Materials Letters 106 (2013) 37–40

Contents lists available at SciVerse ScienceDirect

Materials Letters
journal homepage: www.elsevier.com/locate/matlet

Finite element simulation of accumulative roll-bonding process

Tadanobu Inoue a,n, Akira Yanagida b, Jun Yanagimoto c
a
National Institute for Materials Science, Sengen 1-2-1, Tsukuba 305-0047, Japan
b
Tokyo Denki University, Senju-Asahi-Cho 5, Adachi-ku, Tokyo 120-8551, Japan
c
The University of Tokyo, Komaba 4-6-1, Meguro-ku, Tokyo 153-8505, Japan

ar t ic l e i nf o a b s t r a c t

Article history: The accumulative roll-bonding (ARB) process, which is a severe plastic deformation process, was
Received 11 March 2013 simulated using finite element analysis, including the influence of friction, stress–strain relations, and
Accepted 25 April 2013 roll diameter. The complicated distributions of equivalent strain through the thickness of ARB-processed
Available online 3 May 2013
sheets were quantified. These quantitative strain analyses would be useful for analyzing the evolution of
Keywords: ultrafine-grained structures in the ARB process.
Finite element analysis & 2013 The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license.
Accumulative roll bonding
Strain distribution
Aluminum

1. Introduction in ARB-processed sheets using FEA, including the influence of

Bulk ultrafine-grained (UFG) materials with grain sizes of tens
to hundreds of nanometers, which show improved mechanical
properties without the addition of alloying elements, have
attracted the attention of researchers in materials science. Since
the microstructural evolution of plastically deformed materials is
directly related to the magnitude of the plastic strain, under-
standing of the phenomenon associated with the strain develop-
ment is very important. In accumulative roll-bonding (ARB), which
is a severe plastic deformation (SPD) process for realizing UFG
microstructures in metals and alloys, the microstructure and
texture in a sheet processed by one ARB cycle without a lubricant
dramatically changed depending on the thickness location of the
sheet [1–3]. The embedded-pin method is often employed to
measure the strain through thickness experimentally [4,5], but
magnitude of the strain obtained by this method does not exhibit
an exact value [6]. To control the microstructures, it is essential to
understand the deformation behavior in the ARB-processed sheets
accurately and quantitatively. Although some studies have used
finite element analysis (FEA) for other SPD processes, such as
equal-channel angular pressing (ECAP) [7], high-pressure torsion
[8], and warm caliber rolling [9,10], there have been no reports for
the ARB process. This study aims to quantify the equivalent strain
n
Corresponding author. Tel.: +81 29 859 2238; fax: +81 29 859 2201.
E-mail address: INOUE.Tadanobu@nims.go.jp (T. Inoue).
friction, stress–strain relations and roll diameter.
2. Modeling of ARB process
The two sheets were set in order to analyze three ARB cycles, as
shown in Fig. 1a. The sheet is composed of two sheets with
dimensions of 1 mm t
12 mm L. The finite element mesh in each
sheet with dimensions of 2 mm t
12 mm L shown in Fig. 1b
included 4141 nodes and 4000 elements. The mesh size in the
thickness direction, tel, is constant, tel ¼0.05 mm. The mesh size in
the longitudinal direction, Lel, gradually decreases toward the
center from front and back, and the minimum Lel is 0.025 mm at
mid-length (center element). In the present study, the condition of
a commercial 1100 Al sheet rolled at ambient temperature, as
reported by Lee et al. [1], was referred to as the rolling condition:
initial thickness, t0 ¼ 2 mm; nominal reduction per pass, r ¼50%;
roll diameter, dϕ ¼255 mm; and rolling speed, 170 mm s-1.
In addition, the case of a small roll diameter, dϕ ¼ 118 mm, was
also examined because the strain depends strongly on the roll
diameter as well as the friction [11]. Young's modulus of 70 GPa
and Poisson's ratio of 0.35 were used as the elastic modulus. The
stress s–strain ε relationships of 1100 Al at 301 K employed in the
analysis were described by s¼ 28+105.67ε0.32ε_ 0.017 MPa, but the
flow stresses were assumed to remain constant at ε¼4.0 because
hardness does not vary at equivalent strain of over 4.0 on the basis
of the ECAP studies [12]. In order to investigate the effect of the
s−ε relations, the simulation for the 1100 Al at 473 K was also
conducted using s ¼ s0 þ Kεn ε_ m , where s0 ¼20 MPa, K ¼58.40
MPa, n ¼0.24 and m ¼0.0405 with Young's modulus of 60 GPa

0167-577X & 2013 The Authors. Published by Elsevier B.V. Open access under CC BY-NC-ND license.
http://dx.doi.org/10.1016/j.matlet.2013.04.093
38 T. Inoue et al. / Materials Letters 106 (2013) 37–40

Fig. 1. Schematic illustration of three ARB cycles and finite element mesh used.

