Proc 2nd Int FSW Symp,00-An Exp Validated 3D Model for FSW

8/12/2019 Proc 2nd Int FSW Symp,00-An Exp Validated 3D Model for FSW

1/12

An Experimentally Validated 3D Model for Friction Stir Welding

G.J. Bendzsak*, T.H. North* and C. B. Smith**

* Department of Metallurgy and Materials Science,University of Toronto, Ontario, Canada

** Tower Automotive Inc., Milwaukee, Wisconsin., USA

Introduction The work described in the present paper is part of a multi-year program by Tower

Automotive Inc. aimed at developing a validated process model for friction stir welding.

This program is being carried out in distinct phases, where Phase 1 involves thedetermination of unknown material properties (particularly material viscosity) and Phase 2

involves the development and validation of a 3-D process model for friction stir welding.This paper presents the preliminary results obtained using a recently developed 3-D heatand material flow simulation program for friction stir welding (STIR3DTMsoftware1). For

illustrative purposes the flow structure produced using a tool designed by The WeldingInstitute (TWI) is analyzed. The intent is to demonstrate an approach that successfully

analyses physical phenomena within the weld. For the sake of brevity, mathematical andcomputer related details required for implementation of the simulation will be presentedelsewhere.

The essence of the simulation is that the dynamics within the weld zone can be exploredthrough solutions of the 3-D Navier-Stokes equations and the principles of non-linear

mechanics. Flow patterns are calculated using the equations of fluid mechanics and their

dynamics are then examined by following the motions of tracers embedded within the flow.Concerning heat transfer, viscous dissipation of mechanical energy is the heat source [1,2].

As a result, the simulation does not resort to the artifact of a coefficient of friction for heatgeneration. The approach ensures both consistent and unified application of the relevant

equations of mathematical physics and yields results that correspond with the featuresobserved in actual welds.The inputs to STIR3DTMsoftware involve tool geometry, material properties and operational

conditions such as welding speed, tool rotational velocity, travel (tilt) angle and depth ofimmersion. STIR3DTM software generates a wide range of outputs for analysis and design.

These include:A) 3-D material velocity profiles and pressure distributionsB) temperature and heat generation fieldsC) complete shear and stress mapsD) force distributions on all tool surfacesE) tool forces and torques

In Phase 1 of this program, the parameters of a non-Newtonian, temperature dependentviscosity were derived for the 3-D process simulation [3]. It was also confirmed that the

eutectic temperature prevailed in material immediately adjacent to the tool periphery. Withthis in mind, it is assumed in the initial 3-D simulations that the material surrounding the tool

pin is at the eutectic temperature and that the required value of viscosity corresponds to thistemperature. It is assumed that variations in the thermal profiles around the tool and within

1Trade mark pending


2/12

the nugget have only secondary effects on the flow dynamics. The influence of thermal

profiles on the flow structure will be included in the model and described in a future paper.An Overview of the Model

The flow dynamics produced using the threaded TWI tool design are used to illustratethe numerical solution of the coupled Navier-Stokes and heat transfer equations. The

computational techniques employed in the solution are modifications of those described inreference [3] and will not be presented here

The entire problem is first divided into a convenient number of regions. In turn, each

of these regions is sub-divided into a set of non-orthogonal control volumes. These arerequired by the numerical approximation schemes for the flow and heat transfer equations.Meshes for the regions have varied densities and this allows for a higher mesh density near

the tool in order to reveal the fine structure of the flow. Varied mesh densities allow moredetailed output of the critical regions close to the tool and reveal the fine structures embedded

within the flow. A mesh generated in any region is independent of any other and does notrequire matching at region interfaces. For illustrative purposes, the flow region formed bythe plate and the tool has been divided into 41 separate sub-domains, see Figure 1.

