Journal of Materials Processing Technology 191 (2007) 356–359

CDRX modelling in friction stir welding of AA7075-T6

aluminum alloy: Analytical approaches
G. Buffa a,∗ , L. Fratini a , R. Shivpuri b
a Dipartimento di Tecnologia Meccanica, Produzione e Ingegneria Gestionale Università di Palermo, Viale delle Scienze 90128 Palermo, Italy
b The Ohio State University, Department of Industrial, Welding and Systems Engineering 1971 Neil Avenue,

210 Baker Systems, Columbus, OH 43210, USA

Abstract
Friction stir welding (FSW) is an energy efficient and environmentally “friendly” (no fumes, noise, or sparks) welding process, during which the
workpiece are welded together in a solid-state joining process at a temperature below the melting point of the workpiece material under a combination
of extruding and forging. Significant microstructural evolution takes place during FSW: in particular continuous dynamic recrystallization (CDRX)
phenomena result in a highly refined grain structure in the weld nugget and strongly affect the final joint resistance. In the paper two different
analytical models aimed to the determination of the average grain size due to continuous dynamic recrystallization phenomena in FSW processes of
AA7075-T6 aluminum alloys have been implemented in a 3D FEM model and numerical analyses of the welding processes have been performed
to verify their effectiveness.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Friction stir welding; Continuous dynamic recrystallization; FEM

1. Introduction the composition of the tool spin vector and of the feed rate vec-
tor determines a peculiar metal flow all around the tool contact
Friction stir welding (FSW) of butt joints is obtained insert- surface [2,3]. What is more, considering a section of the joint nor-
ing a specially designed rotating pin into the adjoining edges of mal to the tool movement direction, an asymmetric metal flow is
the sheets to be welded and then moving it all along the welding obtained. An advancing side and a retreating one are observed
line [1]. The tool is characterized by a rather small nuting angle in the joint section: the former is characterized by the “posi-
(θ) limiting the contact between the tool shoulder and the sheets tive” composition of the tool feed rate and of the peripheral tool
to be welded just to about one half of the shoulder surface. As the velocity; on the contrary, in the latter the two velocity vectors are
pin is inserted into the sheets, the blanks material undergoes to opposite. Overall, the tool action determines the material soft-
a local backward extrusion process up to reach the tool shoulder ening and, what is more, the metal flux which allows the blanks
contact. The tool rotation determines an increase of the material welding.
temperature due to the friction forces work. As a consequence Recently, a few research activities have been developed on
the material mechanical characteristics are locally decreased and the numerical simulation of FSW processes [4,5]. In previ-
the blanks material reaches a sort of “soft” state; no melting is ous papers [6,7] the authors presented a 3D fully coupled
observed, a circumferential metal flow is obtained all around thermo-mechanical FEM model in which the tool–workpiece
the tool pin and close to the tool shoulder contact surface. As interaction in FSW of butt joints was investigated. In par-
such material softening is obtained, the tool is moved along the ticular in Ref. [7] the material microstructure evolution was
joint avoiding the pin fracture due to excessive material reac- taken into account through a proper model of grain size evolu-
tion. The tool movement determines heat generation due to both tion due to recrystallization phenomena. It should be observed
friction forces work and material deformation one. Furthermore, that in FSW processes a continuous dynamic recrystallization
phenomenon (CDRX) [8] occurs due to the tool pin action.
The tool stirring action generates fine, equiaxed, recrystal-
∗ Corresponding author. Tel.: +39 091 665 7051; fax: +39 091 665 7039.
lized grains; such new microstructure determines the local
E-mail addresses: [email protected] (G. Buffa), [email protected] material mechanical properties and the overall joint resistance
(L. Fratini), [email protected] (R. Shivpuri). [9].

