FLAWS IN FRICTION STIR WELDS A J Leonard and S A Lockyer ...

4th International Symposium on Friction Stir Welding, Park City, Utah, USA, 14-16 May 2003

FLAWS IN FRICTION STIR WELDS

A J Leonard and S A Lockyer TWI Ltd, Granta Park, Great Abington, Cambridge, CB1 6AL, UK

Abstract One of the major drivers for using friction stir welding for aluminium fabrication is the low incidence of weld flaws compared to that produced by conventional arc welding. However, the process does have its own characteristic flaws. A number of different process variables affect the quality of a joint produced by friction stir welding: tool design, tool rotation and travel speeds, tool heel plunge depth and tilt angle, welding gap, thickness mismatch and plate thickness variation. Successful, reproducible welds may be produced by operating within process “windows”. However, problems may arise when the welding conditions deviate from the standard operating window. In such circumstances, flaws may be generated. In the current work, a number of flaws encountered in friction stir welds, in particular voids, joint line remnants and root flaws were generated in an Al-Cu-Mn-Si-Mg alloy by using welding parameters outside of the established tolerance box for producing flaw-free welds. The welds were characterised using X-ray and ultrasonic non-destructive testing techniques and by metallographic sectioning. The causes of such features are described and recommendations are made to prevent their occurrence. INTRODUCTION Flaws are inherent in most materials joining processes. For example, a number of flaw types are associated with fusion welding of aluminium alloys, most notably weld metal porosity (1), Fig.1, and in certain alloys, weld metal solidification cracking and heat affected liquation cracking (2), Fig.2. Such problems have contributed to the view that some aluminium alloys, in particular some of the 2xxx and 7xxx series alloys, are difficult to fusion weld successfully. Friction stir welding, being a solid state process, has overcome the problems of porosity and hot cracking encountered in fusion welds. In addition, being a largely automated process, it is possible to produce defect-free welds reliably when operating within an

established parameter, or “weldability” envelope (3). The precise details of any operating envelope will depend on the material being welded and the actual quality control criteria specified. The rapid commercial application of the process is testimony to its robustness within defined operating parameters; to date, the authors are aware of no major problems that

2001-3-16-8-34-41-003 2001-1-31-9-38-36-003 Fig.1 Porosity in an aluminium arc weld. Fig.2 Hot cracking in an aluminium arc weld.


have been reported with regard to the occurrence of defects. However, it is inevitable that, as the process becomes more widely used, and operating conditions are pushed to their limits to satisfy needs for improved joint properties and productivity, flaws will be introduced. Two previous studies have reported on flaws in friction stir welds (4,5), in which tool rotation speed and welding speed were evaluated. The authors identified void formation and a root flaw. These were attributed to high travel speeds coupled with slow tool rotation speeds. Inadequate tool plunge depth was attributed to the cause of the root flaw. The objective of the current paper is to characterise the flaws that may, typically, be introduced into friction stir welds when operating outside of the “standard” set of optimised welding parameters. This will assist end users of the technology with an awareness of how process variation may introduce flaws and to enable appropriate process modifications to be made to eradicate them. The results reported were all generated on welds made in 6mm thick plate from an Al-Cu-Mn-Si-Mg alloy 2014A, produced using altered conditions derived from, but outside of, the original TWI development work. Specifically, flaws were generated by the variation of welding speed, welding force (forging pressure), tool pin height and surface oxide thickness.

Table 1 Summary of X-ray, ultrasonic and metallographic inspection of each of the welds in alloy 2014A.

Weld

Welding speed

mm/min Other details X-ray result Ultrasonic

inspection result Sectioning results W9 90 Shortened pin

used No defects recorded

No defects recorded

Root flaw

W10 90 2mm shim inserted between plates

No defects recorded

No defects recorded

No flaws found

W11 90 Total pressure limited to

500psi

Two large voids, 65mm and 140mm

long

Two voids: No 1, 5mm deep, 45mm in length No 2, 4.9 to 5.6mm deep 120mm

in length

Large voids on advancing side of

the weld

W12 180 -

No defects recorded

No defects recorded

Root flaw + void beneath surface of

weld W13 250 -

No defects recorded

No defects recorded

No flaws found

W14 300 -

No defects recorded

No defects recorded

Small voids beneath surface of

weld W15 400 -

Large void and associated small

voids at stop end of weld: area affected

85mm

Void 4.0m deep 11mm in length

Large void on advancing side of

weld

W17 90 20µm anodised

surface on plates

Feint linear indications full

weld length

Intermittent reflector full weld

length, 4.5mm deep

Oxide inclusions (joint line remnant)


