1.Material Flow Behavior During Friction Stir

8/8/2019 1.Material Flow Behavior During Friction Stir

1/9

ABSTRACT. Friction stir weldi ng (FSW) isa new technique for joining aluminumalloys. Invented in 1991 at The WeldingInstitute (Ref. 1), thi s technique results inlow distortion and high joint strengthcompared with other techniques, and iscapable of joining all aluminum alloys.To date, the majority of research has con-centrated on developing the tools andprocedures for making reliable welds ina variety of alloys, on characterizing theproperties of welds and on developingdesign allowables (Refs. 27). However,very little is known about material flowbehavior during welding. The purpose of

the current study is to document themovement of material during friction stirwelding as a means of developing a con-ceptual model of the deformationprocess. In this paper, two new tech-niques for visualizing material flow pat-terns in friction stir welds are presented.Based on measured results in welds of6061 and 7075 aluminum, materialmovement wi thin fri ction stir welds is byeither simple extrusion or chaotic mix-ing, depending on where wi thin the weldzone the material originates. These re-sults impact the development of weldingprocedures and suggest ways to modelthe process for predicting welding toolperformance.

Introduction

Friction stir welding is a new weldingtechnique for aluminum alloys invented

by The Welding Institute, Cambridge,U.K., in 1991 (Ref. 1). This techniqueuses a nonconsumable steel welding toolto generate frictional heating at the pointof welding and to induce gross plastic de-formation of workpiece material whilethe material is in a solid phase, resultingin complex mixing across the joint. A de-tailed account of the process has beenprovided by others (Refs. 1, 3, 7). Al-though friction stir welding can be usedto joi n a number of materials, the primaryresearch and industrial interest has beento join aluminum alloys. Defect-freewelds with good mechanical properties

have been made in a wide variety of alu-minum alloys, even those previouslythought to be unweldable, in thick-nesses from less than 1 mm to more than35 mm. In addition, friction stir weldscan be accomplished in any position.Clearly, friction stir welding is a valuablenew technique for butt and lap jointwelding aluminum alloys.

Of importance to this work, and sub-sequent interpretation of results, is theFSW tool design and how i t interacts wi th

the workpiece. The steel tool is com-prised of a shank, shoulder and pin, asshown in Fig. 1. The welding tool is ro-tated along its longitudinal axis in a con-ventional mil ling machine and the work-piece material is firmly held in place in afixture. The shoulder is pressed againstthe surface of the metal generating fric-tional heat while containing the softenedweld metal. The pin causes some addi-tional heating and extensive plastic flowin the workpiece material on either sideof the butt joi nt. As can be seen in Fig. 1,the pin is equipped with a screw thread.This thread was found to assist in ensur-

ing that the plastically deformed work-piece material is fully delivered aroundthe pin, resulting in a void-free weld. Toachieve full closure of the root, it is nec-essary for the pin to pass very close to thebackplate, since only a l imi ted amount ofplastic deformation occurs below thepin, and then only very close to the pinsurface.

A typical butt joint welding sequenceproceeds as follows:

The workpiece material, with squaremating edges, is fixtured on a rigid back-plate. The fixturing prevents the platesfrom spreading or li fting during welding,and holds the material at a slight anglerelative to the axis of the welding tool.The welding tool, fixed in its holder, isspun to the correct spindle speed and isslowly plunged into the workpi ece mate-rial unti l the shoulder of the welding toolforcibly contacts the upper surface of thematerial and the pin is a short distancefrom the backplate. At this point thewelding tool is forcibly traversed alongthe butt joint, which continues until theend of the weld is reached. The welding

WELDING RESEARCH SUPPLEMENT | 229-s

W E L D I N G R E S E A R C H

SUPPLEMENT TO THEWELDING JOURNAL, JULY 1999Sponsored by the American Welding Society and the Welding Research Council

Material Flow Behavior during Friction StirWelding of Aluminum

BY K. COLLIGAN

Tracers embedded in the weld path and a stop action technique give insight intothe movement of material during friction stir welding

KEY WORDS

Friction Stir WeldingAluminum Alloys6061 and 7075Material FlowWeld ToolFriction HeatingPlastic DeformationAt the time this paper was written, K. COLLI-

GAN was with The Boeing Co., Seattle, Wash.He is currently wi th Lockheed Martin MannedSpace Systems, New Orleans, La.


2/9

tool is then retracted, generally while thespindle continues to turn. Once the toolis completely retracted, the spindle isstopped, and the welded plate can be re-moved from the fixture. It should benoted that when the tool is retracted thepin of the welding tool leaves a hole inthe workpiece at the end of the weld.

