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

Subscribe >
Skip to content

Friction stir welding of magnesium alloys (March 2003)

   
Richard Johnson and Philip Threadgill

Paper presented at TMS Symposium on Magnesium Technology, San Diego, CA, USA, 2-6 March 2003.

Abstract

An investigation has been carried out on the friction stir welding of four magnesium alloys. These consisted of one wrought and three die-cast magnesium alloys, including those containing manganese such as AM50 and AM60, and alsozinc such as AZ91 and AZ31 (wrought material). All of the alloys have each been successfully welded to themselves and also to each other, without any problems from the trapped gases in the cast materials, but the tolerance box of processing parameters to ensure that sound welds are produced has been found to be more restrictive than those seen in friction stir welding aluminium alloys.

Introduction

Magnesium is one of the most abundant elements in the earth's surface, with virtually inexhaustible supplies in the oceans. Over recent years the industrial output of magnesium alloys has been rising by almost 20% per annum, which is faster than that of any other metal. The increased use of aluminium and magnesium alloys is of great interest to the automotive industry, with the goal of reducing the weight of road vehicles to make them more fuel-efficient or to increase the vehicle specification without adversely affecting its fuel efficiency.

In recent years there has been a renewed interest in the use of magnesium parts for body components, many of which have made by pressure die casting. [1,2,3] These have limited ductility, contain gas occlusions, and are frequently difficult to weld satisfactorily by fusion welding techniques. With the major proportion of magnesium alloys being made by casting there has not previously been an extensive need for improved weldability to be developed. The solid state joining technique of friction stir welding (FSW) Fig.1, was patented in 1991, [4] and was initially used to extend the weldability of aluminium alloys, some of which were difficult to join by fusion welding techniques because of cracking and porosity problems. The scope of this new welding process has since been extended to the welding of lead, zinc, copper, titanium and ferrous alloys, with some considerable success.

Fig. 1. Friction stir welding process
Fig. 1. Friction stir welding process

 

Of particular interest to the joining of cast magnesium parts is the success of FSW cast and wrought aluminium materials together. The cast material contained significant porosity, indicating trapped gases to be present, but whereas fusion welding methods would have encountered problems from their presence, FSW created a sound weld with no porosity in the weld bead or the immediate HAZ, [5,6] Fig.2. TWI has been studying the FSW of magnesium alloys, and is currently participating in a European collaborative project, MagJoin, to assess the weldability of magnesium alloys to themselves, to each other, and toaluminium alloys.

Fig. 2. Aluminium cast-wrought weld
Fig. 2. Aluminium cast-wrought weld

 

Experimental results of friction stir welding trials on magnesium alloys

The work undertaken at TWI has concentrated on four magnesium alloys: three cast alloys, AM50, AM60 and AZ91; and one wrought alloy, AZ31. The AM alloys contain aluminium and manganese as the major alloying elements, while the AZalloys have aluminium and zinc as the major alloying elements, and all of these alloys are of interest for applications in the automotive industry.

The trials were performed on small plates, 140 x 100 x 6mm in size, except that the AZ31 plates were slightly thicker at 6.4mm. The latter plates were not machined down to match the other plate thicknesses, but the FSW tool was setfor a full penetration weld in the 6mm materials, and was allowed to plough slightly into the thicker AZ31. The FSW tools used were of the plain threaded pin or the MX Triflute TM design, and a range of rotation and traverse speeds were investigated.

The first trials were performed on the AM50 alloy plates, and the rotation speed was varied from 250-500rpm, and the traverse speed varied from 160-450mm/min. It was determined that 355rpm and 160-224mm/min resulted in sound welds,and 355rpm was adopted as the standard rotation speed for the other magnesium alloys to be welded, so that a direct comparison of the weld quality could be established. In these other alloys the traverse speed was varied in the range160-315mm/min, and 160mm/min selected as the best with which to compare full penetration welds.

It was found relatively easy to weld the AZ91 and AZ31 alloys to themselves, and smooth weld surfaces were achievable, particularly with the AZ31 alloy. However, with the AM50 and AM60 alloys, although some smooth weld beads were achievable, under certain conditions these materials were found to adhere to the FSW tool, and would initially cause a slightly torn appearance to the weld surface, and then the FSW tool began to gouge material out of the plates toleave a surface void, Fig.3. It was also found that the wrought alloy AZ31 and the softest cast alloy AM50 could be processed at slightly higher traverse speeds than the other two alloys.

