Experimental and Computational Investigation of The Roll Forming Process

DOC TOR A L T H E S I S
ISSN: 1402-1544 ISBN 978-91-7439-031-5

Lule University of Technology 2009
Michael Lindgren
Division of Material Mechanics
Experimental and Computational Investigation

of the Roll Forming Process
Experimental and Computational Investigation of the Roll Forming Process
Michael Lindgren
Experimental and Computational Investigation

of the Roll Forming Process
Michael Lindgren
Lule University of Technology

Division of Material Mechanics
Printed by Universitetstryckeriet, Lule 2009

ISSN: 1402-1544
ISBN 978-91-7439-031-5
Lule 2009
www.ltu.se
Preface
This work has been carried out at Dalarna University. The nancial support
was provided by the Swedish Foundation for Knowledge and Competence Development (KK-stiftelsen), ORTIC AB, Jernkontoret and Dalarna University.
I would like to thank the following people:
My supervisor, professor Lars-Erik Lindgren for his experienced guidance and
continual support.
Dr Lars Ingvarsson, for sharing his great experience in roll forming with me.
All colleagues at Dalarna University and ORTIC AB, for their friendship and
for making the workplace a great place to be at.
Finally, I would like to thank my family, Jenny, Fanny and Sandra for always
being there.
Borl
ange, November 2009
Michael Lindgren
ii
Abstract
One of the rst questions to consider when designing a new roll forming line
is the number of forming steps required to produce a prole. The number
depends on material properties, the cross-section geometry and tolerance requirements, but the tool designer also wants to minimize the number of forming steps in order to reduce the investment costs for the customer. There are
several computer aided engineering systems on the market that can assist the
tool designing process. These include more or less simple formulas to predict
deformation during forming as well as the number of forming steps. In recent
years it has also become possible to use nite element analysis for the design
of roll forming processes.
The objective of the work presented in this thesis was to answer the following question:
How should the roll forming process be designed for complex geometries and/or
high strength steels?
The work approach included both literature studies as well as experimental
and modelling work. The experimental part gave direct insight into the process
and was also used to develop and validate models of the process. Starting with
simple geometries and standard steels the work progressed to more complex
proles of variable depth and width, made of high strength steels. The results
obtained are published in seven papers appended to this thesis.
In the rst study (see paper 1) a nite element model for investigating the
roll forming of a U-prole was built. It was used to investigate the eect on longitudinal peak membrane strain and deformation length when yield strength
increases, see paper 2 and 3. The simulations showed that the peak strain
decreases whereas the deformation length increases when the yield strength increases. The studies described in paper 4 and 5 measured roll load, roll torque,
springback and strain history during the U-prole forming process. The measurement results were used to validate the nite element model in paper 1. The
results presented in paper 6 shows that the formability of stainless steel (e.g.
AISI 301), that in the cold rolled condition has a large martensite fraction,
iii
can be substantially increased by heating the bending zone. The heated area
will then become austenitic and ductile before the roll forming. Thanks to
the phenomenon of strain induced martensite formation, the steel will regain
the martensite content and its strength during the subsequent plastic straining. Finally, a new tooling concept for proles with variable cross-sections is
presented in paper 7.
The overall conclusions of the present work are that today, it is possible to
successfully develop proles of complex geometries (3D roll forming) in high
strength steels and that nite element simulation can be a useful tool in the
design of the roll forming process.
iv
The doctoral project has

resulted in following
The Thesis
This thesis consists of a survey and the following appended papers:
1. M. Lindgren, Finite Element Model of Roll Forming of a U-Channel Prole, Presented at the international conferences on technology of plasticity, Verona, Italy October 2005.
2. M. Lindgren, Cold Roll Forming of a U-channel Made of High Strength
Steel, Journal of Materials Processing Technology 186 (2007) 77-81.
3. M. Lindgren, An Improved Model for the Longitudinal Peak Strain in
the Flange of a Roll Formed U-Channel developed by FE-Analyses, Steel
Research Int. 78 (2007) No. 1.
4. M. Lindgren, Experimental Investigation of Roll Load and Roll Torque
when High Strength Steel is Roll Formed, Journal of Materials Processing Technology, 191 (2007) 44-47.
5. M. Lindgren, Validation of Finite Element Model of Roll Forming, Presented at international deep drawing research group, Olofstr
om, Sweden,
June 2008.
6. M. Lindgren, U. Bexell, L.Wikstr
om, Roll Forming of Partially Heated
Cold Rolled TRIP Steel, Journal of Materials Processing Technology,
209 (2009) 3117-3124.
7. M. Lindgren, L-O. Ingmarsson, 3D Roll-forming of Hat-section With
Variable Depth and Width, Presented at the 1st international congress
on roll forming, Bilbao, Spain, October 2009.
Author contribution to the publications

1. Single author
2. Single author
3. Single author
4. Single author
5. Single author
6. Writing and done all work in close co-operation with the co-authors
7. Everything except measuring the shape of the produced sections
vi
Notation
Symbols
a
A
b
B
BB
c
C
DS
D
E
e
ep
et
ej
f
fs
Fn
Ft
h
hl
HC , d
H, g
I
kl
L
LS
n
p
R1
Flange length
Integration constant
Web width
Integration constant
The width of the at strip used to formed the nished prole
Forming length
Relative sliding velocity
Diameter of the shaft
Tolerance distance or horizontal distance between the forming steps
Youngs modulus
Longitudinal engineering strain in the edge of the ange
Longitudinal peak strain in the edge of the ange
Transverse longitudinal strain in the ange
Extra pass.
Yield criterion
Tolerance factor.
Normal force
Tangential force
The distance from the neutral layer to the inner side of the bend
The vertical distance from where the prole will hit the lower
tool to the roll gap between the upper and lower tool
Horizontal distance between two forming steps
Height of the nished prole
Second moment of inertia
Constants used in transverse bending
Deformation length
Length of a shaft
Number of forming steps
Perpendicular moment arm
Female tool radius
vii
r
r0
s
sij
sf
t
Us
VC
vr
Wb
Ws
Wsl
Wt
Y, y
z
x
X
z
1,2
Ls , x
Lb
t
y
pij
, xy
Distance from bend

Radius to the neutral layer
Parallel moment arm
Deviatoric Cauchy stress
Shape factor.
The thickness of the material
Ultimate tensile strength.
The vertical distance between centre of upper and lower tool
Sliding velocity
Plastic work due to transverse bending
Plastic work due to longitudinal bending per unit volume
Plastic work due to longitudinal bending per unit length
Total plastic work for one bending
Yield strength
Coordinate in longitudinal direction
Coordinate in transverse direction
Distance between centre of the roll station and
there the prole will hit the lower tool
Pre-punched hole
Bend angle for the active bend
Bend angle for the female tool
Bend angle increment
The derivate of the bend angle
Total degree of formed angle.
Angle between the web and the innite small element
Bend angle for the inactive bend
Longitudinal stretching of the ange
Longitudinal bending of the ange
Strain due to transverse bending in the bending zone
Transverse strain of the ange
Plastic strain
Plastic parameter
Friction coecient
Eective stress
Shear strain in the ange
Forming angle
viii
List of Figures
1.1
2.1
2.2
2.3
2.4
2.5
2.6
2.7
3.1
3.2
3.3
3.4
3.5
3.6
3.7
Roll formed products have many applications in the automotive

sector, buildings, domestic appliances etc. . . . . . . . . . . . .
The strip is formed in several steps, beginning with an undeformed strip and ending with the nished prole, Lund et al.
[41]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The material in the ange will travel a longer distance than the
material in the bending zone. The dierence will cause strain
(e) in the ange. In this gure (H) represents the ange length,
(L) the deformation length and (e) the strain in the ange, Lund
et al. [41]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dierent defects that can occur in the nished prole if the roll
forming process is not well designed. From left: twist, are and
oil canning, Ingvarsson [27]. . . . . . . . . . . . . . . . . . . . .
Lead gate is used to guide the prole into the tools and checked
lead gate is used to control the strip laterally. . . . . . . . . . .
3D roll formed proles in paper 7, Lindgren and Ingmarsson
[39]. From the top: a straight section, a hat-section with a
waist on one side and nally a hat-section conical in depth and
width. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
U-proles roll formed in the experimental 3D roll forming machine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
a) 3D roll forming simulation of the tooling concept in Paper
7, Lindgren and Ingmarsson [39]. b) The real tools which are
modelled in a). . . . . . . . . . . . . . . . . . . . . . . . . . . .
This gure shows the strain histories of the longitudinal web
strain and the longitudinal membrane ange strain for two forming stations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The denition of the parameters for the bend angle and the
bend angle increment. . . . . . . . . . . . . . . . . . . . . . . .
Deformation length for one forming step. . . . . . . . . . . . . .
The prole will be reversely bent since the prole will be curved
down and then lifted up (hl) by the next roll station. . . . . . .
The forming angle method. . . . . . . . . . . . . . . . . . . . .
The ower method: the engineer starts with a nished prole
and progressively unfolds it to a at strip. . . . . . . . . . . . .
Front view and side view of the roll forming mill used in the
Algorithm, equation 3.6 to 3.11, [41]. . . . . . . . . . . . . . . .
ix
4
5
6
7
10
10
12
13
14
15
15
3.8
3.9
3.10
3.11
3.12
3.13
3.14
3.15
3.16
3.17
3.18
3.19
4.1
4.2
u is the distance from the edge of a at strip to the edge of the

nished prole. In this case u is used to calculate the number
of forming steps, equation 3.11. . . . . . . . . . . . . . . . . . .
An element strip between two roll forming passes. . . . . . . . .
View in z-direction. . . . . . . . . . . . . . . . . . . . . . . . . .
The bend angle is divided into three stages: stage one when the
angle does not change, stage two when the angle changes but
the strip is not in contact with the tool and stage three when
the strip is contact with the tool. . . . . . . . . . . . . . . . . .
The V-prole made of high strength steel and mild steel were
roll formed in four forming steps: 15o , 30o , 45o and 60o . The
V-prole made of high strength steel was straight after forming
but the prole made of mild steel was not. . . . . . . . . . . . .
Delta Flare is the dierence between the springback in the centre of the prole and the end of the prole. This is a plot for 4
dierent ultra high strength steels with various yield strength. .
To the left a C-prole is roll formed in 8 passes and to the right a
top-hat prole is roll formed in 5 passes. The C-prole is formed
mainly by bending whereas the top-hat prole is formed both
by bending and drawing [13]. . . . . . . . . . . . . . . . . . . .
This gure shows three dierent ways of forming a 90o prole.
One way is to let the tools have the same radius in every forming step and dierent arc length. Another way is to let the arc
length be the same in every forming step but decrease the tool
radius. These methods give dierent spring back and longitudinal residual stresses in the bending zone. The constant arc
length is the method most commonly used. . . . . . . . . . . .
The deformed sheet between roll stand (i) and roll stand (i+1).
The neutral layer of the sheet is described by the shape function,
Equation 3.32. . . . . . . . . . . . . . . . . . . . . . . . . . . .
Shape functions for dierent parameters (q). . . . . . . . . . . .
To the left: the unfolded C-prole, the blank of sheet metal.
To the right: two dierent cross-sections of the C-prole. . . . .
A 3D roll-formed U-prole. Tension stress acts on the transition
zone where the U-prole is small and compression stress acts
on the transition zone where the prole is wider. . . . . . . . .
A universal, spindle type roll-forming mill where the lower tools
are driven with universal joint driven shafts. . . . . . . . . . . .
Geometry of the tools used in the experiment. The tools have
a bend angle of 20o , 40o , 60o and 80o . . . . . . . . . . . . . . .
x
17
19
19
22
24
25
26
27
28
28
33
33
36
37
4.3
4.4
4.5
4.6
4.7
4.8
4.9
4.10
4.11
4.12
4.13
5.1
5.2
5.3
5.4
This gure shows only half of the U-channel because the symmetry and the rosette strain gage is bonded on the top surface
close to the edge of one ange. The longitudinal direction is
the same as x-direction and the transverse direction is same as
y-direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The roll load in z-direction was measured with two donut cells.
The torque sensor was used to measure the roll torque and it
was mounted between the lower tool and power transmission.
The equipment measured the forces of each forming step, one
by one. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The spring back, width and depth between the dierent forming
steps were measured with a sliding caliper. . . . . . . . . . . . .
The picture shows the prototype resistance heating machine and
the steel strip which was fed through the machine by wheels
made of copper. . . . . . . . . . . . . . . . . . . . . . . . . . . .
A sketch of the electrical circuit of the prototype resistance
heating machine. Wikstr
om [56]. . . . . . . . . . . . . . . . . .
The tools and the ower pattern. The V-section was roll formed
in 6 forming steps, from 15o to 120o . . . . . . . . . . . . . . . .
The experimental equipment used for research and prototyping
of 3D roll formed proles. . . . . . . . . . . . . . . . . . . . . .
The experimental equipment has 6 forming stands. Each stand
has 4 units (tool holders) that can rotate and translate. . . . .
View from the top: The prole is roll formed in six passes and
to do that the prole must go through the machine two times.
Forming stands number 1, 3 and 5 form the left side and forming
stands number 2, 4, and 6 form the right side. . . . . . . . . . .
View from the back. The geometry of the tools is the same for
all forming stands. The only dierence is that the tools for the
ange are moving up and closer to the tools that hold the web,
for example bend angles 30o and 60o . Tools with a constant
radius have been used, Chaing [9]. . . . . . . . . . . . . . . . .
The xture used. . . . . . . . . . . . . . . . . . . . . . . . . . .
The nite element model described in paper 5, Lindgren [37]. .
The geometry of the tools and the corresponding ower pattern
described in paper 5. . . . . . . . . . . . . . . . . . . . . . . . .
Four forming stands were used. The two rst stands were used
as a belt feeder, the other two forming stands have the same
bend angle. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The tensile test data for the materials used in paper 5. . . . . .
xi
37
38
39
40
41
42
43
44
44
45
46
47
48
49
51
5.5
The used tolerance distance D1 and D2 in paper 5 is 5 % of the

thickness. The bias factor is default 0 in MARC which means
that D1 = D2. In paper 5 is a bias factor of 0.25 used which
means D1 < D2. . . . . . . . . . . . . . . . . . . . . . . . . . .
xii
52
Contents
1 Introduction
1.1 Design of the roll forming process . . . . . . . . . . . . . . . . .
1.2 Research Question and Approach . . . . . . . . . . . . . . . . .
1
1
2
2 The roll forming process

2.1 Traditional roll forming . . . . . . . . . . . . . . . . . . . . . .
2.2 3D roll forming . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
3
5
3 Literature Survey
3.1 Experimental work . . . . . . . . . . . . . . . . . .
3.1.1 Strain histories in roll forming . . . . . . .
3.1.2 Longitudinal membrane strain . . . . . . .
3.1.3 Deformation length . . . . . . . . . . . . . .
3.1.4 Roll load . . . . . . . . . . . . . . . . . . .
3.2 Theoretical work . . . . . . . . . . . . . . . . . . .
3.2.1 Number of forming steps . . . . . . . . . . .
3.2.2 Deformation types . . . . . . . . . . . . . .
3.2.3 Deformation length . . . . . . . . . . . . . .
3.2.4 Longitudinal membrane strain . . . . . . .
3.2.5 Geometrical restriction from the female tool
3.2.6 Roll forming of high strength steel . . . . .
3.2.7 Bending method . . . . . . . . . . . . . . .
3.3 Computer simulation of roll forming . . . . . . . .
3.4 3D roll forming, proles with variable cross-section
4 Experimental setup
4.1 The roll forming experiment . . . . .
4.1.1 The roll forming machine and
4.1.2 The tools . . . . . . . . . . .
4.1.3 Strain measurement . . . . .
4.1.4 Roll load and roll torque . . .
4.1.5 Spring back . . . . . . . . . .
xiii
. . .
data
. . .
. . .
. . .
. . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . . . . . . . . .
acquisition equipment
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
9
9
9
9
11
12
13
13
17
19
21
22
23
25
26
32
35
35
35
35
36
38
39
xiv
4.2
4.3
Contents
Partial heating experiment . . . . . . . . . . .
4.2.1 Furnace experiment . . . . . . . . . .
4.2.2 Prototype resistance heating machine
4.2.3 Roll forming machine and tools . . . .
4.2.4 Sample preparation . . . . . . . . . .
3D Roll forming experiment . . . . . . . . . .
4.3.1 The 3D roll forming machine . . . . .
4.3.2 The tooling concept . . . . . . . . . .
4.3.3 Measuring equipment . . . . . . . . .
5 Computational model
5.1 Nonlinear solution procedure and
5.2 The geometry . . . . . . . . . . .
5.3 Shell element . . . . . . . . . . .
5.4 Material model . . . . . . . . . .
5.5 Contact . . . . . . . . . . . . . .
6 Summary
6.1 Paper
6.2 Paper
6.3 Paper
6.4 Paper
6.5 Paper
6.6 Paper
6.7 Paper
of
1
2
3
4
5
6
7
appended papers
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
convergence
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
criteria
. . . . .
. . . . .
. . . . .
. . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
40
40
41
41
42
42
42
43
45
.
.
.
.
.
47
48
48
50
50
51
.
.
.
.
.
.
.
53
53
53
54
54
55
55
56
7 Discussions and conclusions
57
Bibliography
61
Chapter 1
Introduction
1.1
Design of the roll forming process
Roll forming is a metal forming process which is spread throughout the world.
Today, roll forming products have numerous applications, for example in buildings, airplanes and the automotive sector as well as in furniture and domestic
appliances, see Figure 1.1. Roll forming is a highly productive process and its
use increases every year, Halmos [19]. Compared to other metal forming processes the benet of this process is that auxiliary operations, such as punching,
welding, clenching etc. can be included, which makes it possible to produce
proles that are ready to use directly.
Figure 1.1. Roll formed products have many applications in the automotive sector,
buildings, domestic appliances etc.
Introduction
When a new roll forming machine is designed the tool designer must decide
how many forming steps are required to form the prole. The number of steps
depends on the cross-section, tolerances, nish of the surface and material
properties. Today there are several computer aided engineering (CAE) systems, for example ORTIC System [27], COPRA RF [15], PROFIL [55], that
can support the tool designer in creating tools. CAE systems use more or
less simple formulas and rules of thumb for predicting the number of forming
steps and suggesting geometry of the tools. Some CAE programs also include
computerized simulation techniques, for example based on the nite dierence
method, Duggal[12], for investigating tool design proposed by the program.
The advantage of this type of analysis is that the simulation time can be very
short. A few years ago nite element analysis was not used for designing roll
forming processes or roll formed proles. Today, however, there are examples
of new proles successfully developed with the help of nite element simulations. This research was devoted to nite element simulations as well as
experiments, and great eort was put into building experimental equipment.
The experiments were carried out in parallel with the building of nite element
models and simpler models.
1.2
Research Question and Approach
The research question of the work can be stated as:

The approach used was a combination of literature studies, and experimental
and modelling work. The experimental part gave direct insight into the process
and was also used to develop and validate models of the process. The work
started with simple geometries and standard steel and progressed with more
complex proles of variable depth and width, made of high strength steel.
Chapter 2
The roll forming process

2.1
Traditional roll forming
In roll forming bending is done in several steps, beginning with an undeformed

strip and ending with the nished prole, see Figure 2.1. Forming is a geometrically complex process since the forming takes place not only in the forming
tools, but also between each forming stand. As a result of the latter, the material in the ange will travel a longer distance than the material in the bending
zone, see Figure 2.2. This causes a longitudinal strain in the ange, which
should not be plastic in order to avoid large, longitudinal residual stresses in
the nished prole.
Figure 2.1. The strip is formed in several steps, beginning with an undeformed strip
and ending with the nished prole, Lund et al. [41].
All materials that can be bent, such as aluminium, steel, stainless steel, copper,
can also be roll formed and the material can be pre-painted or pre-coated. The
speed at which a prole can be produced varies between 15 m/min and 185
m/min, Tool and Manufacturing Engineers Handbook [3], depending on the
tolerance of the cross-section, the material and how fast the machine can be
fed with raw material or how fast the nished product can be removed from
the run out table. The thickness of material that can be roll formed ranges
from 0.15 mm to 19 mm, Kolev [31]. The roll forming process is very robust
provided it is set up correctly, and the geometry of the proles produced shows
only small variations. However, the prole can have defects as bows, twist,
are, spring back and oil-canning, see Figure 2.3, if the roll forming process is
not well designed. Many defects depend on the number of forming steps being
too small, which gives residual stresses in the prole that cause unwanted
deformations.
Figure 2.2. The material in the ange will travel a longer distance than the material
in the bending zone. The dierence will cause strain (e) in the ange. In this gure
(H) represents the ange length, (L) the deformation length and (e) the strain in the
ange, Lund et al. [41].
Figure 2.3. Dierent defects that can occur in the nished prole if the roll forming
process is not well designed. From left: twist, are and oil canning, Ingvarsson [27].
2.2 3D roll forming
Today simulations are used more and more in the industry and especially so
in the automotive sector for industrial engineering, crash tests, sheet metal
forming and so on. The need of simulation also includes roll formed products.
Important advantages of the nite element analysis are that not only can it
provide the same information as computerized simulations, it also includes the
eect of lead gates, see Figure 2.4, and end eect of pre-cut material among
other things.
Figure 2.4. Lead gate is used to guide the prole into the tools and checked lead
gate is used to control the strip laterally.
In the last 10 years a number of research papers have been published where
more or less simplied simulation models have been compared with experiments, for instance Hellborg [24], Bui and Ponthot [8], Sukmoo et al. [53] and
paper 5 in this work, Lindgren [37]. These studies show that the nite element
method is a very valuable tool in designing the roll forming process. Some of
the simplied models also show the possibility of shortening the simulation
time, for example by ignoring the friction between sheet and tooling. Today
there are commercial programs that can reduce the time spent on modelling
the process, COPRA RF [15]. In combination with increased computer capacity this makes the nite element modelling a standard tool in designing the
roll forming process.
2.2
3D roll forming
The use of the roll forming process grows every year thanks to the possibility of forming complicated products in dierent metals in combination with
high productivity. Until eight years ago a disadvantage of roll forming was
that only sections with a constant cross-section could be produced. Today
the problem is solved, ORTIC [1], and 3D roll forming technology is used in
the building industry to produce panels with variable cross-sections and variable longitudinal curvature. The method is very exible which means that
panels can be produced with the same set-up of roll forming tools. For ex-
ample, the Budapest Arena is covered with about 4700 dierent, individually
shaped panels. The automotive industry has become interested in 3D roll
forming as a result of the methods exibility making it particularly suitable
for components made of high strength steels. However, the automotive industry demands design rules, simulations and prototypes to be convinced of
the benets. To meet their demands, probably one of the worlds rst 3D
roll-forming machines, with two translational and two rotational degrees of
freedom per axis, was built within this research work (see Chapter 4). Experiences and results obtained are presented in depth in paper 7, Lindgren and
Ingmarsson [39], of this thesis. Examples of manufactured proles are shown
in Figure 2.5 and Figure 2.6. For the same purpose more or less advanced
experimental equipment have been built by other research institutes and universities also, including the Labein Tecnalia [46], the University of Technology
in Darmstadt [16], Takushoku University in Tokyo [44].
Figure 2.5. 3D roll formed proles in paper 7, Lindgren and Ingmarsson [39]. From
the top: a straight section, a hat-section with a waist on one side and nally a hatsection conical in depth and width.
Today, only a few papers describe nite element simulations of 3D roll forming,
Gulceken et al. [18]. The main dierence between traditional roll forming and
3D roll forming is that modelling the 3D process requires much more work
since the motion of every tool must be exactly described. In the nal paper of
this thesis, Lindgren and Ingmarsson [39], a new tooling concept is described
where hat-proles with variable depth and width are roll formed. In parallel
with the tooling concept, a rst nite element model was developed and used
to simulate the process, Sagstrom [48], see Figure 2.7. These results are not
included in the thesis.
2.2 3D roll forming
Figure 2.6. U-proles roll formed in the experimental 3D roll forming machine.
Figure 2.7. a) 3D roll forming simulation of the tooling concept in Paper 7, Lindgren
and Ingmarsson [39]. b) The real tools which are modelled in a).
Chapter 3
Literature Survey
A literature survey of research concerned with the roll forming process is
presented below. The chapter is divided into four parts: experimental works,
theoretical work, computer simulations and 3D roll forming.
3.1
3.1.1
Experimental work
Strain histories in roll forming
The strain history, when roll forming a U-prole, has been measured in several
experiments, [9], [4], [51]. These experiments show, see Figure 3.1, that the
longitudinal ange strain starts between the forming stations and increases
rapidly to a maximum value just before the centre of the forming station. The
web strain exhibits the same behaviour but the strain is more compressive.
These strains decrease rapidly when approaching the roll centre. Some strains
cause residual stresses that give rise to defects on the nished prole. Chiang
[9] also measured the longitudinal membrane strain across the strip when
forming a U-prole and a V-prole. The measurement results showed that the
strain is largest at the ange edge.
3.1.2
Longitudinal membrane strain
Bhattacharyya and Smith [4] investigated the longitudinal strain by use of

a single roll station as well as multiple roll stations. They also investigated
the strain variation with dierent bend angles and bend angle increments, see
Figure 3.2.
The results obtained were as follows:
In the case with a single roll station (0o o1 ) the strain increased almost
linearly with the bend angle, o1 .
9
10
Literature Survey
Figure 3.1. This gure shows the strain histories of the longitudinal web strain and
the longitudinal membrane ange strain for two forming stations.
When multiple roll stations were used (0o o1 0o ), where the rst and
last forming steps were at rolls, the longitudinal strain was reduced by
10 15 % more than in the case with a single station.
With multiple roll stations (0o o1 o2 ) the strain level was on the
same level as in the case (0o o1 0o ).
The level of peak strain depends on the bend angle increment =
(o2 o1 ) and not the roll angle used at the roll station.
Figure 3.2. The denition of the parameters for the bend angle and the bend angle
increment.
Chiang [9] investigated how ange length (a), web width (b) and bend angle
() eect the web strain and the longitudinal membrane strain in the ange of
3.1 Experimental work
11
a U-prole. A single roll station was used to study the bend angle, and the result obtained was similar to those obtained by Bhattacharyya and Smith. The
longitudinal strain increased almost linearly with the bend angle but the variation for the web strain was small. In the experiment with varying ange length
the longitudinal strain decreased when the ange length increased. When the
web width increased the longitudinal strain slightly decreased, see Table 3.1.
Table 3.1. The peak strain in the ange edge for various ange lengths and web
widths.
Web(b)
20
20
20
30
30
30
Flange(a)
10
15
20
10
15
20
Average peak strain (103 )

3.75
3.10
2.70
3.70
2.85
2.60
In an experiment Zhu [57] investigated the inuence of ange length (a), material thickness (t), the bend angle () and bend angle increment () on the
longitudinal strain distribution. The results obtained were as follows:
The longitudinal strain increases in the beginning with a ange length
shorter than 15 mm, but with a ange length over 15 mm the strain
starts to decrease.
When the material thickness increases the longitudinal strain increases
as well.
An increasing bend angle increment increases the longitudinal strain.
An increasing bend angle at a constant bend angle increment decreases
the longitudinal strain. For example, the longitudinal strain is higher for
0o 20o than for 20o 40o . This was not obtained by Bhattacharyya
and Smith [4].
3.1.3
Deformation length
In a study Bhattacharyya et al. [5] formed a U-prole and measured the deformation length, see Figure 3.3, for mild steel and aluminium. The results were
compared with a model, equation 3.1, which they had developed. The derivation of the deformation length will be reviewed in the section on theoretical
work.