Experimental Procedures

Steel Shot Tracer Technique

To better visualize the flow of mater-ial around the welding tool, two newtechniques were used. First, small steelballs (0.38 mm diameter) were used as atracer material embedded at different po-sitions wi thin butt joi nt welds of 6.4-mm-thick 6061-T6 and 7075-T6 plate. A weldwas run along the length of the seededbutt joint and stopped at a point along thetracer pattern. By stopping the forward

motion of the

welding tool whileit is still in theseeded material,the steel shot dis-tribution aroundthe welding tool ispreserved in theend of the weld,revealing the paththat the tracer ma-terial took in trav-eling around thewelding tool. Eachweld was subse-quently radi-ographed to revealthe distribution ofthe tracer materialas it transitioned

from its original position, around thewelding tool and into the welded joint.

A number of techniques for embed-ding the steel tracer material were evalu-ated in preparation for the detailed ex-perimental work in this study. However,the most effective method of embeddingthe tracer material involved machining asmall groove, 0.75 mm high by 0.3 mmdeep, along the butting edge of a plate of

6.4-mm aluminum and fil ling the groovewith steel balls, as shown in Fig. 2. Priorto welding, the plates are forced togetherto imbed the 0.38-mm balls into the 0.3-mm groove. This technique results in ahorizontal line of steel shot arranged atany desired position within the weld bymaking the groove at different depths andby orienting the butt joint at different lat-eral positions relative to the path of thewelding tool. The initial tracer line loca-tions are shown in Fig. 3 for the 6061 andin Fig. 4 for the 7075. Inspection of the

weld by radiography then revealed the

line of tracer material in advance of thewelding tool, as the material was de-forming around the tool, and after pas-sage of the welding tool.

The second technique used in thisstudy involved ending friction stir weldsby suddenly stopping the forward motionof the welding tool and simultaneouslyretracting the tool at a rate that causedthe welding tool pin to unscrew itselffrom the weld, leaving the materialwithin the threads of the pin intact andstill attached to the keyhole. By section-ing the keyhole at the end of a weld thatwas made using this stop action tech-nique, one can study the flow of materialin the region immediately within thethreads of the welding tool. This tech-nique requires the use of a numericallycontrolled (NC) milling machine.

The welds in 6061 were made on athree-axis NC mil ling machine equippedwi th a special fixture for friction stir weld-ing. The use of the NC mill allowed thestop action technique used on the 6061alloy. Due to the higher forces requiredto weld 7075, it was necessary to use adifferent FSW system for those welds.Unfortunately, this equipment was not

capable of performing the stop actionmotion at the end of the weld; therefore,the tracer dispersal patterns in the regionimmediately around the welding tool arenot valid representations of the pathtaken by the steel shot for the 7075 as ittraveled around the pin.

Welding Procedures

The welding parameters and detail s ofthe welding tool geometry are restrictedfrom publication by agreement with the

230-s | JULY 1999

RESEARCH/DEVELOPMENT/RESEARCH/DEV

ELOPMENT/RESEARCH/DEV


ELOPMENT

Fig. 1 Schematic of friction stir welding tool (Refs. 1, 4, 8, 9). The

tool is composed of a steel shaft with a pin and shoulder on one end

and a shank on the opposite end, which is mounted to the spindle of

a milling machine.Fig. 2 The tracer line technique employs a continuous line of 0.38-

mm steel shot tracer material sprinkled into a small rectangular

groove machined in the butting edge of a plate.

Fig. 3 Tracer li ne positions for the 6.4-mm 6061-T6 plate. The groove

containing the steel shot tracer material was oriented at various posi-tions relative to the welding tool pin and at various depths in the work-

piece plate, as shown in this schematic.


3/9

companies that funded industrial devel-opment of FSW at The Welding Institute(TWI). However, it can be said that thewelding tool pin had a threaded surface,as shown schematically in Fig. 1, and theweld travel speed can be in excess of 30cm/min. For reference, some public do-main information about the details ofwelding tool geometry and attitude dur-ing welding is available (Ref. 8). It shouldbe noted that different welding tools andparameters were used for the welds madein the 6061 alloy and the 7075 alloy;however, it is expected that the small dif-ferences between the welding tools useddoes not preclude comparison of the re-sults obtained.

