Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: Friction stir welding (FSW) has proved to be a successful joining technology for aluminum alloys and
Received 3 February 2013 many other metallic materials. The severe plastic deformation of solid-state metal during FSW made it
Received in revised form 14 June 2013 a fully coupled thermo-mechanical process. In order to quantitatively study both the total heat genera-
Accepted 1 July 2013
tion and the spatial distribution of the heat flux, a thermo-mechanical coupled model based on compu-
Available online 3 August 2013
tational fluid dynamics was presented in this study. The heat generation, the temperature field and the
material flow pattern were simulated in a fully coupled way. The simulated temperature distribution
Keywords:
agreed well with the experimental results. The total heat generation was found to be proportional to
Friction stir welding
Heat generation
the 0.75 power of the tool rotating speed. The spatial distribution of the heat flux around the FSW tool
Numerical simulation was almost axisymmetric about the tool axis. A radial distribution function was defined to describe
Computational fluid dynamics the heat flux in different rotating rates. The radial distribution function in the shoulder region was fitted
Heat flux distribution to a parabolic function.
Ó 2013 Elsevier B.V. All rights reserved.
1. Introduction generation, the temperature distribution and the material flow pat-
tern were affected by each other in FSW. Since thermal modeling
Friction stir welding (FSW) is an advanced solid-state joining was the basis of other models, evaluation of both total amount
technology invented at The Welding Institute (TWI) in 1991. It is and distribution of heat flux during FSW was required [4]. In order
a very successful joining technology for aluminum alloys and many to evaluate the heat generation in FSW, the underlying physics, such
other metallic materials [1,2]. Specifically, the process was carried as the material flow, should be included. This paper provided a first
out by plunging a rotating FSW tool into the surface of rigidly quantitative investigation of the total amount and the distribution
clamped sheets and traversed along the weld lines. Large amount of the heat flux by a fully coupled thermal–mechanical model.
of heat was generated during the process, and significant plastic In the published works, experimental methods and theoretical
flow was caused by the rotating tool. In result, dynamic recrystal- analysis had been employed to estimate the heat input during
lization was brought to occur. FSW joints were formed as a result FSW. By using the experimental methods, Pew et al. [23] measured
of the severe plastic deformation and dynamic recrystallization. the torque carried by the tool to estimate the heat input during
Key issues about FSW in the published researches included the FSW. Based on the theoretical analysis, different theoretical heat
heat generation [2–6], material flow [1,2,7–10], microstructure source models had been proposed to estimate the heat flux. In
evolution [1,2,11–13] and defects formation [2,14–16] during the these models, the heat flux was calculated based on the contact
process, the resulted properties[1,2,17–19] of the joint and the de- shear stress on the tool/material boundary. Chao et al. [24] defined
sign of the FSW tool [1,2,20]. the contact stress based on the Coulomb’s law to simulate the tem-
Heat generation during FSW was a fully-coupled thermo- perature distribution. Shercliff et al. [25] set the contact shear
mechanical process [3]. The transient temperature distribution dur- stress as 5% of the yield stress at room temperature. Schmidt
ing FSW was related to both the total amount and the distribution of et al. [5,26] proposed to calculate the total heat generation and
the heat generation. Meanwhile, the heat generation during plastic its distribution analytically from the uniformly distributed contact
deformation was also related to the mechanical property, i.e. the shear stress on the tool/material boundary. In Schmidt et al.’s stud-
flow stress of the metal, and the transient deformation rate ies [5,26], the contact shear stress was determined by the contact
[3,21,22]. Note that the flow stress of aluminum alloy was state. Li et al. [27] provided an auto-adaptive heat source model
significantly temperature and strain rate dependent, the total heat for FSW by assuming a temperature dependent contact state.
However, the material flow pattern was difficult to be taken into
consideration by using the experimental methods and theoretical
⇑ Corresponding author. Tel.: +86 10 62784578; fax: +86 10 62773637.
analysis.
E-mail address: shqy@tsinghua.edu.cn (Q.-y. Shi).
