Accepted Manuscript

Title: A mechanical model of symmetrical three-roller setting

round process: the static bending stage

Author: Jun Zhao Gaochao Yu Rui Ma

PII: S0924-0136(16)30001-2
DOI: http://dx.doi.org/doi:10.1016/j.jmatprotec.2016.01.002
Reference: PROTEC 14682

To appear in: Journal of Materials Processing Technology

Received date: 31-7-2015

Revised date: 25-12-2015
Accepted date: 1-1-2016

Please cite this article as: Zhao, Jun, Yu, Gaochao, Ma, Rui, A mechanical model
of symmetrical three-roller setting round process: the static bending stage.Journal of
Materials Processing Technology http://dx.doi.org/10.1016/j.jmatprotec.2016.01.002

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.
A mechanical model of symmetrical three-roller setting
round process: the static bending stage

Jun Zhaoa,b,*, Gaochao Yua,b,1, Rui Maa,b

a Key Laboratory of Advanced Forging & Stamping Technology and Science (Yanshan University), Ministry of Education

of China, Qinhuangdao City, P.R. China, 066004

b College of Mechanical Engineering, Yanshan University, Qinhuangdao City, P.R. China, 066004

* Corresponding Author.Tel: +86 13933554343

E-mail: zhaojun@ysu.edu.cn(J. Zhao), gch_yu@ysu.edu.cn(G. Yu), mar@ysu.edu.cn(R. Ma).

1
Tel.: +86 15133589652

1
Abstract

The static bending deformation of three-roller setting round process is an elastic-plastic secondary indeterminate problem, and belongs

to the plane bending deformation. This paper adopts a bilinear material model. In consideration of shifting of the tangent points between

the pipe and the lower rollers, the static equilibrium conditions, the physical relationships of elastic-plastic deformation, the deformation

compatibility conditions, and the constraint conditions are analyzed comprehensively. By applying the discrete method, a semi-circular

pipe is meshed equably into N micro-pipe-wall elements with the same geometric parameters. Deformation characteristics of each

micro-pipe-wall element are calculated, and then the deformation history responses of the whole pipe are resolved by applying the load

increment method, based on the variable step size, and the experimental apparatus is further designed to verify the theoretical model.

Finally, the effects of various process parameters on the upper roller load and bending curvature are studied. The results show that the

static bending deformation can be assumed to be a pure bending process; the cross section of the pipe has two positive bending regions

and two reverse bending regions; the upper roller diameter and the lower rollers spacing are the major factors for the upper roller load

and the bending curvature; the dynamic roll bending process can be simplified into a static to-and-fro bending process.

Key words: Three-roller setting round; Secondary indeterminate problem; Mechanical model; Finite element; Process parameters

2
1 Introduction

Metal pipe is one of the essential materials in engineering, widely used in electric power, petrochemical industry,

shipbuilding, construction and other industries. Roll bending is a kind of pipe forming method with the great potential that

has a wide range of processing, large flexibility, and other characteristics. Symmetrical three-roller bending is one of the

simplest and most common roll bendings. According to the principle of a circle determined by three points, with

underdraught of the upper roller and rotation of the two lower rollers, the sheet metal undergoes a continuous three-point

plastic bending, and then becomes a predetermined shape of the workpiece. The forming process includes three stages:

pre-bending, roll bending, and setting round, as shown in Fig.1.

In recent years, many scholars have carried out a lot of research on the mechanical model, forming simulation and

process parameters optimization in the pre-bending and roll bending stages, which has laid foundations for the theoretical

analysis of setting round in this paper.

Pre-bending: Gandhi et al. (2008) developed a mathematical model of upper roller load for pre-bending of

three-roller plate bending process, which can be used for capacity assessment and analysis of roller bending machines.

After considered effects of the various process parameters, Zhang et al. (2014) also developed a mathematical model

of pre-bending, and identified the influence of bottom roller radius, relative curvature, and bending arc length on the

springback angle, as well as on relationship between the springback ratio and the edge pre-bending angle.

Roll bending: With considering the tangent point movement at the interface between the lower roller and the plate,

Gandhi et al. (2008) proposed an analytical and empirical model to estimate the upper roller position explicitly as a

function of the desired radius of curvature for three-roller cylindrical bending of plates. Fu et al. (2013) proposed an

analytical model for three-roll bending forming of sheet metal and carried out the simulation and experiment to

optimize the process parameters and verify the theoretical model. Zeng et al. (2008) established a simulation model to

study the dynamic process of roll plate bending by using the finite element method, in which the continuous three-roll

pyramidal bending configuration with conical rollers is used to bend a thick plate into a conical tube. Chudasama et al.

(2014) developed an analytical model for prediction of the force during the three-roller conical bending process,

which consists of various parameters, such as material parameters and geometrical parameters. Based on the finite

element method, Feng et al. completed a lot of research about the roll bending. They built numerical models for

non-kinematical symmetrical roll bending process with cylindrical rolls (Feng et al., 2009) and conical rolls (Feng et

al., 2012), and they also built a model of asymmetrical three-roll bending process (Feng et al., 2011) under

ANSYS/LS-DYNA environment.