Results

Steel Shot Tracer Results

Before presenting the results, it is nec-essary fi rst to discuss the convention usedfor referring to locations within frictionstir welds. Since friction stir welds areasymmetrical, it is necessary to accu-rately convey which side is intendedwhen referring to specific locations

wi thin a weld with respect to the tool ro-tation and feed di rections. The follow ingconvention will be used in the discus-sions to follow. The side of the weldingtool where the motion of the surface ofthe welding tool is in the same directionas the feed direction is referred to as theadvancing side. The opposite side, wherethe motion of the surface of the weldingtool opposes the feed direction, is re-ferred to as the retreating side. A termi-nology convention that is also used (Ref.10) refers to the advancing and retreating


Fig. 4 Tracer line positions for the 6.4-mm 7075-T6 plate. The groove

containing the steel shot tracer material was oriented at various positions

relative to the welding tool pin and at various depths in the workpiece

plate, as shown in this schematic.

Fig. 5 The plan view radiograph of 6061 tracer position 7weld, its position defined in Fig. 3, shows the unwelded tracermaterial at the top of the figure, as it flows around the weldingtool pin, and as it is dispersed withi n the weld. The outline ofthe welding tool shoulder can just be seen. The hole left by theextracted welding tool pin i s clearly vi sible.

Fig. 6 Drawings generated from each of the plan view radiographs for the 6061 welds are

shown here. The positions are defined relative to the welding tool path in Fig. 3.


4/9

sides as the shear side and the flow side,but since this convention makes assump-tions about the material flow, the moregeneric terminology will be used here.

Another convention used here is torefer to tool movement in indicating thefeed direction, as opposed to workpiecemovement. Also, the welding tool rota-tion direction used in the welds made in7075 was clockwise when viewed from

above; however, the diagrams of tracerdispersal patterns were reversed in orderto make them appear the same as the di-agrams from the 6061 alloy, which had acounterclockwise tool rotation.

The tracer li ne technique produced acomprehensive depiction of materialmovement in friction stir welds. A typi-cal radiograph is shown in Fig. 5. Thissample was taken from 6061 alloy,tracer line position 7 as defined in thelayout in Fig. 3. This plan view showsthe weld keyhole, the steel shot in ad-vance of the keyhole (in the upper por-tion of the figure) and the reoriented

steel shot behind the keyhole (in thelower portion of the figure). Figure 6shows the plan views of all of the 6061tracer line radiographs, Fig. 7 shows theside views from the same specimens,Fig. 8 shows the plan views from the7075 radiographs and Fig. 9 shows theside views from the same specimens.The stop action technique was usedwhen welding the 6061 specimens, butnot with the 7075 specimens.

Results shown in Figs. 6 and 7 confi rmsome conclusions drawn from conven-tional metallographic FSW cross sectionsand also introduce new information inthe development of a comprehensivematerial flow description. Referring toFig. 6, one can immediately observe dif-ferent material movement patterns in dif-ferent parts of the weld. In positions 1through 3, lines of steel shot originatenear the upper surface of the plate. Thetracer material in these positions isbrought around the pin on the retreatingside and scattered behind the pin, thefinal resting place being biased towardthe advancing side, but the material isotherwise randomly scattered. It is evi-dent from Fig. 7 that this tracer material

also rises slightly in front of the pin andis then driven down to a final depth thatis deeper than the original level.

The pattern of movement of tracermaterial in position 2 is representative ofthat seen in each of the first three posi-tions. In position 2 a nearly continuous

232-s | JULY 1999




ELOPMENT

Fig. 7 Drawings generated from each of the side view radiographs for the 6061 welds are

shown here.

Fig. 8 Drawings generated from each of theplan view radiographs for the 7075 welds areshown here. The positions are defined relativeto the welding tool dimensions in Fig. 4.


5/9

line of tracer material in front of the weld-ing tool is lifted and brought counter-clockwise around the pin as it enters theweld zone. Both motions take place wellin advance of the pin, as can be seen byviewing both the plan view and side viewin Figs. 6 and 7. The appearance of tracermaterial on the advancing side of the pinimplies that some of the tracer materialmay make at least one full rotation

around the pin as it is being driven downinto the weld. This observation is re-peated in position 3, where a number ofsteel shot are seen on the advancing side,very close to the pin. It appears that in po-sition 2 the steel shot is driven from aninitial depth of 1 mm to a maximum ob-served depth of about 2 mm, or 30% ofthe plate thickness.

Position 4 appears to fundamentallydiffer from the first three positions. Thematerial is not significantly propelleddownward into the weld and the tracermaterial is not chaotically scattered be-hind the pin. The tracer material in this

position appears to be dragged behindthe pin by the rotation of the shoulderand is deposited behind the shoulderwithin about 1 mm of the upper surfaceof the weld, near its initial depth.

