Thin-Walled Structures 107 (2016) 4956

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

The deformation mechanism analysis of a circular tube under free

inversion
XinMing Qiu a,n, XiaoHuan Yu a, YiLiang Li a, T.X. Yu b,c
a

Department of Engineering Mechanics, Tsinghua University, Beijing 100084, China

Department of Mechanical and Aerospace Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong
c
Y.K. Pao Chair Professor, Ningbo University, Ningbo, China
b

art ic l e i nf o

a b s t r a c t

Article history:
Received 8 December 2015
Received in revised form
10 May 2016
Accepted 23 May 2016

As a popularly adopted design of thin-walled energy absorber, a circular tube under free inversion has
been studied theoretically since 1960s However, in all the existing models, the prole of inversion is
assumed to be a semi-circle, i.e., the curvature in the region of inversion is a constant. In this study, from
equilibrium of a representative element, the curvature is found to be varying, due to bending moment
distribution. The conclusion of varying curvature is veried by the corresponding nite element simulations.
Then the deformation mechanism of a tube under free inversion is theoretically investigated by establishing a curvature varying model. Considering the interaction of bending and tension, the steady
inversion load and knuckle radius predicted by current model agree well with those obtained from FE
simulations as well as the experimental data published in literature.
& 2016 Elsevier Ltd. All rights reserved.

Keywords:
Tube inversion
Energy absorption
Curvature change
Inversion force
Knuckle radius

1. Introduction
Thin-walled metal tube is a kind of commonly used energy
absorber. For a circular tube under axial compression, the energy
could be dissipated by progress buckling [1,2], global buckling [3],
inversion [4], expansion or even splitting [5,6]. Among the above
mentioned different collapse modes, inversion possesses a long
steady stroke, during which the inversion force remains almost
constant, so it is regarded as an ideal energy absorber.
There are two types of tube inversions: free inversion, and inversion with a die. Besides, a circular tube can be inverted externally or internally. In the present paper, only the externally free
inverting is studied. As shown in Fig. 1, a metal tube is xed at one
end and compressed at the other end, then it is inverted freely and
formed into a co-axial cylinder outside the original one.
The theoretical studies of free inversion are aimed at obtaining
a steady inversion load P and a natural knuckle radius b [4,711],
which depend on tube geometries and material properties, but not
the initially induced boundary condition. Most theoretical analysis
are based on two-dimensional models. Guist and Marble [7] proposed the original two-dimensional model and predicted the
steady knuckle radius and inversion force, where rigid, perfectly
plastic material idealization (R-PP) was employed. In their model,
n

Corresponding author.
E-mail address: qxm@tsinghua.edu.cn (X. Qiu).

http://dx.doi.org/10.1016/j.tws.2016.05.019
0263-8231/& 2016 Elsevier Ltd. All rights reserved.

curvature, tube length and thickness in the inversion region were

all assumed to be constant during the inverting process; which is
in conict with the material volume conservation. Compared to
experimental data, their prediction on knuckle radius is almost
doubled, and inversion force is about 2030% lower. After that,
researchers keep on trying to improve this model. For example,
Kinkead [9] considered the additional circumferential bending
energy dissipation, which improved the prediction of steady inversion load P; that is, deviation was reduced to the range of
10%
to 10%. But the prediction on knuckle radius b was not improved.
Calladine [12] modied the positon of plastic hinges; and Reddy
[8] changed the material property to rigid, linearly strain-hardening material; Qiu and He [11] modied velocity at the exit point.
All above attempts have improved the steady inversion force or
knuckle radius to some extent. However, there are still some deviations between theoretical predictions and nite element
simulations.
A two-stage assumption was proposed by Colokoglu and Reddy
[10], in which the inversion region was divided into two parts: in
rst part, tube wall thickness is assumed to be constant; while in
second part, material velocity is assumed to be constant. The
knuckle radius given by this model [10] is almost the same as that
of Reddy's model [8], only 2% higher. Inherited this two-stage assumption [10], Qiu et al. [4] proposed a three-dimensional model
considering the strain variation along wall thickness direction, in
which volume conservation and Von Mises yield criteria were
adopted. This three-dimensional model [4] did not improve the

