Documente Academic
Documente Profesional
Documente Cultură
a r t i c l e i n f o a b s t r a c t
Article history: Coil set is a phenomenon occurring in industrial sheet metal in which the strip retains a residual curvature
Received 18 May 2016 as it comes off the coil. The roll forming of two coils exhibiting coil set is studied; each coil has similar
Received in revised form 18 October 2016 tensile and bending properties. It is shown that there are signicant differences in shape defects in the
Accepted 18 October 2016
industrial roll forming of a particular channel section depending on whether the strip is fed in with
the convex side upwards or downwards. The properties in bending in the longitudinal direction were
Keywords:
determined using a free bending test and found to depend on the direction of bending; when bending
Roll Forming
so that the curvature increased in the direction of the residual curvature, the bending yield stress was
Coil set
Residual Stress
almost 50% lower compared with bending in the direction opposite to the residual curvature, i.e. in
Shape Defects straightening the strip. In the transverse direction, bending properties were independent of the direction
Bending Properties of bending. The defects measured in the roll formed product were twist and are and the magnitude of
both were greatest when the strip was formed with the residual curvature convex upwards in the roll
forming line. The process was simulated using the commercial software package Copra FEA; two sets of
material property data were used both were derived by an inverse method from the bending tests.
One case used bend test data from tests in which the curvature increased in the same direction as the
residual curvature, and the other set for curvature in the opposite direction. The defects predicted by the
numerical analyses reproduced the trends observed in the industrial trials regarding twist and end are
even though the levels predicted were too high. Comparison of the bending test results with other work
suggests that the strip is subjected to plastic deformation (straightening) as it comes off the coil resulting
in an asymmetric longitudinal residual stress distribution through the thickness. Both the experiments
and the results of the simulation strengthen the view that differences in the mechanical behaviour in
bending near the elastic plastic transition indicate the presence of residual stresses that inuence nal
shape in the roll forming process.
2016 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jmapro.2016.10.005
1526-6125/ 2016 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
M. Weiss et al. / Journal of Manufacturing Processes 25 (2017) 815 9
plastic strains in the process. This is why changes in the elastic limit
and the pre-and post-yielding behaviour in the incoming strip can
have a signicant effect on shape defects [14]. Additionally, recent
studies have revealed that residual stresses due to steel processing
inuence roll forming shape defects such as springback, bow and
end are [15]. In processing steel, residual stresses can be intro-
duced by skin pass rolling [16], roller levelling and effects related
to the coiling of the strip [17]. Some previous studies have shown
that residual stresses due to skin pass rolling and roller levelling
alter the level of yield stress in a pure bending tests [18] while
there is only a minor effect on the material response in the conven-
tional tensile test [19]. The major deformation mode in roll forming
is bending and this suggests that material data based on the bend-
ing test may be more appropriate to represent material behaviour
in roll forming if residual stresses are present.
The present work arose because it was found that with certain
Fig. 1. Channel section shape formed in the industrial trials. (Dimensions in mm).
coils of stainless steel strip exhibiting coil set, the magnitude of sev-
eral defects differed when the orientation of the coil was reversed,
i.e. when the coil was turned up-side-down so that the residual cur- 2. Product and Industrial Process
vature of the incoming strip was changed from vertically upwards
to downwards. Coil set is a common phenomenon in which the The product studied was a roll formed channel having the cross-
strip retains a residual curvature on uncoiling and can occur when section shown in Fig. 1.
the material is bent past its yield point during coiling in a cold The channel section was formed in 11 forming steps; lubrication
or hot rolling mill [20]. Pure bending tests revealed that the strip was limited to a 12% water soluble oil coolant. The constant radius
showed asymmetrical bending properties in that the elastic plastic method was used in the roll pass design, where the prole radius is
transition moment for bending in the same direction as the resid- the same at each roll and the bend angle is increased incrementally
ual curvature was less than half of that in the opposite direction, through the forming steps. The forming sequence is shown in Fig. 2.
suggesting the presence of residual stress. A previously published The distance between the roll stations in the roll former was
inverse method [21] to derive a stress strain curve from the bend- 550 mm and the roll gap was set according to the material thickness
ing moment characteristic was used to obtain two different stress of 3.4 mm. In forming stations, 1-6, both the top and bottom rolls
strain curves that were used as input to a nite element simulation were driven; in stations, 7-11, only the bottom shaft was driven. In
of the roll forming process. The nite element system used a single stations 5-11 the bottom cone rolls were split with the outer seg-
solid element for the full thickness so that an initial residual stress ments being free spinning rolls that were positioned on bearings
distribution could not be represented and input was limited to a to more closely match the line speed of the sheet. This signicantly
stress-strain curve. It should be emphasised that the stress strain reduced the torque required and prevented scufng. The material
curve used here in the numerical analysis does not include residual was fed from a un-coiler as shown in Fig. 3a and no levelling equip-
stress information, but reproduces the observed moment curvature ment was used. After the nal forming station (Fig. 3b), sections of
behaviour in plane strain bending of strip that does contain residual 2 m length were cut using the shear cutter shown in Fig. 3c.
