The International Journal of Advanced Manufacturing Technology

https://doi.org/10.1007/s00170-019-04183-2

ORIGINAL ARTICLE

An analytical model of roll bending steel pipe formed by three

rollers
Kyu Won Kim1 · Moon Kyum Kim1 · Woo Yeon Cho2

Received: 24 March 2019 / Accepted: 18 July 2019

© Springer-Verlag London Ltd., part of Springer Nature 2019

Abstract
This study proposes an analytical model to predict the final radius of a pipe manufactured using three-roller bending. This
model accounts for various parameters such as the strength of the steel sheet, the thickness of the sheet, the sizes of the
rollers, and the distance between the bottom rollers. It has been derived in order to calculate the loading stress distribution
and residual stress across the thickness of the plate. Equations to estimate the final diameter of the pipe are developed based
on calculations of the loading and unloading moments about the neutral axis. The position of the neutral plane is assumed
to be identical before and after the plate is bent. The results of the proposed analytical model are compared with those of the
numerical analysis to validate it under various conditions. It is observed that the analytical model estimates the final radius
to be approximately 15% greater than the numerical analysis.

Keywords Roll bending · Finite element analysis · Springback · Pipe manufacturing

1 Introduction To solve the plasticity problem, the equilibrium equa-

tions, constitutive equations, compatibility equations, yield
The quality of a steel pipe depends not only on the quality criteria, flow rules, and hardening rules must be considered
of the material but also on the mechanical deformation in the analysis of the bending of a metallic material such
that occurs during the manufacturing process. A steel sheet as a steel pipe. However, obtaining analytical solutions is
undergoes deformation due to the rollers in the process of mathematically challenging because the equations related to
forming large-diameter roll-bent steel pipes. The stress state the plasticity theory (such as yield criteria, flow rules, and
and amount of deformation depend on the position of the hardening rules) are nonlinear and apply to equilibrium and
product during the manufacturing process. compatibility equations in the form of partial derivatives.
The springback phenomenon occurs in this process, To overcome these challenges, numerical, analytical, and
owing to elastic deformation. Springback depends on the experimental studies have been conducted [1].
forming conditions such as the material strength, thickness, Sidebottom and Gebhardt [2] proposed a plane-strain
shape of the dies, and operating load. Therefore, to form solution that employs a small strain to predict the springback
a target product, it is necessary to predict the amount of of plates formed by bending. Tan et al. [3] presented a
springback by accurately considering various influences. new method to evaluate the distribution of residual stresses.
Springback must be considered while determining the target Herein, properties of bent sheet metals were expressed
diameter of a steel pipe. as a function of geometric parameters and materials. Hua
and Lin [4] presented a mathematical model, assuming an
elastic–perfectly plastic material, for a single-pass steady
 Kyu Won Kim
kyu won@yonsei.ac.kr continuous four-roller thin plate bending. Hu and Wang
[5] proposed an improved roll-bending model that imparts
1
higher flexibility in the formation of large bending parts.
School of Civil and Environmental Engineering, College
of Engineering, Yonsei University, 50, Yonsei-ro,
Gandhi and Raval [6] proposed an analytical and empirical
Seodaemun-gu, Seoul, 03722, South Korea model to estimate the upper roller position explicitly as a
2 Steel Structure Research Group, POSCO,
function of the desired radius of curvature for three-roller
100, Songdogwahak-ro, Yeonsu-gu, Incheon, cylindrical bending. Yu et al. [7] analyzed the elastic–plastic
21985, South Korea deformation process of the roll bending stage. Moreover,
 Int J Adv Manuf Technol

