Journal of Materials Processing Technology 211 (2011) 1948–1956

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

A numerical investigation of the continuous bending under tension test

A. Hadoush a,1 , A.H. van den Boogaard b,∗ , W.C. Emmens c
a
Materials innovation institute, P.O. Box 5008, 2600 GA Delft, The Netherlands
b
University of Twente, P.O. Box 217, 7500 AE Enschede, The Netherlands
c
TATA Steel RD&T, P.O. Box 10000, 1970 CA IJmuiden, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: In this paper the continuous bending under tension test is analyzed by numerical simulation. The ability
Received 8 May 2010 of achieving high strains by combined stretching and bending is considered. This deformation mode has
Received in revised form 14 June 2011 similarities with the deformation that takes place in incremental sheet forming (ISF) and may—at least
Accepted 18 June 2011
partly—explain the high strains that are observed there. The sensitivity of the numerical model to mesh
Available online 25 June 2011
discretization is studied as well as the influence of different material models. An isotropic hardening
material model and two mixed isotropic/kinematic hardening material models are used. The results for
Keywords:
the three models are very similar, for the shape of the load curves, but not for the point of necking. A
Incremental sheet forming
Bending under tension
numerical analysis of the cyclic force–displacement curve of the CBT test is presented. This analysis is
Cyclic bending focused on the pattern of the cycle and the evolution of the cycle during the test. The loss of stability
Force–displacement for inhomogeneous stress distributions is analyzed and the importance of bending in stabilizing the
Instability deformation under tension is demonstrated. Stability is lost if the complete cross section is in a state of
FEM tensile stress.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction Incremental sheet forming is a displacement controlled process

The continuous bending under tension (CBT) test can be seen as
a tensile test on strip material, with additional local bending by a set
of rolls that is travelling over the length of the strip. The main effect
of additional bending is that the required tensile force for the same
amount of elongation is reduced (Marciniak and Duncan, 1992).
Hadoush et al. (2007) presented a simplified 2-dimensional finite
element model for the CBT test to study the contribution of bending
in stabilizing the deformation of a strip to high strain. The CBT test
was proposed by Benedyk et al. (1971) to investigate material prop-
erties at high levels of straining. Emmens and van den Boogaard
(2009a) identified the CBT test as an incremental forming process
and showed experimentally that high levels of strain are obtained
for various materials.
In the CBT test, sheet material deforms incrementally rather
than continuously as in a standard tensile test. The deformation
around rolls in the CBT test bears resemblance with the defor-
mation around the spherical tool in incremental sheet forming
(ISF) and this resemblance motivated the research described in this
paper.
∗ Corresponding author.
E-mail address: a.h.vandenboogaard@utwente.nl (A.H. van den Boogaard).
1 a previously claimed stability criterion for bending under tension
Present address: Faculty of Engineering, The Hashemite University, P.O. Box
330127, Zarqa 13115, Jordan.
performed on a CNC machine. A clamped blank is deformed by the
movement of the tool that follows a prescribed tool path as intro-
duced by Matsubara (1994). An extensive overview of the process
is given by Jeswiet et al. (2005). In ISF, strains are observed that are
often far above the forming limit curve for the material under con-
sideration. In classical sheet forming operations the deformation
would become unstable, leading to the inception of necking, but in
ISF the deformation appears to be stabilized. Several mechanisms
have been proposed in literature to explain the increased forma-
bility: shear, contact stress, bending, and cyclic straining. These
mechanisms are discussed in detail in a recent review paper by
Emmens and van den Boogaard (2009b). To separate the effect of
bending under tension from other stabilizing effects in ISF, the CBT
test is investigated.
In this paper, a 3D finite element model of the CBT test is pre-
sented. The model is used to obtain process knowledge and to
validate assumptions about the observed extended formability. The
relation between the CBT test and ISF operations is not discussed,
but will be the subject of a separate paper (Emmens and van den
Boogaard, 2011).
The basic components of the numerical model to predict the
force–displacement curve of the CBT test are presented in Section
2. Then, explanations on the shape of the force–displacement curve
are presented in Section 3. These explanations are based on pro-
cess mechanics and process characteristics of the CBT test. Finally,
is investigated by numerical analysis.