stresses. The numerical results achieved this way reproduced the Two conditions of feeding the coil were investigated one in
trends observed experimentally for the effect of coil feed position which the lip of the coil and the direction of residual curvature
on the level of twist and end are but signicantly overestimated point upwards (Convex upwards) and the other in which the lip of
the magnitude of shape defects. The results of this study suggest the coil is directed downwards (Convex Downwards) as shown in
that if there is residual stress present in the incoming strip, mate- Fig. 4.
rial input that is based on bending test data may lead to an improved Since it was not possible to change the position of the same coil
representation of material behaviour in the numerical analysis of during the run, two different coils of the same material were inves-
the roll forming process compared with input that is based on the tigated (coil 1 and coil 2), with coils 1 and 2 being fed in convex
conventional tensile test. upwards and the convex downwards positions respectively. Sam-
Fig. 3. Photographs of the roll forming set up showing a) the roll forming line including the un-coiler, b) the nal forming station and c) the material cut-off.
4.1. Twist
Twist per unit length, d/dl, (Fig. 5a.) was analysed by scan-
ning the outer surfaces of the formed sections after cut off using an
ExaScan 3D scanner [22] as shown in Fig. 5b.
The resolution of the scanned surface was 0.05 mm giving an
accuracy of 0.04 mm. The scanned surface was aligned with the
ideal surface generated by SolidWorks as shown in Fig. 5b using the
software package Geomagic [23]. By comparing the cross sections
measured at an offset of 20 mm from the part edge at the front and
the back of the scanned component with that of the ideal surface the
rotation angles at the front, F , and at the back, B , were measured
Fig. 4. The two material feed-conditions investigated in this study. and the overall twist per unit length determined using the relation
d (F B )
= /m (1)
ple material for the tensile and bending tests, which are further dl 2
explained later on, was taken from each of the two coils from the
mid-width of the strip. Three sample sections of 2 m length were 4.2. End Flare
cut close to that region of the coil for both coil feed conditions after
roll forming. End are, s, was determined by measuring the opening of the
section (Fig. 5a) at the front and the back end of the product using
a vernier caliper and applying the relation
3. Material
s = La Lp (2)
The material used in this study was a ferritic chromium-nickel
steel with added Titanium (Thyssen 5CR12 Ti) having a thickness with the actual opening, La , and the prescribed opening, Lp .
M. Weiss et al. / Journal of Manufacturing Processes 25 (2017) 815 11
Table 1
Chemical composition (wt%) of the two coils of Thyssen 5CR12 Ti steel in weight percentage determined with a SPECTRO analytical instrument.
IADS designation C Si Mn P S Cr Ni Fe
Coil 1 (wt%) 0.0487 0.281 2.00 0.0045 <0.0005 11.5 0.655 85.0
Coil 2 (wt%) 0.0453 0.284 1.99 0.0019 <0.0005 11.66 0.672 84.8
Fig. 5. a) Denition of twist per unit length,d/dl, and of the actual opening, La . b) Comparison of the scanned surface with the ideal CAD surface.
the left of the diagram with a 40 mm free bending length. The pins
on the right-hand ends of the bending arms are attached to a ten-
sile testing machine at the places normally occupied by the tensile
grips. The moment curvature characteristic can be calculated from
the load, F, and the displacement of the cross-head as detailed in
[24]. In these tests, a clip-on curvature gauge was used to obtain a
more accurate measure of curvature in the test-piece. This device
is fully described in [25].
The strips were bent in two directions as shown in Fig. 6b. In
one, the residual curvature was increased and in the other, the cur-
vature was changed in the opposite direction so that the test-piece
was initially straightened and then bent in the other direction. In
the tests, the curvature is increased until the outer bre strain was
about 0.014, i.e. 1.4%. Three to four tests were performed for each
condition on test samples oriented in the longitudinal direction and
transverse to the rolling direction.
