Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

The use of bobbin tools for friction stir welding of aluminium alloys

   
P L Threadgill1, M M Z Ahmed2, J P Martin3, J G Perrett3 and B P Wynne2

1TWI Ltd, Cambridge
2IMMPETUS (Institute for Microstructural and Mechanical Process Engineering: The University of Sheffield)
3TWI (Yorkshire) Ltd

Paper presented at Thermec 2009. Berlin, Germany, 25 - 29 August 2009.

Keywords: friction stir welding; aluminium; bobbin tools; EBSD; microstructure; texture

Abstract

The use of a double sided friction stir welding tool (known as a bobbin tool) has the advantage of giving a processed zone in the workpiece which is more or less rectangular in cross section, as opposed the triangular zone which is more typically found when conventional friction stir welding tool designs are used. In addition, the net axial force on the workpiece is almost zero, which has significant beneficial implications in machine design and cost. However, the response of these tools in generating fine microstructures in the nugget area has not been established. The paper presents detailed metallographic analyses of microstructures produced in 25mm AA6082-T6 aluminium wrought alloy, and examines grain size, texture and mechanical properties as a function of processing parameters and tool design, and offers comparison with data from welds made with conventional tools.

Introduction

Friction stir welding has made remarkable progress since its invention in 1991, in particular for welding aluminium alloys. Although the process has many benefits, there also drawbacks, one of them being the risk of root flaws in single sided welds. One of the most promising ways to avoid these is to use a double-sided or bobbin tool, as this removes the root region. This variant was described in the original TWI patent[1], and is shown schematically in Figure 1. Its advantages can be summarised as follows:

  • Eliminates weld roots, and root defects.
  • Low Z forces on fixture and machine.
  • No backing bar required.
  • Low distortion due to uniform heat input
  • Simple control.
  • Tolerance to thickness variation.
Fig.1. Principle of the bobbin tool
Fig.1. Principle of the bobbin tool

Subsequent work in the USA and elsewhere developed the bobbin tool concept by allowing the gap between the shoulders, and hence the force exerted on the workpiece to be controlled, usually to keep a constant value. However, recent work at TWI has shown that a fixed bobbin tool can give very good results without the sophistication of the variable gap, although care is needed with the tool design.

Welding trials

In order to prove the performance of the bobbin tool, welding trials have been carried out on 25mm thick AA6082-T6 alloy over a range of welding speeds and tool rotation speeds. Welds were examined using standard metallographic techniques including hardness surveys, and selected welds were examined in more detail using electron backscattered diffraction. Welds were made over a range of traversing speeds from 200 to 500 mm/min, at a constant rotation speed of 300rev/min. The tool used is illustrated in Figure 2. This shows a Tri-flat based design, with coarse threads. Both shoulders featured scrolled shoulders, which is usually beneficial to zero tilt angle tools. Sound welds were obtained over a range of welding speeds, from 200mm/min to 500mm/min.

Fig.2. Bobbin tool used in this work
Fig.2. Bobbin tool used in this work

Results

Typical macro sections are shown in Figure 3, and are interesting in that the single onion ring pattern observed in the nugget in conventional welds is replaced by series of three onion rings stacked vertically through the thickness in the probe dominated mid section of the weld. This implies that the motion of the flowing material around the tool is more complex than in a simple tool, but there is no firm evidence to support this. The effect of different probe designs on the onion rings is also not known.

Fig.3. Macrosections from bobbin tool welds at 300mm/min (left) and 500mm/min (right)
Fig.3. Macrosections from bobbin tool welds at 300mm/min (left) and 500mm/min (right)

Figure 3 also shows that the weld shapes differ from welds made with conventional tools, and are slightly hourglass shaped. The width of the nugget in this case is ~20mm, which is as expected slightly greater than the mid-thickness probe diameter of 17mm. Thus, if used for friction stir processing, fewer passes would be required to refine the microstructure over a given area. Mean steady state torque and force data obtained with a bobbin and conventional tools at 300mm/min are shown in Table 1. The X (traversing) force and torque levels are virtually the same as found with conventional welds (although the X force is more erratic), but as expected the Z (down) force is much lower than that found in with conventional tools. In the floating tool variation, where the bobbin tool is allowed to move freely in the Z direction, the Z force would be almost zero. Figure 4 compares hardness values of welds made at 300 mm/min in conventional and bobbin welds, and again there is little difference, except for the shape of the nugget. This similarity is not surprising, as the energy input of the two welds was very similar. The estimated heat input for each weld is 2.99 and 3.35kJ/mm for the bobbin and conventional weld respectively, the main difference arising from the slightly higher rotation speed of 400rev/min used in the conventional weld, compared to 350 rev/min in the bobbin tool weld.