0927-0256/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.commatsci.2013.07.004
G.-q. Chen et al. / Computational Materials Science 79 (2013) 540–546 541
In recent years, numerical models based on Computational So- where q was the density, l was the non-Newtonian viscosity, p was
lid Mechanics (CSM) and Computational Fluid Dynamics (CFD) v was the velocity of material flow. The energy
the pressure and ~
had been established to simulate the material flow during FSW. conservation equation was given by:
Based on CSM method, Schmidt et al. [28] established a material
flow model for FSW. The Arbitrary Lagrangian–Eulerian (ALE) v HÞ ¼ r ðkrTÞ þ SV
r ð~ ð3Þ
formulation was adopted to avoid unacceptable element distor- where H was the enthalpy, T was the temperature, k was the ther-
tion. The frictional force given by the Coulomb’s law of friction mal conductivity and SV was a spatial source term.
was applied on the tool/material boundary. Heat was generated
from the sliding friction on the boundary and plastic deformation
in the model. Both the velocity and temperature could be simu- 2.2. Geometric model
lated. Similar models based on CSM were also provided by Zhang
et al. [29] and Grujicic et al. [30]. In these models, however, the The geometric model was shown in Fig. 1. The geometric model
material flow pattern was obtained only when the tool penetrated was obtained by subtracting the FSW tool from the work piece. The
through the work pieces [6–8]. Because of the limitation in the size of the work piece was 60 mm 40 mm 3.1 mm. The tilt an-
above geometric models, the physical processes represented by gle and the concave shoulder were not employed. The FSW tool
was taken as a rigid body, and was not included in the geometric
models based on CSM were significantly different from real FSW.
The CFD method was another important approach employed to model. The geometric model was meshed by hexagonal grids be-
fore simulation. The diameter of the tool shoulder was 12 mm,
simulate the material flow pattern in FSW. The work presented by
Colegrave et al. [9] was one of the earliest 3D thermal–mechanical- and the pin diameter was 3 mm at the root. The pin length was
2.8 mm. The rotating direction and welding direction were shown
coupled models of FSW based on CFD. In the study, the material
viscosity was considered as a function of strain rate and tempera- in the figure. The welding parameters were taken to be 920 rpm,
20 mm min1, the same as Ref. [32].
ture. The material close to the tool was assumed to stick on the
tool. Heat was generated fully by viscous dissipation. Based on this
model, both the temperature and material flow pattern were sim- 2.3. Material properties
ulated, but both the temperature and the size of the deformation
zone were over-predicted. Nandan et al. [10] calculated the heat Aluminum alloy 6061 was selected as the target material. The
input from the contact stress in their material flow model based density of AA6061 was assumed to be constant and was taken as
on CFD. In their model, the 3D distributed heat generation from 2700 kg m3. Temperature dependent thermal physical properties
viscous dissipation was taken as a small fraction of the total heat were taken from Ref. [33]. The viscosity of the material was con-
generation. They reported that the asymmetric feature of the tem- sidered to be both temperature and strain rate dependent. The
perature field would not be simulated if the small part of heat from viscosity was calculated based on the following formulation
the viscous dissipation was ignored. Colegrave et al. [6] established [9,34]:
a 2D thermo-fluid model, coupling with a 3D thermal model to cal- r
culate the heat generation from the axisymmetric flow pattern and l¼ _ ð4Þ
3e
the material properties. In their model, the heat generation was
coupled with the material flow pattern. where r was the flow stress and e_ was the strain rate. The expres-
Though lots of researches have been carried out to simulate the sion of flow stress was given by [35]:
material flow during FSW, quantitative analysis of the total " 1=n #
amount and the spatial distribution of the heat flux during FSW 1 e_ Q
re ¼ sinh1 exp ð5Þ
has not been reported yet. This paper provides a first attempt to a A RT
investigate both the total amount and the distribution of the heat
flux by a fully coupled thermal–mechanical model based on CFD. where A (=8.86 106 s1), a (=0.045 MPa1) and n (=3.55) were
Different from the published researches, this paper improves the material constants, Q (=145 kJ mol1) was the temperature-inde-
constitutive equations in order to predict the flow stresses at high pendent activation energy, and R was the gas constant. The param-
temperatures more accurately. Based on the improved constitutive eters in the above Eq. (5) were taken from Ref. [36].
