materials

Article
Estimating of Bending Force and Curvature of the Bending
Plate in a Three-Roller Bending System Using Finite Element
Simulation and Analytical Modeling
Ionel Gavrilescu 1 , Doina Boazu 1 and Felicia Stan 2, *

1 Department of Mechanical Engineering, Dunarea de Jos University of Galati, Domneasca 47,

800008 Galati, Romania; ionel.gavrilescu@ugal.ro (I.G.); doina.doazu@ugal.ro (D.B.)
2 Center of Excellence Polymer Processing, Dunarea de Jos University of Galati, Domneasca 47,
800008 Galati, Romania
* Correspondence: felicia.stan@ugal.ro






Citation: Gavrilescu, I.; Boazu, D.;
Stan, F. Estimating of Bending Force
and Curvature of the Bending Plate in
a Three-Roller Bending System Using
Finite Element Simulation and
Analytical Modeling. Materials 2021,
14, 1204. https://doi.org/10.3390/
ma14051204
Abstract: Many industries such as shipbuilding require steel bending plates in a wide range of radii,
thus bending machines are often designed and produced on a custom basis in shipyards. From a
design perspective, however, the bending force and the radius of the bending plate as a function of
vertical displacement of the upper roller must be known. In this paper, a hybrid numerical–analytical
approach is proposed to investigate the three-roller bending process for two plates of steel used in
the naval industry. Firstly, the bending process is modeled using the finite element (FE) method and
regression models for the bending force as a function of plate thickness and vertical displacement
of the upper roller were constructed. Then, based on the findings from FE analysis, using the bent
bar theory, two analytical expressions for the bending force were derived. Using geometric and
deformation compatibilities, analytical expressions for the vertical displacement of the upper roller as
a function of the curvature of the bending plate were also developed. The FE results suggest that the
cross section of the plate is practically a plastic hinge in the tangent area of the upper roller and that
the deformation compatibilities must be considered in order to estimate the curvature radius of the
bending plate using analytical formulations. These results are of practical importance in designing
rolling machines to estimate the setting parameters.
Keywords: three-rolling bending process; finite element analysis; bending force; vertical displacement

Academic Editor: Georgios Savaidis

Received: 3 February 2021 1. Introduction

Accepted: 25 February 2021 Cylindrical and conical steel shells are important components used in various engi-
Published: 4 March 2021
neering applications such as bilge planking as a part of the naval hull, cylindrical tanks,
pressure vessels, etc. [1]. Rolling machines with three or four rollers are generally consid-
Publisher’s Note: MDPI stays neutral
ered for manufacturing shells (bending plates) with different curvatures [1–4]. In order to
with regard to jurisdictional claims in
design a rolling machine, it is necessary to calculate the bending force and the curvature
published maps and institutional affil-
radius of the bending plate for given material properties and geometrical parameters.
iations.
Therefore, efforts have been devoted to the prediction of the bending force or curvature
radius by experimental, theoretical, and numerical approaches.
In general, the bending force is estimated using an analytical approach based on the
one-dimensional beam theory combined with experimental analysis for validation. For
Copyright: © 2021 by the authors. example, Chudasama and Raval established an analytical model for the prediction of the
Licensee MDPI, Basel, Switzerland.
bending force for the three-roller conical multi-pass bending process [4]. Furthermore,
This article is an open access article
Chudasama and Raval proposed an analytical model for force prediction in the three-roller
distributed under the terms and
conical bending process, taking into account the effects of various material properties and
conditions of the Creative Commons
geometrical parameters [5], which can also be used to estimate the roller plate-interface
Attribution (CC BY) license (https://
friction. Padgan et al. have conducted an analytical and experimental analysis of bending
creativecommons.org/licenses/by/
forces for five materials, such as aluminum alloy, copper alloy, stainless steel, gray cast iron,
4.0/).

Materials 2021, 14, 1204. https://doi.org/10.3390/ma14051204 https://www.mdpi.com/journal/materials

Materials 2021, 14, 1204 2 of 16

and magnesium alloy [6]. However, the analytical models are not able to fully describe
the complex deformation behavior of the materials during the bending process due to
oversimplified assumptions [2,7]. Therefore, in the attempt to optimize the roll bending
process, in the last years, the finite element method (FEM) was applied to simulate the
bending process and extract different information, such as bending force, the radius of
curvature, stress-strain state in the contact area under various conditions, and the finite
element (FE) environment [8–13]. For example, Tran investigated the forming process of an
asymmetrical roll bending machine using a three-dimensional (3D) dynamic FE model in
the Ansys/LS-DYNA environment and the results were validated by experiments [8]. Feng
and Champliaud reported on the numerical simulation of cylindrical roll bending to predict
the position of the lateral roll [9] and on the conical roll bending [10] using Ansys/LS-
DYNA environment with explicit time integration. Ktari et al. carried out two-dimensional
(2D) modeling of the rolling process in three roll bending pyramid systems using Abaqus
explicit module [11]. Taylor et al. investigated the influence of various parameters of
the three-roll bending process on the final radius of curvature of the bent sheet using an
elastoplastic explicit dynamic FE method under the LS-DYNA environment [12]. Neto
et al. developed an FE model to analyze the stress–strain in the vicinity of the contact area
where the plastic deformation increases due to the forming tool [13]. Moreover, several
attempts have been made to combine the FE simulation with the analytical modeling of
different bending systems. Patel et al. proposed a moment curvature-based model for
elastoplastic micro-beam bending to solve a micro-cantilever subjected to a normal load
undergoing large deflection [14]. Pandit and Srinivasan presented an explicit numerical
approach for three-point bending of a thin elastoplastic beam undergoing large deflection
supported on cylindrical rollers with a radius comparable to the deflection [15]. Using both
analytical and finite element approaches, Zemin et al. studied the three-roll bending of a
large workpiece using a model based on one dimension beam theory to determine the inner
radius displacement and the Abaqus FE model for optimizing the process parameters [16].
It was shown that, in general, the FE analysis provides more accurate results for the
bending force and radius of the curvature compared with the experimental data than the
oversimplified analytical models based on bending beam theory.
In summary, even though a number of researches have been carried out to improve
the roll bending technology, estimation of the bending force and the curvature radius of
the bending plates as a function of the displacement of the upper roller is still difficult at
this stage, especially for large sheets and mainly relays on expensive physical experiments.
Thus, the three-roller bending process needs further investigation both numerically and
analytically. Moreover, the FE analysis, in combination with existing analytical theories,
may offer a cheap and easy alternative to the trial-and-error, experimental approach.
The goal of the paper is to derive simple yet effective analytical expressions for
calculating the bending force and the vertical displacement of the upper roller as a function
of the radius of curvature, taking into account the findings from the FE analysis of the
three-roller bending process. The novelty of this paper consists of coupling the plastic
hinge condition observed during the FE analysis with the bent bar theory.

