Accepted Manuscript

Title: Double-stage forming using critical pre-bending radius

in roll bending of pipe with rectangular cross-section

Author: Do-Sik Shim Kee-Poong Kim Ki-Yong Lee

PII: S0924-0136(16)30137-6
DOI: http://dx.doi.org/doi:10.1016/j.jmatprotec.2016.04.033
Reference: PROTEC 14807

To appear in: Journal of Materials Processing Technology

Received date: 22-10-2015

Revised date: 25-4-2016
Accepted date: 28-4-2016

Please cite this article as: Shim, Do-Sik, Kim, Kee-Poong, Lee, Ki-Yong,
Double-stage forming using critical pre-bending radius in roll bending of
pipe with rectangular cross-section.Journal of Materials Processing Technology
http://dx.doi.org/10.1016/j.jmatprotec.2016.04.033

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 Double-stage forming using critical pre-bending radius in

roll bending of pipe with rectangular cross-section

Do-Sik Shim*, Kee-Poong Kim, Ki-Yong Lee

Green Manufacturing Process Technology Center, KITECH, Wolchul-dong, Gwangju, South

Korea, 500-460

*Correspondence to be addressed to:

Dr. Do-Sik, SHIM

Green Manufacturing Process Technology Center

KITECH

Cheomdan-venturero 108beon-gil, Buk-gu, Gwangju, South Korea

E-mail: : imagineer@kitech.re.kr

C.P: +82-10-4844-4455

Tel: +82-62-600-6200

Fax: +82-62-600-6099

Abstract

Bent pipes are widely used in many high-end industries as a structural component. Consequently,

pipe roll bending has become an attractive manufacturing technology for forming lightweight products.

For industrial applications, bent pipes should be accurately shaped into 2D or 3D shapes without

defects. In this study, the behavior of pipes with rectangular cross-sections under roll bending was

considered in order to improve product quality. When a pipe product is bent with a large curvature, the

pipe cross-section shrinks under the bending moment. In order to minimize defects and improve
product quality, this paper proposes double-stage forming to regulate the shrinking load, which

causes the axial wrinkles and poor strength, as structural members. In double-stage forming, the

bending radius at the pre-bending stage is assumed to be a key parameter to avoid cross-section

distortion. In order to determine the pre-bending radius, material behavior was analyzed to figure out

the reason for the cross-section shrinkage. The bending stresses on the cross-section of the pipe

material subject to bending moments bring about the distribution of inward forces, which induce

shrinkage moments on the cross-section. The main idea in this paper is that the cross-section

distortion can be minimized by regulating the pre-bending radius at the first stage. The proposed

process design was evaluated through nonlinear finite element simulation using a commercial

program and experiments.

Keywords: Pipe; Roll bending; Bending moment; Shrinkage

1. Introduction

In transport-related industries such as aircraft, ship, train, and automobile manufacturing, the

development of lightweight technology is being promoted in order to improve energy efficiency. In

addition to optimizing the structural shape, the development of lightweight materials has led to

advances in lightweight technology; in particular, the application of materials with high specific

strength has increased significantly. These technological developments have led to demand for

suitable forming and manufacturing techniques.

In recent years, pipe forming techniques have been developed to apply pipes as lightweight

structures. Unlike conventional manufacturing, where a structural member for an automobile is

completed by sheet metal forming and welding, using bent pipes for an automobile structure reduces

the manufacturing costs as well as the weight. Thus, pipe forming techniques are receiving renewed

attention.

Many types of forming technologies can be used to manufacture structural pipes. One technique that

has received the maximum attention is the three-roll bending process, where structural pipes are
produced with various curvatures using a single set of rolls. Three-roll bending is a flexible forming

process because various curvatures are produced by adjusting the distance between rollers.

Over the last decade, many researchers have studied the three-roll bending process. Because it is

used extensively in the manufacture of simple shaped structural parts, the roll bending process has

been studied for a long time by academic and industrial experts. Previous work can be divided into

two major areas: (i) The relation between bending moment and curvature in the determination of the

positions of each bending roll, and (ii) the prediction of springback behavior and residual stress

distributions. With regard to the former, Yang and Shima (1988) discussed the distribution of the

curvature and bending moment in accordance with the displacement and rotation of the rolls by

simulating the deformation of a workpiece with a U-shaped cross-section in the three-roller bending

process. Hardt et al. (1982) proposed a process model for use in simulating the manufacture of

cylindrical shells from flat plates that require sequential bending by incorporating the prior bend

history. Yang et al. (1994) performed a controlled finite element method (FEM) simulation wherein the

movement of the roll is adjusted by feedback control to approach the desired curvature distribution.

Hua et al. (1999) proposed a mathematical model for determining the internal bending resistance at

the contact point of the front roller for thin-plate bending, along with the principle mechanisms of the

bending process for single- and multi-pass bending. Shin et al. (2001) explored a kinematics-based

symmetric approach to the process of assigning the region of the plate to be rolled, in an effort to form

smoothly curved plates. Gandhi and Raval (2006) developed an analytical model for estimating the

position of the central roller as a function of the desired radius in consideration of material behavior

and the change in its Youngs modulus. Kajrup and Flamholz (2003) proposed an economical and

efficient approach for obtaining the desired curvature through control of the rollers position in real

time; for which the roll bending machine is required to be equipped with expensive electric

components and workpiece-handling devices, as well as a computer numerically controlled (CNC)

machine that can control the entire production process.

Several researchers also investigated residual stresses on roll-formed square pipes. Residual

stresses play a significant role in determining the behavior and strength of these pipes when roll-

formed steel columns are used, especially in structural steel members (Chen and Chang, 1993).

Some researchers have also focused on springback problems, in which the inevitable elastic release

phenomenon of different degrees occurs when the tooling is removed. They derived analytical
formulae for predicting the springback behavior and residual stress distributions in the bending by

considering the beam bending theory and ignoring the wrinkling phenomenon. On the basis of tube

bending theories and experiments they established empirical formulae for determining the springback

and stretching (Rasmussen and Hancock, 1998). Researchers have also deduced the springback

angle by considering the relationship between the bending moment and the bending curvature

(Landolfo and Mazzolani, 1991). Empirically, a numerical-analytical method was used to rapidly

calculate the springback angle of a thin-walled tube. Despite these previous researches, most of the

focus in research on the pipe roll bending process has been on predicting the stroke of the central

roller and determining the roll set positions to fabricate a precise product with a desired curvature.