However, there are a few of publications on the setting round stage which directly determines quality of the product,

3
without a mechanical model having been reported. In addition, with rapid development of the industrial technology, not

only the bending capacity of rolling machine needs to be higher, but also the bending precision needs to be higher.

Therefore, theoretical guidance of the three-roller setting round process has become an urgent need from pipe production

enterprises.

Due to closure of the welded pipe, the static bending in the setting round process is an elastic-plastic secondary

indeterminate problem. Currently, analysis of the statically indeterminate problem is mainly limited to the plastic collapse

limitation, and focuses on effect of the elastic deformation on the stiffness of whole structure. However, very few

particular studies are available on the mechanical analysis for statically indeterminate beams following a plastic

deformation. With combining the limit analysis theory and elastic-plastic fracture mechanics, Yoo and Ando (2003), as

well as Liu et al. (2010), analyzed and evaluated the plastic collapse and the leak-before-break behavior of a statically

indeterminate piping system, and provided plastic collapse limitation of the statically indeterminate pipe system. Paul et al.

(2009) assessed influence of the crack stability in statically indeterminate beams, in which the load was within the plastic

range, and the results showed that in many cases a fully plastic collapse must lead to a completed failure. Based on the

springback equation of small curvature (Zhao et al., 2011) and the meshing idea, the authors have presented a

quantificationally analytical method for the statically indeterminate problem, such as the circle-oval process (Zhao et al.,

2014), which has provided a way to solve the three roller setting round problem in this paper.

In this paper, movement of the tangent points between the pipe and lower rollers is taken into account. On the basis of

a comprehensive analysis on the static equilibrium condition, the physical relationship of elastic-plastic deformation, the

deformation coordination condition, and the constraint condition, a mechanical model of the static bending stage in the

three-roller setting round process is finally established. With the help of the computer programming, the deformation

history response of the whole pipe is resolved by using the load increment method that is based on a variable step size, and

the experimental apparatus is designed to verify the theoretical model. Finally, effects of the various process parameters on

the upper roller load and the bending curvature are studied.

2 Three-roller setting round process

The three-roller setting round process can be divided into three stages: static bending, roll bending, and unloading. A

non-round pipe is supported by two lower rollers, and is pressed by an upper roller, which will result in an elastic-plastic

bending deformation in the contact area between the pipe and the rollers. Under a constant bending curvature, the rotated

lower rollers drive the pipe rotating, so that arbitrary pipe-wall longitudinal section alternately experiences multiple times

of positive and reverse bending deformation. With rotation of the lower rollers, the upper roller is unloaded gradually to

eventually make the curvature of each pipe-wall section be consistent. The maximum bending curvature, minimum

4
bending curvature, and maximum load in the static bending stage can be approximated as the maximum bending curvature,

minimum bending curvature, and maximum load for the entire three-roller setting round process. Therefore, the study

summarized in this paper mainly focus on the static bending stage.

The symmetrical three-roller setting round device is mainly composed of an upper roller and two lower rollers.

Tangent points between the pipe and the lower rollers are not fixed, but changed with the reduction of the upper roller. As

shown in Fig. 2, at the initial state, the tangent point is at P; while the reduction is H, the tangent point moves to Q. Point

Q ' is the intersection of the normal extending line at point Q and the geometric center layer of pipe, is the central

angle of
A ' Q ' . Because the tangent point Q shifts with change of the reduction, will also change.

3 Establishment and solution of the mechanical model

3.1 Static analysis (outer bending moment)

The three-roller setting round process is a small deformation problem. According to the symmetry of static bending

deformation, a semi-circular pipe is selected as the study object. The mechanical model of the static bending is shown in

Fig.3.

Considering the static equilibrium condition,

Fa +Fb Fn sin 0 Ft cos 0

(1)
F / 2 Fn cos 0 Ft sin 0

Ft Fn (2)

Where, 0 is the initial angle between y axis and the normal line in which the tangent point locates, F is the upper roller

load, Ft is friction, Fn is the support force, Fa is the top horizontal section force, Fb is the bottom horizontal section

force, is friction coefficient.

Assuming that the relationship between the top horizontal section force and the bottom horizontal section force is

Fa / Fa Fb (3)

By the Eqs.(1), (2) and (3), the outer bending moment of arbitrary pipe-wall section can be expressed as

FR sin 0 cos 0 1 sin

Ma 0 / 2 / 2
2 cos 0 sin 0
M (4)
FR sin 0 cos 0 1 sin FR cos 0 + FR 1 sin 0
M a / 2 0 / 2
2 cos 0 sin 0

L
0 =arcsin (5)
2 R 2 R2 t

Where, is the angle between x axis and arbitrary pipe-wall section, M is outer bending moment, M a is outer
5
bending moment of the top section, R is the geometric center layer radius of the pipe, L is the lower roller spacing, t is wall

thickness, R2 is the lower roller radius.