2. Numerical Simulation of the Three-Roller Banding Process

2.1. Research Methodology
Considering the complexity of the deformation during the rolling process, a hybrid
numerical–analytical approach is proposed to investigate the three-roller bending process,
as shown in Figure 1. First, the three-roller bending process is modeled using the FE
method and Ansys Workbench Static Structural program (version 19.0, 2019, Ansys, Inc.,
Canonsburg, PA, USA), under plane strain conditions. In addition, regression models for
the bending force as a function of plate thickness and vertical displacement of the upper
roller are derived for two types of structural steel commonly used in the shipbuilding
industry. Second, based on the findings from FE analysis, according to which the cross
section of the plate is practically a plastic hinge in the tangent area of the upper roller, two
 Materials 2021, 14, x FOR PEER REVIEW 3 of 16

the bending force as a function of plate thickness and vertical displacement of the upper
roller are derived for two types of structural steel commonly used in the shipbuilding
Materials 2021, 14, 1204
the bending force as a function of plate thickness and vertical displacement of the upper
industry. Second, based on the findings from FE analysis, according to which the 3cross
roller are derived for two types of structural steel commonly used in the shipbuilding
of 16

section
industry. of the platebased
Second, is practically a plasticfrom
on the findings hingeFE in analysis,
the tangent area of to
according thewhich
uppertheroller,
crosstwo
analytical
section of approaches for estimating
the plate is practically thehinge
a plastic bendingin theforce in different
tangent hypotheses
area of the (frictional
upper roller, two
and frictionless)
analyticalapproaches
analytical were derived
approaches for estimatingbased
estimating the on the
the bending bent
bendingforce bar theory and
forceinindifferent compared
differenthypotheses
hypotheses with the FE
(frictional
(frictional
results.
and Third, using
andfrictionless)
frictionless) were
were geometric
derived and
derived based
baseddeformation
on
on the
thebent
bentcompatibilities,
bar
bartheory
theoryandandanalytical
compared
compared expressions
with
withthethe for
FEFE
the vertical displacement
results. Third, using
results. of
usinggeometric the
geometricand upper
and roller
deformation
deformation as a function
compatibilities,of the radius
analytical
compatibilities, of curvature,
expressions
analytical for by
expressions
taking
thethe
for into
vertical
verticalaccount the
displacement thickness
displacement of the of
upper
of the the sheet,
roller
upper were
as a as
roller derived
function and verified
of theofradius
a function against
of curvature,
the radius the
by FE
of curvature,
results.
taking
by taking into account
into accountthethe
thickness
thicknessof the sheet,
of the werewere
sheet, derived and verified
derived against
and verified the FEthe
against
results.
FE results.

Figure
Figure1.
1. Flow
Flow chart
chart for the hybrid
for the hybridnumerical–analytical
numerical–analyticalapproach.
approach.

2.2.
2.2. Finite Element Model
2.2.Finite
Finite Element
Element Model
Model
Figure
Figure 22 shows a schematic representation of the three-roller bending process with
Figure 2 shows aa schematic
shows schematic representation
representationof ofthe
thethree-roller
three-roller bending
bending process
process with
with
identical
identical cylindrical
identical cylindrical rollers. This
rollers. This
cylindrical rollers. system
This system employs
systememploys one
employsone upper
oneupper roller
upperroller
rollerand two
and
and two
two lower rollers
lower
lower as
rollers
rollers
aasforming
asaaforming tool. The radius
forming tool. The radius of curvature
radius ofof curvatureof the
curvatureofofthebending
thebending plate
bendingplate is
platecontrolled
is is by
controlled
controlled changing
byby the
changing
changing
vertical
the position
thevertical
vertical of the
position
position upper
the upper
of the roller.
upper To To
roller.
roller. reach a desire
Toreach
reach (final)
a adesire
desire curvature
(final)
(final) radius,
curvature
curvature the
radius,
radius, upper
thethe
up-up-
roller
per moves
perroller down
roller moves in
moves down in a vertical direction
in aa vertical
verticaldirectionpressing
directionpressing down
pressingdown the
downtheplate.
the plate.
plate.

Figure 2. Schematic representation of the three-roller bending process (d – the outer diameter of
Figure
Figure 2.2. Schematic
Schematic representation
representationof
ofthe
thethree-roller
three-rollerbending
bendingprocess
process(d—the
(d – theouter
outerdiameter
diameterofofthe
the rollers, L – the distance between the lower rollers).
rollers, L—the
the rollers, L – distance between
the distance the lower
between rollers).
the lower rollers).
Three design parameters, i.e., the vertical displacement of the upper roller, the force
Three
Three design parameters, i.e., the vertical
vertical displacement of
of the upper roller, the force
applied to design parameters,
the upper i.e.,the
roller, and thetorsional displacement
moments at the the
axesupper
of theroller, theare
rollers, force
applied to the upper roller, and the torsional moments at the axes of the rollers,
applied to the upper roller, and the torsional moments at the axes of the rollers, are are required
for design a three-roller bending system according to the imposed specifications, such as
the plate thickness, final radius of the bending plate, and material properties.
To determine the design parameters for the three-roller bending system, in this paper,
a static finite element analysis was carried out using the Ansys Workbench Static Structural
program [17] with specific nonlinear settings. It should be pointed out that modeling large
plastic deformations, which is the case of the three-rolling process, is based on the use of
 Materials 2021, 14, 1204 4 of 16

aterials 2021, 14, x FOR PEER REVIEW

logarithmic deformations (Hencky), the Jaumann derivative of the Cauchy stress tensor,
4 of 16
and the von Mises flow condition [17].
The bending system, particularly used in the naval shipyards, consists of three iden-
tical rollers with an outer diameter of 189 mm and a distance between the centers of the
required for design
lower a three-roller
rollers of 320 mm. bending systemstudy
A previous according to the imposed
[18] indicated that, specifications,
for the displacements of
such as thetheplate thickness, final radius of the bending plate, and material
upper roller equal to the roller radius, the applied force increases asymptotically and, properties.
To determine
therefore, thethedesign
moments parameters for the
at the axes three-roller
of rollers increase bending system, in indicating
asymptotically, this pa- that the
per, a statictechnological solution may be inappropriate for the specified conditions. Static
finite element analysis was carried out using the Ansys Workbench
Structural program [17] with
To simulate thespecific
deformation nonlinear settings.
behaviors, It should
because be pointed
the analysis of theout thatthree-roller
entire
modeling large
bendingplastic deformations,
process which is the case
is very complicated, of the three-rolling
a two-dimensional process,
(2D) model is based
was considered.
on the use Because
of logarithmic deformations
the plate’s width to (Hencky),
thickness ratio the Jaumann
is more than derivative
200, andof the
withCauchy
the assumption
stress tensor, and the von
of uniform Misesalong
loading flow the condition
width [17].of the plate, the plane strain model is suitable. The
The bending
simulationsystem, particularly
process was broken useddown
in theinto naval shipyards,
three stages asconsists
a functionof three
of timeiden-
(as a pseudo-
tical rollersvariable),
with an outerwhichdiameter
are (i) stage of 1:
189Themm timeand a distance
varies between
in the range of 0–1thes centers
in whichofthe theupper roller
lower rollershasofan imposed
320 vertical downward
mm. A previous study [18] displacement;
indicated that,(ii) forstage 2: The time varies
the displacements of thein the range
upper roller of equal
1–2 s in towhich the rollers
the roller radius,havethe an imposed
applied forcerotation;
increases andasymptotically
(iii) stage 3: Theand, time varies in
the moments
therefore, the range of 2–3 s in axes
at the which ofthe upper
rollers roller has
increase an imposed vertical
asymptotically, indicatingupward displacement.
that the
Figure may
technological solution 3a shows the 2D planefor
be inappropriate strain
the model
specifiedfor conditions.
the three-roller bending process corre-
sponding
To simulate to simulationbehaviors,
the deformation stages 1 and 2. In addition,
because the analysis a symmetrical
of the entire 2D plane strain model
three-roller
bending process is very complicated, a two-dimensional (2D) model was considered.was
corresponding to stage 1 (i.e., the vertical displacement of the upper roller) Be- developed
to gainwidth
cause the plate’s information that could
to thickness ratiobeisused
moreinthan the analytical
200, and approach (on a symmetrical
with the assumption of model)
for thealong
uniform loading bending theforce
width calculation
of the plate, (Figure 3b). Instrain
the plane this particular case, the center
model is suitable. The sim- of the bottom
rollerwas
ulation process wasbroken
fixed, and down theintoforcethree
is applied
stages in as the center ofofthe
a function timeupper
(as aroller
pseudo-along the axis
of symmetry.
variable), which are (i) stage 1: The time varies in the range of 0–1 s in which the upper
The models
roller has an imposed verticalwere drawn indisplacement;
downward the xy plane using the Design
(ii) stage 2: The timeModeler
varies(version
in the 19.0, 2019,
range of 1–2 s in which the rollers have an imposed rotation; and (iii) stage 3: The time [17]. The
Ansys, Inc., Canonsburg, PA, USA) included in the Static Structural module
varies in theorigin
rangeof the reference
of 2–3 s in whichsystem theis upper
in the center of the
roller has anupper
imposed roller, whereas
vertical the x and y axes
upward
displacement.define the plane (in plane strain), where x is the horizontal axis and y is the vertical axis.
The plate was
Figure 3a shows the discretized
2D plane strain withmodel
2D eight-node quadrilateral
for the three-roller higherprocess
bending order elements
cor- (type
PLANE 183 (Ansys, Inc., Canonsburg, PA, USA)
responding to simulation stages 1 and 2. In addition, a symmetrical 2D plane strain model [17], with large deflection and contact
correspondingconditions
to stageincluded. These elements
1 (i.e., the vertical displacement account forupper
of the the nonlinearities introduced by the
roller) was developed
nonlinear behavior of the material, large displacements,
to gain information that could be used in the analytical approach (on a symmetrical and the use of contact elements
between the roller surfaces and the plate [17]. The finite element size was 1 mm, resulting
model) for the bending force calculation (Figure 3b). In this particular case, the center of
in 10 elements with that thickness for the 10-mm plate thickness in Figure 4. A total number
the bottom roller was fixed, and the force is applied in the center of the upper roller along
of 15,000 elements were used to model the plate.
the axis of symmetry.