8a3
(3.1)
L=
3t
12
Literature Survey
Figure 3.3. Deformation length for one forming step.
Their conclusions were as follows:

There is a good agreement between the experiment results and Equation
3.1. In most cases the discrepancy was around 6 %.
Based on the good agreement they concluded that the deformation length
(L) depends on three variables: ange length (a), bend angle increment
() and material thickness (t), whereas it is independent of material
properties.
3.1.4
Roll load
Bhattacharyya et al. [6] also investigated the roll load when U-proles of mild
steel with thicknesses of 0.6 mm, 1.0 mm and 1.2 mm, as well as a U-prole
of aluminium with a thickness of 0.6 mm, were roll formed. The investigated
bend angles () ranged from 20o to 40o at intervals of 5o , and to simulate
multiple roll pass situations a roll schedule of 0o o 0o where o was
chosen as above. The width of 40 mm to 60 mm was formed depending on
which ange length (a) was used. The roll load was measured with a load
dynamometer. The roll clearance was set to give an initial bite of 200 N and
this roll load was subtracted from the measured roll load. The result from
the experiment was compared with a model, equation 3.2, which they had
developed. The derivation of equation 3.2 is extensive and can be found in
Bhattacharyya et al. [6]. The following is a brief explanation: the rst term
in equation 3.2 is load caused by bending and stretching of the prole. The
second term is from the reverse bending casing by the prole will curve down
after the rst roll station and then be lifted up by the next roll station, see
Figure 3.4. The second term comes from small deection theory.
3.2 Theoretical work
13
P =Y
2t3 3 a
3hl EI
+
2
3 sin (D X)3
(3.2)
Their conclusions were as follows:

The experiments and equation 3.2 agreed quite well, the discrepancy
being 0 - 20 %.
The theory showed that the roll load depends on yield strength (Y ) of
the material, thickness (t), bend angle () and ange length (a).
The theory also showed that when multiple roll stations are used the
roll load depends on an extra load, the second term in equation 3.2 ,
This term includes the Youngs modulus (E) of the material, the second
moment of inertia (I) of the prole and machine data (D, hl, X).
Figure 3.4. The prole will be reversely bent since the prole will be curved down
and then lifted up (hl) by the next roll station.
3.2
3.2.1
Theoretical work
The number of forming steps that a prole requires depends on material properties and the complexity of the prole. Other factors such as the part width,
horizontal centre distance between the passes and the part tolerances must also
be taken into account. However, a standard method to calculate the number
14
Literature Survey
of forming steps does not exist, Halmos [19], so below a number of dierent
approaches are described. A common method used to calculate the number of
forming steps is the forming angle method [3], see Figure 3.5. In the following
example the calculation of a simple section is described.
To determine the number of passes you take the height (g) of the nished
prole and the distance between the forming stands (d) and use the derived
equation 3.5 below. The method is based on the amount of forming per length.
That a conservative forming angle (C) is 1.5o can be kept in mind as rule of
thumb.
Figure 3.5. The forming angle method.
The following notation and equations 3.3 - 3.5 are used to calculate the number of forming steps:
n
c
g
=
=
=
=
=

The forming length
Forming angle
Horizontal distance between the forming steps
c = g cot .
(3.3)
15
c = (n 1)d.
(3.4)
3.3 and 3.4

g
cot + 1.
(3.5)
d
If the prole has multiple bends the forming angle method must be applied
to every bend independently and then combine the bends where it is possible
and approximate how many passes that are needed to nish the prole.
g cot = (n 1)d n =
Figure 3.6. The ower method: the engineer starts with a nished prole and
progressively unfolds it to a at strip.
In the ower method, [9] the tool engineer starts with a nished section and
gradually unfolds the prole to a at strip, see Figure 3.6. The way the prole
will be unfolded and the number of forming steps is up to the tool designer
to decide. However, a successful use of this method requires a great deal of
experience, and the method can be very time consuming when applied to more
complex shapes.
Figure 3.7. Front view and side view of the roll forming mill used in the Algorithm,
equation 3.6 to 3.11, [41].
A method to roughly estimate the machine design and number of forming step
for a prole is proposed in [41]. Figures 3.7 - 3.8 and the section below present
16
Literature Survey
the notations used to calculate the machine design.

HC
VC
LS
DS
H
BB
n
u
=
=
=
=
=
=
=
=
Horizontal distance between the forming stands

Vertical distance between the centre of upper and lower tools
Length of the shaft
Diameter of the shaft
Width of the at strip used to form the prole
Distance from the edge of the at strip to the edge
of the nished prole, see Figure 3.8.
The logic Algorithm, Equation 3.6 - 3.11 in [41] consists of the rules below
that should be fullled by a design. It is used to roughly estimate the machine
design.
LS 1.2BB .
DS
LS
min 30mm.
7
(3.6)
(3.7)
VC 1.5DS + 2H.
(3.8)
HC 2VC (pre-cut material).
(3.9)
HC 2.5VC (post-cut material).
(3.10)
n>
75u
.
HC
(3.11)
The drawback to this method is that thickness and yield strength of the material used are not taken into account.
To predict the number of forming steps Halmos [19] proposes an empirical
equation to be used as a guide. The equation 3.12 was not fully tested at the
publishing date of [19].
n = [0.237H 0.8 +
Y 2.1 0.15
0.834 d
][
+
] s(1 + 0.5z) + ej + fs + 5zsf . (3.12)
t0.87
90 0.003Us
17
Figure 3.8. u is the distance from the edge of a at strip to the edge of the nished
prole. In this case u is used to calculate the number of forming steps, equation 3.11.
where
n =
Y =
H =
t =
Us =
d =
sf =
fs =
z =
ej =
3.2.2
Number of forming steps.

Yield strength.
Height of the prole.
Thickness of the material.
Ultimate tensile strength.
Total degree of formed angle.
Shape factor.
Tolerance factor.
Pre-punched hole
Extra pass.
Deformation types
In a bending operation the major part of the deformation occurs in a transverse direction. In roll forming, which also is a bending operation although
the bending is made gradually, other deformation types besides deformation in
the transverse direction can occur. Paton et al. [45] propose four fundamental
deformation types: longitudinal stretching, longitudinal bending, transverse
bending and shear.
The following assumptions were made to derive the deformation types, see
Figure 3.9 and Figure 3.10:
The thickness of the strip is small compared to other geometrical dimensions.
Bending only takes place in the fold line of the active bend.
Deformation at the inboard side is neglected.
18
Literature Survey
The outboard region remains constant in the cross-section and rotates
around the active bend.
Transverse sections of the strip remain plane and the prole is bent as
a beam. The latter means that cross-sections remain orthogonal to the
centreline along the prole.
With help of the innitesimal element of length dz, see Figure 3.9 and 3.10,
the following strain models were derived:
Longitudinal stretching
1
d
Ls = r2 ( )2 .
2
dz
(3.13)
where (r) is the distance from the bend.

Longitudinal bending
Lb = h(s(
d2
d
p( )2 )).
2
dz
dz
(3.14)
where (h) is the distance from the neutral layer to the inner side of the bend,
(s) is parallel moment arm and (p) is the perpendicular moment arm.
Shear
= p(
d
).
dz
(3.15)
h
.
r0
(3.16)
Transverse bending
t =
where (r0 ) is the radius to the neutral layer.

The conclusion drawn by Paton et al. [45] was that the shear strain is larger
than the longitudinal strain in sections with inactive bends and must therefore
be considered as a factor in the roll forming process. The shear strain is small
for a simple section such as a U-prole without inactive bends.
19
Figure 3.9. An element strip between two roll forming passes.
Figure 3.10. View in z-direction.
3.2.3
Deformation length
Bhattacharyya et al. [5] derived an expression for the deformation length (L),
see Figure 3.3, by minimising the total plastic work (Wt ) for the bending in
transverse and longitudinal direction. They made the following assumptions:
The material is rigid perfectly plastic.
The bend only takes place along the fold line.
Out of plane bending of the ange and the longitudinal bending of the
web are neglected.
The ange adopts a shape that minimises the plastic work.
Plastic work due to transverse bending
20
Literature Survey
1
Wb = Y t2 .
4
(3.17)
where (Y) is the yield strength.

Plastic work due to longitudinal bending per unit volume
1
d
Ws = Y x2 ( )2 .
2
dz
where (x) is coordinate in the transverse direction.
Plastic work per unit length
a
1
d
1 2 d 2
Y x ( ) (tdx) = Y a3 t( )2 .
Wsl =
dz
6
dz
0 2
(3.18)
(3.19)
The total work done for one bend is

Wt =
1
1
d
[ Y t2 + Y a3 t( )2 ]dz.
4
6
dz
(3.20)
The function (z) that minimises the Equation 3.20 satises the Euler Equation 3.21
dF
d F
d2
3t
( ) 2 3 = 0.
d
dz
dz
4a
where F = F (z, , ) =
t
4
(3.21)
a3 d 2
6 ( dz )
the general solution is

(z) =
3t 2
z + Az + B.
8a3
(3.22)
The end conditions give the integration constants (A) and (B), (0) = 0,
(L) = , (0) = 0 and (L) = 0.
The end conditions and Equation 3.22 give the deformation length

8a3
.
L=
3t
(3.23)
21
where () is the bend angle increment, (t) is the thickness of the material
and (a) is the ange length.
3.2.4
Longitudinal membrane strain
Chiang [9] developed two models, (A) and (B) for the longitudinal membrane
strain. Model (A) was based on a publication by Bhattacharyya et al. [5],
equation 3.24, and a model (B) that Chiang derived by geometrical considerations, equation 3.27.
equation 3.13 and equation 3.22 give model (A)
e=
9 t2 2 2 0ra
( )r z |0zL .
32 a6
(3.24)
The peak strain and transverse strain be obtained from Equation 3.24 in model
(A). The peak strain for the ange edge, (r = a, z = L), is written:
3 t
ep = ( ).
4 a
(3.25)
The transverse longitudinal strain at any position z is written:

et = k1 r2 .
where k1 =
(3.26)
9 t2
2
32 ( a6 )z
Chiang compared model (A) with an experiment and concluded that the model
overestimated the strain three times when approaching the roll station.
Model (B) based on geometry for the engineering strain in the ange edge is
written:

e=
1+
3t
(1 cos ) 1.
4a
(3.27)
The model predicts that the strain is uniform in the deformation zone but gives
no information about the transverse strain. Compared with an experiment,
model (B) gave a closer approximation of the value of the peak strain than
model (A).
22
3.2.5
Literature Survey
Geometrical restriction from the female tool
Zhu [57] studied how the geometry of the tool inuences the bend angle in the
ange of a U-prole. The forming was divided into three stages, see Figure
3.11, the rst stage being when the bend angle does not change, stage two
when the bend angle changes but the strip is not in contact with the tool and
the third stage being when the strip is contact with the tool.
Figure 3.11. The bend angle is divided into three stages: stage one when the angle
does not change, stage two when the angle changes but the strip is not in contact
with the tool and stage three when the strip is contact with the tool.
Zhu studied both horizontal and vertical rolls, but it is the study of horizontal
tool that is discussed here. Zhu developed models applicable to three cases:
a) the roll fully overlaps the outer edge of the ange.
b) the outer edge of the ange overlaps the roll at any position.
c) the outer edge of the ange overlaps the roll initially.
Only case a) is presented below since in practise case b) and c) are not utilized.
The model obtained for case a) is

a2 cos2
sin(2 )
z =L
+ 2aR1
a2 .
(3.28)
cos2 2
cos 2
Zhu assumed that the derivative for equation 3.28 could be used to predict
forming severities at the point where the strip get in contact with the female
tool for the rst time. The derivative is
23
d
=
dz
Lz
a2
sin 2
2 cos2 2
2)
+ aR1 cos(
cos 2
(3.29)
The theoretical work was compared with the experiment and the conclusions
were as follows:
A concept, the bend angle curve, was proposed and there was a good
agreement between the predicted bend angle distribution and the experimental results.
The longitudinal membrane strain reaches a maximum when the strip
gets in contact with the female tool for the rst time.
An increased ange length will decrease the longitudinal membrane strain.
Increased tool radius will decrease the longitudinal membrane strain.
The longitudinal peak membrane strain increases with increasing bend
angle increment.
Constant bend angle increment and an increasing bend angle decreases
the longitudinal peak membrane strain. However, when the bend angle
is close to 90o the opposite eect will occur.
3.2.6
Roll forming of high strength steel
One project [54] involved the roll forming of both a V-prole made of high
strength steel, yield strength 1000 MPa, and a V-prole made of mild steel,
yield strength 250 MPa. In both cases the V-channel was roll formed in four
steps: 15o , 30o , 45o and 60o . The V-prole made of high strength steel was
straight after the forming whereas the prole made of mild steel was curved,
see Figure 3.12.
Ingvarsson [28] derived a simple formula, equation 3.30, based on geometry
for the longitudinal engineering strain in the ange of the V-channel used in
the project [54].
The simple formula was:

a2 sin2 Vc2 Vc
.
(3.30)
e=
Vc
where (Vc ) is the vertical distance between the centre of upper and lower tool.
In this case the deformation length (L) was simplied with (Vc ).
24
Literature Survey
Figure 3.12. The V-prole made of high strength steel and mild steel were roll
formed in four forming steps: 15o , 30o , 45o and 60o . The V-prole made of high
strength steel was straight after forming but the prole made of mild steel was not.
Ingvarsson calculated the longitudinal strain, equation 3.30 for the case with
the V-prole and compared it with the maximum elastic strain, equation 3.31,
for both mild and high strength steel used in the experiment [54]. The result
showed that in the case with mild steel the strain in the ange was larger than
the maximum elastic strain, and therefore both elastic and plastic strain were
present during the forming. In the case with high strength steel only elastic
strain was present. Plastic strain will give residual strain and in this case
the channel made of mild steel was curved. The conclusion drawn was that
it is advantageous to use high strength steel in roll forming since it will have
smaller plastic strain in the anges and as a result smaller residual strain as
well.
Y
.
(3.31)
E
where E is Youngs modulus and Y is the yield strength for the material.
eelastic =
Sagstr
om et al. [49] studied the inuence of roll forming on the tolerances of
pre-punched holes in a hat-prole made of dierent ultra high strength steels.
25
Furthermore, the amount of springback and are for these steels was observed.
They concluded by measurements that the holes maintained their shape and
that they could be pre-punched, with high tolerances requirements, close to the
ange edges before the roll forming started. Their investigation also showed
that the are in the end of a prole will be smaller with increasing yield
strength, see 3.13, and a parameter for this phenomenon called Delta Flare
was introduced. The denition of Delta Flare is the dierence between the
springback at the end of a prole and the springback in the centre of the same
prole. A suggested explanation to decreasing are was that increasing yield
strength gives less residual stresses that could cause are in the proles.
Figure 3.13. Delta Flare is the dierence between the springback in the centre of
the prole and the end of the prole. This is a plot for 4 dierent ultra high strength
steels with various yield strength.
Galdos et al. [14] studied the inuence of material properties on the roll
forming process when a U-prole was formed. The main objectives were to
investigate the longitudinal strain, forming force, springback and longitudinal
bow for steels ranging from mild to ultra high strength steels. Both nite
element simulations and experiments were used in the study. Their conclusion was that the maximum longitudinal peak strain decreases when the yield
strength increases. The experiment also showed that increasing yield strength
gives a slight, longitudinal bow, which possibly is the result of a smaller peak
strain in the prole.
3.2.7
Bending method
Two dierent proles are shown in Figure 3.14: one C-prole and one top-hat
prole. The C-prole can be roll formed mainly by bending, but forming the
top-hat prole requires both bending and drawing in the bending zone [13].
When bending is used to form a prole the spring back will be large and the
longitudinal residual stresses in the bending zone will be small. In the case
26
Literature Survey
when both bending and drawing are used to form the prole the spring back
will be smaller whereas the longitudinal residual stresses in the bending zone
will be larger, Ingvarsson [26], Lindgren [32]. In roll forming the eect of
drawing can be reduced depending on which bending method that is used,
see Figure 3.15. For example, if the most common method called constant
arc length is used, the thinning of the material will be reduced. This gives
less longitudinal residual stresses in the bending zone, than when the constant
radius-method is used. The latter method, on the other hand, gives more
stretch forming and thus a smaller spring back. Another method called the
variable radius and variable arc length- method, gives a large spring back but
small longitudinal residual stresses.
Figure 3.14. To the left a C-prole is roll formed in 8 passes and to the right a
top-hat prole is roll formed in 5 passes. The C-prole is formed mainly by bending
whereas the top-hat prole is formed both by bending and drawing [13].
3.3
Computer simulation of roll forming
Kiuchi et al. [30] introduced a normalized shape function, equation 3.32, to

describe a spatial locus of an element of the strip between two forming stands
(roll stand = i to roll stand = i + 1), see Figure 3.16.
U q
S(U ) = sin( (
) ).
2 HC
(3.32)
HC = (U2 U1 ).
(3.33)
where HC is the horizontal distance between two forming stands and U is the
coordinate in longitudinal direction, Figure 3.16
3.3 Computer simulation of roll forming
27
Figure 3.15. This gure shows three dierent ways of forming a 90o prole. One
way is to let the tools have the same radius in every forming step and dierent arc
length. Another way is to let the arc length be the same in every forming step but
decrease the tool radius. These methods give dierent spring back and longitudinal
residual stresses in the bending zone. The constant arc length is the method most
commonly used.
The shape function has a built in parameter (q, q > 1), i.e. a low value of
q gives a surface with a curvature that gradually increases when the strip is
approaching the forming tools whereas a larger value describes a curvature
that increases slowly in the beginning and faster just before the tools, see Figure 3.17. To determine the value of (q) a minimization of the total power of
deformation in the strip between the forming stands is done through a mathematical procedure.
To describe the deformed curved surface, the following equation is used:
U = U (u, v).
(3.34)
W = W1 (v) + [W2 (v) W1 (v)] S(U ).
(3.35)
V = V1 (v) + [V2 (v) V1 (v)] S(U ).
(3.36)
U = U1 U2 .
(3.37)
28
Literature Survey
Figure 3.16. The deformed sheet between roll stand (i) and roll stand (i + 1). The
neutral layer of the sheet is described by the shape function, Equation 3.32.
Figure 3.17. Shape functions for dierent parameters (q).
29
where
W1 , V1 = W and V coordinates of the cross section of the prole at the roll
stand (i) at U = U1
W2 , V2 = W and V coordinates of the cross section of the prole at the roll
stand (i + 1) at U = U2
Boundary conditions and Equation 3.35 and Equation 3.36 give:
When U = U1 , S(U ) = 0, V = V1 (v) and W = W1 (v)
When U = U2 , S(U ) = 1, V = V2 (v) and W = W2 (v)
The equations 3.34 - 3.37 are purely geometric descriptions of the 3D shape of
the deformed strip, which do not take into account the stresses. To calculate
stress and strain a steady-state deformation is assumed between roll stand (i)
and roll stand (i + 1). The strip is divided into a suitable number of elements
and a incremental theory of plasticity is used to calculated the stress and
strain in the strip. The details are given in Hallmos [19]. Kiuchi et al. [30]
developed a computed aided design system that includes the above mention
method. The system has successfully been used to design tools for circular
tubes.
Duggal et al. [12], [11] used the computer aided simulation program RFPASS in a study. The program is based on the mathematical model developed
by Kiuchi et al. [30] and a nite dierence method. The program uses an
elastic-plastic formulation and can be used for analysing multiple roll stations. In the study a simulation of a U-prole was compared with an experiment done by Bhattacharrya and Smith [4]. Duggal et al. concluded that
the peak longitudinal strain deviated less then 2 % between the simulation
and the experiment. The longitudinal residual strains obtained from the RFPASS dropped more then measured strains, which can be explained by the
assumption of a sinusoidal shape for the movement of the sheet. RFPASS can
be used, at least for simple sections, to assist the tool designer. Furthermore,
other surface models than the sinusoidal shape function will be investigated
in future studies.
Han et al. [22] developed a B3 -spline nite strip method to simulate the roll
forming process. The displacement of the strip was modelled with two types
of shape functions. In the longitudinal direction the strip was modelled with
B3 -spline functions and in the transverse direction it was modelled with Hermitian cubic polynomials. Han et al. [22] introduced the updated-Lagrangian
method of the nite element theory to the nite strip method. The model
used accounted for large deformation strain and displacement. Prandtl-Reuss
plasticity ow theory and the von Mises yield criteria were also applied. Han
30
Literature Survey
et al. have written several articles, for example [20] and [21], where they describe how the method was used to analyse simple sections and parameters
that eect the design. The conclusion drawn was that the results from the
simulations agree with observations of roll forming in practise.
Rebelo et al. [47] compared the relative eectiveness of implicit and explicit
nite element analysis in metal forming. The simulated U-prole was modelled with 4-node shells - 20 through the width and 40 through the length. The
strip was pulled through three roll stations and the rigid rolls rotated freely.
The material was modelled as an elastic-plastic material with a yield strength
of 229 MPa and with Youngs moduls of 206.7 GPa. The implicit nite element analysis was almost three times faster (47 CPU hours) than the explicit
analysis (125 CPU hours). The conclusion was the implicit formulation has a
relative advantage, since the problem is rather one dimensional and therefore
has a small wave front.
Brunet et al. [7] developed a master 2D cross-section model with a slave
3D analysis. The 2D analysis was a generalised plane-strain analysis and for
the 3D analysis a thick shell element was used. The tools were modelled as
rigid surfaces and instead of rotating tools they were modelled as rigid surfaces
continuously moving from one forming station to the next station. The friction
between the sheet and the tools was modelled as Coulomb friction. Both the
3D and the 2D analyses included Hills anisotropic model of initial anisotropy
with isotropic hardening. The computed longitudinal deection was compared
with the measured deection in order to validate the model. Brunet et al.
concluded that there was a reasonable dierence between the model and the
experiment.
Heislitz et al. [23] used the explicit code PAM-STAMP to simulate roll
forming. The strip was pulled through the rolls with constant speed. The
rolls were not rotating and the friction between the strip and the rolls was
ignored. Two dierent elements were tried: 8-node brick elements and four
node shell elements. In the nal simulations, however, they used 8-node brick
element. The tools were modelled with rigid 4-node shell elements. The mass
density was increased with a factor 100 without inertia eects aecting the
result. The material model used was Swifts isotropic strain hardening and
Hookes law. The simulations were compared with an experiment and the
maximum deviation was about 10%. The conclusions from the simulations
were as follows:
The adaptive mesh renement can help to speed up the simulation.
At the current status of development, the simulation of roll forming by
using PAM-STAMP is not very ecient due to the required CPU time.
The simulation time for a U-channel was 250 CPU hours.
31
FEM code PAM-STAMP can be used to accurately simulate roll forming. It is possible to produce both the strain distribution and the nal
geometry after spring back.
Sukmoo et al. [53] used the nite element program (COPRA FEA-RF) to
simulate the roll forming of a U-prole and compared the deformation length
with an experiment described in literature. The FE-program COPRA FEARF is a rigid-plastic nite element analysis that uses a combined 2D and a
3D algorithm. One conclusion from the study was that the work hardening
exponent has the most signicant eect on the forming length. Increasing
work hardening exponent gives increasing forming length. Sukmoo et al. also
concluded that it was a good agreement between the simulation results and
the experiment.
Alsamhan et al. [2] utilised the FE-code EPFEP3, which is a 3D implicit
elastic-plastic FE program, to developed a remeshing technique for simulating
roll forming. A dual mesh was used: one for storing the deformation history
and another for FE computational. The conclusion drawn from the simulation
was that the computational time was much less compared to a conventional
FE-model. The result from the simulation was also acceptable compared to
computations without remeshing.
Sheu [52] used LS-DYNA explicit nite element code to simulate a roll
formed U-channel. The study comprised the inuence of friction, corner radius, rolling speed and changing rate of bend angle on the ange angle deviation and the ange length deviation of a 90o U-prole. In the simulation
the front end of the strip was xed and the tools were moved towards the
strip with a given velocity curve. To minimise the number of trails, in this
case 9 trail runs, a reduced factorial design L9 (34 ) was used. The conclusion
drawn from this study was that the bend angle changing rate and rolling speed
are important factors for the bend angle deviation. As concerns the length
deviations all factors are equally important.
Salmani et al. [50] used ABAQUS implicit FE-package to investigate localized edge buckling in a symmetric channel section when dierent bend angles
were used. Here only the model itself will be described. The tools were modelled as rigid surfaces and they rotated to drive the strip forward. A friction
coecient of 0.2 was assumed. The strip was modelled with a four node
shell element and ve integration points through the thickness were used. An
elastic-plastic material model was implemented as a table in the software. A
comparison was made between the model and an experiment done by Bhattacharyya and Smith [4]. The conclusion was that the deformation history in
the ange edge of the prole was similar to the deformation history in experiment and the error between the membrane peak strain ranged from 0 to 12
% depending on the bend angle.
32
Literature Survey
Bui and Ponthot [8] used the in-house nite element code Metafor to simulate a roll formed U-prole and compared the result with experiments in
literature. The strip was modelled with a 8-noded brick element and with
renement of the mesh in the bending zone. The tools were modelled as rigid
surfaces. The mechanical behaviour of the material was described by Swift
isotropic strain hardening law. The speed of the rolls and the friction between
them and the sheet were studied. Coulomb friction = 0 and = 0.2 were
tested. The result showed that the friction did not inuence the springback
and the longitudinal strain to any great extent, and the speed inuenced the
result only moderately. Bui and Ponthot concluded that the nite element
models can be simplied especially for complex proles, for example by ignoring the friction and thereby saving computer simulation time. They also
concluded that the study conrmed the potential of the nite element simulation in designing the roll forming process.
3.4
3D roll forming, proles with variable crosssection
In 2001 ORTIC AB [1] developed a new roll forming technology, 3D roll forming, which made it possible to roll form proles with variable cross-sections.
Since then the technology has been used to produce panels that cover buildings
world wide, Ingvarsson [29].
In [33] Lindgren describes how to produce a C-prole with 3D roll forming
technology. Figure 3.18 shows two cross-sections for a C-prole with variable
depth and width and the corresponding blank of sheet metal. The automatic
control system controls the tools so they are always perpendicular to the tool
paths. The tool path is a mathematical description of the bending line which
is generated from the cross-section in the longitudinal direction of the channel.
One way to roll form the C-prole according to Figure 3.18 is to start forming
along the outer bending line and then continue along the inner bending line.
Groche et al. [16] integrated a single exible roll forming stand with a
traditional roll forming line. The objective was to produce U-proles with
variable cross-sections. The stand had tools that could rotate and translate
with the help of servomotors. The control of the motors depended on the feed
rate of the prole which was measured by an encoder. The tools followed a
bending line created in a CAD-System which the control program could read.
In [44] Ona describes a single exible roll forming stand and a slitter roll
stand. The units were controlled with a CPU. The tools and the slitter could
rotate and translate in and out. The equipment was used for producing Uproles with variable cross-sections in the longitudinal direction. The experiments showed that the material in the ange were compressed or stretched,
see Figure 3.19, if the ange was too high.
3.4 3D roll forming, proles with variable cross-section
33
Figure 3.18. To the left: the unfolded C-prole, the blank of sheet metal. To the
right: two dierent cross-sections of the C-prole.
Figure 3.19. A 3D roll-formed U-prole. Tension stress acts on the transition zone
where the U-prole is small and compression stress acts on the transition zone where
the prole is wider.
Groche et al. [17] also developed a semi-empirical model for designing 3D roll
formed U-channels. The model was based on mechanics of buckling of plates
and nite element analyses. The model focused on the compressed area in the
ange of a U-prole with variable cross-section, see Figure 3.19, and it can be
used for feasibility checks without simulation or experimental tests.
34
Literature Survey
G
uleceken et al. [18] used COPRA RF [15] coupled with the nite element
module MSC.Marc to simulate the 3D roll-forming process. The main objective was to study how the exible roll forming process can be modelled and
simulated. The simulated U-channel was modelled with a number of 1274 full
integrated 8-noded hexahedral volume elements. An elastic-plastic material
model where the swift extended power law describes the material hardening
was used. In addition, Von Misses yield criteria and the associated ow rule
were applied. The surface of the tools was modelled using rigid shells. To
control the rotating tools the load control option in the MSC.Marc was used.
The simulation time was only 3-4 CPU hours since the model was relatively
rough. The result showed that the dimensions of the simulated part were close
to the dimensions of the desired prole.
Chapter 4
Experimental setup
The experimental equipment developed and built for the work presented in
the appended papers 4 to 7, is described below.
4.1
The roll forming experiment
Section 4.1 describes the experimental equipment used in the study presented
in paper 4 and 5. The aim of these experiments was to measure strain history,
roll load, roll torque and springback during the roll forming of a U-channel
and to compare the results with nite element simulations.
4.1.1
The roll forming machine and data acquisition equipment
A universal, spindle type roll-forming mill, Eckhardt [13], with six roll forming
stands and universal joint driven shafts, was used in the experiment, see Figure
4.1. Only the lower shafts were driven. The mill has a three-phase frequency
controlled motor. The horizontal distance between the roll forming stands
is 450 mm and, in this experiment, the vertical distance was 150 mm plus
material thickness. The rst and last forming stands were used as feeder rolls.
The data acquisition equipment came from NATIONAL INSTRUMENTT M
and the software used for measurement was LabVIEWT M .
4.1.2
The tools
The investigated prole was a U-channel where the forming steps had the bend
angles: 20o , 40o , 60o and 80o , see Figure 4.2. The bending method used to
form the prole was the constant arc length method, Chaing [9], which means
that the radius of the neutral layer of the bend decreases in every forming step.
For the bend angles listed above the radii of the neutral layer of material with
35
36
Experimental setup
Figure 4.1. A universal, spindle type roll-forming mill where the lower tools are
driven with universal joint driven shafts.
a thickness of 1mm are 19.8, 9.9, 6.6 and 4.95 mm respectively. The tools
were designed for materials with a thickness of 1 mm but they were used for
materials with thicknesses up to 1.47 mm when roll torque and roll load were
measured. The results imply that the material in the ange will be squeezed
when the bend angle increases. To avoid this phenomenon, with material
thicker than 1 mm the upper tool for a bend angle of 80o was also used as an
upper tool for a bend angle of 60o . The roll pitch radius for the upper and
lower tools was 75 mm.
The lower tools in the two rst forming steps had guides, see Figure 4.2, to
align the material with the machine. These guides inuence the longitudinal
and transversal strain in the strip.
The surface nish of the tools is RZ = 1m, and an abundance of lubrication was applied to have a similar low friction coecient between the dierent
materials and the tools.
4.1.3
Strain measurement
Strain gages were used [25] to measure the biaxial strain history on the top
surface of the ange during the forming process. The strain gage was a
45o /90o /135o rosette, relative to longitudinal direction, with a measuring grid
of 1.5 mm. It was bonded close to the edge of the ange and the grid was
allocated about 2.5 mm from the edge, see Figure 4.3.
4.1 The roll forming experiment
37
Figure 4.2. Geometry of the tools used in the experiment. The tools have a bend
angle of 20o , 40o , 60o and 80o .
Figure 4.3. This gure shows only half of the U-channel because the symmetry and
the rosette strain gage is bonded on the top surface close to the edge of one ange.
The longitudinal direction is the same as x-direction and the transverse direction is
same as y-direction.
The grids, 45o /90o /135o , were placed with 2.4 mm in longitudinal direction
(x-direction) between each, to make them follow the same path during the
forming process. The position of the sheet was measured with an encoder,
and with the help of the position the strain measurement results from the
dierent grids were adjusted in time to make them measure the same point in
the process.
The strain in longitudinal and transverse direction was determined with
strain transformation, Dahlberg [10], equation (4.1).
38
Experimental setup
() = x cos2 () + y sin2 () + xy cos() sin().
(4.1)
The equation (4.1) and the measurements in 45o /90o /135o give the equation
(4.2) and the equation (4.3) for the longitudinal and transverse strain.
4.1.4
x(longitudinal) = 45 + 135 90 .
(4.2)
y(transverse) = 90 .
(4.3)
Roll load and roll torque
The roll load was measured with two donut load cells from Honeywell Sensotec
Sensors. They were mounted in the forming stands on both sides of the upper
tool, see Figure 4.4, and were able to measure a total roll load of 17800 N in
z-direction, Figure 4.3.
Figure 4.4. The roll load in z-direction was measured with two donut cells. The
torque sensor was used to measure the roll torque and it was mounted between the
lower tool and power transmission. The equipment measured the forces of each forming step, one by one.
4.1 The roll forming experiment
39
The roll torque was measured with a torque sensor from Burster, which can
measure up to 75 Nm. It was mounted in the universal joint shaft between
the lower tool and the power transmission, see Figure 4.4.
The measurement of the load and torque was made in one forming step at
a time. For example the bend angle 20o was measured when the rst feeder
roll and the roll for a bend angle of 40o also were engaged with the strip, for
measurement of 40o are the rolls for bend angle 20o and 60o engaged at the
same time.
The load and torque are dependent on the tolerance between the thickness
of the roll formed material and the roll gap between upper and lower tools.
The roll gap was adjusted to be 10 % to 15 % wider than the thickness of the
material in order to avoid the material to be squeezed between tools causing
very high load and torque. The roll gap was checked with a feeler gage. This
method also takes into account shaft deection and bearing slackness.
4.1.5
Spring back
The prole will experience some spring back after each forming step and then
deform more and more as it approaches the next step. However, between
springback and deformation, the prole will have a shape that does not change
provided the distance between the forming steps is large enough. In the present
experiments this occurred about 100 mm to 150 mm before and after the
forming step depending on material properties and geometry conditions of the
prole.
The width (y-direction) and depth (z-direction) of the prole were measured, see Figure 4.5, between the forming steps in the area where the shape
does not change. This was done with a sliding caliper, and a measurement
tolerance of about 0.1 mm.
Figure 4.5. The spring back, width and depth between the dierent forming steps
were measured with a sliding caliper.
40
4.2
Experimental setup
Partial heating experiment
Section 4.2 describes the experimental equipment used in the study described
in paper 6, Lindgren et al. [38]. The aim of this study was to partially
heat up the bending area on cold rolled AISI 301 stainless steel and then roll
form the strip. The heating reduced the amount of martensite in this area.
The plastic forming increases the martensite fraction again restoring the yield
strength of the material. The material properties in the bending zone were
then investigated before and after the roll forming process and compared with
heat treated samples from furnace experiments.
4.2.1
Furnace experiment
Samples, with a size of 0.7 15 84 mm, of the cold rolled stainless steel:
AISI 301, were heated in an electrical batch furnace and then quenched in
water. The heat treatment times were 15, 30, 60, 90 and 120 s, respectively,
with temperatures increasing from 500o C to 1200o C, in steps of 100o C each.
Figure 4.6. The picture shows the prototype resistance heating machine and the
steel strip which was fed through the machine by wheels made of copper.
4.2 Partial heating experiment
4.2.2
41
Prototype resistance heating machine
The prototype resistance heating machine and a sketch of the electrical circuit are shown in Figure 4.6 and Figure 4.7. The machine consists of a programmable logical control (P LC) system where the rotation speed of the heating wheels can be adjusted. The power used to heat up the material was controlled by a thyristor that was controlled by a trigger signal from the PLC.
The transformer increases the current in the material. The voltage over the
material can be varied between 5 and 7 V (voltage) by adjusting the transformer. For all experiments 7 V, a speed of 1 m/min and a contact pressure of
5 Bar between the wheels and the sheet were used. The diameter of the wheels
was 122.5 mm and the width of the contact zone was 4 mm. Three dierent
powers called P 10, P 20 and P 30 adjusted by the thyristor were investigated;
the results described Paper 6, Lindgren et al. [38].
Figure 4.7. A sketch of the electrical circuit of the prototype resistance heating
machine. Wikstr
om [56].
4.2.3
Roll forming machine and tools
The roll forming machine in the experiment was the same as that described in
section 4.1.1, see Figure 4.1. A V-section was roll formed in six forming steps:
15o , 30o , 60o , 90o , 105o and 120o . The prole had an inner radius of 0.4 mm
and the material was 0.7 mm thick, see Figure 4.8. The forming velocity was
3.5 m/min and the forming load about 500 N. The diameter of the upper tools
was 140 mm and of the lower tools 100 mm. The tools were designed with the
constant radius method , Chaing [9], which means all forming steps had the
same inner radius.
42
Experimental setup
Figure 4.8. The tools and the ower pattern. The V-section was roll formed in 6
forming steps, from 15o to 120o .
4.2.4
Sample preparation
Cross-sections of all samples from the furnace heat treatment as well as partially heated samples from before and after the roll forming, were examined
with light optical microscope and with Vickers microhardness tests with an
applied load of 500 g. The samples were hot mounted using a thermosetting resin followed by mechanical grinding and polishing with 1 m diamond
suspension. To reveal the microstructure of the samples from the furnace experiment, an etching solution of 60 ml deionised water, 62 ml concentrated
HCl, 3 ml concentrated HN O3 and 5 ml concentrated H2 SO4 was applied.
The etching time was approximately 60 s in room temperature. To etch the
heated zone of the partially heated samples, a solution consisting of 25 ml
concentrated HCl and 5 ml 10 wt% chrome acid solution in deionised water
was used. In these cases etching time was 5-10 s.
4.3
3D Roll forming experiment
Section 4.3 describes the experimental equipment used in the study presented
in paper 7, Lindgren and Ingmarsson [39]. The aim of the study was to
investigate a new tooling concept for a production of hat-channels with variable
cross-sections.
4.3.1
The 3D roll forming machine
An experimental 3D roll forming machine was built in this study, see Figure
4.9. The machine was equipped with slitter heads to t the metal sheet,
and six forming stands with four units each, see Figure 4.10. The units have
4.3 3D Roll forming experiment
43
servo control axis, two translations and two rotations axis. A unit can be
moved up and down, in and out and rotate, and the speed of the tools can
be controlled individually. To control the servo-axis the position of the sheet
metal is measured by encoders.
The desired cross-section (depth and width) can be created in a CADprogram and read by the control system. Depending on the tooling the machine can produce for example U-proles, C-proles, hat-proles etc. with
variable cross-sections.
Figure 4.9. The experimental equipment used for research and prototyping of 3D
roll formed proles.
4.3.2
The tooling concept
In the study described in paper 7, a hat-section with variable depth and width
was investigated and a special tooling concept was used. This concept demands
two forming stands per pass, see Figure 4.11. Forming stands number 1, 3 and
5 form the left side of a channel and forming stands number 2, 4, and 6 the
right side. For example, a channel is roll formed in six forming steps with
the bend angles 15o , 30o , 45o , 60o , 75o and 85o and therefore the section
demands two laps in the machine. The rst lap forms 15o - 45o , and then the
control program of the machine switches to the next lap, 60o - 85o .
The tools are the same for all stands, for example Figure 4.12 shows that
44
Experimental setup
Figure 4.10. The experimental equipment has 6 forming stands. Each stand has 4
units (tool holders) that can rotate and translate.
Figure 4.11. View from the top: The prole is roll formed in six passes and to do
that the prole must go through the machine two times. Forming stands number 1,
3 and 5 form the left side and forming stands number 2, 4, and 6 form the right side.
the tools for the left and right side are the same for both the 30o and 60o . The
dierence between the passes is that the tools that hold the ange have moved
up closer to the tools that hold the web. The ower pattern is a function of
the length of the section. This exibility makes it possible to use as many
passes as the cross-section requires without making more tools. It also makes
it possible to produce a section not only of variable width but also of variable
4.3 3D Roll forming experiment
45
depth, in material of varying thickness and with dierent properties.
Figure 4.12. View from the back. The geometry of the tools is the same for all
forming stands. The only dierence is that the tools for the ange are moving up and
closer to the tools that hold the web, for example bend angles 30o and 60o . Tools
with a constant radius have been used, Chaing [9].
4.3.3
Measuring equipment
An optical scanner based on the principle of triangulation was used to measure