Fig. 2. Deformation histories of the center part of the FE mesh in the ARB process against roll diameter dϕ and friction coefficient μ.

and Poisson's ratio of 0.35. The Coulomb condition was used as the imposed by rolling is defined as follows:
frictional condition between the rolls and the sheet. Assuming the Z tðsteadyÞ
dεeq
Coulomb law, the minimum friction coefficient, μ, is 0.13 for εeq ¼ dt ð1Þ
dt
dϕ ¼118 mm and 0.09 for dϕ ¼255 mm. Here, as the lubricated 0

conditions, μ¼ 0.14 for dϕ ¼118 mm and μ¼0.1 for dϕ ¼255 mm
were adopted; furthermore, as the unlubricated condition, μ¼ 0.25
was adopted for both diameters.
In the analysis, the classical metal plasticity models with a von
Mises yield surface, (n)PLASTIC, HARDENING¼ISOTROPIC as key-
word in ABAQUS, were employed. The equivalent strain, εeq,
where dεeq/dt denotes the incremental equivalent strain, and
t(steady) is the rolling time [11].
In the ARB simulation shown in Fig. 1a, first, a 2 mm-thick sheet
was rolled to a thickness of 1 mm by 50% (1st ARB cycle). Subse-
quently, the two 1 mm-thick rolled sheets were stacked to be 2 mm
in thickness and rolled again to a thickness of 1 mm (2nd ARB cycle).
 T. Inoue et al. / Materials Letters 106 (2013) 37–40
Fig. 3. Distribution of the equivalent strain along the sheet thickness against roll diameter dϕ and friction coefficient μ.
In order to repeat this procedure, a symmetry condition in the y-axis
was set before the next rolling, and the sheet was rolled (3rd ARB
cycle). Hence, for the 1st and 2nd ARB cycles, a 1/1 model was
adopted, and, for the 3rd ARB cycle a 1/2 model was adopted. As the
bonding conditions during the simulation, (n)EQUATION keyword in
ABAQUS was employed in order to define linear multi-point con-
strains for the bonded surfaces before the 2nd ARB cycle; further-
more, no slip condition of μ¼1.0 was given on these surfaces.
3. Results and discussion
Fig. 2 shows deformed meshes of a center part in each sheet of the
FE mesh during the ARB cycles. Under μ¼0.25, the flection is clearly
larger in a sheet rolled at dϕ ¼255 mm than in one rolled at
dϕ ¼ 118 mm (Fig. 2a and b). This tendency is larger with increasing
the number of ARB cycles, N. A similar deformation behavior can be
seen in a difference of μ, and the flection under dϕ ¼ 255 mm becomes
larger in a sheet rolled at μ¼ 0.25 than in one rolled at μ¼0.1 (Fig. 2b
and c). The result means that the evolution of the microstructure and
hardness through thickness in the sheet processed by ARB is different
depending on the dϕ even if the conditions of reduction and friction
are the same.
Fig. 3 represents the distributions of the εeq along the sheet
thickness. Under the lubricated conditions of a small μ, although the
εeq has a slight distribution through the sheet thickness, it is almost
constant, and its magnitude corresponds to a value of 0.8
N
regardless of the dϕ. On the other hand, under the unlubricated
condition, the εeq has a distribution whose magnitude becomes larger
than 0.8. In dϕ ¼ 118 mm, the εeq at the surface reaches 1.2 after the 1st
cycle. The εeq after the 2nd cycle shows the maximum value of 2.4 at
the surface, and it has a peak of 2.0¼1.2 (for the 1st cycle)+0.8 (for the
2nd cycle) at the center. The maximum εeq at the surface represents
the value of 3.8 after the 3rd cycle, and the three peaks exist within
the 1 mm thickness. The positions with indicated peaks at y¼0.25 and
0.75 correspond to the surface in the 1st cycle. The peak position at
the center (y¼0.5) corresponds to the surface in the 2nd cycle; hence,
the εeq at the center is represented by 3.2 (¼2.4+0.8). In a larger roll,
dϕ ¼ 255 mm, the εeq at the surface is 2.2 after the 1st cycle. After the
2nd cycle, the εeq shows a distribution with the maximum value of
4.6 at the surface, and it has a peak of 3.0 (¼2.2+0.8) at the center. The
maximum εeq at the surface showed a value of 7.0 after the 3rd cycle.
Fig. 4 shows variations of εeq at the surface against the N, at
ambient temperature. For comparison, the results in the Al rolled at
473 K for dϕ ¼118 mm and μ¼ 0.25 are shown in the figure. The
result of 0.8
N is also shown as a broken line in the figure.
Although, under dϕ ¼255 mm and μ¼0.1, the εeq at the surface is
39
Fig. 4. Variations of equivalent strain at the surface against the number of ARB cycles at
ambient temperature. Here, dϕ and μ denote the roll diameter and friction coefficient
used, respectively. The open symbol indicates the results at 473 K. The “
” is the results
of one-pass rolling shown in Ref. [11].
slightly larger than 0.8
N, these values are almost the same. Under
μ¼0.25, the εeq increases approximately linearly with respect to N,
and the gradients are larger in dϕ ¼255 mm than in dϕ ¼118 mm.
This proportionality means that the strain at the surface in the multi-
cycle ARB-processed sheet can be estimated by the conventional
one-pass rolling analysis. Furthermore, in Fig. 4, the results at 301 K
(solid symbol) are almost the same as those at 473 K (open symbol),
under dϕ ¼118 mm and μ¼0.25. Hence, the effect of the s–ε relations
of the rolled sheets on the strain is very small in comparison with the
effect of friction and dϕ. A similar feature is seen in the IF-steel sheets
rolled with the embedded-pin method [13]. The symbol “
” in the
figure denotes the results of one-pass rolling analysis against (dϕ, μ)
shown in Fig. 6c of [11]. Using these accurate εeq data, we are able to
know the magnitude of the εeq in the multi-cycle ARB-processed
sheet against various (dϕ, μ), such as the thin lines shown in Fig. 4.
As a result, the strain at the surface, quarter (1/4t), and center
(1/2t) in the thickness of the multi-cycle ARB-processed sheet is
estimated by the following equations:
εeq;surðNÞ ¼ εeq;surðFEAÞ N for surface
εeq:;1=4tðNÞ ¼ εeq:;1=4tðFEAÞ þ εeq;1=2tðN−1Þ for 1=4t
pffiffiffi  
εeq:;1=2tðNÞ ¼ 2= 3 ln 1=ð1−rÞ þ εeq;surðFEAÞ ðN−1Þ for 1=2t
where εeq;surðFEAÞ and εeq;1=4tðFEMÞ denote the equivalent strain at the
surface and 1/4t, respectively, obtained using FEA for one-pass
rolling, and these magnitudes are necessary to accurately analyze
40 T. Inoue et al. / Materials Letters 106 (2013) 37–40