5.1. Tensile tests Inverse analysis in combination with a numerical model of the
pure bending test was applied to generate material input data for
The tensile tests were performed in a 100 kN Instron machine the subsequent numerical analysis of the roll forming process. The
in accordance with Australian Standard AS 13911991; specimens inverse modelling technique is described in detail in [21]. It pro-
were oriented along and transverse to the rolling direction. A non- vides a tensile true stress strain curve that when used in a simple
contact extensometer with a gauge length of 50 mm was used and numerical bending analysis will give an optimised t to the experi-
the cross-head speed was 2 mm/min. A total of three to four tests
The tting was prescribed over the
mental bending characteristics.
were performed for each condition. range of curvature of o R1 8m1 for test data generated on
longitudinally oriented samples for the bending with and against
5.2. Bend tests the coil set direction. This stress strain curve does not contain
directly information on residual stress; it is a true stress strain
Pure bending tests were performed using the apparatus shown curve that will predict a moment curvature relation similar to that
in Fig. 6a. Specimens were clamped between the grips shown on observed experimentally.
12 M. Weiss et al. / Journal of Manufacturing Processes 25 (2017) 815
Fig. 7. FEA model of the roll forming process and material cut off.
A 3-D nite element model was developed for the industrial roll
forming process using Copra FEA [26], which is an implicit solver
(Fig. 7). The strip length was limited to 3 m and the roll forming
from a coil was simulated by xing the movement of the sheet at
both ends in the z-direction and at 6 centre nodes at the back end
of the strip in x direction.
Eight-node, isoparametric, arbitrary hexahedral elements (Ele-
ment Type 7 in the MSC Marc Element library [27]) were used to
mesh the metal strip with a single element through the thickness
to reduce calculation time. The forming rolls were dened as rigid
surfaces and their movement was xed in ve degrees of freedom
allowing only the movement in the z-direction. The same station
distance as used in the industrial trials was chosen and frictionless
contact, = 0, was dened with the touching contact condition.
The elastic material properties applied were a Youngs Modulus Fig. 8. True stress strain curves obtained from tensile tests on each coil in the
of 210 GPa and a Poissons ratio of = 0.33. To dene the material longitudinal and transverse directions.
behaviour in the plastic range, the von Mises yield condition was
applied. Two sets of material input (Curve A and Curve B) were
utilised. Those were derived from the experimental bend test data plastic transition is only about one-half of that for bending in the
using the inverse routine and are shown in Fig. 10b. During form- opposite direction (against coil set and further straightening the
ing, the rolls were moved in the negative z-direction over the sheet sheet). No effect of the bending direction on the bending behaviour
to form it into shape. To simulate the cut off of the 2 m section at was observed for transversely oriented samples (Fig. 9b).
the end of the simulation, the back end of the part was xed at four The true stress strain curves extracted from the experimental
nodes in the bottom in the x, y and z direction (this restricted the moment curvature data shown in Fig. 9a using the inverse method
movement and the rotation of the part after cut off) and all elements are compared to the experimental tensile curve in Fig. 10b; the t
situated outside the 2 m boundary deleted using the Deactivation achieved with the experimental moment curvature data is illus-
command. trated, for bending against and with coil set, in Fig. 10a. Notice that
To measure Twist and End Flare, section cuts were taken at the the curves in Fig. 10b are shown in an expanded scale at the low
front and the back end of the part (after cut off) and evaluated in strain range as the difference between them at higher strain is not
the CAD package, AutoCad using the same procedure as applied in great. The curves are given as they were later applied to dene
the experimental trials and described in Section 4. material behaviour in the numerical model of the roll forming pro-
cess, i.e., without the elastic component.
A clear difference, especially in the yield point and the material
7. Results hardening at low plastic strain, can be seen between the tensile
curves generated by the inverse method based on the two sets of
Both coils showed almost identical tensile properties in the lon- experimental bend test data (Fig. 10b). For tests in which the curva-
gitudinal and transverse directions as can be seen in Fig. 8. ture was increased in the same direction as the residual curvature
The moment curvature diagrams obtained in the pure bending after uncoiling, Curve A, the tensile test curve derived by the inverse
tests are shown in Fig. 9. For samples oriented longitudinal to the method showed a low yield (YP = 163 MPa) and high initial material
rolling direction (Fig. 9a) it may be seen that when the curvature is hardening. This is in marked contrast to that obtained from bending
increased in the direction of the coil set (with coil set), the elastic against the coil set direction (Curve B); the latter curve is compara-
M. Weiss et al. / Journal of Manufacturing Processes 25 (2017) 815 13
Fig. 9. Moment curvature characteristics obtained from bend tests in two different directions. a) Samples oriented in the longitudinal direction. b) Samples oriented transverse
to the rolling direction.
Fig. 10. a) Numerical simulation result for the moment curvature diagram based on tensile input generated by the inverse routine compared to the experimental results. b)
Tensile input generated by the inverse analysis for bending with and against coil set compared to the experimental tensile test data.
ble to that obtained in the experimental tensile test with an initial relation between the numerical predictions and the experimental
yield stress of YP = 310 MPa and a low initial strain hardening rate. values can be observed for end are in the front section (Fig. 12b).