50

X. Qiu et al. / Thin-Walled Structures 107 (2016) 4956

Fig. 1. The external free inversion of a circular tube under axial compression [4].

predictions accuracy on knuckle radius, but did improve the prediction of steady inversion load.
By comparing the initial state and nal state, a two-dimensional deformation theory model was then proposed by Yu [13]. In
this study, three different deformation mechanisms were analyzed, and among them the results based on a two-stage assumption about tube thickness [10] agree well with those obtained
from FE simulations as well as the experimental data.
In summary, the existing theoretical models of free inversion
tube are all based on the assumption of inversion region is a semicircle [4,713]. Therefore, an interesting question arises; that is, in
the deformation process of a free inverting tube, is the inversion region really a semi-circle with constant curvature?
In latest research on tube inversion over a die, the curvature
change in inversion region was observed. For example, behavior of
the free deformation under die-less constraint was studied by He
et al. [14] by employing FE method. It was found that before
reaching the steady-state, the curvature at each point in the free
deformation zone were changing continuously with time. Employing an energy method, the curling behavior of inverted tubes
was analyzed by Leu [15], which led to the determination of inversion force and curling radius. In the case of a quarter circle die
for tube inversion, although tube is intimate contact with die, it
will curl up when becomes free with a curvature different from
that of the llet die radius. Compared to the constant curvature
models [16,17], energy dissipation due to curvature change was
considered in Ref. [15].
Inspired by studies of tube inversion over a die [14,15], the
deformation mechanism of inversion region of a free inversion
tube will be revisited in this study. Firstly, the variation of thickness and curvature in reversion region will be checked. Then based
on previous work, a two-dimensional theoretical model is proposed in order to analyze the free inverting metal tube. In Section
2, the two-dimensional deformation theory model proposed by Yu
[13] will be briey reviewed. In Section 3, the distribution of
bending moment in inversion region will be analyzed by the

equilibrium of a representative element, and the deformation

prole will be studied by nite element analysis. In Section 4, a
theoretical model considering curvature change and interaction of
bending-tension will be established. And nally, the prediction of
current model will be discussed in Section 5.

2. Review of the theoretical model based on deformation

theory
In this section, Yu's two-dimensional deformation theory [13]
will be briey reviewed. The axial prole of an external free inversion tube is shown in Fig. 2. Fully clamped at the end near point
B, a circular tube with original thickness t0 and radius R0 is
compressed along axial direction under applied load P; then is
inverted from inside to outside. The inversion region AB is assumed to be a semi-circle with radius b. The tube radius is enlarged from R0 at A to R0 + 2b at B after inversion. Thereafter, a
toroidal small segment dl is changed to dl, after the deformation
consists of bending at point A (in axial direction), expansion and
bending across region AB (in circumferential direction); and
bending at point B (in axial direction). Without considering the
deformation history of stress/strain of each material point, by
comparing nal state with initial state, that is dl and dl , a simple
deformation theory is established based on the energy
conservation.
To calculate the work rate of applied load, suppose the velocity
of small segment dl is v0 , and velocity of dl is v. The external work
rate done by the applied compression load is,

W = P (v0 + v)

(1)

Then to calculate the energy dissipation rate along the circumferential direction E and along the axial direction El , due to
energy conservation,

X. Qiu et al. / Thin-Walled Structures 107 (2016) 4956

51

Compared with corresponding FE and experimental results, the

two-stage assumption was found to be more accurate, with the
simplied equation given as follows [13],

3b
1
1
P =
+
+
2R 0
4R 0
4b

(5)

where P P /( 2YR0 t0 ), R 0 R0/t0 , b b /t0 .

