Friction stir welds in magnesium alloys - just how good are the mechanical properties?

TWI Bulletin, March/April 2001

After obtaining his BEng degree and PhD in Metallurgy and Materials Science at Liverpool University in 1992, Scott Lockyer worked as a Research Fellow at both Nottingham University and Oxford University before joining TWI in 1999. He is currently a Senior Project Leader in the Structural Assessment section and is, among other things, responsible for the structural integrity of non-ferrous materials within the Structural Integrity Technology Group.

Friction stir welding (FSW) was invented and patented by TWI in 1991, and has emerged as a very important process for the welding of aluminium alloys. As part of the ongoing research into this, FSW trials have been performed on a number of other materials, including magnesium alloys. Scott Lockyer reports on the results of a recent investigation into the mechanical properties of friction stir welds in two magnesium alloys.

Magnesium consumption for structural applications in 1997 was 101 000 tonnes, approximately 30% of the total consumption, ^[1] and all predictions indicate that this will rise. One major reason for this predicted increase is the need to reduce automobile fuel consumption by reducing vehicle weight. For example, Brown ^[2] has estimated that a 10% saving in vehicle weight could result in an increase in fuel economy of 5.5%. The weight saving made possible by using magnesium alloys (instead of cast iron) for the crankcase and transmission housing for the Volkswagen Beetle was estimated at 50kg. It can therefore be seen that dramatic weight savings are possible. The average amount of magnesium alloys currently used in automobiles is 3kg ^[1] and estimates of future usage predict this to increase.

Most current applications for magnesium alloys make use of its excellent castability to produce housings and casings. The main mechanical property requirements for such applications are stiffness and static strength. Other potential applications for magnesium alloys are, however, more structural and will involve fatigue loading; examples are engine cradles, dash panels, door liners and possibly suspension components. These components can be produced by die-casting, however large castings can be prohibitively expensive due to high capital costs. The ability to fabricate such components from smaller sub-assemblies extends the opportunity to exploit the benefits of magnesium alloys. Welding processes are commonly used, due to their flexibility and the better distribution of stress achieved when compared to that caused by mechanical fasteners.

Magnesium alloys are currently welded using arc techniques such as gas tungsten arc welding (GTAW or TIG) and gas metal arc welding (GMAW or MIG). Whilst reasonable welding speeds can be achieved, problems can be experienced with high welding residual stresses, and changes in metallurgical structure due to melting and resolidification. High purity shielding gases are also necessary to prevent weld contamination. Magnesium alloys can also be laser or electron beam welded, although similar melt zone problems can occur, in addition to porosity in laser beam welding. A further complication is that electron beam welding generally has to be performed under a vacuum. The novel process of friction stir welding offers considerable promise for joining magnesium alloys. As part of TWI's ongoing programme of research into FSW, welds have been produced in a range of materials including aluminium, steel, titanium, zinc, copper, lead and magnesium alloys.

Friction stir welding

Friction stir welding was invented and patented by TWI in 1991, ^[3] and has been applied to a wide range of welding applications, to date mainly in aluminium alloys. The principle of the process is summarised in Fig.1. The process involves rotating a tool consisting of a cylindrical shoulder, and a probe feature. During welding the shoulder of the tool acts on the top surface of the plates, with the smaller probe feature plunged between the plate edges. Heating occurs initially by friction, and subsequently by adiabatic or material shear and generates considerable quantities of heat, which cause the plate materials under the tool to be softened. As the rotating probe is moved through the joint line, this softened material is forced from the region in front of the probe to the region directly behind the probe and creates a consolidated solid-phase joint. The shoulder contains the hot plasticised material under the tool, assisting the forging action of the probe. Heat generation by material shear relies on the resistance to shear of the parent material, which reduces as the melting point is approached. FSW is therefore a self regulating non-fusion process, and is suitable for joining dissimilar materials, and alloys which are generally difficult or impossible to weld by fusion processes such as 2xxx and 7xxx aluminium alloys.

Friction stir welding is now well established, and industrially accepted, for the joining of Al alloys and is now being applied to many other alloy systems, including copper, titanium, steels, lead, zinc and magnesium. It is believed that FSW has the potential to make a significant impact in the field of magnesium joining.

