Thin-Walled Structures 42 (2004) 1103–1137

www.elsevier.com/locate/tws

Experimental and numerical investigation of

static and dynamic axial crushing of circular
aluminum tubes
D. Al Galib
, A. Limam
URGC Structure, INSA de Lyon, Domaine Scientifique de la Doua, Bâtiment Joseph Charles Augustin
Coulomb, 34 avenue des Arts, Villeurbanne, France
Received 28 May 2003; received in revised form 13 February 2004; accepted 19 March 2004

Abstract

A comprehensive experimental and numerical study of the crash behavior of circular

aluminum tubes undergoing axial compressive loading is performed. Non-linear finite
element analyses are carried out to simulate quasi-static and dynamic test conditions. The
numerical predicted crushing force and fold formation are found to be in good agreement
with the experimental results. A summary of available analytical solutions is presented in
order to estimate the mean crushing load and establish a comparison between these analyti-
cal loads and the experimental one. Some observations are made on the influence of geo-
metrical imperfections and material strain rate effect.
# 2004 Elsevier Ltd. All rights reserved.

Keywords: Thin-walled circular tube; Aluminum extrusions; Axial crushing; Dynamic; Static; Numerical
simulation; Imperfections

1. Introduction

As automotive manufacturers have to take environmental issues into consider-

ation, the aluminum space frame concept has become more and more attractive in
the design of vehicles. When using aluminum in the body structure of a vehicle,
weight reductions of as much as 25% can be achieved compared to conventional
steel structures. Aluminum is quite a new material in modern car body design. This

Corresponding author. Fax: +33-4-7243-8523.
E-mail addresses: [email protected], [email protected] (D. Al Galib).

0263-8231/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tws.2004.03.001
1104 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Nomenclature

A material constant [12]

amax maximum amplitude of imperfection
B material constant [12]
D diameter of cylinder
E elastic modulus
fe frequency
2H initial distance between plastic hinges at the top and bottom of a
basic folding element [5]
K impact energy or energy absorbed by the specimen
L initial length of tube (mm)
M mass
M0 full plastic moment of the tube wall per unit length, ð1=4Þr0 t2 .
N number of circumferential corners in non-axisymmetric mode of
deformation
n hardening parameter, material constant [12]
PExperimental
md mean dynamic crushing load in experimental tests, defined by
Ðd
PExperimental
md ¼ ð1=dÞ 0 PðdÞdd
PAnalytical
ms analytical prediction for mean static crushing force (Eqs. (1)–(10))
PExperimental
ms mean static crushing load in experimental tests, defined by
Ðd
PExperimental
ms ¼ ð1=dÞ 0 PðdÞdd
R mean radius of circular extrusion
T time
t wall thickness
vo impact velocity of projectile
d axial crushing displacement of cylindrical tube
deff effective crushing distance for the formation of a complete lobe
e total strain
eu ultimate strain
rl initial yield stress
ru ultimate stress
r0.2 stress at 0.2% strain (0.2% proof stress)
r0 energy equivalent flow stress

means that its energy absorption capabilities have to be evaluated such that the
integrity of the passenger compartment is ensured and minimal accelerations are
transferred to the occupants.
For application in the design of devices for absorbing energy in case of a crash
or accident, the elasto-plastic behavior of structural elements like tubes of circular
and non-circular cross-section has received considerable attention [6,13,14,16,20,21]
during the last four decades.
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1105