The problem, as defined by the sub-domains, is divided into two separate inter-linkedparts:

A)An inclined, sliding grid system bounded by a tilted cylinder that both encloses thetool and rotates with it. The effects of the changing geometry caused by the rotation ofthe inclined tool are automatically accounted for by a set of continually deforming

grids. This allows the simulation to account for variations in the travel angle.B) A fixed frame geometry for the plate that holds the tool region. The cylinder in a)

above translates along a fixed surface. After each step of the tool rotation thegeometrical configuration at the interface is re-visited to determine the contactrelationships between the control volumes on either side of the translating surface.

Bead-on-plate welds were made using a threaded TWI tool design shown in Figure 2. Thegeometry of the tool shoulder was simplified to reduce the complexity of an already

complicated model. The welding parameters were:A) welding speed of 20 mm/secB) counter-clockwise tool rotation at 1200 rpmC) angle of inclination of 30from the verticalD) the tool heel was 0.2 to 0.3 mm below the plate surfaceE) 7 mm thick 6061-T6 aluminum plate was used throughout

Assumptions

The assumptions comprise:A) The material properties of 6061-T6 are those indicated in reference [3]B) Constant temperature conditions prevail in every region in Figure 1. This

simplification is not inherent in the formulation of the problem; rather it was broughtabout by computer resource limitations associated with the solution of a complex fluid

flow problem. Simply put, there was not enough memory available to include anotherset of equations. This problem is presently being eliminated by an implementation ofSTIR3DTM software on a large parallel computational facility to eliminate memory

constraints and reduce the solution time.C) Since temperature effects on viscosity are not included in the analysis, viscosity is

assumed to be constant and equal to that at the eutectic temperature. Comparisonsbetween computed results and the observed flow patterns in actual welds reveal that

this assumption does not result in unrealistic predictions of flow dynamics.


3/12

D) The solution was initiated by assuming that a tool embedded within the aluminumplate was impulsively started and instantaneously attained the welding velocity.

Typical Resul ts

Large variations in the magnitude of velocity present difficulties in presentation and

interpretation of the test results. For this reason, all vector magnitudes in the paper areuniform and are set equal to 1.5% of the maximum shoulder velocity. This simple expedientallows the examination of the desired flow features without obstruction by large vectors in

the vicinity of tool surface. Also, a better understanding of the flow can be obtained byviewing what happens on cylindrical surfaces around the tool since this is the natural

coordinate system within the nugget. Viewing of cylindrical surfaces readily highlights theeffects produced by several singularities within the flow.Plan Vi ews

Figures 3a to 3c illustrate the flow features observed in different through-thicknesssections. Counterclockwise flow occurs immediately below the shoulder (see Figure 3a) and

is consistent with published results indicating the transfer of material (from the advancing tothe retreating side of the weld) as the tool rotates [4,5,6]. However, at a distance of 3 mm

below the shoulder the flow direction is not counterclockwise, it is changed to extrusion of

material past the rotating pin (see Figure 3b). A transition region separates thecounterclockwise and extrusion flow regimes (see Figure 3c), is about 600 m wide and is

located 1.5 mm below the shoulder.On the advancing side of the weld, at a point within the transition zone the component

of rotational velocity will be opposite to that of tool motion, reducing the net vectorcomponent in the welding direction to zero. Above and below this location the flow will bein opposite directions generating a vortex. The physical situation in a friction stir weld is

consequently similar to the establishment of the reverse flow typically found under the

rotating helicopter wing. Tool rotation will shift the zone of instability in the throughthickness direction and small markers released in close vicinity of each other on the leading

side of the weld will follow drastically different paths. Another way of describing thisphenomenon is to say that the dynamics within a particular region in the weld are very

sensitive to the initial conditions and the flow can become chaotic in nature [7]. For thisreason the transition region located at the advancing side of the weld is termed the chaotic

flow zone for the remainder of this paper. In Figure 3c, the line AB represents the

demarcation between the stable and chaotic regions. Line AB will be displaced towards theleading edge of the weld for planes above Figure 3c; it will be moved towards the trailing

edge of the weld on planes below Figure 3c.Figure 4 shows a cylindrical surface located at a radial distance 3 mm from the axis

of tool rotation. Two spirals (S1 and S2) and an unstable saddle (C) are observed and thesewill noticeably affect the flow structure. For example, spiral S1 will direct material towardsthe pin surface (see later) and spiral S2 forces material in the nugget to form the vertical wall

feature typically observed on the advancing side of friction stir welds.Tr ansverse Sections

No significant features in the flow are observed at the retreating side of the weld.