In summary, for tubes with t0/R0 0.04 , the knuckle radius
given by Yu's model [13] agrees well with experimental values, the
maximum difference is about 15% compared with that of the FE
simulations [4]. For steady inversion load, the theoretical prediction [13] is about 815% higher than that of FE. It should be pointed
out that for tubes with thin wall thickness, t0/R0 < 0.04 , all the
existing theoretical predictions have large deviations in predicting
knuckle radius and inversion load. In this region, the constant
curvature assumption is estimated to be different with the actual
deformation.
Fig. 2. The axial prole of a free inversion tube in the deformation theory [13].

W = E + El

(2)

The steady inversion load is obtained as a function of knuckle

radius b,

P = f ( b)

(3)

The real knuckle radius b should render the lowest applied

force P,

dP
=0
db

(4)

By substituting knuckle radius of Eq. (4) into Eq. (3), the steady
inversion load is given as a function of tube geometries.
Based on volume conversation during plastic deformation,
three different deformation mechanisms were analyzed in Ref.
[13]:
a) Constant velocity; i.e. material velocity along axial direction is
constant;
b) Constant thickness; i.e. tube wall thickness tube remains
constant;
c) Two-stage assumption; that is, in segment AC, tube thickness
remains constant; while in segment CB, material velocity is
constant; which inherits the two-stage assumption proposed
by Colokoglu and Reddy [10].

3. Deformation mechanism analysis in inversion process

In this section, the deformation mechanism of inversion region
will be investigated, rstly by equilibrium analysis of a representative element; and then by checking the deformation
prole along midline of tube.
3.1. Distribution of bending moment
The top view of a circular tube under free external inversion is
illustrated in Fig. 3(a). In view of the axial symmetry of structure, a
representative element AB with respect to angle increment d is
analyzed, as shown in Fig. 3(b). During expanding deformation
along circumferential direction, distributed tensional force T (y ) is
applied on both sides of element AB. It should be noted that T (y ) is
varying with material position (y), as illustrated in Fig. 4, which is
the rigid body diagram of representative element AB of inversion
region.
In Fig. 4, supposing point C is located at the bottom of inversion
region. Representative element AB with respect to angle increment d , is compressed at point A along y direction, and is
stretched at point B with applied force,

PA = P B =

P
d
2

(6)

Obviously, at any point (x, y ) inside segment AC, the resultant

Fig. 3. Top view of a circular tube under free external inversion.

52

X. Qiu et al. / Thin-Walled Structures 107 (2016) 4956

increases with y due to tube expansion. Therefore, MK increases

monotonically from point A to point C.
If cross-section K locates in segment CB, as shown in Fig. 5(b),
from the equilibrium of segment AK,
MK = PA (x K x A ) +

Fig. 4. Rigid body diagram of the representative element AB.

force T (y ) of the distributed force T (y ) on both sides, is along x

direction,

T (y) = 2T (y) sin

d
T (y)d, 0 y yC
2

(7)

Similarly, inside segment CB, the resultant force T (y ) is also

along x direction,

T (y) T (y)d,

0 y yC

(8)

It should be pointed out that, T (y ) increases with y while T (y )

decreases with y. Besides, T (y ) T (y ) for same value of y.
Suppose K is a material particle representing an arbitrary crosssection in segment AC, with coordinate (xK , yK ). In steady stage of
inversion, the internal force components at point K is shown in
Fig. 5(a), where bending moment MK is obtained from the equilibrium of segment AK,

MK = PA (xK xA ) +

yK

T (y)(yK y)dy

(9)

If material particle K moves from point A to point C, the values

of xK and yK all increase. Besides, resultant force T (y ) also

yK

T (y )(yK y )dy

y C T (y) + T (y) (y yK )dy

K

(10)