In position 5, tracer material is signif-icantly lifted as it passes the pin and isslightly pushed away from the weldzone. A small amount of tracer materialis deposited outside of the general trendof the rest of the tracer material. Referringto the plan view of position 5, a smallcluster of tracer material was depositedtoward the back of the shoulder on theretreating side. Despite this exception, alarge majority of the tracer material ispushed away from the pin and lifted asthe pin passes. A very simil ar trend is fol-lowed in position 9, originating at thesame lateral position as 5 but just belowit in the weld. As can be seen in the planand side views, about half of the tracermaterial is lifted and pushed away fromthe pin but the balance of the material, alarger percentage in this position than inthe previous one, is deposited on the re-treating side of the pin. In both positions5 and 9, the material deposited on the re-treating side is left at the same depth in

the weld as that on the advancing side.Similar tracer deposition patterns arefound in positions 6 through 8 and 10through 12. Generally, tracer materialfrom these welds was brought around theretreating side of the pin, while being si-multaneously lifted to a higher positionwi thin the weld. The tracer material wastypically deposited adjacent to the re-treating flank of the pin, or slightly be-hind this side of the pin.

In positions 13 and 14 the steel shotpasses under the pin and does not signif-


Fig. 9 Drawings generated from each of the side view radiographs for the 7075 welds are

shown here.

Fig. 10 In this 6.4-mm 6061 specimen the stop action technique was used to capture ma-

terial flow patterns during welding near the welding tool pin and shoulder. A Initi al butting

surface intersection wi th section plane; B material fi ll ing thread space; C fi ll ed thread space

in front of pin; D void behind upper portion of pin; E filled thread space behind pin; F

material fl owing upward behind pin; G material extruded under pin; H material extruded

under shoulder.


6/9

icantly change its depth or lateral posi-tion. It appears that there is only slightmaterial movement in this region.

Referring to Figs. 8 and 9 for welds in7075, it can be seen that similar resultswere observed in this alloy. It should benoted that these tests were done on awelding machine that could not performthe stop action pin withdrawal, and asa result, the tracer dispersal patterns verynear the pin should not be interpreted asbeing representative of those present dur-ing actual welding. Positions 1 through 3show the same chaotic dispersal patternas was seen in the 6061 radiography re-sults. The steel shot is slightly li fted in ad-vance of the welding tool pin and is thendriven to a greater depth wi thin the weld.For example, in position 3 the steel shotoriginates at a depth of 1 mm and is de-posited behind the pin as deep as 4 mm,or 23 of the plate thickness (twice as deepas that observed in the 6061 for a simi lartracer line position).

In position 4 the steel shot also rises as

the pin passes but is not dri ven down intothe welded plate. Instead, the steel shotremains just below the surface and is re-oriented counterclockwise behind thepin by a small distance. The majority ofthe steel shot is deposited in a straightline with a percentage of the tracer ma-terial being deposited in a somewhat dif-ferent pattern. This pattern is somewhatdifferent from position 4 in the 6061.

In positions 5 and 9 the dispersal pat-terns observed are quite similar to thoseobserved in the 6061 results. In position5 the steel shot is li fted in advance of thepin to very near the upper surface of thematerial under the shoulder of the weld-ing tool and is slightly pushed away fromthe pin. In positi on 9 the shot is also liftedin advance of the pin and most of thetracer material passes the pin withoutmuch lateral movement, except for a per-centage of the tracer material that i s cap-tured by the pin, moved counterclock-wise and deposited on the retreating sideof the pin.

The tracer material dispersal patternsseen in positions 6 through 8 and 10 and11 are similar to each other, as were thecorresponding positions in the 6061 re-sults. In these positions the tracer mater-ial is delivered counterclockwise aroundthe pin and delivered to different posi-tions behind the welding tool. At thesame time, the tracer material is liftednearly to the upper surface of the weld,

just behind the pin and pushed backdown into the weld as the shoulderpasses. In all cases the final vertical posi-tion is above the original vertical posi-tion. Positions 6, 7, 10 and 11 differ fromthe 6061 counterparts only in that thetracer material is left more directly be-hind the pin in the 7075 case than in the6061 and the lifting as the pin passes ismore pronounced in the 7075 case thanin the 6061. Based upon the regularity ofthe dispersal pattern behind the pin ineach case, it is likely that the tracer ma-terial does not make more than a partialrevolution around the pin in all cases ex-

234-s | JULY 1999




ELOPMENT

Fig. 11 View of the 6.4-mm 6061 stop action keyhole trailing

edge longitudinal section. A Partiall y fil led threads; B full y fil led

threads; C dark band; D downward distortion of material; E

upward distortion of material.

Fig. 12 An enlargement of the specimen from Fig. 10 in the region

behind the pin. A Area with no material filling behind the pin; B

full y fil led threads; C material fi ll ing behind the pin from below.