When a pipe product is bent with a large curvature, the cross-section of the pipe shrinks under the

bending moment. During the loading process, another type of failure may also occur; the compressed

region of the bent pipe can wrinkle axially and produce a localized instability. Many researchers have

investigated the distortion of the cross-section of hollow structural members during the bending

process.

Vaze and Corona (1998) studied the pure bending of rectangular section profiles by experiment, and

proposed that the combined action of inner flange and side wrinkling led to the collapse of rectangular

tubes. Okude et al. (2008) studied wrinkling limit improvement of aluminum square tubes in large

bending radius during draw bending process. Zhao et al. (2007) mainly studied the effects of process

parameters on wrinkling of rectangular wave-guide tubes in rotary-draw bending based on the

platform ABAQUS/Explicit. Gantner et al. (2005) proposed a bending technique Free-Bending,

particularly suitable for bending profile and tube cross-sections. They investigated with the aim of

improving its applicability in the automotive industry. Goodarzi et al. (2005) investigated on the shear

bending to realize a considerable small bending radius. And, the stresses and strains of the bending

of a large scale pipe using local induction heating were investigated under loading and unloading by

Hu (2000). Chatti et al. (2010) developed a torque-superposed-spatial (TSS) bending process and

machine for three-dimensional bending of profiles with symmetrical and asymmetrical cross-sections.

They analyzed several important factors that influence bending accuracy and validated their research

through experiments and FE simulation.

Recently, there are many studies on the bending technique with an inner mandrel to improve the

product quality in draw bending, pressure bending, and rotary-draw bending. The steel balls as the
pressure transferring medium can effectively restrain the wrinkling and section distorting of bent pipe

(Zhu et al., 2012). Zhao et al. (2010) studied the effect of the clearance of mandrel die-tube on

wrinkling distribution along the bending direction of rectangular pipe. The experimental and analytical

methods were employed to study the wall thinning and section ovalization of the small-diameter tube

bended without mandrel (E et al, 2008). Most studies as stated above are limited to the rotary draw

bending process. However, unlike the draw bending of pipes with a single bending point, the mandrel

is unavailable in the three-roll bending of a product with locally various curvatures because the

curvature varies along the longitudinal direction of the pipe. Therefore, to improve the quality of the

cross-section, the authors investigated the cross-section shrinkage in terms of the process design

without using additional tools. To the best of our knowledge, there is no study on the design of

multiple forming steps for bending of rectangular pipes in the three-roll bending processes. Our

current work is the first to deal with the design of multi-stage forming method to improve the product

quality. In the present study, the process design considering the material behavior and deformation of

rectangular cross-section pipe under roll bending was developed. In actual production, the final

product is manufactured based on the workers experience without any engineering criterion or

consideration of the mechanical behavior of the pipe subjected to the bending moment. For these

reasons, the final bent product inevitably has defects such as surface wrinkling and cross-section

shrinkage. This paper proposes double-stage forming, where the material is deformed in a pre-

bending stage prior to the final product bending. Mechanical analysis was performed to determine the

reference value of the pre-bending radius in order to minimize product defects. In order to design the

pre-bending radius, material behavior was analyzed to determine the reason for the cross-section

shrinkage. The stresses acting on the pipe material subject to bending moments were examined to

estimate changes in the pipes cross-section. The bending stresses on the pipe cross-section bring

about the distribution of inward forces, which induce shrinkage moments on the cross-section.

Considering the distribution of inward forces, the elastic limit condition is determined by using a

classical theory. With the application of the plastification conditions, the cross-section distortion is

minimized by regulating the pre-bending radius at the first stage. The proposed double-stage forming

using the analytically calculated pre-bending radius was verified through experiments and finite

element (FE) analysis.

2. Roll bending process for structural part forming

2.1. Bent pipe as structural member

The application of pipes as structural members has been gradually expanding in the automotive

industry, especially for commercial vehicles such as large buses. As shown in Fig. 1, the entire

structure of the next-generation bus adopts bent pipes with various cross-sections and shapes.

Because the next-generation bus requires both safety and aesthetics to be satisfied at the same time,

a pipe bending technique that guarantees design flexibility and shape accuracy is a requisite.

Fig. 1. Structure of next-generation vehicle

2.2. Roll bending process

The roll bending process is a flexible forming process used to produce pipes or beams with various

curvatures. Fig. 2 shows the configuration of the pyramid-type three-roll bending process. The

workpiece is bent into the desired shape by moving the central roller while the positions of the two

front rollers are fixed. At the start of bending, the central roller is located near the front roller and is

moved upwards to bend the front end of workpiece without rotating the three rollers. Next, the central

roller approaches the workpiece while the front roller rotates until the critical stroke of the central

roller; no slip occurs between the rollers and workpiece. To produce the desired curvature, the central

roller is moved up and down. Finally, the central roller is moved downwards without rotating when the

rear end of the workpiece comes to the right front roller.

Fig. 2. Schematic diagram of three-roll bending process

3. Occurrence of cross-section distortion

The main defect produced in a pipe under bending is the distortion or shrinkage of its cross-section

(Fig. 3) because of the inward stress components. The shrinkage of a bent pipe results in a loss of

stiffness in the form of limit point instability; this is referred to as ovalization instability. Furthermore,

the increased axial stress at the compression side due to shrinking may cause bifurcation instability

(buckling) in the form of longitudinal wavy-type wrinkles, usually before a limit moment is reached.

During the assembly of bent pipes in the welding process, the final formed product must have high-

quality geometric accuracy at the end of product; otherwise, difficulties will ensue during assembly.
Ensuring that the final product is fabricated in accordance with product specifications with regard to

geometric accuracy and without any defects is very important and is the motivation of this paper.