0924-0136/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2011.06.013
 A. Hadoush et al. / Journal of Materials Processing Technology 211 (2011) 1948–1956
Fig. 1. CBT setup (Emmens and van den Boogaard, 2009a).
2. Process description and numerical model
In this section, the CBT experiment will be introduced briefly and
a FEM model will be developed and validated. In the next section
results from the model will be compared with experimental results.
The experimental setup and results of the considered CBT test have
been presented in more detail in Emmens and van den Boogaard
(2009a).
2.1. The CBT experiment and FEM model
A photograph of the CBT setup is shown in Fig. 1. In the right-
hand part of the picture a vertical tensile strip is visible that is
clamped in a tensile machine. A set of 3 rolls bends the strip, of
which 1 roll is visible in the picture. The rolls are freely rotat-
ing in bearings and the complete roll set is moving up and down
repeatedly while the strip is stretched.
In the FEM model, the tensile strip is placed horizontally and
the roll set is moving repeatedly from left to right, as presented in
the 2-dimensional schematic in Fig. 2. The roll set is modeled by 3
analytical, frictionless, cylinders of 15 mm diameter. In longitudinal
direction, the rolls are separated from each other by 17.5 mm. The
roll set can travel in the longitudinal direction only. First, the central
roll is placed such as to fit the specimen in between the rolls with-
out deforming it. Then the bending of the specimen is introduced
by the movement of the central roll downwards over 2.3 mm plus
the initial sheet thickness. The cyclic movement of the roll set in
longitudinal direction introduces bending at continuously chang-
Fig. 2. Continuous bending under tension test description.
1949
Fig. 3. Schematic of the specimen (dimensions in mm, the drawing is not to scale)
(Emmens and van den Boogaard, 2009a).
ing positions. The roll set moves with a stroke of 100 mm and a
velocity of 66 mm/s, while the cross bar moves with 2.5 mm/s.
The specimen used in the CBT experiment is schematically
shown in Figure 3. It has a uniform thickness of 1 mm and piecewise
uniform width. The middle part of the specimen has the smallest
width (20 mm). The cyclic bending is performed only in the middle
part of the specimen. Experimentally, it is observed that the part
that experiences combined tension and bending deforms plastically
as shown in Fig. 4. Because of the geometry and applied loading con-
ditions, plastic deformation of the wider parts can be neglected for
mild steel. Because of symmetry along the longitudinal axis, only
half of the middle part of the specimen is modeled. The modeled
part of the specimen is 200 mm in length and 10 mm in width. It
can be observed in Fig. 4 that there is still lateral contraction and
therefore a 2D plane strain state cannot be assumed in the CBT
simulation.
In the next two sections the influence of the mesh discretization
and the material model on the predicted force–displacement curve
is studied. First, three different FE mesh sizes are used to investigate
the mesh dependency, then a study on the effect of the applied
hardening model is presented.
2.2. Mesh size dependency
A regular mesh is used with 8 triangular shell elements used
to discretize the 10 mm half-width of the specimen. The shell ele-
ment is a combination of a discrete Kirchhoff triangular element for
bending and a constant strain membrane element as proposed by
Batoz et al. (1980). A 3-point integration scheme in the plane, com-
bined with 7 integration points in thickness is applied, resulting
in 21 integration points per element. The triangles are large at the
longitudinal symmetry line and small at the free edge of the strip
with an element size ratio of 4 to 1. This distribution represents the
non-uniform stress distribution along the specimen width that is a
result of the stress-free edge. Because the stress distribution near
the symmetry line is quite uniform in width-direction a high aspect
ratio (width/length) is not detrimental. Force–displacement curves
were virtually equal for uniformly and nonuniformly distributed
meshes.
In longitudinal direction a uniform element length is used that
should be small enough to represent the bending by the rolls, even
1950 A. Hadoush et al. / Journal of Materials Processing Technology 211 (2011) 1948–1956

Fig. 4. Unused and used specimens: untested (top), tensile tested (middle) and CBT tested (bottom). A uniform deformation is observed in the white rectangle.