Characterisation was performed using two non-destructive techniques: X-ray radiography and ultrasonic inspection, the latter being performed manually utilising a compression wave at 5MHz with a 10mm diameter probe; metallographic sections were also prepared to illustrate the features found by non-destructive examination. Table 1 summarises each of the welds produced and the flaws encountered. Three main flaw types are discussed: voids, root flaws and joint line remnants. VOIDS A number of the welds contained voids, which ran along the advancing side between the weld nugget and the remainder of the thermomechanically affected zone (TMAZ). Limiting the forging pressure generated a void along the entire length of weld W11, which was predominantly, although not exclusively, surface-breaking (Fig.3). If welding pressure were to vary momentarily during welding, voids may be generated that visual appraisal alone may not detect. A similar void in terms of appearance was generated in a weld produced at a speed of 400mm/min, over a factor of four times faster than a conventional friction stir weld in alloy 2014A using established technology. In this particular case, void formation was, in part, caused by the plates moving apart and the tool lifting (Fig.4).

Advancing side

2000-7-25-11-12-59-002a

Fig.3 A void on the advancing side of a weld (W11, mm scale).

67363_01

Fig.4a A surface-breaking void in a weld produced at 400mm/min (W15).


Advancing side

2000-7-25-11-45-33-002a

Fig.4b A surface-breaking void in a weld produced at 400mm/min (W15, mm scale). The location and shape of the voids provide useful information concerning the development of the characteristic microstructure observed in aluminium alloy friction stir welds. In each case, the voids were on the advancing side of the weld and showed a clear delineation between the region of the nugget and the remainder of the TMAZ. The fine-grained nugget exhibited clear scallops (Fig.5) which indicated that this region was formed by stirring and forging material either side of the joint line. The almost vertical edge of the advancing side of the void which was not at the original joint interface, indicates that material had been swept away by the rotation of the tool, the final joint being then formed by forging the nugget material into the created void. The tool shoulder plays an important part in forging the material that has been plasticised by the rotating pin in the joint line. If insufficient forging pressure is applied to the tool, the workpiece may ride up and prevent complete consolidation. When welding at higher travel speeds, the material receives less work per unit of weld length, i.e. fewer tool rotations per mm. Under such conditions, the plasticised material may be cooler, and less easily forged by the shoulder, resulting in voids remaining unconsolidated.

2000-8-4-14-16-16-003a

Fig.5 A void on the advancing side of weld W11. Bendzsak et al (6) and Colegrove (7) have developed mathematical models for material flow during friction stir welding. Both models predict a region of transition between a region of


rotational motion of material immediately beneath the tool shoulder and a region in which material is extruded past the rotating tool pin. Bendzsak et al (6) described the motion of the transition region as chaotic. Both models predict flow singularities in this region on the advancing side of the weld. Bendzsak et al attributed these to be the source of weld defects (voids). The current work supports the theoretical models in that voids were generated at the locations in which they were predicted to occur. Small voids were also observed elsewhere in the welds. In some instances, they were due to inadequate forging. Others were present intermittently, below the top of the weld when the welding speed was increased (Fig.6). The features corresponded to the region in which the tool shoulder left a mark on the top surface, in the form of a series of advancing semi-circles. It may be noted that all of the voids in the current work, with the exception of the small voids below the weld surface, were detected by conventional X-ray radiography and ultrasonic inspection. The small voids below the top of the weld may be removed by light machining of the weld surface.

2000-3-30-10-31-16-003a

Fig.6 A void beneath the upper surface of a weld produced at 180mm/min (W12). Christner and Sylva (8) investigated the effect of the presence of a joint gap on the mechanical properties of friction stir welds in 6.4mm thick alloy 2014A-T6. As in the current work, joint gaps were achieved by the insertion of shims at each end of the weld panel assembly. The length of weld produced and the distance between the shims was not presented, but it was recorded that a joint gap of 2.3mm, which represented 36% of the plate thickness, could be tolerated without a significant reduction in joint strength. A joint gap of 3.2mm, or 50% of the thickness, resulted in incomplete joint consolidation and the presence of a cavity along the advancing side of the weld. In the current work, a joint gap of 2mm (33% of the plate thickness) was tolerated, over a span of ~260mm, without the presence of any discernible flaws. This is consistent with the findings of Christner and Sylva, although in the current work the actual joint properties were not measured. ROOT FLAWS AND JOINT LINE REMNANTS Two principal examples of joint line remnants were generated in the current work. Firstly, by use of a shortened pin, a root flaw was created. During the formation of the friction stir weld, it is vital that the oxide interface between the two butting plates is adequately disrupted in order to form a bond. The correct depth of penetration of the tool pin is essential to ensure