In this study, the shrinking of elastic pipes was examined. The pipes were initially straight and bent

with a longitudinal radius of curvature R that was significantly larger than the cross-sectional

dimensions (i.e., width and height).

Fig. 3. Schematic presentation of shrinkage

4. Double-stage forming

In actual production, if the desired pipe product has a small bending radius, the inner side

(compression region) of the highly bent pipes is wavy, and cross-section shrinkage occurs, as shown

in Fig. 3. The conventional manufacturing process based on trial-and-error results in product defects

and a long lead time for design and production. For this reason, this paper proposes double-stage

forming in which the initially straight pipe is deformed incrementally into a final shape, as shown in

Fig. 4. In double-stage forming, the bending radius at the pre-bending stage is assumed to be a key

parameter to avoid cross-section distortion and improve the quality of the final product. Thus, the

bending radius at the pre-bending stage should be designed in terms of material deformation and not

depend on the workers knowledge, which ignores the material properties and dimensions of the pipe

and the final curvature.

The reference value for double-stage forming in the pipe roll bending process is necessary to

minimize changes in the pipe cross-section due to the deformation that the pipe undergoes in each

stage. The next section presents an analysis of the longitudinal stresses generated by pipe bending

and the shrinking moment caused by the stresses that induce the cross-sectional change based on

material mechanics.

Fig. 4. Double-stage forming of pipe with large curvature

5. Mechanics of pipe bending

Conceptually, the cross-section change occurs when the curvature from bending creates components

of bending forces that act in a direction perpendicular to the neutral axis, thereby creating bending

moments in the pipe wall. Eventually, the increase in bending forces exceeds the ability of the
shrinking section to resist the bending in the wall, and an instability occurs. As the bending continues,

a kink (or local buckling) is observed. This process is nonlinear, and hollow tubes under bending have

unique and complex behaviors. All modeling and analyses described in this paper were based on the

following general assumptions:

(i) The material is homogeneous and isotropic.

(ii) The strain at a point in the plate is linearly proportional to its distance from the neutral plane.

(iii) The moment distribution is linear between the two rollers in the three-roll bending so that the

shear force is not negligible.

(iv) Stress is uniform through thickness t (assuming t is much smaller than d in Fig. 5)

First, the stresses acting on the pipe material must be examined to estimate changes in the pipes

cross-section caused by longitudinal bending. In the analysis, a straight pipe was considered with a

thickness t, width w, and height d that was subject to the bending moment M* (Fig. 5). When a

straight pipe is deformed into a curved product with the bending radius R, the tensile stresses in the

longitudinal direction act near the upper region of the neutral plane; in contrast, there are compressive

stresses near the lower region of the neutral plane, as shown in Fig. 6.

Fig. 5. Curved pipe under longitudinal bending

Fig. 6. Internal force under longitudinal bending

Fig. 6 shows the bending-induced opposing longitudinal forces. These forces induce an external force

dv on both sides of an infinitesimal segment. The internal forces dFlongi and vertical forces dv can be

expressed as

= = (1)

and

d =2 (2)

where is the longitudinal stress, and is the bending angle in the longitudinal direction after

the pipe has been deformed by the bending moment. For small angles, /2 can be replaced by

/2 so that the inward force dv becomes

=2 =2 = (3)

The force dv acts on the cross-section in the y-z plane, which is perpendicular to the x-z bending

plane.

As shown in Fig. 7, the uniform load acts on a channel-section bridge. The worst shrinkage in the

cross-section occurs in the top half. To determine the start condition for shrinking, see Fig. 7.

Assuming the top surface of the channel-section bridge to be a simply supported beam, , which is

the shrinking moment acting at the midpoint of the top surface, can be given as

= = (4)

where V is the vertical load (summation of dv) acting at the midpoint (with y=0) of the top surface.

Elchalakani et al. (2002) found that a square pipe subject to bending did not achieve the full plastic

moment owing to shrinkage and the relatively low strain-hardening capacity of cold-formed tubes. The

critical point (point at which the shrinking begins) is attained when the bending reaches the elastic

limit. Therefore, it is assumed that the longitudinal stress equals the yield stress ( ) of the

material when the shrinkage occurs. For a half section, we have from Eq. (3)

As shown in Fig. 7, consider a thin rectangular band of thickness where = ! (where the

bending radius R is much larger than the height d). Then

 "
= (5)
#

The second moment of area I of the band cross-section is introduced as

%& " ' (

$ =
'
or =
%&
(6)

Using Eqs. (4), (5), and (6), the bending (shrinking) moment at the mid-point of the width is given by

) *(
= = (7)
#% *

In this work, the critical value is derived considering the situation when the initial yielding occurs at the

top surface of the midsection (with y = 0) in Fig. 7. For this derivation, two assumptions were made: (i)

at the moment of initial yielding, R is finite (not infinite); (ii) stress is uniform through the thickness t

and width w. Since the maximum tensile and compressive stresses acting at any given cross section

occur at points located farthest from the neutral axis, the flexure formula ( = ,$) between the

stresses and the bending moment can be expressed as

./ (
= %
(8)

The above equation can be used to predict the elastic bending limit of a hollow tube and the

corresponding bending radius (Elchalakani et al. (2002)).

Then we can obtain from Eqs. (7) and (8)

) *
#%
=1

Finally, the critical bending radius can be determined by the following criterion:
 ) *
! %
(9)

Fig. 7. Distribution of inward forces on cross-section

To provide a clearer understanding of this analysis, the upper bounds of the pre-bending radius that

were developed to avoid excessive shrinking of the pipe cross-section under roll bending is discussed

for several cases of cross-section dimensions t and w. From Eq. (9), the critical bending radii that

represent the beginning of shrinkage were determined; the results are depicted in Fig. 8. According to

the critical bending radii at dimension, a larger width for the same thickness allows plastic deformation

to occur more easily and quickly for a small bending moment. The critical value increases with

decreasing pipe thickness; that is, the pipe cross-section shrinks with a small curvature. A numerical

simulation based on the FE method was conducted to verify the double-stage roll bending using the

above critical bending radii. The FE simulation results are discussed in the following section.