Fig. 5. Samples of mesh densities: coarse (left), intermediate (middle) and fine (right). The samples represent 2 mm strip length and 10 mm half-width, with the free edge
on the bottom and the symmetry line on the top of each sample.

after considerable elongation. Three longitudinal element lengths 2.3. Material model
are used and they are classified as: coarse (1 mm), intermediate
(0.5 mm) and fine (0.25 mm). The modeled strip does not require In this work, material models are used with parameters that
a geometrical imperfection—as often used in the simulation of a are representative for mild steel DC06, taken from Van Riel (2009).
tensile test—because a nonuniform deformation is induced anyhow The anisotropic plastic behavior is modeled with the quadratic
by the local bending action. Samples of the different mesh densities Hill’48 yield function (Hill, 1948) with parameters based on the
are shown in Fig. 5. R-values. Hardening is modeled with a combination of isotropic
The predicted longitudinal force at the clamped edge versus hardening governed by a power law relation and kinematic hard-
the cross bar displacement is shown in Fig. 6 for the three dif- ening represented by the Armstrong–Frederick model as presented
ferent meshes. The different meshes predict the same pattern of by Chaboche (1991). Three parameter sets are used for the hard-
the force–displacement curve and an almost equal force level. The ening model to see the effect on the simulation results. The first
difference in the predicted forces is a result of the spatial dis- model (isotropic) only uses the isotropic part of the model. Because
cretization. As expected, a higher level of oscillation is observed bending and unbending of the strip results in cyclic tensile and com-
in the coarse mesh. The fine mesh simulation finishes 15 complete pressive stresses—at least in the outer fibres—it can be assumed
cycles and fails during cycle 16. More stable cycles are predicted that the Bauschinger effect plays a role in the analysis. To inves-
with the coarse and the intermediate meshes. A larger element is tigate the influence of the Bauschinger effect on the predicted
expected to smooth the achieved strain and that results in delaying force–displacement curve two different sets of parameters were
the localization of deformation. In the next section, the intermedi- used (iso/kin 1 and iso/kin 2). Both sets were fitted to the same
ate mesh is used for time efficiency. For the detailed discussion of cyclic shear tests, but the second set used a higher weighing fac-
the force–displacement curve in Section 3 the fine mesh is applied. tor for the transient zone after load reversal. The implementation

5000 4000
Predicted longitudinal force (N)
Predicted longitudinal force (N)

3800
4000
Coarse
3600
3000
Coarse
3400
2000
Fine
3200

1000
3000 Fine
Intermediate Intermediate

0 2800
0 50 100 150 200 90 91 92 93 94 95
Cross bar displacement (mm) Cross bar displacement (mm)