the influence of the upper roller reduction and of the initial In this paper, an analytical model is proposed to predict
ellipse and lower roller interval on the residual ellipticity the final radius of a pipe manufactured using three-roller
was verified through experiments. bending. Analytical equations for three-roller bending are
Zeng et al. [8] developed a general simulation model of derived to calculate the loading stress distribution and
the dynamic process of conical roll plate bending using the residual stress across the thickness of the plate. The
finite element method. Yang and Shima [9] introduced the equations are based on the calculation of the loading and
distribution of curvature and bending moments following unloading moments about the neutral axis. The position
the displacement and rotation of the rollers by simulating of the neutral plane is assumed to be identical before
the deformation of the workpiece with a cross-section in and after plate bending. The proposed model considers
a three-roller bending process. Numerical analyses were various parameters such as the strength of the steel sheet,
performed to simulate steel sheet bending with a three- thickness of the sheet, size of the rollers, and distance
roller machine. In addition, the numerical and experimental between the bottom rollers. The final radius of the roll
results were compared to demonstrate the validity of the bending pipe is calculated after considering springback
finite element model [10–13]. Salem et al. [14] proposed and plastic deformation caused by the load owing to the
an analytical model to predict roll bending force, residual rollers. Finite element analysis of the three-roller bending
stresses, and power for a roll bending process. They also pipe forming, based on Ktari et al. [10], is performed to
performed asymmetrical three-roll bending experiments to verify the model proposed in this paper. In addition, the
verify the developed model. Wang et al. [15] established results of the numerical analysis and those of the proposed
a four-roll bending forming process model based on the model are compared. Finite element analysis is performed
rebound theory of thick plates. Furthermore, the theoretical using Abaqus FEA 6.14 (SIMULIA, Johnston, RI, USA)
calculations were compared to the simulation results for finite element software. Material nonlinearity, geometric
verification. nonlinearity, and contact nonlinearity are considered in the
Most studies have suggested empirical models that can finite element model.
predict the final diameter of a pipe based on experimental
or numerical analysis. However, these models are difficult
to be commonly used as the model proposed under specific 2 Analytical model of pipe forming using
conditions. On the other hand, theoretically approaching three-roll bending
studies to predict the final diameter of a pipe use a complex
equation to predict the final diameter of the pipe. Therefore, The three-roller plate bending process that is applied to
it is difficult to confirm the final deformation of the pipe bend and roll a plate using three rollers is shown in Fig. 1.
intuitively, and it is difficult to use it directly by the The plate is supported by two bottom rollers which have
practitioner. a constant spacing between them. The initial curvature of

Fig. 1 Three-roller bending

pipe-forming process
Int J Adv Manuf Technol

Thus, it is necessary to consider springback to predict the

final diameter of the cylinder formed by the deformed plate.
If the material property of the steel is assumed as
isotropic hardened and the plate undergoes bending without
tension, the strain and stress distribution inside the plate
before and after springback is as shown in Fig. 3. If the
plate is bent by a pure moment without applying tension,
the neutral axis will be at the mid-thickness. The strain of
the layer at a distance of y from the neutral plane can be
Fig. 2 Plate deformation owing to vertical displacement of the top
roller expressed as y/Ri . Moreover, the distribution of the strain
generated by the curvature of the plate exhibits a linear
relationship with the distance y, as shown in Fig. 3a. If
the plate is formed by the vertical displacement of the top the strain in the plate is in the elastic range, the stress π
roller, and the rotation of the top roller forms a continuous corresponding to the strain follows Hooke’s law, which is
cylindrical shape of the steel sheet. In Fig. 1, the line expressed as:
indicates the initial state of the plate and rollers before the
load of the top roller is applied to the plate. Meanwhile, Ey
σ = Eε = Eκi y = (2)
the dotted line indicates the position after the vertical Ri
displacement of the top roller, showing the state in which where ε is the strain of the plate caused by the bending
the top roller deforms the plate. moment, E is the elastic modulus of the steel, κi is the initial
If springback does not occur at the plate after the curvature of the plate, and y is the distance from the neutral
conclusion of the force of the top roller that was applied plane.
to bend the plate, the final curvature of the plate relative to If the vertical displacement of the top roller generates
an arbitrary point Oa is as shown in Fig. 2. If the initial a stress distribution in the plate, as shown in Fig. 3b, the
curvature of the plate formed by the top roller consists moment acting on the plate is given by:
of the neutral axis of the deformed plate, the relationship   t/2 
ye Ey 2 Et y 2 ye
between the initial curvature of the plate and the vertical M = 2w dy + + (E − Et ) dy
displacement of the top roller is given by the following 0 Ri ye Ri Ri
equation: 
−4ERi3 σy3 + 4Et Ri3 σy3 − 3E 2 Et Ri σy t 2 w