Fig. 8. Critical bending radii with respect to thickness (t) and width (w) of rectangular pipes

6. FE simulation of double-stage forming

6.1. 3D geometric model of tool set

Fig. 9 shows a schematic diagram of the roll bending tool set used in the experimental and numerical

analyses. All of the rollers had the same dimensions. The distance between the two front rollers has a

significant effect on the cross-section deformation shape. In this study, taking consideration of product

specifications and empirically optimized tool setting conditions, the distance between the two front

rollers was set to be 320 mm. There was a 1 mm clearance between the workpiece and roll die to

avoid jamming and induce smooth feeding of the workpiece. The rotating speed could be changed

within the range of 1100 mm/s. In the experiment, the average feed rate was set to 30 mm/s. This

not only shortened the production time but also promoted the dimensional accuracy of the product by

allowing reasonable control of the metal flow.

Fig. 9. Tooling dimensions for pipe bending

6.2. FE modeling and its validation

FE modeling was carried out to simulate the behavior of a rectangular pipe in the roll bending

process, and the accuracy and efficiency were evaluated.

The general-purpose commercial software ABAQUS v.6.12 (implicit) was used. Fig. 10 illustrates the

FE model used for the pipe roll bending simulation. The central roller and front rollers were modeled

as rigid surfaces, and the pipe workpiece was modeled as a finite element S4R. The S4R is a four-

noded, doubly curved quadrilateral shell element. It includes large rotations, transverse shear

deformations, and finite membrane strains (ABAQUS Inc., 2011). Two initial pipe sections were used:

40 mm 40 mm and 40 mm 80 mm. For the initial pipe, the element size was 2 mm 2 mm. The

central roller was modeled to rotate about its instantaneous rotation center. The central roller was

driven, and the Coulomb friction coefficient between the central roller and pipe was set to 0.3 (Lange,

1985). The von Mises plasticity model was used. The front rollers, which were fixed, were treated as

frictionless surfaces. The forming procedure was modeled as real workingloading of the central

roller toward the workpiece, feeding the workpiece by rolling, and unloading. In the FE simulations,

after passing through the roll dies by rolling, the deformed pipe material undergoes a slight elastic

recovery as a result of the springback phenomenon, which is inevitable during the cold-forming

process. ABAQUS/Implicit code was used to simulate the springback phenomenon.

Fig. 10. FE model for roll bending simulation

Tensile tests were carried out to obtain mechanical properties of the pipes. The tests were carried out

according to the ASTM D638 standard for tube specimens using a Shimadzu AG-IS (50 kN) testing

machine. The pipe materials were the general-purpose structural steels STKR490-JIS (2.0 mm

thickness). The tests were performed with stroke control up to tensile failure of the samples. The

loading speed was set to 100 mm/min; the strain rate effect was not considered.

Table 1 summarizes the mechanical properties of the tensile test specimens. Fig. 11 presents the

results in terms of stressstrain curves and stressstrain data for ABAQUS input. The obtained

material properties were used for the FE simulations.

Fig. 11. True plastic stress-strain data for ABAQUS input

For qualitative evaluation of the FE model, the simulation results were compared with experimental

results. In the experiments, rectangular pipes with cross-sections of 40 mm 40 mm and 40 mm 80

mm were used. Each experiment was conducted using the same central roller displacement Dc. Fig.

12 shows comparisons of the experimental and simulation results for the 40 mm 40 mm and 40 mm

80 mm cross-section pipes, respectively. The simulated wrinkles on the inner surface of the pipes

and the pipe deformations were nearly identical to those in the experiments. The experimental and

simulation results showed that the proposed FE model was found to predict the pipe deformation by

roll bending with great accuracy. In particular, the wrinkles on the inner surface of the deformed pipe

showed good similarity when shell elements were employed. In the case of using solid elements,

application of three layers of elements in the thickness direction would improve the analysis accuracy,

but the required computation time would be excessive. Therefore, shell elements were found to be

more appropriate than solid elements in terms of accuracy and efficiency for the present study. Based

on these results, it can be said that the forming of different pipe materials and cross-sections can be

simulated by the proposed FE model with sufficient efficiency and accuracy.

(a)

(b)

Fig. 12. Comparison between experimental and FE simulation results: (a) 40 mm 40 mm square

pipe and (b) 40 mm 80 mm rectangular pipe

6.3. Initial wrinkling

As the bending continues with larger curvatures, a kink (longitudinal wavy or wrinkles) is observed in

the compression region. As stated above, the increase in bending forces exceeds the ability of the

shrinking section to resist the bending in the wall, and an instability occurs. Many researchers studied

wrinkling limit of square pipes with large bending radius in analytical approach. However, the

analytical solution has a limit for practical uses due to various bending conditions and pipe

geometries. Hence, in this study, the FE simulations are used for the determination of the wrinkling

limit. Five pipe widths were used for simulations: 40 mm, 50 mm, 60 mm, 70 mm, and 80 mm. The

pipe bending simulations are conducted with decreasing the bending radii until the wrinkling initiates.

The waviness of the product can be quantified well by the inwardoutward deflection of the inner

surface due to wrinkling. Thus, the waviness of the formed pipe was evaluated by using the deviation

(R) as the difference in the radii of two reference concentric arcs, as shown in Fig. 13(a). The arc
with the best fit was determined using the least-squares method; it contained all the measured points.

In this study, the initial wrinkling is defined as the moment when the R is less than 0.5 mm.

The simulation results for each pipe width are shown in Fig. 13. The initial wrinkling has a similar

trend with the cross-section shrinkage. The upper bound for the pre-bending radius is given as the

critical bending radius at the cross-section shrinkage, and now the lower bound is limited by the

bending radius at the initial wrinkling.