Fig. 6. The predicted force–displacement diagrams for the entire CBT test (left) and a zoomed in part of the process (right).
 A. Hadoush et al. / Journal of Materials Processing Technology 211 (2011) 1948–1956
600
400
200
Isotropic
Iso / kin 1
x
0
σ
Iso / kin 2
−200
−400
−600
0 0.2 0.4 0.6
ε
x the deceleration and acceleration of the roll set in the simulation
Fig. 7. Stress–strain curves for the components in x-direction.
of the models follows the procedure as given in Zienkiewicz and
Taylor (2005).
The cyclic stress–strain response for the 3 parameter sets is pre-
sented in Fig. 7. Due to the different fitting procedures, a small
difference in the stress–strain curve is observed even for the first
part of the test that represents monotonic loading. After a load
reversal, the first isotropic/kinematic hardening model shows a
non-sharp elastic/plastic transition compared to the sharp elas-
tic/plastic transition that is observed in the isotropic material
model. The second isotropic/kinematic model clearly shows early
re-yielding and an increased hardening rate after load reversal.
Because the hardening rate is an important parameter in the sta-
bility of deformation, it could be expected that the last parameter
set shows extended stability.
The predicted force at the clamped edge for each parameter set
is plotted versus the cross bar displacement in Fig. 8, together with
experimental results. For this analysis, the model with intermediate
mesh size was used because of time efficiency. In general, the dif-
ferent parameter sets predict the experimental force–displacement
curve successfully. Due to a finite time, needed for reversing the
movement of the roll set in the experiment and the instantaneous
reversal in the simulation, the experimental and simulation results
are shifted slightly. In Fig. 8 the start and end of a cycle are indicated
with ‘S’ and ‘E’ respectively. The Bauschinger effect that is included
in the isotropic/kinematic material models shows no significant
difference in the pattern of the predicted force compared to the
predicted force by the isotropic material model, but a small effect
on the force level can be seen. The combined isotropic/kinematic
material models shows earlier localization than the isotropic mate-
rial model. This contradicts with the expectation that kinematic
hardening models can stabilize the process by the increased hard-
ening rate after stress reversal.
It is concluded that a standard shell element description and
an isotropic hardening material model is sufficient to predict
the experimental force–displacement curve of the CBT test for
many cycles with good agreement for stable cycles. The numerical
model does not predict the formability limit of CBT test correctly
because of material and mesh dependency. Most surprisingly, mod-
eling the Bauschinger effect with kinematic hardening has no
significant influence on the predicted force–displacement relation,
although it does influence the numerical stability. The relatively
low influence of the material model shows that the pattern of the
force–displacement curve is mainly determined by the mechan-
ics of the process rather than the material behavior. The process
mechanics will be discussed in the following section.
1951
3. Force–displacement curve
The purpose of this section is to analyze the shape of the
force–displacement curve and validate an hypothesis about the
force peaks that was presented in the experimental work of
Emmens and van den Boogaard (2009a). The analyses are per-
formed with the fine mesh as presented in Section 2.2.
The force–displacement curve includes a particular number of
cycles depending on the experimental settings and the tested mate-
rial. Three representative cycles of the force–displacement curve
are plotted in Fig. 9. A cycle consists mainly of two peaks and two
relatively steady parts. The peaks are observed after the moment
that the roll set changes its traveling direction. The steady parts
represent the main part of the cycle. As explained earlier, ignoring
results in predicting a smaller time interval of a cycle in the sim-
ulation compared to the experimental time interval. This explains
the increasing lag of the measured force profile compared to the
predicted force profile.
3.1. Semi-steady part
In the CBT test, only material regions that are currently being
bended deform plastically. The force–displacement curve of a com-
plete test was presented in Fig. 8. Because of the bending action, a
lower tensile force is required to stretch the strip compared to the
force required to stretch the same cross-section under tension only.
Although the tensile force is decreased by bending action, material
hardening will increase the tensile force after some deformation,
just like in an ordinary tensile test. Because the cross-section area
reduces upon stretching, the growth of the tensile force reduces
and finally the tensile force will decrease because the cross-section
area reduces faster than the material hardening increases the force.
This is, again, just like in an ordinary tensile test, but at a lower
level because of the additional bending action. At high elongation
the material gets thinner and the relative proportion of bending
compared to stretching reduces. Then, also the reduction of tensile
force by bending becomes smaller.
Within one cycle it can be observed that every first half of a cycle
shows a higher average force than the second half of the same cycle.
This can be explained by the additional positive/negative contribu-
tion of the left/right (down/up) movement of the roll set. Because
material is pulled out of the deformation zone from one side, the
part between the peaks is not truly in a steady state. Due to the
dimension of the roll set, the strip is bent and unbent up to dif-
ferent levels. Three zones are identified based on the frequency
of bending/unbending as shown in Fig. 10. In zone 1, the mate-
rial is bent and unbent once for each pass of the roll set. Similarly,
the material experiences 2 sets of bending/unbending and 3 sets
of bending/unbending deformation in zone 2 and zone 3, respec-
tively. The cyclic forward/backward movement of the roll set shifts
the active location of the plastic deformation through these zones
assuming that only the bent material deforms plastically.
The numerically achieved results of the longitudinal force for
the different cyclic patterns are plotted versus the cumulative dis-
placement of the roll set in Fig. 11. At 0 mm displacement, the roll
set is at the right near the fixed clamp and the force transducer and
at 100 mm the roll set is at the left near the moving cross bar. A com-
plete cycle consists of moving from right (0 mm) to left (100 mm)
and back to right again.
The material migrates from one zone to another in the direc-
tion of the cross bar. The calculated displacements of several nodes
through the length of the strip is shown in Fig. 12. The cross bar pulls
the material to the left, while the three rolls are moving between
the positions indicated by [1–3] in the initial position near the fixed
clamp, where the tensile force is measured, and initial position
1952 A. Hadoush et al. / Journal of Materials Processing Technology 211 (2011) 1948–1956

Fig. 8. Force–displacement curves for different material models: the entire test (left) and one cycle (right).