 2   2 =
t t d 12E 3 Ri
dy = Ri + + Rb − Ri + + R b − (1)

2 2 2 E T w 3Ri σy + Et t
3 2
+ (3)
where dy is the vertical displacement of the top roller, Ri is 12E 3 Ri
the initial radius of the deformed plate, t is the thickness of where Et is the modulus of strain hardening (rate of strain
the plate, Rb is the radius of the bottom roller, and d is the hardening), ye is the position of the elastic region in the plate
distance between the bottom rollers. However, springback (ye = Ri σy /E), σy is the yield strength of the plate, and w
occurs when the metal is formed. Therefore, when the load is the width of the plate.
of the top roller is removed from the plate, the final radius If the strain in the plate exceeds the elastic range as
of the deformed plate becomes larger than its initial radius. a result of the excessive bending moment due to perfect

Fig. 3 Strain and stress distribution before and after springback in pure bending: a strain distribution; b stress distribution; c residual stress
distribution
 Int J Adv Manuf Technol

plasticity, the stress does not increase and remains in the shown in Fig. 4. The radius of curvature of the plate after
yield stress state. Thus, the moment acting on the plate can springback can be expressed as:
be expressed as follows:
1 1 Mi
  = − (6)
ye  Rp Ri EI
Ey 2 t/2
M = 2w dy + σy ydy
0 Ri ye Substitution of Eq. 6 into Eq. 4 yields:
wσy t 2 wRi2 σy3
= − (4) 12E 3 I Ri
4 3E 2 Rp =  (7)
12E 3 I − 3E 2 σy t 2 − 4Ri2 σy3 Ri
While bending, the internal moment obtained by Eq. 4
is in equilibrium with the external moment applied from where I is the moment of inertia of the cross-section.
the top roller. In addition, when the external moment is Thus, if the diameter of the bottom roller and the
removed, the internal moment of the plate should also thickness of the steel plate are determined, the vertical
disappear. That is, the sum of the moments resulting displacement of the top roller for forming the initial
from the residual stresses present in the plate after elastic curvature of the steel plate can be obtained using Eq. 2.
recovery should be 0. Therefore, the deformed plate is Furthermore, the final diameter of the steel pipe according
subjected to elastic recovery until the sum of the internal to the initial curvature of the steel plate can be predicted by
moments becomes 0. Ultimately, the compressive residual substituting the thickness, width, and material properties of
stress remains at the outer surface of the plate after elastic the steel plate in Eq. 7.
recovery, and the tensile residual stress remains near the
neutral plane (Fig. 3c). Furthermore, a different shape
appears at the inner surface of the plate, and the curvature of 3 Finite element analysis
the plate alters as the moment changes, owing to springback.
The change of the curvature owing to springback can The forming analysis of the three-roller bending pipe
be explained by the moment–curvature diagram shown in must consider the material nonlinearity and geometric
Fig. 4. Assuming that the deformation of the plate by the nonlinearity owing to the large deformation of a plate
rollers is of the elastic and plastic types, the initial curvature as well as the contact nonlinearity owing to the change
of the plate can be expressed as follows: in the contact condition between the roller and plate. In
order to simulate the pipe forming while considering these
1 1 1 nonlinearities, 3D numerical simulations of the three-roller
= + (5)
Ri Re Rp bending process were performed using Abaqus FEA 6.14.
The three-roller bending pipe analysis has steps for bending,
where Re is the radius of curvature caused by the elastic rolling, and springback. The explicit algorithm was used to
deformation of the plate and Rp is the radius of curvature conduct dynamic and nonlinear analysis of the bending and
caused by the plastic deformation of the plate. rolling processes. Moreover, the static analysis (obtained
For a plate that has been bent to fully plastic moment, from the results of the sheet-metal forming process and
the unloading curve is parallel to the elastic loading line, as considered as the initial condition) was used to simulate the
springback process.

