Materials and Design 31 (2010) 2001–2010

Contents lists available at ScienceDirect

Materials and Design

journal homepage: www.elsevier.com/locate/matdes

Coupling effects of material properties and the bending angle on the

springback angle of a titanium alloy tube during numerically controlled bending
Z.Q. Jiang a, H. Yang a, M. Zhan a,*, X.D. Xu b, G.J. Li b
a
School of Materials Science and Engineering, State Key Laboratory of Solidification Processing, Northwestern Polytechnical University, Xi’an 710072, China
b
Chengdu Aircraft Industry (Group) Corporation Ltd., Chengdu 610092, China

a r t i c l e i n f o a b s t r a c t

Article history: The significant springback after the numerically controlled (NC) bending of a titanium alloy tube has an
Received 15 June 2009 important influence on the precision of the shape and size of the bent tube. This springback depends on
Accepted 16 October 2009 the material properties of the tube, the bending angle, and especially their coupling effects. The influence
Available online 21 October 2009
of some material properties and the bending angle on the springback angle in the NC bending of a TA18
tube were investigated using a three-dimensional (3D) elastic–plastic finite element model. Using mul-
Keywords: tivariate and stepwise analyses, the coupling effects of the bending angle and the material properties on
Titanium alloy tube
the springback angle during NC bending were revealed. It was observed that Young’s modulus, yield
NC bending
Springback
stress, the strain hardening coefficient and exponent, and the thickness anisotropy exponent, as well
Numerical simulation as interactions of these parameters with the bending angle, have a significant influence on the springback
Coupling effects angle. The bending angle, yield stress, and hardening coefficient have positive effects on the springback
Stepwise regression analysis angle, and Young’s modulus, the hardening exponent, and the thickness anisotropy exponent have neg-
ative effects. The influence of the material properties of the titanium alloy increases with the bending
angle. Young’s modulus and the strain hardening coefficient and exponent have the greatest influence
on the springback angle. The results will be very useful in predicting, compensating for and controlling
the springback of titanium alloy tubes during NC bending.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction For the bending of TA18 M titanium alloy tubes with the same
bending radius but different values of outside diameters D and wall
The demand for bent tubes with high pressure resistance, a high thickness t using an NC bending machine, it was found that the
strength/weight ratio and long life, as well as the demand for bend- variations in springback angle with the bending angle for different
ing processes with high precision and manufacturing efficiency for tubes are complicated, as shown in Fig. 1. This results from the
the aviation and aerospace industries, have led to increased inter- complicated coupling effects between material properties and the
est in and application of numerically controlled (NC) tube bending bending angle. Though these tubes have a similar geometric ratio
technology for the rotary bending of titanium alloy tubes [1,2]. of D/t, they have different material properties (including Young’s
Titanium alloys exhibit poor plastic deformation, high strength, modulus, yield stress, strain hardening coefficient and exponent,
and a low Young modulus [3]. Owing to elastic recovery of the Poisson’s ratio, and thickness anisotropy exponent) owing to the
material, significant springback occurs after bending [4], which complexity of tube manufacturing process, even within the same
leads to an increase in the bending radius and a decrease in the batch. Thus, it is necessary to research the coupling effects be-
bending angle. This has an important influence on the precision tween material properties and bending angle and determine the
of the shape of the bent tube, and the accuracy of the bent tube of- significance rank of these material parameters to accurately pre-
ten does not satisfy requirements for assembly and application. Re- dict the springback angle and control the bending angle of the tita-
search by Lou and Stelson revealed that springback is the major nium alloy tube during NC bending.
source of error in bent tube geometry and is hard to predict. There- Research into springback during tube bending has involved
fore, efforts should be made to improve springback prediction to experiments, theoretical analysis, numerical–analytical analysis,
reduce errors in bending angles [5], especially for the bending of and numerical simulation. Lou and Stelson [6] used springback
a titanium alloy tube. data measured online for the same batch to predict and compen-
sate for springback. E et al. [7,8], and Murata et al. [9] investigated
* Corresponding author. Tel.: +86 29 88460212x805; fax: +86 29 88495632.
the springback in the draw-bending and press bending, respec-
E-mail address: [email protected] (M. Zhan). tively, of aluminium alloy and stainless steel tubes by experiments.