(a)

Figure 3. Cont.
 Materials 2021, 14, 1204 5 of 16
rials 2021, 14, x FOR PEER REVIEW 5 of 16

(b)
Figure 3. (a) A two-dimensional (2D) model for the three-roller bending process for stages 1 and 2 and (b) a 2D symmetric
model for stage 1.

The models were drawn in the xy plane using the Design Modeler (version 19.0, 2019,
Ansys, Inc., Canonsburg, PA, USA) included in the Static Structural module [17]. The
origin of the reference system is in the center of the upper roller, whereas the x and y axes
define the plane (in plane strain), where x is the horizontal axis and y is the vertical axis.
The plate was discretized with 2D eight-node quadrilateral higher order elements (type
PLANE 183 (Ansys, Inc., Canonsburg, PA, USA) [17], with large deflection and contact
conditions included. These elements account for the nonlinearities introduced by the non-
linear behavior of the material, (b)large displacements, and the use of contact elements be-
Figure 3. (a) A two-dimensional (2D) modeltween the
for the roller surfaces
three-roller and the
bending process plate [17]. 1The
for stages finite
2 and element
andthree-roller size was 1 mm, resulting in
(b) abending
2D symmetric
Figure 3. (a) A two-dimensional (2D) model for the process for stages 1 and
model for stage 1. 10 elements with that thickness for the 10-mm plate thickness in Figure 4. A total number
2 and (b) a 2D symmetric model for stage 1.
of 15,000 elements were used to model the plate.
The models were drawn in the xy plane using the Design Modeler (version 19.0, 2019,
Ansys, Inc., Canonsburg, PA, USA) included in the Static Structural module [17]. The
origin of the reference system is in the center of the upper roller, whereas the x and y axes
define the plane (in plane strain), where x is the horizontal axis and y is the vertical axis.
The plate was discretized with 2D eight-node quadrilateral higher order elements (type
PLANE 183 (Ansys, Inc., Canonsburg, PA, USA) [17], with large deflection and contact
conditions included. These elements account for the nonlinearities introduced by the non-
linear behavior of the material, large displacements, and the use of contact elements be-
tween the roller surfaces and the plate [17]. The finite element size was 1 mm, resulting in
10 elements with that thickness for the 10-mm plate thickness in Figure 4. A total number
of 15,000 elements were used to model the plate.

Figure 4. Discretization of the three-roller bending model.

Figure 4. Discretization of the three-roller bending model.

The contact between the plate and the rollers was considered frictional in the sense of
Coulomb with a coefficient of friction of 0.3. In the 2D model, for the surface-to-surface
contact, the “contact” and “target” elements are CONTA192 (Ansys, Inc., Canonsburg, PA,
USA) and TARGE169 (Ansys, Inc., Canonsburg, PA, USA) for contact bodies and target
bodies, respectively. The augmented Lagrangian formulation with asymmetric behavior
was considered and the small sliding was considered off. The model worked without
activating the interface treatment Adjust to Touch [17].
The simulation of the bending process was performed by increasing the vertical
displacement of the upper roller from 0 to the maximum physical displacement imposed
by the geometry
Figure 4. Discretization of the system
of the three-roller (seemodel.
bending Table 1). However, to avoid numerical instability (i.e.,
 PA, USA ) and TARGE169 (Ansys, Inc., Canonsburg, PA, USA) for contact bodies and
target bodies, respectively. The augmented Lagrangian formulation with asymmetric be-
havior was considered and the small sliding was considered off. The model worked with-
out activating the interface treatment Adjust to Touch [17].
Materials 2021, 14, 1204 The simulation of the bending process was performed by increasing the vertical 6 ofdis-
16

placement of the upper roller from 0 to the maximum physical displacement imposed by
the geometry of the system (see Table 1). However, to avoid numerical instability (i.e., the
applied
the applied force increases
force asymptotically
increases asymptotically with increasing
with increasingthethe
radius of curvature
radius of curvature[18]), the
[18]),
vertical displacement was limited to 50 mm. The simulations were carried
the vertical displacement was limited to 50 mm. The simulations were carried out for two out for two
typesofofstructural
types structuralnaval
navalsteels,
steels,for
forwhich
whichananelastoplastic
elastoplasticmaterial
materialmodel
modelwith
withisotropic
isotropic
hardeningwas
hardening wasconsidered,
considered,asasininTable
Table1.1.ItItshould
shouldbebenoted
notedthat
thatthe
thespring-back
spring-backofofthe the
bendingplate
bending platewas
wasnot
nottaken
takeninto
intoaccount,
account,andandthetherollers
rollerswere
wereassumed
assumedtotobebedeformable
deformable
ininthe
thelinear-elastic
linear-elasticdomain.
domain.

Table1.1.Material
Table Materialproperties
propertiesand
andmodel
modelparameters.
parameters.