Fig. 4 Moment–curvature diagram for an elastic-perfectly plastic plate Fig. 5 Finite element model of three-roller bending
Int J Adv Manuf Technol

3.1 Finite element model and its vertical displacement was applied to the plate so
that the plate was flexed. The rotation of the top roller
The finite element models were developed by considering is constrained to prevent slipping when the plate and top
three rollers and a plate. The plate was considered to be a roller come in contact. The displacement of the plate in
shell element because the D/t ratio of the pipe produced the z-axis was constrained to prevent the plate from being
by the rollers is large. The rollers were formed using twisted by the rollers. In the rolling step, the displacements
shell elements. However, the rollers applied rigid constraint of all the rollers were constrained, and the top roller rotated
conditions based on their respective reference points. That around the z-axis with an angular velocity of 0.23 rad/s.
is, the rollers were assumed to be rigid bodies. Generally, Meanwhile, the boundary conditions of the bottom rollers
the behavior of the steel is assumed to be perfect plasticity remained as defined in the previous step. In the springback
or isotropic hardening, except in exceptional cases such as step, all the displacements and rotations at one end of the
steel subjected to repetitive loads. Also, it is considered plate were constrained to enable the deformation associated
that the influence on the additional moment generated with elastic recovery, depending on the internal stress of
in the plasticity like hardening is small because the roll the plate. Furthermore, the rollers were deactivated by a
bending pipes covered in this paper are pipes for energy function of the model change in Abaqus FEA, and the
transportation which have more than 610 mm in diameter. rollers did not interfere with the deformation of the plate.
Therefore, the material property was assumed to be elastic– The weight of the plate was not considered. The finite
perfectly plastic. The elastic modulus of the steel is 210 element models are summarized in Fig. 5.
GPa, and Poisson’s ratio is 0.3. The contact algorithm
was used to simulate the interfaces between the plate 3.2 Verification
and rollers. The contact properties were defined in two
perpendicular directions: tangential and normal. Coulomb’s A 2D numerical analysis of the roller pipe based on Ktari
friction defines the tangential behavior, and the static and et al. [10] was performed to verify the finite element model.
dynamic friction coefficients are constant. The friction To verify the results of the numerical model of three-roller
coefficients between the rollers and plate in the bending bending, the stress distributions inside the steel plate were
step were assumed to be 0.2. In the rolling step, the friction compared using von Mises stress. Moreover, the diameter
coefficients between the top roller and plate were assumed of the pipe was verified after all the deformations had
to be 1.0, and the friction coefficients between the bottom occurred. Figure 6 represents the residual stress distribution
rollers and plate were assumed to be 0.2. The normal owing to the forming of three-roller bending. The maximum
behavior was assumed to be a hard contact, so that the stress during pipe forming was 365.7 MPa, which varies by
contacting parts between the rollers and plate could not approximately 1% from Ktari’s result [10]. The minimum
penetrate. In the bending step, the displacements of the stress was 187 kPa, which varies by approximately 70%.
bottom rollers were constrained in the x-, y-, and z-axes, However, this could be considered negligible in the case
whereas rotation was constrained in the x- and y-axes. of the minimum stress because it is less than 1 MPa. The
The lateral displacement of the top roller was constrained, diameter of the pipe was 983 mm, and the error in the

Fig. 6 Residual stress

distributions owing to roll
bending forming: a Ktari et al.
(2012); b roll forming analysis
 Int J Adv Manuf Technol

diameter represents less than 1%. Therefore, it can be said

that the numerical analysis for evaluating the residual stress
of a roller pipe due to the forming process is reasonable.