0261-3069/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2009.10.029
2002 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010

nium alloy tube, the complexity of its bending and its high price,
it is difficult to investigate coupling effects of the bending angle
and the material properties using purely analytical or experimental
methods. Therefore, it is essential to systematically study the
bending of titanium alloy tube using FE numerical simulation. In
the present study, we used a combination of explicit and implicit
elastic–plastic FEMs to investigate the effects of bending angle
and the material properties on the springback angle of titanium al-
loy tube. The material properties include Young’s modulus, yield
stress, the strain hardening exponent, the strain hardening coeffi-
cient, Poisson’s ratio, and the thickness anisotropy exponent. Using
multivariate and stepwise regression analysis methods, the cou-
pling effects between material properties and the bending angle
were identified and the significance of their influence was as-
sessed. This study may serve as an important guide for predicting,
compensating and controlling the springback in the NC bending of
titanium alloy tubes.

Fig. 1. Variation of springback angle with bending angle for tubes with similar
values of D/t but different material properties. 2. Method

2.1. Analysis procedure

Tang [10] employed plastic-deformation theory to investigate the
tube bending and obtained stresses and bending moment in the Fig. 2 presents the analysis procedure. Since several factors af-
process, which is help for springback analysis. Using the theory fect the NC bending process, simulations were designed using the
of beam bending, Al-Qureshi and Russo [11,12] presented a theo- single factorial experimental design (SFED), and multiple polyno-
retical analysis model of springback for tube bending assuming mial regression analysis (MPRA) and stepwise regression analysis
elastic–perfectly plastic material, plane strain conditions, and the (SRA) [20] were chosen to assess the effects of the parameters.
absence of defects and Bauschinger effects. Using the theoretical The significance of the material properties was assessed using
model of tube bending, the bending radius and residual stress after standard methods. A three-dimensional (3D) FE model was
springback can be obtained. Using the same method and taking the adopted to simulate the bending and springback processes and im-
work hardening behavior of the tube into consideration, El Meg- prove the analysis accuracy and efficiency and the universal appli-
harbel et al. [13] developed improved theoretical models for cability of the results.
springback during tube bending. Their theoretical results for
springback of aluminium, stainless steel, copper, and titanium
tubes showed remarkable agreement with experimental data, but
the springback was only dependent on the yield stress, bending ra-
dius, diameter and wall thickness of the tube, and hardening expo-
nent. Moreover, some inevitable behavior during tube bending was
not considered, such as wall thickness variations, cross-section
flattening, strain neutral axis shift, and differences in stress and
strain fields under various bending conditions. Considering the
above deformation behaviors, Zhan et al. [14,15] proposed a
numerical–analytical method for predicting the springback angle
involving rigid-plastic finite element (FE) simulation of the tube
bending and theoretical analysis of the springback. The method
was more computationally efficient with satisfactory accuracy for
analysis of the bending and springback of thin-walled tubes of alu-
minium alloy and stainless steel. It helped in identifying the source
of springback and changes in features and in multifactorial analy-
sis. However, the numerical–analytical method may not be suit-
able for the springback analysis of the bending of a titanium
alloy tube. Because titanium alloys have high strength, poor plas-
ticity, and a low Young modulus, greater elastic deformation oc-
curs during bending [16,17], so there might be an obvious
difference between the stress–strain fields obtained using a rigid-
plastic finite element method (FEM) and real values. In recent
years, a combination of explicit and implicit FEMs has been used
to investigate springback for thin-walled stainless steel and alu-
minium alloy tubes [18,19]. In addition, Murata et al. [9] investi-
gated the influence of hardening exponent on the springback in
the press bending of tubes by means of explicit FEM.
However, much of this work has focused on the bending of
stainless steel and aluminium tubes [7–9,18,19]. To date, very
few reports on the springback of titanium alloy tubes during bend-
ing have been published. Owing to the material features of tita- Fig. 2. Analysis procedure.
 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010 2003

pression–torsion method and an MMX-2 friction and abrasion test-

ing machine.
In this study, two parameter types were investigated: the bend-
ing angle and the material properties (E, rs, n, K, m, and r). Variable
values are listed in Table 1, where values in bold are the default
values adopted in simulations when investigating the influence
of other parameters on the process.