3.2 Elastic-plastic analysis (inner bending moment)

3.2.1 Basic assumptions

(1) Pure bending assumption: the static bending process is considered to be a pure bending process, which means to

ignore the influence of normal stress that the upper roller load F causes and to consider the bending strains only.

(2) Neutral layer coinciding assumption: the strain neutral layer, stress neutral layer and geometric neutral layer

always coincide during the deforming process.

(3) Bilinear hardening material model assumption: Effects of initial strain and change of material properties are

neglected. The relationship between strain and stress is

E s / E
(6)
D 0 s / E

0 1 D / E s (7)

Where, E is Young's modulus, s is yield strength, D is plastic tangent modulus.

(4) Plane section assumption: Any plane section remains plane after deforming and no aberrance occurs. The effective

strain is expressed as

Kw K 0 w (8)

Where, K 0 is the initial curvature of the pipe section neutral layer ( K 0 1 / R ), K is the curvature of the pipe section

neutral layer after loading, w is the distance from the neutral layer.

(5) Uniaxial stress state assumption: any particle on the pipe is uniaxially stretched or compressed when a

deformation occurs.

3.2.2 Relationship between bending curvature and inner bending moment

It is defined that the initial curvature of neutral layer is positive. If the applied bending moment M makes the

curvature larger, the bending is called positive, the region is called positive bending region, and M is also positive. On the

contrary, the bending is called reverse, the region is called reverse bending region, and M is negative. Therefore, the

bending moment and the curvature are both vectors.

By the assumption (4), the maximum strain of pipe-wall section occurs in the outer surface

t
max ( K K 0 ) (9)
2

By the assumption (3), when max s / E , all pipe-wall sections are in an elastic deformation state, so the elastic

limitation curvatures of positive bending and reserve bending are expressed as

6
 K Pe lim K 0 2 s / tE (10)

K Re lim K 0 2 s / tE (11)

Where, K Pe lim is the elastic limitation curvature of positive bending, and K Re lim is the elastic limitation curvature of the

reverse bending.

By the assumptions (2) to (5), the inner moment M can be quantificationally given with integrating the bending stress

over the pipe-wall section by

DI K K 0 0 t 2 / 4 0 ws2 / 3 K K P elim

M wdA EI K K 0 K R elim K K Pe lim (12)
DI K K t 2 / 4 w2 / 3 K K e
0 0 0 s R lim

s
ws (13)
E K K 0

Where, ws is the demarcation point of elastic region and plastic region in the pipe-wall section, I is moment of inertia,

K is bending curvature, M is inner moment.

According to the Shengjin formula (Fan, 1989), K can be expressed in terms of M as

Q
K0 0

Q W cos 3 3 sin 3


W Bt 2 s
K0 0 M
3DI 6
Q
K0 0

2
2

3 Q 3DIQ (3DIQ) 2 4QW 3 3 Q 3DIQ (3DIQ ) 2 4QW 3


(14)

M Bt s
2
Bt 2 s
K K 0 M
EI 6 6

Q
Q W cos 3 3 sin 3 K 0 0


W Bt 2 s
K0 0 M
3DI 6
Q
K0 0
3Q

2
3DIQ (3DIQ) 4QW
2 3 3
Q
2

3DIQ (3DIQ ) 4QW
2 3

3DIQ 2
Where, W M 0 t 2 4 , Q =( E D) s3 , 9 D 2 I 2 Q 4W 3 , T , arccosT .
QW
3
2

3.3 Discretization

Adopting the discrete method, the semi-circular pipe is meshed into N parts uniformly along with the circumferential

direction, which means that the semi-circular pipe is composed of N micro-pipe-wall elements with the same geometric

parameters. The neutral layer of each micro-pipe-wall element is connected with nodes. As shown in Fig.4, starting from

the origin of coordinate x0 o0 y0 , the element number is followed by , , N , and the node number is followed by
7
0, 1, 2, 3 ... N. Let node i is the origin of coordinates, the tangential direction of the neutral layer is xi axis, and the

normal direction is the yi axis, the local follow-up coordinate system xi oi yi is built.

Taking a micro-pipe-wall element as the study object, as shown in Fig.5, in the coordinate system xi 1oi 1 yi 1 the

coordinate of end node i in the initial state can be expressed as

xii01 R sin i 0
i 1 (15)
yi 0 R 1 cos i 0

i 0 = / N (16)

Where, xii01 is the initial x coordinate value of node i in the coordinate system xi 1oi 1 yi 1 , yii01 is the initial y

coordinate value of node i in the coordinate system xi 1oi 1 yi 1 , i 0 is the initial angle between the yi 1 axis and the

pipe-wall section in which the node i locates.

The coordinates after loading are ( xii 1 , yii 1 ) with the value

i 1 1
xi K sin i
i
(17)
1 cos i
y i 1 1
i Ki

Where, K i is the curvature of micro-pipe-wall neutral layer after loading, i is the angle between the xi 1 axis and

the pipe-wall section in which the node i locates after loading.