A circular tube of mean diameter D and thickness t which undergoes stable pro-
gressive collapse and deforms in an axisymmetric concertina, diamond or mixed
mode is an efficient energy absorbing element. However, when its length is greater
than the critical length for the given tube, it deforms in the overall Euler (beam)
buckling mode, which is an inefficient mode of energy absorption and needs to be
avoided in crashworthiness applications. Modes of deformation and load–displace-
ment curves of round aluminum tubes of D=t ¼ 4 63, L=D ¼ 0:17 8:75, without
any discontinuities, were studied experimentally by Andrews et al. [6]. They pre-
sented a classification chart, which reveals that the concertina mode of deformation
occurs in tubes of D/t varying from 10 to 63, and for L/D ratios which vary for
different D/t values, the maximum being about 5 for D=t ¼ 18. Under slow loading
(quasi-static conditions), post-buckling collapse could have begun anywhere in the
specimen and often took the form of discrete folds. Under dynamic conditions, col-
lapse began at the struck end and the fold tended to be compact.
The following paragraphs summarize the available literature on the plastic col-
lapse behavior of thin-walled circular tubes. It is arranged broadly in chronological
order for each type of mode of deformation.
A theoretical study to compute the mean collapse load for cylindrical tubes and
columns deforming in concertina mode was first given by Alexander [5]. This sim-
ple model assumes the formation of four plastic hinges, and that the collapsing
length of the tube consists of two straight arms between the hinges. Abramowicz
and Jones [2,3] later reworked Alexander’s rigid-plastic analysis, and they proposed
an improved model by introducing curvature in the deforming fold length.
Abramowicz and Jones [2,3] observed that reasonable agreement existed between
predictions of average crush loads based on their equations (Eq. (4) in Table 4)
and experimental results for steel tubes with D=t ¼ 9 25. In subsequent work,
Abramowicz and Jones [4] reported on further tests and summarized their findings
for both static and dynamic loading cases in two failures mode maps, adding to the
previous work by Andrews et al. [6].
In 1993 and later in 1997, Gupta et al. [8,10] performed a series of quasi-static
axial compression tests on thin-walled aluminum and mild steel circular tubes, with
D/t ranging from 12 to 30 in both annealed and as received conditions. They com-
bined all their results and developed an empirical model for determining the aver-
age crushing force and the load history.
An empirical expression was derived by Guillow et al. [9] to find the average
crush load (see Eq. (11) in Table 4). From comparison with experimental results,
carried out on circular aluminum tubes with D=t ¼ 10 450, they found that this
formula was applicable (surprisingly) for both axisymmetric and non-axisymmetric
modes.
Pugsley and Macaulay [24] were among the first researchers to consider the non-
axisymmetric folding mode, their study being largely empirical. They proposed an
expression (Eq. (7) in Table 4), which was further modified by Jones and Abramo-
wicz [15] (Eq. (9) in Table 4).
However, as aluminum is quite a new material in vehicle design, and has differ-
ent properties than steel, which is commonly used, validation of the computer code
1106 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

models is necessary. For different categories of aluminum, the reported experi-

mental data for validation studies as well as the experimental details are still lim-
ited or insufficient.
The main objectives of this research are: (1) to present further experimental data
and to validate an existing theoretical prediction for static and dynamic axial
crushing of circular aluminum extrusions in alloy A6060 temper T5; (2) to study
the static and dynamic behavior of circular thin-walled tubes subjected to axial
loading having variable mass and impact velocity values; (3) to study any relation-
ship between the dynamic and static behavior; (4) to validate the numerical predic-
tion of the crushing behavior using the experimental results obtained previously;
and (5) to gauge the effect of initial imperfections on the behavior and the
deformed shape of these tubes.

2. Material properties

The aluminum extrusions used in the tests were made up of alloy A6060 temper
T5. This is an Al–Mg–Si-alloy with the following chemical composition (Table 1).
After extrusion, the profiles were air or water cooled and stretched to an elonga-
tion of 1–2%. Temper 4 was naturally aged at room temperature, whereas T6 was
v v
aged for 8 h at 175 C or 6 h at 185 C. In order to obtain temper T5, the alloy
A6060 was ‘‘aged on press’’ by cooling by blowing air or mist.
Typical engineering tensile stress–strain curves of the material were found by
standard tensile testing. The test specimens were taken from the sidewalls of a
square extrusion parallel to the direction of extrusion.
Fig. 1 shows a typical engineering stress–strain curve, while Table 2 compiles the
most relevant mean mechanical data (the 0.2%—proof stress and the ultimate
stress had a standard deviation of less than 2%). Parameters A, B and n in the
table are the material constants in the Johnson–Cook [12] equation
r ¼ A þ Bðep Þn ð1Þ
which is used here to represent the stress–strain curve. These constants were
obtained by curve fitting to the experimental stress–strain curve, which gave good
agreement up to 1% strain (see Fig. 1).
The effect of strain rate on the mechanical properties of this aluminum has not
been studied, but it is known that this effect is negligible in the range of tested velo-
city (7–9 m/s), according to the studies of McGregor et al. [23].