Parameters
Parameters S235JR
S235JR S275JR
S275JR
Yield
Yield strength(MPa)
strength (MPa) 235235 275
275
Young
Youngmodulus
modulus(Pa)
(Pa) 2.1 2.1
× 10×111011 2.1 × 10
2.1 × 11
1011
Tangent
Tangent modulus(Pa)
modulus (Pa) 2.1 2.1 ×9 109
× 10 2.1 × 1099
× 10
Poisson ratio
Poisson ratio (-) (-) 0.3 0.3 0.3
0.3
Sheet thickness (mm) 8, 10, 12 8, 10, 12
Sheet thickness (mm) 8, 10, 12 8, 10, 12
Maximum vertical displacement (mm) 50 50
Maximum vertical displacement (mm) 50 50
2.3. Numerical Simulation Results
2.3. Numerical Simulation Results
Figure 5 shows the variation of the normal stress σX for the 10-mm plate thickness of
Figure 5 shows the variation of the normal stress for the 10-mm plate thickness
S235JR steel. The results illustrate that the normal stress in each point of the cross section
of S235JR steel. The results illustrate that the normal stress in each point of the cross sec-
in the tangent area of the upper roller is equal to the yield strength.
tion in the tangent area of the upper roller is equal to the yield strength.

Figure 5. Distribution of the normal stress (in Pa) for the 10-mm plate thickness (S235JR steel).
Figure 5. Distribution of the normal stress σX (in Pa) for the 10-mm plate thickness (S235JR steel).

ItItshould
shouldbe be noted
noted that
that the
the normal
normal stress,
stress, σX, ,isisguided
guidedalong alongthe Ox-axis
the Ox-axis that is ori-
that is
ented along
oriented alongthe theneutral
neutraldeformed
deformedsurface
surfaceofofthe
theplate,
plate,i.e.,
i.e.,inineach
eachsection,
section,thetheOxOxaxis
axisis
is“parallel”
“parallel”with withthe faces
the facesof of
thethe
deformed
deformedplate. Furthermore,
plate. Furthermore, it was found
it was that the
found thatdistri-
the
bution of stresses
distribution on aoncross
of stresses section
a cross depends
section depends ononthethemesh
mesh size when
size when the
thecalculation
calculationisisin
inthe
theelastoplastic
elastoplasticdomain.
domain.
The distributionofofnormal
The distribution normalstress
stressalong
alongthethecross
crosssection
sectionininthethetangent
tangentareaareaofofthe
the
upperroll
upper rollisisshown
shownininFigure
Figure6 6forforthe
theS275JR
S275JRsteel.
steel.AsAscan
canbe beseen,
seen,practically,
practically,thetheentire
entire
crosssection
cross sectionofofthe theplate
plateisisyielded,
yielded,indicating
indicatingthatthatthe
thedeformation
deformationstate statecorresponds
correspondstoto
the“plastic
the “plastichinge.”
hinge.”
Figure 7 shows the variation of the equivalent plastic strain in the cross section of
the 10-mm plate in the tangent area of the upper roller. As can be seen, for almost the
entire cross section, the plastic strain is greater than the σY /E ratio, which is 0.0013 for the
S275JR steel.
Figure 8a shows the deformation of the 10-mm plate thickness at the end of stage 2.
In order to estimate the final radius of the bending plate, the displacement of three points
located on the inner surface of the plate was monitored during the FE analysis, as shown in
Figure 8b. These points are approximately located on a circle with center (x, y) and radius r.
 Materials 2021, 14, x FOR PEER REVIEW 7 of 16

Materials 2021, 14, 1204 7 of 16

Materials 2021, 14, x FOR PEER REVIEW 7 of 16

Figure 6. Distribution of the normal stress (in Pa) in the tangent area of the upper roller
(S275JR steel, 10-mm plate thickness).

Figure 7 shows the variation of the equivalent plastic strain in the cross section of the
10-mm
Figure plate in the tangent
6. Distribution area ofstress
of the normal the upper
(in roller. Astangent
Pa) in the can bearea
seen,
of for
the almost the entire
upper roller
Figure
(S275JR
Distribution
6.steel, 10-mm
of the normal stress σX (in Pa) in the tangent area of the upper roller (S275JR
platestrain
thickness).
cross section, the plastic is greater than the /E ratio, which is 0.0013 for the S275JR
steel,
steel. 10-mm plate thickness).
Figure 7 shows the variation of the equivalent plastic strain in the cross section of the
10-mm plate in the tangent area of the upper roller. As can be seen, for almost the entire
cross section, the plastic strain is greater than the /E ratio, which is 0.0013 for the S275JR
steel.

Figure Equivalent
Figure 7.7.Equivalent
Materials 2021, 14, x FOR PEER REVIEW plastic
plastic strain
strain in the
in the tangent
tangent area area
of theofupper
the upper
roller roller
for thefor the plate
1-mm 1-mm plate thickness
thick- 8 of 16
ness (S275JR
(S275JR steel).
steel).

Figure 8a shows the deformation of the 10-mm plate thickness at the end of stage 2.
InFigure
order7.toEquivalent
estimate plastic
the final radius
strain oftangent
in the the bending
area ofplate, the displacement
the upper of three
roller for the 1-mm platepoints
thick-
ness (S275JR
located on the steel).
inner surface of the plate was monitored during the FE analysis, as shown
in Figure 8b. These points are approximately located on a circle with center ( , ) and ra-
dius r.Figure 8a shows the deformation of the 10-mm plate thickness at the end of stage 2.
In order to estimate
By using the final radius
the least-square method, ofthe
thecenter
bending
andplate, the displacement
the radius were determinedof three
by points
min-
located on
imizing the sum the inner surface of the plate was monitored during the FE analysis, as shown
in Figure 8b. These points are approximately located on a circle with center ( , ) and ra-
dius r. ( , , )=∑ ( − ) , (1)
By using the least-square method, the center and the radius were determined by min-
in which , = 1, 2, 3, is the radius corresponding to each point in Figure 8b, i.e., the
imizing the sum
distance from the coordinates of each point ( , ) to the center ( , ),
( , , )=∑ ( − ) , (1)
= ( − ) +( − ) . (2)
(a) in which , = 1, 2, 3, is the radius corresponding to each point (b) in Figure 8b, i.e., the
distance from the coordinates of each point ( , ) to the center ( , ),
Figure
Figure 8.
8. (a)
(a) Total
Total deformation (in m)
deformation (in m) at
atthe
theend
endofofstage
stage22and
and(b)
(b)location
locationofofthe
the points
points forfor
thethe estimation
estimation of of
thethe radius
radius for
for the 10-mm plate thickness of S275JR steel. = ( − ) +( − ) . (2)
the 10-mm plate thickness of S275JR steel.

Figure 9 shows
By using the variationmethod,
the least-square of the inner radius and
the center of the
theplate as awere
radius function of time for
determined by
three vertical displacements.
minimizing the sum It can be seen that the inner radius decreases with increasing
y, r ) = ∑
vertical displacement and tends Sto( x,stabilize with 3 increasing 2 vertical displacement.
i =1 ( r − r i ) , (1)
An approximate value for the inner radius related to a certain vertical displacement can
in obtained
be i =averaging
which ri , by 1, 2, 3 , isthe
thedata
radius
overcorresponding
time. Therefore, to each pointradius
the inner in Figure 8b,bending
of the i.e., the
distance from the coordinates of each point ( x i , y i ) to the center
plate was found to be 218.6 ± 4.8 mm, 158.4 ± 2.9 mm, and 117.6 ± 1.5 mm for 30 mm, 40 ( x, y ) ,
mm, and 50 mm vertical displacement, q respectively.
r i = ( x − x i )2 + ( y − y i )2 . (2)
 (a) (b)
Figure 8. (a) Total deformation (in m) at the end of stage 2 and (b) location of the points for the estimation of the radius
for the 10-mm plate thickness of S275JR steel.