the channels. The accuracy of the measurement equipment and the xture
were tested and showed a variation less than 0.07 mm.
A xture positioned the hat-prole at the bottom, z-direction, with 3 reference points and 1 support point. The distance between the points in the
longitudinal direction was 1300 mm. In y-direction the section was positioned
with 2 reference points and 2 support points and in the x-direction one reference point was used, see Figure 4.13.
46
Experimental setup
Figure 4.13. The xture used.
Chapter 5
Computational model
Details about the nite element models described in paper 1-3 and paper 5
are presented in the following chapter. The roll forming process was modelled
with the implicit FE package MARC/MENTAT [43], see Figure 5.1. Large
displacement and large additive strains were taken into account in an UpdatedLagrangian approach. Nonlinear plastic material behaviour as well as contacts
with friction contributed to the nonlinearities in the models.
Figure 5.1. The nite element model described in paper 5, Lindgren [37].
47
48
5.1
Computational model
Nonlinear solution procedure and convergence

criteria
To solve the nonlinear equilibrium equations the full Newton-Raphson method

was used. Convergence is assumed when the relative unbalance force norm
becomes less than 0.05 or if the relative displacement norm is less than 0.01.
The analysis was performed with a direct sparse solver. The models described
in the appended papers dier with regard to the time step. For the model in
paper 5 a constant time step of 0.006 s was used whereas for the other models
the constant time step was between 0.002 and 0.008 s.
5.2
The geometry
The geometry of the model in paper 5

The geometry of the roll forming tools, see Figure 5.1 and 5.2, was generated
with the ORTIC CAE system [27], and the model comprised four forming
steps. The tools were modelled as analytically described rigid surfaces that
rotated with an angular speed of 0.75 rad/s which gave the prole a speed of
3.375 m/min. The horizontal distance between the tools was 450 mm. Length
and width of the unformed strip were 950 mm and 41.9 mm respectively. The
number of elements was 3840 and the size of the elements was 4 1 mm to
4 2 mm in the bending zone. In other areas the element size was 4 5 mm.
Five layers of integration points were used in the thickness direction
Figure 5.2. The geometry of the tools and the corresponding ower pattern described
in paper 5.
5.2 The geometry
49
The geometry of the model in the paper 2 and 3

The nite element model consisted of two forming stands and two stands where
no forming took place, see Figure 5.3. The latter were used as a belt feeder to
the other two, where the forming was done. The element size was 4 1 mm
in the bending zone and in other areas the element size was 4 4 mm. Three
layers of integration points were used in the thickness direction. The number
of thick shell elements ranged from 1600 to 2400. Due to its symmetry only
half the geometry had to be modelled. The initial angular speed of the tools
was 10 rad/s, which gave the strip a speed of 0.6 m/s.
Figure 5.3. Four forming stands were used. The two rst stands were used as a belt
feeder, the other two forming stands have the same bend angle.
The geometry of the model in the paper 1

This model comprised six forming stands and two stands where no forming
took place. The latter were used as a belt feeder to the six forming stands.
The forming began with an undeformed strip and ended with the nished
symmetrical U-prole. The distance between each of the forming stands was
330 mm.
The strip was 1.502 mm thick; the width was 31.8 mm and the length
1000 mm. The strip was modelled by the use of 2750 thick shell elements.
The element size was 4 1 mm in the bending zone whereas in other areas it
was 4 4 mm. Three layers of integration points were used in the thickness
direction. Due to its symmetry, only one half of the geometry was modelled.
The initial angular speed of the tools was 10 rad/s which gave the strip a
speed of 0.6 m/s.
50
5.3
Computational model
Shell element
The strip was modelled with a bilinear thick shell element type number 75,
MARC elements Library [42]. This is a four node element that calculates
membrane strains and the curvatures at the middle surface. The element can
be used in a curved shell analysis and thanks to the simple formulation the
element is not so expensive. A nite element model, with 3, 5, 7, 9 and 11
integrations points in the thickness direction, was evaluated for the work in
paper 5. It was found that at least ve integration points must be used to
predict the springback decently. The integrations points were spaced equally
throughout the thickness with one point both on the top and bottom surface.
To calculate the stress and strain MARC uses numerical integration through
the thickness.
5.4
Material model
An elastic-plastic material model with the von Mises yield criterion and isotropic
hardening was assumed together with the associated ow rule. The yield criterion is written as

3
f = y =
sij sij y .
(5.1)
2
where sij is the deviatoric Cauchy stress giving the eective stress . The
deformation is elastic as long as this is lower than the yield limit y .
The associated ow rule is written as
pij =
=
.
ij
sij
(5.2)
where is a plastic parameter to be determined by the consistency condition

that f 0 during plastic ow.
The tensile test data for the materials used in paper 5, see Figure 5.4. The
stress versus plastic strain was implemented as a table in the nite element
program. The work hardening slope is described in Marc with a piecewise linear representation in a table. Heating was ignored in all models and therefore
the ow stress at room temperature was sucient. In the studies presented in
paper 1-3 material data were implemented in same way as described above.
5.5 Contact
51
Figure 5.4. The tensile test data for the materials used in paper 5.
5.5
Contact
The friction between the prole and the tools was modelled as a type of
Coulomb friction. MARC uses
Ft = Fn
where
vr
C
Fn
Ft
=
=
=
=
=
vr
2
arctan( ).
(5.3)
Sliding velocity
Relative sliding velocity, in the study is C = 0.01
Normal force
Tangential force
Friction coecient, in the study = 0.085
The tools were modelled as rigid bodies and the material was deformable.
MARC uses a method called direct constraint method for this contact situation. During the procedure, the node seldom hits the surface of the rigid body
exactly. In order to make this situation easier a contact tolerance around the
rigid surface is given. MARC detects contact if a node in the deformable material comes inside this contact tolerance area (D) and it is default 25 % of
the thickness of the smallest shell element used, see Figure 5.5. The contact
tolerance is also default symmetrically (D1 = D2) placed around the surface
of the rigid body. Greater tolerance gives less accuracy when calculating and
52
Computational model
a smaller tolerance means that MARC will have more problems discovering
contact, which causes higher computational costs. In this thesis a tolerance
of 4 - 5 % of the shell thickness was used. As a result of rotating tools, many
nodes are in almost contact in the roll forming process. This means that many
contact separations can occur thereby causing an extended calculation time.
In order to avoid this it is possible to displace the tolerance distance closer to
the surface of the rigid body using a bias factor which means that D1 < D2.
The default is a bias factor of 0 which means that D1 = D2 but in this thesis
a bias factor of 0.25 was used.
Figure 5.5. The used tolerance distance D1 and D2 in paper 5 is 5 % of the thickness.
The bias factor is default 0 in MARC which means that D1 = D2. In paper 5 is a
bias factor of 0.25 used which means D1 < D2.
Chapter 6
Summary of appended papers

6.1
Paper 1
Finite element model of roll forming of a U-channel prole

This article describes the design of a nite element model for simulating roll
forming of a U-channel prole, whereafter the results from one simulation are
compared with previous knowledge on the process as described in literature.
The simulation starts with an undeformed strip and ends with the nished
U-channel. The model includes friction and the tools used are rigid rotating
surfaces, which is an improvement to other models in literature. The strip
was modelled with thick shell elements, which take into account transverse
shear stress. An elastic-plastic material model was used, the hardening was
isotropic. The von Mises yield surface was used and the associated ow rule
was applied.
The conclusion from the simulation is that this model of the roll forming
process works very well, but whereas the simulation time is acceptable for
research purposes it is still too long for industrial use. The results obtained
agree well with existing knowledge about the process, and the model can be
used to further investigate important parameters. Thereby, the model can be
used to improve existing design rules.
6.2
Paper 2
Cold roll forming of a U-channel made of high strength steel

Paper 2 looks into changes of the longitudinal peak membrane strain at the
edge of the ange and of the deformation length when yield strength increases.
This was done by nite element analyses using the simulation model presented
in paper 1. The results show that longitudinal peak membrane strain decreases
and deformation length increases when the yield strength is increased.
53
54
The conclusion is that it is advantageous to use high strength steel in roll

forming since it will make the longitudinal peak membrane strain decrease.
Less strain will give less residual stresses in the ange of the prole, and quality
problems such as wave edges, longitudinal curvature, end are and so on, will
decrease. The simulation results of this study can explain why Ingvarsson [28]
obtained a straight V-section when roll forming ultra high strength steel and
a curved prole when mild steel was used. However, this is inconsistent to
some existing design rules stating that the yield strength of the material does
not aect the ange strain in roll forming.
6.3
Paper 3
An improved model for the longitudinal peak strain in the ange of

a roll formed U-channel developed by FE-analyses
In this study factorial design and nite element analysis were combined to
investigate which parameters aect the longitudinal peak membrane strain
and the deformation length. The results were used to create simple models
which predict the strain and the length. The models were also compared with
the result from paper 1 and paper 2 and they agree well. The conclusion is
that these simple models can easily be implemented in a CAE system and
improve the design of U-channel roll forming lines. The same approach can
also be used to develop design rules for forming other proles.
6.4
Paper 4
Experimental investigation of the roll load and roll torque when

high strength steel is roll formed
Paper 4 presents an experimental investigation of roll loads and roll torque
when mild steel and high strength steel are roll formed. The roll load was
measured with load cells and the roll torque with a torque sensor mounted
between the lower tool and the power transmission.
The results obtained will increase the understanding of the specic conditions for roll forming steel with increasing yield strength. They can be used to
validate nite element models and support the machine designer in the choice
of power unit and dimensioning the shaft on which the tools are mounted.
The conclusion from this study is that load and torque increase with t2.1
(t = thickness of the material) and Y 0.57 (Y = yield strength of the material).
The power needed in a forming step for production speed (30 40 m/min) is
about 0.5 kW when conditions are similar to those in the study.
6.5 Paper 5
6.5
55
Paper 5
Validation of nite element simulation of roll forming

Finite element simulation is commonly used for studying sheet metal forming
but not in the case of roll forming. In this study the nite element model of
roll forming (paper 1) was validated with experiments. Measurements of roll
loads, roll torque, geometry and strain history in the ange of a U-channel
were compared with computed results.
The agreement between measurements and simulations were good except
for the spring-back, which was larger in the experiments than in the simulation.
The conclusion is that nite element simulations will increasingly be used
to analyse roll forming but it will not replace the computer aided engineering
(CAE) programs in the future. Today a CAE program can be used to give an
initial proposal to how many forming steps a prole needs. The program also
proposes the design of the steps. Still it requires a toolmaker with experience
to decide if the proposed tool design is good enough. Here is where the nite
element simulations can support the toolmaker before taking a decision for
manufacture the tools.
6.6
Paper 6
Roll forming of partially heated cold rolled stainless steel

Today you can nd roll formed details in many dierent products, such as
buildings, household appliances and vehicles. The industry tends to use more
and more high strength steel in order to save weight. The disadvantage of
these materials is that they can be dicult to form due to reduced ductility.
Partially heating the steel is one way of increasing ductility in the forming
areas.
This study shows that partial heating substantially increases the ductility
of high strength steel and makes it possible to roll form large bend angles.
When roll forming, the material will work hardening almost to the as-received
condition in the outer and inner radius of the bending zone. The neutral
layer will work hardening moderately. Furthermore, the heating power decides
what bend angle will be obtained. Finally, the partially heated material shows
three dierent regions: one unaected region, one transition region and nally
a recrystallized region. Recrystallization starts in the centre and grows with
higher power input.
56
6.7
Paper 7
3D roll forming of hat-prole with variable depth and width

The use of roll-formed products in the automotive sector, in furniture and
buildings and so on, increases every year due to the low part-production cost
and the complicated cross-sections that can be produced today. Until recent
years one disadvantage of roll forming has been that only proles with constant cross-sections in the longitudinal direction could be made. About eight
years ago, however, ORTIC AB developed a machine in which it was possible
to produce proles with variable cross-sections (3D roll-forming) for the building industry. Experimental equipment was recently built for researching and
prototyping proles with variable cross-sections for the automotive industry.
The objective of the study presented here was to investigate a new tooling
concept that makes it possible to roll-form hat-sections, made of ultra high
strength steel, with cross sections variable in depth and width. The results
show that, under conditions similar to those in the study, the new tooling
concept makes it is possible to produce 3D roll-formed sections with close
tolerances.
Probably one of the rst 3D roll-forming machines in the world, with two
translational and two rotational degrees of freedom per axis, was built and
used in the study. Based on a specic tooling concept, hat-proles with crosssections variable in depth and width were produced using simple, cylindrical
shaped rolls.
The main conclusions are as follows:
The new tooling concept makes it possible to roll-form hat-proles with
cross-sections variable both in depth and width, and with tolerances at
the same level as for straight proles.
The tools make it possible to roll-form material of dierent thickness by
software changes due to their simple, cylindrical shape. The ange of
the hat-prole must be at least 10 mm to avoid that it slips in the tools,
thereby causing variations in width.
Three dierent types of proles have been identied in the study: type
1 which are completely straight, type 2 which are conical proles and
type 3 which are proles with transitions zones. The rst two types
give similar residual stresses in the longitudinal direction and they are
easy to produce. Type 3 proles get residual stresses that can give wavy
edges as well as distortion of the web and ange. This kind of proles
therefore requires a carefully designed process.
Chapter 7
Discussions and conclusions