(a)

(b)

Fig. 13. (a) Detection of the waviness and (b) prediction of initial wrinkling by simulations

6.4. Double-stage forming simulation

To reduce the shrinking moment resulting in cross-section distortion, double-stage roll bending was

proposed. Using the analytical model and simulations explained above, the critical bending radius and

pre-bending radii were calculated. Here, the pre-bending radius was set to be below 95% of the

critical bending radius, which was revealed from several cases of bending simulations and

comparisons of the defects occurred after the final bending. This criterion can design the pre-bending

radius Rpre to satisfy the requirements.

The design process explained above was applied to the manufacture of a bent pipe with a large

curvature. For the double-stage design, the pre-bending radii in the first stage, as presented in the

previous section (Fig. 8), were employed. An FE simulation was performed to verify the double-stage

forming using the critical pre-bending radii. The single-stage bending and the double-stage bending

with different pre-bending radii were designed to allow for a comparison, which is shown in Table 2.

To determine the actual minimum and compare it with double-stage bending with analytically

calculated pre-bending (1.2 and 2.7 m), pre-bending radii of 1.0, 1.1, 1.3, and 1.4 m (for the 40 mm

40 mm section) and 2.4, 2.6, 2.8, and 3.0 m (for the 40 mm 60 mm section) were checked in

addition to the analytically calculated pre-bending radii of 1.2 and 2.7 m.

Prior to pipe bending, the central roller position (Dc) should be determined to obtain the designed

bending radius. In previous research, a numerical approach was applied to calculate the central roller

position to achieve the required curvature. Based on the efficiency and accuracy of the proposed FE

model, forming simulations with different pipe materials and cross-sections were conducted. Through

a series of bending simulations, a database was obtained on the relationship between the central

roller stroke (Dc) and the bending radius. The database was used to calculate the central roller stroke

for a given desired radius.

Moreover, the compressive load on the lateral wall transmitted by contacting with roll dies also

accelerated the deflection of the lateral walls. These two factors (rotation moment and compressive

load) were not considered in the analytical model. This is the reason why the analytically calculated

bending radii overestimated the deformation of the cross-section profiles. Furthermore, the several

assumptions regarding the analytical prediction of the critical pre-bending radius (i.e., the rigid plastic

behavior of the material, stress condition of inner/outer surfaces, and disregarding the deformation of

the lateral wall) influenced the optimal pre-bending radius. However, although it was not the actual

optimal value, the analytically calculated pre-bending radius in the study improved the product quality

more than the arbitrarily chosen pre-bending radii.

The cross-section shrinkage was also found to be more intense with a wider cross-section for the

same bending radius. As discussed in the introduction, another type of failure may occur during the

loading process; that is, the local buckling (or kinks) in the compressed region of the bent pipe. The

kinks were observed in the 40 mm 40 mm 2 mm workpiece, and the 40 mm 60 mm 2 mm

workpiece showed the local buckling because of the increased axial stress at the compression side

due to shrinking.

Based on the numerical analyses, the experiment was conducted under the same conditions. The

deformations of cross-sections with various pre-bending radii, including comparisons with the single-

stage bending, are investigated in the next chapter.

Fig. 14. FE simulation results: (a) single-stage bending and (b) double-stage bending (Left: 40 mm

40 mm 2 mm, Right: 40 mm 60 mm 2 mm)

 Fig. 15. Cross-sections of deformed pipes: (a) 40 mm 40 mm 2 mm and (b) 40 mm 60 mm 2

mm

There were substantial discrepancies between the peak longitudinal strains during the forming stages,

and longitudinal strain was excessively allocated at only one stage of the single-stage bending (case

1). This suggests that, under these conditions, a sound product can be produced with double-stage

bending, but single-stage bending would produce more critical geometric defects in the roll bent pipe.

Moreover, a noticeable difference between the two cases (case 2 and case 3) is the deviation of peak

longitudinal strains between the two successive forming stages. For the case of simple double-stage

bending without considering critical Rpre (case 2), there is a considerable deviation of the peak strains

between the two stages, corresponding to an excessive allocation of strains in the pre-bending stage

and therefore indicating a greater propensity to shrinkage. However, in the case of double-stage

bending considering critical Rpre (case 3), there is very little deviation in the peak longitudinal strain

between the pre-bending and the final bending stage where a distorted surface is most likely to be

reduced. From the above numerical analysis, it can be concluded that the inner surface distortion

should be associated with patterns of longitudinal strains, and peak longitudinal strains should be

balanced for each forming stage in order to prevent the shrinkage as a result of an excessive strain in

a particular stage.

At this point, it is necessary to pay attention to the CRF (cold-roll-forming) process, which is similar to

the roll-bending process in terms of the incremental sheet metal bending. Hua et al. (1995)

investigated the longitudinal strain distribution during the deformation, as it is of great importance in

determining the deformation behavior of the pipe and its relationship to the occurrence of defects.

Theoretical and experimental observations reveal that there is a limit to longitudinal strain beyond

which defects will occur. Considering these theoretical and experimental studies of the CRF process,

if the increment of bending between any two successive steps exceeds a particular limit, the required

straightness of the output product will be compromised and, consequently, local edge buckling may

occur during the subsequent step of the process. Therefore, to avoid edge buckling, a limit should be

applied to the bending radius (or strain) increment.

In this study, the shrinkage of the cross-section was also assumed to occur as a result of the inward

force induced by the longitudinal stress. Therefore, the maximum longitudinal strain could be related

to the shrinkage and buckling defect of the formed pipe. Given these expectations, the authors tried to

relate the maximum variation of longitudinal strains between two successive stages (not the total

amount of strain) to the quality of the final cross-section.

Fig. 16. Location of a node for history output of longitudinal stains

Fig. 17. Change in longitudinal strain during forming stages: (a) single-stage bending, (b) simple

double-stage bending, and (c) double-stage bending considering critical Rpre

7. Experimental results and discussions

7.1. Computer numerically controlled (CNC) roll bending machine

Fig. 18 shows the configuration of a typical CNC bending machine: the pyramid-type three-roll

bending machine. Two front rollers and one central roller are rotated by a motor, and their positions

are controlled in the y-direction by an actuator. A digital encoder and proximity sensor are used to

control the exact movement and rotation of the rollers. Two- or three-dimensionally shaped structural

members with a final radius of curvature Rfinal are fabricated by controlling the position of the central

roller and rotating the front rollers. A workpiece is automatically fed by the two front rollers and bent to

an arbitrary curvature by the worker adjusting the position of the central roller through a control panel;

parts with varying curvatures are easily produced. Thus, this machine is capable of producing

different parts with a single set of rolls. Therefore, this process is much more flexible than other

bending processes, such as press bending.