7/9

cept for positions 1, 2 and 3.In positions 13 and 14 the tracer ma-

terial approximately remains at its origi-nal depth in the plate, but the lateral lo-cation is slightly reoriented to the

retreating side of the weld.

Stop Action Test Results

To gain further insight into the mech-anisms of material flow in friction stirwelds, an analysis of weld sections takenfrom the keyhole region of welds in 6061was undertaken. The welds studied wereproduced using the stop action tech-nique described above. A number ofweld keyholes specimens were exam-ined, produced by taking a longitudinalsection in 6061 and etching for 60 s in aconcentrated Kellers etch. These weldswere produced using a two-piece weld-ing tool that had an H13 tool steel bodyand a tungsten carbide pin with smoothlyground threads. Figure 10 illustrates atypical microstructural morphology.Starting from the right side of the photo-graph, the unwelded base metal in ad-vance of the welding tool i s horizontallybanded, wi th the upper and lower thirdsbeing darker than the center third. As thewelding tool advances, these bands re-veal an upward distortion of material,contacting the leading edge of the weld-ing tool shoulder. Behind the welding

tool, the weld can be seen to consist of anumber of horizontal layers. The toplayer consists of a thin band of horizon-tally striated material extending from theupper portion of the back edge of the pin,along the welding tool shoulder surface,and on down to the fully developed weld(feature H in Fig. 10). The next layerdown, comprising about half of the thick-ness of the weld, consists of vertically ori -ented bands, which appear to emanatefrom the lower portion of the back edgeof the pin (feature F). Previous investiga-

tions have shown this banded structure tobe associated with differences in grainsize (Ref. 7). The bottom half of the weldconsists of a layer that is also verticallybanded but much fainter. The very bot-

tom layer of the weld is very dark in ap-pearance and is not as distinctly struc-tured as other portions of the weld.

As can be seen in Fig. 10, the hole leftby the threaded pin is bounded by alu-minum that occupied the thread spaceswhen the welding tool was retracted.Close inspection of the hole left by thepin reveals that the profile in front of thepin (the right side of the hole) is differentfrom the trailing side. Looking from topto bottom along the front side of the hole,the thread form appears to graduall y de-velop from curls of aluminum that in-crease in size, progressing downward(detail B in Fig. 10) until the fully devel-oped thread form is observed at abouthalfway through the thickness. Lookingfrom top to bottom along the trail ing sideof the hole, the hole is smooth sided inthe top half of the thickness (detail D) andhas fully formed thread shapes in the bot-tom half of the weld (detail E).

Figure 11 shows a higher magnifica-tion view of the leading edge of the key-hole. The progressive fil li ng of the threadform can clearly be seen in the top halfof the weld (detail A), with the bottomhalf of the hole having fully developed

thread forms (detail B). In addition, a darkvertical band (detail C) is seen just in ad-vance of the pin, leading from the darkhorizontal band in the lower portion ofthe base metal (Fig. 10), up along theleading margin of the pin and into the firstfully formed thread form. Material infront of the pin in the upper portion of theplate is seen to be distorted first upwardand then downward near the tool (detailD), whi le in the lower portion of the weldthe material is only distorted upward (de-tail E). Finally, the fully formed thread

forms are vertically banded with lightand dark material.

Figure 12 shows a higher magnifica-tion view of the trailing edge of the key-hole from the weld in Fig. 10. The upper

portion of the keyhole does not containthread-formed material, but instead isrelatively smooth (detail A). The lowerportion of the keyhole contains fullthread forms (detail B). The depth atwhich fully formed threads are seen onthe trailing side closely matches thedepth of the first full y formed threads ob-served on the leading edge, as can bebetter seen in Fig. 10. The vertical lightand dark banding behind the pin can beseen in this view (detail C) emanatingfrom the region of the fully formedthreads on the trailing side of the key-hole.

Discussion of Results

Steel Shot Tracer Results

The tracer line tests performed onboth 6061 and 7075 give insight into theflow of material in the region around thewelding tool pin. First, the material flowsindicated from the various positions inFigs. 6 through 9 can be divided into twogeneral categories: those where the con-tinuous line of steel shot was reorientedand deposited as a roughly continuous

line of steel shot behind the pin, andthose where the line of steel shot was de-posited chaotically behind the pin. InFigs. 6 through 9, positions 1, 2 and 3 (forboth alloys) represent situations wherethe material was chaotically deposited,and positions 4 through 14 representcontinuous line deposits of tracer mater-ial. In this context, chaotic dispersal oftracer material refers to production of atracer pattern in which no evidence ofthe original l ine of tracer remains. In thistype of dispersal, individual tracer ele-


Fig. 13 A schematic view of the average final positi ons of steel shot

tracer material in a 6061-T6 transverse section for positions 5 through

14. The initial positions are shown in Fig. 3.

Fig. 14 A schematic view of the average final positions of steel

shot tracer material in a 7075-T6 transverse section for positions 4

through 14. The initial positions are shown in Fig. 4.