The experimental and numerical work described in this thesis started with
simple geometry, a U-prole, and standard steel. Initial experiments and modelling progressed to investigating roll forming of high strength steels, which
ultimately led to the design of a machine for 3D roll forming, see paper 7,
Lindgren and Ingmarsson [39], in thesis.
The research was based on the following question:
The subsequent discussion, including the main conclusions from this work,
aims to answer the question posed above.
Modelling the roll forming process
There is a limited understanding of the roll forming process and available science on the process is not extensive. A plausible reason is that the complex
geometry of the process has made it dicult to develop simple models of how
the strip is formed. The literature contains only few analytical models for
strip deformation that can be used for computing the number of forming steps
and proposing tool design, Lindgren [35]. Simplied simulations programs,
ORTIC System [27], PROFIL [55], for evaluating tool design, based on numerical methods such as the nite dierence method or the B-spline nite
strip method, have previously been developed. The drawback to this type of
simulations is that they do not include the eect of lead-in anges and the
end eects on the prole when pre-cut material is used. However, the present
studies as described in papers 1 and 5, show that it is possible to use nite element simulations to design roll forming processes with accurate results. This
kind of models were less common in 2003 when the work started.
57
58
A prole can be made in up to 40 forming steps. In earlier years, a simulation

of the process could take as long as 10 days, which made it impractical for
the industry to use. Since then the computer capacity has increased, and still
is increasing, immensely, which means that proles that were impossible to
simulate a few years ago now are possible to handle. For example, the simulation time of the model described in paper 1, Lindgren [34], was 34 h in 2004.
Two years later a simulation with the model described in paper 5, Lindgren
[37], was made in a new and cheaper computer, in about half the time. In
addition, other research articles, Bui and Ponthot [8], have shown that the
friction between tool and sheet can be ignored without losing much information from the simulation. Thereby the simulation time can be reduced even
further. The main conclusions in regard to modelling roll forming processes
are that in the present work a suitable technique has been established, and
that fairly complex processes can be simulated in reasonable time. The results
from the simulations, plastic strains, etc. will also be of interest for reuse in
subsequent analyses of the in-service behaviour of the formed component. As
described in papers 2, Lindgren [35], and 3 , Lindgren [36], nite element simulations can also be used for numerical experiments to create simple formulas
that are quick to use in the early design phase of a roll forming process. For
example Groche et al. [17] developed a semi empirical model with the help of
nite element simulation, for quick use in feasible studies aimed to nd out if
a radius change in the width is possible when a U-prole is 3D roll formed.
There is reason to expect that nite element simulation will become an
indispensable tool in the design of roll forming processes. Some CAE-systems,
for example COPRA RF [15], have simplied the modelling of traditional roll
forming so that tool geometry, position and motion are automatically generated. As concerns 3D roll forming it is vital that this part is developed further
since for the process requires extensive work to describe the position and the
motion of the tools.
Roll forming of complex geometries and advanced materials
In order to save weight, the industry tends to use more high strength steels.
Problems related to forming high strength steels are their low ductility, increasing springback, increasing forming force, among other things. However,
the investigation described in Lindgren [35], VAMP [54] and Sagstr
om [49],
shows that roll forming high strength steel has some advantages since the
residual stresses will be smaller in the formed prole with the result of less
end are, distortion of holes in the prole, wavy edge etc. For other forming
processes the increased springback can be a problem, but in roll forming good
tool design can take care of that.
It has been observed that with roll forming high strength steel can be
formed to smaller minimum bend radius than with bending. The main dier-
59
ence between the two forming operations is that in bending plain strain can
be expected whereas in roll forming there is a more three dimensional strainstate. Lundberg and Melander [40] have compared air bending to roll forming
with the help of nite element simulations. As a part of their study a damage
criterion was proposed, which exhibited a higher damage factor for air bending than for roll forming. They concluded that this could explain why high
strength steel can be roll formed but not air bent to the same small minimum
bending radius.
One way of getting around such problems as the minimum bend radius or
the increased forming force and springback is to use partially heated material. The results presented in paper 6, Lindgren et al. [38], show that it is
possible to substantially increase the formability by using austenitic stainless
steel, e.g. AISI 301, in a cold rolled condition and increase the ductility in
the forming areas by partial heating. Furthermore, the heated areas will be
deformation hardened and regain some of the strength during the roll forming.
It was steel that had an initial high fraction of martensite. The heated region,
that became austenitic, regained the martensitic structure during the plastic
straining. Paper 2, Lindgren [35] shows that the longitudinal strain decreases
when the yield strength increases. It is true that the springback also will increase, but this can be reduced by partial heating so that the strip can easily
enter the next forming step. The results from these two studies indicate that
it would be possible to roll form a prole with just a few forming steps without
introducing residual strain, for example in the ange of a V- or U-prole, that
could cause quality problems.
The 3D roll forming process has great potential when combining its exibility with the advantages of traditional roll forming. With only one set of
tools it is possible to produce proles with varying cross sections in a highly
productive process including high material utilization and low tooling costs.
Research concerning traditional roll forming is limited compared to many other
research areas. However, research on 3D roll forming is still in its cradle. Beside the things we need to know when designing a traditional roll forming
machine, such as the number of forming steps, the tool design and springback,
it is equally important to know the limitations of radius changes in both depth
and width when designing a 3D roll forming machine or a prole. In order to
start investigating these factors a 3D roll forming machine was built within
this research, which proved it possible to produce hat-proles of variable depth
and width with a new tooling concept, see paper 7, Lindgren and Ingmarsson
[39]. Thanks to the simple shape of the new tools it is also possible to roll
form a prole of varying thickness without changing the tools. This tooling
concept has been patented by ORTIC [1]. Furthermore, two dierent proles were identied in the study: conical proles and proles with transitions
zones. Conical proles get residual stresses in the longitudinal direction sim-
60
ilar to those in straight proles and they are easy to produce. Proles with
transitions zones get residual stresses that can give wavy edges and distortion
of the web and ange, which means that a particularly careful design of the
process is required.
Conclusion
In order to answer the research question initially posed, this study, as it progressed, was divided in three parts: modelling, high strength steels and 3D
roll forming. However, when designing a roll forming process these research
areas go hand in hand. On the basis of research of others and from the present
work, the following overall conclusions can be drawn:
Today, it is possible to successfully develop proles of complex geometries (3D
roll forming) in high strength steels and nite element simulations can be a
useful tool in the design of the roll forming process.
Listed below are some suggestions about future work in the above mentioned
areas in order to further improve the competitiveness of the roll forming process and enable the development of new, advanced products:
Develop nite element models for the 3D roll forming process.
Develop material models of advanced high strength steels.
Investigate, through simulations, the potential of partially heating materials. The aim is to enable a production of proles in high strength steel,
in a process based on small bending radius and fewer forming steps.
Establish, through simulations and experiments, the minimum radius
in width and depth that 3D roll formed proles can have, without any
quality problems arising. Alternatively with the same techniques nd
out how these problems can be dealt with.
Bibliography
[1] ORTIC AB. www.ortic.se. internet.
[2] A. Alsamhan, I. Pillinger, and P. Hartely. The development of real time
re-meshing technique for simulating cold roll forming using fe methods.
Journal of Materials Processing Technology, 2004. Article in Press.
[3] Corporate author. Tool and Manufacturing Engineers Handbook, volume 2. SME, Dearborn, Michigan, 4 edition, 1984.
[4] D. Bhattacharyya and Smith P. D. The development of longitudinal strain
in cold roll forming and its inuence on product straightness. Advanced
Techology of Plasticity, 1:422427, 1984.
[5] D. Bhattacharyya, P. D. Smith, and I. F. Collins. The prediction of
deformation length in cold roll forming. Journal of Mechanical Working
Technology, 9(2):181191, Mars 1984.
[6] D. Bhattacharyya, P.D Smith, S. K. Thadakamalla, and Collins L. F.
The prediction of roll load in cold roll forming. Journal of Mechanical
Working Technology, 14(3):363379, June 1987.
[7] M. Brunet, B. Lay, and P. Pol. Computed aided design of roll forming of
channel sections. Journal of Materials Processing Technology, 60:209214,
1996.
[8] Q.V. Bui and J.P. Ponthot. Numerical simulation of the cold roll-forming
processes. Journal of Materials Processing Technology, 202:275282, 2008.
[9] K.F. Chaing. Cold roll forming. Masters thesis, Aukland, Aukland, New
Zeeland, 1984.
[10] T. Dahlberg. Teknisk H
allfasthetsl
ara. Studentlitteratur, 2 edition, 1997.
[11] N. Duggal. Process simulation of roll forming and roll pass design. Masters thesis, Ohio state University, 1995.
61
[12] N. Duggal, M. A. Ahmetoglu, G. L. Kinzel, and T. Altan. Computed

aided simulation of cold roll forming - a computer program for simple
section proles. Journal of Materials Processing Technology, 59:4148,
1996.
[13] T.K. Eckhardt. Cold roll forming, theory and application. Bulletin
709/872, page 24.
[14] L. Galdos, J. Larranaga, L. Uncilla, H. Lete, and G. Arrixabalaga. Process
simulation and experimental tests of cold roll forming of a u-channel made
of dierent ultra high strength steel. In rollform09. 1st international
congress on roll forming, 2009.
[15] Data M Sheet Metal Solution GmbH. COPRA RF. internet, 2009.
[16] P. Groche, G. von Breitenbach, M. J
ockel, and A Zettler. New tooling
concepts for future roll forming applications. In Proc. ICIT 2003,Celje,
Slovenia, 2003.
[17] P. Groche, A. Zettler, and S. Berner. Development of a one-step-model
for the design of exible roll-formed parts. In The 9th International Conference on Material Forming; Glasgow, United Kingdom, April 2006.
[18] E. Gulceken, A. Abee, A. Sedlmaier, and H Livatyali. Finite element
simulation of exible roll forming: A case study on variable width uchannel. In 4th International Conference and Exhibition on Desing and
Production of MACHINES and DIES/MOLDS, Cesme, TURKEY,, 2007.
[19] G.T. Halmos, editor. Roll Forming Handbook. Number 67 in Forming and
Forging. Talylor and Francis Group, New York, USA, 1 edition, 2006.
[20] Z. W. Han, C. Liu, W. P. Lu, and L. Q. Ren. Simulation of a multistand roll forming process for thick channel section. Journal of Materials
Processing Technology, 127:382387, 2002.
[21] Z. W. Han, C. Liu, W. P. Lu, L. Q. Ren, and J. Tong. Spline nite strip
analysis of forming parameters in roll forming a channel section. Journal
of Materials Processing Technology, 2004. Article in Press.
[22] Z. W. Han, L. Q. Ren, C. Liu, and W. P. Lu. Modelling of cold roll forming
of steel strip. Material Science and Technology, 17:415418, April 2001.
[23] F. Heislitz, H. Livatyali, M. Ahmetoglu, G. L. Kinzel, and A. Taylan.
Simulation of roll forming process with the 3-d fem code pam-stamp.
Journal of Materials Processing Technology, 59:5967, 1996.
62
[24] S. Hellborg. Finite element simulation of roll forming. Masters thesis,

Link
oping University, Link
oping, Sweden, 2007.
[25] K. Homan. An Introduction to Measurments Using Strain Gages. Hottinger Baldwin Messtechnik GmbH, Darmstadt.
[26] L. Ingvarsson. Cold-forming Residual Stresses and Box Columns Built
Up By Two Cold-formed Channel Sections Weleded Together. Phd thesis 121, Institutionen f
or byggnadsstatik, Kungliga tekniska h
ogskolan,
Stockholm, Sweden, 1977.
[27] L. Ingvarsson. Rullformningsteknik - datorprogram f
or verktygskonstruktion. Verkst
aderna, 1983.
[28] L. Ingvarsson. F
orenklad teori f
or rullformning av element
ar v-prol,
j
amf
orelse mellan normalt och h
ogh
allfast st
al. www.ortic.se, April 2001.
[29] L. Ingvarsson. Innovativa st
alprodukter-praktikfall baserad p
a rullformningstekniken. In Proceeding St
al 2004, 2004.
[30] M. Kiuchi and Koudabashi T. Automated design system of optimal roll
proles for cold roll forming. In Proceedings of the 3th International
Conference on Rotary Metal Working Process, pages 423436, 1984.
[31] D. Kolev. Roll forming - general overview. KOLEV Engineering Inc.
[32] M. Lindgren. Residual stresses in longitudinal direction in a bending
operation. Presented at 4th ICFG Workshop in process simulation in
metal forming industry, Shanghai China, May 2004.
[33] M. Lindgren. Rullformning av osymetriska sectioner. In Proceeding St
al
2004, 2004.
[34] M. Lindgren. Finite element model of roll forming of a u-channel. In
Advanced technology of plasticity. ICTP, 2005.
[35] M. Lindgren. Cold roll forming of a u-channel made of high strength steel.
Journal of Materials Processing Technology, 186:7781, 2007.
[36] M. Lindgren. An improved model for the longitudinal peak strain in the
ange of a roll formed u-channel developed by fe-analyses. Steel research
international, 78(1):8287, 2007.
[37] M. Lindgren. Validation of nite element model of roll forming. In Best
in Class Stamping, pages 397406. IDDRG, June 2008.
63
[38] M. Lindgren, U. Bexell, and L. Wikstr

om. Roll forming of partially heated
stainless steel. Journal of Materials Processing Technology, 209:3117
3124, 2009.
[39] M. Lindgren and L.O. Ingmarsson. 3D roll forming of hat-proles with
variable depth and width. 1st internatinal congress in rollforming, RollFORM09, 2009.
[40] M. Lundberg and A. Melander. Finite element analysis of roll forming of
uhss compared to traditional bending. In Best in Class Stamping, pages
331338. IDDRG, June 2008.
[41] H. Lundh, B. Carlsson, G. Engberg, L. Gustafsson, and R. Lidgren. Formningshandboken-Stycksk
arande bearbetning och plastiskformning. SSAB Tunnpl
at AB, Borl
ange, Sweden, 1 edition, 1997.
[42] MSC.MARC. Element libary, volume b. User documentation, 2007.
[43] MSC.MARC. Msc software. internet, 2009.
[44] H. Ona. Study on development of intelligent roll forming machine. In
Proceeding of 8th ICTP. ICTP, October 2005.
[45] S. M. Panton, S. D. Zhu, and J. L. Duncan. Fundamental deformation
types and sectional properties in roll forming. International Journal of
Mechancial Science, 36(8):725735, August 1994.
[46] Proform. An innovative manufacture process concept for a exible and
cost eective production of the vehicle body in white: Prole forming.
internet, 2009.
[47] N. Rebelo, J. C. Nagtegaal, L. M. Taylor, and R. Passman. Comparsion
of implicit and explict nite element methods in the simulation of metal
forming processes. In Proceeding of the ABAQUS User Conference, 1992.
[48] E. Sagstr
om, M. Lindgren, R. Moshfegh, and T. Jansson. 3D-roll forming
of ultra high strength steel proles with varying cross-sections. Poster at
the International Deep-Drawing Research Group, Olofstr
om 2008, Sweden, June 2008.
[49] E. Sagstr
om, B. Wadman, R. Moshfegh, M. Lundberg, and A. Melander.
Geometrical robustness of roll formed proles in ultra high strength steels.
In Best in Class Stamping, pages 339348. IDDRG, June 2008.
[50] M. T. Salmani, H. N. Moslemi, P. Hartley, and H. Khademizdeh. Localized edge buckling in cold roll forming of symmetric channel section.
Thin-walled Structures, 44:184196, 2006.
64
[51] R.S. Senanayake, I.M. Cole, and S. Thiruvarudchelvan. The application of

computational and experimental techniques to metal deformation in cold
roll forming. Journal of Materials Processing Technology, 45(1-4):155
160, September 1994.
[52] J. Sheu. Simulation and optimization of the cold roll forming process. In
Simulations and Applications. NUMIFORM, 2004.
[53] H. Sukmoo, L. Seungyoon, and K. Naksoo. A parametric study on forming length in roll forming. Journal of Materials Processing Technology,
113:774778, 2001.
[54] SWEREA-IVF. Vamp 15, Nov 2001. Gothenburg, Sweden.
[55] UBECO. Prol. internet, 2009.
[56] L. Wikstr
om. Partial annealing of stainless steel before roll forming.
Masters thesis, Dalarna University, Dalarna, Sweden, 2007.
[57] S. Zhu. Theoretical and experimental analysis of roll forming. Phd thesis,
Ackland, Ackland, New Zeeland, 1993.
65
PAPER 1
Finite element model of roll forming of a U-channel profile
Michael Lindgren
Advanced Technology of Plasticity (2005)

Proceedings of the 8th ICTP, Oct. 9-13, 2005. Verona, Italy
FINITE ELEMENT MODEL OF ROLL FORMING OF A UCHANNEL PROFILE

M. Lindgren
Dalarna University Sweden
Summary
Cold roll-forming (CRF) is a highly productive process and therefore an interesting metal forming
process. CRF products can be found in many applications like buildings, furniture and vehicles.
However, it is a geometric complex process and therefore the knowledge that has been obtained
through simple models is limited. The use of FE models to simulate sheet metal forming processes
in general is common but not in case of CRF. Finite element models can be used to enhance the
understanding of the process and as design tools. They can also be used to create simpler design
rules.
The objectives of this study are to create a model that can be used to predict the longitudinal
membrane strain in the flange and to analyse the contact between the tools and the strip. This strain
is important as it determines the number of forming steps needed to form the profile.
The created finite element model accounts for friction and elasto-plastic deformations of the strip.
The rotating tools are assumed to be rigid. The strip is modelled with thick shell elements and the
simulations start from an undeformed strip to a finished U-channel.
Keywords: Cold roll forming, Finite element analysis, Computer simulation
Introduction
In cold roll forming (CRF) the sheet is continuously and progressively formed in several forming
stands from sheet metal to a finished cross section, Figure 1. The number of forming steps needed
to obtain a wanted cross-section is the first choice when designing a CRF machine. They depend on
the wanted cross-section, thickness and the material. Other important parameters are spring back,
deformation length and longitudinal membrane strain in the flange. The longitudinal membrane
strain develops as the profile is successively formed. The material in the edge of flange will travel a
longer distance 'L than the material in the bending zone between the forming stands, Figure 2.
This strain should not be plastic in order to avoid wave edge or another defect on the finished
profile. If plastics strains are present, then more forming stands are needed.
Today there are several computer programs available on the market that can support the design of
CRF machines. The programs are based on thumbs of rules and simplified formulas. However,
only limited conclusions can be drawn based on these programs due to the complex geometry of the
formed strip and the simplifications in their design rules. The finite element method can be used to
increase the knowledge of the roll forming process.
M. Lindgren
L + 'L
Rolling direction
Flange
Figure 1. The strip is successively formed in

several steps, from an undeformed strip to a
profile, from [1].
Figure 2. The material in the edge of the flange

will travel a longer way 'L than the material in
the bending zone.
Previous work
Rebelo et al. [2] studied implicit and explicit finite element formulation used in metal forming
process simulations. The advantage of explicit formulation is that the analysis cost increases in
direct proportion to the size of the mesh. Whereas in the implicit formulation it increases with the
square of the wave front times the degrees of freedom. They concluded that in roll forming the
wave front is small due to the problem being very one dimensional. Therefore the simulation was
faster with implicit formulation when CRF was simulated.
McClure and Li [3] simulated roll forming with the ABAQUS software and compared the result
with experiments from Bhattacharrya and Smith [4]. The model ignored friction between the tools
and a horizontal force was applied in the leading edge to pull the material through the roll stations.
The obtained membrane strain was similar to the experiments by the latter.
Heislitz et al. [5] used the code PAM-STAMP to simulate the roll forming process. The strip was
pulled through the rolls with a constant speed, without friction and without rotating tools. The
explicit FEM code was used. They tried both eight node brick element and shell element and they
also concluded that re-meshing reduced the simulation time when they used shell elements. They
concluded that at the current status of development, the simulation of roll forming by using PAMSTAMP is not very efficient due to the required CPU.
Brunet et al. [6] tried a specially developed FEM code, PROFIL. It used a master 2D cross-section
analysis with a slave 3D shell analysis between two or four successive roll stands. The tools were
modelled as rigid surfaces and they were not rotating. The forming of one thin channel, one thick
channel and a circular tube were simulated. The calculated peak strain in the flange overestimated
the experimental values with 10-30%.
Alsamhan et al. [7,8] have developed a re-meshing technique to simulate CRF. The result from the
simulation showed that re-meshing can reduce the computer simulation time.
2
M. Lindgren
Finite element model
In this study a CRF process has been modelled with the FE package MARC/MENTAT [9].
3.1 The geometry
The model consists of six forming stands and two stands where no forming occurs. The latter are
used as a belt feeder to the six forming stands. The forming starts from an undeformed strip and
ends with a finished symmetrical U channel, Figure 3. The distance between each of the forming
stands is 330 mm.
To simulate the CRF process the geometry was generated with the ORTIC CAE system [10]. The
tools are then modelled as rigid surfaces in the FE program.
The strip is 1.502 mm thick, the width is 31.8 mm and the length is 1000 mm. To obtain a steady
state condition three rolls shall be engaged in the same time, therefore the chosen length of the
model. Due to the symmetry, only one half of the geometry is modelled.
The strip is modelled with 2750 thick shell elements. The elements size is 4*1 mm in the bending
zone and in other regions the elements size is 4*4 mm.
The strip is modelled with a bilinear thick shell element type number 75 [9]. This is a four node
element that calculates membrane strains and the curvatures at the middle surface. The element can
also account for transverse shear strains. The element can be used in curved shell analysis and due
to the simple formulation the element is not so expensive [9]. Therefore is the shell element 75
attractive for the CRF simulations. Three layers of integration points are used in the thickness
direction.
The initial angular speed of the tools is 10 rad/s which gives the strip a speed of 0.6 m/s. The speed
is increased by 0.5% in every forming stand in order to counteract buckling between the stands.
Dogal 350 YP
550
Stress
500
450
400
350
300
0
10
15
Plastic strain %
Figure 3. Flower pattern for the used

geometry, six forming stands are used. The
forming steps are from 0 - 90 degrees in
steps of 15 degrees.
Figure 4. Stress - strain curve for Dogal 350 YP.
M. Lindgren
3.2 Material model

An elastic-plastic material model where the material hardening is isotropic and the von Mises yield
surface and the associated flow rule is used. The tensile test data for the used material is shown in
Figure 4. The material is cold forming steel, Dogal 350 YP. Stress versus plastic strain for the
material is implementing as a table in the FE program.
3.3 Contact
In roll forming the tools are lubricated to minimise the wear on the tools and the strip. The
lubrication type is boundary lubrication. This is assumed to correspond to a friction coefficient of
0.1. The friction model is Coulomb friction.
The clearance between the tools is 1.5 mm and the strip is 1.502 mm thick. That gives a contact
pressure of 280 MPa.
3.4 Reduction of CPU time
Two solvers were tested, a direct sparse solver and a multifrontal sparse solver with bandwidth
optimisation metis. The direct sparse solver is the better one. The computer simulation time was
reduced by 30% with this solver.
The rolling contact between the tools is small in roll forming which leads to a number of iterations
to find contact. Separation between the tools and the strip also occurs regularly. These contact
situations make the required CPU time sensitive to the choice of the contact tolerance. It also
depends on the tool radius and the element size. After trial and error a value of about 4% of the
thickness was found to be appropriate.
Result and discussions
The simulation has been performed on a 2 GHz computer with 512 Mb RAM. Thirty-four CPU
hours were needed to finish the simulation above.
The longitudinal membrane strains in the edge of the strip increase between the stands then drop
just before the contact between the upper and lower tools, Figure 5. This result is similar to earlier
papers that have been written about FE simulations of CRF [3], [5].
Longitudinal membrane strain vs displacement in longitudinal direction
Longitudinal membrane strain [%]
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
-0.05
0
0.5
1.5
2.5
Displacement in longitudinal direction [m]
Figure 5. The longitudinal membrane elastic - plastic strain in the edge of the strip
in six forming stands. Constant forming steps are used (15 - 30 - 45 - 60 - 75 - 90).
4
M. Lindgren
When the strip is in contact with the lower tool for the first time the strain will drop from a
maximum value and compress until the contact zone between the upper and lower tools is reached.
This is in agreement with what Bhattacharyya and Smith observed [4].
It has been known that the tool radius influences the membrane strains in the flange. A smaller
radius gives a more severe forming as the material will not roll smoothly into the tool. Zhu [11]
obtained a model for the lower tool geometry and proposed a bend angle curve for the lower tool.
The derivative of the bend angle curve at the point where the strip is in contact with the lower tool
for the first time, shows how severe the forming is.
The tool radius is included in his model and it predicts more severe forming when the lower tool
radius decreases. Zhu also concluded that the peak strains will decrease when constant forming
steps are used. This agrees with the result from the simulation in this study, Figure 5.
Conclusion
A model of the CRF process, which includes friction and rotating tools, has been successfully
implemented. The simulation time is acceptable for research but still too long for industrial use.
The results from the simulations agree well with existing knowledge about the process. The model
can be used to further investigate important parameters as flange length, thickness of the material,
Youngs modules, yield strength and contact conditions. Thereby, the model can be used to improve
existing design rules.
Acknowledgements
The author thanks ORTIC AB, Knowledge foundation, Jernkontoret and Dalarna University for
their technical and financial support.
Reference
[1]
[2]
Lars Ingvarsson, Rullprofilering teknik och ekonomi, Aktuell produktion nr 3 1991 39 - 42

N. Rebelo, J.C. Nagtegaal, L.M. Taylor, R. Passman, Comparison of implicit and explicit
finite element methods in the simulation of metal forming processes,
ABAQUS User
Conf., Newport, RI, 1992
[3] C.K. McClure, H. Li, Roll forming simulation using finite element analysis,
Manuf. Rev.
8 (1995) 114.
[4] D. Bhattacharyya, P.D. Smith, The Development of Longitudinal Strain in Cold
Roll
Forming and its Influence on Product Straightness, First Inter. Conf. on Tech. of Plasticity,
Tokyo, The Japan Soc. For Tech. of Plasticity, 422-427, 1984.
[5] F. Heislitz, H. Livatyali, M.A. Ahmetoglu, Simulation of roll forming process with the 3-D
FEM code PAM-STAMP, J. Mater. Process. Technol. 59, 1996, 59-67
[6] M. Brunet, S. Mguil, P. Pol, Modelling of a roll forming process with a combined 2D and
3D FEM code, J. Mater. Process. Technol. 80-81 (1998) 213 219
[7] A. Alsamhan, P. Hartely, I. Pillinger, The computer simulation of cold-roll-forming using FE
methods and applied real time re-meshing techniques, J. Mater. Process. Technol., 2003
[8] A. Alsamhan, I. Pillinger, P. Hartely, The development of real time re-meshing
technique
for simulating cold-roll-forming using FE methods, J. Mater. Process.
Technol. 2004
[9] MARC, Element Library Volume B, Marc Analysis Research Corporation,USA
[10] ORTIC CAE SYSTEM, Datoriserad rullformnings - teknik, Srtryck, Verkstderna nr 1 1983
[11] S.D. Zhu, Theoretical and experimental analysis of roll forming, PhD thesis, Feb.
1993,
Auckland New Zealand
PAPER 2
Michael Lindgren
Errata
Page 1 (77)
Notation in the text for bend
angle
Notation in the text for
degree
replace the text

mlg@du.se
should be
mld@du.se
Journal of Materials Processing Technology 186 (2007) 7781

M. Lindgren
Dalarna University, Sweden
Received 16 June 2005; received in revised form 3 November 2006; accepted 6 December 2006
Abstract
Cold roll forming is a bending process where the bending occurs gradually in several forming steps from an undeformed strip to a nished
prole. The process is very interesting for the sheet metal industry due to the high speed in which the prole can be produced. High strength steel
has, in recent years, become more common in cold roll forming. These materials have advantages but also disadvantages that affect the design of
the process.
Simple models in literature [K.F. Chiang, Cold roll forming, ME Thesis, University of Auckland, August 1984] predict that the longitudinal
peak membrane strain in the ange of a prole is independent of the material properties. However, Ingvarsson [L. Ingvarsson, Forenklad teori
for rullforming av elementar v-prol, jamforelse mellan normalt och hoghallfast stal, VAMP 15- rullforming 23 april 2001] compared mild and
ultra high strength in a roll forming experiment and the conclusion was that the material properties will affect the nished prole. This paper is a
fundamental study performed in order to understand the observation by Ingvarsson [L. Ingvarsson, Forenklad teori for rullforming av elementar
v-prol, jamforelse mellan normalt och hoghallfast stal, VAMP 15- rullforming 23 april 2001].
The objectives of this study are to investigate the change in the longitudinal peak membrane strain at the ange edge and the deformation length
when the yield strength increases. These are important since they can be used to determine the number of forming steps and the distance between
them when designing the cold roll forming machine. The result from the simulations show that the longitudinal peak membrane strain decreases
and the deformation length increases when the yield strength is increased.
2007 Published by Elsevier B.V.
Keywords: Cold roll forming; High strength steel; Finite element analysis
1. Introduction
In cold roll forming a prole is formed in several forming
steps from an undeformed strip to a nished prole (Fig. 1). The
forming process is geometrically complicated due to the fact that
the forming does not only occur in the tools but also between
each forming stand. When creating the tools the tool designer
must decide how many forming steps the prole demands. The
number of steps is dependent on the shape of the cross-section,
tolerance, thickness and the material properties.
It is important to minimise the number of steps as this reduces
the cost of the cold roll forming machine. Then the process can
be a competitive alternative also for smaller production volumes.
Therefore the knowledge of how high strength steel affects the
number of forming steps is important.
Existing relations between the longitudinal peak membrane
strain, deformation length and the yield strength of the mate-
E-mail address: mlg@du.se.