Fig. 18. CNC roll bending machine

7.2. Experimental results

For experimental verification of the double-stage forming, experiments were carried out using 40 mm

40 mm 2 mm STKR490 specimens. CNC code was created for double-stage forming and input

into the CNC bending machine for experiments. As explained in Fig. 4, the pipe was pre-bent starting

from one end while being transferred to the opposite side, and the final bending radius was formed as

it returned.
Fig. 19 shows the final formed pipes using single-stage bending at a final radius of 0.5 m, double-

stage bending with pre-bending radii of 1.0 and 1.2 m. The figure indicates that the single-stage

bending resulted in a wrinkling defect on the inner surface with the shrinkage.

Fig. 20 shows a comparison of the numerical and experimental results. The formed pipes were

measured using a coordinate-measuring machine (CMM) along the outer surface, and the measured

profiles were offset by half the pipe thickness to obtain the central lines. Moreover, the simulated

cross section profile is identical to the central line. The comparisons provide several meaningful points

to discuss. First, the cross-sections of the simulated profiles are more distorted than the experimental

ones. Indeed, the behavior of the lateral wall was neglected in this study since it was found from the

preliminary experiments that the deformation of the lateral wall was not significant. Meanwhile, in the

simulations, there was a 1-mm clearance between the workpiece and roll die to avoid jamming and to

induce the smooth feeding of the workpiece (see Fig. 9). This allowed the lateral walls to be free of

deformation, and then the cross-section corners were allowed to rotate. Unlike the FE simulations, in

the experiments the lateral walls were completely constrained by the tools, and therefore the lateral

walls of the experimental profiles were less deflected than the numerical ones.

Moreover, in this study, the S4R element type (which has reduced integration including large

rotations, transverse shear deformations, and finite membrane strains) was used in the FE simulation.

The S4R element has one integration location per element compared with four integration locations

for S4, which makes the S4R element computationally more efficient. However, since the S4R

elements have only one integration point, it is possible for them to distort in such a way that the

strains calculated at the integration point are all zero. This, in turn, leads to uncontrolled distortion of

the mesh.

Regarding these characteristics of the S4R elements, S4R elements could not accurately consider the

in-plane bending of the lateral walls, and the cross-section corners are too stiff in bending. In turn, the

corners are rotated. Therefore, the total energy was dissipated by the deformation of the lateral walls

and the rotation of the cross-section corners, so that the inward "shrinking moment" of the outer/inner

walls of the profile seems significantly underestimated by the simulation. Given this point of view, the

finer mesh in the region of the corners and S4 elements (which give accurate solutions to in-plane

bending problems, are not sensitive to element distortion, and avoid parasitic (shear) locking

(ABAQUS Inc., 2011)) seem to outperform in an exact simulation of cross-section distortion.

The cross-sections of the deformed pipes are presented in Fig. 21. The experimental results showed

that the cross-section distortion was reduced for double-stage roll bending compared with single-

stage bending. As with the simulation results, the optimal pre-bending radius seems to be

approximately 1050 mm, which is smaller than the calculated values. Nevertheless, the predicted pre-

bending radius was considerably close to the actual optimal radius, and the overall trends of the FE

simulations are well reproduced in the experiments.

Consequently, the experimental shrinkages in the double-stage bending using the analytically

calculated pre-bending radius and the actual optimum radius were in the allowable range of

geometrical requirements for sound assembly in actual practice. As validated through FE simulations

and experiments, it seems reasonable to conclude that the proposed double-stage bending with an

analytically calculated pre-bending radius can be used for practical applications since this is an

acceptable guideline for the proper amount of the first pre-bending for hollow rectangular tubes, even

though this approach cannot provide the optimal value to minimize the shrinkage of the cross-section.

When the pipe was bent with large curvatures, the cross-section of the formed pipe lacked a perfectly

rectangular shape despite the incremental increase in the amount of deformation. When a pipe is

formed predominantly by stretching or contraction, as in-plane deformation, then the phenomenon

described above is not observed. In a bending-dominant process, however, this phenomenon occurs

naturally. In double-stage forming, since the straight pipe is deformed into the final product via a pre-

bent shape, the deformations caused by bending and in-plane strain progressed incrementally and

gradually, so possible defects were reduced. Therefore, to avoid defects in the forming of a bent pipe

with large curvatures, an effective approach is double-stage roll bending employing the critical radius,

which presents the upper bound of pre-bending.

Although more numbers of forming steps (3 or 4 passes) reduce the cross-section distortion of the

bent pipe, but excessive numbers of forming passes result in a waste of time and cost. Double-stage

roll bending for desired products is now possible by using the critical value of the pre-bending radius.

From the above comparisons, the pre-bending stage can reduce the cross-section shrinkage caused

by excessive longitudinal stress. Moreover, the product quality and process productivity can be

improved through the use of the proposed pre-bending radius.

Fig. 19. Formed pipes

Fig. 20. Comparison of simulations and experiments

 Fig. 21. Experimental results (cross-sections of deformed pipes)

8. Conclusions

Hollow tubes or pipes experience significant changes to their cross-section profile along the pipes

length through shrinkage. This phenomenon makes pipes under bending have unique and complex

behavior, which was the motivation for the current research effort. Conceptually, shrinkage is caused

when curvature in bending creates components of bending forces that act in a direction perpendicular

to the neutral axis, which creates circumferential bending moments in the pipe wall. This behavior is

nonlinear and in turn creates a biaxial stress condition. Eventually, the increase in the bending forces

exceeds the ability of the distorted section to resist the bending in the wall, and an instability occurs.