Materials 2021, 14, 1204 Figure 9 shows the variation of the inner radius of the plate as a function of time for 8 of 16
three vertical displacements. It can be seen that the inner radius decreases with increasing
vertical displacement and tends to stabilize with increasing vertical displacement.
An approximate value for the inner radius related to a certain vertical displacement can
Figure 9by
be obtained shows the variation
averaging the data of thetime.
over innerTherefore,
radius ofthe
theinner
plateradius
as a function of time for
of the bending
three
plate vertical
was founddisplacements.
to be 218.6 ± It
4.8can
mm,be158.4
seen ±that
2.9 the
mm,inner
and radius
117.6 ±decreases
1.5 mm forwith increasing
30 mm, 40
vertical
mm, and displacement
50 mm verticaland tends to stabilize
displacement, with increasing vertical displacement.
respectively.

Figure 9.
Figure 9. Variation of the
the inner
innerradius
radiusofofthe
thebending
bendingplate
plateasasa afunction ofof
function time forfor
time different verti-
different vertical
cal displacements
displacements (10-mm
(10-mm plate
plate thickness
thickness andand S275JRsteel).
S275JR steel).

3. Analytical Modeling
An approximate of the
value forThree-Roller Bending
the inner radius Process
related to a certain vertical displacement
can
3.1. be obtained
Radius of the by averaging
Bending Plate the data over time. Therefore, the inner radius of the bending
plate The
wasradius
foundoftocurvature ± the
be 218.6 of 4.8 bending
mm, 158.4 ± (i.e.,
plate 2.9 mm, of the±neutral
and 117.6
the radius 1.5 mm for 30
layer) canmm,
40 mm, and 50 mm vertical displacement, respectively.
be determined based on the geometric compatibility of the three-roller bending system,
shown in Figure 10, as follows:
3. Analytical Modeling of the Three-Roller Bending Process
3.1. Radius of the Bending Plate = + , (3)

whereThe radius of curvature

is the inner radius ofofthe
the bending
bending plate
plate, and(i.e.,
t isthe
theradius of the neutral layer) can
plate thickness.
be determined based on the geometric compatibility of the three-roller bending system,
shown in Figure 10, as follows:
t
R = Ri + , (3)
Materials 2021, 14, x FOR PEER REVIEW 2 9 of 16
where Ri is the inner radius of the bending plate, and t is the plate thickness.

ϕ
R
Ri
w

t
y d/2

L
Figure 10. 10.
Figure Schematic representation
Schematic ofof
representation the
thegeometric
geometriccompatibilities
compatibilities of
of the
the three-roller bendingsystem.
three-roller bending
system.

By invoking the geometric compatibilities of the bending system, taking into account
that + = 1, the following relation can be derived:

+ = 1, (4)
Materials 2021, 14, 1204 9 of 16

By invoking the geometric compatibilities of the bending system, taking into account
that sinφ2 + cosφ2 = 1 , the following relation can be derived:
!2 !2
L
2 Ri + t + y
d
+ d
= 1, (4)
Ri + t + 2 Ri + t + 2

where y is the distance between the horizontal axis of the lower rollers to the deformed
outer layer of the plate and d is the diameter of the rollers.
After some mathematical manipulation and rearranging, the distance y reduces to
v
u !2 
L 
u
2 d
y = t1 − d
Ri + t + − Ri − t. (5)
Ri + t + 2
2

Based on Figure 10, the vertical displacement, w, of the upper roller can be written as

d
w= − y. (6)
2
Substituting Equation (5) in Equation (6), the final expression for the vertical displace-
ment as a function of the inner radius of the bending plate can be written as
v
u !2 
L 
d u 2 d
w = − t1 − d
Ri + t + + Ri + t. (7)
2 Ri + t + 2
2

The analytical model as in Equation (7), obtained only from geometrical compatibility
and without taking into account the deformation mode of the plate or the spring-back,
overestimates the inner radius of the bending plate as compared with the values predicted
by the FE analysis. For the vertical displacement of 30 mm, 40 mm, and 50 mm, the relative
error between the inner radius calculated by Equation (7) and FE values is 34%, 33%, and
33%, respectively. It should be pointed out that the greater the radius of the deformation,
the greater the effect of the spring back. Thus, taking into account the deformation behavior
from the FE analysis, which is similar to the deformation behavior of the plate during the
bending process presented in [19], a new approach was proposed for calculating the inner
radius as a function of the vertical displacement, as illustrated in Figure 11. It should be
noted that the parameters marked by “*” stand for the correction according to the finding
Materials 2021, 14, x FOR PEER REVIEW 10 of 16
from the FE analysis.

w* Internal surface of
ϕ1* the bending plate
t Ri* ϕ* Ri*

ϕ* d/2 ϕ1*
w* d/2

y*

L/2 L/2

Figure 11. Schematic representation of the deformation compatibilities of the three-roller bending
Figure 11. Schematic representation of the deformation compatibilities of the three-roller bending
system based on FE observations.
system based on FE observations.

Based on the new framework described in Figure 11, taking into account that
( ∗) ∗)
+( = 1, the following relation can be derived:
∗ ∗
∗ ∗ ∗
∗
+ ∗
= 1, (8)

where
Materials 2021, 14, 1204 10 of 16

2
Based on the new framework described in Figure 11, taking into account that sinφ1∗ +

2
cosφ1∗ = 1 , the following relation can be derived:
   2    2
L
2 − Ri∗ − d2 sinφ∗ y ∗ − 2d + Ri∗ − d2 cosφ∗
  +  = 1, (8)
Ri∗ + t + d2 Ri∗ + t + d2

where v
u !2
d u d
2 2
cosφ∗ = d
, sinφ∗ = t1 − d
. (9)
2 +t 2 +t
After some mathematical manipulation and rearranging, the expression for y* takes
the form
v
u    2
L ∗ − d sinφ∗
u
2 − R i 2

d

d

d

∗
cosφ∗ .
u
y∗ = 1 −
t   Ri + t + − − Ri − (10)
R∗ + t + d2
i
2 2 2

As before, the vertical displacement is

d
w∗ = −y∗. (11)
2
Substituting Equation (10) into Equation (11), in term of the inner radius, following
some manipulation of the resulting expression, the vertical displacement can be written as
v
u    2
L ∗ d ∗
2 − Ri − 2 sinφ
u    
 Ri∗ + t + d + Ri∗ − d cosφ∗ .
u
w ∗ = d − t1 −  (12)
Ri∗ + t + d2 2 2

Equation (12) allows for a more accurate prediction of the inner radius of the bending
plate as a function of vertical displacement.
Figure 12 compares the vertical displacement as a function of the inner radius of
the bending plate corresponding to the geometric compatibility (as in Equation (7)), the
deformation compatibility (as in Equation (12)), and the FE results, for the 10-mm plate
thickness. As can be seen, the geometric compatibility model overestimates the vertical
displacement, as compared with the deformation compatibility model, especially at the
higher inner radius in which the effect of the spring-back is very important. The relative
errors between the geometric model and deformation model are 45%, 35%, and 25% for
30 mm, 40 mm, and 50 mm vertical displacement, respectively. It can be seen that the higher
the inner radius is, the higher the relative error. On the other hand, the vertical displacement
of the upper roller can be predicted by Equation (12) relatively well. Compared with the
FE results, the relative errors of the deformation compatibility model are 19%, 2%, and 12%
for 30 mm, 40 mm, and 50 mm vertical displacement, respectively.