Fig. 9. Evolution of the cyclic force–displacement curve during the CBT test: the third cycle (top left), the sixth cycle (top right) and the twelfth cycle (bottom).

Fig. 10. Different deformation history within the cyclic roll set movement. The number of the zones define the number of bending/unbending operations. The zones are
defined from one roll center to another roll center.
 A. Hadoush et al. / Journal of Materials Processing Technology 211 (2011) 1948–1956
4500
Clamped longitudinal force (N) First half
of the cycle
4000
Cycle 6
Cycle 3
3500
Cycle 12
3000
2500
0 35 65 100 65
Roll set position (mm)
Fig. 11. Cyclic force evolution versus the absolute displacement of the roll set.
[7–9] near the moving cross bar. Node 1 hardly moves (less than
1 mm) for 12 cycles. This means that material in zone 1 near the
fixed edge has the lowest cyclic history and consequently is less
hardened. Material migrates from zone 3 and fills gradually zone 2
and subsequently zone 1 on the left-hand side. At the sixth cycle,
the left zones 1 and 2 are completely filled with material originally
deformed in zone 3. The material is almost stretched up to the max-
imum force, where the force is relatively insensitive to the strain
(the strain hardening and cross-section reduction cancel out). This
explains the almost constant level of force. The sixth cycle finishes
with a decreasing force because the roll set goes back to zone 2 and
zone 1 near the clamped edge.
Performing more cycles, material keeps migrating to the left. The
material (between node 3 and node 5) has experienced the highest
number of cycles and consequently has the highest hardening and
the most significant reduction of thickness along the strip. In this
range the average force is reduced compared to the top-value in
cycle 6. When the roll set returns to the position on the right where
the material is less stretched, in zone 2 and zone 1, the force is still
higher, resulting in the concave-like pattern of the twelfth cycle.
3.2. Peak
In Emmens and van den Boogaard (2009a) it was proposed that
the peaks in the force may be explained by the deceleration and
acceleration of the roll set to change the direction of movement.
If the amount of bending is temporarily reduced because of the
lower roll set velocity at the same stretching velocity, the force
must increase. However, the numerical model does not include the
deceleration and acceleration of the roll set but changes the veloc-
ity direction of the roll set instantaneously, and the peaks in the
simulations are still very similar to the peaks in the experiments.
Fig. 12. Migration of material from one zone to another toward the moving edge: initial location (top) then at the beginning of the third cycle, the sixth cycle and the twelfth
cycle (bottom).
1953
−4
x 10
Second half 6
of the cycle
5
Incremental plastic strain
4
Steady
3 Peak
2
1
0
35 0
−50 0 50 100 150 200
Current length geometry (mm)
Fig. 13. A comparison of the plastic increment for two positions of the roll set, one
during the steady part (dotted lines) and one just after the reversal of the travel-
ing direction (dashed line). The cross bar displacement rate is the same for both
measurements. Because of the deformation, the current length of the strip at the
beginning of the fifth cycle is 230 mm.
Therefore, it can be concluded that the explanation of the peaks by
deceleration and acceleration is insufficient.
During the steady part of the cycle, the active plastic zone is
shifted with the traveling roll set. During the peak, an increase of
the longitudinal force is observed. This indicates that the reduc-
tion of the tensile force by bending is momentarily less active. As a
result of the increased tensile force, the strip plastically deforms in
the region of the roll set under bending and tension, and partially
under tension further away from the roll set. The resulting plastic
deformation can be seen in Fig. 13. The plastic strain per increment
is plotted for two positions of the roll set, one during the steady
state (dotted lines) and one just after the traveling direction rever-
sal of the roll set (dashed line). It can be clearly seen that during the
steady state high plastic strain increments appear, but only near the
position of the rolls. Just after the reversal of the traveling direction,
the peak plastic strain increments are much lower but they extend
over a larger region.
The longitudinal force cannot increase suddenly as long as a sig-
nificant change of the curvature dominates the process as explained
in the previous section. Therefore, the following study is carried
out to investigate the curvature change during the peak. The local
geometry of the strip in the FEM model just before reversing the