4 Analytical model versus finite element

analysis

In this paper, the final radius of the deformed plate is

compared with the results of the 3D finite element analysis
of the three-roller bending process to verify how effectively
the proposed analytical model can demonstrate the actual
pipe forming. In the three-roller plate bending process, the
degree of springback is altered by the strength of the steel,
the thickness of the plate, the distance between the bottom Fig. 7 Final radius of deformed plate according to steel type
rollers, the initial curvature of the bending plate, and the
size of the top and bottom rollers. A numerical analysis of
deviation between them can be explained by the effect of
the three-roller bending process considering each of these
springback on the steel plate, as shown in Fig. 8.
parameters was performed in this paper.
The analytical model showed that springback occurred
In order to verify the effect of each factor on the final
only after the deformation of the steel plate was complete.
radius of the deformed plate, a numerical analysis was
However, in the case of the numerical analysis, springback
performed by changing the material properties and shapes
occurred even during the deformation of the steel sheet.
of the rollers and plate. The basic material properties and
Therefore, as the steel plate shifted to the right bottom
shapes used in the numerical analysis are as follows. The
roller, springback occurred in the plate. Moreover, in the
strength of the steel was 485 MPa, thickness of the plate
model proposed in this paper, only pure bending of the steel
was 6.35 mm, distance between the bottom rollers was
sheet occurred during roll bending pipe formation. It was
650 mm, initial radius of the deformed plate was 322 mm,
assumed that the steel sheet was circular. However, in case
and radii of the top and bottom rollers were 300 mm and
of the actual roll bending pipe formation, the steel sheet is
250 mm, respectively. The results of numerical analysis
generated a compressive force in the rolling direction of the
were compared with the analytical model.
roller by the speed and friction of the roller in the section
where the direction of the steel plate is changed such as the
4.1 Steel strength
center of the top roller. Under these effects, the steel plate
exhibited a curvature larger than the target initial curvature,
Numerical analyses of the various steel types such as API
and the analytical model estimated the final radius to be
5L X52, X60, X70, X80, and X100 were carried out to
approximately 15% greater than the results of the numerical
verify the influence of steel strength on pipe formation.
analysis.
To maintain the initial radius of the deformed plate at
322 mm, the top roller was set at a vertical displacement of
100.62 mm, and the shortest distance between the surface
of the top roller and that of the bottom roller was 9.74 mm.
The strength of the steel was defined as 360, 416, 485, 693,
and 831 MPa. The final radius of the deformed plates for
each strength level is shown in Fig. 7.
In the graph presented in Fig. 7, the horizontal axis
represents the strength of the steel, and the vertical axis
represents the final radius of the cylinder formed from the
steel plate. The line represents the final radius of the steel
sheet as predicted by the analytical model, and each point is
the final radius of the steel sheet measured in the numerical
analysis. The analytical model is modified by multiplying
it by 0.85 and the modified model is represented by a
dashed line. The result reveals good correlation between the Fig. 8 Stress distribution of X70 steel plate (t = 6.35 mm) in the
modified analytical model and the numerical analysis; the analytical model and numerical analysis
Int J Adv Manuf Technol

Fig. 9 Final radius of deformed plate according to thickness of plate

Fig. 10 Stress distribution of X70 steel plate (t = 30 mm) in the
analytical model and numerical analysis