model, the heat generation, material flow pattern and the temper- It could be found from Fig. 2 that large discrepancy existed be-
ature distribution are simulated. The computed temperature distri- tween the predicted flow stress by Eq. (5) and the measured yield
bution is compared with a set of published measured data. Special stress. In this paper, the expression of flow stress in Eq. (5) was
attention is paid to quantitative evaluation of the total heat gener- modified by multiplying a temperature dependent factor. Fig. 2
ation and its spatial distribution. Simulations using different weld- showed that the modified flow stress agreed well with the yield
ing parameters are carried out to investigate the effect of welding stress especially at high temperatures. Similar modifications of
parameter on both the total heat generation and spatial distribu-
tion of the heat flux.
2. Simulation model
v~
r ðq~v Þ ¼ rp þ r ðlðr~
v þ r~
v T ÞÞ ð2Þ Fig. 1. Geometric model in the simulation. (0, 0, 0) was the origin of coordinates.
542 G.-q. Chen et al. / Computational Materials Science 79 (2013) 540–546
3. Simulation results
where r was the flow stress, g was a coefficient. g ¼ 0:80 [2] was 3.3. Heat generation
assumed in this paper, which meant that 80% of plastic work was
dissipated into heat, and e_ was the effective strain rate. Both the total amount and the distribution of the heat flux were
critical for understanding of the FSW process. As the tool rotating
2.5. Boundary conditions contributed much more to the material flow in FSW than the weld-
ing speed, the heat generation rate was found to be influenced lar-
The boundary conditions described in this part were set to ap- gely by the rotating rate [2]. In order to investigate how the heat
ply the welding direction and the thermal conditions. As shown generation varied with the welding parameters, simulations were
in Fig. 1, a mass flow at the welding speed was defined to enter carried out using different rotating rates. The total heat generation
the computation domain from the velocity inlet. Pressure p was and the heat flux distribution were discussed in this part.
set to be a constant value, 0 Pa, in the pressure outlet in the model.
At the boundary between the tool and material, the velocity vector 3.3.1. Total heat generation
was defined as the same as the tool in order to simulate tool rotat- The total heat generation was obtained by integrating the spa-
ing. The wall velocity was set to be the opposite speed of welding tially distributed heat flux over the entire computing domain. Fig. 6
speed on the back surface and two sides. The bottom and side work showed the total heat generation and the highest temperature in
piece were contact with the fixtures, hence heat was transferred to different welding parameters. The total heat generation increased
the fixtures. In the model, the contact heat transfer coefficient was greatly with the tool rotating rate. It was clear that the total heat
defined to simulate the heat transferred to the fixtures. The contact generation increased with the rotating rate. As shown in Fig. 6,
pressure on the back surface under the tool was much higher than the total heat generation was found to be proportional to x0.75 in
other area, thus the contact heat transfer was defined to be magnitude, given as,
15,000 W (m2 K)1 under the tool and 800 W (m2 K)1 in other
area. Q total ¼ Q C x0:75 ð7Þ
G.-q. Chen et al. / Computational Materials Science 79 (2013) 540–546 543
Fig. 4. The simulated temperature with different welding parameters. (a) Temperature distribution of the entire workpiece, (b) Temperature distribution on the traverse
section.
Fig. 5. Comparison between the simulated peak temperature and the measured
peak temperature at the mid-thickness plane. The measured temperature data was
taken from Ref. [32]. Fig. 6. The simulated total heat generation with different rotating rates.
Fig. 7. The simulated volumetric heat flux inside the work piece with different welding parameters.