Table 1
Chemical composition of aluminum A6060 temper T5
% Si Fe Cu Mn Mg Cr Zn Ti Other Aluminum
a
Minimum 0.30 0.10 0.10 0.10 0.35 0.05 0.15 0.10 0.15 The rest
Maximuma 0.60 0.30 0.10 0.10 0.60 0.05 0.15 0.10 0.15
a
All in weight %.
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1107

Fig. 1. Typical engineering stress–strain curves for aluminum A6060 T5.

Table 2
Mechanical properties (based on the true stress–strain curve)
r0.2 (MPa) ru (MPa) eu (%) r0 (MPa) E (MPa) A ¼ rl (MPa) B (MPa) n
218.8 264 7.1 243 69 500 99.02 207.524 0.092

3. Experimental test program

Fig. 2 illustrates the test specimen geometry and boundary conditions. The main
objective was to study the peak load and energy absorption behavior of circular
thin-walled aluminum extrusion subjected to a quasi-static axial compression as
well as under axial crushing.
Tubes of 58 mm diameter and 2.0 mm thickness were obtained commercially and
later cut to make specimens of the required lengths. The D/t ratio was 29 and the
L/D ratio was 3.44. In this range, Andrews et al. [6] and Abramowicz and Jones
[2] showed that for these ratios, the tubes should deform in axisymmetric mode.

Fig. 2. Geometry and experimental details.

1108 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

3.1. Static tests

Quasi-static tests were carried out in a Man 200 kN Universal Tension-Com-

pression testing machine. The specimen was fixed at the lower end by means of
steel pieces (sleeves) embedded a distance of 18 mm inside the tube matching the
inside periphery of the specimens (Fig. 3), which was fixed to the moving plate of
the machine. The load was applied through a thick steel plate pinned axially to the
crosshead as shown in Fig. 2. The quasi-static test was carried out in order to
study any relationship between the dynamic and static behavior.
3.2. Dynamic tests

The test rig of INRETS—France (Fig. 4) was used to perform the impact tests.
This rig was composed of a carrying truck on rails, which moved by means of an
inertia wheel. The wheel was connected to the carrying truck by a cable over a
launching distance of 20 m. The maximum speed was 60 km/h (1 km/h). The
truck was hooked to the impacting truck with a pin, which broke when the steel
beams hit a concrete block (Fig. 4). Its speed was then transmitted to the impact-
ing masses, which varied from 30 to 200 kg.
This boundary condition was identical to that used for the static test. The decel-
eration of the impacting truck was measured directly by an accelerometer, which
was fixed to the impacting truck. A displacement cell (laser sensor) was placed on
the concrete wall to measure the axial shortening of the impacted extrusions. In
order to evaluate the displacement of the impacting truck, the axial shortening was
also calculated from double integration of the measured deceleration and verified
with that measured by the laser sensor. Data acquisition was done by means of a

Fig. 3. Boundary conditions used for circular tubes.

 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1109

Fig. 4. Dynamic test rig arrangement.

32 track (channel) TECHNIPHONE case equipped with an Ethernet card enabling

the data to be transmitted to a recording computer during the whole test. Load–
time, displacement–time and acceleration–time characteristics were then obtained.
In order to smooth out the recorded curves, all the recorded signals during the
test were filtered with a cut-off frequency of 2.5 kHz, and the data were re-sampled
at a frequency of fe ¼ 10 kHz. All the values were recorded at this frequency every
0.1 ms.