4.2 Plate thickness 4.3 Initial curvature of bending plate

The influence of the plate thickness on pipe formation was In order to determine the effect of the initial radius of the
studied. The thickness of the plate was set as 6.35, 8, 9, bending plate, three initial radii were selected: 322, 350, and
10, 14, 22, and 30 mm. The material properties, geometrical 420 mm. The plate was pressed by the top roller, which went
shape, and boundary conditions are identical to those of the down by 100.62, 95.05, and 83.65 mm to form each initial
underlying conditions. To maintain the initial radius of the radius. Furthermore, the distance between the surface of the
deformed plate at 322 mm, the top roller was imparted a top roller and that of the bottom roller was 9.74, 14.29, and
vertical displacement of 100.62 mm. Moreover, the shortest 23.65 mm, respectively.
distance between the surface of the top roller and that of Figure 11 shows the final radius of the cylinder
the bottom roller was 9.74 mm. The vertical displacement corresponding to the initial radius of the plate. The distance
of the top roller for maintaining the initial radius at 322 mm between the surface of the top roller and the surface of the
was 100.62, 100.45, 100.34, 100.24, 99.82, 98.99, and bottom roller to thickness ratios were 1.35, 2.25, and 3.72
98.18 mm, respectively. Furthermore, the distance between and prediction errors have 1.51%, 11.29%, and 27.72%,
the surface of the top roller and that of the bottom roller respectively. That is, the numerical analysis results are
was 9.74, 11.23, 12.13, 13.03, 16.65, 23.90, and 31.18 mm, only similar to those of the modified analytical model at
respectively. The final radius of the deformed plates for each 322 mm. The increase in the distance to thickness ratio
thickness is shown in Fig. 9. indicates that the space between the top roller and the
As described above, the analytical model estimated the bottom roller is larger. As the space between the top and
final radius to be approximately 15% greater than the
numerical results. When the thickness of the steel plate was
10 mm or less, the modified analytical model exhibited
a similar trend to the numerical results. As the plate
became thicker, the error between the numerical analysis
and modified analytical model increased. When a thick plate
was formed, the position of the neutral axis of the plate
shifted to the outer surface, as shown in Fig. 10. Moreover,
the bending moment acting on the plate and the moment of
inertia of the cross-section were changed by the position of
the neutral axis. In constrast, in the analytical model, the
neutral axis was kept constant at the center of the cross-
section regardless of the plate thickness. However, the plate
thickness was very small compared with the final radius of
the deformed plate. Therefore, the change in the neutral axis
owing to the increase in thickness appears to be irrelevant in Fig. 11 Final radius of the deformed plate according to initial radius
the prediction of the final diameter of the cylinder. of the plate
 Int J Adv Manuf Technol

bottom rollers increases, the curvature of the steel sheet has plate, the error converges to a value as the size of the top
an unstable arc. The unstable arc violates the assumptions roller decreases.
of the proposed model in this paper.These errors were
caused by the assumption of the analytical model that the 4.5 Size of the bottom roller
deformation of the plate was circular. The analytical model
assumed that the profile of the deformed plate was a circle Parametric studies were performed to evaluate the effect
passing through three points, which are the points of contact of the size of the bottom rollers on the deformed plate.
between the plate and the rollers. However, it was necessary Figure 13 shows the final radius of the deformed plate when
to reduce the vertical displacement of the top roller to the sizes of the bottom rollers are 150, 200, 250, 300, and
increase the initial radius. As a result, the distance between 350 mm. The vertical displacement of the top roller is set to
the upper roller and lower rollers was increased so that the 88, 78, 70, 64, and 59 mm to maintain the initial radius of
plate did not assume a circular shape. In addition, the errors 322 mm. The distances between the surface of the top roller
between the results of the numerical analysis and modified and that of the bottom roller are 9.95, 9.23, 8.71, 8.33, and
analytical model enlarged. 8.03 mm, respectively.
A comparison of the modified analytical model with
4.4 Size of top roller the numerical analysis results revealed that most of the
results were in concurrence. As the size of the bottom
The radius of the top roller was set at 150, 200, 250, and roller decreased, the error increased. When the radius
300 mm to determine the relationship between the size of of the bottom roller decreased to 150 mm, an error of
the top roller and the final radius of the deformed plate. The approximately 5% occurred. These errors occurred as the
initial radius was 322 mm, and the vertical displacement of distance between the upper and lower rollers became larger.
the top roller was 70.00 mm regardless of the size of the top In addition, it is considered that the final radius of the plate
roller. The distance between the surface of the top roller and was influenced more by the size of the top roller than by that
that of the bottom roller was 34.46, 24.23, 15.78, and 8.71 of the bottom roller.
mm, respectively.
Figure 12 shows the relationship between the size of 4.6 Distance between bottom rollers
the top roller and the final radius of the deformed plate.
As the top roller becomes smaller, the distance between The relationship of the distance between the two bottom
the top roller and bottom roller increases. As a result, the rollers with the final radius was observed. The analysis was
shape of the deformed plate does not follow the curvature performed using five bottom rollers positioned in the range
of the circle. It is also verified that the error between the of 550 to 750 mm. A vertical displacement was applied to
analytical model and the result of the numerical analysis the top roller to maintain the initial radius at 322 mm. The
increases. However, as the size of the top roller decreases, distances between the surface of the top roller and that of the
the plate is sharply bent and forms a larger curvature at an bottom roller were 8.71, 9.20, 9.74, 10.35, and 11.04 mm,
equal initial curvature of the plate. Therefore, unlike the respectively.
phenomenon observed at the initial curvature of the bending