3.2. Bending Force

For estimating the bending force, in this study, the findings from the FE analysis were
considered, namely, (i) the normal stress for each point of the cross section reaches the yield
strength in the tangent area of the upper roller and (ii) the bending deformations of the
plate beyond the tangent area are small. Therefore, in order to estimate the bending force,
two scenarios were considered as given below.
Materials
Materials2021,
2021,14,14,
1204
x FOR PEER REVIEW 11
11ofof1616

Figure 12.Vertical
Figure12. Verticaldisplacement
displacementofofthe
theupper
upperroller
rollerasasa afunction
functionofofthe
theinner
innerradius
radiusofofthe
thebending
bending
plate with 10 mm thickness (S275JR steel).
plate with 10 mm thickness (S275JR steel).

3.2.1. Case 1 Scenario

3.2. Bending Force (C1 )
The
Forassumptions
estimating the are as follows:
bending force, (i)inthe plate
this hasthe
study, a finite thickness
findings from theandFEwidth; (ii) were
analysis the
friction
considered, namely, (i) the normal stress for each point of the cross section reaches of
plays an important role in the three-roller bending process, i.e., the coefficient the
friction betweenin
yield strength the rollers
the andarea
tangent plateofisthe
taken as 0.3;
upper andand
roller (iii)(ii)
thethe
bending
bendingmoment M p in theof
deformations
cross section
the plate of the the
beyond plate situated
tangent in the
area are tangent area of thein
small. Therefore, upper
orderroller corresponds
to estimate to the
the bending
so-called plastic hinge condition, i.e., the
force, two scenarios were considered as given below.normal stress for each point of the cross-section
becomes equal to the yield strength. Therefore, the bending moment can be expressed by
3.2.1. Case 1 Scenario (C1)
l × t2
The assumptions are as follows: M p(i)=the
σY plate has , a finite thickness and width; (ii) (13)
the
4
friction plays an important role in the three-roller bending process, i.e., the coefficient of
where is the yielding
σYbetween
friction strength,
the rollers l isisthe
and plate width
taken of the
as 0.3; andplate, and
(iii) the t is themoment
bending thickness ofin
the plate.
the cross section of the plate situated in the tangent area of the upper roller corresponds
The
to the initial configuration
so-called of the three-roller
plastic hinge condition, bendingstress
i.e., the normal system for iseach
given in Figure
point 13a,
of the cross-
while thebecomes
section configuration
equalattoa given timestrength.
the yield is presented in Figure
Therefore, the13b. From the
bending notations
moment can given
be ex-
inpressed
Figure by
13b, it follows that
d
t1 = z × cosφ1 − × − t, (14)
= 2 , (13)
and
where L
is the yielding strength, l is the width of the plate, and t is the thickness of the
sinφ1 + zcosφ1 − d − t = 0, (15)
plate. 2
takingTheintoinitial
accountconfiguration
that of the three-roller bending system is given in Figure 13a,
while the configuration at a given time z
tanαis1 presented
= L. in Figure 13b. From the notations (16)
given in Figure 13b, it follows that 2
Based on Equations (15) and (16), =
following
× some− manipulation,
− , it can be shown that
(14)
" #
and d+t
φ1 + α1 = arcsin L
× cosα1 . (17)
+ −
2 − = 0, (15)
Furthermore, with the geometry of Figure 13b, the function b is given by

L
b= cosφ1 − zsinφ1 , (18)
2
Materials 2021, 14, 1204 12 of 16
Materials 2021, 14, x FOR PEER REVIEW 12 of 16

F1

d/2 d/2 b t
µN
t ϕ1
Z0
Z N ϕ1
d/2 MP t1 d/2
ϕ1
L/2
(a) (b)

FigureFigure 13. Configuration

13. Configuration of the three-roller
of the three-roller bendingbending
systemsystem for1:Case
for Case 1: (a) initial
(a) initial positionposition of the rollers
of the rollers and undeformed
and undeformed
plate
plate and (b)and (b) position
position of the rollers
of the rollers and theand theatplate
plate at atime
a given given time
and theand the diagram
diagram of applied
of forces forces applied on the plate.
on the plate.

Bytaking
takinginto
intoaccount
accountthat
the forces represented in Figure 13b, the bending moment M p ,
which corresponds to reaching the “plastic hinge” condition in the cross section, can be
written as = .
2
(16)
l × t
M p = σY = N × 2b, (19)
4
Based on Equations (15) and (16), following some manipulation, it can be shown that
considering the balance of forces in the vertical direction (in the y-axis direction), which
implies that + = × . (17)
F1 = Ncosφ1 + µNsinφ1 , (20)
Furthermore,
and the bending force is with the geometry of Figure 13b, the function b is given by
simply
F = 2 × F1 . (21)
= − , (18)
3.2.2. Case 2 Scenario (C2 )
By taking into account the forces represented in Figure 13b, the bending moment ,
The assumptions
which corresponds areto
as reaching
follows: (i)
thefor estimating
“plastic hinge” thecondition
deformedinposition
the of section,
cross the plate,can be
Materials 2021, 14, x FOR PEER REVIEW 13 of 16
the thickness of the plate is negligible; (ii) the effect of friction is also negligible; (iii) the
written as
bending moment (Figure 14) corresponds to the high condition of plastic bending (i.e.,
×
the normal stress for each point of the cross-section= is=equal
× to
,the yield strength) and is (19)
given by Equation (13); and (iv) the moment in Section × 3 is zero. It should be noted that, in
Figureconsidering thelines
14, the dotted balance of forces
represent the in = verticalplate,
the
undeformed direction
, while(in thethe y-axis
solid line direction),
denotes the which
(23)
implies that
final position of the upper roller.
where b is the arm of the reaction force, and is the yield strength.
= + , (20)
F
and the bending force is simply
= 2 × . (21)
Initial position of
w α 2) b the plate
3.2.2. Case 2 Scenario (C
α
The assumptions 1are 3
2 as follows: (i) for estimating the deformed position of the plate,
N N
the thickness of the plate is negligible; (ii) the effect of friction is also negligible; (iii) the
bending moment (Figure 14) corresponds to the high condition of plastic bending (i.e., the
normal stress for each point of the cross-section is equal to the yield strength) and is given
by Equation (13); and (iv)L the moment in section 3 is zero. It should be noted that, in Figure
14, the dotted lines represent the undeformed plate, while the solid line denotes the final
Figure
Figure 14.14. Schematic
Schematic representation
representation of of
thethe three-roller
three-roller bending
bending system
system and
and thethe forces
forces applied
applied to
to the
position of the upper roller.
the sheet for
sheet for Case 2.Case 2.
Considering the equilibrium of the forces that act on the three-roller bending system
in Figure
Considering 14,thecontext,
In the present the bending
equilibrium forcethat
of the forces can act
be calculated by
on the three-roller bending system
in Figure 14, ×
= 2 × =2 . , (24)(22)
l × t
N × b = σY , (22)
the reaction force
Assuming N is
that the displacement of 4the upper roller can be approximated by
given by
vertical

−2 = , (25)

then, the angle α (in radians) is

Materials 2021, 14, 1204 13 of 16

the reaction force N is given by

l × t2 1
N = σY , (23)
4 b
where b is the arm of the reaction force, and σY is the yield strength.
In the present context, the bending force can be calculated by