In this paragraph, the simulation result in 920 rpm, was relatively high. So the second highest volumetric heat flux was
20 mm min1 was taken as an example to analyze the simulated obtained. It could be found from Fig. 8a that the heat flux in the
distribution of the volumetric heat flux, strain rate and flow stress, area around the corner between the shoulder and the pin was rel-
as shown in Fig. 8. It could be found that the simulated distribution atively low, because the strain rate and the flow stress were rela-
of the volumetric heat flux, strain rate and flow stress were shown tively low in this area (Fig. 8b and c).
in Fig. 8. The volumetric heat flux exceeded 1 1010 W/m3 in the In order to investigate the radial distribution of heat generation,
following two areas. One of the areas was located under the shoul- a global heat flux, qglobal , at different distances from tool axis was
der periphery. In this area, the strain rate was extremely high, as calculated. The heat flux was defined as
shown in Fig. 8b, resulting in the high flow stress (Fig. 8c) despite
the high temperature (Fig. 4). Thus the highest volumetric heat flux dQ total
was obtained. The other area was around the pin tip. In this area, qglobal ¼ ð8Þ
2pr dr
the temperature (Fig. 4) was relatively low due to the heat trans-
ferred into the back plate, thus the flow stress was as high as more
than 50 MPa (Fig. 8c). Besides, the strain rate in this area (Fig. 8b) where r denoted the distance from the tool axis. The global heat flux
described the heat generation as a function of r. The plot of the heat
flux was provided in Fig. 9. In the plot of the radial distribution of
the global heat flux, it could be easily found that the heat input in-
creased with the rotating rate. There were two peaks in the plotted
curve. One corresponded to the heat generation from the pin side,
and another corresponded to the heat generation at the periphery
of the shoulder. As shown in Fig. 9, the curve could be departed into
three regions, which were the pin-bottom region, the pin-side re-
gion and the shoulder region. In the pin-bottom region and the
shoulder region, the heat flux increased gradually as the distance
increased. Due to the plastic deformation caused by the pin side,
the heat flux was very high at the distance corresponding to the
pin radius. In addition, the heat flux decayed rapidly when the dis-
tance exceeded the shoulder periphery.
Different distribution functions describing the distribution of
the heat flux in welding process had been proposed for fusion
welding, such as Gaussian distribution function [40] and double
ellipsoid [41]. Note that the global distributions in different weld-
ing parameters were similar to each other (Fig. 9). We defined a
Fig. 8. The simulated distribution of the volumetric heat flux, strain rate and flow
stress on the traverse section (920 rpm, 20 mm min1). (a) Volumetric heat flux, (b)
Stain rate, (c) Flow stress. Fig. 9. The global heat flux in different welding parameters.
G.-q. Chen et al. / Computational Materials Science 79 (2013) 540–546 545
distribution function and expressed the heat flux as the following proper form of the analytical function had not been established
expression, based on the research in this stage.
References
[1] R.S. Mishra, Z.Y. Ma, Mater. Sci. Eng. R 50 (2005) 1–78.
[2] R.S. Nandan, T. DebRoy, H.K.D.H. Bhadeshiab, Prog. Mater. Sci. 53 (2008) 980–
1023.
[3] K.J. Colligan, R.S. Mishra, Scripta Mater. 58 (2008) 327–331.
[4] H.B. Schmidt, J.H. Hattel, Scripta Mater. 58 (2008) 332–337.
[5] H. Schmidt, J. Hattel, J. Wert, Modell. Simulat. Mater. Sci. Eng. 12 (2004) 143–
157.
[6] P.A. Colegrove, H.R. Shercliff, R. Zettler, Sci. Technol. Weld. Joi. 12 (2007) 284–
297.
[7] Y. Li, L.E. Murr, J.C. McClure, Mater. Sci. Eng. A 271 (1999) 213–223.
[8] T.U. Seidel, A.P. Reynolds, Metall. Mater. Trans. A 32 (2008) 2879–2884.
[9] P.A. Colegrove, H.R. Shercliff, J. Mater. Process. Technol. 169 (2005) 320–327.
[10] R. Nandan, G.G. Roy, T.J. Lienert, T. Debroy, Acta Mater. 55 (2007) 883–895.
[11] J.Q. Su, T.W. Nelson, R.S. Mishra, M. Mahoney, Acta Mater. 51 (2003) 713–729.
[12] P.B. Prangnell, C.P. Heason, Acta Mater. 53 (2005) 3179–3192.
[13] W. Woo, Z. Feng, X.L. Wang, K. An, B. Clausen, T.A. Sisneros, J.S. Jeonng, Mater.
Lett. 85 (2012) 29–32.