Fig. 12 Final radius of the deformed plate according to the size of the Fig. 13 Final radius of the deformed plate according to size of the
top roller bottom roller
Int J Adv Manuf Technol

Fig. 15 Predictions of the final radius of the pipe for the analytical
Fig. 14 Final radius of the deformed plate according to the distance model under arbitrary conditions
between the bottom rollers

Figure 14 illustrates the final radius as a function of shape, and roller distance, numerical analysis of three-roller
the distance between the bottom rollers. In the analysis, bending was performed by setting an arbitrary condition.
considering other factors such as the thickness, the initial The various constant parameters used in the numerical
radius, and the size of the rollers, errors between the analysis are reported in Table 1. To obtain the initial radius
results of the numerical analysis and the modified analysis of each analysis case, the steel plate was bent by pressing
model increased as the distance between the top and bottom the top roller to impart vertical displacements of 112, 103,
rollers increased. However, regardless of the change in 141, 124, and 49 mm. The distances between the surfaces of
the distance between the two bottom rollers, the modified the top and bottom rollers were maintained at 59.99, 46.92,
analytical model is effectively represented by the results of 32.55, 24.86, and 31.23 mm.
the numerical analysis. The results of the numerical analysis The final radius of the deformed plate according to each
and the modified analytical model both corresponded to condition is shown in Fig. 15. Errors between the modified
the distance between the bottom rollers being 700 mm. analytical model and the result of the numerical analysis
Moreover, the error increased as this distance increased or were 30%, 9%, 11%, 3%, and 2% in C1, C2, C3, C4,
decreased from 700 mm. Although the distance between the and C5, respectively. In C1 and C2, the shortest distances
top roller and bottom roller increased, the error was more between the surface of the top roller and the surface of
or less marginal. Therefore, the analytical model proposed the bottom roller are 59.99 mm and 46.92 mm, which are
in this paper can be effective for predicting the final radius relatively far. However, the shortest distance-to-thickness
of a cylinder fabricated using three rollers, as the distance ratio is 6 to 1.56. Therefore, the deformation of the plate
between the lower rollers increases. does not form a circular shape in C1, so that more errors
occur. In C2 and C3, the shortest distance-to-thickness ratio
4.7 Application examples is 1.56 to 2.71. However, C2 is thicker than C3. Therefore,
since the neutral plane moves to the inner surface of the
In order to verify the applicability of the analytical model plate when the pipe is formed, C2 causes more errors than
under various conditions of strength, thickness, rollers C3. In C4 and C5, the shortest distance-to-thickness ratio is

Table 1 Conditions for forming three-roller bending pipe

Case Yield strength (MPa) Thickness Distance between Initial radius Radius of top Radius of bottom
(mm) bottom rollers (mm) (mm) roller (mm) roller (mm)

C1 360 10 700 350 150 250

C2 831 30 750 420 300 300
C3 554 12 720 300 225 225
C4 485 15 800 400 340 300
C5 554 13 550 350 310 255
 Int J Adv Manuf Technol