0924-0136/$ see front matter 2007 Published by Elsevier B.V.
doi:10.1016/j.jmatprotec.2006.12.017
rial have been investigated and compared with nite element

analysis in this study.
2. Notations
The notations are given below and in Fig. 2:
L, deformation length;
a, ange length;
t, thickness of the strip;
Y, bend angle;
r, distance from the bending zone;
z, distance from where the bending starts;
e, the longitudinal membrane engineering strain at the ange.
3. Background
The prole is formed in several steps and that will cause longitudinal strain in the ange. The strain develops as the material
in a ange of a prole will travel a longer distance than the
78
M. Lindgren / Journal of Materials Processing Technology 186 (2007) 7781
Chiang [3] derived a model for the longitudinal engineering

strain in the ange based on minimising of the plastic work
due to stretching and bending of the prole. It is written as:
e=
9
32
t2
a6
0<r<a
r 2 z2
(2)
0<z<L
The model overestimated the strain when the strip approached

the tool. Therefore Chiang derived an improved expression by
a geometrical consideration for the peak strain, leading to:

a 2
e= 1+2
(1 cos ) 1
(3)
L
Fig. 1. The prole is formed in several forming stands from an undeformed strip
to a nished prole.
material in the bending zone. Panton et al. [1] concluded that

the longitudinal peak strain occurs when the strip is in contact
with the rolls for the rst time. The peak strain should not be
plastic as plastic strain will give a residual stress that causes
defects on the prole as wave edges, longitudinal curvature, etc.
Bhattacharyya et al. [2] created a model of the deformation
length by minimising the total plastic work for a U-channel. The
obtained model predicts that the deformation length is independent of yield strength. It is written as:

8a3
(1)
L=
3t
This model showed that the longitudinal strain is uniform in

the deformation zone. All models predict that the behaviour is
independent of the material properties.
Ingvarsson [4] compared mild steel with ultra high strength
steel in an experiment where a V-section was cold roll formed
with the same number of forming stands. The ultra high strength
steel gave a straight prole, but not the mild steel. The conclusion
was that fewer forming stands could be used when ultra high
strength steel is being roll formed.
4. Approach
Finite element simulations are used in the current study to
evaluate the yield strength inuence on peak strain and deformation length. Since the analytical formulas for longitudinal
peak membrane strain do not account for the yield strength.
Several papers have been written about the nite element
simulation on cold roll forming for example [57].
Fig. 2. One half of the U-channel. The strip is formed by a bend angle (Y8) in the contact zone between the rolls. It will cause a longitudinal membrane strain (e) in
the ange.
79
Table 1
Series of experiments where the forming steps are 08 (feeder roll)108108 and
08 (feeder roll)208208, when seven different yield strengths are used
200 MPa
400 MPa
600 MPa
800 MPa
1000 MPa
1200 MPa
1400 MPa
08108108
08208208
5. The nite element model

Fourteen different simulations have been carried out with
varying yield strengths and bend angles (Table 1). The nite
element package MARC/MENTAT was used to perform the
simulations.
The simulation starts from an undeformed strip and stops
when the material reaches the second forming step. The total-,
plastic- and elastic longitudinal peak membrane strain at the
edge of the ange, are evaluated.
5.1. The geometry
The model consists of four forming stands where the rst
two stands are used as a belt feeder to the others (Fig. 3). The
length of the strip is 800 mm so that the forming steps 08 (feeder
rolls)108108 and 08 (feeder rolls)208208 are engaged at
the same time when the simulation has come to an end. An
evaluation of the longitudinal peak membrane strain and the
deformation length is then performed for the rst forming step
where bending occurs.
The cross-section geometry of the strip is, width 40 mm,
ange length (a) 10 mm, bending radius 3 mm and the thickness (t) of the strip is 1.5 mm. Due to the symmetry only half of
the strip is modelled.
Fig. 3. Four forming stands are used. The two rst stands are used as a belt
feeder. The other two forming stands have the same bend angle (Y).
The strip is modelled with 1600 thick shell elements. The

mesh size is 4 mm 1 mm in the bending zone and in other
regions the mesh size is 4 mm 4 mm.
The strip is modelled with a bilinear thick shell element type
number 75 [8]. This is a four-node element that calculates the
membrane strain in the middle surface. Three layers of integration points are used in the thickness direction.
The rolls are modelled as rigid surfaces and they rotate, giving
the strip an initial speed of 0.6 m/s. The speed is then increased
by 0.5% in each forming stand to counteract buckling.
5.2. Contact
The clearance between the tools is 1.5 mm and the strip thickness is 1.504 mm, which gives a contact pressure of 560 MPa.
The friction is modelled as Coulomb friction and the friction
coefcient is 0.1.
Fig. 4. The ctive tensile test data for seven different materials. They have been implemented in the nite element program as a table.
80
Fig. 5. When the virgin yield strength increases the longitudinal peak membrane
strain will decrease. When the plastic strain goes to zero the total strain will
atten out and the strain is purely elastic.
5.3. Material model

The material is modelled as an elasto-plastic material. The
material hardening is isotropic and the von Mises yield surface
and the associated ow rule are used. A ctive tensile test data is
implemented in the nite element program as a table. The yield
strength is then scaled so the material starts to yield in seven different points from 200 to 1400 MPa (Fig. 4). The model account
for large deformations and strains. An additive decomposition
of total strain rate into elastic and plastic strain rates is assumed.
6. Result and discussion
Fig. 5 shows the peak longitudinal membrane strain of the
edge due to bending from an originally at strip to a 108 angle
for the different simulations. The result for each simulation is
denoted by the virgin yield strength of the ow stress in Fig. 4
used in respective simulation. This value is given on the horizontal axis in Fig. 5. The simulations show for a bend angle of
108 that the total (plastic and elastic) strain decreases when the
virgin yield strength increases. The total strain decreases more
in the beginning when plastic strain is present.
Fig. 6 is the bend angle 208 and the behaviour is similar as
in Fig. 5. The difference between the cases is that the strain
becomes purely elastic at higher material virgin yield strength.
The strain is also greater for a bending angle of 208, which
agrees with Chiang [3].
The deformation length, Eq. (1), is the distance between
where the transverse bending starts and the forming stands. In
this study a strain based deformation length is used. It is dened
as the distance from the forming stand to the point in the ange
edge where the strain is greater than 2e5. The strain based
deformation length for the cases 108 and 208 will increase when
the yield strength increases (Fig. 7). The length is greater for the
Fig. 6. The behaviour for the longitudinal peak membrane strain is similar to
the case with a forming step of 108. But now is the curve for the elastic and
plastic longitudinal peak membrane strain displaced to a higher level of virgin
yield strength.
bend angle 208 than 108, which agrees with Bhattacharyya et

al. [2].
Figs. 7 and 8 are the simulation results for the peak strain
and the deformation length compared with the models, Eqs. (1)
and (3), that Bhattacharyya et al. and Chiang derived. One can
see that they do not agree due to the large simplications in the
assumed pattern and material behaviour in their models [2,3].
The assumptions for the deformation length were:
The material is rigid perfectly plastic.
Bending takes place only along the fold-line.
The longitudinal bending of the web and the out-of-plane
bending of the ange can be neglected.
The ange adopts the shape that minimises the plastic work.
Fig. 7. The deformation length for bend angle 108 and 208 will increase when
the yield strength increases. The length is greater for bend angle 208.
81
the explanation why Ingvarsson [4] obtained a straight V-prole

when ultra high strength steel was used. Ingvarsson [4] used the
same number of forming steps for both the mild and the ultra
high strength steel. For mild steels were the forming steps too
few and larger residual stresses were present after the forming.
The latter gave a V-prole with a longitudinal curvature.
Increasing deformation length leads to the horizontal distance
between the forming steps must increase when high strength
steel is used.
Acknowledgements
The author thanks ORTIC AB, Swedish Knowledge Foundation, Jernkontoret and Dalarna University for their technical
and nancial support.
References
Fig. 8. The simulations are compared with the model Chiang derived, Eq. (3),
and one can see that they do not agree well.
The model for the peak strain, Eq. (3), has the additional assumption that the ange edge remains straight during the deformation.
7. Conclusions
The simulations show that the longitudinal peak membrane
strain decreases, the deformation length increases for materials with higher yield strength. This information has not been
possible to obtain from simple models as in Eqs. (1)(3).
Decreasing longitudinal peak membrane strain gives less
residual stresses in the ange of the prole and quality problems as wave edges, longitudinal curvature, end are, etc. will
decrease. It will make it possible to use fewer forming steps
for proles made of high strength steel. However, high strength
steel has larger spring back that has to be accounted for. This is
[1] S.M. Panton, S.D. Zhu, J.L. Duncan, Geometric constraints on the forming
path in roll forming channel sections, Proc. Inst. Mech. Eng. 206.
[2] D. Bhattacharyya, P.D. Smith, C.H. Yee, L.F. Collins, The prediction of
deformation length in cold roll forming, J. Mech. Work. Tech. 9 (1984)
181191.
[3] K.F. Chiang, Cold roll forming, ME Thesis, University of Auckland, August
1984.
[4] L. Ingvarsson, Forenklad teori for rullforming av elementar v-prol,
jamforelse mellan normalt och hoghallfast stal, VAMP 15- rullforming 23
april 2001.
[5] M. Brunet, S. Mguil, P. Pol, Modelling of a roll-forming process with a
combined 2D and 3D FEM code, J. Mater. Process. Technol. 8081 (1998)
213219.
[6] M. Farzin, M.S. Tehrani, E. Shameli, Determination of buckling limit of
strain in cold roll forming by nite element analysis, J. Mater. Process.
Technol. 125126 (2002) 626632.
[7] M. Lindgren, Finite element model of roll forming of a U-channel prole,
To be presented at International Conference on Techn. of Plasticity, Verona,
2005.
[8] MARC, Element Library, vol. B, Marc Analysis Research Corporation,
USA.
PAPER 3
An improved model for the longitudinal peak strain in the
flange of a roll formed U-channel developed by FE-analyses
Michael Lindgren
Errata
Page 4 (85)
Page 1 (86)
replace the text

A study of Figure 5 indicates
that five variables should be
included, the thickness, bend
angle, flange length, yield
strength and the tool radius,
Equation (7)
Equation (6) and Equation
(9), are determined so the
model fits the results from
the simulations done in the
factorial design.
should be
A study of Figure 5 indicates
that five variables should be
included, the thickness, bend
angle, flange length, yield
strength and the tool radius,
Equation (8)
Equation (7) and Equation
(10), are determined so the
model fits the results from
the simulations done in the
factorial design.
Metal Forming
An Improved Model for the Longitudinal Peak Strain in the Flange of a Roll
Formed U-Channel developed by FE-Analyses
Michael Lindgren
Material Science, Dalarna University, 78188 Borlnge, Sweden, mld@du.se
Today one can find cold roll forming (CRF) products in many applications, for example vehicles, furniture and in the building industry. Though
CRF is a well known sheet metal process, it is still not entirely understood due to the geometrically complex forming. There are several computer aided engineering (CAE) programs on the market that can assist the tool maker when designing a new CRF machine. However, they
are often based on simple formulas when predicting the stress and the strain in the strip.
The main objective of this study is to improve formulas for the longitudinal peak membrane strain and the deformation length when a U-channel is formed. These are important since they can be used to determine the number of forming steps and the distance between them. A twolevel factorial design is done using the finite element analysis to investigate which parameters affect the peak strain and the deformation
length. The parameters are then used to build models for the peak strain and the deformation length.
Keywords: cold roll forming, factorial design, finite element analysis
1RWDWLRQV
D
W
6
T
T
'T
5
U
.L
$%
]
/
/6
\
H
)ODQJHOHQJWK
7KLFNQHVVRIWKHVWULS
<LHOGVWUHQJWK
%HQGDQJOH
%HQGDQJOHRIWKHIHPDOHWRRO
%HQGDQJOHLQFUHPHQW
)LFWLYHWRROUDGLXV
7RROUDGLXV
&RQVWDQWV
,QWHJUDWLRQFRQVWDQWV
&DUWHVLDQFRRUGLQDWHLQORQJLWXGLQDOGLUHFWLRQ
'HIRUPDWLRQOHQJWK
6WUDLQEDVHGGHIRUPDWLRQOHQJWK
'LVWDQFHIURPWKHEHQGLQJ]RQHWRWKHIODQJHHGJH
(QJLQHHULQJVWUDLQ
,QWURGXFWLRQ
,Q FROG UROO IRUPLQJ &5) WKH VKHHW LV FRQWLQXRXVO\
IRUPHGLQVHYHUDOIRUPLQJVWHSVIURPWKHVKHHWPHWDOWRD
ILQLVKHG FURVV VHFWLRQ 'XH WR WKH JHRPHWULFDOO\ FRPSOH[
IRUPLQJWKHUHVHDUFKDERXW&5)LVOLPLWHG7KHNQRZOHGJH
RIKRZWKHWRROVVKDOOEHGHVLJQHGLVPDLQO\EDVHGRQWKH
H[SHULHQFHRIWKHWRROPDNHU7RGD\WKHUHDUHVHYHUDOFRP
SXWHU DLGHG HQJLQHHULQJ &$( SURJUDPV DYDLODEOH RQ WKH
PDUNHWWKDWFDQVXSSRUWWKHWRROPDNHUZKHQGHVLJQLQJWKH
WRROVIRUWKH&5)PDFKLQH7KHFRPSXWHUSURJUDPVDUHXVX
DOO\EDVHGRQVLPSOLILHGIRUPXODV
6RPHUHVHDUFKKDVEHHQGRQHWRVLPXODWHWKH&5)SURFHVV
E\ XVLQJ WKH ILQLWH HOHPHQW DQDO\VLV >@ 7KHUH DUH DOVR
&$(SURJUDPVWKDWFDQRIIHUDPRGXOHEDVHGRQWKHILQLWH
HOHPHQWPHWKRG)(0WKDWFDQVLPXODWHWKH&5)SURFHVV
EDVHGRQWKHJHRPHWU\JHQHUDWHGE\WKH&$(SURJUDP7KH
GLVDGYDQWDJHRI)(0LQ&5)VLPXODWLRQVLVWKDWWKHSURILOH
FDQKDYHDFRPSOLFDWHGFURVVVHFWLRQDQGVRPHWLPHVWKHUH
DUHXSWRIRUW\UROOVWDQGVQHHGHGWRFRPSOHWHWKHILQLVKHG
VHFWLRQWKXVWKHVLPXODWLRQWLPHEHFRPHVORQJ+RZHYHU
)(0DQDO\VLVZLOOEHDFRPSHWLWLYHDOWHUQDWLYHLQWKHIXWXUH
DQGLVWRGD\DQH[FHOOHQWWRROWRXVHLQUHVHDUFKWRLQFUHDVH
WKHNQRZOHGJHDERXWWKH&5)SURFHVV7KHREMHFWLYHRIWKLV
ZRUNLVWRXVH)(PRGHOWRLPSURYHH[LVWLQJIRUPXODVIRU
WKH ORQJLWXGLQDO SHDN VWUDLQ LQ WKH IODQJH RI D IRUPHG 8
VKDSH FURVVVHFWLRQ 7KHVH DUH LPSRUWDQW FRQFHUQLQJ WKH
QXPEHURIIRUPLQJVWHSVDQGWKHGLVWDQFHEHWZHHQWKHP
%DFNJURXQG
Figure 1. The material will travel a longer distance 'z in the flange
than in the bending zone.
82
/RQJLWXGLQDO VWUDLQ W\SLFDOO\ GHYHORSV LQ WKH

IODQJH RI D UROO IRUPHG FURVVVHFWLRQ GXH WR WKH PDWHULDO
WUDYHOOLQJ D ORQJHU GLVWDQFH '] LQ WKH IODQJH WKDQ LQ WKH
EHQGLQJ]RQHDVVKRZQLQ)LJXUH7KHVWUDLQGHSHQGVRQ
WKH QXPEHU RI IRUPLQJ VWHSV XVHG PDWHULDO SURSHUWLHV
ZDQWHG FURVVVHFWLRQ DQG WKLFNQHVV RI WKH PDWHULDO %KDW
WDFKDU\\DHWDO>@VWXGLHGDFKDQQHOVHFWLRQDQGGHULYHGD
PRGHOIRUWKHEHQGDQJOHGLVWULEXWHGEHWZHHQWZRUROOVWD
WLRQV E\ PLQLPLVLQJ WKH WRWDO SODVWLF ZRUN GXH WR EHQGLQJ
DQGVWUHWFKLQJ,WLVZULWWHQDV
steel research int. 78 (2007) No. 1
Metal Forming
?] $
W
] # $] # %
D
ZKHUH] LV&DUWHVLDQFRRUGLQDWHLQORQJLWXGLQDOGLUHFWLRQ
WLVWKHWKLFNQHVVDQGDLVWKHIODQJHOHQJWK
)URP WKLV DQG WKH NQRZQ ERXQGDU\ FRQGLWLRQV %KDW
WDFKDU\\D HW DO FRXOG REWDLQ D PRGHO IRU WKH GHIRUPDWLRQ
OHQJWK
R
D ?
/$

W
ZKHUH'TLVWKHLQFUHPHQWLQWKHEHQGDQJOH
7KHPRGHOSUHGLFWVWKDWWKHPDWHULDOSURSHUWLHVGRQRWDI
IHFWWKHOHQJWK,QWKHVDPHVWXG\WKH\GHULYHGDPRGHOIRU
WKHHQJLQHHULQJORQJLWXGLQDOVWUDLQLQWKHIODQJHRIDSURILOH
,WLVJLYHQE\
r s
G?

H $ \

G]
ZKHUH \ LV WKH GLVWDQFH IRUP WKH EHQGLQJ ]RQH WR WKH
IODQJHHGJH
&KDLQJ >@ FUHDWHG DQRWKHU UHODWLRQ IRU WKH ORQJLWXGLQDO
VWUDLQLQWKHHGJHRIWKHIODQJHZLWKKHOSRI(TXDWLRQ
7KLVH[SUHVVLRQLV
l\D
r s
l
W
l
\
]
H$

l
D
]/
&KDLQJFRPSDUHGWKHPRGHO ZLWK DQ H[SHULPHQW DQG KH
FRQFOXGHGWKDWWKHPRGHOSUHGLFWVDSHDNVWUDLQDERXWWKUHH
WLPHVODUJHUWKDQWKHPHDVXUHG
=KX>@VWXGLHGWKHIHPDOHWRROJHRPHWU\DQGREWDLQHGDQ
LPSOLFLWEHQGDQJOHFXUYHPRGHOIRUWKHIHPDOHWRROZKLFKLV
S
]$/U
D FRV ?]
VLQ? U ?]
# DU
U D
FRV ?
FRV ?

ZKHUHTLVWKHEHQGDQJOHRIWKHIHPDOHWRRO
=KX FRQFOXGHG WKDW WKH ORQJLWXGLQDO SHDN VWUDLQ RFFXUV
ZKHQWKHVWULSLVLQFRQWDFWZLWKWKHIHPDOHWRROIRUWKHILUVW
WLPH=KXDOVRFRQFOXGHGWKDWWKHGHULYDWLYHRIWKHEHQGDQ
JOHFXUYHDWWKDWVSHFLILFFRQWDFWPRPHQWFDQEHXVHGWRGH
WHUPLQHWKHVHYHULW\RIWKHIRUPLQJ7KHGHULYDWLYHLV
G?
/U]
$
G]
D VLQ ?]
FRV?] U ?
# DU
FRV ?
FRV ?
7KH PRGHO SUHGLFWV WKDW WKH IRUPLQJ VHYHULW\ GHFUHDVHV

ZKHQ WKH WRRO UDGLXV LQFUHDVHV DQG WKHUHE\ GHFUHDVLQJ WKH
ORQJLWXGLQDO SHDN VWUDLQ =KX DOVR REVHUYHG WKDW WKH SHDN
PHPEUDQHVWUDLQGHFUHDVHVZKHQDIL[HGIRUPLQJLQFUHPHQW
LVXVHGLQHDFKUROOVWDWLRQIRUH[DPSOH LQVWHSVRI

3DQWRQHWDO>@SURSRVHGDEHQGDQJOHFXUYHPRGHOE\
FRPELQLQJ (TXDWLRQ DQG (TXDWLRQ 7KH\ DOVR FRQ
FOXGHG WKDW WKH ORQJLWXGLQDO SHDN PHPEUDQH VWUDLQ RFFXUV
ZKHQWKHVWULSLVLQFRQWDFWZLWKWKHIHPDOHWRROIRUWKHILUVW
WLPH7KH\IRXQGWKDWWKHLUFRPELQHGPRGHOKDGDJRRGFRU
UHODWLRQEHWZHHQWKHRU\DQGH[SHULPHQWIRUT ]
=KX HW DO >@ LQYHVWLJDWHG WKH HIIHFWV RI JHRPHWU\ ZLWK
KHOSRIWKHEHQGDQJOHFXUYHWKDW3DQWRQHWDO>@SURSRVHG
7KHPRGHOSUHGLFWHG
7KH SHDN ORQJLWXGLQDO VWUDLQ LQFUHDVHV ZLWK WKH PDWHULDO
WKLFNQHVV
7KHSHDNORQJLWXGLQDOVWUDLQDQGWKHGHIRUPDWLRQOHQJWK
LQFUHDVHZLWKWKHEHQGDQJOHLQFUHPHQW
:KHQIL[HGIRUPLQJLQFUHPHQWLVXVHGWKHSHDNORQJLWX
GLQDOVWUDLQIDOOVEXWWKHQUDSLGO\LQFUHDVHVZKHQWKHDQ
JOHILQDOO\DSSURDFKHV
,QFUHDVLQJWKHIHPDOHUDGLXVGHFUHDVHVWKHSHDNORQJLWX
GLQDOVWUDLQGXHWRORQJHUFRQWDFWOHQJWKZLWKWKHWRRO
:KHQ WKH IODQJH OHQJWK LQFUHDVHV WKH SHDN ORQJLWXGLQDO
VWUDLQLQFUHDVHVLQLWLDOO\EXWDWDFHUWDLQIODQJHOHQJWKLW
VWDUWVGHFUHDVLQJ
7KHVLPSOHIRUPXODVIRUWKHGHIRUPDWLRQOHQJWKDQGWKH
ORQJLWXGLQDOVWUDLQ(TXDWLRQVDQGGRQRWLQFOXGHWKH
HIIHFWRIPDWHULDOSURSHUWLHVDQGWKHWRROJHRPHWU\+RZHY
HUWKHVWXGLHVE\=KXHWDO>@,QJYDUVVRQ>@DQG/LQGJUHQ
>@VKRZWKH\GRDIIHFWORQJLWXGLQDOVWUDLQDQGWKHGHIRU
PDWLRQOHQJWK
,QJYDUVVRQ>@UROOIRUPHGD9VHFWLRQRIPLOGDQGXOWUD
KLJKVWUHQJWKVWHHOZLWKWKHHTXDOQXPEHURIIRUPLQJVWHSV
7KHXOWUDKLJKVWUHQJWKVWHHOJDYHDVWUDLJKWSURILOHZKHUHDV
WKHPLOGVWHHOGLGQRW7KHFRQFOXVLRQZDVWKDWIHZHUIRUP
LQJVWHSVDUHQHHGHGZKHQKLJKVWUHQJWKVWHHOLVXVHG
/LQGJUHQ>@LQYHVWLJDWHGWKH\LHOGVWUHQJWKLQIOXHQFHRQ
WKH ORQJLWXGLQDO SHDN VWUDLQ DQG WKH VWUDLQ EDVHG GHIRUPD
WLRQ OHQJWK 7KH FRQFOXVLRQ ZDV WKDW LQFUHDVHG \LHOG
VWUHQJWKJLYHVLQFUHDVHGGHIRUPDWLRQOHQJWKDQGGHFUHDVHG
ORQJLWXGLQDOSHDNVWUDLQ
$SSURDFKDQG6FRSH
,QWKLVSURMHFWD&5)SURFHVVKDVEHHQPRGHOOHGZLWKWKH
)( SDFNDJH 0$5&0(17$7 7KH REMHFWLYHV ZLWK WKLV
VWXG\DUH
:LWK WKH KHOS RI WZROHYHO IDFWRULDO GHVLJQ XVLQJ )(
DQDO\VLV WR LQYHVWLJDWH ZKLFK SDUDPHWHUV LQIOXHQFH WKH
ORQJLWXGLQDO SHDN PHPEUDQH VWUDLQ DQG WKH VWUDLQ EDVHG
GHIRUPDWLRQOHQJWK
7RFUHDWHDVLPSOHPRGHOWKDWFDQSUHGLFWWKHSHDNVWUDLQ
DQGWKHGHIRUPDWLRQOHQJWK
7KHGHYHORSHGVLPSOLILHGPRGHOLVFRPSDUHGZLWKHDUOLHU
)( DQDO\VLV RI WKH &5) SURFHVV >@ >@ DQG ZLWK WKH
IRUPXODVLQ(TXDWLRQDQG
)LQLWH(OHPHQW0RGHO
7KHJHRPHWU\7KHILQLWHHOHPHQWPRGHOFRQVLVWVRIWZR
IRUPLQJ VWDQGV DQG WZR VWDQGV ZKHUH QR IRUPLQJ RFFXUV
)LJXUH7KHODWWHUDUHXVHGDVEHOWIHHGHUWRWKHRWKHUWZR
ZKHUHWKHIRUPLQJWDNHVSODFH7ZRGLIIHUHQWVWULSVDUHVLP
XODWHG7DEOH
7KHVWULSLVPRGHOOHGZLWKDIRXUQRGHWKLFNVKHOOHOHPHQW
W\SHQXPEHULQ0$5&>@7KHHOHPHQWVL]HLV PP
LQWKHEHQGLQJ]RQHDQGLQRWKHUUHJLRQVWKHHOHPHQWVL]HLV
PP7KUHHOD\HUVRILQWHJUDWLRQSRLQWVDUHXVHGLQWKH
WKLFNQHVVGLUHFWLRQ
83
Metal Forming
Table 1. The strip models.
.GPIVJ
6JKEMPGUU
9KFVJ
0WODGTQHGNGOGPV
5VTKR
OO
OOOO
OO

5VTKR
OO
OOOO
OO

Table 2. The factors and there levels.