According to the assumption (1), the length of the micro-pipe-wall element is unchanged after deformation

i K i R i 0 (18)

To obtain the profile curve of pipe section neutral layer after loading, the coordinates of nodes are needed to be

unified as relations of the local coordinate systems. The coordinates of node i in the coordinate system x j o j y j ( j i ) are

( xij , yij ) . According to definition of the local coordinate system, when j i , the values of ( xij , yij ) are

xi 0
i

i (19)
yi 0

By Eqs.(15), (16), (17), (18) and (19), the coordinates of each node are unified to the local coordinate system x0 o0 y0 ,

the iterative formula is given by

xi0 xi01 xii 1 cos i 1 yii 1 sin i 1

0 (20)
yi yi 1 xi sin i 1 yi cos i 1
0 i 1 i 1

i 1
i 1 R 0 K i (21)
1

8
Where, i 1 is the angle between the y0 axis and the pipe-wall section in which the node i-1 locates after loading.

3.4 Solution

3.4.1 Compatibility equation of deformation

Since the static bending deformation is a secondary statically indeterminate problem, it is known that two

compatibility equations of deformation are required for a solution.

According to the continuity and symmetry of the static bending process, the pipe-wall section in which the node 0

locates is always parallel to the x0 axis, and the pipe-wall section in which the node N locates is always parallel to the

y0 axis in the deformation process, thus

i 1
i (22)

Substituting Eqs.(16) and (18) into Eq.(22) and simplifying, a compatibility equation of deformation is given by

K
i 1
i N/R (23)

Moreover, the lowest and the highest point (Node 0 and Node N) of the neutral layer are always on the vertical

symmetry plane, so the second compatibility equation of deformation is given by

xN 0 (24)

3.4.2 Constraint equation

When deformation of the pipe is large enough, inner wall of the pipe will fit with the upper roller surface, so the pipe

section can be divided into two regions: the non-attaching-roller region and the attaching-roller region. In the

attaching-roller region, because the pipe-wall element is constrained by the upper roller, the curvature of the pipe inner

wall can not be greater than the curvature of the upper roller, that is, the curvature of the pipe neutral layer should meet the

following condition

2
Ki (25)
2 R1 +t

Where, R1 is the upper roller radius.

Therefore, when the calculated K i by Eq.(14) is not satisfied by Eq.(25), K i should be

2
Ki (26)
2 R1 +t

3.4.3 The load increment method

In order to improve the accuracy and speed, the response relationship among the upper roller load F, the bending

curvature K, and the reduction H is obtained by using the load increment method that is based on a variable step size. The

9
load F can be divided into many small increments to apply gradually, and then iteration is used to solve the curvature of

each element and the coordinates of each node in each incremental step.

Setting up each load increment as f, when the initial incremental step (k=1) is loaded, according to the Eq.(4), the

outer bending moment of arbitrary pipe-wall section is given by

fR sin 0 cos 0 1 sin i 1

M a 0 / 2 i / 2
2 cos 0 sin 0

1 fR sin 0 cos 0 1 sin i fR cos 0 i
M i (27)
2 cos 0 sin 0 2 cos 0 sin 0
fR 1 sin
+ 0 i 1
Ma / 2 i 0 / 2
2 cos 0 sin 0

i 1 / 2 i / N (28)

1
Where, i is the angle between the x axis and the pipe-wall section in which the node i locates, M i is the outer

1
bending moment of the pipe-wall section in which the node i locates after the first loading, M a is the outer bending

moment of the top section after the first loading.

When k 2 , because shape of the pipe is changed with increase of F, point P is also changed. According to

geometric relations, search each node (0~ N / 2 ) to determine node number of the tangent point by the following equation

L t
i i N MIN xi0 R2 sin i 1 , i 0,1, 2...N / 2 (29)
T
2 2

Where, NT is node number of the tangent point.

Furthermore, the outer bending moment of the pipe-wall section in which the node i locates after the k times loading

is expressed as

kf sin
NT k 1
cos NT
k 1
yN0
k 1
yi0
k 1
Ma
k
NT k 1 i N

2 cos NT k 1
sin
NT k 1



kf sin NT k 1 cos NT k 1 yN k 1 yi k 1
0 0




2 cos NT sin NT
k 1

k 1
k
M i y NT
0
yi 0
yN0 T yi0 (30)

kf cos NT ar tan 0 k 1 k 1
kf sin ar tan k 1 k 1

T k 1
N
k 1
x NT xi0 xN0 T xi0
k 1 k 1

k 1 k 1

+ *
2 cos NT sin NT 2 cos NT sin NT
k 1 k 1 k 1 k 1





2 2 k
xNT k 1 xi k 1 yNT k 1 yi k 1 M a 0 i NT k 1
0 0 0 0

Where, NT k 1
is the node number of the tangent point after the k-1 times loading.

10
 In the deformation process, the inner and outer bending moments of the arbitrary section are equal.

M M i (31)

Replacing Eqs.(27), (28), (30) and (31) in Eq.(14), K i in the non-attaching-roller region should be
K i , M a , F (32)

When F is known, Eqs.(23), (24) and (32) are composed of an equation set about M a , and K i (i=1, 2,, N),

which includes N+2 equations. In each incremental step, taking Eqs.(23) and (24) as the criterion, the equation set can be

solved by two-dimensional search of M a and . The initial iteration values of M a and are set to zero. The

iterative algorithm of variable step size is adopted ( ' / 2 , ' / 2 ), as shown in Fig.6, and then M i , K i and

coordinates in the local coordinate system x0 o0 y0 can be given by Eqs.(27), (28), (32), (20) and (21).