that this occurs. Where a shortened pin is used, where the tool plunge depth is incorrectly set, or where there is poor alignment of the tool relative to the joint line, a root flaw is produced; the original plate interface is only partially disrupted and can still be identified on a cross-section, Fig.7b. In one weld, which essentially had the correct pin height and plunge depth, a root flaw was still produced. In this instance, either the plunge depth cannot have been adequate, or the joint line was badly misaligned relative to the tool axis of rotation, generating the flaw.

Advancing side

2000-3-28-11-36-31-002a

Fig.7a A root flaw in weld W9 (arrowed, mm scale).

2000-4-3-11-36-40-003

Fig.7b Higher magnification photomicrograph of weld W9, showing the root flaw (arrowed). Such root flaws do not necessarily represent an absence of any bond, indeed some regions of weak bonding may be present. However, work to date at TWI (9) on welds in alloy 5083-H321, which contained similar features, did show a reduction in tensile properties and a loss in fatigue strength. Therefore, such features should not be regarded as innocuous. Currently, NDT techniques such as radiography and dye penetrant inspection, as well as visual examination, are not reliable for detecting root flaws, and, indeed, the root flaws were not detected in the current work. The only definitive method that is currently available is a destructive bend test with the root in tension (10), although efforts are being applied to the


development of NDT techniques that are capable of detecting them. In the case of critical applications, machining the weld root may be an effective measure to remove such features. A second type of joint line remnant is the distribution of oxide particles through the thickness of the weld, Fig.8. In the current work, the oxide coating, which normally covers the surface of aluminium, was deliberately thickened by anodising. In this instance, a visible dispersed oxide line was produced in the weld, which was also of sufficient size to be detected by both radiography and ultrasonic inspection. Previous work at TWI (9) produced joint line remnant flaws in welds in alloys 5083-O and 6082-T6. In the former alloy, the feature was introduced by increasing the welding speed, resulting in less disruption of the oxide per mm advance of the tool; in the case of the latter, an oversize tool shoulder was employed, resulting in more surface oxide being swept into the weld. In each of these welds the tensile properties were not affected by the features, with the exception of a small drop in ductility in the 5083 weld. The fatigue strength did not appear to be affected by the presence of the flaws in the welds examined and fracture mechanics calculations were conservative in comparison with the test results. The TWI work concluded that such a flaw resulting in a loss in mechanical properties should be viewed as a crack and therefore not tolerable. Even if there is no loss in mechanical properties, the feature is clearly undesirable and is best minimised by careful machining of the butting faces to reduce the quantity of oxide prior to welding and control of welding speed. Improvements in tool design may also be of benefit in disrupting oxidised layers.

2000-7-27-15-25-7-002a

Fig.8a Advancing side

2000-7-28-9-38-57-002a

Fig.8b Joint line remnant in weld W17 (mm scale). PRACTICAL IMPLICATIONS Table 2 summarises the main flaw types in friction stir welds identified in this work and other published information. It also includes details of the factors causing their formation. It should be noted in the case of voids produced in the current work that only some of the welding parameters were varied, namely welding pressure, travel speed and joint fit-up; other parameters, such as tool design and rotation speed, may also influence flaw formation.


Comprehensive mathematical modelling of the welding process, as adopted by Bendzsak et al (6) and Colegrove (7), may result in a greater understanding of how individual welding parameters interact, permitting weld procedures to be developed without recourse to a large matrix of test welds in order to establish a tolerance window. The Table provides an initial resource for the identification of problems encountered in production welds, and an aid to their eradication. Table 2 Flaws that may be present in friction stir welds and their causes.

Flaw Type Location Causes Void Advancing side at edge of weld

nugget. 1. Reduced forging pressure. 2. Welding speed too high. 3. Plates not clamped close

enough together. Joint gap too wide.

Void Beneath top surface of weld Welding speed too high Joint Line Remnant Weld nugget, extending from the

root of the weld at the point where the original plates butted together.

1. Inadequate removal of oxide from plate edges.

2. Inadequate disruption and dispersal of oxide by tool.

3. Increase in welding speed. 4. Increase in tool shoulder

diameter. Root flaw Weld nugget, extending from the

root of the weld at the point where the original plates butted together.