2.3. FE model
Fig. 3. Tension test specimen and plug.
A 3D FE model was developed to simulate the NC bending of a
titanium alloy tube using ABAQUS/Explicit and Implicit code [16],
as shown in Fig. 5. The Explicit dynamic FE code was used to sim-
800
ulate the bending process, and the Implicit static FE code was used
to simulate the unloading springback process. This approach
700 avoids long computation times and improves the computation effi-
ciency. In the model, the tube is defined as a 3D deformable solid
600 body, and the bending, clamp and pressure dies are defined as dis-
Nominal stress / MPa

crete rigid bodies. Each rigid body is assigned a reference point to

500 represent rigid motion for all degrees of freedom. The constitutive
equation for the tube used in the models is r ¼ K en , where r is the
400 true stress and e is the true strain.
Specimen 1 The model was validated in terms of the bent tube configura-
300 Specimen 2 tion, wall thickness and springback angle Dh [16].
Specimen 3
200
2.4. MPRA
100
After simulating NC bending of TA18 alloy tubes under different
0
conditions, Dh was calculated. Regression equations for the param-
eters and indexes were identified using MPRA, which comprises
0.00 0.05 0.10 0.15 0.20
both mathematical and statistical techniques and is beneficial for
Nominal strain
investigating phenomena having several independent variables.
Fig. 4. Relationship between nominal stress and strain for titanium alloy tubes. MPRA considers the influence of independent variables on a spe-
cific dependent variable (response). Initially, the independent vari-
ables of a problem are denoted by x1, x2, . . ., xk. These variables are
2.2. Computation condition presumed to be continuous and can be controlled with negligible
error in the experiments. The response (Y) is postulated to be a ran-
Research was carried out on a titanium alloy tube with an initial dom variable. For most MPRA problems, the relationship between
outside diameter (D) of 8 mm and a wall thickness of 0.8 mm. The the response and independent variables is unknown. Therefore, the
final bending angle (h) was 120°. The bending radius was 24 mm primary step in MPRA is to postulate a suitable approximation for
(i.e., 3D). The material properties of the tube determined by ten- the relationship between Y and the independent variable set. If the
sion tests were Young’s modulus E = 94.41 GPa, yield stress response is a linear function of the independent variables accord-
rs = 569.58 MPa, hardening coefficient K = 931.0 MPa, hardening ing to the visual analysis of simulation results, then first-order
exponent n = 0.113, Poisson’s ratio m = 0.39, and thickness anisot- equations can be used; otherwise, higher-order equations are
ropy exponent r = 4.0. Fig. 3 shows the specimen and tube plug applied.
used in the tests and Fig. 4 shows the curves obtained for nominal
stress vs. nominal strain. The coefficient of friction between the
tube and pressure die was 0.25 and that between the tube and
bending die was taken as 0.1, which were obtained using a com-

Table 1
Parameter values for NC bending of TA18 tube.

Parameter Value Value Value Value Value

1 2 3 4 5
Bending angle h (°) 30 60 90 120
Young’s modulus E (GPa) 84.41 94.41 104.41
Yield stress rs (MPa) 539.58 569.58 599.58
Hardening coefficient K 831.0 931.0 1031.0
(MPa)
Hardening exponent n 0.063 0.113 0.163
Poisson’s ratio m 0.31 0.39 0.47
Thickness anisotropy 1.0 2.0 3.0 4.0 5.0
exponent r
Fig. 5. 3D FE model for NC bending of a titanium alloy tube.
2004 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010

In this study, the bending angle and the material properties response for Dh of the bent tubes. Once the simulation results
were chosen as independent process parameters, and they are de- are obtained, the coefficients of these functions can be calculated.
noted in Eq. (1) as xi. The coefficients b0, bi, bii and bij were deter- Significance tests for the regression equations and coefficients
mined using the least squares method, and Y is the estimated were performed using MPRA software that we developed.

Fig. 6. Flowchart of the stepwise regression.