4. Test results

4.1. Static tests

Fig. 5 illustrates typical load–displacement curve of a tube deforming in the con-
certina mode. Alternate high and low peak loads characterize this curve. These
peaks, marked E and F, correspond to the formation of the outward and inward
parts of the folds. Prior to buckling, an axisymmetric deformation mode with one
or two axisymmetric buckles was observed near the top and/or bottom ends of the
specimens. Without change of this mode, the equilibrium curve reaches the
maximum load level at point A. At this level, and if the loading was continued, one
of the buckles localized at a position that varied from test to test. The wall begins
to bend outwards and the force falls sharply until the complete formation of the
first fold. The force reaches its minimum at D, and starts to increase again. The
wall near the deformed zone tends to bend inwards and another force peak E
appears at the moment of appearance of the internal buckling pattern. The force
1110 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 5. Static axial crushing behavior of a thin-walled aluminum circular tube (Test SR000).

then decreases as the wall bends inwards. After a small internal bending of the wall
F, the force begins to increase again until point G, and the pattern is repeated.
Fig. 6 illustrates typical load–displacement and energy–displacement curves of
the tested extrusions in the two cases (axisymmetric and non-axisymmetric), while
Table 3 summarizes the most important data from the static tests carried out.
Although the mode of loading and the boundary conditions were identical, the
modes of deformation are different from one test to another. Two principal modes
of deformation have been observed for these extrusions, as illustrated in Fig. 7.

Fig. 6. Force, mean load and energy curves versus displacement (static test): (a): axisymmetric mode
(static test SR002) (b): mixed mode (static test SR001).
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1111

Table 3
Static test results
Tests no. PExp
max dExp
max dExp 2H mm dExp PExp
mean
Mode of deformation
eff f
ðkNÞ ðmmÞ ðmmÞ ðmmÞ ðkNÞ

SR000 75 3.322 16.69 21.74 120 38 7 Axisymmetric lobes

SR001 73 3.410 17.28 22.42 82 32 3 Axisymmetric lobes; then
ovalization (mode mixed)
SR002 71 3.502 17.22 22.35 82 36 5 lobes (Axisymmetric)
SR003 73 3.550 17.29 22.43 82 33 2 lobes at the top and 3 lobes
at the bottom
SR004 70 3.606 18.48 23.82 82 36 2 Axisymmetric lobes; then
ovalization (mode mixed)
SR005 71 4.356 18.79 24.18 58 33 2 Axisymmetric lobes then
ovalization (mode mixed)
SR01 71 2.0 16.96 22.05 102 33 5 lobes (Axisymmetric)
SR02 72 2.4 16.96 22.05 103 34 5 lobes (Axisymmetric)
SRTER 69 1.962 99 39.5 Non-axisymmetric (diamond)

4.1.1. Axisymmetric mode of deformation

As shown in Fig. 6, certain specimens exhibit a progressive axisymmetric defor-
mation mode. When the specimen (Test SR002) was fully compressed, five lobes
were found. The first lobe formed at the uppermost section. Each peak in the load–
displacement curves shown in Fig. 6(a) was associated with the formation of one
lobe.
The average value of the first peak was measured to be 72 kN (the standard
deviation is less than 2%), while the corresponding values for the other peaks
were around 60 kN which is 15% smaller than the value of the first peak. This is
due to the fact that for the formation of the first peak, the structure is free of all
distortions (deformations), the deformations already formed during this fold going
to influence the formation of the next one by producing a local bending at the plas-
tic hinges level; so the other peaks will be, due to this fact, lower than the first one,

Fig. 7. Deformed modes (axisymmetric and mixed) of static tests.

1112 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

because the second lobe is already initiated. The average crushing distance to form
a complete lobe was measured to be 18 mm.

4.1.2. Mixed mode of deformation

For the test SR001, the specimen developed three axisymmetric folds, and then
there was ovalization of the right section of the tube leading to the formation of
mode 2 of deformation, as shown in Fig. 7 and Table 3. In certain cases, the ovali-
zation occurred after the second fold, which always remains axisymmetric as in the
case of the test SR005. This phenomenon indicates that mode 2 will be likely to
appear during the formation of the second and third folds.
For the same displacement, one can see that the energy absorbed by kinematics
of ‘‘mixed’’ type is approximately 7% higher than that corresponding to an axisym-
metric mode. However, the mode of mixed type induces asymmetry, which can
encourage the global bending mode (Euler mode), which would strongly restrict
the capacity for energy absorption. The axisymmetric mode of deformation guar-
antees a progressive collapse; in this sense, it is preferable even though the energy
assessment is not favorable.
In order to verify the influence of loading conditions on the critical mode, the
uniformity of the load applied at the upper section has been modified. The new
mode of loading was assimilated to a loading at three equidistant points in the
uppermost section. The obtained kinematics corresponded to the ‘‘diamond’’ mode
at three harmonics (see Fig. 8).

