Beruflich Dokumente
Kultur Dokumente
This paper describes the application of the computational fluid dynamics (CFD) code, FLUENT, to
modelling the two-dimensional metal flow in friction
stir welding (FSW). The primary goal is to assist in
the development of new welding tools, though the
models also improve the understanding of the deformation mechanism in FSW, and enable a first order
visualisation of the flow round the probe. The paper
describes a quantitative method of comparing the flow
round different tool shapes. A novel slip model was
developed, where the interface conditions were governed by the local shear stresses. This revealed
significant differences in behaviour when compared
with the common assumption of material stick. The
two-dimensional model has demonstrated the viability
of the approach for investigating the flow round
practical tool shapes, and gives confidence that the
development of more computer intensive threedimensional models will be justified.
STWJ/413
Keywords: frictions stir welding, modelling, material
deformation
Drs Colegrove (pac44@cam.ac.uk) and Shercliff
(hrs@eng.cam.ac.uk) are in the Dept. of
Engineering, Cambridge University, Trumpington
Street, Cambridge CB2 1PZ, UK. At the time this
work was carried Dr Colegrove was also with TWI
Ltd, Granta Park, Great Abington, Cambridge,
CB1 6AL, UK. Manuscript received 16 January
2003; accepted 12 June 2003.
# 2004 Institute of Materials, Minerals and Mining.
Published by Maney on behalf of the Institute.
INTRODUCTION
Friction stir welding (FSW) is a relatively new welding
process, patented in 1991 by Thomas et al.1,2 This process
has great advantages in welding aluminium alloys that are
difficult to weld. The process gives low distortion, can weld
thick sections in a single pass and produces welds with
excellent mechanical properties.
A schematic diagram illustrating the process of FSW is
shown in Fig. 1. The key components of the FSW tool are:
(i) The Shoulder. This is the primary means of
generating heat during the process, and it prevents
material expulsion and assists material movement
around the tool
(ii) The Pin. The pins primary function is to deform the
material around the tool and its secondary function
is to generate heat.
Macrosections from friction stir welds exhibit four
distinct regions. First, there is the dynamically recrystallised
region (DRZ) where the material is recrystallised from
the high temperature deformation. Outside this is the
DOI 10.1179/136217104225021832
DESCRIPTION OF MODEL
Basic model
A computational fluid dynamics package, FLUENT,4 was
selected for the modelling work because it enabled
modelling of the high strain rate flow beside the tool and
permitted the application of a slip boundary condition.
Figure 2a shows the standard set-up for the models. The
physical size of the model was kept small to keep solution
times low. The model is divided into two regions: a rotating
region beside the tool where the mesh moves at the rotation
speed of the tool and a surrounding stationary region,
which models the flow in the far field. All walls on the
boundary of the stationary flow region are given a velocity
equal to the welding speed. Note that the material flows
from left to right and the tool rotates in a clockwise
direction in all the models. The four pin profiles that were
examined are shown in Fig. 2b.
2004
Vol. 9
No. 6
483
484
Model types
Slip model
Most of the published flow models have assumed that
material is stuck to the tool surface. Although Xu et al.6
demonstrated a slip boundary condition with a cylindrical
two-dimensional tool, such a boundary condition has not
been demonstrated on profiled tools.
There are several reasons that justify the development of
the slip model
(i) Based on the material properties for 7075-T6
provided by Jin et al.,5 a three-dimensional stick
model of the process (without significant softening
near the melt temperature) requires a power input
of 1518 kW for the 6.35 mm welds described in
Colegrove and Shercliff.7 This compares with the
measured heat input of 2.63.5 kW. Therefore the
model significantly overpredicts the power requirement. It is believed that one of the causes of this
discrepancy is the assumption that all the material
sticks to the tool surface.
(ii) The temperature measurements in Colegrove and
Shercliff7 and Song and Kovacevic8 show that the
peak temperature near the tool approached the
solidus. Therefore it is likely that a thin partially
liquid layer exists at the tool/material interface
which reduces the shear stress that can be applied
to the material. This phenomenon has been
discussed extensively by North et al.9
(iii) Colegrove and Shercliff7 demonstrated that the
tool material type affected the weld temperature
and quality. Since heat loss through the tool is
quite minimal (thermal models suggest 2% in
steady state), this observation is best explained
Science and Technology of Welding and Joining
2004
Vol. 9 No. 6
All the models used a rotation speed of 250 rev min1 and a
welding speed of 300 mm min21, which are typical conditions for welding thick (16 mm) 7075 aluminium alloy.
Each of the four tool types was examined with a stick
boundary condition, and a slip boundary condition where
the shear stress was limited to 40 or 80 MPa. These values
are used for demonstration purposes only and compare
with a typical shear strength of 90 MPa (approx) for the
unsoftened material at this temperature.
Two model types were analysed. The first was a steadystate model where the solution is a snap-shot of the flow at
a particular instant in time. Provided the time-dependent
terms in the NavierStokes equation remain relatively
small, this approximation will be reasonably good. The
second type of model was a full transient analysis and was
used only for the mixing models and the force analyses.
485
Results for Tool_1: velocity vectors and strain rate52 s21 contour for a stick model and b slip model (limiting shear
stress540 MPa); streamline plots for c stick model and d slip model (limiting shear stress540 MPa)
the distance between the material at the tool surface and the
undeformed region is particularly small on the advancing
side. Nevertheless, the same amount of material still needs
to get past the pin so the deformation region enlarges on the
retreating side. In the stick model, the tool is able to stir
the material more effectively.