roll set traveling direction is plotted in Fig. 14. This is the current
geometry at the end of the fourth cycle, it is used to investigate the
peak at the beginning of the fifth cycle. The current traveling direc-
tion of the roll set is toward the clamped edge (from left to right)
and it is going to reverse. A mirror image of the current geometry
shows that the roll set can travel easier to the left than to the right.
1954 A. Hadoush et al. / Journal of Materials Processing Technology 211 (2011) 1948–1956
Fig. 14. The local geometry of the strip in the FEM model (solid line) just before
reversing the travelling direction of the roll set. The dotted line is a mirror image of
the current geometry of the strip.
Furthermore, reversing the traveling direction of the roll set (mov-
ing to left) will not result in a sudden and significant change in the
curvature of the strip. This geometry will last for an interval of time.
The increase of the longitudinal force when the roll set reverses
its traveling direction is caused by the continuous movement of the
cross bar. A comparison of the predicted longitudinal force during
the reversal of the roll set direction with and without displacement
of the cross bar is shown in Fig. 15. Without displacement of the
cross bar, the roll set travels for 5 mm at almost constant force (a
slight drop of the force is observed at the very beginning) with
no significant change in geometry. Then, further movement of the
roll set corrects the geometry and the bending/unbending starts
producing significant plastic deformation (the strip is elongated)
which reduces the tension in the strip. This supports the claim that
the peak results from the deformation of the strip by the continuous
cross bar movement with no significant bending.
4. Stability analysis
It has already become clear in the above that in the CBT test
levels of uniform elongation can be obtained that are much higher
than in a conventional tensile test. Apparently the instabilities that
limit a conventional tensile test do not operate in the CBT test. The
question is now where this increased stability is derived from.
A well-known stability criterion is the so-called maximum force
condition, that states that a tension process is stable if and only
if the tensile force increases with elongation. A quick look at the
force–displacement curves presented in Fig. 8 shows that this does
not apply to the CBT test: frequently the force decreases with
increasing elongation without causing instability. In the CBT test
only material being bent does deform, and the zones of deforma-
4000
Clamped longitudinal force (N)
Reversing with tension
3500
3000 Reversing without tension
2500
2000
1500
0 5 10
Roll set travelling distance (mm)
Fig. 15. Longitudinal force evolution for the first part of the fifth cycle with and
without cross bar displacement.
e·R
t
Fig. 16. The strain state in a strip that is bent and stretched at the same time.
tion continuously travel through the specimen as the roll set moves.
This means that when comparing the force at two different elon-
gations, one in fact compares forces originating at two different
moments, and consequently the forces obtained from two differ-
ent parts of the specimen. It should be clear that this is not correct
if one wants to assess stability.
The maximum force condition is more correctly defined as: a
tension process is stable if and only if the tensile force increases
with local strain increment. If by whatever reason two neighbour-
ing cross-sections have a different strain but are subjected to the
same tensile force, the process is stable if the cross-section with the
larger strain also requires a larger force to elongate, or:
dF
>0 (1)
dε
and it is of no relevance what mechanisms makes the force change.
In a previous paper (Emmens and van den Boogaard, 2009a) the
CBT test has been linked to the situation known as bending under
tension, where it is implied that part of the cross-section is actually
in compression. This will reduce the effective pulling force without
having to alter the flow stress. Obviously, the pulling force strongly
depends on how much of the cross-section is in compression; if
one half is in compression the force will be zero (in first approxi-
mation). If there is a way to influence the part of the cross-section in
compression, then also the net pulling force will be influenced and
this can create a mechanism to raise the pulling force with strain. In
other words, the tensile force depends on the position of the neu-
tral line, and stability depends on how the neutral line shifts with
increasing strain. The situation is shown schematically in Fig. 16,
this figure presents the strain state in a strip of thickness t that is