Fig.3a) AM50, 250rpm-224mm/min
Fig.3a) AM50, 250rpm-224mm/min
Fig.3b) AZ31, 355rpm-160mm/min
Fig.3b) AZ31, 355rpm-160mm/min
Fig.3c) AM50, 355rpm-224mm/min
Fig.3c) AM50, 355rpm-224mm/min
Fig.3d) AM60, 355rpm-315mm/min
Fig.3d) AM60, 355rpm-315mm/min

Fig. 3. Weld surfaces of magnesium alloys welded to themselves

 

The weld macrosections revealed that at a traverse speed of 160mm/min the AZ91 weld nugget was quite parallel-sided, whereas the other three alloy weld nuggets were broader in shape. At 224mm/min the AZ31 weld nugget remained broad whereas the AZ91 and AM60 weld nuggets were now more parallel-sided and possibly beginning to leave a potentially unwelded root at the bottom of the weld. The AM50 weld nuggets also became more parallel-sided at the higher traversing speeds, and at 315mm/min the FSW tool penetration was quite evidently reduced and leaving a potentially unwelded root, Fig.4.

Fig.4a) AZ91, 355rpm-160mm/min
Fig.4a) AZ91, 355rpm-160mm/min
Fig.4b) AM60, 355rpm-160mm/min
Fig.4b) AM60, 355rpm-160mm/min
Fig.4c) AZ31, 355rpm-224mm/min
Fig.4c) AZ31, 355rpm-224mm/min
Fig.4d) AM50, 355rpm-315mm/min
Fig.4d) AM50, 355rpm-315mm/min

Fig. 4. Macrosections of 6mm plate magnesium alloys welded to themselves

 

The welding of the dissimilar magnesium alloys to each other were initially performed at a traverse speed of 160mm/min, the speed at which the full penetration of the FSW tool had been verified in all of the alloys. In general the weld surfaces were quite smooth, except that the AZ91-AZ31 weld appeared somewhat rougher. Selected macrosections of these welds, as seen in Fig.5, show that the AM50-AM60 weld is quite parallel-sided, as are the welds between these alloys and AZ91, but all of the welds with the AZ31 revealed a significantly broader weld nugget.

Fig.5a) AZ91-AZ31 showing rough weld surface
Fig.5a) AZ91-AZ31 showing rough weld surface
Fig.5b) AZ91-AM50 weld
Fig.5b) AZ91-AM50 weld
Fig.5c) AZ91-AZ31weld
Fig.5c) AZ91-AZ31weld
Fig.5d) AM60-AZ31 weld
Fig.5d) AM60-AZ31 weld

Fig. 5. Dissimilar alloy 6mm plate weld macrosections, 355rpm-160mm/min


When welding the dissimilar magnesium alloys together, it was found that in some pairings the alloys could be stirred into each other with equal ease, but in others there was a marked difference in the appearance of the weld surfaces depending on the which alloy was stirred into the other. In order to emphasise this effect, a series of welds was performed at a higher traverse speed than the initial dissimilar alloy welding trials, this time at 250mm/min,and the results can be seen in Fig.6. In particular it was found preferable to stir the other alloy into AZ31 rather than the opposite way round, while this orientation effect was not seen to be as critical with the welds made without AZ31. It should benoted that a similar effect has been noted in the FSW processing of aluminium alloys, [7] and is probably caused by the asymmetric nature of the FSW process and the different strengths and flow stresses of the two materials being welded at the elevated temperature.

Fig.6a) AM50-AM60
Fig.6a) AM50-AM60
Fig.6b) AM50-AZ31
Fig.6b) AM50-AZ31
Fig.6c) AM50-AZ91
Fig.6c) AM50-AZ91
Fig.6d) AM60-AZ31
Fig.6d) AM60-AZ31
Fig.6e) AM60-AZ91
Fig.6e) AM60-AZ91
Fig.6f) AZ91-AZ31
Fig.6f) AZ91-AZ31

Fig. 6. Weld surfaces from dissimilar magnesium alloy welding trials

 

The macrosections of the dissimilar alloy welded joints showed a fine intermixing of the two alloys. In particular some aluminium-manganese intermetallic particles were clearly visible in the AZ31 alloy, and were relatively unaffected by the FSW process. These particles served to show clearly that they remained in the AZ31 material even when the intermixing was extremely fine, such as at the bottom of the weld nuggets, although occasionally they were found to be situated along the interfaces between the alloys, Fig.7.