4 Experimental design

In order to verify correctness of the theoretical model, a small scale three-roller bending device is developed, which

composes of upper roller, lower roller, U-shape plate, slider and frame, as shown in Fig.7 (a). Through bearings, the two

lower rollers are fixed on the frame, and the upper roller is fixed on two slides, which are connected with the punch

through the U-shape plate and are able to slide along the vertical direction. The process parameters are shown in Tab.1.

The experiment is carried out on the WDD-LCT-150 electronic tension-torsion multifunctional test machine, whose

displacement accuracy is 0.01mm and maximum allowable load is 150kN. The outline of pipe is measured by using

3000iTM series portable coordinate measuring machine (CMM) that has a measurement accuracy of 0.01mm, as shown in

Fig.7 (b). The pipe material is AISI 1045 medium carbon steel in accordance with ASTM A29/A29M-05. The material

properties used in the theoretical analysis are obtained experimentally from tensile tests using ASTM-E8 standard tension

test specimens that are cut off circumferentially from the pipe and then flatted, as shown in Fig.8. Two sets of wall

thickness pipes are selected for the experiment, and their geometrical dimensions and mechanical properties are shown in

Tab.2.

5 Results and discussions

Based on the above mechanical model, it is assumed that the semi-circular pipe is composed of 1000 micro-pipe-wall

elements, namely N=1000. Results of the section forces, bending moments, bending curvatures and upper roller load are

presented in this section.

Fig.9 shows curves of the horizontal section forces and moments with reduction. It can be seen that Fa and Fb

both increase with the increase of reduction H, and the two curves obviously appear the turning points, that are the

elastic-plastic boundary points, where H is approximately 0.74mm when t=2.5mm, and 0.47mm when t=3.8mm.

According to establishment and solution of the mechanical model, normal stress of the symmetric section at bottom of the
11
pipe reaches the maximum. When t=2.5mm, H=3.93mm, the maximum stress is 5.03MPa s , and when t=3.8mm,

H=4.06mm, the maximum stress is 8.18MPa s . Thus it proves correctness of the pure bending assumption. M a and

M b increase with the increase of reduction H, and then M b becomes a fixed value, which indicates that the

attaching-roller stage starts.

Fig.10 shows the circumferential distribution of the bending curvature and moment along the semi-circular pipe. It

can be observed that there are two positive bending regions and two reverse bending regions in whole section of the pipe.

The bending curvature and moment of elements at the beginning local area are the same, indicating that this area is the

attaching-roller region. In addition, the maximum bending curvature K max (maximum bending moment M max ) appears

in the bottom symmetric section of the pipe, while the minimum bending curvature K min (minimum bending moment

M min ) appears in the pipe-wall section in which the tangent point locates.

The curves of maximum bending curvature K max and minimum bending curvature K min with the reduction H are

given in Fig.11. It can be seen that the maximum bending curvature increases slowly with increase of the reduction in the

elastic range, and then increases rapidly after the elastic-plastic boundary point, and finally keeps unchanged until the local

inner wall of the pipe attaches to the upper roller. In addition, the minimum bending curvature decreases with increase of

the reduction. It is noteworthy that the reduced speed of the minimum curvature is less than the increased speed of the

maximum curvature, which indicates that the plastic deformation occurs in the positive bending region earlier than in the

reserve bending region.

Fig.12 shows the outline of pipe after loading. According to the curvature value of the micro-pipe-wall element after

loading, the outline of pipe obtained by the calculation is divided into the positive bending region and the reverse bending

region. Compared with the initial ideal circle, the outline of pipe becomes longer along the vertical direction and shorter

along the horizontal direction, resembles an egg shape. Compared with the experimental measurement points, the tendency

of the calculational fitting curve is in good agreement with that of the experiment. Unfortunately, due to the limitation of

the rollers, a small part of the bottom of the pipe failed to be measured. Fig.13 shows the relationship between the upper

roller load and the reduction. It can be seen that the load increases with increase of the reduction. The trends from

experimental curves and from theoretical points are in the same way, and the overall error is very small. When t=2.5mm,

the maximum relative error is 12.4%, and when t=3.8mm, the maximum relative error is 5.3%. The error of predicted

results is mainly caused by the bilinear mechanical model, initial round shape assumption and program calculation. The

12
experimental results can prove reliability of the theoretical mechanical model for the engineering application.

6 Process parameters

Based on the established three-roller static bending mechanical model and the above discussions, the effects of

process parameters on the load, maximum bending curvature and minimum bending curvature are studied in this

section. Taking an engineering real pipe as the research object, its outside diameter is 492mm, the wall thickness

12mm, and the length 3000mm. The material mechanical properties are as listed in Tab.2, and values of the process

parameters are shown in Tab.3. According to Eqs.(10) and (11), K Pe lim and K Re lim are 4.4510-3mm-1 and 3.89

10-3mm-1, respectively.