Weld microsections are also relatively featureless at the retreating side of the weld. Thiscorrelates with the fact that it is difficult to produce weld defects on the retreating side of the

weld and that cross-sections show few distinctive features on the retreating side of the weld.For this reason, the calculated results will only be presented for the advancing side of theweld. Figure 5 shows a transverse section 5 mm behind the tool centerline. A number of

features are exhibited, e.g., the spiral S1 is 3-dimensional and has its midpoint 3 mm from theweld centerline. This spiral forces material towards the tool periphery and a wall-like feature


4/12

is produced by the upward flow of material (W1 in Figure 5). A boundary forms between

two flow regimes (line AB BC) and has similar features to those in weld micrographs of lappenetration welds (see Figures 6 and 7). This vertical flow is also responsible for the

interface deformation that is seen in lap penetration welds, using the tool with a threaded pin.Pin Thr eads

Figure 8 shows the flow behavior associated with the threads on the TWI tool (at thelocation 2 mm below the shoulder, in the chaotic flow zone). Material contained within thetool thread space moves with the tool as it rotates in a counterclockwise direction. However,

at the advancing side of the weld material is removed from the thread space and forms astream at the trailing edge of the tool (P in Figure 8). It is suggested that this effect, incombination with spiral S2, produces the flaw shown in Figure 9.

Tracer Paths

Figure 10a shows a 3-D view of typical particle paths for release points A to E, which

are located 5 mm from the centerline of the weld at the leading edge of the tool. SingularityS1 dramatically alters the particle paths depending on their release point. For example,

particles can proceed along the length of the weld (see A and B in Figure 10a). Alternately

they may follow complicated spiraling motions towards the tool surface and eventuallybecome trapped in the region below the tool shoulder (see C, D, E in Figure 10a). In this

connection, Colligan [5] also observed erratic particle paths when he introduced steel tracerparticles into Al 6061 and Al 7075 friction welds, e.g., particles located 2 mm below thesurface of the aluminum plate were significantly lifted as they passed the pin.

In a similar manner, Figure 10b shows the particle paths produced for release points 1mm ahead of the traversing tool (at the location 1 mm below the shoulder). Spiral S1 causes

a suddenly reversal in the direction of particle motion, so that a particle moving towards theadvancing side of the weld suddenly reverses its direction and moves in a counterclockwisedirection with the tool. Flow reversal test results were also observed when Al 5054 inserts

were introduced during friction stir welding of Al 2195 base material [6].The apparently erratic particle motions observed when insert materials were

introduced in Al 6061, Al 7075 and Al 2195 friction stir welds [5,6] are symptomatic ofsingularities embedded within the flow structure. It is worth emphasizing that the use ofinsert can illustrate the general features of the flow structure in friction stir welds. However,

numerical modeling is required to fully understand the detailed features of the flow structurein completed joints.

Discussion

The present paper has provided evidence that STIR3DTMsoftware can model the flowstructure and flow irregularities that correspond with microstructural features observed in

actual welds. The formation of a chaotic flow region at the location immediately below the

shoulder (on the advancing side of the weld) is of particular importance. This is the regionwhere spiral and saddle singularities are embedded within the flow structure and thesecorrespond with the location of defects in completed welds. Possible mechanisms forsingularity formation comprise:

a) The interaction between the flow generated by the shoulder and material introducedon the advancing side of the stir weld.

b) Motion of material within the tool thread space and the resulting vorticity, whichcould destabilize the local flow regime.

c) Some combination of a) and b) above.It must be emphasized that the present paper examines the output produced using a

single set of welding parameters and a particular tool design. It is suggested that the form of

the chaotic flow zone and the flow irregularities contained within it will be affected by therotational speed selected, the travel speed during welding and, particularly, by the geometry


5/12

of the tool (the relative diameters of the shoulder and pin regions, the thread dimensions and

thread geometry). The combination of fluid mechanics and non-linear dynamics can predictthe features of the chaotic flow zone. This may provide information that will allow the

identification of operating zones where defect-free friction stir welds can be produced in aconsistent manner.