Fig.7a) intermixing of alloys
Fig.7a) intermixing of alloys
Fig.7b) intermetallic particles in AZ31 alloy
Fig.7b) intermetallic particles in AZ31 alloy

Fig. 7. AM60-AZ31

 

Hardness measurements were made on selected welds at a load of 2.5kgf, and traverses made at both 2mm and 4mm below the top surfaces of the welds. This was to check if there were significant differences between the top of each weldnugget, where the tool shoulder affects the weld nugget width, and the lower region where the pin alone dictates the nugget width. On the single alloys welded to themselves the plots showed some degree of scatter, with in general alittle hardening discernible in the weld nugget region, presumably due to the fine grain size there, but no great differences between the 2mm and 4mm readings. The exception, as expected, was AZ31, which derives much of its strength from being rolled as a wrought product, and here the weld nugget showed a reduced hardness with respect to the parent material. Hardness measurements were also made on the dissimilar alloy joints, where there was again a degree of scatter seen in the plots. In general there did not appear to be any significant hardening other than by the grain size refining of the weld nugget, and only slightly lower values in the HAZ regions, which can be seen in Fig.8.

sprjmar2003f8a.gif
sprjmar2003f8b.gif
sprjmar2003f8c.gif
sprjmar2003f8d.gif

Fig. 8. Hardness traverses of like-like and dissimilar alloy welds

 

Additionally some tensile tests were made on the magnesium alloys welded to themselves, and these showed that for AM50, AM60 and AZ91 the proof stresses of the welded test specimens were quite similar to those of the parent plates,and that the ultimate tensile stresses were about 10-15% lower for AM50 and AZ91, but slightly higher for AM60. With the AZ31 welds, because much of the parent strength is derived from work hardening during fabrication, it was not surprising to see that both proof and ultimate tensile stresses in the welded specimens were appreciably lower. In all cases the elongations were lower in the welds, as expected when introducing a slightly harder weld nugget and twoHAZ regions.

Discussion

The preliminary FSW processing trials on the four selected magnesium alloys have shown that, although three of them are made by high pressure die casting, they can be welded together satisfactorily and without any problems from the trapped gases contained in the materials. Welding each alloy to itself could be readily performed, although the processing parameters were somewhat lower than would be the case in aluminium alloys of the same thickness. The AM alloys were found to tend to adhere to the FSW tool during welding, and this could lead to surface defects on the resultant weld surfaces, and this was seen with plain tool steel FSW tools and also with those coated with titanium nitride. Itis intended to assess whether or not other tool coatings might reduce the effect.

When welding dissimilar magnesium alloys together, again there were some problems with the AM alloys adhering to the FSW tool. Additionally, it was found that placing one alloy on the advancing side of the weld was not always as successful as placing it on the retreating side of the weld. This was especially found to be the case with the wrought AZ31 alloy, where better quality weld surfaces resulted from that alloy being placed on the retreating side of the weld. The FSW process is asymmetrical, because of the tool rotation, and the heating effects on the advancing and retreating sides of the joint will necessarily be different. It is interesting that the next phase of the MagJoinprogramme is to try to weld aluminium and magnesium alloys together, and the first trials have shown the same effect - that placing the aluminium on one side of the joint can result in a different weld quality from placing it on the other side. However, a more systematic study of the FSW processing parameters is required before definite conclusions can be drawn as to how these dissimilar materials may best be welded together.

This work has demonstrated the suitability of friction stir welding for joining magnesium, and it is expected that interest in the method will increase. The process is suitable for adaptation to robotic welding, and there is anincreasing volume of literature that has demonstrated this, for a variety of robot designs. Similarly, 5 or 6 axis CNC systems can also be used, and at least one such system for complex three dimensional welds has been delivered intothe aerospace industry. The use of such flexible systems will be of value to the automotive industry, as few joints are one-dimensional, as used here.

This work has only described butt welds, although work on aluminium alloys has demonstrated very well that the process can be adapted for use with lap welds, although greater care is required in specifying procedures and developing parameters and tool designs to avoid various defects which can form at the interface. Preliminary studies on lap welds in magnesium have proved encouraging, and there is no reason to assume that this should not be reduced to commercial practice.