1. Tool pin too short. 2. Incorrect tool plunge depth. 3. Poor joint to tool alignment.

Currently, no national or international inspection criteria exist for evaluating the quality of friction stir welds. Although, an AWS standard is in preparation (10). Lloyds Register of Shipping has issued guidance notes for weld qualification (11), which have been based largely on the requirements for arc welds in British Standard BS EN288 part 4 (12). This latter document specifies 100% visual examination, 100% radiographic or ultrasonic inspection and 100% penetrant inspection, together with bend tests, tensile tests and metallography. Lloyds have specified penetrant inspection on only the weld root for friction stir welds. The acceptance levels for weld imperfections in BS EN288 Part 4 are described as level B – stringent, as defined in BS EN30042 (13). In the current work, flaws were identified by a combination of non-destructive and metallographic inspection techniques. No attempt was made to determine the limits of detectability of flaws by these techniques. However, if weld qualification standards are to be developed for friction stir welds, further work will be necessary in determining these limits. CONCLUSIONS 1. Voids may be formed when insufficient forging pressure is applied to the weld. Welding

too fast may also result in similar effects. 2. A joint gap of up to 2mm may be tolerated when welding 6mm thick 2014A plate without

the formation of weld flaws.


3. Joint line remnants in the form of inadequately dispersed oxide may be produced if inadequate pre-weld cleaning is performed or the welding speed is too high. Evidence suggests that joint line remnants may be tolerated in certain circumstances, but are best avoided. Machining the plates prior to welding and control of welding speed are effective in restricting their appearance.

4. Joint line remnants in the form of root flaws are introduced when either insufficient pin

depth or tool plunge depth is selected for the joint, or when there is poor tool to joint alignment. These flaws have been found to be damaging to both tensile and fatigue properties. Appropriate selection of the above parameters may eliminate these flaws. However, for critical applications, machining of the weld root may be advisable.

5. Cross weld tensile testing and root bend tests are recommended to identify damaging joint

line remnants and root flaws. ACKNOWLEDGEMENTS This work was funded by Industrial Members of TWI as part of the Core Research Programme. The authors are grateful for the assistance of colleagues at TWI who assisted with the work. REFERENCES 1. Gingell A B D and Gooch T G: ‘Review of factors influencing porosity in aluminium

arc welds’, TWI Members Report 625/1997 October 1997. 2. Gittos M F and Scott M H: ‘Selection of filler metals for arc welding aluminium

alloys’, The Welding Institute Research Bulletin 1987 28 (8) 259-263. 3. Midling O T, Oosterkamp L D and Bersaas J: ‘Friction stir welding aluminium -

process and applications’, Cambridge, UK, TWI, 16 April 1998, 161-169. 4. Hashimoto T, Jyogan S, Nakata K, Kiu Y G and Ushio M: ‘FSW joints of high

strength aluminium alloy’, Proc 1st International Symposium on Friction Stir Welding, 15-16 June 1999, Thousand Oaks, CA, USA.

5. Nakata K, Kim Y G, Ushio M, Hashimoto T and Jyogan S: ‘Weldability of high

strength aluminium alloys by friction stir welding’, ISIJ International 2000 40 515-519.

6. Bendzsak, G J, North T H and Smith C B: ‘An experimentally validated 3D model for

friction stir welding’, Proc 2nd International Symposium on Friction stir welding, 26-28 June 2000, Gothenburg, Sweden.

7. Colegrove P: ‘3 Dimensional flow and thermal modelling of the friction stir welding

process’, Master of Engineering Science Thesis, The University of Adelaide, Australia, January 2001.


8. Christner B K and Sylva G D: ‘Friction stir welding development for aerospace applications', Proc. Conf "ICAWT '96", 6-8 November 1996, Columbus, OH, USA, 359-368.

9. Dickerson T and Przydatek J, ‘The significance of root flaws in friction stir welds in

aluminium alloys’ TWI Members Report 714/2000, September 2000. 10. AWS D17.3 ‘Specification for friction stir welding of aluminum alloys for aerospace

applications’, - Draft. 11. Pryzdatek J: ‘A ship classification view on friction stir welding’, Proc 1st International

Symposium on friction stir welding, 15-16 June 1999, Thousand Oaks, CA, USA. 12. BS EN288-4:1992 ‘Specification and approval of welding procedures for metallic

materials, Part 4: Welding procedure tests for the arc welding of aluminium and its alloys’.

13. BS EN30042:1994 ‘Arc welded joints in aluminium and its weldable alloys –

guidance on quality levels for imperfections’.

FLAWS IN FRICTION STIR WELDS A J Leonard and S A Lockyer ...

Documents