0924-0136/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2007.03.033
 G. Buffa et al. / Journal of Materials Processing Technology 191 (2007) 356–359 357

In the present paper in order to predict the average grain

size occurring in FSW of AA7075-T6 sheets due to the CDRX
phenomena and to all the metallurgical effects occurring due to
the welding process, two different analytical approaches have
been followed. In particular, two analytical models of grain size
evolution have been implemented in the FE model. The models,
once characterized for the utilized materials, are able to calculate
the local value of the average grain size on the basis of the most
relevant process field variables, namely temperature, strain and
strain rate. The effectiveness of the two proposed tools was tested
on the basis of experimental evidences, i.e. the measured grain
sizes, for further welding processes.
Fig. 2. The FEM model.
2. Experiments and FE model
constant interface heat exchange coefficient of 11 N/mm s ◦ C was utilized for
The FSW of butt joints between 3 mm thick AA7075-T6 blanks were con- the tool sheet contact surface.
sidered. The utilized alloy was characterized at room temperature by a yield The tool was modeled as rigid body and meshed, for the thermal analysis.
stress of 460 MPa and an ultimate tensile stress (UTSb) of 530 MPa. At room As far as the modeling of the workpiece is regarded, a “single block” continuum
temperature the blank material showed a microhardness equal to 160 HV and model (sheet blank without a gap) is used in order to avoid contact instabilities
grains of about 72 ␮m (D0 ). Square specimens of 100 mm edges were jointed. due to the intermittent contact at the sheet–sheet and sheet–tool interfaces. The
The utilized tool was made in H13 steel quenched at 1020 ◦ C, characterized by rotating tool moves forward and welds a crack left behind the pin as it advances
a 52 HRc hardness; a cylindrical pin was used with a pin diameter of 3.00 mm along the welding line. The sheet blank, 3 mm in thickness, was meshed with
and a pin height equal to 2.80 mm; the shoulder diameter was equal to 10 mm. about 10,000 tetrahedral elements with single edges of about 0.75 mm (Fig. 2);
During the tests a tool sinking of 2.90 mm has been utilized. Fig. 1 shows in this way about four elements were placed along the sheet thickness. A non-
the developed experimental plane: six different tests, namely FSW1–6, char- uniform mesh with adaptive re-meshing was adopted with smaller elements
acterized by different values of rotating speed (R) and advancing speed (Vf ), close to the tool and a re-meshing referring volume was identified all along the
were carried out on a milling machine. Each experiment was repeated three tool feed movement. A constant shear friction factor of 0.46 was used for the
times and macro and micrography were developed on the joints transverse tool–sheet interface [6].
sections.
The commercial FEA software DEFORM-3DTM , Lagrangian implicit code
designed for metal forming processes, has been utilized to model the FSW
3. The CDRX phenomenon modeling
process (Fig. 2). The workpiece was modelled as a rigid viscoplastic material,
and the welding tool was assumed rigid. It should be observed that the rigid- It is well known that a detailed observation of the material
viscoplastic finite element formulation is based on the variational approach; a few microstructure of a friction stir welded joint section allows to
details on the equations governing the material behavior can be found in Ref. [6]. distinguish a few different areas: starting from the parent mate-
The FSW numerical simulation was divided into two stages: the sinking stage,
in order to reach high enough temperature for the subsequent welding process,
rial, the heat affected zone (HAZ), characterized by an enlarged
and the welding (advancing) one, modeled to investigate the thermo-mechanical average grain size, is first found out. Then the so-called thermo-
phenomena in the formation of the weld nugget. As far as the mechanical prop- mechanically affected zone (TMAZ) is encountered: in this area,
erties of the material are regarded, a tabular material flow stress data at the the material has been plastically deformed by the tool, and the
varying of the temperature and of the strain rate were utilized on the basis of heat flux has also exerted some influence on the material. Finally,
literature data [6]. Furthermore, for the thermal characteristics of the consid-
ered AA7075-T6 aluminum alloy, the following values were utilized: thermal
close to the welding line, the nugget area is discovered: a recrys-
conductivity k = 180 N/s ◦ C and thermal capacity c = 2.4 N/mm2 ◦ C taken from tallized area in which the original grain and subgrain boundaries
literature; no variation of k and c with temperature was taken into account. A appear to be replaced with fine, equiaxed recrystallized grains
characterized by a nominal dimension of few ␮m [8].
In the present research two different analytical formulations
have been taken into account to predict the final recrystallized
grain dimension. The first analytical model was implemented
by the authors and used for AA6082-T6 alloys in Ref. [7]. The
model takes into account the local value of strain, strain rate and
temperature as well as a few material constants:
 