bent and stretched at the same time. The net elongation is e, the
elongation of the centre fiber and the bending radius of the cen-
tre line is R, leading to a bending strain εb = y/R with the origin of
the y-coordinate at the centre line. The elongation then shifts the
4000
Tensile force in fixed clamp (N)
3000
2000
1000
Cycle 13 Cycle 14 Cycle 15 Cycle 16 Cycle 17
0
15 90 95 100 105 110 115 120 125
Cross bar displacement (mm)
Fig. 17. Force–displacement curve for the last five cycles of the CBT test using the
fine mesh.
 A. Hadoush et al. / Journal of Materials Processing Technology 211 (2011) 1948–1956
1000
Stress (MPa)
500
0 middle
−500
−1000 Cycle 13
90 95
Fig. 18. The evolution of  x through thickness for the failed element at the symmetry line.
neutral line over a distance e · R. The lower part of the strip (at the
concave side) is in compression, but that is only the case when the
bending radius is not too large, more specific when: e · R < t/2. It is
not too difficult to visualize from Fig. 16 that when the net elon-
gation e increases, there will be less material in compression and
consequently the pulling force will be higher, but only as long as the
radius R does not change too much at the same time. Of course this
only works as long as the neutral line is still within the material,
or: as long as a part of the cross-section is still in compression.
We can now hypothesize that stability in the CBT test is related
to the occurrence of a part of the cross-section being in compres-
sion, this has already been stated by Hadoush et al. (2007). This will
now be verified in more detail based on the numerical simulation
with the fine mesh and isotropic hardening model (see Section 2). It
should be realized that the numerical analysis is used here to inves-
tigate the underlying mechanisms of stability in the CBT test, not
to predict the actual maximum achievable strain because a more
accurate material model would be needed for that purpose.
Fig. 17 presents the simulated pulling force in the final part of
a test for five consecutive cycles. Specimen failure occurs in the
last cycle, no. 17, but the used material model cannot simulate
actual failure. During these five cycles the stress in three integra-
tion points, one in the centre (middle) and two near the surfaces
(upper and lower), has been tracked, and the results are presented
in Fig. 18. In the graphs, the repeated passing of the three rolls can
clearly be recognized. The middle point stress is always in tension,
the upper stress and the lower stress vary between compression
and tension during the pass of the roll set. Each cycle starts with a
force peak as described above, the stress situation during that peak
can be observed in the very first part of the individual cycles in
Fig. 18. In cycles 13 and 14 there is still compression of the upper
point. This is an important observation because it shows that the
force peak itself does not create an instability. In cycle 15 com-
pression is just lost; the stress in the upper point is considerable
lower than in the middle and lower point, but it is just in tension.
In cycle 16 all three points are clearly in tension. The same develop-
ment can be observed in other places as well. For example, in cycle
14 there is a small zone at approximately 103.7 mm displacement
where the whole cross-section is in tension, albeit barely. In cycle
15 this tensile zone is more pronounced. In cycle 16 more zones are
fully in tension, and cycle 17 is completely in tension. If indeed the
whole cross-section is in tension and remains in tension, the CBT
test has become a conventional tensile test and is subjected to the
well-known instabilities. As the material has strained considerably,
and consequently hardened considerably, fracture will occur very
fast once the process has become unstable. This can also be noticed
in Fig. 8 that presents an actually measured curve, and shows that
fracture is almost instantaneous.
This analysis confirms the hypothesis stated above that the pro-
cess becomes unstable if there is no more part of the cross-section
in compression.
1955
lower
upper
Cycle 14 Cycle 15 Cycle 16 Cycle 17
100 105 110 115 120 125
Cross bar displacement (mm)
5. Conclusions
The continuous bending under tension (CBT) test is an incre-
mental sheet deformation process. The advantages of investigating
bending under tension in the CBT setup rather than on a typ-
ical ISF process are the simple stress field around the rolls
and the absence of doubly curved shapes. The essentially 3-
dimensional complex bending in ISF is reduced by the CBT setup
to a merely 2-dimensional case. Using the CBT test, the effect
of bending on the stability of sheet deformation can be iso-