(CEVQTU
(NCPIGNGPIVJ
C
6JKEMPGUU
V
$GPFCPIGNKPETGOGPV
'T
6QQNTCFKWU
T
;KGNFUVTGPIVJ
5
.GXGNU
.QY*KIJ
OO
OO
OO
OO

OO
/2C
OO
/2C
7KH WRROV DUH PRGHOOHG DV ULJLG VXUIDFHV DQG GXH WR WKH
V\PPHWU\RQO\KDOIRIWKHJHRPHWU\LVPRGHOOHG7KHLQL
WLDODQJXODUVSHHGRIWKHWRROVLVUDGVZKLFKJLYHVWKH
VWULSDVSHHGRIPV7KHVSHHGLVVWHSSHGXSE\LQ
WKHVW DQGQG IRUPLQJVWDQG7KLVLVGRQHLQRUGHUWRFRXQ
WHUDFWEXFNOLQJEHWZHHQWKHVWDQGV
0DWHULDOPRGHO $QHODVWLFSODVWLFPDWHULDOPRGHOZKHUH
WKHYRQ0LVHV\LHOGVXUIDFHDQGWKHDVVRFLDWHGIORZUXOHLV
XVHG7KHKDUGHQLQJLVLVRWURSLF7KHXVHGGDWDIRUWKHPD
WHULDOLVVKRZQLQ)LJXUH)ORZVWUHVVYHUVXVSODVWLFVWUDLQ
IRUWKHPDWHULDOLVLPSOHPHQWHGDVDWDEOHLQWKH)(SURJUDP
&RQWDFW&RXORPEIULFWLRQLVDVVXPHGLQWKHPRGHODQG
WKHIULFWLRQFRHIILFLHQWLV7KHFOHDUDQFHEHWZHHQWKHXS
SHUDQGORZHUWRROVLVUHVSHFWLYHO\
)DFWRULDO'HVLJQ
Figure 2. The finite element model consists of two forming stands

and two stands where no forming occurs. The latter are used as belt
feeder to the other two.
Figure 3. Stress strain curve for the material hardening.
84
$ WZROHYHOIDFWRULDOGHVLJQ>@KDVEHHQXVHGWRLQYHV
WLJDWHZKLFKIDFWRUVDUHPRVWLQIOXHQWLDORQWKHORQJLWXGLQDO
SHDNPHPEUDQHVWUDLQDQGWKHGHIRUPDWLRQOHQJWK7KHLQ
YHVWLJDWHGIDFWRUVDUHIODQJHOHQJWKDVWULSWKLFNQHVVW
WRROUDGLXVU\LHOGVWUHQJWK6DQGEHQGDQJOHLQFUHPHQW
'T 7KH IDFWRUV KDYH WZR OHYHOV WKH ORZOHYHO DQG WKH
KLJKOHYHO7DEOH
7KHWHVWVHULHVDUHVHWXSLQDVWDQGDUGRUGHULQDWDEOHRI
FRQWUDVWFRHIILFLHQWV>@QRWVKRZQKHUH7KHILYHSDUDP
HWHUVQHHGWKDWLVVLPXODWLRQV7KHORQJLWXGLQDOSHDN
PHPEUDQH VWUDLQ DQG WKH VWUDLQ EDVHG GHIRUPDWLRQ OHQJWK
DUH WKH UHVXOWV IRU HDFK VLPXODWLRQ ,W LV DVVXPHG WKDW SD
UDPHWHUV WKDW KDYH D VPDOO LQIOXHQFH RQ WKH UHVXOWV LV OLNH
QRLVH DQG ZLOO IROORZ D QRUPDO GLVWULEXWLRQ7KH IDFWRUV
WKDWKDYHDODUJHLQIOXHQFHRQWKHUHVXOWVZLOOWKHUHIRUHGH
YLDWHVWURQJO\IURPDQRUPDOGLVWULEXWLRQ7KHVHSDUDPHWHUV
FDQ EH HDVLO\ LGHQWLILHG E\ SORWWLQJ WKH UHVXOWV LQ QRUPDO
SUREDELOLW\SORWDVLQ)LJXUHDQG7KLVSORWLVVXFKWKDW
UHVXOWVDORQJWKHLQGLFDWHGVWUDLJKWOLQHGRIROORZDQRUPDO
GLVWULEXWLRQ7KH\KDYHDVPDOOLQIOXHQFHRQWKHUHVXOWV2QH
FDQVHHWKDWILYHFDVHVLQ)LJXUHDQGVHYHQFDV
HVLQ)LJXUHGHYLDWHIURPQRUPDOGLVWULEXWLRQ
7KXVWKH\VKRXOGEHLQFOXGHGLQPRGHOVIRUWKH
SHDNVWUDLQDQGWKHGHIRUPDWLRQOHQJWK7KHSD
UDPHWHUV WKDW DIIHFW WKHVH FKDQJHV KDYH EHHQ
LGHQWLILHG EDVHG RQ WKH WDEOH RI FRQWUDVW FRHIIL
FLHQWV7KHVH SDUDPHWHUV DUH ZULWWHQ DV LGHQWLIL
FDWLRQWH[WLQWKHILJXUHVQHDUWKHVHSRLQWV
3URSRVHG PRGHOV IRU WKH ORQJLWXGLQDO SHDN
VWUDLQDQGWKHGHIRUPDWLRQOHQJWK $ VWXG\RI
)LJXUHLQGLFDWHVWKDWWKHVWUDLQEDVHGGHIRUPD
WLRQOHQJWKPRGHOVKRXOGLQFOXGHIRXUYDULDEOHV
EHQGDQJOHIODQJHOHQJWKWKHWKLFNQHVVDQG\LHOG
VWUHQJWK7KH EHQG DQJOH WKH IODQJH OHQJWK DQG
WKH\LHOGVWUHQJWKGRKDYHSRVLWLYHHIIHFWVLQ)LJ
XUH DV WKH\ DUH LQ WKH SRVLWLYH SDUW RI WKH DE
VFLVVD ,QFUHDVLQJ WKH WKLFNQHVV UHGXFHV WKH GH
IRUPDWLRQOHQJWKDQGWKHUHIRUHLWVHIIHFWLVQHJD
WLYH LQ WKH ILJXUH 3RVLWLYH HIIHFWV DUH SODFHG LQ
WKHQXPHUDWRUDQGQHJDWLYHHIIHFWVLQWKHGHQRP
Metal Forming
LQDWRU RI WKH SURSRVHG PRGHO LQ (TXD

WLRQ
/ V $ .
D . ? . 6 .
W .
$ VWXG\RI)LJXUHLQGLFDWHVWKDWILYH
YDULDEOHVVKRXOGEHLQFOXGHGWKHWKLFN
QHVV EHQG DQJOH IODQJH OHQJWK \LHOG
VWUHQJWK DQG WKH WRRO UDGLXV (TXDWLRQ
2QH FDQ DOVR VHH LQWHUDFWLRQV EH
WZHHQ WKH SDUDPHWHUV LQ )LJXUH DQG
)LJXUH DQGFRQVLGHUDWLRQVIRULQWHUDF
WLRQVZLOOEHLQFOXGHGLQWKHPRGHOV,Q
WHUDFWLRQV DUH WKRVH VLPXODWLRQV ZKHUH
GLIIHUHQWSDUDPHWHUVLQWKHWDEOHRIFRQ
WUDVWVFRHIILFLHQWVDUHYDULHGVLPXOWDQH
RXVO\7KHILQDORXWFRPHRIWKHUHOHYDQW
SDUDPHWHUVLV
SHDN $ .
W . ? .
D . 6 . U .
7KHIHPDOHWRROUDGLXV =KX>@FRQ
FOXGHG WKDW ZKHQ FRQVWDQW IRUPLQJ LQ
FUHPHQWVDUHXVHGIRUH[DPSOH
LQ VWHSV RI WKH ORQJLWXGLQDO SHDN
PHPEUDQHVWUDLQZLOOGHFUHDVHLQHYHU\
IRUPLQJVWHS=KXSURSRVHGDPRGHOIRU
WKHIHPDOHJHRPHWU\(TXDWLRQWKDW
LQFOXGHVWKHWRROUDGLXVEXWWKHPRGHO
KDV WKH GLVDGYDQWDJH WKDW RQH PXVW
NQRZZKHUHWKHVWULSLVLQFRQWDFWZLWK
WKH WRRO WKH ILUVW WLPH 7KH VXJJHVWLRQ
JLYHQLQWKLVSDSHULVWRXVHDQHIIHFWLYH
UDGLXV:KHQWKHHGJHRIWKHIODQJHLV
LQFRQWDFWZLWKERWKWKHXSSHUDQGORZ
HUWRROWKHGLUHFWLRQRIWKHQRUPDOYHF
WRUWRWKHSRLQWZKHUHWKHHGJHLVFRQ
WDFWZLWKWKHWRROVFDQEHXVHGDVDQHI
IHFWLYHUDGLXV)LJXUH
7KH PRGHO IRU WKH HIIHFWLYH WRRO UD
GLXV5LV
5 $ D WDQ ? #
U
FRV ?
Figure 4. Normal probability plot for the deformation length.
7KHDGYDQWDJHRIWKHHIIHFWLYHUDGLXV
LVWKDWWKHUDGLXVZLOOLQFUHDVHZKHQWKH Figure 5. Normal probability plot for the longitudinal peak membrane strain.
EHQG DQJOH LQFUHDVHV 7KH PRGHO FDQ
DOVREHLPSOHPHQWHGLQD&$(V\VWHP7KHGLVDGYDQWDJHRI
WKHHIIHFWLYHUDGLXVLVWKDWWKHPRGHOZLOOSUHGLFWDQLQILQLWH
UDGLXVZKHQWKHIODQJHUHDFKHV LQWKHIRUPLQJSURFHVV
,Q>@WKDWSUREOHPZLWKWKHODVWIRUPLQJVWHSZDVDOVRRE
VHUYHG DQG FRQFOXGHG WKDW LW KDV WR GR ZLWK WKH VHYHUH
ERXQGDU\FRQGLWLRQVEHWZHHQWKHWRRODQGWKHVWULS
7KHWRROUDGLXVULQ(TXDWLRQUHSODFHVZLWKWKHHIIHF
WLYHWRROUDGLXV5(TXDWLRQWKHSHDNVWUDLQPRGHOZLOO
QRZDFFRXQWIRUFRQVWDQWIRUPLQJLQFUHPHQW,WLVZULWWHQDV
SHDN $ .
W . ? .
D . 6 . 5 .

Figure 6. The female tool and the effective tool radius.
85
Metal Forming
5HVXOWVDQG'LVFXVVLRQ
7KH REWDLQHG PRGHOV IRU WKH VWUDLQ
EDVHGGHIRUPDWLRQOHQJWKDQGWKHORQ
JLWXGLQDOSHDNPHPEUDQHVWUDLQDUH
/ V $
D ? 6

W
SHDN $
Figure 7. The models Equation (2) and Equation (11) are compared with the work done by
Lindgren [10]. One can see that Equation (11) agrees well with Lindgren [10] but Equation
(2) does not.

W ?
D 6 5

ZKHUH WKH YDULDEOHV DUH JLYHQ LQ WKH
XQLWVPLOOLPHWUHUDGLDQVDQG1HZWRQ
7KHPRGHOWKDW%KDWWDFKDU\\DHWDO
>@ REWDLQHG IRU WKH GHIRUPDWLRQ
OHQJWK(TXDWLRQSUHGLFWVWKDWWKH
GHIRUPDWLRQ OHQJWK LQFUHDVHV ZLWK LQ
FUHDVLQJIODQJHOHQJWKEHQGDQJOHDQG
GHFUHDVLQJ WKLFNQHVV (TXDWLRQ
SUHGLFWV WKH VDPH EXW LW DOVR SUHGLFWV
LQFUHDVLQJ GHIRUPDWLRQ OHQJWK ZKHQ
WKH\LHOGVWUHQJWKLQFUHDVHV7KHLQIOX
HQFH RI WKH \LHOG VWUHQJWK RQ WKH GH
IRUPDWLRQOHQJWKDOVRDJUHHVZLWKWKH
)(VLPXODWLRQV E\ /LQGJUHQ >@ ,Q
)LJXUH (TXDWLRQ DQG (TXDWLRQ
DUH FRPSDUHG ZLWK WKHVH UHVXOWV
7KH DJUHHPHQW EHWZHHQ WKH (TXDWLRQ
DQG WKH VLPXODWLRQ UHVXOW WKDW
/LQGJUHQ >@ REWDLQHG LV JRRG 7KH
(TXDWLRQ DQG WKH UHVXOWV IURP WKH
VLPXODWLRQV GR QRW DJUHH ZHOO GXH WR
WKH ODUJH VLPSOLILFDWLRQV LQ (TXDWLRQ
,W VLPSOLILHG ERWK WKH PDWHULDO DV
ZHOO DV WKH GHIRUPDWLRQ EHKDYLRXU
7KHDVVXPSWLRQVIRUSUHGLFWLRQRIWKH
GHIRUPDWLRQOHQJWKDUH
7KH IODQJH DGRSWV WKH VKDSH WKDW
PLQLPLVHVWKHSODVWLFZRUN
7KHPDWHULDOLVULJLGSHUIHFWO\SODV
WLF
%HQGLQJ WDNHV SODFH RQO\ DORQJ WKH
IROGOLQH
7KHORQJLWXGLQDOEHQGLQJRIWKHZHE
DQG RXWRISODQH EHQGLQJ LQ WKH
IODQJHFDQEHQHJOHFWHG
7ZRDGGLWLRQDOGLIIHUHQFHVEHWZHHQ
WKHPRGHOVIRUWKHGHIRUPDWLRQOHQJWK
/DUH
(TXDWLRQ LV EDVHG RQ WKH EHQG
Figure 8. The model Equation (12) is compared with the simulation result that Lindgren [10]
DQJOHRIWKHIODQJHIRUDVWULSZKHUH
obtained. One can see that they agree very well.
HODVWLFVWUDLQVDUHQHJOHFWHGZKHUH
DV WKH FXUUHQW FRPSXWDWLRQDO PRGHO
LQFOXGHVWKHVH
'HWHUPLQHWKHPRGHOSDUDPHWHUV 7KHFRHIILFLHQWVDQG
7KH VWUDLQ EDVHG GHIRUPDWLRQ OHQJWK (TXDWLRQ LV
H[SRQHQWV.LLQWKHSURSRVHGPRGHOVIRUVWUDLQEDVHGGH
GHILQHG GLIIHUHQWO\ IURP WKH GHILQLWLRQ XVHG E\ %KDW
IRUPDWLRQ OHQJWK DQG WKH SHDN VWUDLQ (TXDWLRQ DQG
WDFKDU\\DHWDO>@,WZDVIRXQGFRQYHQLHQWWRLGHQWLI\
(TXDWLRQ DUH GHWHUPLQHG VR WKH PRGHO ILWV WKH UHVXOWV
LW LQ WKH ILQLWH HOHPHQW PRGHO DV WKH OHQJWK IURP WKH
IURPWKHVLPXODWLRQVGRQHLQWKHIDFWRULDOGHVLJQ
SRLQW ZKHUH WKH ORQJLWXGLQDO VWUDLQ DW WKH HGJH RI WKH
86
Metal Forming
IODQJH LV JUHDWHU WKDQ WR

WKH SRLQW ZKHUH WKH XSSHU DQG
ORZHU WRROV DUH LQ FRQWDFW ZLWK
WKHVWULS
)LJXUH FRPSDUHV(TXDWLRQ
IRUWKHSHDNVWUDLQZLWKWKHVLPXOD
WLRQUHVXOWZKLFK/LQGJUHQ>@RE
WDLQHG 2QH FDQ VHH WKDW WKH PRGHO
SUHGLFWVWKHGHFUHDVHLQSHDNVWUDLQ
ZKHQ \LHOG VWUHQJWK LQFUHDVHV ZHOO
(TXDWLRQ JLYHV D SHDN VWUDLQ RI
IRUWKHEHQGDQJOHRI DQG
DSHDNVWUDLQRIIRUEHQGDQJOH
LQGHSHQGHQW RI WKH \LHOG
VWUHQJWK7KLVLVSHDNVWUDLQLVODUJHU
WKDQLQ/LQGJUHQ>@DQGLQ(TXD
WLRQ &KDLQJ >@ FRQFOXGHG DV
ZHOO WKDW (TXDWLRQ RYHUHVWLPDW
HGWKHSHDNVWUDLQ
=KX >@ FRQFOXGHG WKDW E\ LQ
FUHDVLQJ WKH IHPDOH WRRO UDGLXV WKH
SHDN VWUDLQ ZLOO GHFUHDVH (TXDWLRQ
Figure 9. The model Equation (12) is compared with the simulation result by Lindgren [3]. The
model predicts decreasing peak strain well when constant forming increment is used. In this
7KH IDFWRULDO GHVLJQ LQ WKLV
case the constant forming increment is 15. For the 90 forming step Equation (12) is not valid.
VWXG\JLYHVWKHVDPHUHVXOW=KXHW
DO>@LQYHVWLJDWHGWKHHIIHFWRIWKH
JHRPHWULFDO YDULDEOHV DQG WKH FRQFOXVLRQV ZHUH WKDW WKH
$FNQRZOHGJHPHQWV
SHDNVWUDLQLQFUHDVHVZKHQWKLFNQHVVLQFUHDVHVWKHEHQGDQ
JOH LQFUHDVHV DQG WKH IHPDOH WRRO UDGLXV GHFUHDVHV 7KLV
7KH DXWKRU WKDQNV 257,& $% 6ZHGLVK .QRZOHGJH
DJUHHV ZLWK WKH SHDN VWUDLQ PRGHO (TXDWLRQ 7KH\
)RXQGDWLRQ -HUQNRQWRUHW DQG 'DODUQD 8QLYHUVLW\ IRU WKHLU
DOVR FRQFOXGHG WKDW WKH SHDN VWUDLQ ZLOO GHFUHDVH ZKHQ D
WHFKQLFDODQGILQDQFLDOVXSSRUW
FRQVWDQWIRUPLQJLQFUHPHQWZDVXVHG7KHPRGHO(TXDWLRQ
$RULJLQDOO\VXEPLWWHGWR6FDQGLQDYLDQ
DOVRDFFRXQWVIRUWKDW,Q)LJXUH WKHSHDNVWUDLQPRG
-RXUQDORI0HWDOOXUJ\DFFHSWHGRQ$SULO
HOLVFRPSDUHGZLWKWKHVLPXODWLRQUHVXOWVWKDW/LQGJUHQ>@
REWDLQHG,QWKLVFDVHDFRQVWDQWIRUPLQJLQFUHPHQWRI
LVXVHGIURP2QHFDQVHHWKDWWKHPRGHOSUHGLFWV
WKH GHFUHDVLQJ SHDN VWUDLQ ZHOO H[FHSW IRU WKH EHQG DQJOH
ZKHUHWKHSHDNVWUDLQPRGHO(TXDWLRQLVQRWYDOLG
5HIHUHQFH
DVWKHHIIHFWLYHWRROUDGLXV(TXDWLRQZLOOEHLQILQLWH
&RQFOXVLRQ
)(DQDO\VLVDQGIDFWRULDOGHVLJQKDYHEHHQXVHGZLWKVXF
FHVVWRFUHDWHVLPSOHIRUPXODVIRUWKHVWUDLQEDVHGGHIRUPD
WLRQ OHQJWK DQG WKH ORQJLWXGLQDO SHDN PHPEUDQH VWUDLQ ,W
VKRXOGEHQRWHGWKDWWKHPRGHOLVGHYHORSHGIRUWKHIRUPLQJ
RIDUHFWDQJXODU8VKDSH
7KHPRGHOIRUWKHORQJLWXGLQDOSHDNPHPEUDQHVWUDLQLV
LPSURYHGFRPSDUHGZLWKWKHDQDO\WLFDOPRGHO7KHPRGHO
FDQQRZSUHGLFWGHFUHDVLQJSHDNVWUDLQZKHQFRQVWDQWIRUP
LQJ LQFUHPHQWV DUH XVHG DV ZHOO DV GHFUHDVLQJ SHDN VWUDLQ
ZKHQWKH\LHOGVWUHQJWKLQFUHDVHV
7KH PRGHO IRU WKH VWUDLQ EDVHG GHIRUPDWLRQ OHQJWK LQ
FOXGHVSDUDPHWHUVVLPLODUDVIRU(TXDWLRQEXWWKHPRGHO
DOVRDFFRXQWVIRULQFUHDVLQJOHQJWKZKHQWKH\LHOGVWUHQJWK
LQFUHDVHV
7KHVLPSOHPRGHOVWKDWKDYHEHHQGHYHORSHGFDQHDVLO\
EHLPSOHPHQWHGLQD&$(V\VWHPWRHVWLPDWHWKHORQJLWXGL
QDOSHDNPHPEUDQHVWUDLQDQGWKHVWUDLQEDVHGGHIRUPDWLRQ
OHQJWKDQGWKHUHE\JLYLQJEHWWHUGHVLJQ
>@ &.0F&OXUH+/L0DQXI5HY

>@ 0%UXQHW60JXLO33RO-0DWHU3URFHVV7HFKQRO

>@ 0/LQGJUHQ)LQLWH(OHPHQW0RGHORI5ROO)RUPLQJRID8FKDQQHO
3UHVHQWHGDW,QW&RQIRQ7HFKRI3ODVWLFLW\9HURQD2FW
>@ ' %KDWWDFKDU\\D 3 ' 6PLWK &+ <HH ,) &ROOLQV - RI 0HFK
:RUN7HFKQRO
>@ .)&KDLQJ&ROGUROOIRUPLQJ0(7KHVLV8QLYHUVLW\RI$XFNODQG
$XJ
>@ 6 =KX 7KHRUHWLFDO DQG H[SHULPHQWDO DQDO\VLV RI UROO IRUPLQJ 3KG
7KHVLV)HEUXDU\$XNODQG1HZ=HDODQG
>@ 603DQWRQ6'=KX-/'XQFDQ3URF,QVWQ0HFK(QJUV

>@ 6' =KX 60 3DQWRQ -/ 'XQFDQ 3URF ,QVWQ 0HFK (QJUV

>@ /,QJYDUVVRQ)|UHQNODGWHRULI|UUXOOIRUPQLQJDYHOHPHQWlU9SURILO
MlPI|UHOVHPHOODQQRUPDOWRFKK|JKnOOIDVWVWnO9$03 UXOOIRUPQ
LQJDSULO
>@ 0/LQGJUHQ&ROGUROOIRUPLQJRID8FKDQQHOPDGHRIKLJKVWUHQJWK
VWHHO6XEPLWWHGWR-0DWHU3URFHVV7HFKQRO-XQH
>@ 0$5&(OHPHQW/LEUDU\9ROXPH%0DUF$QDO\VLV5HVHDUFK&RU
SRUDWLRQ86$
>@ *(3%R[:*+XQWHU-6+XQWHU6WDWLVWLFVIRUH[SHULPHQWHUV
-RKQ:LOH\DQG6RQV,6%1
87
PAPER 4
Experimental investigation of roll load and roll torque when
high strength steel is roll formed
Michael Lindgren
Journal of Materials Processing Technology 191 (2007) 4447
Experimental investigations of the roll load and roll torque

when high strength steel is roll formed
Michael Lindgren
Material Science, Dalarna University, 781 70 Borlange, Sweden
Abstract
The cold roll forming process is a highly efcient process used to produce proles for many applications, for example vehicles, buildings,
domestic machines, etc. Therefore, its market share is increasing every year. Many of the above products are already today made of high strength
steel and the usage of these materials will likely continue to increase.
The objectives of this project are to nd how the roll load and roll torque are inuenced by the yield strength of the material. Full-scale experiments
have been performed. U-channels made of different materials from mild to ultra high strength steels have been formed. The roll torque is measured
during the process using a torque sensor mounted between the tool and the power transmission. Used power is also calculated with help of the
motor current. The roll load is measured with load cells mounted on both side of the roll forming tool.
The experimental result will increase the understanding of the specic conditions for roll forming steels with increasing yield strength. The
result can be used to support the roll machine designer to choose machine elements and power unit for these applications. Furthermore, the result
can also be compared with nite element simulations in order to improve and validate simulation models.
2007 Elsevier B.V. All rights reserved.
Keywords: Cold roll forming; High strength steel; Roll torque; Roll load
1. Introduction
Roll forming is a bending operation where the bending occurs
in several forming steps from an undeformed strip to a nished
prole. Every forming step consists of a pair of forming tools that
rotate and drive the strip forward during the forming. Today one
can nd roll formed products in many different elds as buildings, vehicles, furniture, etc. The use of the process increases
continuously due to the efciency of the process.
Despite the fact that roll forming is a common sheet metal
forming process and used worldwide there are relatively few
publications about the process. The explanation may be the complexity of the deformation behaviour in the process despite its
visual simplicity. There are numerous factors which can inuence the roll forming design, for example unlimited variation
of the shape of the prole. Nevertheless, some researchers have
developed simple models [1,2], and in a newly released book
Halmos [3] described how this process can be used successfully.
Finite element simulations have been used a long time in
sheet metal forming to develop forming tools. However, this has
E-mail address: mld@du.se.

0924-0136/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2007.03.041
not been the case for the roll forming processes due to limited
computer capacity. Only in recent years has this became possible [46], and today roll forming computer aided engineering
system [7], can offer simplied nite element models for simulations. From these models the roll load and roll torque can be
received and compared with measurement and thereby validated.
Bhattacharyya et al. [8] suggested an analytic model for the
roll load in a single roll station, Eq. (1), when a U-channel was
formed:

2t 3 3 a
F =s
(1)
3 sin()
where t is the thickness, s the yield strength, a the ange length
and is the bend angle.
The model was compared with measurements for mild steel
and aluminium for different thickness and bend angle. The
conclusion was that the error between the model and the measurements was within 020%.
The current work is an investigation of the roll forming of a
U-channel of different sheet metals, ranging from mild to ultra
high strength steel. Roll load and roll torque have been measured
during the forming. The results from these measurements can
be used to validate nite element models, can support the roll

machine designer to choose machine elements and power unit
when the strength of the material increases.
2. Experimental details
2.1. The roll forming mill
A standard roll forming mill [3] (Fig. 1) is used to investigate the roll load and
roll torque. The mill has six forming stands and universal joint driven shafts are
used to run the tools. The power unit is a 15 kW frequency controlled three-phase
motor. No driven top shafts are used.
45
Table 1
The materials that have been tested are Docol, DC01 (carbon steel) and HyTens,
SS2333 (stainless steel)
Material
Thickness (mm)
RP0.2 (MPa)
DC01
Docol 800DP
Docol 1200M
Docol 1200M
HyTens 1200
SS2333
HyTens X
1.46
1.48
1.46
1.22
1.0
1.0
1.0
193
533
1129
1119
1018
301
286
2.2. The measurement equipment

2.5. The experimental set-up
Two load cells of donut type from Honeywell Sensotec Sensors are mounted
between the tools and the forming stands in order to measure the roll load
(Fig. 1). The maximum roll load that the load cells can measure is 17,800 N.
The roll torque is measure with a torque sensor from Burster, which can
measure up to 75 Nm. The torque sensor is mounted between the power transmission and the lower forming tool (Fig. 1). An encoder is used to measure
the velocity of the prole. The motor current is measured by a frequency
changer.
The data acquisition equipment is from NATIONAL INSTRUMENTTM
where the used software is LabVIEWTM .
2.3. The prole

The U-channel is formed in four steps from an undeformed strip to a bend
angle of 80 . The undeformed strip has a width of 84 mm and a thickness ranging
from 1.0 to 1.5 mm. The bending method is constant arc length [9] which
means that the bend radius will be different for every forming step but the arc
length of the bend corner will be constant. The nal prole, bend angle 80 , has
a ange length of 23.3 mm, a web of 23.3 mm and a nal radius of 5.55 mm at
the neutral layer when 1 mm strip is formed.
2.4. The material

Seven different materials have been tested, four carbon steels and three
stainless steels (Table 1).
The strip is formed through a single forming step with varying bend
angles, 20 , 40 , 60 and 80 , during the forming the load and torque are
measured. The forming passes before and after the single step are only
used as at rolls (0 - -0 ). This is done in order to eliminate the inuences on torque from other forming steps. The bending angle increment
is dependent on the amount of spring back from the previous forming
step.
The load and the torque are dependent on the tolerance between the thickness
of the formed material and the roll gap between the tools. The load and the torque
will be very large if the clearance between the upper and lower tool is too small.
To avoid this, the prole is rolled in between the tool and the roll load is increased
until the roll gap is between 0.1 and 0.15 mm larger then the thickness of the
formed material, this is checked with a feeler gage.
The measurement starts when the prole rests on both the at rolls before
and after the forming step. The velocity of the prole that is used during the
measurement is 9.7 m/min. A lubricant is used in abundance and the surface
nish is RZ = 1 m.
3. Result
In Fig. 2 are the measured roll load and the roll torque presented for Docol 1200 M, thickness 1.46 mm. The mean value is
calculated for every experiment in Fig. 2 during the sample time
30 to sample time 60. A new mean value and standard deviation
is then calculated for the mean value from experiments 1 to 3
for the load and the torque. These values are presented for all
used materials in Table 2.
3.1. Models
In the experiments, for example bend angle 60 , have materials with different thickness and virgin yield strength been used,
if one assumes that only these parameters inuence the load and
the torque a model can be created as Eqs. (2) and (3):
FLoad = k1 ()t k2 sk3
k5 k6
TTorque = k4 ()t s
Fig. 1. A standard roll forming mill is used to do experiments, the roll load is
measured with load cells and the roll torque is measured with a torque sensor
during a forming of different steels.
(2)
(3)
where t is the thickness of the material, s the yield strength of

the material, ki the constants and exponents and is the used
bend angle, in this case 60 .
In Fig. 3 are the models, Eqs. (2) and (3), compared with
experiments when the bend angle 60 is used. The constants
and the exponents in the Fig. 3 are k1 () = 88, k4 () = 0.38,
k2 = k5 = 2.1 and k3 = k6 = 0.57. This is somewhat different from
46
Fig. 2. The roll load and roll torque is measured for three different experiments. The used material in the gure is Docol 1200 M, thickness 1.46, at the bend angle
60 .
the analytic model for the load as in Eq. (1), Bhattacharyya et

al. [8].
3.2. Power
The motor current is measured during the forming of Docol
1200 M, thickness 1.46 mm. All forming passes are engaged at
the same time and the current increases during the forming with
0.6 0.1 A.
The three-phase power (Pengine ) used to forming the sheet

can be calculated as:
Pengine = 3U I cos()
(4)
where U = 400 V, I = 0.6 0.1 A and cos() = 0.83.
Which gives a power of:
PEngine = 0.345 0.057 kW
(5)
Table 2
Seven different materials have been formed with the bend angles 20 , 40 , 60 and 80
Material
HyTens X
SS2333
HyTens 1200
Docol 1200 M
Docol 800 DP
DC01
Docol 1200 M (1.22 mm)
HyTens X
SS2333
HyTens 1200
Docol 1200 M
Docol 800 DP
DC01
Docol 1200 M (1.22 mm)
HyTens X
SS2333
HyTens 1200
Docol 1200 M
Docol 800 DP
DC01
Docol 1200 M (1.22 mm)
HyTens X
SS2333
HyTens 1200
Docol 1200 M
Docol 800 DP
DC01
Docol 1200 M (1.22 mm)
Bend angle (degrees)
20
20
20
20
20
20
20
40
40
40
40
40
40
40
60
60
60
60
60
60
60
80
80
80
80
80
80
80
The forming load and the forming torque are measured for all cases.
Mean value
Standard deviation
Load (N)
Torque (Nm)
Load (N)
Torque (Nm)
1818
2039
2857
8503
5247
3846
5399
2228
2418
4500
10001
8203
4877
6825
2038
2121
4689
10148
7946
4992
6521
2752
2628
4536
9575
8368
4543
7260
9.2
13.3
11.4
30.1
28.4
28.8
21.5
12.1
12.9
21.2
54.1
42.4
31.1
33.5
8.4
9.0
16.8
45.7
36.2
23.0
29.5
15.4
16.3
25.2
61.3
51.9
27.0
47.1
59
591
299
412
258
199
417
141
26
195
750
547
158
1168
206
286
394
704
384
863
522
205
275
344
372
42
527
1109
0.6
1.7
1.6
1.6
1.2
3.8
0.7
0.7
0.2
1.0
2.6
2.4
1.2
4.8
0.3
1.1
1.3
2.8
1.0
1.6
2.4
1.2
1.3
0.7
2.0
0.3
11.0
2.1
Fig. 3. The models, Eqs. (2) and (3), are compared to the experimental result
for the used materials.
The power calculated with help of measured total torque