including not only the friction and roll bite geometry but also the References
mesh division.
[1] Lee SH, Saito Y, Tsuji N, Utsunomiya H, Sakai T. Scr Mater 2002;46:281.
[2] Li X, Samman TA, Gottstein G. Mater Lett 2011;65:1907–10.
4. Conclusions [3] Kamikawa N, Sakai T, Tsuji N. Acta Mater 2007;55:5873.
[4] Cui Q, Ohori K. Mater Sci Technol 2000;16:1095.
[5] Hashimoto S, Tsukatani I, Kashima T, Miyoshi T. Kobe Steel Eng Rep
The exact magnitude and distribution of the equivalent strain in 1998;48:14.
the Al sheet processed during three ARB cycles using FEA were [6] Inoue T, Tsuji N. Comput Mater Sci 2009;46:261.
shown in the present study. These quantitative strain analyses would [7] Nagasakhar AV, Hon YT. Comput Mater Sci 2004;30:489.
[8] Yoon SC, Horita Z, Kim HS. J Mater Process Technol 2008;201:32.
provide useful guidelines for understanding the quantitative correla- [9] Inoue T, Yin F, Kimura Y. Mater Sci Eng A 2007;466:114.
tion between the microstructures and strain in the ARB process. [10] Inoue T, Somekawa H, Mukai T. Adv Eng Mater 2009;11:654.
[11] Inoue T. In: Moratal David, editor. Finite element analysis. Croatia: SCIYO;
2010 pp. 589–610.
Acknowledgment [12] Horita Z, Kishikawa K, Kimura K, Tatsumi K, Langdon TG. Mater Sci Forum
2007;558–559:1273.
[13] Um KK, Jeong HT, An JK, Lee DN, Kim G, Kwon O. ISIJ Int 2000;40:58.
This study was financially supported in part by the Grant-in-Aid for
Scientific Research on Innovative Area, “Bulk Nanostructured Metals”,
through MEXT, Japan (No. 2210205), whose support is gratefully
appreciated.