Experimental method

The materials under investigation were the commonly used die casting alloys AZ91D and AM50A. All welds were made on a converted milling machine, using a relatively simple FSW tool design with a shoulder diameter of 15mm. The principal parameters studied were the rotation speed of the tool and the travel or welding speed. The test pieces consisted of die cast coupons, nominally 3mm x 100mm x 140mm in size. The edges of the test pieces to be welded were milled shortly before welding to give a clean surface that was perpendicular to the plate surface. For the initial trials, two test pieces were welded together, but for the majority of this work, four pieces were welded together to maximise usage of material. At the start of the weld, the rotating tool was plunged into the workpiece until the shoulder made contact with the upper plate surface, a dwell time of a few seconds was then allowed to generate sufficient heat to start welding. Welds were subjected to a visual examination before being sectioned for mechanical and metallurgical testing.

The effects of the different welding conditions on the microstructure and mechanical properties of the material were investigated by metallographic investigation, hardness and tensile testing. Metallographic sections were taken from each panel to investigate the microstructural changes caused by welding. The sections were polished and etched for examination using optical microscopy. Hardness surveys were performed at (mid-thickness) in the parent material, the thermomechanical affected zone (TMAZ) and the weld centre using a Vickers indentor with a 2.5kg load. Tensile tests were also performed to identify the most promising welding conditions. Fatigue tests were performed on specimens taken from the best performing weld conditions, and on parent material samples. The weld specimens were dressed to remove any welding flash to encourage failure within the weldment. The specimens were tested to failure on an Amsler Vibrophore at a test frequency of approximately 100Hz and a stress ratio of R = 0.1. Samples that achieved lives of greater than 10 million cycles were classed as runouts, and were retested at a higher stress. The specimens were examined post-failure, using optical and scanning electron microscopy to determine the cause and location of failure ie weld defect, parent material casting defect, etc.

Experimental results

An initial programme was undertaken to develop welding procedures that gave welds of reasonable quality. The parameters used in this work are based on these results and summarised in Table 1. The appearance of the welds was generally good, with a smooth finish, and minimum flash formation: a typical example is shown in Fig.2. The metallographic sections are shown in Figs 3 and 4 for AZ91 and AM50 respectively. The triangular weld region is typical of a friction stir weld and corresponds approximately to the region mechanically stirred by the welding tool.

Table 1: Preliminary welding parameters (non-optimised)

The microstructure within the weld region consists of fine equiaxed grains, with no evidence of the original dendritic structure, as can be seen in Fig.5. This refined central zone, known as the weld 'nugget' is thought to be formed by dynamic recrystallisation caused by the high strains and thermal treatment that occur during the welding process. There is some evidence of slightly larger grains near to the top surface of the weld, these are probably due to the more severe thermal cycles experienced in this region.

The hardness surveys ( Fig.6) show that the two alloys appear to exhibit similar behaviour. The AZ91 results show slightly increased hardness in the weld region, and the minimum hardness measured is that of the parent material. The AM50 results also show an increase in the hardness through the TMAZ and into the weld region, but to greater extent than seen in the AZ91 alloy.

The tensile properties of the parent material are detailed in Table 2. The results for the parent material exhibit some scatter, especially in the AM50 alloy where reduced proof stress and ductility were observed for one specimen. The effect of welding speed on the tensile properties of friction stir welds in AZ91 and AM50 are shown in Figs 7 and 8, and the tensile properties are detailed in Tables 3 and 4 respectively. The tensile properties of the welded AZ91 alloy start lower than those of the parent material and then improve with increasing welding speed up to 120 mm/min, after which they appear to decrease slightly. In all but two of the AZ91 tensile tests, failure occurred in the parent material. The situation was different for the AM50; the tensile properties were lower than in the parent material, especially the % elongation but were not found to vary significantly with welding speed. The failure locations in this alloy were equally distributed between the parent material, weld and TMAZ.