If point K moves from C toward B, xK will increase and yK will

decrease. The rst and second term of Eq. (10) are positive, where
rst term is increasing and second term is decreasing. The third
term is negative and its absolute value is increasing. Therefore, the
bending moment distribution in segment CB depends on the deformation prole y ( x ) and also the material property that may
affect T (y ) and T (y ).
The above analysis is based on equilibrium analysis, which is
independent of the traditionally adopted deformation mechanism
assumption [4,713]. Then relationship between bending moment
and curvature will be considered. Under pure bending, bending
moment M varying with curvature is illustrated in Fig. 6. For
elastic-plastic material, bending moment increases monotonically
with curvature during loading process; while for idealized rigid,
perfectly plastic material (R-PP), bending moment remains as a
constant, that is the fully plastic bending moment, M = MP , for any
curvature. Usually metal materials will have an elastic stage as
well as strain hardening effect. Therefore, across inversion region
ACB, curvature should be varying in relation to varying bending
moment, as given in Eq. (9) and Eq. (10).
The premise condition of above conclusion should be noted;
that is, y monotonically increases with x in region AC; while y
decreases with x monotonically in region CB. Clearly, a semi-circle
satises this condition. It is found that at point A, the bending
moment is zero. A contradiction occurs in the assumption of inversion region being a semi-circle [4,713]: discontinuity in curvature appear at point A where bending moment is zero.
If inversion region is not a semi-circle, then what will be the
deformed prole? In the following, we will try to nd it from FE
simulations.
3.2. Deformation prole of inversion region
An axis-symmetric FE model is built for the inversion process
employing software ABAQUS, similar to Ref. [4]. Tube is assumed
to be made of elastic, perfectly plastic material (E-PP) with density
2800 kg/m3, Young's modulus 210 GPa, Poisson ratio 0.3, and yield
strength 288 MPa. Tube length is 100 mm and initial radius is

T ( )

Fig. 5. Internal force components at point K inside inversion region of (a) AC or (b) CB.

X. Qiu et al. / Thin-Walled Structures 107 (2016) 4956

M
MP

Elastic plastic

RPP

unloading

Fig. 6. Bending moment M varying with curvature for different material.

25 mm. One end (B) of tube is fully constrained, while the other
end (A) is compressed by a rigid surface without friction, at very
low velocity. Along tube thickness direction, totally 4 or 8 square
elements are divided.
In Fig. 7, contours of strain distribution of the inverting tubes
are given for t0/R0 = 0.04 and 0.06. From Fig. 7, the inversion region can be distinguished being slightly different to a semi-circle
which it looks like. Taking tube of t0/R0 = 0.04 as an example, the
deformed prole of tube midline line (along thickness direction) is
traced and plotted in Fig. 8(a) as the black solid line. Here O is
mid-point of line AB, where the length of OA , OC , and OB are
equal, i.e., OA = OB = OC = b. For comparison, the semi-circle
with radius b through points A, B, C is plotted in dot dash line. In
view of the difference between two lines, inversion region does
not completely coincidence with a semi-circle.
For an arbitrary point K within inversion region, the angle between OA and OK is denoted as . Clearly, = 0 at point A, and
= 180 at point B. From Fig. 2, before entering the inversion region ACB, two points at the same horizontal line, but on inner and
outer surface respectively, is regarded as a point-pair, e.g., pointpair A+ and A . Clearly, the initial distance between point-pair is t0 .
During inverting process ACB, the distances between point-pairs
are recorded as tube thickness t that vary with position . Within
region ACB, the tube wall thickness changes, and the curvature of
midline varies with angle as plotted in Fig. 8(b) and (c), respectively. It is seen from Fig. 8(b) that the tube wall thickness
rstly increases and then decreases with angle ; while the
maximum changing ratio is about 6%. Therefore, the thickness
change during inversion could be neglected.
It is seen from Fig. 8(c) that within 0 < < 60 , curvature of
midline increases rapidly; within 60 < < 175 , curvature almost
remains constant 1/b; while dramatic changes occur in a small