8/9

ments are scattered in an unpredictableway within a relatively broad zone be-hind the welding tool pin.

It should also be noted that in thosecases were the tracer material was chaot-ically deposited, the tracer material wasalso moved from its original depth andscattered through the thickness to a some-what greater depth. Conversely, in thosecases where the tracer material was de-

posited in a nearly continuous line of ma-terial behind the pin, the tracer materialwas moved from its original position to afinal position that was somewhat closer tothe upper surface of the material.

Based on these observations, it is ap-parent that those cases where the tracermaterial was deposited as a roughly con-tinuous line of material behind the pinrepresent cases where the aluminum im-mediately surrounding the original tracerline positi on was simply extruded aroundthe pin. In this case, extrusion refers to aflow of material where plastic deforma-tion takes place without mixing, i.e., ad-

jacent elements of material are deformedbut remain adjacent to each other.

If the lateral and vertical positions ofthe lines of tracer material are measuredfrom the drawings in Figs. 6 through 9and plotted on a cross section of the weldschematic, the final positions can becompared w ith the original positions, asshown in Fig. 13 for the 6061 and Fig. 14for 7075. The data from positions 1through 4 are omitted from these draw-ings since the final dispersal patterncould not be represented as a point on across section drawing, as was the casewi th the extruded li nes of tracer material.These drawings reveal that in the case ofthe 6061, material from positions 6through 11 is deposited in an area on theretreating side of the pin just behind theedge of the pin. Material from positions5, 9 and 12 are moved upward and out-ward away from the pin, while materialthat passes under the pin was onlyslightly displaced. Similar results wereobserved for the 7075, but the zonewhere material from positions 4, 6, 7, 9,10 and 11 was deposited is higher in theweld and closer to the centerline of theweld than in the 6061 case.

It seems clear from these results thatas the welding tool approaches, materialfrom the mid-section of the plate directlyin the path of the pin is lifted and ex-truded around the pin in the direction ofpin rotation. Material from the mid-sec-tion of the plate, which is aligned withthe margins of the pin, is also lifted, butis less affected by the shear forces im-parted by the rotating pin. On the ad-vancing side of the weld, this marginalmaterial may be captured by the pins ro-tational flow and carried around to the

retreating side, or it may simply lift andbe pushed away from the weld center-line. On the retreating side the mid-thick-ness material, which is aligned with themargin of the pin, is simply lifted as thesurrounding material extrudes by the pinwi thout much displacement laterally.

Material that passes below the pin,from positions 13 and 14 in both alloys,is not greatly displaced, although in the

case of the 7075, the tracer material didappear to be under the influence of therotation of the pin and was displaced ina manner similar to that of the mid-sec-tion material (positions 5 through 12). Inthe case of the 6061, this effect was less,and differences in the pin geometry be-tween these two welds is likely responsi-ble. For example, the welding tool usedfor the 6061 had a smaller pin diameterand had a relatively flat end when com-pared to the tool used for the 7075 welds,which would certainly alter materialmovement patterns in the bottom portionof the weld.

Stop Action Weld Results

The welding tool affects material fromthe upper portion of the plate in a muchdifferent manner than in the case of theother positions within the weld. Basedupon the stop action welds, it appearsthat material from positions 1, 2 and 3 iscurled directly into the threads of thewelding tool pin and is then delivereddeeper into the weld. Figure 11 clearlyshows material that was fil li ng the threadspace at the time the tool was retracted.

This mechanism of filling the threadspace shows that the material curl ing intothe thread space is still attached to thebase metal directly in front of the pin asit i s extruded into the thread space. If thisis the case, the material being curled intothe thread space is not rotating with thepin, but is sliding against the rotatingthread surface. Therefore, relative to thematerial i n front of the pin, the thread ismoving downward as it rotates and ismoving forward with the overall motionof the welding tool; thus, the curl of ma-terial continuously grows as it i s extrudedinto the thread space. Also, the existence

of a void behind the pin, seen in Fig. 12,detail A, supports the notion that the ma-terial curling into the thread space is at-tached to the base metal in front of thepin and not traveling around the pin w iththe rotation of the threads, although otherinterpretations are possible.