Q
DCDRX = C1 εk ε̇j D0h exp − (1)
RT
where DCDRX is the average grain size due to the continuous
dynamic recrystallization phenomena, ε the equivalent plastic
strain, ε̇ the strain rate, D0 the initial grain size, Q the material
continuous recrystallization activation energy, R the gas con-
stant, T the absolute temperature and C1 , k, j and h are the
material constants. In the technical literature a few references
Fig. 1. The investigated processes. were found out on the metallurgical phenomena occurring in
358 G. Buffa et al. / Journal of Materials Processing Technology 191 (2007) 356–359

Fig. 3. Comparison between the measured and the calculated average grain sizes at y = 1 mm for: (a) FSW5 and (b) FSW6.

aluminum alloys at high temperatures. In such papers a recurring being a = 3.63, b = −1.62, Z = ε̇ eQ/RT , and Q the gas constant
value of 140 kJ/mole referred to the aluminum alloys continuous [10]. The local values of the grain size were then calculated
recrystallization activation energy (Q) was found out. In order for each weld test on the basis of the local values of the field
to characterize the AA7075-T6 aluminum alloy, for each of the variables derived from the numerical model.
four experimental tests FSW1–4, a grain size map was traced
considering 65 different loci of measurement referred to a x–y 4. The obtained results
coordinate system [8]. The numerical simulations of the FSW
processes characterized by the investigated operative conditions Both the models were implemented as subroutine in the FEM
(FSW1–4) permitted to highlight the temperature, strain and code; in order to test their effectiveness the FSW5 and FSW6 pro-
strain rate distributions occurring during the processes; the field cesses indicated in Fig. 1 were developed both experimentally
variables values obtained by the numerical simulations in section and numerically. The obtained numerical results were compared
AA (Fig. 2), in the positions corresponding to the ones consid- with the experimental ones in terms of grain sizes: for each of
ered in the experiments, were acquired for each test. A matrix the 2 tests, 65 different measurement loci with 5 depths through
of results was then obtained made of the experimental values the specimen thickness were then considered, both in the exper-
of the grain size and of the numerical values of the considered iment and in the numerical simulation. Fig. 3a and b shows the
field variables. A constant D0 value of 72 ␮m was considered. average grain size for FSW5 and FSW6 considered at a height
An inverse approach permitted to determine the proper material of y = 1 mm from the bottom of the blanks. As it can be seen
coefficients (C1 , k, j, h) to be introduced in the model minimizing from the figure, both the models are able to correctly predict the
the error between the predicted grain size and the experimen- grain size close to the welding line, i.e. at a distance of about
tally measured one leading to the following grain size evolution 1 mm. As far as the distance from the welding line increases, the
model for the AA7075-T6 material: prediction becomes less accurate, especially for FSW5. Experi-
 
Q mental evidence demonstrates that the operative parameters that
DCDRX = 100ε−0.1648 ε̇−0.322 D0−0.104 exp − (2) characterize FSW5 (R and Vf , see again Fig. 1) result in a smaller
RT
nugget area of the welded joint. At a distance of 2 mm from the
The second analytical model takes into account the Zener- welding line, no more CDRX phenomena are observed, result-
Hollomon parameter and a few material constants as follows: ing in a poorer performance of both the analytical model. In turn,
the presence of a wider nugget area in FSW6 results in a better
1 performance of both the analytical models, even at a distance
DCDRX = (3)
a + b ln(Z) from the welding line of 2 mm.