6.1 Upper roller diameter

As shown in Fig.14 (a), in the non-attaching-roller stage the upper roller diameter has no obvious effects on the load,

while in the attaching-roller stage an increase of the upper roller diameter results in the load increase. When the upper

roller diameter is 396mm and the reduction H reaches 6mm, the load rises exponentially under the rigid constraint of lower

rollers. It can be seen from Fig.14 (b) that the upper roller diameter has little effect on the maximum bending

curvature K max and the minimum bending curvature K min in the elastic range. While after the plastic deformation occurs,

the upper roller diameter is greater, the maximum bending curvature K max and the minimum bending curvature K min are

smaller.

6.2 Lower roller diameter

As shown in Fig.15, the lower roller diameter has little effect on the load and bending curvature, which indicates that

position change of the tangent point between the lower roller and the pipe is very small and can be neglected, through the

lower roller is with different diameters.

6.3 Lower roller spacing

As shown in Fig.16 (a), the load decreases with the increase of the lower roller spacing. When the lower roller

spacing L is 450mm and the reduction H reaches 4.5mm, the load rises rapidly under the rigid constraint of lower roller. It

can be observed from Fig.16 (b) that the lower roller spacing is larger, the maximum bending curvature K max is smaller,

and the required reduction is larger in order to attach to the upper roller. Therefore, an increase of the lower roller spacing

is useful to delay the attaching-roller. The lower roller spacing has no obvious effect on the minimum bending curvature in

the elastic range, while the minimum bending curvature increases with an increase of the lower roller spacing, after the

plastic deformation occurs.

13
6.4 Friction coefficient

As shown in Fig.17 (a), the load has a small amount of increase with the increase of the friction coefficient. It can be

observed from Fig.17 (b) that the friction coefficient has no obvious effect on the bending curvature. It can be predicted

that, in the follow up dynamic roll bending stage, the friction force plays a key role in driving rolling of the pipe, and its

effect on the bending curvature can be neglected. Therefore, under the same roller configuration and same reduction, the

static bending stage of the closed pipe can be approximated to the transient state of the dynamic roll bending. That is, in

the roll bending stage, any micro-pipe-wall element alternately experiences positive-reverse bending for many times, and

values of the bending curvature are results from the static bending analysis. In this way, the dynamic roll bending process

is simplified to a static to-and-fro bending process.

7 Conclusions

In this paper, the static bending deformation in the symmetrical three-roller setting round process is mainly studied.

Due to closure of the welded pipe, the static bending is an elastic-plastic secondary indeterminate problem. Based on the

bilinear material model and the shifting of tangent points between the pipe and lower rollers, the static equilibrium

conditions, physical relationship of elastic-plastic deformation, deformation compatibility conditions and boundary

constraint conditions are all taken into account. The discrete method and the load increment method based on a variable

step size are used to resolve the problem programmatically, and the physical experiment is designed. The conclusions are

as follows

(1) The mechanical model of the static bending deformation in the symmetrical three-roller setting round process is

established, and the quantitative relationships between the upper roller load, bending curvature of each micro-pipe-wall

element and the reduction are predicted. This not only lays a theoretic foundation for the development of the three-roller

special setting round machine and control strategies, but also provides an idea for resolving a many degrees of statically

indeterminate problem with an elastic-plastic deformation.

(2) It is proved that the static bending deformation in the symmetrical three-roller setting round process can be

assumed to be a pure bending process, namely the normal stress caused by the upper roller load can be neglected.

(3) In the static bending deformation, there are two positive bending regions and two reverse bending regions in

whole section of the pipe. In addition, the maximum bending curvature appears in bottom symmetric section of the pipe,

and the minimum bending curvature appears in the pipe-wall section in which the tangent point locates.

(4) The upper roller diameter and the lower roller spacing are the major factors for the upper roller load and the

bending curvature, while the lower roller diameter and the friction coefficient have no obvious effects.

(5) Under the same roller configuration and the same reduction, the static bending stage of the closed pipe can be

approximated to the transient state of the dynamic roll bending. That is, the dynamic roll bending process is simplified to a

14
static to-and-fro bending process.

Acknowledgements

This project was funded and supported by the National Natural Science Foundation of China (Grant No.

51575473and 51175452), the Natural Science Foundation of Hebei Province in China (Grant No. E2015203244), and the

Youth Talent Support Plan of Hebei Province in China.