1.66 to 1.29, and the result shows a good agreement between References
the modified analytical model and the numerical analysis.
1. Ablat MA, Qattawi A (2017) Numerical simulation of sheet metal
forming: a review. Int J Adv Manuf Technol 89(1–4):1235–1250
2. Sidebottom OM, Gebhardt CF (1979) Elastic springback in plates
5 Conclusion and beams formed by bending. Exp Mech 19(10):371–377
3. Tan Z, Li WB, Persson B (1994) On analysis and measurement
In this study, an analytical model was proposed that of residual stresses in the bending of sheet metals. Int J Mech Sci
accounts for various parameters, such as the strength of the 36(5):483–491
4. Hua M, Lin YH (1999) Large deflection analysis of elastoplastic
steel sheet, the thickness of the sheet, the size of the rollers,
plate in steady continuous four-roll bending process. Int J Mech
and the distance between the bottom rollers. To confirm Sci 41(12):1461–1483
the validity of the proposed model, the analytical model 5. Hu W, Wang ZR (2001) Theoretical analysis and experimental
was compared with the results of numerical analysis under study to support the development of a more valuable roll-bending
process. Int J Machine Tools and Manuf 41(5):731–747
various conditions.
6. Gandhi AH, Raval HK (2008) Analytical and empirical modeling
As a result of comparing the final radius of the of top roller position for three-roller cylindrical bending of plates
pipe using the results of the analytical model and the and its experimental verification. J Mater Process Technol 197(1–
numerical analysis model, it was found that the analytical 3):268–278
7. Yu G, Zhao J, Zhai R, Ma R (2018) Theoretical analysis and
model overestimates the final radius by approximately 15%
experimental investigations on the symmetrical three-roller setting
compared with the numerical analysis. The reason for this round process. Int J Adv Manuf Technol 94(1–4):45–46
difference is that a compression force is generated on the 8. Zeng J, Liu Z, Champliaud H (2008) FEM Dynamic simulation
upper surface of the steel plate by rolling the top roller, and and analysis of the roll-bending process for forming a conical tube.
J Mater Process Technol 198(1–3):330–343
springback occurs on the plate in the roller-forming process.
9. Yang M, Shima S (1988) Simulation of pyramid type three-roll
The compression force generated on the upper surface is bending process. Int J Mech Sci 30(12):877–886
related to the rolling speed and the coefficient of friction of 10. Ktari A, Antar Z, Haddar N et al (2012) Modeling and
the top roller. Therefore, it is necessary to calibrate the pipe computation of the three-roller bending process of steel sheets. J
diameter more precisely to predict the diameter of the pipe. Mech Sci Technol 26(1):123–128
11. Fu Z, Tian X (2013) Chen Wet al, Analytical modeling and
For the case of a thick plate, the position of the neutral numerical simulation for three-roll bending forming of sheet
axis of the plate can change. For this reason, an error in the metal. Int J Adv Manuf Technol 69(5–8):1639–1647
final radius can occur. Therefore, it is necessary to calibrate 12. Tran QH, Champliaud H, Feng Z, Dao TM (2014) Analysis
the proposed model when predicting the diameter of a roller of the asymmetrical roll bending process through dynamic FE
simulations and experimental study. Int J Adv Manuf Tech 75(5–
bending pipe with a thick wall. 8):1233–1244
To apply the proposed model correctly, it is essential 13. Zhao W, Liao TW, Kompotiatis L (2017) Stress and springback
to make the curvature of the plate by the roller as close analyses of API x70 pipeline steel under 3-Roller bending via
as possible to a circle. To accomplish this, it is necessary finite element method. Acta Metall Sin 30(5):470–482
14. Salem J, Champliaud H, Feng ZK, Dao TM (2016) Experimental
to reduce the shortest distance between the surface of the analysis of an asymmetrical three-roll bending process. Int J Adv
top roller and the surface of the bottom roller. The shortest Manuf Tech 83(9–12):1823–1833
distance-to-thickness ratio should be kept below 1.5. 15. Wang Y, Zhu X, Wang Q, Cui X (2018) Research on multi-roll
roll forming process of thick plate. Int J Adv Manuf Tech:1–10
Funding information This research was supported by a grant
(No.17IFIP–B067091–05) from the Industrial Facilities and Infras- Publisher’s note Springer Nature remains neutral with regard to
tructure Research Program funded by Ministry of Land, Infrastructure jurisdictional claims in published maps and institutional affiliations.
and Transport of Korea government, and was supported by the Korea
Institute of Energy Technology Evaluation and Planning (KETEP)
granted financial resource from the Ministry of Trade, Industry &
Energy, Republic of Korea (No.20171510101910)