One drawback of friction stir welding is the need for some backing system to react the downforce. This is simple in straight line one-dimensional welds, but clearly more complex in three dimensional welds. However, with high part numbers expected in automotive applications, the cost of this is likely to be acceptable. For lap welds, which are never fully penetrating through all the layers, the components themselves may contribute to the rigidity, thus reducing, although not eliminating, this requirement.

At present, travel speeds which can be achieved with magnesium alloys are very much less than obtainable with aluminium alloys. This is believed to be a consequence of the close packed hexagonal crystal structure, and limited slip systems compared to aluminium alloys. However, the speeds currently obtainable are economically viable, particularly when assessed against the high quality obtainable in the welds. Efforts are in place to improve these speeds, and hence the productivity which can be obtained. A further minor contribution to process economics is the tool life. There is no critical evaluation available to determine expected tool life, but experience gained at TWI has indicated that this is not likely to be a problem. However, the need for periodic tool cleaning in some alloys, as mentioned earlier, is noted.

Concluding remarks

A study of friction stir welding four common magnesium alloys has been conducted, in which alloys have been welded to themselves and to each other. It has been possible to develop procedures giving sound welds for all combinations,and initial indications are that mechanical properties will meet expectations.

Acknowledgements

The author acknowledges that this welding research has been possible through the internal TWI Corporate Research Programme. This has also enabled TWI to be a partner in a three-year European collaborative project, MagJoin (GROWTH Project No. GRD1-1999-10918), to investigate the joining of magnesium alloys to themselves and to aluminium alloys, particularly for potential automotive applications. TWI has worked solely on the development of FSW in this project,while other partners have investigated different welding techniques.

 

Table 1: Tensile test results of parent and welded magnesium alloy plates.

SampleProof Stress
[MPa]
Ultimate Stress
[MPa]
Elongation
[%]
Fracture location of welded specimens
AM50, typical [8] 125 230 15  
AM50, parent 104 218 10.0  
AM50, parent 117 215 10.5  
AM50, FSW 115 180 4.5 Parent/HAZ
AM50, FSW 110 164 4.0 Weld
 
AM60, typical [8] 130 240 13  
AM60, parent 121 173 *  
AM60, parent 77 103 1.0  
AM60, FSW 118 190 4.5 Parent/HAZ
AM60, FSW 110 198 6.5 Parent/HAZ
 
AZ91, typical [8] 160 250 7  
AZ91, parent 144 203 2.0  
AZ91, parent 150 201 2.0  
AZ91, FSW 157 177 2.0 Parent/HAZ
AZ91, FSW 153 183 1.5 Parent/HAZ
 
AZ31, typical [8] 200 255 12  
AZ31-H24, typical [8] 220 290 15  
AZ31, parent 219 288 6.5  
AZ31, parent 205 292 10.0  
AZ31, FSW 127 201 3.5 Weld
AZ31, FSW 124 201 4.0 Weld

* fractured outside gauge length
-H24 indicates the material has been strain-hardened and partially annealed

References

  1. S. Juttner: Return of the Light Alloy Brigade - Welding of Magnesium Alloys, Welding & Metal Fabrication, Vol 66, No 1 (1998).
  2. G. S. Cole: The Potential for Magnesium to Reduce Vehicle Mass by 100kg, Automotive Light Metals, Vol 1, Issue 1 (2001).
  3. H. Friedrich and S. Schumann: Turning Vision to Reality for the Second Age of Magnesium, Automotive Light Metals, Vol 1, Issue 1 (2001).
  4. W. M. Thomas et al: Friction Stir Butt Welding, GB Patent Application 9125978.8, December 1991, (1991).
  5. R. Johnson: Friction Stir Welding for Castings, Aluminium Castings Conference, Northampton (1998).
  6. S. W. Kallee and A. Mistry, Friction Stir Welding in the Automotive Body-in-White Production, 1 st International Symposium on Friction Stir Welding, Thousand Oaks (1999).
  7. C. J. Dawes et al: Development of the New Friction Stir Technique for Welding Aluminium, TWI GSP 5651 Project, Phases I, II and III (1994-97).
  8. M. M. Avedesian and H. Baker: Magnesium and Magnesium Alloys, ASM Speciality Handbook, ASM (1999).

For more information please email:


contactus@twi.co.uk