It should be noted that no materialwas detected clinging to the welding toolpin after welding. In order for this tech-nique to work, it is important that thethreads of the welding tool pin should befree of aluminum alloy, with the excep-

tion of the thin film of aluminum thatcommonly clings to friction stir weldingtools after use.

Material curls into the thread spaceuntil the space is filled, at about the mid-point of the plate thickness. The evidenceproduced by the stop action tests is lessclear about what happens at this point,but Fig. 12 provides two clues that allowspeculation about how material is trans-

ported around the pin. First, the thread-formed material that is seen behind thepin occurs only at the depths where fullyformed threads exist on the leading edgeof the pin. Second, i t appears that the ma-terial rising behind the pin (detail C) orig-inates from the thread-formed material onthe rear of the pin. It can be speculatedthat once the thread space is filled, thecontinued downward motion of thethread relative to the material in advanceof the pin causes the material captured in-side the thread space to be separated fromthe surrounding material by a narrowband of excessively strained material. The

material inside of the thread space thenbegins rotating with the welding tool pin.As the material within the thread spacecomes around the back of the pin, it isthen deposited behind the pin.

Validity of the Techniques Used

The above discussion describes ma-terial movement in the upper portion ofthe weld as being caused by materialcurli ng into the welding tool s threads ata rate dictated by the amount of forwardmotion of the tool in each revolution.

Since this rate is about1

3 the diameter ofthe steel shot tracer material, it is likelythat the tracer material does not flow intothe thread space in the same way as thealloy. Therefore, it is expected that in theportions of the plate where this type ofmaterial movement takes place thetracer does not accurately reflect move-ment of the aluminum alloy. Instead ofcontinuously curling into the threadspace, it is likely that the steel shot ispushed out of the thread space by thethreads downward motion. However,the downward movement of tracer ma-terial seen in positions 1 through 3 for

both alloys does support the conclusionreached based on stop-action testing of6061 plates: that material from the upperportion of the weld is pulled down by thethread form and deposited low in theweld behind the pin. Although using asmaller tracer material would improveits ability to accurately reflect materialdeformation, the smaller tracer materialswould also be more difficult to detectusing radiography.

In the middle and lower portions ofthe plate, aluminum appears to move by

236-s | JULY 1999




ELOPMENT


9/9

a more orderly deformation process. Thisis reflected in the way that the tracer ma-terial largely remains in a line even afterpassage of the welding tool. It is expectedthat in this area the geometry of the tracermaterial used is adequate as an indicatorof deformation in the surrounding alloy,and conclusions can be inferred from thetracer dispersal patterns observed.

The use of steel shot as a tracer mate-

rial could also be an inaccurate indicatorof material deformation due to the differ-ence in thermal conductivity and massbetween the tracer and surrounding alu-minum alloy. The line of closely packedsteel shot will certainly reduce heat con-duction along its length in advance of thewelding tool, and will therefore disturbthe temperature distribution in the sur-rounding aluminum. However, the smallsize of the steel shot should have pre-vented this inhomongeneity from alter-ing the main heat conduction propertiesof the alloy. Also, the high thermal con-ductivity of the aluminum would tend to

reduce thermal gradients in the vicinityof the line of tracer material.

The difference between the mass ofthe tracer material and the aluminumalloy causes inertial forces to develop be-tween the materials in cases where theyare accelerating. This effect is greatestwhere the acceleration is greatest, suchas where the tracer and alloy are beingpulled around the welding tool pin. Sincethe steel tracer material has a higher massthan the surrounding alloy, it would beexpected that the tracer material shouldlag behind the surrounding alloy for pos-itive accelerations. However, at the tem-peratures expected in the weld zone (Ref.6), the aluminum alloy still has high vis-cosity and would l ikely resist relative mo-tion between the alloy and the tracer ma-terial. Therefore, it is expected that thedifference in masses between the alloyand tracer material is not significant forthis study.

Normally, when making a butt jointweld using the friction stir technique thebutt joint is located on the centerline ofthe welding tool path. The tracer methodused in this study requi res that the buttingsurfaces be located in various lateral po-

sitions relative to the welding path inorder to place the tracer material as de-sired relative to the welding tool path.The butting surfaces form a heat transferbarrier due to incomplete contact whencompared to continuous material. In thefriction stir welding process, there is alarge thermal gradient in the radial direc-tion, along the length of the butting sur-faces, but one would not expect a signif-icant thermal gradient across the buttingsurface. As a result, it is reasonable to ex-pect that moving the butting surfaces lat-

erally a small distance from their normalposition should not significantly disturbthe heat flow radially away from thewelding tool.