 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010 2005

X
k X
k X
k X
k
3. Results and discussion
Y ¼ b0 þ b i xi þ bii x2i þ bijði<jÞ xi xj þ    þ e ð1Þ
i¼1 i¼1 i j
3.1. Effects of material properties and the bending angle

2.5. SRA
3.1.1. Effects of Young’s modulus and the bending angle
Fig. 7 shows variations in Dh with E and h. From Fig. 7a, it is evi-
SRA was used to identify the best expression that includes only
dent that the increase in Dh with h decreases with increasing E,
parameters with a significant effect on Dh for the NC bending of a
indicating that the interaction E  h might have an effect on Dh.
TA18 tube and to understand the coupling effects of the parame-
Springback after tube bending mainly depends on the stress and
ters. Parameters are included in each step according to their signif-
Young’s modulus of the tube material. For bending of tubes that
icance rank. The significance of the regression is then assessed and
only differ in E, there should be no obvious difference in the value
if some parameters have become non-significant on the inclusion
and state of stresses before unloading. Thus, according to the elas-
of a new parameter, they are excluded from the regression equa-
tic unloading springback principle (as shown in Fig. 8), greater
tion one by one. The process is repeated for all parameters until
springback occurs for a tube with lower E [8].
all those that remain in the equation are significant. A flowchart
Fig. 7b shows that the decrease in Dh with increasing E is re-
of the method is shown in Fig. 6.
lated to h and the trend is similar to a quadratic relationship. MRPA
The advantage of SRA is its shorter computation time and com-
revealed that the relationship among Dh, h, E, the interaction E  h,
putation trials when compared with MPRA if a large number of
and the quadratic term E2 can be expressed as Eq. (2) with a corre-
regression variables are considered. The SRA regression expression
lation coefficient of 0.999577. The partial correlation coefficients
is similar to the MPRA expression given as Eq. (1). In SRA, we solve
for the terms are 0.999994, 0.999983, 0.999940, and 0.99997. All
the normal equation system using elimination without back sub-
the coefficients are close to 1, which means the regression equation
stitution and with a suitable choice of significant factors. The elim-
and regression coefficients are significant. Therefore, Eq. (2) can be
ination sequence is determined by the significance of the
used to predict and control Dh for the NC bending of a TA18 tube
parameters, which in turn reflects their contribution in reducing
for a given 3D bending radius and other material properties, except
the residual sum of squares. The critical values Fin and Fout for
for E.
including or excluding a parameter are determined by looking up
F-test tables for a given confidence level and degree of freedom. Dh ¼ 11:0087  0:160791h þ 0:10899E
 0:000613883Eh þ 0:000739375E2 ð2Þ

(a) 10.0 3.1.2. Effects of yield stress and the bending angle
9.5
Fig. 9 shows variations in Dh with rs and h. From Fig. 9a, it is
9.0 E =104.41GPa evident that the increase in Dh with h increases slightly with rs,
8.5 E =94.41GPa which means that the interaction rs  h might have an effect on
o

8.0 E =84.41GPa
Springback angle Δθ /

Dh. With increasing rs, the value of the stress at the same strain in-
7.5
creases, and there is greater elastic deformation for the same strain
7.0
[8]. Thus, according to the elastic unloading springback principle
6.5
(as shown in Fig. 10), there is greater springback during unloading
6.0
of a tube with greater rs.
5.5
Fig. 9b shows that the increase in Dh with rs is nearly linear.
5.0
MPRA revealed a relationship among h, rs and the interaction
4.5
rs  h according to Eq. (3) with a correlation coefficient of
4.0
0.999564. The partial correlation coefficients for the terms are
3.5
0.999598, 0.999316, and 0.999837. The correlation coefficients
3.0
20 40 60 80 100 120 are close to 1, which means the regression equation is significant,
o
Bending angle θ /

(b) 9.5
9.0
8.5
o
Springback angle Δθ /

8.0
7.5
7.0
6.5
6.0
5.5
5.0
4.5
4.0 θ = 30
o
3.5 o
3.0 θ = 60
o
2.5 θ = 90
2.0 o
1.5 θ = 120
85 90 95 100 105
Young's modulus E / GPa
Fig. 7. Relationship between the springback angle and (a) the bending angle and (b) Fig. 8. Effect of Young’s modulus on springback according to the elastic unloading
Young’s modulus. principle.
2006 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010