(TTotal ) and measured forming angle velocity is:
Ptorque = TTotal
(6)
where total torque and forming angle velocity for bend angles
20 , 40 , 60 and 80 (Table 2) is:
TTotal = 30.1 + 54.1 + 45.7 + 61.3 = 191 Nm,
= 2.156 rad/s
(7)
Which gives a power of:

PTorque = 0.411 kW
(8)
4. Conclusion
The results from the experiments can be used to validate
and calibrate nite element models. Calculated force will agree
fairly well with the measured load provided the material model is
correct. The material model can be calibrated by separate material testing. Then the friction coefcient can be adjusted until
the torque also agrees well. A lubricant is used in abundance
and the surface nish is RZ = 1 m which should give a friction
coefcient between 0.05 and 0.1 for Coulomb friction.
Eqs. (2) and (3) give the torque and force required for the
bending in a forming step as a function of thickness, material
47
virgin yield strength and bending angle. The calibration of the

model gave a good agreement with measurements according to
Fig. 3. A consistency check of the measured torque by power
comparisons (Eqs. (5) and (8)) in Section 3.2 was done and
showed reasonable agreement.
A low production velocity used in the experiments (v = r,
r = 0.075 m = 2.156 rad/s, v = 9.7 m/min). A more common
production speed is 3040 m/min. A speed of 40 m/min gives a
power of 1.7 kW. Thus about 0.5 kW is required for every forming step when normal production speed is used under conditions
similar to those in the current study. It is possible to use of Eq. (3)
to estimate the increase in power consumption when changing,
for example, the yield strength of the formed material.
Acknowledgements
The author thanks ORTIC AB, Swedish Knowledge Foundation, Jernkontoret and Dalarna University for their technical
and nancial support.
References
[1] S.M. Panton, S.D. Zhu, J.L. Duncan, Fundamental deformation types and
sectional properties in roll forming, Int. J. Mech. Sci. 26 (8) (1994) 725
735.
[2] M. Kiuchi, T. Koudabashi, Proceeding of the 3rd International Conference
on Rotary Metal Working Processes, IFS Ltd and North Holland, September
1984, pp. 423436.
[3] G.T. Halmos, Roll Forming Handbook (Manufacturing Engineering and
Materials Processing), CRC Press, Taylor & Francis Group, 2006, ISBN
0-8247-9563-6.
[4] C.K. McClure, H. Li, Roll forming simulation using nite element analysis,
Manuf. Rev. 8 (1995) 114.
[5] M. Brunet, S. Mguil, P. Pol, Modeling of a roll forming process with a
combined 2D and 3D FEM code, J. Mater. Process. Technol. 8081 (1998)
213219.
[6] M. Lindgren, Finite element model of roll forming of a U-channel prole, in:
Presented at International Conference on Technology of Plasticity, Verona,
Italy, 2005.
[7] PROFIL, Rollform Design Software, UBECO, www.ubeco.com.
[8] D. Bhattacharyya, P.D. Smith, S.K. Thadakamalla, I.F. Collins, The prediction of roll load in cold roll forming, J. Mech. Work. Technol. 14 (1987)
363379.
[9] K.F. Chaing, Cold roll forming, M.E. Thesis, University of Auckland,
August, 1984.
PAPER 5
Validation of finite element model of roll forming
Michael Lindgren
International Deep Drawing Research Group

IDDRG 2008 International Conference
16-18 June 2008, Olofstrm, Sweden
VALIDATION OF FINITE ELEMENT MODEL OF ROLL FORMING

M. Lindgrenc
c
National Postgraduate School of Metal Forming

Dalarna University
Department of Material Science
SE-781 88, Borlnge, Sweden
e-mail: mld@du.se
Keywords: Roll Forming, Roll Load, Roll Torque, Finite Element Simulation
ABSTRACT. Roll forming is a common sheet metal forming process and it is used for
producing profiles to vehicles, buildings, furniture etc. The use of the process increases every
year due to the high production speed that can be used, 10 - 40 meters per minute is common.
Though roll forming is a well known sheet metal process it is still not entirely understood due
to the geometrically complex process. The use of finite element simulation in sheet metal
forming in general is common but not in the case of roll forming, due to the profile can have a
complicated cross-section and up to 40 forming steps can be needed and the simulations time
can be several days. However, the computers have become faster and the demand for finite
element simulations increases from the industry. The objectives with this study is to compare
finite element simulations with experimental data for the roll load, roll torque, strains and
spring back when a U-channel is roll formed.
M. Lindgren
1. INTRODUCTION
1.1. Background
Roll forming is a bend operation but the bending occurs in several steps from an
undeformed sheet to a finished profile. The rate of production is very high and in most cases
one can also find secondary operations in a roll forming line such as punching, slitting,
notching etc.
Roll forming is a complex process and experience is required in order to be successful with
the machine and the tool design. Today one can find computer aided engineering programs,
for example [1-2], that can assist the designer and get a proposal for the tool design. The
programs use more or less simple formulas to calculate the number of forming steps, the
strain and spring back in the profile during the forming [10]. In the past few years some
computer aided engineering programs can also offer simplified finite element models.
The objectives with this study are to validate a finite element model by comparing it with
earlier published experimental results [9]. The focus is on roll load and torque, strains and
spring back during the roll forming.
1.2. Previous work
Rebelo et al. [3] studied sheet metal forming with the help of finite element analysis and
they compared implicit and explicit finite element formulations when simulating roll forming.
They concluded that implicit formulation was faster then explicit due to roll forming having a
small wave front and the problem is assumed one dimensional.
McClure and Li [4] simulated roll forming with ABAQUS. The model ignored friction
between the tools and the profile. A horizontal force was applied in the leading edge to pull
the profile through the roll stations. The membrane strain in the profile was compared with
experiments available in literature. The strain in the profile was similar to the experiment.
Sukmoo et al. [5] used COPRA FEA-RF to simulate roll forming and compared the
forming length with an experiment available in literature. The finite element model used a
combined 2D and a 3D algorithm. A rigid-plastic material model was used. The conclusion
from the simulation was that increased work hardening exponent gives increased forming
length. It was also good agreement between the simulation result and the experiment.
Lindgren [6] modelled the roll forming process with the finite element program
MSC.MARC. The tools were modelled as rigid surfaces and they rotated to drive the profile.
A Coulomb friction model was used with the friction coefficient 0.1. The profile was
modelled with thick shell elements. An elastic-plastic material model was used. The result
from the simulation was similar to earlier papers that have been written about finite element
simulations.
Chiang [7] used the strain gage technique to measure the strain history in the profile. The
strain gage was bounded onto both the top and bottom surface at the edge of the profile. The
recorded result was used to calculate the membrane strain in the edge of the profile. One of
the conclusions was that the membrane strain increases rapidly when approaching the tools
and the maximum value is right before the centre of the tools. In the study cold rolled mild
steel was used.
Bhattacharyya et. al. [8] developed a semi-empirical model for calculating the forming
load when a U-channel was formed. The model was compared with measurements for mild
steel and aluminium with a different thickness and a different bend angle. The error between
the measurement and the model was about 0-20%.
Lindgren [9] measured the roll load and the roll torque with a load cell and a torque sensor
when a U-channel was formed. The used steels had a thickness from 1 mm to 1.46 mm and a
yield strength from 193 MPa to 1129 MPa. The conclusion was that about 0.5 kW was needed
2
M. Lindgren
in every forming step for normal production speed and under conditions similar to those in the
study. It was also concluded that the roll load and torque required for a bending in a single
forming step can be approximated as a function of material virgin yield strength, thickness
and geometry.
2. EXPERIMENTAL PROCEDURE
2.1. The roll forming machine and tools
A standard roll forming machine [10] is used to form the U-channel. The machine has six
forming stands and universal joint driven shafts run the tools, the top shafts are not driven.
Five forming steps are used to form the U-channel, 20, 40, 60, 80 and 97.
Measurements are performed between the forming steps. The geometry for the used tools is
shown in Figure 1. The horizontal distance between the forming stands is 450 mm. The roll
load and the roll torque will be very large if the clearance between the lower and upper tools
is too small. To avoid this, the roll gap is about 10% larger then the thickness of the formed
material, this is checked with a feeler gage.
The velocity of the profile that is used during the measurements is 3.375 m/min. A
lubricant is used in abundance and the surface finish for the tools is RZ = 1 m.
Figure 1. The geometry of the used roll forming tools during the strain measurement experiment.
From the left, bend angle 20, 40, 60 and 80.
2.2. The profile and material

The undeformed strip is 84 mm and the thickness is 1.0 mm. The bending method is
constant arc length [7] and the final radius is 4 mm for the U-channel. The web and the
flange are both 23.9 mm.
Two different stainless steels are roll formed, an ultra high strength steel, HyTens 1200,
and a mild steel, SS2333.
2.3. Longitudinal and transverse strain
The biaxial strains at the upper surface of the flange are measured with a strain gage
rosette. The angles between the measuring grids and the forming direction are 45, 90 and
135, Figure 2, and the grid length is 1.5 mm. The strain gage is mounted close to the edge of
the flange, the centre of the grids is located about 2.5 mm from the edge.
M. Lindgren
Figure 2. One half of the U-channel with localiser of the strain gage rosette indicated.
2.4. Spring back

A sliding caliper is used to measure the width and depth of the profile after every forming
step where the profile has spring back, Figure 3. The measurement is compared with the finite
element simulation.
Figure 3. The spring back is measured with a sliding caliper. The width in y-direction and depth in zdirection of the profile after every forming step is compared with finite element simulation.
3. FINITE ELEMENT MODEL

The FE-package MARC/MENTAT 2005 [11] has been used to model the roll forming
process.
3.1. The geometry
The model consists of five forming steps and the geometry of the tools is the same as in the
experiment, Figure 1. The tools are modelled as a rigid surface and rotate with an angular
speed of 0.75 rad/s which gives the profile a speed of 3.375 m/min. The friction between the
profile and the tools is modelled as Coulomb friction and the friction coefficient is 0.085. The
horizontal distance between the tools is 450 mm.
Due to the symmetry only one half of the U-channel is modelled. The length and width of
the unformed strip are 950 mm and 42 mm. The number of elements are 3840 and the size of
the element is 41 mm to 42 mm in the bending zone and in other regions the elements size
is 45 mm. The element type is a thick shell element 75 [11].
In this study a finite element model with 3, 5, 7, 9 and 11 integrations points in the
thickness direction has been tested and the conclusion from this is that 3 integrations points
are too few to model the spring back when a U-channel is formed. 5 - 11 integrations points
gave equal result but it was an improvement compared with 3 integrations point. In this
investigation have five integration points been used.
4
M. Lindgren
3.2. Material model

An elastic-plastic material model where the hardening is isotropic and the von Mises yield
surface and the associated flow rule is used. The tensile test data for the materials have been
extrapolated, Figure 4, and stress versus plastic strain is implemented as a table in the finite
element program. The maximum plastic strain in the flange is 1% and in the bending zone
10% in the simulations.
Figure 4. The extrapolated tensile test data for the materials SS2333 and HyTens 1200 are
implemented in the finite element program as a table.
4. RESULT AND DISCUSSION

4.1. Longitudinal and transverse strain
In Figures 5-6 the experimental results for longitudinal strain (x-direction) and the
transverse strain (y-direction), at the surface at the edge of the flange, are compared with the
finite element simulation. The experimental and simulation results for longitudinal strain
agree well. The measured transverse strains have the same pattern but the strain from the
simulations is larger than the experiment. These deviations trigged a second round of
experiment for SS2333. The measured and computed transverse strains have the same pattern
but with a sudden deviation, a shift, between the curves after the first roll. The shift became
about 50% smaller when using a thinner adhesive. It is known [12] that a too thick or too soft
adhesive will not fully transfer the strain from the sheet to the strain gage. It is likely that this
difference between computed and measured transverse strain depends on the imperfect
adhesion between the strain gage and the sheet. The measurements were only performed once
for HyTens 1200 in Figure 5 and then with a thicker adhesive.
M. Lindgren
Figure 5. Measured and computed longitudinal and transverse strains. The used material is HyTens
1200.
Figure 6. Measured and computed longitudinal and transverse strains. The used material is SS2333.
4.2. Roll load and roll torque

In Figures 7-8 the roll load from the experiments for the used materials is compared with
the finite element simulations. The computed results are within 17% of the measured for
HyTens 1200 and for the material SS2333 the computed results are within 35% of the
6
M. Lindgren
measured for the worst case. The measured results [9] are mean values of three different
measurements and the standard deviations between these measurements are 200 N to 400 N
(error bars in the figure).
Figure 7. Measured and computed roll load versus bend angle for HyTens 1200
Figure 8. Measured and computed roll load versus bend angle for SS2333
The agreement between the finite element simulation and the experiment for the roll torque
is within 25% for HyTens 1200 and 47% for SS2333, Figure 9-10. The roll torque is
dependant both on the roll load and the friction between the sheet and the tools so the
experimental result can vary more between the different measurements. The standard
deviation for the measurements is between 0.7 Nm to 1.7 Nm (error bars in the figure).
M. Lindgren
Figure 9. Measured and computed roll torque versus bend angle for HyTens 1200.
Figure 10. Measured and computed roll torque versus bend angle for SS2333.
4.3.
Spring back
The measured and computed width and depth of HyTens 1200 is shown in Figure 11. The
definitions of width and depth of the U-channel is shown in Figure 3. The upper curve in
Figure 11 starts at 84 mm, which is the original width of the strip. The corresponding square
symbols show measured width between the forming steps. They are joined by a dashed line.
The lower curve and corresponding triangle symbols show change in the depth of the Uchannel.
M. Lindgren
Figure 11. Measured and computed width and depth is compared between every forming step. The
used material is HyTens 1200.
The ultra high strength steel, HyTens 1200, Figure 11, shows satisfying agreement
between simulation and experiment for the forming step one and two. The difference
increases for the third and fourth step. The final error in the simulations is about 1 2 degrees.
The same results are shown for SS2333 in Figure 12. The agreement between the
simulation and the experiment is better than in the case of HyTens 1200 and the error in the
bend angle is about 0.5 1.0 degrees for the last step.
Figure 12. Measured and computed width and depth is compared between every forming step. The
used material is SS2333.
M. Lindgren
5. CONCLUSION
Results from the simulations have been compared with experimental results. The
simulations and experiments agree satisfactorily for the longitudinal strain, transverse strain,
roll load and torque for both HyTens 1200 and SS2333. The spring back agrees decently for
SS2333 but for the HyTens 1200 is the difference about 1 2 degrees, which is not
acceptable if the model will be used to design the forming process.
Finite element simulations will not replace the computer aided engineering (CAE)
programs in the future, but it will be a useful complement to the CAE programs. Today a
CAE program can be used to give an initial proposal to how many forming steps a profile
needs. The program also proposes the design of the steps. Still it requires an engineer with
experience to decide if the proposed tool design is good enough and if not, modify it. The
toolmaker also has to design guides on the tools so that the materials can be lead in to the
tools properly. Here is where the finite element simulations can support the toolmaker before
taking a decision for manufacture the tools.
6. ACKNOWLEDGEMENT
The author thanks ORTIC AB, Swedish Knowledge Foundation, Jernkontoret and Dalarna
University for their technical and financial support.
7. REFERENCE
1. Ubeco, Profile-Roll forming design software.
2. Data M software, COPRA-Roll forming.
3. N. Rebelo, J. C. Nagtegaal, L. M. Taylor, R. Passman: Comparsion of implicit and
explicit finite element methods in the simulation of metal forming processes, ABAQUS
User Conference, Newport, RI, (1992).
4. C. K. McClure, H. Li: Roll forming simulation using finite element analysis,
Manufacturing Review 8 (1995) 114-119.
5. H. Sukmoo, L. Seungyoon, K. Naksoo: A parametric study on forming length in roll
forming, Journal of Materials Processing Technology 113 (2001) 774-778.
6. M. Lindgren: Finite Element Model of Roll Forming of a U-channel, International
Conference on Technology of Plasticity, Verona, (2005).
7. K. F. Chiang: Cold roll forming, M. Sc. Thesis, University of Auckland (1984).
8. D. Bhattacharyya, P. D. Smith, S. K. Thadakamalla, I. F. Collins: The Prediction of Roll
Load in Cold Roll Forming, Journal of Mechanical Working Technology 14 (1987) 363
379.
9. M. Lindgren: Experimental Investigations of the Roll Load and Roll Torque When High
Strength Steel is Roll Formed, Journal of Materials Processing Technology 191 (2007)
4447
10. G. T. Halmos Roll Forming Handbook, first ed. (2005), CRC Talylor & Francis, New
York.
11. MSC.MARC., Element Library Volume B, Marc Analysis Research Corporation,
USA.
12. K. Hoffmann: An Introduction to Measurements using Strain Gages, Hottinger Baldwin
Messtechnik GmbH, Darmstadt (1989).
10
PAPER 6
Michael Lindgren, Ulf Bexell, Lars Wikstrm
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 9 ( 2 0 0 9 ) 31173124
journal homepage: www.elsevier.com/locate/jmatprotec
Review

Michael Lindgren , Ulf Bexell, Lars Wikstrm
Dalarna University, 781 88 Borlnge, Sweden
a r t i c l e
i n f o
a b s t r a c t
Article history:
Today you will nd roll formed details in many different products, for example buildings,
Received 2 July 2007
household appliances and vehicles. The industry, in order to save weight, tends to use more
Received in revised form
and more high strength steel. The disadvantage with these materials is that they can be
18 July 2008
difcult to form due to reduced ductility. A way to increase the ductility in the forming
Accepted 29 July 2008
areas is by partially heat the steel.

It is shown that partial heating substantially increases the ductility of high strength steel
and make it possible to roll form large bend angles. When roll forming, the material will
Keywords:
work hardening almost to the as-received condition in the outer and inner radius of the roll
TRIP steel
formed prole. Furthermore, the heating power decides the bend angle obtained. Finally,
High-strength steel
the mechanical properties after heating and roll forming are discussed.
Roll forming
2008 Elsevier B.V. All rights reserved.
Partial heat treatment
Contents
1.
2.
3.
4.
5.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1. Material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2. Furnace experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3. Partial heating experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4. Roll forming experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5. Sample preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1. Furnace experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2. Partial heating experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3. Roll forming experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Corresponding author. Tel.: +46 23 778741; fax: +46 23 778601.

E-mail address: mld@du.se (M. Lindgren).
0924-0136/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2008.07.041
3118
3118
3118
3118
3118
3119
3119
3120
3120
3120
3121
3122
3123
3123
3123
3118
1.
Introduction
Roll forming is a forming process for sheet metal, which has

gained in popularity due to high production rates and the possibility to form complex geometries. By including processes, as
welding, riveting and punching in the same production line, it
is possible to manufacture near net shape products from sheet
metal. In order to save weight, the industry tends to use more
high strength steels. One problem when forming high strength
steels is their low ductility. When a high strength steel product of complex geometry is desired it is possible to use an
austenitic stainless steel, e.g. AISI 301, in its annealed condition. This type of steel has an excellent ductility and strain
hardening during forming resulting in high strength in the
formed areas, also called TRIP (TRansformation Induced Plasticity) steels. However, the unformed parts will remain soft.
An alternative method to produce a complex geometry is to
start with this type of steel in a cold rolled condition and
increase the ductility in the forming areas by partial heating.
This approach will give products with high strength properties
also in undeformed areas. The latter approach is preferable in
roll forming, where it is desired to concentrate plastic deformation to the bending areas. Furthermore, the heated areas
will strain hardening and regain some of the lost strength
during roll forming.
In a study conducted by Mumtaz et al. (2004) it was concluded that specimens of a 304 type austenitic stainless
steel rolled to 55% reduction in thickness produced 74% martensite phase. In this study cold rolled AISI 301 steel, rolled
to a reduction of 57.5% in thickness, has been partially heated
and roll formed. Thus, the as-received material in this study
should have similar high strain induced -martensite phase
content. Upon heating the material will soften due to recovery, recrystallization and phase transformations. The recovery
is manly attributed to the annihilation and rearrangement of
dislocations. A complete recovery is seldom achieved without recrystallization, a process where grains with high defect
density is replaced with new defect free grains. Johannsen et
al. (2006) concluded, for annealing temperatures higher than
800 C, that the microstructure consists of recrystallized grains
of austenite and minor carbides, majority of the carbides were
found in the grains and not at the grain boundaries. Di Schino
et al. (2002) studied the austenite-martensite transformation
under cold rolling deformation followed by annealing in a
304 stainless steel. The purpose with the investigation was
to study the effect of thermo-mechanical processes on grain
renement. It was concluded that an increased reduction in
thickness gives a more homogenous grain size and a more
recrystallized microstructure using the same annealing time
and temperature.
When the heated areas of the material are roll formed it will
transform back to -martensite from austenite, -phase. The
phase transformation is a nucleation and shear process where
austenite, transforms into -martensite and -martensite. It
is believed that the -martensite acts as an intermediate phase
in the transformation sequence as discussed in the
papers by Mszros and Prohszka (2005) and Spencer et al.
(2004). It should be noted that this transformation sequence
is far from fully understood as discussed by Petit et al. (2007),
Fig. 1 The tensile test data for an as-received HyTens

1600.
particularly the role of -martensite on the formation of martensite.

The aim of the present investigation is to apply partial heating to a cold rolled AISI 301 stainless steel. The inuence of
resistance heating on the material in the bending area before
and after roll forming will be investigated.
2.
Experimental
2.1.
Material
The investigated material, HyTens 1600, is a cold rolled stainless steel. The composition satises the grade of AISI 301. The
total reduction after the last annealing step is 57.5% resulting
in a yield strength Rp0.2 of 1460 MPa, an average hardness of
580 HV0.5 and a thickness of 0.7 mm. In Fig. 1 a tensile test
of the material is presented and in Fig. 2 the microstructure
is shown. The microstructure shows a typical cold deformed
structure with heavily deformed austenite grains which are
transformed to -martensite.
2.2.
Furnace experiment
Flat samples of 0.7 mm 15 mm 84 mm in size were heat

treated in an electrical batch furnace followed by quenching in water. The heat treatment time was 15, 30, 60, 90, and
120 s, respectively, with temperatures ranging from 500 C up
to 1200 C, in steps of 100 C.
2.3.
The partial heating was performed in a prototype resistance

heating machine, see Fig. 3. The diameters of the heating
wheels are 122.5 mm and the width of the contact zone
between the wheel and the material is 5 mm. The power was
3119
Fig. 2 Light optical image of the microstructure of an

as-received HyTens 1600. The image width is 150 m.
Fig. 4 Schematics of a standard roll forming machine
equipped with load cells.
adjusted as high as possible without any, for the naked eye,
detectable surface defects. This power setting is noted P30.
Also, two lower power settings were investigated noted P10
and P20, were P10 is the lowest. The velocity used to heat the
material was 1 m/min for all power setting.
2.4.
Roll forming experiment
Halmos (2005) described different types of roll forming

machines where the most popular type is the standard inboard
machine. This due to the design of the machine, they can be
built and used for materials with any thickness and width.
In this study such a machine, see Fig. 4, has been used to
roll form a V-section. The horizontal distance between the
passes is 450 mm. The radius for the upper tools is 70 mm and
for the lower tools 50 mm. The forming velocity used in the
experiments was 3.5 m/min. The forming load was measured
with load cells. The measured load was about 500 N. The Vsection is roll formed in six passes where the forming steps
are 15 , 30 , 60 , 90 , 105 and 120 , see Fig. 5. Chiang (1984)
Fig. 5 The ower pattern for the roll formed V-section.
The forming steps are 15 , 30 , 60 , 90 , 105 and 120 .
discussed three different ways to bend a corner, constant arc

length method, constant radius method and a mixture of
both methods. In this study the constant radius method is
used to bend the V-section, which means that the inner radius
is the same for all passes, in this case 0.4 mm, see Fig. 6. The
width of the unformed material is 54 mm and the thickness is
0.7 mm.
2.5.
Fig. 3 Prototype resistance heating machine used for

partial heating experiments.
Sample preparation
Cross-sections, perpendicular to the rolling directions, of

all samples were hot mounted using a thermosetting resin
followed by mechanical grinding and polishing with 1 m diamond suspension in the last step. Chemical etching was used
to reveal the microstructure of the samples. The etching solu-
3120
tion consisted of 60 ml deionised water, 62 ml concentrated

HCl, 3 ml concentrated HNO3 and 5 ml concentrated H2 SO4 .
The etching time was approximately 60 s at room temperature. To reveal the heated zone in the partially heated samples
etching was performed in an etching solution consisting of
25 ml concentrated HCl and 5 ml 10 wt% chrome acid solution
in deionised water. In this case, the etching time was 510 s.
The effect on the material due to the heat treatments and
the roll forming was evaluated with Vickers microhardness
tests with an applied load of 500 g. All microhardness testing was performed before etching of the samples. The etched
cross-sections were examined with light optical microscopy
(LOM).
3.
Results
3.1.
Furnace experiment
The hardness for different heat treatment temperatures and

times are presented in Fig. 7. The results show that at temperatures of 500 C and 600 C the hardness is almost independent
of heat treatment time, i.e. the hardness is the same as for
the as-received material. For temperatures in the interval
7001200 C the resulting hardness depends both on time and
temperature.
In Fig. 8 the tensile test for a sample heat treated at 1100 C
for 60 s is presented. Tensile tests for all heat treatments over
1000 C and at different times have similar appearance. Thus,
the formability of the material has substantially increased
compared with as-received material, cf. Fig. 1. The microstructure of samples heat treated at 1200 C for 15 s and 120 s,
respectively, is shown in Fig. 9. As compared to the as-received
material, cf. Fig. 2, it can be seen that the material has started
to recrystallize already after 15 s and small austenitic grains
has started to grow. Longer heat treatment times give a substantial grain growth as illustrated in Fig. 9b.
3.2.
Fig. 7 Hardness of the material, HyTens 1600, after

different heat treatment temperatures and times.
Hardness proles in the centre and in a near surface region of

cross-sections through the partial heated zone for the P10, P20
Fig. 8 Tensile test of the material, HyTens 1600, after a

heat treatment at 1100 C for 60 s.
Fig. 6 Different methods to form a V-section. In the

constant radius method the tool radius is the same in all
forming passes while in the constant arc length method
the arc length is the same. In the experiments tools with
constant radius have been used.
and P30 settings, respectively, is presented in Fig. 10. For the