Although smaller forming increments or more forming stages reduce the product defects, excessive

forming stages wastes time and cost for mass production and small batch production. In some

industries, mandrels are utilized inside pipes to avoid cross-section shrinkage and inner surface

waves. In other industries, the draw bending technique, in which drawing forces are applied to the

workpiece, is employed to eliminate wrinkling in the compression region by stretching. However,

these techniques are limited to particular industries and products or have high capital costs.

This paper presents a novel pipe-forming technique in terms of process design that can be applied

without incurring a long manufacturing time and high equipment costs. This paper mainly explained

the cause of the cross-sectional distortion and proposed a double-stage bending using an analytically

calculated pre-bending radius to improve the deformed cross section. For the cold bending of hollow

pipes, many researchers have analyzed the longitudinal strain distribution as it is of great importance

in determining the deformation behavior of the pipe and its relationship to the occurrence of defects.

In this study, the shrinkage of the cross section was also assumed to occur because of the inward

force induced by the longitudinal stress. Moreover, increasing the amount of bending (or decreasing

the bending radius) increases the longitudinal stresses. The first instance of buckling instability may

occur at the location of maximum stress. Therefore, the maximum longitudinal strain could be related

to the shrinkage and wrinkling defects of the formed pipe. Given these expectations, this study tried to

relate the maximum variation of longitudinal stresses to the quality of the final cross section.

Theoretical and experimental observations reveal that longitudinal stretching or compressive strain

occurs at the outer or inner surface of the deformed pipe. Once the pipe surface has been plastically
deformed by excessive compression, it is difficult to stretch the metal back into a straight section. This

inevitably leads to shrinkage or wrinkling.

In this work, the authors proposed a simpler methodology for the practical design of pipe-bending

metalworking. It can be cautiously stated that the analytical solution was useful in determining the

reason for the cross-sectional shrinkage and in designing the pre-bending radius using the

engineering approaches considering the behavior of the material. Considering these theoretical and

experimental studies on the pipe-bending process, it can be noted that if the amount of bending

exceeds a particular limit, the required smoothness of the output product will be compromised and,

consequently product defects may occur during the subsequent steps of the process. Double-stage

roll bending of desired products can now be used by employing the critical value of the pre-bending

radius. In order to design the forming process during the three-roll bending of highly rounded pipes,

the critical bending radius was defined as the design variable for double-stage forming. The

effectiveness of the double-stage bending was investigated by FE simulations and experiments. From

the above comparisons, it was observed that the pre-bending stage can reduce the cross-sectional

shrinkage caused by excessive longitudinal stress. Moreover, the product quality and process

productivity can be improved using the analytically calculated pre-bending radius.

Our final objective is a design for multistage bending, rather than double-stage bending, for the most

improvement. In the near future, the authors plan to optimize the bending radii at each forming stage

for a more robust process design rather than for only a better solution. As stated in the previous

section, a higher number of forming steps would be more effective for the reduction of the cross-

sectional distortion. Therefore, it is necessary to optimize the allocation of bending radii at each pass

in the event of multi-pass bending (more than three passes). To improve productivity and the quality

of the products, the forming step should be optimized to assure larger and more appropriate

longitudinal strains in each step, and, consequently, to minimize the number of steps. For this

optimized step design, it is necessary to develop a series of forming increments to satisfy the

requirements, including maintaining the magnitude of compression strain below a critical value. We

have planned to develop a method of optimizing the forming steps during the incremental bending of

a cylindrical pipe considering this criterion.

Acknowledgement
This work was supported by Grant No. JA130021 from the Ministry of Science, ICT and Future

Planning in Korea. The author would also like to express sincere gratitude to KITECH (Korea Institute

of Industrial Technology) for supporting this research.