Table 1. Comparison of force and torque data

ToolZ force kNX force kNTorque Nm
Bobbin 14.73 11.72 408.6
Conventional 62.38 10.65 400.2
Fig.4. Hardness maps of welds made with bobbin tool (upper) and conventional tool
Fig.4. Hardness maps of welds made with bobbin tool (upper) and conventional tool

Figure 5 shows the macrosection of the bobbin tool weld investigated by electron back scattered diffraction (EBSD) with the areas investigated indicated at the top and mid-thickness. In addition to the onion rings two further regions can be seen within the nugget (NG), 1) a narrow dark region adjacent to the advancing side (AS) and 2) a wide bright region to the retreating side (RS) of the weld. For EBSD analysis samples of 30mm x 10mm were cut as shown in Figure 5 which were mechanically ground and polished using 6 and 1 mm diamond pastes and a 0.05 µm colloidal silica suspension, followed by electropolishing with 30% HNO3 in CH4OH for 60s at about -15°C at 14 V to ensure a strain free surface. EBSD data were acquired using an FEI Sirion FEGSEM equipped with an HKL Technology EBSD attachment, and operated at 20 kV. Orientation mapping was then performed with a step size of 4 mm. Initial data processing was carried out using HKL Channel 5 software.

Fig.5. Optical macrograph showing the areas analysed by EBSD, indicated by rectangles at the top and the midsection of the weld
Fig.5. Optical macrograph showing the areas analysed by EBSD, indicated by rectangles at the top and the midsection of the weld

Figure 6 shows the orientation image map (OIM) of the area analysed at the top of the NG using inverse pole figure (IPF) colouring with respect to the normal direction (ND) of the weld and corresponding (111) pole figures of 1200 µm wide segments of the map. The crystallographic texture was examined with respect to both the probe and the shoulder shear directions. Literature shows[2-4] that the deformation in the FSW process is dominated by simple shear; however, the shear plane and shear plane normal are different in case of the shoulder and the probe dominated regions. In the former, the shear plane normal is parallel to ND and the shear direction is aligned with shear flow lines generated by the rotating shoulder. In the probe dominated region the shear plane is parallel to the probe surface and the shear direction is aligned with the shear flow lines of the rotating tool. The pole figures (PFs) above the map are in the probe shear configuration and below the map are in the shoulder shear configuration. The OIM shows that the AS side is dominated by a mixture of (001) and (111) orientations whereas the RS is dominated by (110) orientations. In terms of texture it appears that the AS part of the map is more dominated by the shoulder as the seven PFs below the map from the AS are close to the simple shear texture with respect to the shoulder. On the other hand, the RS part of the map is clearly dominated by the probe as the rest of the PFs above the map are showing the standard off-axes simple shear texture as observed at the midsections of conventional tool FSWed AA6082. [5,6]

Fig.6. EBSD IPF map with respect to ND for the upper region in Figure 5. (111) pole figures in 1200 µm steps are shown across the whole map, with ND in the centre above the map and WD in the centre below
Fig.6. EBSD IPF map with respect to ND for the upper region in Figure 5. (111) pole figures in 1200 µm steps are shown across the whole map, with ND in the centre above the map and WD in the centre below

Figure 7 shows the OIM map of the whole NG acquired at the weld midsection. The OIM map is not dominated by a single orientation but rather a mixture of orientations with some areas dominated by (111) grain orientations at the RS. In terms of texture the RS part of the map is dominated by a strong off-axes simple shear texture which progressively weakens towards the AS but is still clearly probe dominated shear texture.

Fig.7. EBSD IPF map with respect to ND for the mid-section region in Figure 5. (111) pole figures in 1200 µm steps are shown across the whole map. Note the pole figures are shown with the ND in the centre and the WD at the bottom
Fig.7. EBSD IPF map with respect to ND for the mid-section region in Figure 5. (111) pole figures in 1200 µm steps are shown across the whole map. Note the pole figures are shown with the ND in the centre and the WD at the bottom

The grain refining observed is shown in high resolution maps of the TMAZ and NG regions in Figure 8. The base material is highly deformed, and a very high level of substructure can be seen in the grain boundary map in Figure 8b for the TMAZ and also in the misorientation angle distribution histogram for the same area (Figure 9). In the NG region the fine grain structure with low level of substructure can be observed from Figure 8d and also from the misorientation angle distribution of the same area shown in Figure 9. The 110 and 111 PFs of the TMAZ part and the NG part are shown in Figure 10 a and b respectively. It can be seen that the rolling texture of the base metal is slightly rotated due to the passage of the tool in the TMAZ. The NG part is mainly simple shear texture.