15
References

Chudasama, M.K., Raval, H.K., 2014. Bending force prediction for dynamic roll-bending during 3-roller conical bending process. Journal of
Manufacturing Processes. 16, 284-295.
Fan, S., 1989. A new extracting formula and a new distinguishing means on the one variable cubic equation. Nature Science Journal of Hainan Teachers
College. 2, 91-98.
Feng, Z., Champliaud, H., Dao, T.M., 2009. Numerical study of non-kinematical conical bending with cylindrical rolls. Simulation Modelling Practice
and Theory. 17, 1710-1722.
Feng, Z., Champliaud, H., 2012. Investigation of non-kinematic conical roll bending process with conical rolls. Simulation Modelling Practice and
Theory. 27, 65-75.
Feng, Z., Champliaud, H., 2011. Modeling and simulation of asymmetrical three-roll bending process. Simulation Modelling Practice and Theory. 19,
1913-1917.
Fu, Z., Tian, X., Chen, W., Hu, B., Yao, X., 2013. Analytical modeling and numerical simulation for three-roll bending forming of sheet metal.
International Journal of Advanced Manufacturing Technology. 69, 1639-1647.
Gandhi, A.H., Gajjar, H.V., Raval, H.K., 2009. Asme. MATHEMATICAL MODELLING AND FINITE ELEMENT SIMULATION OF PRE-BENDING
STAGE OF THREE-ROLLER PLATE BENDING PROCESS. Msec 2008: Proceedings of the Asme International Manufacturing Science and
Engineering Conference 2008, Vol 1, 617-625.
Gandhi, A.H., Raval, H.K., 2008. Analytical and empirical modeling of top roller position for three-roller cylindrical bending of plates and its
experimental verification. Journal of Materials Processing Technology. 197, 268-278.
Liu, S.P., 2010. Remaining life assessment and optimal design of statically indeterminate pipe system with circumferential crack. International
Communications in Heat and Mass Transfer. 37, 266-273.
Paris, P.C., Palin-Luc, T., 2009. The behavior of statically-indeterminate structural members and frames with cracks present. Engineering Fracture
Mechanics. 76, 1920-1929.
Yoo, Y.S., Ando, K., 2003. Leak-before-break behavior of statically indeterminate piping system. Nuclear Engineering and Design. 224, 253-263.
Zeng, J., Liu, Z., Champliaud, H., 2008. FEM dynamic simulation and analysis of the roll-bending process for forming a conical tube. Journal of
Materials Processing Technology. 198, 330-343.
Zhang, Z.Q., Song, J.L., Fu, J.H., Li, Y.T., Guo, Y.N., 2014. A Refined Model of Three-roller Elastoplastic Asymmetrical Pre-bending of Plate. Journal of
Iron and Steel Research International. 21, 328-334.
Zhao, J., Yin, J., Ma, R., Ma, L., 2011. Springback equation of small curvature plane bending. Science China-Technological Sciences. 54, 2386-2396.
Zhao, J., Zhan, P., Ma, R., Zhai, R., 2014. Prediction and control of springback in setting round process for pipe-end of large pipe. International Journal
of Pressure Vessels and Piping. 116, 56-64.

16
Figure legends

(a) (b)

(c) (d)

Fig. 1 Whole process of three-roller bending. (a) pre-bending; (b) initial stage of roll bending; (c) end stage of roll bending

(d) setting round

17
 Before loading
After loading

y
Pipe Upper Roller, R1

H O
O' Lower Roller, R2
Q'
A P
Q
A'
O1 O2
L

Fig. 2 Schematic of static bending by three rollers

18
 Fa

Ma

0 Fn
F
Ft
2

M b Fb

Fig. 3 Mechanical model of the semi-circular pipe

19
 xN N ( oN )

yN N

y0
y1
y2
x2

2 x1
1 2 o2
1 o1
0 o0 x0

Fig. 4 The number of nodes and elements and the local coordinate systems

20
 yi 1

i 0

1
Ki

oi 1 xi 1

Fig. 5 The micro-pipe-wall element before and after deformation

21
x0 [i ], , y0 [i]

N
N

i1
K i
R

N
N

i 1
K i
R

x0 [i ] y0 [i ]

x0 [ N ]

x0 [ N ]

H k H goal

Fig. 6 Program flow chart

22
U-shape plate

Pipe

Upper roller Slide

Lower roller

Flame

(a)

(b)

Fig. 7 Experimental equipment

23
 100.05
0
R2

20
(a) (b)

Fig. 8 Uniaxial tensile specimen. (a) Specimen preparation; (b) Dimensions

24
 1.5 60 3.5 150

50 3.0
1.2 120

Bending moment/(kNmm)
2.5
40
Section force/kN

Fa

Moment/(kNmm)
Section force/kN
0.9 90
Fa 2.0 Fb
Fb 30 Ma
0.6
Ma 1.5 Mb
60
Mb 20
1.0
0.3 30
10
0.5

0.0 0 0.0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Reduction H/mm Reduction H/mm

(a) (b)

Fig. 9 Section forces and moments with reduction curve. (a) t=2.5mm; (b) t=3.8mm

25
 4.0 80 4.0 160
Curvature K H=3.93mm Curvature K H=4.06mm
3.5 Moment M 60 3.5 120
Moment M

Positive bending region

Positive bending region
Positive bending region
Positive bending region
3.0 40 80
Curvature K/(10 mm )
3.0