References
ABAQUS Inc., 2011, ABAQUS 6.12 Users Manual, Version 6.12. ABAQUS Inc., Paw-
tucket, RI, USA.
Chatti, S., Hermes, M., Tekkaya, A.E., Kleiner, M., 2010. The new TSS bending process: 3D
bending of profiles with arbitrary cross-sections. CIRP Ann. Manuf. Tech. 59, 315-318.
Chen, S.J., Chang, S.C., 1993. Residual stresses in welded jumbo box columns. J. Constr.
Steel Res. 25, 201-209.
E, D.X., Ning, R.X., Li, Y.M., Tian, X., 2008. Experimental study on the wall thickness
thinning of the curved tube affected by material property. Mat. Sci. Tech. 16, 200203 (in
Chinese).
Elchalakani, M., Zhao, X.L., Grzebieta, R.H., 2002. Plastic mechanism analysis of circular
tubes under pure bending. Int. J. Mech. Sci. 44, 1117-1143.
Elchalakani, M., Zhao, X.L., Grzebieta, R.H., 2002. Plastic slenderness limit for cold-formed
circular steel hollow sections. Australian J. of Struct. Eng. Steel Issue 3, 116.
Gandhi, A.H., Raval, H.K., 2006. Analytical modeling of top roller position for multiple pass
(3-roller) cylindrical forming of plates. In: International mechanical engineering congress and
exposition, Chicago, IL, 107-116.
Gantner, P., Bauer, H., Harrison, D., Silva, A., 2005. Free-bending a new bending
technique in the hydroforming process chain. J. of Mat. Process. Tech. 167, 302308.
Goodarzi, M., Kuboki, T., Murata, M., 2005. Deformation analysis for the shear bending
process of circular tubes. J. Mat. Process. Tech. 162, 492497.
Hardt, D.E., Roberts, M.A., Stelson, K.A., 1982. Closed-loop shape control of a roll-bending
process. J. Dyn. Sys. Meas. Control 104, 317-322.
Hu, Z., 2000. Elasto-plastic solutions for spring-back angle of pipe bending using local
induction heating. J. Mat. Process. Tech. 102, 103108.
Hua, M., Baines, K., Cole, I.M., 1999. Continuous four-roll plate bending: a production
process for the manufacture of single seamed tubes of large and medium diameters. Int. J.
Mach. Tool. Manu. 36, 905-939.
Hua, M., Sansome, D.H., Baines, K., 1995. Mathematical modeling of the internal bending
moment at the top roll contact in multi-pass four-roll thin-plate bending. J. Mat. Process.
Tech. 52, 425-459.
Kajrup, G., Flamholz, A., 2003. Bending the wind-roll bending challenges in fabricating
conical cylinder wind towers. The Fabricator.
Landolfo, R., Mazzolani, F.M., 1991. The influence of the variation through the thickness of
residual stresses in tubular columns. J. Constr. Steel Res. 19, 49-57.
Lange, K.,1985. Handbook of metal forming. McGraw-Hill.
Okude, Y., Sakaki, S., Yoshihara, S., 2008. Improving working limit in draw bending process
of aluminum square tubes in large bending radius. Adv. Technol. Plast. 564569.
Rasmussen, K.J.R., Hancock, G.J., 1998. Deformations and residual stresses induced in
channel section columns by presetting and welding. J. Constr. Steel Res. 11, 175-204.
Shin, J.G., Park, T.J., Yim, H., 2001. Kinematics based determination of the rolling region in
roll bending for smoothly curved plates. J. Manuf. Sci. Eng. 123, 284-290.
Vaze, S.P., Corona, E., 1998. Degradation and collapse of square tubes under cyclic bending.
Thin-Walled Struct. 31, 325441.
Yang, G., Mori, K., Osakada, K.,1994. Determination of forming path in three-roll bending
using FEM simulation and Fuzzy reasoning. J. Mater. Process. Tech. 45, 161-166.
Yang, M., Shima, S., 1988. Simulation of pyramid type three roller bending process. Int. J.
Mech. Sci. 30, 877-886.
Zhao, G.Y., Liu, Y.L., Yang, H., 2007. Numerical simulation on influence of clearance and
friction on wrinkling in bending of aluminum alloy rectangular tubes. Mater. Sci. Forum 546,
833838.
Zhao, G.Y., Liu, Y.L., Yang, H., Lu, C.H., 2010. Cross-sectional distortion behaviors of thin-
walled rectangular tube in rotary-draw bending process. Trans. of Nonferrous Metal Soc. of
China 20, 484489.
Zhu, Y.X., Liu, Y.L., Yang, H., 2012. Sensitivity of springback and section deformation to
process parameters in rotary draw bending of thin-walled rectangular H96 brass tube. Trans.
of Nonferrous Metal Soc. of China 22, 22332240.

Table 1 Material properties

E (Young's Tensile Uniform Total

Yield stress
modulus) stress elongation elongation
(MPa)
(GPa) (MPa) (%) (%)

STKR 490 175 517.3 533.0 9 24

Table 2 Test conditions for FE simulations (unit: mm)

Pipe dimension (d w t): 40 40 2 and 40 60 2 (Material: STKR490)

Final radius: 500

Pipe Pre- Final

Dimension Initial state bending bending Remarks
(d w t) (Rpre) (Rfinal)
Single Straight
N/A 500
stage pipe
1000 500
1100 500
40 40 2
Double Straight Proposed value of
1200 500
stage pipe Rpre
1300 500
1400 500
Single Straight
40 60 2 N/A 500
stage pipe
 2400 500
2600 500
Double Straight Proposed value of
2700 500
stage pipe Rpre
2800 500
3000 500

Fig. 14 shows the simulation results for each case. The inner surface waves decreased in the pipe
formed from double-stage roll bending considering the critical bending radius. The results of the
numerical analyses indicated that the assumptions for double-stage forming and the critical pre-
bending radius were valid. Fig. 15 compares the cross-sections of deformed pipes for each case. In
addition, the amount of shrinkage (vertically sunk down of the inner surface at a mid-point) for each
case was measured and summarized in Table 3. For the two cross-section geometries, single-stage
bending resulted in the largest shrinkage, as expected. The double-stage bending with analytically
calculated pre-bending (with radii of 1.2 and 2.7 m) improves the shrinking significantly compared with
single-stage bending. However, the actual minimal shrinking appeared at a pre-bending radius of
approximately 1.1 m for the 40 mm 40 mm section, and at approximately 2.4 m for the 40 mm 60
mm section. For the two geometries, the actual optimal pre-bending radii were smaller than the
calculated values. Here, it is useful to take a closer look at these results. As can be seen in the
deflected cross-section profiles, the deformation of the lateral wall was not negligible. As the vertically
sunk down of the outer surface increases, the corner of the lateral wall is subjected to a rotation
moment so that the upper end of the lateral wall deflects inward in the cross-section.
Table 3 Amount of shrinkage for each case (unit: mm)

Pipe Pre-
dimension bending Shrinkage Remarks
(d x w x t) (Rpre)
Single
N/A 5.16
stage
1000 3.16

1100 1.83 (minimum)

40 x 40 x 2
Double
1200 2.30
stage
1300 3.03

1400 3.71
Single
N/A 11.76
stage
2400 3.84 (minimum)
2600 5.76
40 x 60 x 2
Double
2700 5.84
stage
2800 6.96

3000 8.72

For more detailed analysis of the behavior of the material, including the occurrence of shrinkage,
further investigation into the distribution of strains on the deformed material during the process is
needed. Furthermore, to observe the correlation between longitudinal strain and product defect, it is
necessary to look into the longitudinal strains on the inner surface of the deformed pipe based on the
simulation results. The changes in longitudinal strain for the three simulation cases (40 mm 40 mm
cross section) are depicted in Fig. 17. The plots were obtained from the history output of the
longitudinal total strain at a node located on the mid-layer of the shell element (Fig. 16). The
longitudinal strains in the elements on the inner surface of the deformed pipes, before and after
contacting with the central roller in the pre-bending and final bending stages, were therefore,
compared for each case, in Figs. 17(a)(c). For all cases, prior to contacting with the central roller the
elements on the inner surface are subjected to tensile longitudinal strain, when the elements are
deformed by contacting with the central roller, the longitudinal strain drops drastically. This must be
the result of compressive stressed induced by the bending moment. After departing from the central
roller, the tensile strain increases.