53

range of 175 < < 185 , where curvature drops from a positive
peak to a negative peak. Hereby, the curvature in inversion region
ABC is not a constant as that of a semi-circle. Within 60 < < 175
that is about to be segment CB, the prole of inversion region
agrees with a semi-circle better, as conrmed by Fig. 8(a). Furthermore, a transition region AA with non-zero curvature is found
to exist besides the main inversion region ACB, corresponding to
angle range 20 < < 0 .
Hereby, combined with the equilibrium analysis illustrated in
Section 3.1, inversion region could be roughly divided into three
segments:
1. Segment AC , where bending moment and curvature both increase, corresponding to the loading process shown in Fig. 6. In
early stage of deformation, the elasticity of material is estimated
to have large effect over transition region AA .
2. Segment CB, where bending moment varies and curvature decreases slightly, corresponding to unloading process as showed
in Fig. 6. During loading process, plastic deformation is larger
than elastic deformation, therefore curvature change during
unloading process is relatively small so as to be negligible.
3. Near point B, where bending moment and curvature both vary
rapidly. Although the variations are hardly visible in Figs. 7 and 8
(a), a deformation localization zone with S shape exists, as
suggested by curvature curve shown in Fig. 8(c). Considering the
equilibrium analysis given in Section 3.1, due to the existence of
this S shape prole near point B, a non-zero bending moment at
this point is produced, which re-bend the tube to be straight again.
In summary, from observation of FE results, following remarks
could be made: (1) the thickness change in reversion region could
be neglected; (2) in segment AC , curvature increases continuously
from zero to a constant value, i.e., there is no curvature jump at
point A; (3) in segment CB, curvature could be regarded as a
constant, i.e., segment CB is close to a 1/4 circle; (4) at point B,
discontinuities occur in both curvature and bending moment.

4. Modied theoretical model

Based on the assumptions drawn from FE analysis, the two dimensional theoretical model of Yu [13] will be modied in this study.
4.1. Basic assumptions
As shown in Fig. 2, a tube with initial thickness t0 and radius R0
is compressed under axial force P, and the deformation region of

Fig. 7. Contours of strain distribution of the inverting tube.

54

X. Qiu et al. / Thin-Walled Structures 107 (2016) 4956

4. Tube thickness remains unchanged during the inversion process, i.e. t t0 .

4.2. Deformation analysis
According to volume conservation and constant thickness assumption, material velocity at point B is

v =

R0
v0
R 0 + 2b

(12)

External work rate done by the applied compression load P is,

W = P (v0 + v)

(13)

4.2.1. Yield surface

If a beam or plate is loaded by both bending moment M and
tension force N, the stress on the cross-section is superposition of
bending and tension. Considering the interaction of bending and
tension, the exact yield surface for R-PP material is,

M
N
+
MP
NP

=1

(14)

Obviously, the above relation is a nonlinear function of M and

N, which will be complicate in application. A simplied independent yield criterion is usually adopted,

M
= 1,
MP

N
=1
NP

(15)

As shown in Fig. 9, Eq. (15) is the circumscribing yield locus of

the exact yield surface Eq. (14), which will give an upper bound of
the load capacity. Similarly, a lower bound is given by taking the
factor of 0.618 to criterion of Eq. (15), that is,

M
= 0.618,
MP

N
= 0.618
NP

(16)

In current study, the upper bound of load capacity will be given

by using approximate yield square Eq. (15), while lower bound is
reached directly by multiply a factor of 0.618. And the average of
upper bound and lower bound is regarded as current theoretical
prediction.

circumscribing
yield locus
1

M MP

Fig. 8. (a) Prole of midline of the inversion region; (b) thickness change and
(c) curvature changing with angle . ( t0/R0 = 0.04 ).

0.618

tube under free external inversion, ACB, will be studied. In steady

stage of inversion, the distance between midlines of points A and B
is denoted as 2b.
In current theoretical analysis, the following assumptions are
adopted:

inscribing
yield
0 locus

1. The tube is made of rigid, perfectly plastic material;

2. The deformed region ACB is not symmetric about OC . Segment
CB is 1/4 of a circle with radius b, and segment AC is an unknown curve with increasing curvature.
3. The material is incompressible, so volume conversation leads to

2R 0 v0 t0 = 2Rvt

(11)

-1

exact
yield locus

0.618

N NP

-1

Fig. 9. Yield surfaces considering the interaction of bending and tension.

