtained without a static shoulder.

Fig.4. Macro section of a weld made at 500rev/min and 80 mm/min travel speed

A more detailed examination of the weld microstructure of a weld made at 400rev/min and 60mm/min using high resolution electron backscattered diffraction (EBSD) is shown as an orientation image map in Figure 5a. The EBSD data was acquired using a 3µm step size in the plane perpendicular to the travel direction with an FEI Sirion FEGSEM equipped with an HKL Nordlys CCD camera controlled by HKL Channel 5 acquisition software. The map, which covers the top 4 mm of the whole weld, shows distinct orientation regions across the weld which are labelled A to K and a surface layer labelled L. Regions A and K are in parent material and contain a strong rolling texture and a spheroidised alpha microstructure. There appears to be some slight coarsening of the microstructure on both sides of the weld as the weld is approached, but there is no evidence of deformation. This is confirmed by the texture in region J (retreating side) which is almost identical to that of the parent material. Just beyond J there is a subtle orientation change that is approximately 250µm wide (Region I). The pole figure of this region has changed but has many similarities with the parent material suggesting that this region has perhaps been heated above the beta transus, (~990°C), but has experienced no deformation. To examine this further the map in Figure 5a has been plotted ( Figure 5b) showing only boundaries with a misorientation greater than 12° and which also do not correspond to a misorientation between alpha transformation variants from the same beta grain 2). This highlights the position of the prior beta grain boundaries. Thus it can be concluded from Figure 5b that region I has experienced significant beta grain growth.

Fig.5 a) An Euler contrast orientation image map and b) A beta grain boundary construction map using Burgers misorientation. {0002} pole figures represent alpha phase textures in the regions highlighted in the map. Pole figure L is the texture for the surface layer of the weld. Each contour in the pole figures represents 1x random intensity. Regions A and K are on the advancing and retreating side, respectively.

This is further confirmed by high resolution maps of regions I and J shown in Figures 6a and 6b. These maps clearly show the as-rolled microstructure in region J and a fully acicular alpha structure transformed from large equiaxed beta grains in region I. The adjoining region, H, also has a similar transformed microstructure but the texture has become diffuse suggesting that the prior beta grains have been slightly plastically deformed before transformation back to alpha. Hence, we believe region H defines the edge of the thermo-mechanically affected zone (TMAZ). A similar situation also exists on the advancing side (the detailed analysis is not shown in Figure 5) except that the equivalent regions are slightly larger. Work is now ongoing to confirm these observations with emphasis on static heat treatments above and below the beta transus of the parent material to identify the textures of heat affected only material.

Fig.6 a) High resolution orientation image map (Euler contrast) of retreating side of the weld in Figure 5 b) Beta grain boundary construction of a) using Burgers misorientation. Highlighted boundaries represent a misorientation >12°.

For material closer to the weld centre the microstructure again consists of an acicular alpha microstructure but the width of the similarly orientated regions is of the order of 2mm. In all cases these regions have relatively distinct textures clearly different from the parent but with intensities at the same level of the parent's texture. It is unclear at this stage, however, if these textures are transforming from deformed or recrystallised beta grains. Work is now ongoing to reconstruct the beta morphology and texture in order to gain evidence for distinct TMAZ (non-recrystallised beta) and nugget (recrystallised beta) zones. Information already in hand ( Figure 5b), however, shows the beta grain size to be relatively uniform at approximately 50µm throughout the whole cross-section of the weld suggesting that there are no major regions of accelerated grain growth within the weld.

For material at the surface of the weld, region L, there is a significant change in texture from the bulk with the development of a strong transverse basal texture to a maximum depth of about 250µm. The microstructure is also changed with a considerable reduction in prior beta grain size. It should also be noted that this surface layer extends beyond the pin affected region of the weld to fit exactly the area underneath the stationary shoulder and, therefore, we believe this layer has its origin in the shear force generated by the forward motion of the stationary shoulder. Moreover, below this layer the texture immediately returns to the textures described previously indicating that the effect of the stationary shoulder on microstructure evolution is limited to a relatively thin surface layer.

On a more bulk scale, the complete textures of the advancing and retreating sides have been analysed and they show an interesting symmetry relationship. If there was no forward motion in the weld then the advancing side's texture would look the same as the retreating side's texture when viewed from the opposite face of the viewing plane. This was not the case for the current weld but the textures were similar and they were made nearly identical, as shown in Figure 7, by a rotation of about 30° about the vertical direction of the weld.

Fig.7 {0002} pole figures of a) advancing side and b) retreating side of weld looking from behind the acquisition plane and rotated 30° about the vertical direction of the weld

A deeper analysis of these relationships will determine if these texture symmetries can be related to a macroscopic symmetry within the welding process itself.

4. Discussion

The weld examined in this paper is a development weld, and modifications to the tool/process design and/or welding parameters may result in significant differences in microstructure. It is arguably the most detailed analysis of a friction stir weld in Ti-6Al-4V, and further analysis of the data is in progress. However, it is a valuable exercise to take a good development weld and examine it in detail as a yardstick against which possible improvements can be measured. The work has demonstrated that friction stir welds of very high quality can now be made in Ti-6Al-4V, and it is reassuring to see that there are no unexpected features in the microstructure. Moreover, it has been shown that EBSD is the ideal tool for generating high quality statistically quantitative data of benchmark microstructures in order to help in the development of this technology.

5. Summary

This paper has described the principal points of the static shoulder variant of friction stir welding for titanium alloys, and the advantages it confers over more traditional approaches. Key aspects of the microstructures produced in Ti-6Al-4V have also been highlighted as follows:-

1. the microstructure is relatively uniform throughout the whole cross-section of the weld with a prior beta grain size of the order of 50µm which produces a fine transformed acicular microstructure.
2. the effect of the shoulder is limited to a very thin surface layer.
3. the TMAZ/HAZ interface most likely occurs in a weld section with temperature above the beta transus.
4. there are distinct texture zones across the weld which are approximately 2 mm thick.

References

1. M J Russell and C Blignault: Proc 6th Int. Symp on Friction Stir Welding, ed. by P L Threadgill, TWI, (Saint-Sauveur, Quebec, Canada), 10-13 Oct 2007, Paper 14.
2. N. Gey and M. Humbert: Acta Materialia 50 (2002) pp 277-287.