(a) 9.5 (a) 10.0

9.0 9.5
9.0 K=1031MPa
8.5
σs= 599.58MPa 8.5 K= 931MPa
8.0 K= 831MPa

o
o

σs= 569.58MPa 8.0

Springback angle Δθ /

Springback angle Δθ /
7.5
σs= 539.58MPa 7.5
7.0
7.0
6.5 6.5
6.0 6.0
5.5 5.5
5.0 5.0
4.5 4.5
4.0
4.0
3.5
3.5
3.0
3.0 20 40 60 80 100 120
20 40 60 80 100 120 o
o Bending angle θ /
Bending angle θ /
o
θ = 30
(b) 9.0 (b) 9.5 o
θ = 60
8.5 9.0 o
θ = 90
8.0 8.5 o
θ = 120
o
Springback angle Δθ /
8.0
o

7.5
Springback angle Δθ /

7.5
7.0
7.0
6.5
6.5
6.0
o 6.0
θ = 30
5.5 o 5.5
θ = 60
5.0 o 5.0
θ = 90
4.5 o 4.5
θ = 120
4.0 4.0
3.5
3.5
800 850 900 950 1000 1050
530 540 550 560 570 580 590 600
Strain hardening coefficient K/MPa
Yield stress σs / MPa
Fig. 11. Relationship between the springback angle and (a) bending angle and (b)
Fig. 9. Relationship between the springback angle and (a) bending angle and (b) strain hardening coefficient.
yield stress.

Dh ¼ 0:664238 þ 0:00312675h þ 0:0166415rs þ 0:0000599006rs h

ð3Þ
3.1.3. Effects of the strain hardening coefficient and bending angle
Fig. 11 shows variations in Dh with K and h. From Fig. 11a, it is
evident that the increase in Dh with h increases with K, indicating

Fig. 10. Effects of yield stress on springback according to the elastic unloading
principle.

and the regression coefficient for each term is significant. There- Fig. 12. Effects of the strain hardening coefficient on springback according to the
fore, Eq. (3) can be used to predict and control Dh. elastic unloading principle.
 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010 2007

that the interaction K  h might have an effect on Dh. The cause of

the increase in K leading to an increase in Dh is similar to the case
for yield stress [8], as shown in Fig. 12.
Fig. 11b shows that the increase in Dh with K is related to h, and
the trend is linear. MPRA revealed that the relationship among h, K
and the interaction K  h can be expressed as Eq. (4) with a corre-
lation coefficient of 0.999738. The partial correlation coefficients
for each term are 0.999503, 0.999248, and 0.999969. The correla-
tion coefficients are close to 1, which means the regression equa-
tion is significant, and the regression coefficient for each term is
significant. Therefore, Eq. (4) can be used to predict and control Dh.

Dh ¼ 1:11793 þ 0:00140495h  0:0129907K þ 0:0000689042Kh

ð4Þ
3.1.4. Effects of the strain hardening exponent and bending angle
Fig. 13 shows variations in Dh with n and h. From Fig. 13a, it is
evident that the increase in Dh with h decreases with increasing n,
indicating that the interaction n  h has a slight influence on Dh.
The variation in Dh with n during the NC bending of a TA18 M tube
is different from that Murata et al. obtained for the press bending Fig. 14. Relationship between the hardening exponent and springback according to
the elastic unloading principle.
of tubes in Ref. [9]. It is found that the strain hardening exponent n
has little effect on the springback of the press bending of tubes
with similar properties as aluminium alloy except for varied n in the hardening exponent, the lower is the stress needed to achieve
Ref. [9]. This is may be result from different bending processes of the same deformation during the NC bending of tubes. Thus,
tubes and material properties. It is well known that the larger according to the elastic unloading springback principle (as shown
in Fig. 14), there is greater springback for a tube with lower n.
Fig. 13b shows that the decrease in Dh with increasing n is lin-
(a) 10.0 ear. MRPA revealed the relationship among h, n, and the interaction
9.5
n  h can be expressed as Eq. (5) with a correlation coefficient of
9.0 n=0.163 0.9997. The partial correlation coefficients for each term are
8.5 n=0.113
Springback angle Δθ / o