X. Qiu et al. / Thin-Walled Structures 107 (2016) 4956

v
2b
=
v0
R 0 + 2b

l = 1

4.5

3.5
3.0

1
( MP l + NP l )
t0

2.5
2.0

(17)

1.5

The density of plastic energy dissipation in axial direction per

unit volume is,

El = Elb + Els =

Current model
Yu's model, 2015
FE
Qiu's experiment
Guist's experiment

4.0

4.2.2. Energy dissipation along axial direction

In segment AC , curvature continuously is changed from 0 to
1/b. At point B, the localized curvature change is also 1/b. Hence
the total curvature change in axial direction is l = 2/b.
Axial compression/tension will contribute to energy dissipation. By comparing the initial and nal states, the stretching strain
in axial direction l is

55

1.0
0.5
0.00

(18)

0.02

0.04

where MP = Yt02/4 and NP = Yt0 are the fully plastic bending moment and fully plastic axial force per unit length, respectively.
4.2.3. Energy dissipation along circumferential direction
Tube expansion makes a dominant contribution to energy
dissipation. From the change of tube radius, strain along circumferential direction is,

(19)

From Fig. 2, a material point passing through point A will pass

through point B after inversion, as a result of inside-out inversion. Therefore, bending deformation exists in circumferential direction. Curvature of the midline (along thickness direction) is
adopted for an estimation. The curvature in direction is 1/R0 at A
and 1/(R0 + 2b) at B, so that curvature change is,

1
1
2
+

R0
R 0 + 2b
R0

E = Es + Eb =

1
( NP + MP )
t0

(21)

(23)

where P P U/( 2YR0 t0 ), R 0 R0/t0 , b b /t0 .

The real value of b should produce the lowest inversion force,
that is, dP U/db = 0. Omitting high order quantities, b is given by,

1
2 2

R 0 0.3536 R 0

(24)

Substituting Eq. (24) into Eq. (23), and omitting high order
quantities results in an expression of the upper bound of steady
inversion force as,

P U

2
3
+
8R 0
R 0

Similarly,

0.0
0.00

0.02

0.04

0.06

t0/R0

0.08

0.10

Fig. 10. (a) The normalized knuckle radius and (b) steady inversion load versus
t0/R0 .

inversion force is obtained by,

(26)

(22)

0.1

Here V = 2R0 t0 v0 denotes the volume rate, P U is upper bound

of load.
By substituting Eq. (18) and Eq. (21) into Eq. (22), the steady
inversion force is given in a normalized form,

1 R 0 + 2b 1
2b
2b
1
P U =
+
+
+

2 R 0 + b 2b
R0
2R 0
R 0 + 2b

Current model
Yu's model, 2015
FE
Qiu's experiment
Guist's experiment

1 U
1.144
0.303
P =
P + P L
+
2
R 0
R 0

4.2.4. Energy conservation

According to energy conservation,

P U ( v0 + v) = ( El + E ) V

0.2

(20)

Hereby the density of plastic dissipation in circumferential direction per unit volume is,

0.10

0.3

0.08

0.4

2b
=
R0

0.06

t0/R0

(25)

P L = 0.618P U . Therefore, the normalized steady

5. Results and discussions

The normalized knuckle radius and inversion load in steady
stage given by current model are plotted in Fig. 10, where Yu's
predictions and the previous experimental data are also shown. In
Fig. 10(a), experimental data are lower than FE; while in Fig. 10(b),
experimental data are higher than FE. These differences could be
explained by the difference in material property, as real material
adopted in experiments had slight strain hardening effect.
Compared with both FE and experimental data, it is clear that
current study largely improves the prediction accuracy of knuckle
radius. While for steady inversion load, curve of current prediction
is in the middle of FE and experimental data, which is accurate
enough for engineering application.