F = 2 × Ncosα. (24)

Assuming that the vertical displacement of the upper roller can be approximated by

L d
− 2 α = w, (25)
2 2
then, the angle α (in radians) is
L
2−w
α= . (26)
d
Considering the three-roller bending system in Figure 14, the geometrical considera-
tion leads to
d L
2 sinα + bcosα = , (27)
2 2
and the arm b of the reaction force is given by
 
1 L
b= − dsinα . (28)
cosα 2

Substituting Equation (28) into Equation (23), the reaction force N becomes

l × t2 cosα
N = σY h i. (29)
4 L
− dsinα
2

The expression for N is then substituted into Equation (24) to obtain the bending force
in the following form:
l × t2 cos2 α
F = 2σY h i. (30)
4 L
− dsinα
2

4. Predictive Models for Bending Force

In order to determine a predictive model for the bending force, FE simulations were
carried out for different geometrical and material properties, as shown in Table 1. The FE
simulations were performed with the model described in Section 2 for vertical displacement
of the upper roller varying from 5 mm to 78 mm. The FE results for the bending force as a
function of vertical displacement of the upper roller and the sheet thickness are shown in
Figure 15 for S235JR and S275JR steels.
The results obtained from the FE simulations were used to determine a predictive
model for the bending force as a function of two variables, i.e., vertical displacement of
the upper roller and plate thickness. The FE results were fitted using Matlab curve fitting
toolbox (version R2018a, The MathWorks, Inc., Natick, MA, USA) [20,21].
The regression models for the bending force (in Newton) for the two plates of steel are
expressed by the following equations:
F (w, t) = 25.96 − 0.06384w + 0.0542 exp(0.001454 + 0.5549t) + 0.5711 w exp(−1.852 + 0.2448t) (31)
for the S235JR steel, and
F (w, t) = 33.56 − 22.08w + 0.008891 exp(−0.002553 + 0.7136t) + 1.123w exp(2.909 + 0.0128t) (32)
for the S275JR steel.
Materials
Materials 2021,
2021, 14,
14, x1204
FOR PEER REVIEW 1414of
of 16
16

(a) (b)
Figure 15. The bending
Figureforce from
15. The FE analysis
bending force for
from(a)FE
S235JR steelfor
analysis and
(a)(b) S275JR
S235JR steel.
steel and (b) S275JR steel.

The
The results
accuracy obtained from the models
of the regression FE simulations were used
was analyzed throughto determine
the coefficienta predictive
of deter-
model
mination for(R 2 ). bending
the The R2 coefficient
force as aforfunction
the S235JRof two
andvariables, i.e., vertical
S275JR is 1.0% and 0.999%, displacement
respectively,of
the upper roller
indicating a very andgoodplatecorrelation
thickness. The FE results
between the FE were fittedand
results using Matlab curve
regression models.fitting
In
toolbox
addition, (version R2018a, of
the adequacy The
theMathWorks,
derived models Inc., Natick,
was also MA, USA) [20,21].
analyzed using the root-mean-
squareTheerror (RMSE).
regression The RMSE
models for thevalues for the
bending forceS235JR and S275JR
(in Newton) for theare two × 10−5of%steel
4.06 plates and
4.786 − 1
× 10 %,byrespectively. These results show that the regression models are reliable and
are expressed the following equations:
can be used to calculate the bending force.
( , ) = 25.96 − 0.06384 Figure + 0.0542
16 compares ( 0.001454 + 0.5549
the bending force )as+ a0.5711 
function of (vertical
− 1.852displacement
+ 0.2448 ) of(31) the
upper roller for
for the S235JR steel, and the two analytical models and the FE regression model (as in Equation (31))
for a plate thickness of 10 mm and S235JR steel. It can be seen that the C1 analytical model
( , ) = 33.56 − (frictional,
22.08 + 0.008891as in Equation ( −(21))
0.002553 + 0.7136the
over predicts ) +bending
1.123 force ( 2.909 + 0.0128with
as compared ) the(32) C2
analytical model (frictionless, as in Equation (30)) and FE model. The relative errors for
for the S275JR steel.
Materials 2021, 14, x FOR PEER REVIEW 15 of 16
bending forces between the C1 and C2 models vary between 19% and 36%, and the higher
The accuracy of the regression models
the vertical displacement, the greater the relative error. was analyzed through the coefficient of deter-
mination (R2). The R2 coefficient for the S235JR and S275JR is 1.0% and 0.999%, respec-
tively, indicating a very good correlation between the FE results and regression models.
In addition, the adequacy of the derived models was also analyzed using the root-mean-
square error (RMSE). The RMSE values for the S235JR and S275JR are 4.06 × 10−5 % and
4.786 × 10−1 %, respectively. These results show that the regression models are reliable and
can be used to calculate the bending force.
Figure 16 compares the bending force as a function of vertical displacement of the
upper roller for the two analytical models and the FE regression model (as in Equation 31)
for a plate thickness of 10 mm and S235JR steel. It can be seen that the C1 analytical model
(frictional, as in Equation (21)) over predicts the bending force as compared with the C2
analytical model (frictionless, as in Equation (30)) and FE model. The relative errors for
bending forces between the C1 and C2 models vary between 19% and 36%, and the higher
the vertical displacement, the greater the relative error.
Compared with the FE regression model, the average relative error, the maximum
relative error, and the RMSE for the C1 analytical model are 27%, 54%, and 18.2%, respec-
tively, while the average relative error, the maximum relative error, and RMSE for the C2
model are 13.5%, 25%, and 12%, respectively. It should be noted that the maximum rela-
Figure
Figure
tive 16.Bending
16.
error ofBending
54% for force
forcethe versus
versus vertical
vertical
C1 model and displacement
displacement
25% for the predicted
predicted
C2 model based
based onon
occurs the the 5analytical
atanalytical models
models
mm vertical and
dis-
and
the FEthe FE
analysis analysis
for a for
10-mm a 10-mm
plate plate thickness
thickness and and
S235JR S235JR
steel. steel.
placement. For the C1 model, it appears that the presence of friction in the analytical for-
mulation has a noticeable effect on the deformation behavior; thus, it predicts higher val-
ues5. Conclusions
Compared with the FE regression model, the average relative error, the maximum
as compared with the C2 model. However, the bending force values calculated based
relative error,study,
and thehybrid
RMSEapproach
for the C1based
analytical model are 27%, 54%, and 18.2%,andrespec-
on theIn C2this
model areain better agreement withonthe theFEfinite element
values (FE) analysis
than those calculated analyt-
by the
tively, while the average relative error, the maximum relative error, and RMSE for the
Cical modeling
1 model. Therefore,of theit three-roller bending
can be concluded thatprocess was hinge
the plastic proposed
conditionin order to predict
without frictionthe
Cbending
2 model force
are 13.5%,
and 25%,
the and 12%,
curvature respectively.
radius of the It should
bending beThe
plate. noted that the of
simulation maximum
the three-
is adequate to estimate the bending force with reasonable accuracy.
relative error of process
roller bending 54% forwas the C 1 model
carried outand
using25%theforAnsys
the C2Workbench
model occurs at 5Structural
Static mm vertical pro-
displacement. For
gram, under plane strain the C model, it appears that the presence of friction
1 conditions. Starting from the observations following the in the analytical
FE anal-
ysis, analytical approaches for estimating the bending force and vertical displacement of
the upper roller were derived. Based on the FE analysis and the analytical modeling, the
following main findings and conclusions can be drawn:
(i) A 2D FE model was established to analyze the three-roller bending process for the
Materials 2021, 14, 1204 15 of 16

formulation has a noticeable effect on the deformation behavior; thus, it predicts higher
values as compared with the C2 model. However, the bending force values calculated
based on the C2 model are in better agreement with the FE values than those calculated
by the C1 model. Therefore, it can be concluded that the plastic hinge condition without
friction is adequate to estimate the bending force with reasonable accuracy.