0.998043, 0.999961, and 0.998378. The correlation coefficients

8.0 n= 0.063
are close to 1, which means the regression equation is significant,
7.5
7.0 and the regression coefficient for each term is significant. There-
6.5 fore, Eq. (5) can be used to predict and control Dh.
6.0 Dh ¼ 3:0909  5:77295h þ 0:0596148n  0:07721nh ð5Þ
5.5
5.0 3.1.5. Effects of Poisson’s ratio and the bending angle
4.5 Fig. 15 shows the variation in Dh with h for different values of m.
4.0 The results demonstrate that m has hardly any effect on Dh.
3.5
3.0 3.1.6. Effects of the thickness anisotropy exponent and bending angle
20 40 60 80 100 120
o
Fig. 16 shows variations in Dh with r and h. Fig. 16a shows that
Bending angle θ / the increase in Dh with h decreases with increasing r. The value of r
reflects the ratio of the hoop strain to thickness strain. Under the
o
θ = 30
o
(b) 9.5 θ = 60
o 9.0
9.0 θ = 90
o 8.5
8.5 θ = 120
Springback angle Δθ / o

8.0 8.0 ν =0.47

Springback angle Δθ /

7.5 7.5 ν =0.39

7.0 7.0 ν =0.31
6.5 6.5
6.0
6.0
5.5
5.0 5.5

4.5 5.0
4.0 4.5
3.5 4.0
0.06 0.08 0.10 0.12 0.14 0.16 3.5
Strain hardening exponent n 20 40 60 80 100 120
o
Bending angle θ /
Fig. 13. Relationship between the springback angle and (a) bending angle and (b)
hardening exponent. Fig. 15. Relation of springback angle with Poisson ratio and bending angle.
2008 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010

However, the effects of the material properties and h on Dh for

titanium alloy tube bending are very complicated. There are inter-
actions between h and each material property except for m. More-
over, quadratic terms for some parameters affect Dh to some
degree. Thus, we identified relations reflecting the coupling effects
among these parameters to predict and control Dh during the
bending of a titanium alloy tube.
Considering the manufacturing effects for a tube billet, the
material properties of tubes are not perfectly consistent, even
within the same batch, and thus it is necessary to obtain a more
universal relation to reflect the coupling effects of parameters.
To avoid a large computation effort for multiple variables and
their interactions, SRA was employed to identify the best regres-
sion equation including only parameters with a significant influ-
ence on Dh:

Dh ¼ b0 þ b1 h þ bj xj þ b1j hxj þ b22 x22 þ b66 x26 ðj ¼ 2;    ; 6Þ ð7Þ

where x2 is E, x3 is rs, x4 is K, x5 is n, and x6 is r.