6. Conclusions
In this paper, theoretical analysis of the free inversion of a
circular tube is revisited. Different from all previous theoretical
analysis, in which inversion region was assumed to be a semicircle directly, the variation of bending moment and curvature in

56

X. Qiu et al. / Thin-Walled Structures 107 (2016) 4956

inversion region is investigated in current study. In rst stage of

inversion, the curvature of deformation prole is found to increase
continuously, accompanied by varying bending moment.
Then the previous theoretical prediction of free inversion tube
is improved by considering the interaction of bending and tension.
The predicted steady inversion load and knuckle radius, which are
in very simple form, are found to be accurate compared with FE
simulations and previous experimental data.

Acknowledgement
The authors gratefully acknowledge the nancial support from
National Science Foundation of China No.11372163.

References
[1] X. Luo, J.Y. Xu, J.S. Zhu, Y. Gao, L.X. Nie, W.M. Li, A new method to investigate
the energy absorption characteristics of thin-walled metal circular tube using
nite element analysis, Thin Wall Struct. 95 (2015) 2430.
[2] S. Salehghaffari, M. Tajdari, M. Panahi, F. Mokhtarnezhad, Attempts to improve
energy absorption characteristics of circular metal tubes subjected to axial
loading, Thin Wall Struct. 48 (2010) 379390.
[3] T.G. Ghazijahani, H. Jiao, D. Holloway, Plastic buckling of dented steel circular
tubes under axial compression: an experimental study, Thin Wall Struct. 92
(2015) 4854.

[4] X.M. Qiu, L.H. He, J. Gu, X. Yu, A three-dimensional model of circular tube
under quasi-static external free inversion, Int. J. Mech. Sci. 75 (2013) 8793.
[5] J. Tanaskovic, D. Milkovic, V. Lucanin, G.V. Franklin, Experimental investigations of the shrinking-splitting tube collision energy absorber, Thin Wall
Struct. 86 (2015) 142147.
[6] J.L. Yang, M. Luo, Y.L. Hua, G.X. Lu, Energy absorption of expansion tubes using
a conical-cylindrical die: experiments and numerical simulation, Int. J. Mech.
Sci. 52 (2010) 716725.
[7] L.R. Guist, D.P. Marble, Prediction of the inversion load of a circular tube, Natl.
Aeronaut. Space Adm. (1966).
[8] T. Reddy, Guist and marble revisitedon the natural knuckle radius in tube
inversion, Int. J. Mech. Sci. 34 (1992) 761768.
[9] A. Kinkead, Analysis for inversion load and energy absorption of a circular
tube, J. Strain Anal. Eng. Des. 18 (1983) 177188.
[10] A. Colokoglu, T. Reddy, Strain rate and inertial effects in free external inversion
of tubes, Int. J. Crashworthiness 1 (1996) 93106.
[11] X.M. Qiu, L.H. He, J. Gu, X.H. Yu, An improved theoretical model of a metal tube
under free external inversion, Thin-Walled Struct. 80 (2014) 3237.
[12] C.R. Calladine, Analysis of Large Plastic Deformations in Shell Structures,
Spring, Berlin, Heidelberg, 1986.
[13] X.H. Yu, X.M. Qiu, T.X. Yu, Analysis of the free external inversion of circular
tubes based on deformation theory, Int. J. Mech. Sci. 100 (2015) 262268.
[14] Y. He, Z.H. Sun, Y.J. Jin, FEM analysis of mechanism of free deformation under
dieless constraint in axial compressive forming process of tube, J. Mater.
Process. Technol. 115 (2001) 367372.
[15] D.K. Leu, The curling characteristics of static inside-out inversion of metal
tubes, Int. J. Mach. Tool Manuf. 40 (2000) 6580.
[16] P.C. Miscow, H.A. AlQureshi, Mechanics of static and dynamic inversion processes, Int. J. Mech. Sci. 39 (1997) 147161.
[17] A. Niknejad, M. Moeinifard, Theoretical and experimental studies of the external inversion process in the circular metal tubes, Mater. Des. 40 (2012)
324330.