[14] Y.G. Kim, H. Fujii, T. Tsumura, T. Komazaki, K. Nakata, Mater. Sci. Eng. A 415
(2006) 250–254.
[15] W.J. Arbegast, Scripta Mater. 58 (2008) 372–376.
[16] R. Crawford, G.E. Cook, A.M. Strauss, D.A. Hartman, M.A. Stremler, Sci. Technol.
Weld. Joi. 11 (2006) 657–665.
[17] M.W. Mahoney, C.G. Rhodes, J.G. Flintoff, R.A. Spurling, W.H. Bingel, Metall.
Mater. Trans. A 29 (1998) 1955–1964.
[18] M. Peel, A. Steuwer, M. Preuss, P.J. Withers, Acta Mater. 51 (2003) 4791–4801.
[19] A.P. Reynolds, W. Tang, T. Gnaupel-Herold, H. Prask, Scripta Mater. 48 (2003)
1289–1294.
[20] R. Rai, A. De, H.K.D.H. Bhadeshia, T. DebRoy, Sci. Technol. Weld. Joi. 16 (2011)
325–342.
[21] M. Zhou, M.P. Clode, Mech. Mater. 27 (1998) 63–76.
[22] R.L. Goetz, S.L. Semiatin, J. Mater. Eng. Perform. 10 (2001) 710–717.
[23] J.W. Pew, T.W. Nelson, C.D. Sorensen, Sci. Technol. Weld. Joi. 12 (2007) 341–
347.
[24] Y.J. Chao, X. Qi, Heat transfer and thermomechanical analysis of friction stir
joining of AA6061-T6 plates, in: 1st Int. Symp. On Friction Stir Welding, CA,
USA, 1999.
[25] H.R. Shercliff, P.A. Colegrove, Math. Model. Weld. Phenom. 6 (2002) 927–974.
Fig. 10. The radial distribution function of heat generation in different welding [26] H. Schmidt, J. Hattel, Sci. Technol. Weld. Joining 10 (2005) 176–186.
parameters. (a) The global heat flux divided by the total heat generation. (b) The [27] H.K. Li, Q.Y. Shi, T. Li, W. Wang, Mater. Sci. Forum 580–582 (2008) 267–270.
global heat flux divided by QC x0.75. [28] H. Schmidt, J. Hattel, Model. Simulat. Mater. Sci. Eng. 13 (2005) 77–93.
546 G.-q. Chen et al. / Computational Materials Science 79 (2013) 540–546
[29] Z. Zhang, J.T. Chen, J. Mater. Sci. 43 (2008) 222–232. [35] T. Sheppard, D.S. Wright, Metal. Technol. 6 (1979) 215–223.
[30] M. Grujicic, G. Arakere, B. Panduangan, J. Mater. Eng. Perform. 20 (2011) 1097– [36] T. Sheppard, A. Jackson, Mater. Sci. Technol. 13 (1997) 203–209.
1108. [37] T.U. Seidel, A.P. Reynolds, Sci. Technol. Weld. Joining 8 (2003) 175–183.
[31] FLUENT, Release 6.3.26, Fluent Inc., Lebanon, NH, 2006. [38] J.R. Davis, ASM Specialty Handbook – Aluminum and Aluminum Alloys,
[32] Y.M. Hwang, Z.W. Kang, Y.C. Chiou, H.H. Hsu, Int. J. Mach. Tools Manuf. 48 Materials Park, OH, 1993.
(2008) 778–787. [39] J.W. Qian, J.L. Li, J.T. Xiong, F.S. Zhang, W.Y. Li, X. Lin, Sci. Technol. Weld. Joining
[33] K.C. Mill, Recommended Values of Thermophysical Properties for Selected 17 (2012) 338–341.
Commercial Alloys, Woodhead Publishing Ltd and ASM International, [40] V. Pavelic, R. Tanbakuchi, O.A. Uyehara, P.S. Myers, Weld. Joining 48 (1969)
Cambridge, 2002. 295s–305s.
[34] O.C. Zienkiewicz, I.C. Cormeau, Int. J. Numer. Methods Eng. 8 (1974) 821–845. [41] J. Goldak, A. Chakravarti, M. Bibby, Metall. Trans. B 15B (1984) 299–305.