The coefficients in Eq. (7) are listed in Table 2. A coefficient with
a value of zero indicates that the corresponding term has no signif-
icant effect on the target variable and does not need to be included
in the regression equation. Thus, according to Table 2, only E, rs, K,
n and r, as well as the interactions of these parameters with h, have
a significant influence on Dh, and the final best regression equation
should be as Eq. (8).
Dh ¼ b0 þ b1 h þ bj xj þ b1j hxj ðj ¼ 2;    ; 6Þ ð8Þ
As observed from Table 2, Dh decreases with E, n, and the inter-
actions E  h, n  h and r  h on one hand, and increases with the
other terms on the other hand. The maximum error for regression
and measured values is not more than 3.5%, indicating that the
regression equation obtained is of high precision and can thus be
used to predict Dh for a titanium alloy tube with an outside diam-
eter of 8 mm and thickness of 0.8 mm according to the require-
ments of the bending angle and variations in material properties.
For a desirable bending angle of h0 ,
Fig. 16. Relationship between the springback angle and (a) bending angle and (b)
thickness anisotropy exponent. Dh ¼ h  h 0 : ð9Þ
By relating Eqs. (8) and (9), expressions for the practical bend-
same deformation conditions, lower thickness deformation occurs ing angle with springback compensation under different precondi-
for tubes with greater r, and thus smaller springback occurs during tions can be obtained (as shown in Eq. (10)) and used to control the
unloading owing to a larger constraint function. bending angle of a titanium alloy tube of a given gauge.
Fig. 16b shows that the decrease in Dh with increasing r is re- h 0 þ b 0 þ b j xj
lated to h and the decreasing trend is quadratic. MPRA revealed h¼ ðj ¼ 2;    ; 6Þ ð10Þ
1  b1  b1j xj
that the relationship among h, r, the interaction r  h, and the qua-
dratic term r2 can be expressed as Eq. (6) with a correlation coeffi-
cient of 0.998495. The partial correlation coefficients for each term 3.3. Significance analysis
are 0.988879, 0.999831, 0.995928, and 0.992452. All the correla-
tion coefficients are close to 1, which means the regression equa- To assess the significance of the influence of the material
tion is significant, and the regression coefficient for each term is parameters on Dh for NC bending of TA18 alloy tube, the rate of
significant. Therefore, Eq. (6) can be used to predict and control Dh. change in Dh as a function of standardized parameters [Eq. (11),
after standardization, x0j 2 ½0; 1 and xj ¼ xjmin þ x0j  Dxj ] is proposed
Dh ¼ 2:62414  0:257297h þ 0:0553478r  0:00520335rh as Eq. (12).
þ 0:118359r 2 ð6Þ
Table 2
3.2. Analysis of coupling effects Stepwise regression coefficients for the springback angle of titanium alloy tube
bending.
From the above analysis, we know that the relationships be- Parameter Value Parameter Value
tween the springback angle of a TA18 M titanium alloy tube and
b0 1.87613 b12 0.000612383
Young’s modulus, yield stress, the hardening coefficient, and the b1 0.02391219 b13 0.0000606923
hardening exponent are similar to those for stainless steel b2 0.0212939 b14 0.0000690218
[7,8,15,19] and aluminium alloy tubes [7,18] for NC bending, but b3 0.00306682 b15 0.0771639
the springback angle of the titanium alloy tube under the same b4 0.00139599 b16 0.00552268
b5 5.77669 b22, b66 0.0
bending conditions is much greater than the springback angles of
b6 0.205171 Max error (%) 3.493
the latter tubes.
 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010 2009

2.0

1.8 E
σs

Variation of springback angle Δθ / o

1.6
K
1.4 n
r
1.2

1.0

0.8

0.6

0.4

0.2

0.0
30 60 90 120
o
Bending angle θ /

Fig. 17. Significance of the influence of material properties on springback during NC bending.