The magnitude of twist predicted by the numerical analysis is
shown in Fig. 11 and compared with the experimental values. 8. Discussion
It was found that using Curve A (Fig. 10b.) to model the con-
vex upwards coil position and Curve B (Fig. 10b.) for the convex The industrial trials showed a clear difference in the magnitude
downwards coil position gave a similar trend to that observed of shape defects twist and are for two coils of the same mate-
experimentally, although the magnitudes of the twist predicted rial and tensile properties (Fig. 8) depending on the orientation of
were much greater than experimentally observed. the incoming strip. The overall level of twist as well as end are
The same selection of stress strain curves associated with the in the back of the section after cut off were higher if the coil was
two forming conditions shown in Fig. 11 was also used in the fed convex upwards compared to the convex downwards position
numerical modelling to predict are. For are in the back section, (Figs. 11 and 12a); the opposite trend was observed for end are in
Fig. 12a, the predicted values are lower than those determined the front section of the roll formed prole (Fig. 12b).
in the industrial trials, but the reduction in end are observed in Free bend tests on the uncoiled strip showed marked asymmetry
the experiments for the change in coil feed condition from convex in the moment curvature characteristics; the elastic plastic transi-
upwards to convex downwards is reproduced. A very good cor- tion moment in strip bent in the same direction as the residual
curvature (with coil set) was only about one-half of that for strip
14 M. Weiss et al. / Journal of Manufacturing Processes 25 (2017) 815
Using this data, shape defects were predicted that followed the
trends observed in the industrial trials. The tensile tests did not
show any difference in the stress strain curves of each coil (Fig. 8)
and therefore numerical analysis based on tensile data would not
have revealed any difference in the conditions for coil feeding. On
the other hand, when two different stress strain curves were used in
the simulation that were based on the tests in bending in two dif-
ferent directions, differences were revealed in shape defects that
followed the trends observed. This suggests that material input
based on the bend test may lead to improved model accuracy in
roll forming simulations if residual stresses are present in the metal
strip. The simulation results achieved in this way only reproduced
the general trends observed in the industrial trials but signicantly
overestimated twist (Fig. 11) and underestimated end are at the
back end of the section (Fig. 12a). A possible explanation for this
is that since the experimental roll forming trials in this study were
performed in an industrial environment the exact tool set up may
not have been reproduced in the numerical model in sufcient
detail. In particular it is possible to correct twist by rotating the
tooling at the exit shown in Fig. 3c. It is likely that during tool setting
of the roll forming line, slight modications were done by the oper-
Fig. 11. Measured and simulated twist in the section where Curves A and B (see ators to reduce twist in the section to an acceptable limit and these
Fig. 10b) were used for material inputs for coil feed conditions Convex upwards and are not included in the model input. It also should be pointed out
downwards respectively. that the nite element model used to analyse the roll forming pro-
cess in this study employs a single element in the through-thickness
direction and cannot take account of a residual stress prole in the
material. Likewise, the inverse modelling method also assumes no
bent in the opposite direction (Fig. 9a), i.e. when further straighten- residual stress in the bending element. The trends shown in the
ing the strip. Similar results have been reported in previous studies numerical analyses are therefore purely related to the reduction in
that investigated the bending response of roller levelled and skin bending yield stress when changing the material input from curve
passed metal strip [18]; here the change in bending yield stress B to curve A (Fig. 10b); in roll forming shape defects are often the
was related to the presentence of a residual stress prole through result of very small magnitudes of permanent longitudinal defor-
the material thickness [19]. In a later study it was observed that mation and have been found to strongly depend on material yield
residual stress and the resulting change in bending yield stress can strength [28].
signicantly inuence shape defects that are common in the roll Nevertheless, a recent study has shown that to accurately cap-
forming process such as bow and end are [15]. This suggests that ture the effect of residual stress on nal shape in roll forming with
the effect of the coil feed position on twist and end are observed numerical analysis, material input that includes detailed informa-
in the industrial trials of this study may be the result of residual tion in regard residual stress and the stress and strain history from
stress and previous investigations have conrmed that the coiling the preceding metal production processes (skin passing, roller lev-
and uncoiling of metal strip can lead to the development of residual elling or coiling/uncoiling) is required [15]. This suggests that part
through thickness stress in metal strip [17]. of the overestimation of the magnitude of shape defects by the
An inverse method was employed to obtain optimised tensile numerical model of this study may be due to the fact that the stress
stress strain curves that tted the observed bending characteristics. strain curve used did not have information on residual stresses, but
These were used as input to a nite element simulation package.
Fig. 12. Measured and simulated are for a) Back Section b) Front Section. Curves A and B were used for material inputs for coil feed conditions convex upwards and
downwards respectively.
M. Weiss et al. / Journal of Manufacturing Processes 25 (2017) 815 15
Acknowledgements