P10 setting, see Fig. 10a, the hardness prole indicates that
the material is heat affected in a 2.2 mm wide zone. The
heated zone consists of a transition region were the hardness
decreases. The transition region is about 0.6 mm wide on both
sides of a middle region, approximately 1 mm wide, where the
hardness values are at the same level. In the middle region the
hardness in the centre of the material is softer, 270300 HV0.5 ,
than the material near the surface, 380420 HV0.5 . For the P20
and P30 samples the transition region has the same width as
for the P10 sample. The difference between the samples, i.e.
3.3.
3121
Roll forming experiment
The as-received and the partially heated materials have been

roll formed to 120 , in steps of 15 , 30 , 60 , 90 , 105 and
120 . The as-received material could be roll formed to 30
but ruptured in the bending zone at 60 . The partially heated
materials could be roll formed to 120 without rupture, see
Fig. 14. Micro fractures were visible in the bending zone for
the P10 sample after forming to a bend angle of 105 .
The hardness in the bending zone after forming to 90 , 105
and 120 for the P30 sample is presented in Fig. 15. The hardness in the bending zone has increased in the outer and inner
radius to 550600 HV0.5 . In the neutral layer has the hardness
increased 50100 HV0.5 as compared to the not roll formed
partially heated P30 sample, cf. Fig. 10c. The thickness reduc-
Fig. 9 LOM images of the microstructure of material heat

treated at a temperature of 1200 C and a time of 15 s (a) and
120 s (b). The image widths are 150 m (a) and 375 m (b),
respectively.
different power input, is that the higher power input the wider
is the middle region and the hardness in the centre and near
the surface approaches the same hardness values. For the P20
sample, see Fig. 10b, the hardness is 250270 HV0.5 in the centre
and 300320 HV0.5 at the near surface. For the P30 sample the
hardness is 260 HV0.5 in both the centre and at near surface
and the middle region is around 2.2 mm wide, see Fig 10c.
Fig. 11 shows LOM images of etched cross-sections of the
partially heated material for the P10, P20 and P30 settings,
respectively. It can be seen that the heating zone starts in the
centre of the material and grow towards the surface when the
power increases, which also is apparent from the hardness
proles in Fig. 10.
Figs. 12 and 13 show LOM images of the microstructure in
the transition and in the middle regions, respectively, for the
P30 sample. It can be seen that the material has started to
recrystallize in the transition region although the austenite
grains are barely visible. In the middle region the material is
fully recrystallized and the characteristic structure of austenite grains is seen.
Fig. 10 Hardness proles from the power settings (a) P10,

(b) P20 and (c) P30 in the centre and at the near surface
region of cross-sections of partially heated material.
3122
Fig. 13 The microstructure in the middle region. The

image width is 150 m.
Fig. 11 Cross-sections of partially heated samples P10,
P20 and P30. The image width is 3 mm.
of heating and cooling is very rapid. The furnace experiments
give knowledge about hardness and microstructure changes in
the material due to different times and temperatures. When
comparing the partially heated material (see Fig. 10) with furnace experiments (see Fig. 7) it was found that the hardness
and the microstructure (see Fig. 9a) for a furnace experiment at
1200 C for 15 s (see Fig. 13) agree most with the middle region
in all partially heated samples. This implies that the temperature in the middle region must at least reach a temperature
of 1200 C. Furthermore, there is no indication of melting, i.e.
a temperature over 1400 C, thus it can be estimated that the
temperature in the middle region is between 1200 and 1400 C.
It should be noted that the furnace and the partially heated
experiments are not on the same time scale, but the results
clearly show that a relevant comparison is possible. Although
the partially heated material reaches a temperature were one
Fig. 12 The microstructure in the transition region. The

image width is 150 m.
tion caused by roll forming is about 10% in the bending zone

for all samples as compared to the original thickness. This, in
combination with an increasing hardness in the neutral layer,
indicates that combined bending and tension has occurred.
Fig. 16 shows the hardness response due to roll forming
to a bend angle of 120 for the P10, P20 and P30 samples,
respectively. It can be seen that due to a lower power input the
hardness prole is displaced to higher values for the P10 and
P20 samples as compared to the P30 sample. The behaviour is
similar for the other bending angles.
4.
Discussion
The velocity, 1 m/min, used to partially heat the material

implies that the time the material is in contact with the heating wheels is very short, for example during 1 s the material
has travelled 16 mm, thus it can be suspected that the moment
Fig. 14 The as-received material ruptured in the 60

forming step. The partially heated material can be roll
formed to 120 without rupture.
Fig. 15 The hardness change of the partially heated

cross-section due to bending to angles of 90 , 105 and 120
for sample P30.
can suspect sensitization to occur it is not probable in this

case due to the very short heating time, i.e. fractions of a second, which Trillo et al. (1995) also conclude in their study of the
combined effect of deformation, grain size and carbon content
on carbide precipitation and corrosion sensitization.
The recrystallisation starts in the centre of the material and
almost reaches the surface for the P30 sample, cf. Fig. 11, which
is also seen in the hardness proles in Fig. 10. For power inputs
lower than P30 it can be seen that the hardness increases
towards the surface, i.e. the formability decreases. In a bending operation, as in roll forming, the bending deformation
will be largest in the outer (tensile strain) and inner radius
(compress strain). This implies that as high power input as
necessary should be used without reaching the melting temperature or that surface damage occur to reach full formability.
However, depending on the wanted bending angle a lower
input can be used. For example, using the power settings for
the P10 sample is enough for a bending angle of 90 but not
for 105 where some micro fracture appeared.
Fig. 15 shows that the hardness is almost independent
of bending angle when considering the three last forming
steps, i.e. 90 , 105 and 120 . The hardness in the neutral
layer has increased about 50100 HV0.5 as compared to the
not roll formed partially heated samples. These observations
can be derived to the used bending method, i.e. the constant
3123
radius method where it is a constant distance from the bending point to the arc centre in all forming steps. This method
will bend a part of the arc length in every forming step to its
nal shape. This means that the middle part of the arc length
is nished already in the 15 forming step. The method will
also give a superimposed tension on the bending. Therefore,
the hardness is almost the same for all bending angles since
the forming has occurred already in the rst forming step.
As a consequence of the combined bending and tension this
bending method will cause thinning of the sheet metal in the
bending zone. In this case, the thickness reduction was about
10%. The increased hardness in the neutral layer veries the
superimposed tension. If the constant arc length method
(see Fig. 6), which not superimpose tension in the material to
the same extent, had been chosen it probably would have been
possible to roll form the material with a lower power input to
a larger bending angle.
In this study, the only parameter that is varied is the power
input. Other parameters such as thickness and yield strength
of the material, and the velocity of the heating etc. can be varied. It can be expected that these parameters also will have an
inuence on the resulting forming properties. However, this
study has shown that by partially heating a cold rolled TRIP
steel the formability properties has increased substantially
thus making it suitable for roll forming.
5.
Conclusions
In the present study, partial heating of cold rolled TRIP steel

has been investigated. The main conclusions are:
Partial heating substantially increases the formability.
Too much power input causes surface defects.
Depending on required bend angle, the power input should
be adjusted.
The partially heated material shows three different regions,
unaffected-, transition- and a recrystallized region. The
recrystallization starts in the centre and grows with higher
power input.
The material will work hardening almost to the as-received
condition in the outer and inner radius of the bending zone
during roll forming. The neutral layer will work hardening
moderately.
Acknowledgements
The authors wish to thank ORTIC AB, The Knowledge
Foundation (KK-stiftelsen), Jernkontoret, Triple Steelix and
Outokumpu Stainless AB for their technical and nancial support.
references
Fig. 16 The hardness change of the partially heated

cross-section due to a bending of 120 for sample P10, P20
and P30, respectively.
Chiang, K.F., 1984. Cold roll forming. M.Sc. Thesis. Mechanical

Engineering Department, University of Auckland.
Di Schino, A., Salvatori, I., Kenny, J.M., 2002. Effects of martensite
formation and austenite reversion on grain rening of AISI
304 stainless steel. J. Mater. Sci. 37, 45614565.
3124
Halmos, G.T., 2005. Roll Forming Handbook, rst ed. CRC Talylor
& Francis, New York.
Johannsen, D.L., Kyrolainen, A., Ferreira, P.J., 2006. Inuence of
annealing treatment on the formation of nano/submicron
grain size AISI 301 Austenitic stainless steels. Metall. Mater.
Trans. A 37, 23252338.
Mszros, I., Prohszka, J., 2005. Magnetic investigation of the
effect of -martensite on the properties of austenitic
stainless steel. J. Mater. Process. Technol. 161, 162168.
Mumtaz, K., Takahashi, S., Echigoya, J., Kamada, Y., Zhang, L.F.,
Kikuchi, H., Ara, K., Sato, M., 2004. Magnetic measurements of
the reverse martensite to austenite transformation in a rolled
austenitic stainless steel. J. Mater. Sci. 39, 19972010.
Petit, B., Gey, N., Cherkaoui, M., Bolle, B., Humbert, M., 2007.
Deformation behavior and microstructure/texture evolution
of an annealed 304 AISI stainless steel sheet. Experimental
and micromechanical modelling. Int. J. Plast. 23, 323341.
Spencer, K., Embury, J.D., Conlon, K.T., Vron, M.T., Brchet, Y.,
2004. Strengthening via the formation of strain-induced
martensite in stainless steels. Mater. Sci. Eng. A. 387389,
873881.
Trillo, E.A., Beltran, R., Maldonado, J.G., Romero, R.J., Murr, L.E.,
Fisher, W.W., Advani, A.H., 1995. Combined effects of
deformation (strain and strain state), grain size, and carbon
content on carbide precipitation and corrosion sensitization
in 304 stainless steel. Mater. Charct. 35, 99112.
PAPER 7
3D roll-forming of hat-profile with variable depth and width
Michael Lindgren, Lars-Olof Ingmarsson
1st International Congress on RollForming, RollFORM09

14th -15th October. 2009, Bilbao, Spain
3D roll-forming of hat-profile with variable depth and width

Michael Lindgrencd
cORTIC AB
Rgker 47, Borlnge, SE-781 93, Sweden
dDepartment of Material Science, Dalarna University
SE-781 88, Borlnge, Sweden
Lars-Olof Ingmarsson
SWEREA IVF AB
Argongatan 30, SE-431 22, Mlndal, Sweden
Abstract
The use of roll-formed products in automotive, furniture, buildings etc. increases every year due to the low
part-production cost and the complicated cross-sections that can be produced. The limitation with
roll-forming until recent years is that one could only produce profiles with a constant cross-section in the
longitudinal direction. About eight years ago ORTIC AB [1] developed a machine in which it was possible to
produce profiles with a variable width (3D roll-forming) for the building industry. Experimental equipment
was recently built for research and prototyping of profiles with variable cross-section in both width and
depth for the automotive industry. The objective with the current study is to investigate the new tooling
concept that makes it possible to roll-form hat-profiles, made of ultra high strength steel, with variable
cross-section in depth and width. The result shows that it is possible to produce 3D roll-formed profiles with
close tolerances.
Keywords: 3D roll-forming, Variable cross-section, Flexible roll-forming, Profile, Ultra high strength steel.
1. Introduction
1.1 Background
Roll-forming is a sheet metal forming process where the forming occurs with rolls in several steps, often
from an undeformed sheet to a product ready to use. This is a highly productive process and the speed by
which the profiles can be formed is between 5-60 m/min depending on a second operation such as welding,
punching, etc that often is done in the same line. The use of the process increases due to the possibility to
produce complex products in material as ultra high strength steel.
The limitation with the process until eight years ago was that only a profile with constant cross-section
was possible to produce. At that time ORTIC AB [1] developed a method, 3D roll-forming, that could
produce panels to buildings where the cross-section was variable in the longitudinal direction, Figure 1.
Figure 1.
Roll formed panels with variable cross-section (3D roll-forming).
The method used is very flexible which means that panels with different geometry in the longitudinal
direction can be produced with the same set-up of roll-forming tools. The Budapest Arena for example is
covered with 4700 different individually shaped panels. Today many buildings worldwide have been covered
with panels produced using this method [2]. The success of the forming method has made other industries
interested.
The automotive industry is one area where 3D roll-forming is of great interest since the industry can utilise
the flexibility of the method together with the use of high strength steel with low part-production cost. A 3D
roll-forming experimental machine has been built to use for research and to fulfil the needs of prototypes for
the automotive industry. The difference compared to the machines for the building industry is that a profile
with variable depth can also be produced.
The objective with this study is to investigate a new tooling concept that makes it possible to roll form
hat-profiles with variable depth and width in longitudinal direction. To evaluate the new tooling concept
three different hat-profiles, one with constant cross-section and two with variable cross-section in depth and
width, are roll-formed and the tolerances from fifty profiles of each kind are compared.
1.2 Previous work

In 2001 ORTIC AB [1] developed a 3D roll-forming machine for conical profiles, Ingvarsson [3]. The
machine was built in a mobile container and moved around the world to different construction sites. The
machine was used for covering straight and circular roofs. In 2002 the technology was further developed, a
3D roll-forming machine, not only for conical profiles, and a curving mill were built for profiles with
curvature and variable width in longitudinal direction. This technology made it possible to cover for example
the Budapest arena and many other buildings world wide, [2].
Groche et. al. [4] described a new tooling concept for flexible roll-forming. A single flexible stand was
integrated in a conventional roll-forming line. Through tests with finite element simulations and experiments
it was shown that the flexible frame should be perpendicular to the bending edge to give a profile with
quality. A CAD-system was also developed where the bending edge can be drawn and read by the control
program for the machine.
Ona [5] studied the 3D roll-forming process with a single forming stand, that made it possible to rotate,
turn and move the tools in and out. The movements were depending on feed rate of the sheet metal, which
was measured with a rotary encoder. A U-profile with variable cross-section was studied. It was shown in the
experiments that the material in the edge of the flange were compressed or stretched, Figure 2, which gave
buckling or distortion, if the flange was too high. Ona also concluded that the increasing number of forming
steps would decrease the buckling and distortion.
Groche et. al. [6] developed an analytic one-step-model to use for the design of wrinkle free 3D roll
formed U-profiles. The model is semi-empirical and based on mechanics of buckling of plates and a series of
finite element analyses. The model focuses on the compressed area, Figure 2, and is used as a feasibility
check without simulation or experimental tests.
Gleceken et. al. [7] used COPRA RF [8] coupled with finite element module MSC.Marc [9] to
simulate the 3D roll-forming process. The objective with the study was to explore the process using finite
element simulation. The results from the simulations show end-flare in the back and front of the U-profile,
which is common for roll-forming of pre-cut material. Due to the programmability of the tools they
concluded that end-flare might be possible to compensate for by increasing the bending in the end and the
beginning of the profile. The simulations also show a variation in the leg height of r1 mm, they conclude it
could originate from relatively rough finite element mesh.
Figure 2. A 3D roll-formed U-profile. Tension stress acts in transition zone where the U-profile is small and compression stress in the
transition zone where the profile is wider.
2. Experimental procedure
2.1 The roll-forming machine and new tooling concept
A 3D roll-forming experimental machine is used to form profiles with variable depth and width. The
machine has slitter heads, to fit the metal sheet, and six forming stands where every stand has four units. The
units have servo control axis, two translations and two rotations axis. The unit can be moved up and down, in
and out, rotate and the speed of the tool can be controlled individually.
The geometry of the tools is simple, i.e. the tools are completely cylindrical. This also means that the
thickness of the material can vary without having to install new tools. The size of the machine decides which
thickness of the material that can be roll-formed.
In this study a hat-profile with variable depth and width has been roll-formed and for this type of profile
the new tooling concept demand two forming stands per pass, Figure 3. Forming stand number 1, 3 and 5
form the left side of the profile and the forming stand number 2, 4, and 6 the right side. For example the
profiles are roll-formed in six forming steps with bend angles 15q, 30q, 45q, 60q, 75q and 85q and therefore
the profile demands two laps in the machine, the first 15q-45q and then the control program of the machine is
switched to the next lap 60q-85q.
The tools are the same for all stands, for example in Figure 4 one can see that the tools for the left and
right side are the same for both 30q and 60q. The difference between the passes is that tools that hold the
flange have moved up and moved closer to the tools that hold the web. The used flower pattern is a
function of the length of the profile. This flexibility makes it possible to use as many passes as the
cross-section requires without making more tools. It also makes it possible to produce profiles with, not only
variable width, but also, variable depth in different thickness of the material.
Figure 3. View from the top. The profile is roll-formed in six passes and to do that the profile must go through the machine two laps.
Forming stand number 1, 3 and 5 formed the left side and forming stand number 2, 4, and 6 formed the right side.
Figure 4. View from the back. The geometry of the tools is same for all forming stands. The only difference between, for example,
bend angles 30 and 60 is that the tools for the flange are moving up and closer to the tools that hold the web. Tools with constant
radius have been used, Lindgren et. al.[10].
The horizontal distance between the forming stands is 400 mm. The production speed that is used during
the tolerance tests is 2.6 m/min. A simple input and run out table is used and the profiles are hand fed both
the first and the second lap. First all profiles with the same cross-section go through their first roll-forming
lap. Then is the control program switch to the second lap and all profiles of same cross-section are finished.
2.2 The profiles and material

The used material is ultra high strength dual phase steel, Docol 1000 DP. The thickness of 30 sheets has
been measured with micrometer and the mean value is 0.970 mm and with a standard deviation of 0.008 mm.
The width and the length of the as-received material are 400 mm and 1500 mm respectively. The slitter heads
in the beginning of the machine are used to give the blanks the right width.
Three hat-profiles with different cross-section have been studied, Figure 5, a straight, a conical in depth
and width and a profile with a waist on one side. The length, the thickness, the inner radius of the
cross-sections and the bend angles are the same for all profiles and they are 1500 mm, 1 mm, 2 mm and 85q
respectively. The number of hat-profiles, of each kind, that has been produced and measured are 50.
Figure 5. Three different hat-profiles have been roll formed.
2.3 The measurement equipment

An optical scanner, ATOS, [11], based on the principle of triangulation is used to measure the variation of
the profiles. The accuracy of the measurement equipment and the fixture are tested and the variation is less
than 0.07 mm.
The fixture positions the hat-profile in the web, z-direction see Figure 6, with three reference points and
one support point, the distance between the points in the longitudinal direction is 1300 mm. In the y-direction
the profile positions with two reference points and two support points and in the x-direction one reference
point is used.
Figure 6. Points (light dots) on the flanges (F), at the sides (S) and in the bottom (B) are measured and this is done in eleven
different cross-sections in the longitudinal direction.
Eleven different cross-sections in the longitudinal direction have been measured, five cross-sections in the
middle part of the hat-profile and three cross-sections in the ends, Figure 6. The distance between the
cross-sections is 100 mm. In every cross-section eleven points are measured, two on respective flange, two
on respective side and three points in the web, notations for the points see Figure 6.
3. Result and discussion

3.1 Result
In the study 50 hat-profiles of each type, Figure 7, have been measured. The results from these
measurements are compared to see if the new tooling concept works and can produce profile with same
tolerances both for straight ones and ones with variable depth and width.
The variations between the measured points are presented with a range and standard deviation plot, Figure
8 Figure 10. The surfaces in these plots are a function of position of the measured points (light dots) in
Figure 6. In this figure the coordinate system is also defined for the profiles.
Figure 7. The produced hat-profiles. The top profile is straight, the middle one has a waist on one side and the bottom is a conical
profile in both depth and width.
In Figure 8 the range, the difference between the maximum and minimum value, and the standard
deviation are presented for the straight hat-profile. The result shows that most points are below 1.2 mm in
range and a standard deviation of 0.3 mm or less. The maximum value is on the right flange, 1.46 mm, with a
standard deviation of 0.26 mm. The web has less variation than the flanges and the maximum value is 0.78
mm with a standard deviation of 0.19 mm. It can also be seen that the profiles have low variation in the
points, (B7,-600) and (B5, -600), this is where the profiles are fixed to the measurement fixture.
Figure 8. To the left is the range and to the right is the standard deviation for each point in longitudinal direction of the straight
hat-profile. The coordinate system, x = longitudinal direction, y = measured point, see Figure 6.
The result for the conical hat-profiles in depth and width, Figure 9, shows that the range for most of the
points is below 1.2 mm and with a standard deviation less than 0.3 mm. The highest value (range) is 1.52
mm with a standard deviation is 0.31 mm. This point is located in the left flange in the beginning of the
profile.
In Figure 10 the range and standard deviation are presented for the hat-profile with a waist on one side.
The range is less than 1.2 mm and the standard deviation is less than 0.3 mm for almost every point. The
maximum value is on the right flange in the beginning of the profile. The value is 1.67 mm and the standard
deviation 0.33 mm.
Figure 9. To the left is the range, the difference between the maximum and minimum value, and to the right the corresponding
standard deviation for the hat-profile with conical width and depth. The coordinate system, x = longitudinal direction, y = measured
point, see Figure 6.
Figure 10. The measured result for the hat-profile with a waist on one side is presented. To the left is the range, the difference
between the maximum and minimum value, and to the right is the corresponding standard deviation. The coordinate system, x =
longitudinal direction, y = measured point, see Figure 6.
3.2 Discussion
The focus in the current study is to investigate the new tooling concept and see if the cross-section
tolerances are similar for different types of 3D roll-formed profiles.
The profiles can be dived into three types:
1. Straight profiles. The tooling is not translating or rotating in z-direction (Figure 6) during the
forming. This is the same as in traditional roll-forming.
2. Conical profiles in depth and width. The tooling translates in z- and y-direction (Figure 6). This is
only still a bending in longitudinal direction. Then the residual stresses are similar as for straight
profiles
3. Profiles with transition zones, similar to the profile in Figure 3. The tooling will translate and rotate
in all direction during the forming.
It can be expected to be more difficult to fulfil tolerance requirements when forming conical profiles than
for straight profiles. However, the study shows that the tolerances are at the same level, Figure 8 and Figure
9. The range for most points is under 1.2 mm with a standard deviation of less than 0.3 mm. This means that
only translation of the tools does not make it more difficult to fulfil the tolerances.
Comparing the result of the forming of the profile of the third type with the straight profile show that the
tolerances are also in this case about the same in level, Figure 8 and Figure 10. The difference, apart from
that the tools translate and rotate in all directions, is that the profile will get residual stresses after forming
completely different from profile 1 and 2. This is due to complex material flow in this forming process and it
may warp the web and the flange, Figure 11. In the study the goal was to get a flange that was in the same
plane throughout the complete profile. To do that the length of the leg has been made longer in this area, see
Figure 12.
Figure 11. The complex material flow warps the web and the flange so the flange is not on the same plane in the longitudinal
direction.
Figure 12. In the study was a goal to get the flange in the same plane in the longitudinal direction. To do that the leg of the profile
was made longer in the warping zone. This is possible with the new tooling concept.
The tool concept requires one to hold on to the flange and the web in each forming station, Figure 4. To
be able to do so the flange has to have a certain width so that the material does not slip away from the tool
which would lead to variations of the width of the flange. During the test it has been noted that a width of at
least 10 mm is needed to avoid this problem.
4. Conclusion
Probably one of the first 3D roll-forming machines in the world with two translational and two rotational
degrees of freedom per axis has been built and used in the current study. Based on a specific tooling concept
hat-profiles with variable cross-section in depth and width has been produced using simple cylindrical
shaped rolls.
The main conclusions are:
x The new tooling concept makes it possible to roll-form hat-profiles with a variable cross-section
both in depth and width with tolerances on the same level as for straight profiles.
x The tools make it possible to roll-form different thickness of the material with only software
changes due to their simple, cylindrical shape.
x The flange of the hat-profile must be at least 10 mm to avoid that the flange slips in the tools and
thereby cause variations of the width
x Three different profiles have been identified in the study, type 1, which are completely straight,
type 2 are conical profiles and type 3, profiles with transitions zones. The first two give similar
residual stresses in the longitudinal direction and they are easy to produce. Type 3 gets residual
stresses that can give wavy edges, distortion of the web and flange, these profiles require careful
design of the process.
5. Reference
[1] ORTIC AB, (www.ortic.se)
[2] BEMO SYSTEM, (www.bemo.com)
[3] L. Ingvarsson: Innovativa stlprodukter-praktikfall baserad p rullformningstekniken, In: Proc. Stl
2004, May 2004, Borlnge, Sweden
[4] P. Groche, G. von Breitenbach, M. Jckel, A. Zettler: New tooling concepts for future roll forming
applications. In: Proc. ICIT 2003, Celje, Slovenia
[5] H. Ona: Study on development of intelligent roll forming machine, In: Proc. 8th ICTP 2005, Verona,
Italy
[6] P. Groche, A. Zettler, S. Berner: Development of a one-step-model for the design of flexible
roll-formed parts, The 9th International Conference on Material Forming; Glasgow, United Kingdom
April 26 - 28, 2006
[7] E. Gulceken, A. Abe, A. Sedlmaier and H. Livatyali: Finite element simulation of flexible roll
forming: A case study on variable width U channel. 4th International Conference and Exhibition on
Desing and Production of MACHINES and DIES/MOLDS, Cesme, TURKEY, 21-23/6/2007
[8] COPRA RF Software, (www.datam.com)
[9] MSC.Marc, (www.mscsoftware.com)
[10] M. Lindgren, U. Bexell, L. Wikstrm: Roll Forming of partially heated cold rolled stainless steel,
Journal of materials processing technology 209 (2009) 3117-3124
[11] gom, optical measuring techniques, (www.gom.com)
Acknowledgements
The author wishes to thank ORTIC AB, Swedish Knowledge Foundation, Jernkontoret, Dalarna University,
SSAB, VOLVO, SAAB, Bendiro and Swerea/IVF for their technical and financial support.

Experimental and Computational Investigation of The Roll Forming Process

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Experimental and Computational Investigation of The Roll Forming Process

Uploaded by

Copyright:

Available Formats

DOC TOR A L T H E S I S

ISSN: 1402-1544 ISBN 978-91-7439-031-5

Division of Material Mechanics

Experimental and Computational Investigation

Experimental and Computational Investigation of the Roll Forming Process

Experimental and Computational Investigation

Lule University of Technology

Printed by Universitetstryckeriet, Lule 2009

The doctoral project has

Author contribution to the publications

Distance from bend

Roll formed products have many applications in the automotive

u is the distance from the edge of a at strip to the edge of the

The used tolerance distance D1 and D2 in paper 5 is 5 % of the

2 The roll forming process

7 Discussions and conclusions

Design of the roll forming process

Research Question and Approach

The research question of the work can be stated as:

The roll forming process

Traditional roll forming

In roll forming bending is done in several steps, beginning with an undeformed

The roll forming process

2.2 3D roll forming

The roll forming process

2.2 3D roll forming

The roll forming process

Longitudinal membrane strain

Bhattacharyya and Smith [4] investigated the longitudinal strain by use of

3.1 Experimental work

Average peak strain (103 )

Figure 3.3. Deformation length for one forming step.

Their conclusions were as follows:

3.2 Theoretical work

Their conclusions were as follows:

Figure 3.5. The forming angle method.

Number of forming steps

3.2 Theoretical work

3.3 and 3.4

the notations used to calculate the machine design.

Horizontal distance between the forming stands

HC 2VC (pre-cut material).

HC 2.5VC (post-cut material).

3.2 Theoretical work

Number of forming steps.

where (r) is the distance from the bend.

where (r0 ) is the radius to the neutral layer.

3.2 Theoretical work

Figure 3.9. An element strip between two roll forming passes.

Figure 3.10. View in z-direction.

where (Y) is the yield strength.

The total work done for one bend is

the general solution is

3.2 Theoretical work

Longitudinal membrane strain

The transverse longitudinal strain at any position z is written:

Geometrical restriction from the female tool

3.2 Theoretical work

Roll forming of high strength steel

3.2 Theoretical work

Computer simulation of roll forming

Kiuchi et al. [30] introduced a normalized shape function, equation 3.32, to

3.3 Computer simulation of roll forming