xj  xjmin xj  xjmin the springback angle. However, when taking the effects of
x0j ¼ ¼ ð11Þ
xjmax  xjmin Dxj all material properties into consideration, the quadratic
@ðDhÞ terms are not distinguishable.
¼ ðbj þ b1j hÞ  Dxj ð12Þ
@x0j (4) The influence of the material properties of titanium alloy
tubes increases with the bending angle. Regardless of the
The significance of the effect of different material parameters on bending angle, Young’s modulus and the strain hardening
Dh for NC bending of TA18 M tubes is shown in Fig. 17. For a bend- coefficient and exponent are the three parameters that have
ing angle of 30°, the significance of the effect decreases in the order the greatest influence on the springback angle.
n > E > K > rs > r. For a bending angle of 60°, the significance de-
creases in the order E > K > n > rs > r. For a bending angle of 90°,
Acknowledgements
the significance decreases in the order E > K > n > r > rs. For a bend-
ing angle of 120°, the significance decreases in the order
This research was supported by the Program for New Century
K > E > n > r > r s.
Excellent Talents in University (NCET-08-0462) and the Fund of
the State Key Laboratory of Solidification Processing in NWPU
4. Conclusions (No. KP200919), for which the authors wish to express their
gratitude.
Using a 3D elastic–plastic FE model and multivariate and step-
wise analyses, the coupling effects of the bending angle and the
material properties on the springback angle during the NC bending References
of a titanium alloy tube were revealed. The conclusions are as
[1] Hashmi MSJ. Aspects of tube pipe manufacturing processes: meter to
follows: nanometer diameter. J Mater Process Technol 2006;179:5–10.
[2] Yang H, Zhan M, Liu LY. Some advanced plastic processing technologies and
(1) Compared with stainless steel and aluminium alloy tube NC their numerical simulation. J Mater Process Technol 2004;151:63–9.
[3] MacKenzie PM, Walker CA, Mckelvie J. A method for evaluating the mechanical
bending, the springback angle of the titanium alloy under performance of thin-walled titanium tubes. Thin-Walled Struct
the same bending conditions is much greater. The relation- 2007;45:400–6.
ships between the springback of a titanium alloy tube and [4] Wang G, Zhang KF, Wu DZ, Wang JZ, Yu YD. Superplastic forming of bellows
expansion joints made of titanium alloy. J Mater Process Technol
Young’s modulus, yield stress, and the strain hardening coef-
2006;178:24–8.
ficient and exponent are similar to those for stainless steel [5] Lou HZ, Stelson KA. Three-dimensional tube geometry control for rotary draw
and aluminium alloy tubes in NC bending. The larger the tube bending, part 2: statistical tube tolerance analysis and adaptive bend
correction. J Manuf Sci Eng 2001;123:266–71.
thickness anisotropy exponent, the smaller the springback
[6] Lou HZ, Stelson KA. Three-dimensional tube geometry control for rotary draw
angle is. The Poisson ratio has hardly any influence on tube bending, part 1: bend angle and overall tube geometry control. J Manuf
springback angle. Sci Eng 2001;123:258–65.
(2) Young’s modulus, yield stress, the strain hardening coeffi- [7] E DX, He HX, Liu XY, Ning RX. Experimental study and finite element analysis
of spring-back deformation in tube bending. Int J Miner Metall Mater
cient and exponent, and the thickness anisotropy exponent, 2009;16:177–83.
as well as coupling effects of these parameters with the [8] E DX, Liu YF. Springback and time-dependent springback of 1Cr18Ni9Ti
bending angle, have a significant influence on the springback stainless steel tubes under bending. Mater Des 2010;31:1256–61.
[9] Murata M, Kuboki T, Takahashi K, Goodarzi M, Jin Y. Effect of hardening
angle. exponent on tube bending. J Mater Process Technol 2008;201:189–92.
(3) The quadratic terms of Young’s modulus and thickness [10] Tang NC. Plastic-deformation analysis in tube bending. Int J Press Vessels
anisotropy exponent affect the springback angle to some Piping 2000;77:751–9.
[11] AL-Qureshi HA. Elastic–plastic analysis of tube bending. Int J Mach Tool Manuf
degree when only considering their individual effects on 1999;39:87–104.
2010 Z.Q. Jiang et al. / Materials and Design 31 (2010) 2001–2010

[12] Al-Qureshi HA, Russo A. Spring-back and residual stresses in bending of thin- [17] Hur S, Park JS. The 360° cold bending of Ti–6Al–4V large-diameter seamless
walled aluminum tubes. Mater Des 2002;23:217–22. tube. J Miner 1999;51(6):28–30.
[13] El Megharbel A, El Nasser GA, Domiaty Aly El. Bending of tube and section [18] Gu RJ, Yang H, Zhan M, Li H. Springback of thin-walled tube NC precision
made of strain-hardening materials. J Mater Process Technol 2008;203:372–80. bending and its numerical simulation. Trans Nonferr Met Soc China
[14] Zhan M, Yang H, Huang L. A numerical–analytic method for quickly predicting 2006;16:s631–8.
springback of numerical control bending of thin-walled tube. J Mater Sci [19] Gu RJ, Yang H, Zhan M, Li H, Li HW. Research on the springback of thin-walled
Technol 2006;l22(5):713–20. tube NC bending based on the numerical simulation of the whole process.
[15] Zhan M, Yang H, Huang L, Gu RJ. Springback analysis of numerical control Comput Mater Sci 2008;42(4):537–49.
bending of thin-walled tube using numerical–analytic method. J Mater Process [20] Yang H, Zhan M. Experimental modeling methods used for material forming
Technol 2006;177:197–201. processes. Xi’an, Northwestern Polytechnical University press; 2008 [in
[16] Jiang ZQ, Yang H, Zhan M, Yue YB, Liu J, Xu XD, et al. Establishment of a 3D FE Chinese].
model for the bending of a titanium alloy tube. Int J Mech Sci 2009.
doi:10.1016/j.ijmecsci.2009.09.029.