ConclusionsThis paper shows that computational fluid mechanics can be successfully used to

analyze flows in friction stir welds. When using the TWI tool design, the calculated flow

structures using STIR3DTM software correspond with features observed in actual weldmicrographs after there has been considerable forward motion of the tool. The present resultsindicate that three quite distinct flow regimes are formed below the tool shoulder, namely,

i) A region of rotation immediately below the shoulder (where flow occurs in thedirection of tool rotation)

ii) A region where material is extruded past the rotating tool; this occurs towards thebase of the piniii) A region of transition in between i) and ii) where the flow has chaotic behaviorThe development of these regions within the weld is thought to be the root cause for creationof defects. The essential features of the flow structure are established at the initiation of tool

forward motion and instabilities contained within the flow structure will remain for the wholelength of the welded joint. Bearing this in mind, a defect produced when the forward motionof the tool is initiated will be continually extended as the tool traverses the workpiece. If a

defect develops at some location during the friction stir welding operation, it will then becontinually extended as the component is fabricated.

References

1. Bendzsak, G.J., North, T.H. and Li, Z. 1997, Numerical model for steady-state flow infriction welding, Acta Metall. Mater., 45(4), pp. 1735-1745.

2. Bendzsak, G.J. and North, T.H.; 1997, Numerical Modelling of Fluid Dynamics and HeatTransfer in Friction Welding Proc. International Seminar, Numerical Analysis of

Weldability, Graz, Austria, Sept. 30th-Oct. 3rd.3. Smith, C.B., Bendzsak, G.J., North, T.H., Hinrichs, J.F., Noruk, J.S. and Heideman, R.J.

1999, Heat and Material Flow Modeling of the Friction Stir Welding Process,

Proceedings of the 9th International Conference in Computer Technology in Welding,Detroit, MI, Sept. 28 30.

4. Thomas, W.M., Threadgill, P.L. and Nicholas, E.D. 1999, Feasibility of friction stirwelding steel, Science and Technology of Welding and Joining, vol. 4, no.6, pp. 365 372.

5. Colligan, K, Welding Journal, 1999, Material flow behavior during friction stir weldingof aluminium, July, pp. 229s - 237s.6. Reynolds, A., 1999, Visualisation of material flow in autogenous friction welds, Scienceand Technology of Welding and Joining, vol.5, No. 2, pp.120 124.

7. Otino, J.M. The kinematics of mixing; stretching, chaos and transport, CambridgeUniversity Press, Cambridge, 1989.


6/12

Figure 1:Schematic of the simulated FSW process


7/12

Figure 2: Schematic of the right-hand threaded TWI tool

Figure 3a Counter-clockwise rotation of the nugget approximately 1 mm below

shoulder


8/12

Figure 3b Extrusion approximately 3 mm below shoulder

Figure 3c Plan view in the chaotic region approximately 1.5 mmbelow shoulder


9/12

Figure 4 Flow structure on a cylindrical surface 3mm from tool centerline

Figure 5 Flow features in a transverse cut 3 mm behind tool centerline


10/12

Figure 6 Transverse weld micrograph at location approximately corresponding to thatof Figure 5 showing vertical wall on advancing side

Figure 7 Weld micrograph exhibiting features along lines AB BC in Figure 5


11/12

Figure 8 Flow features between tool threads within the chaotic zone

Figure 9 Micrograph exhibiting weld defect in the vicinity of spiral S1


12/12

Figure 10a 3D particle dynamics at the tool leading edge

Figure 10b Reverse particle motion on the tool leading edge

Proc 2nd Int FSW Symp,00-An Exp Validated 3D Model for FSW

Documents