Table 2: Parent material tensile properties

Table 3: AZ91 FSW tensile properties

Table 4: AM50 FSW tensile properties

The results of the fatigue tests are plotted in Figs 9 and 10 as S-N curves and are also detailed in Tables 5 and 6. The data from the parent material and weld are plotted for comparison. There is scatter in the fatigue data for both alloys. The AZ91 and AM50 parent materials exhibited similar fatigue behaviour, with an endurance limit between 60-70 N/mm ². The weld behaviour, however, was different, with the AZ91 results being very close to the AZ91 parent material, if slightly lower. Also plotted is the scatterband of fatigue results for AZ91E from Geary. ^[4] It can be seen that both the parent and weld results lie either side of the lower limit of the scatter band. The welded AM50 results were much lower than those of the AM50 parent material and a great deal of scatter is observed. Also shown are data from Draugelates et al ^[5] on parent and non-vacuum electron beam (NVEB) welds in AM50A. The parent material results are much lower than those found in this study, whereas the weld results lie close to the NVEB weld results.

Table 5: AZ91 FSW fatigue data

Table 6: AM50 FSW fatigue data

Discussion

The hardness data for the two alloys showed similar behaviour: both the AZ91 and AM50 alloys exhibited an increase in hardness from the parent material to the weld centreline. The increase in hardness is most likely to be due to refinement of the as-cast microstructure. Prior to welding, the parent material consists of a dendritic as-cast microstructure with both porosity and intermetallic particles at the dendrite boundaries. However, after welding, a very fine equiaxed microstructure is present; the intermetallic particles are much reduced in size and no evidence of porosity is observed. This microstructural refinement will result in an increase in the proof stress and therefore hardness. This behaviour is in contrast to that observed in friction stir welded AZ31 hot rolled plate by Nagasawa and Otsuka ^[6] where no increase in strength occurred. This was attributed to both the parent material and weld metal possessing a fine grain structure.

The scatter in the parent tensile test results reflected the presence of casting defects within the die cast parent material blanks. It would appear from the tensile test results and the occurrence of parent material failures in the tensile tests, that welding conditions used for the AZ91 alloy were close to optimum for this material and tool design. This was not the case for the AM50 alloy however, where difficulties were experienced in producing high quality welds in this limited study. The proof strengths of the AM50 welds produced were found to be reasonable, but the UTS and % elongation were some way short to the parent material values. It should be noted that a relatively simple tool was used to manufacture the welds in this preliminary study and current studies at TWI suggest that more complex tools can produce improved weld properties, as well as faster welding speeds for both alloys.

The fatigue behaviour of the two alloys correlates with the trends identified in the tensile tests. The results for the AZ91 parent material and weld were very close, which is a further indication that the welds were very good and demonstrated the potential for friction stir welding of Mg alloys. Although the parent data were slightly low when compared to the scatter band for AZ91E from Geary, ^[4] this is probably a result of small casting defects that were observed on the surface of the die cast coupons used in this study. The fatigue cracking in the welded specimens was found to initiate either at casting defects in the parent material or at welding defects if any were present. In the few cases when failure occurred in the weld, the joint still possessed reasonable fatigue strength. The AM50 parent material results were similar to the AZ91 parent results, although the welded results were some 40-50% lower and reflect the non-optimised weld procedure. The weld procedure for this material requires further development to obtain improved properties. The current results do however, compare favourably with NVEB weld data from Draugelates et al. ^[5]

Conclusions

• The magnesium alloys AZ91 and AM50 can be successfully joined using FSW, using welding speeds comparable to conventional fusion processes.
• FSW refines the as-cast microstructure and results in improved mechanical properties within the weld.
• The mechanical properties and fatigue strength of friction stir welded AZ91 and AM50 can approach those of the unwelded parent materials.

Further investigation is required to optimise weld procedures for magnesium alloys using new tool designs to improve welding speeds and mechanical properties.

Acknowledgements

The author would like to thank Mike Russell for supplying the friction stir welds and for his assistance writing this paper, Phil Threadgill and Keith France for their assistance in writing this paper, Richard Dolby for authorising the funding for this investigation, Lee Smith and the staff of the Materials department for their help with the metallography and fractography, and Bill Noonan and Alistair Collins for performing the fatigue testing.

References