lated from other sources of extended formability, observed in
ISF.
A 3-dimensional FE model is created for simulation of the CBT
test, using triangular shell elements based on discrete Kirchhoff
theory. Three different meshes are investigated: coarse, intermedi-
ate and fine. The three meshes show no significant difference in the
prediction of the numerically stable part of the force–displacement
curve, although the coarser models show some waviness, related
to the element size. The predicted strain at the onset of instabil-
ity, however, is highly influenced by the mesh size and with the
fine mesh, specimen failure is predicted too early with the current
model.
Three different sets of material parameters are used to model
the CBT test. The models represent isotropic hardening and com-
bined isotropic/ kinematic hardening. Despite the large differences
between these models in uniaxial stress–strain curves after load
reversal and the significant cyclic loading in the CBT test, only
small differences between the models are observed in the predic-
tion of the numerically stable part of the force–displacement curve.
A major difference between the material models is observed in pre-
dicting failure of the specimen, but due to the mesh-dependency of
the failure prediction no conclusions can be drawn on which model
should be preferred.
The force–displacement curve for each half-cycle of the roll
set consists of two parts: a semi-steady part and a peak. The
force level and trend of the semi-steady part is influenced by
hardening of the material, travelling of test material out of the
deformation zone and especially travelling of less deformed mate-
rial from the fixed clamp into the main deformation zone. The
peaks result from increasing the deformation of the strip without
significant bending. The peak is observed directly after reversing
the traveling direction of the roll set because the geometry of the
strip just before the reversal is not symmetric with respect to the
rolls.
The effect of bending in stabilizing the deformation under
tension is numerically investigated. A stable deformation can
be achieved as long as local stretching requires an increase of
the tensile force. Irrespective of plastic hardening, a relatively
large change in the force occurs if a part of the cross-
section still experiences compressive stresses. This stabilizes the
deformation.
1956 A. Hadoush et al. / Journal of Materials Processing Technology 211 (2011) 1948–1956
Acknowledgement
This research was carried out under the project number
MC1.05227 in the framework of the Research Program of the
Materials innovation institute M2i (www.m2i.nl), the former
Netherlands Institute for Metals Research.
References
Batoz, J.L., Bathe, K.J., Ho, L.W., 1980. Study of three-node triangular plate bending
elements. International Journal for Numerical Methods in Engineering 15 (12),
1771–1812.
Benedyk, J., Stawarz, D., Parikh, N., 1971. A method for increasing elongation values
for ferrous and nonferrous sheet metals. Journal of Materials 6 (1), 16–29.
Chaboche, J., 1991. On some modifications of kinematic hardening to improve the
description of ratchetting effects. International Journal of Plasticity 7, 661–678.
Emmens, W.C., van den Boogaard, A.H., 2009a. Incremental forming by continu-
ous bending under tension—an experimental investigation. Journal of Materials
Processing Technology 209 (14), 5456–5463.
Emmens, W.C., van den Boogaard, A.H., 2009b. An overview of stablizing deforma-
tion mechanisms in incremental sheet forming. Journal of Materials Processing
Technology 209 (8), 3688–3695.
Emmens, W.C., van den Boogaard, A.H. Formability in incremental sheet forming
and cyclic stretch-bending. Steel Research International, in press.
Hadoush, A., van den Boogaard, A.H., Huétink, J., 2007. Stable incremental deforma-
tion of a strip to high strain. Key Engineering Materials 344, 615–620.
Hill, R., 1948. A theory of the yielding and plastic flow of anisotropic metals. In: The
Royal Society of London; Series A, vol. 193 , pp. 281–297.
Jeswiet, J., Micari, F., Hirt, G., Bramley, A., Duflou, J., Allwood, J., 2005. Asymmet-
ric single point incremental forming of sheet metal. Annals of the CIRP 54 (2),
88–114.
Marciniak, Z., Duncan, J., 1992. The Mechanics of Sheet Metal Forming. Edward
Arnold, Great Britain.
Matsubara, S., 1994. Incremental backward bulge forming of a sheet metal with a
hemispherical tool. Journal of the Japan Society for Technology of Plasticity 35,
1311–1316.
Van Riel, M., 2009. Strain path dependency in sheet metal, experiments and models.
Ph.D. thesis, University of Twente.
Zienkiewicz, O.C., Taylor, R.L., 2005. The Finite Element Method: For Solid and Struc-
tural Mechanics, 6th ed. Elsevier Butterworth, Barcelona.