5. Conclusions
In this study, a hybrid approach based on the finite element (FE) analysis and analytical
modeling of the three-roller bending process was proposed in order to predict the bending
force and the curvature radius of the bending plate. The simulation of the three-roller
bending process was carried out using the Ansys Workbench Static Structural program,
under plane strain conditions. Starting from the observations following the FE analysis,
analytical approaches for estimating the bending force and vertical displacement of the
upper roller were derived. Based on the FE analysis and the analytical modeling, the
following main findings and conclusions can be drawn:
(i) A 2D FE model was established to analyze the three-roller bending process for the
S235JR and S275JR steels used in the naval industry. For the system under consideration,
the FE results suggest that the cross section of the plate is practically a plastic hinge in the
tangent area of the upper roller;
(ii) Taking into consideration the assumption of reaching the “plastic hinge” condition
in the cross section of the plate, two analytical models for the bending force were derived
based on the bent bar theory. These models can be used to first-hand estimate the bending
force independently of FE analysis;
(iii) Using geometric and deformation compatibilities, simple analytical models for
the vertical displacement of the upper roller as a function of the curvature radius of the
plate were developed and verified against the FE results. However, in order to estimate the
curvature radius of the bending plate using the analytical formulation, the deformation
compatibilities must be considered;
(iv) FE-based regression models for estimation of the bending force for the S235JR
and S275JR steels were formulated. The models take into consideration the plate thickness
(8–12 mm) and vertical displacement of the upper roller (up to 78 mm). For other values
of the yielding stress, the bending force can be obtained by interpolation (because the
three-roller bending system has a diameter of 189 mm and the distance between the axis of
the lower rollers of the system is 320 mm).
These results are of practical importance for the industry to estimate the setting
parameters required in designing a three-roller bending machine. However, future research
will address the experimental validation of the derived models.

Author Contributions: Conceptualization, D.B. and F.S.; methodology, I.G., D.B., and F.S.; software,
I.G. and D.B.; validation, D.B. and I.G.; formal analysis, D.B. and F.S.; writing—original draft
preparation, D.B. and F.S.; writing—review and editing, F.S.; funding acquisition, F.S. All authors
have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
Acknowledgments: The authors would like to thank Nicoleta-Violeta Stanciu for her support with
the drawings.
Conflicts of Interest: The authors declare no conflict of interest.
Materials 2021, 14, 1204 16 of 16

References
1. Roggendorff, S.; Haeusler, J. Plate bending: Three rolls and four rolls compared. Weld. Met. Fabr. 1979, 47, 353–357.
2. Hua, M.; Lin, Y.H. Large deflection analysis of elastoplastic plate in steady continuous four-roll bending process. Int. J. Mech. Sci.
1999, 41, 1461–1483. [CrossRef]
3. Shin, J.G.; Lee, J.H.; Kim, Y.I.; Yim, H. Mechanics-based determination of the center roller displacement in three-roll bending for
smoothly curved rectangular plates. J. Mech. Sci. Technol. 2001, 1, 1655–1663. [CrossRef]
4. Chudasama, M.K.; Raval, H.K. Analytical model for prediction of force during 3 Roller multipass conical bending and its
experimental verification. Int. J. Mech. Eng. Robot. Res. 2012, 1, 91–105.
5. Chudasama, M.K.; Rava, H.K. An approximate bending force prediction for 3-roller conical bending process. Int J. Mater. Form.
2013, 6, 303–314. [CrossRef]
6. Padgan, N.P.; Deshpande, P.D.; Sakhale, C.N. Force Analysis of Metal Sheet in Bending Operation on Sheet Bending Machine. Int
J. Eng. Res. Tech. 2015, 4, 267–270.
7. Jadhav, C.S.; Talmale, P.S. Verification and Analysis of Stress-Strain Curve of Sheet Metal by using Three Roller Bending Machine.
Int. J. Eng. Dev. Res. 2017, 5, 368–374.
8. Tran, H.Q. Asymmetrical Roll Bending Process Study: Dynamic Finite Element Modeling and Experiments. Ph.D. Thesis,
Université du Québec, Québec, QC, Canada, November 2014.
9. Feng, Z.; Champliaud, H. Modeling and simulation of asymmetrical three-roll bending process. Simul. Model. Pract. Theory 2011,
19, 1913–1917. [CrossRef]
10. Feng, Z.; Champliaud, H. Three stage process for improving roll bending quality. Simul. Model. Pract. Theory 2011, 19, 887–898.
[CrossRef]
11. Ktari, A.; Antar, Z.; Haddar, N.; Elleuch, K. Modeling and computation of the three-roller bending process of steel sheets. J. Mech.
Sci. Technol. 2012, 26, 123–128. [CrossRef]
12. Taylor, V.K.; Gandhim, A.H.; Moliya, R.D.; Raval, H.K. Finite element analysis of deformed geometry in three-roller plate bending
process. In Proceedings of the ASME 2008 International Manufacturing Science and Engineering Conference, Illinois, IL, USA,
7–10 October 2008; Evanston: Illinois, IL, USA, 2008; pp. 1–7.
13. Neto, D.M.; Martins, J.M.P.; Oliveira, M.C.; Menezes, L.F.; Alves, J.L. Evaluation of strain and stress states in the single point
incremental forming process. Int J. Adv. Manuf. Tech. 2016, 85, 521–534. [CrossRef]
14. Patel, B.N.; Pandit, D.; Srinivasan, S.M. Large elaso-plastic deflection of micro-beams using strain gradient plasticity theory.
Procedia Eng. 2017, 173, 1064–1070. [CrossRef]
15. Pandit, D.; Srinivasan, S.M. Large Elasto-Plastic Deflection of Thin Beams with Roller Support Contact, 11th International
Symposium on Plasticity and Impact Mechanics. Procedia Eng. 2017, 173, 1079–1084. [CrossRef]
16. Fu, Z.; Tian, X.; Hu, B.; Yao, X. Analytical modeling and numerical simulation for three-roll bending forming of sheet metal. Int. J.
Adv. Manuf. Technol. 2013, 69, 1639–1647. [CrossRef]
17. Ansys Workbench 2018, Version 19; Software For Technical Computation; Ansys, Inc.: Canonsburg, PA, USA, 2018.
18. Gavrilescu, I.; Boazu, D. Simulation of roll bending with three rollers pyramid system using FEM. Ann. Univ. Dunarea de Jos of
Galati Fascicle VI 2017, 35, 21–28.
19. Hardt, D.E.; Constantine, E.; Wright, A. A model of a sequential bending process for manufacturing simulation. J. Eng. Ind. Trans.
ASME 1992, 114/181, 181–187. [CrossRef]
20. MATLAB 2018, Version R2018a; Software For Technical Computation; The MathWorks, Inc.: Natick, MA, USA, 2018.
21. Kamiński, K.; Solecka, M. Optimization of the truss-type structures using the generalized perturbation-based Stochastic Finite
Element Method. Finite Elem. Anal. Des. 2013, 63, 69–79. [CrossRef]