Fig. 4. Comparison between the measured and the calculated average grain sizes at y = 2.5 mm for: (a) FSW5 and (b) FSW6.
 G. Buffa et al. / Journal of Materials Processing Technology 191 (2007) 356–359 359

Fig. 4a and b shows the average grain size for the same two lead to an even more accurate grain size prediction. What is
tests at y = 2.5 mm from the bottom of the blanks, i.e. close to more, the use of an artificial intelligence tool could be par-
the upper surface of the joint. Again, an overall good agreement ticularly effective in correlating simple field variables, easily
between the measured and the calculated results is found for both deliverable from the FEM model, and the average material grain
the analytical models. However, a slightly worse performance is size, tanking into account both static and dynamic metallurgical
obtained by the A2 model in FSW6. phenomena occurring in FSW and thus resulting in an overall
In order to quantitatively compare the obtained results of the grain size prediction in the welded joint section.
two analytical models for the investigated material, a simple
evaluation of the total quadratic error between the numerical Acknowledgments
data (num) – evaluated with the FEM code either utilizing the
A1 model or the A2 – and the experimental ones (exp) was This work has been performed with funding from MIUR
utilized as expressed in the next Eq. (4): (Italian Ministry for Instruction, University and Research) and
 supported from the Center for Excellence in Forging Technology
Err = (expij − numij )2 (4) at the Ohio State University.
i j

where j is the index indicating the measurement loci of the nugget References
area (j = 1–65) and i is the index indicating the FSW test process
[1] H.J. Liu, H. Fujii, M. Maeda, K. Nogi, Tensile properties and fracture
conditions (i equal to 5 or 6). The calculated total quadratic error
locations of friction-stir-welded joints of 2017-T351 aluminum alloy, J.
was equal to 899 for the A1 model, and 1548 for the A2 model. Mater. Process. Technol. 142 (2003) 692–696.
The better overall performance of the A1 model is due to the fact [2] I. Shigematsu, Y.J. Kwon, K. Suzuki, T. Imai, N. Saito, Joining of 5083
that it takes into account also the local value of the average effec- and 6061 aluminum alloys by friction stir welding, J. Mater. Sci. Lett. 22
tive strain, resulting both in a more accurate characterization of (2003) 343–356.
[3] W.B. Lee, Y.M. Yeon, S.B. Jung, The improvement of mechanical prop-
the considered measurement loci, and in a larger robustness of
erties of friction-stir-welded A356 Al alloy, Mater. Sci. Eng. A355 (2003)
the model to small oscillations of the process variables values 154–159.
due to the FEM simulation. [4] M. Song, R. Kovacevic, Thermal modeling of friction stir welding in a
moving coordinate system and its validation, Int. J. Mach. Tools Manuf.
5. Conclusions 43 (2003) 605–615.
[5] W.D. Lockwood, A.P. Reynolds, Simulation of the global response of a
friction stir weld using local constitutive behavior, Mater. Sci. Eng. A339
In the paper two different analytical models, namely A1 and (2003) 35–42.
A2, are implemented in a FE model in order to calculate the [6] G. Buffa, J. Hua, R. Shivpuri, L. Fratini, A continuum based FEM model
final average grain size due to the CDRX phenomena in FSW for friction stir welding—model development, Mater. Sci. Eng., A419/1–2
of AA7075-T6 aluminum alloys. A quadratic error has been (2006) 381–388.
[7] G. Buffa, L. Fratini, CDRX modeling in friction stir welding of aluminum
utilized in order to compare the performances of the two utilized
alloys, J. Mach. Tools Manuf. 45 (10) (2005) 1188–1194.
approaches. Both the models returned calculated average grain [8] K.V. Jata, S.L. Semiatin, Continuous dynamic recrystallization during fric-
size values in the nugget area close to the experimental ones. tion stir welding of high strength aluminum alloys, Scr. Mater. 43 (2000)
The A1 model showed better performances due to the presence 743–749.
of the effective strain, which results in a better characterization [9] A. Barcellona, G. Buffa, L. Fratini, Process parameters analysis in fric-
tion stir welding of AA6082-T6 sheets, in: VII ESAFORM Conference,
of the considered locus and in a larger robustness of the model
Trondhaim, 2004, pp. 371–374.
itself. [10] E. Cerri, E. Evangelista, A. Forcellese, H. McQueen, Comparative hot
It can be easily predicted that further improvements in the workability of 7012 and 7075 alloys after different pretreatments, Mater.
FEM model and in the specific grain size evolution model would Sci. Eng. A197 (1995) 181–198.