With respect to the stop actionmethod, the main concern about the va-lidity of the technique relates to the ac-celeration and deceleration profiles ofthe axes of the machine tool used tomake the welds. In order for this tech-

nique to be valid, the longitudinal axismust quickly decelerate while the out-of-plane axis accelerates to full retractspeed, leaving the material that waswi thin the threads during welding intactand undisturbed. The theoretical accel-eration and deceleration rates of the ma-chine tool can be calculated based on in-formation from the machine toolmanufacturer. One revolution of thewelding tool takes about 39 ms. The lon-gitudinal axis takes only 1.07 ms to de-celerate from the full welding speed tozero, representing 9.9 deg of weldingtool rotation. The vertical axis acceler-

ates at the same rate but must reach ahigher speed. This transition from zerospeed to maximum retract rate takesabout 8 ms, or 74-deg rotation, which re-sults in pushing the material within thethreads down by about 0.1 mm. This dis-tortion of the thread form left within thekeyhole is much less than the threadspacing, about 1 mm. While there areother sources of axis motion error, suchas backlash in the drive system and de-flection of each axis, the welding tooland the tooling, it is expected that thissmall distortion does not invalidate theconclusions suggested by this technique.

Conclusions

Based upon welds in 6061 and 7075aluminum, it is apparent that in frictionstir welds not all material influenced bythe pin is actually stirred in the weld-ing process. Much of the material move-ment takes place by simple extrusion.The material that is stirred originates fromthe upper portion of the path of the weld-ing tool pin. The stirred material is forceddown in the weld by the threads on thepin and is deposited in the weld nugget.

Other material in the weld zone simplyextrudes around the retreating side of thewelding tool pin, rising in the weld as itgoes around the pin.

The objective of this study was to per-form tests to reveal the material move-ment patterns that take place in frictionstir welds. The conceptual model pro-posed here is preliminary, and addit ionalwork must be undertaken to refine theconceptual model and give a more de-tailed understanding of how friction stirwelds are made, especially to investigate

the behavior of other aluminum alloys.The ultimate goal is to gain sufficient un-derstanding to formulate idealized mod-els of the process, which can be used topredict welding tool forces, temperatureprofi les and other parameters of interest,thereby, all owi ng researchers to evaluatechanges to welding tool design and op-erating parameters and possibly suggestnew i mprovements to the process.

Acknowledgments

The author would like to acknowl-edge the assistance of Lane Ball ard of theBoeing Co. who performed and analyzedthe tracer material welds in 6061, the as-sistance of Chris Dawes of The WeldingInstitute, Cambridge, U.K., for his help inpreparing the welds in 7075 and the as-sistance of Dr. Denny Graham of the Uni-versity of Adelaide and Dr. Murray Ma-honey of the Rockwell Science Center fortheir reviews of the fi rst draft of this paper.

References

1. Thomas, W. M., et al. 1991. Friction StirButt Welding. U.S. Patent No. 5,460,317.

2. Dawes, C. J., and Thomas, W. M. 1995.Friction Stir Joining of Alumi num All oys. TWIBulletin 6, p. 124.

3. Dawes, C. J., and Thomas, W. M. 1996.Friction stir process welds aluminum alloys,Welding Journal75 (3): 41.

4. Christner, B. K., and Sylva, G. D. 1996.Friction stir weld developments for aerospaceapplications. International Conference on Ad-vances in Welding Technol ogy, Joining ofHigh Performance Materials, Columbus,Ohio.

5. Ranes, M., Kluken, A. O., and Midling,O. T. 1995. Fatigue properties of as-welded6005 and 6062 aluminum all oys in T1 and T5temper condition. 4th International Confer-ence on Trends in Welding Science and Tech-nology, Gatlinburg, Tenn.

6. Rhodes, C. G., Mahoney, M. W., andBingel, W. H. 1997. Effects of friction stirwelding on microstructure of 7075 aluminum.Scripta Materialia36(1): 6975.

7. Mahoney, M. W., et al. 1998. Propertiesof friction stir welded 7075 aluminum. Metal-lurgical and Materials Transactions29A.

8. Midl ing, D. T., Morley, G. J., and San-vick A. 1994. Friction Stir Welding. Interna-tional Patent Application WO 95/26254.

9. Thomas, W. M., et. al. 1997. Friction Stir

Welding, U.K., Patent Application GB 2-306366-A.10. H aagensen, P. J., and Midli ng, O. T.

1997. The friction stir welding process me-chanical strength and fatigue performance ofa 7xxx series aluminum alloy. Faculty of Civi lEngineering, The Norw egian University of Sci-ence and Technology, Trondheim, Norw ay.


Documents

1.Material Flow Behavior During Friction Stir