A streamline plot of the stick model is shown in Fig. 3c
and for the slip model in Fig. 3d. Note that streamlines are
instantaneously parallel to the flow and are shown for a
particular tool orientation. Particle tracks are much more
difficult to calculate and require a full transient analysis,
however they give a better indication of the flow around the
tool. An attempt to implement particle tracks in FLUENT
was unsuccessful. Both streamline plots are similar to those
reported in Siedel and Reynolds.12 The streamlines are
much nearer the tool for the slip model. In the stick model,
there is an essentially stagnant region, which (although
shearing) does not permit the entry or exit of flow material
because it is shearing parallel to the tool surface. (Note that
flow material refers to the material which flows past the
tool.) This region of parallel shear does not occur with the
slip model and hence the flow material can come into
contact with the tool.
Figure 4a and b show an arrow and deformation plot of
the flow for Tool_4 at two orientations. The flow around
Tool_4 is less circular than that observed for Tool_1,
however it is broadly the same size. There is however
greater perturbation around the features, with the flow
being more dependent on the tool orientation. The
perturbation around the flats leads to the bulging of the
streamlines seen in Fig. 4c and d. The velocity field is
investigated later, but in essence, the material beside the
flats is stuck to the tool surface and therefore has a high
rotational velocity. In comparison the velocity of the
material at the outer circumference is near zero. This
causes the enlargement of the flow field around the flats and
the bulging of the streamlines in this area.
2004 Vol. 9
No. 6
486
Results for Tool_4: a velocity vectors and strain rate52 s21 contour for Tool_4 slip model (limiting shear
stress540 MPa) and b rotated 60u; c streamlines for Tool_4 slip model (limiting shear stress540 MPa) and d rotated
60u
2004
Vol. 9 No. 6
487
a Pathline markers reproduced from London et al.,13 b transverse copper marker in 2024 aluminium alloy mid-section
(courtesy T. Dickerson); c horizontal planar section through a FSW in 2014-T6 (scale on left shows 1 mm intervals);
d dissimilar weld between 2024-T3 and cast AlSiMg (courtesy T. Dickerson)
in Fig. 5b. Both of these models used the slip model with a
limiting shear stress of 40 MPa and were incremented 5u
per iteration. An extra fine mesh was used near the tool so
that the boundary between the two materials could be
tracked.
In models where the two materials were separated along
the joint line, the light material in the flutes was trapped
during start-up, i.e. the model started with the dark
material below the y axis and the light material above.
The discontinuous flow of material round the tool can be
clearly seen in the stick (Fig. 7a) and slip (Fig. 7b) models
2004 Vol. 9
No. 6
488
Material mixing model using Tool_2 and transverse marker; a 4.9 rotations, b 9.9 rotations and c 19.4 rotations. All
are from slip model with limiting shear stress of 40 MPa
2004
Vol. 9 No. 6
489
Surface velocity and shear stress results for Tool_4: a surface velocity and b resultant shear stress for slip model ()
with limiting shear stress of 40 MPa and stick model (- - -); c surface velocity and d resultant shear stress for slip
model () with limiting shear stress of 80 MPa and stick model (- - -)
CONCLUSIONS
The aim of the current work has been the development of a
model that enables the comparison of different tool shapes
and a better understanding of the science behind friction
stir welding. The application of the CFD code FLUENT,
and the development of a novel slip model have made
some significant progress toward that aim.
Good visualisation is provided by plots of the flow. In
particular, the streamlines correlated well with embedded
marker experiments. Qualitatively, the model compared
favourably, though the deformation zone size predicted by
the model was greater than that indicated by the experiments. Great quantitative accuracy is not sought given the
assumptions in the present model, i.e. isothermal conditions
and 2D flow. Finally, the mixing model indicated oscillating
flow which was thought to be the cause of the banded
onion-ring microstructure.
To demonstrate the modelling approach, four different
tool profiles were evaluated, with three types of boundary
conditions: the stick model and two slip models with
limiting shear stresses of 40 and 80 MPa, respectively. The
slip model revealed significant differences with different
tool shapes, which was not evident with the conventional
stick model. Comparison of the traversing force and
welding torque for the constant shear stress models
indicated that Tool_4 was the most efficient, followed by
Tool_2 and Tool_3 and finally Tool_1. It was considered
Science and Technology of Welding and Joining
2004 Vol. 9
No. 6
490
10
Quarter sections of a surface velocity, b velocity vectors and c shear stresses for slip model shown in Fig. 9
Appendix
Applying the slip boundary condition in two dimensions
Defining shear stress and relative velocity in two
dimensions
Define three vectors
n~(xn ,yn ,0)
v
~(0,0,{1)
jvj
dtan ~n|
11
2004
Vol. 9 No. 6
v
~({yn ,xn ,0)
jvj
The first is the normal unit vector, which points toward the
tool surface. The second is the unit direction vector for the
rotation. Its value is negative for clockwise rotation.
Finally, dtan gives the direction of the shear stress, which
will always be parallel to the surface in the direction of
rotation. These vectors are shown in Fig. 12a.
491
vrelt f0
vrelt > 0
: :
: (1)
: :
: (2)
t0 ~tzCvrel
ACKNOWLEDGEMENTS
The authors wish to express their thanks to the University of
Adelaide, The Welding Institute (TWI), the UK Department
of Trade and Industry (DTI), the Post-Graduate Training
Partnership (PTP), the University of Cambridge, and the
Cambridge Commonwealth Trust for their financial support.
The images provided by Terry Dickerson and Blair London,
and the advice given by Philip Threadgill, Chris Dawes,
Nikos Nikiforakis, Wayne Thomas, Mike Russell, and
Richard Johnson were greatly appreciated.
Science and Technology of Welding and Joining
2004 Vol. 9
No. 6
492
10.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
2004
Vol. 9 No. 6
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.