Curvature K/(10 mm )
-1

Moment M/(kNmm)
-1

Moment M/(kNmm)
Reverse bending region Reverse bending region 40
2.5 20
-2

2.5

-2
0
2.0 0 2.0
-40
1.5 -20 1.5
-80

1.0 -40 1.0 -120

0.5 -60 0.5 -160

0 200 400 600 800 1000 0 200 400 600 800 1000
Node ID Node ID

(a) (b)

Fig. 10 Distribution of bending curvature and moment. (a) t=2.5mm; (b) t=3.8mm

26
 3.5

3.0

Curvature/(10 mm )
-1
2.5
Kmax.t=2.5mm
Kmin.t=2.5mm

-2
Kmax.t=3.8mm
2.0
Kmin.t=3.8mm

1.5

1.0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Reduction H/mm

Fig. 11 K max and K min with reduction curve

27
 90 90
Positive bending region Positive bending region

60 60

30 H=3.93mm 30 H=4.06mm
Calculation Calculation
Experiment Experiment
Initial ideal circle Initial ideal circle
y/mm

y/mm
0 0

Reverse bending region Reverse bending region

-30 -30

-60 -60

Positive bending region Positive bending region

-90 -90
30 60 90 30 60 90
x/mm x/mm

(a) (b)

Fig. 12 Outline of pipe after loading. (a) t=2.5mm; (b) t=3.8mm

28
 25

Exp. t=2.5mm
20 Cal. t=2.5mm
Exp. t=3.8mm
Cal. t=3.8mm
15

Load F/kN
10

0
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Reduction H/mm

Fig. 13 Load with reduction curve

29
 1400 6.5
Kmax.D1=328mm
D1=328mm Kmin.D1=328mm
1200 6.0
D1=359mm Kmax.D1=359mm
Kmin.D1=359mm
1000 D1=396mm

Curvature/(10 mm )
-1
5.5 Kmax.D1=396mm
Kmin.D1=396mm
800
Load F/kN

-2
e e
5.0 KPlim KRlim

600
4.5
400

4.0
200

0 3.5
0.0 1.5 3.0 4.5 6.0 7.5 9.0 0.0 1.5 3.0 4.5 6.0 7.5 9.0
Reduction H/mm Reduction H/mm

(a) (b)

Fig. 14 Effect of upper roller diameter. (a) load; (b) curvature

30
 1000 6.5
Kmax.D2=240mm
D2=240mm Kmin.D2=240mm
6.0
D2=280mm Kmax.D2=280mm
800
Kmin.D2=280mm
D2=320mm

Curvature/(10 mm )
-1
5.5 Kmax.D2=320mm
600 Kmin.D2=320mm
Load F/kN

-2
e e
5.0 KPlim KRlim

400
4.5

200 4.0

0 3.5
0.0 1.5 3.0 4.5 6.0 7.5 9.0 0.0 1.5 3.0 4.5 6.0 7.5 9.0
Reduction H/mm Reduction H/mm

(a) (b)

Fig. 15 Effect of lower roller diameter. (a) load; (b) curvature

31
 1400 6.5
Kmax.L=450mm
L=450mm Kmin.L=450mm
1200 6.0
L=500mm Kmax.L=450mm
L=550mm Kmin.L=500mm
1000

Curvature/(10 mm )
Kmax.L=550mm

-1
5.5
Kmin.L=550mm
800
Load F/kN

-2
e e
KPlim KRlim
5.0
600
4.5
400

4.0
200

0 3.5
0.0 1.5 3.0 4.5 6.0 7.5 9.0 0.0 1.5 3.0 4.5 6.0 7.5 9.0
Reduction H/mm Reduction H/mm

(a) (b)

Fig. 16 Effect of lower roller spacing. (a) load; (b) curvature

32
 1000 6.5
Kmax.=0.1
=0.1 Kmin.=0.1
=0.2 6.0
800 Kmax.=0.2
=0.3 Kmin.=0.2

Curvature/(10 mm )
-1
5.5 Kmax.=0.3
600 Kmin.=0.3
Load F/kN

-2
e e
5.0 KPlim KRlim

400
4.5

200
4.0

0 3.5
0.0 1.5 3.0 4.5 6.0 7.5 9.0 0.0 1.5 3.0 4.5 6.0 7.5 9.0
Reduction H/mm Reduction H/mm

(a) (b)

Fig. 17 Effect of friction coefficient. (a) load; (b) curvature

33
 Table 1 Process parameters

Upper roller diameter Lower roller diameter Lower roller spacing Friction coefficient
D1/mm D2/mm L/mm
60 45 85 0.1

Table 2 Geometric dimensions and mechanical properties of pipe. (D0 is outer diameter of pipe, l is length of pipe.)

D0/mm t/mm l/mm E/GPa D/MPa s /MPa

2.5
160 100 206 2160 345
3.8

Table 3 Process parameters

Upper roller diameter Lower roller diameter Lower roller spacing Friction coefficient
D1/mm D2/mm L/mm
328
359 280 500 0.2
396
240 474
359 280 500 0.2
320 526
450
359 280 500 0.2
550
0.1
359 280 500 0.2
0.3
Note: when the lower roller diameter is changed, the lower roller spacing is unchanged.

34