Fig.8a) IPF map with high angle boundaries (>15°) superimposed: b) grain boundary map with red lines representing low angle boundaries (5 to 15°) and black lines high angle boundaries (>15°) for the TMAZ: c) and d) similar maps for the NG
Fig.8a) IPF map with high angle boundaries (>15°) superimposed: b) grain boundary map with red lines representing low angle boundaries (5 to 15°) and black lines high angle boundaries (>15°) for the TMAZ: c) and d) similar maps for the NG
Fig.9. Misorientation angle distribution for the TMAZ and NG shown in Figure 9a and c
Fig.9. Misorientation angle distribution for the TMAZ and NG shown in Figure 9a and c
Fig.10. 110 and 111 PFs for a) TMAZ b) NG shown in Figure 9a and c
Fig.10. 110 and 111 PFs for a) TMAZ b) NG shown in Figure 9a and c

Discussion

The crystallographic texture was examined near the top surface and the midsection of the bobbin FSW (BFSW) tool welded AA6082 to compare the results with those obtained with conventional FSW (CFSW) tool welds. Typically the NG region of CFSW AA6082 consists of three regions from the surface to the bottom[5]: (1) shoulder affected region, (2) probe and shoulder affected region, and (3) probe affected region for the rest of the weld. In the case of 32 mm thick CFSWs the effect of the shoulder extends to about 8 mm below the top surface.[6] From the texture results obtained at the top surface of the current BFSW tool weld shown in Figure 6 the weld texture has a significant component of probe dominated shear texture. This result would suggest that the probe is almost dominating the deformation process along the thickness of the weld with a minimum effect for the shoulders at both surfaces. The small shoulder effect for the BFSW tool weld could possibly be attributed to the low axial force required in comparison to conventional welds as shown in Table 1, although a significant force is generated by the differential thermal expansion of aluminium and steel, but it is likely that the force is distributed more evenly through the thickness. The texture results obtained at the BFSW midsection clearly show a probe dominated, off axis simple shear texture as found in conventional thick section welds. It should be noted that the texture progressively weakens towards the AS but is still probe dominated shear texture. This appears to correlate with the two vertical regions that were observed on the optical macrograph of the weld shown in Figure 5. This slight variation on the texture across the NG indicates a slight variation on the deformation mode across the transverse cross section of the NG. In contrast, this weakening in the texture is not observed in case of the CFSW welds at the same level where a rather strong shear texture was found all the way across the NG. [6]

In terms of microstructure, the NG of the BFSW consists of a fine grain structure with grain size ranges from 6 to 8µm as illustrated in Figure 8c. It is probable that further development could further reduce this size. With the dominance of the probe on the deformation process through the NG thickness, as confirmed from the texture data, a uniform grain size is formed through the NG in comparison to a varied grain size from the top to the bottom the CFSW tool welds [5], and this should be beneficial in friction stir processing. A further advantage of the BFSW is that the almost rectangular NG profile will mean that fewer passes would be required than for a single sided cone-shaped probe when processing larger areas. The travel speed achieved, and the energy input per unit length, are similar to those found in conventional single sided welds, and so productivity will be similar in both cases.

Conclusions

Detailed study of the microstructure and crystallographic texture of a friction stir weld in 25mm AA6082-T6 shows a favourable fine grained microstructure, in which the complex shoulder dominated region is significantly reduced. There appear to be no drawbacks to using the process, and it is likely that this variant will be of considerable benefit in friction stir processing of aluminium alloys.

Acknowledgements

Work at TWI was funded by Yorkshire Forward. MMZA is indebted to the Egyptian Government for funding. Hardness tests were done by R Kitchen (IMMPETUS).

References

  1. W.M Thomas, et al. 'Improvements relating to friction welding' US 5,460,317 and EP 0 615 480 B1 (1991)
  2. R.W. Fonda, J.F. Bingert, K.J. Colligan. Scripta Materialia 2004; 51, 243.
  3. R.W. Fonda, J.F. Bingert. Scripta Materialia 2007,: 57, 1052.
  4. P.B. Prangnell, C.P. Heason. Acta Materialia 2005, 53, 3179.
  5. M.M.Z. Ahmed, B.P. Wynne, W.M. Rainforth, P.L. Threadgill. 7th International Friction Stir Welding Symposium, Awaji Island, Japan, 20-22 May 2008, TWI.
  6. M.M.Z. Ahmed, B.P. Wynne, W.M. Rainforth, P.L. Threadgill. Scripta Materialia 2008; 59, 507.

For more information please email:


contactus@twi.co.uk