Fig. 8. Deformed mode (non-axisymmetric) of static tests.

 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1113

Fig. 9. (a) Initial imperfections, (b) Composition in Fourier series.

The comparison of the energies shows that the diamond mode absorbs 9% more
energy (E ¼ 3899 J) than the axisymmetric mode (E ¼ 3541 J).

4.1.3. Influence of initial imperfections

The analysis of the initial imperfections of the round tubes showed that the pro-
cess of manufacture by the extrusion of liquid aluminum systematically leads to
some geometric imperfections. Fig. 9(a) represents a summary of the shapes of
imperfections measured with a laser sensor. This 3D view permits one to localize
the geometric imperfections situated on the surface of the analyzed tube.
Decomposition by the Fourier series on the 30 first harmonics shows a predomi-
nance of the amplitude of the corresponding defect to the second harmonic
(Amax =t ¼ 0:03), as one can see in the histogram in Fig. 9(b). The other amplitudes
remain weak, but it is interesting to note that this defect is inherent in the process
of manufacture because this mode is similar (close to 10 measured tubes) in terms
of amplitude and wavelength.
When the tube is subjected to axial compression, the first fold formed progress-
ively in mode 0. The amplitude of the initial imperfections will influence the forma-
tion of the other folds. Mode 2 will be therefore likely to appear during the
formation of every fold leading to the ovalization of the right section.

4.1.4. Comparison of the analytical models

Fig. 10 shows the ratio between the test data given in Table 3 and the average
crushing force calculated using the analytical models mentioned in Table 4 for tes-
ted extrusions, as a function of the D/t ratio. It was found from Fig. 10(a) that test
1114 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 10. Comparison between analytical models and experimental tests. (a) Axisymmetric mode; (b)
Non-axisymmetric mode.

results were in good agreement with the analytical model proposed by Abramowicz
and Jones in 1986 for an axisymmetric mode of deformation. The same conclusion
stays valid for the non-axisymmetric mode shown in Fig. 10(b).
Taking strain-hardening effects into account, the energy equivalent flow stress
proposed by Wierzbicki and Schneider [31] and Wierzbicki and Abramowicz [29]
(Table 5) was used. In the calculation, either the engineering or the true stress–
strain curves have been used. Fig. 11 and Table 6 show again that the present
experimental data overestimate the calculated mean crushing force by 4–16%.
However, a much better agreement between the experimental data and the model
proposed by Abramowicz and Jones [3] is found by using a flow stress, which is
calculated from the true stress–strain curve. The analytical mean crushing force can
be predicted with the same accuracy by introducing the definition of the flow stress
as the mean value between the 0.2% proof stress and the ultimate stress (Table 5).

5. Dynamic tests

The test data are listed in Table 7, while some representative photographs of
deformed specimens are presented in Fig. 12. The localization of the axisymmetric
and mixed mode of deformation for all tested specimen, started at the impacted end.
Impact energy versus axial crushing displacement for all specimens is shown in
Fig. 13. An approximately linear relationship between energy and displacement
was found for all tests.
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1115

Table 4
Analytical models
Eq. Model Pm H de ¼ 2H  2xm  t
No. derived by M0 R
xm ¼0:28H2 from [1]
rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi
(2) Alexander [5] Pm 2R H t de t
¼ 20:725 þ 6:283 ¼ 1:905 ¼ 0:86  0:525
M0 t R 2R 2H 2R
¼ 117:9
rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi
(3) Abramowicz and Pm 2R H t de t
¼ 22:366 þ 11:766 ¼ 1:618 ¼ 0:86  0:618
Jones [2] M0 t R 2R 2H 2R
¼ 132:2
rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi
(4) Abramowicz and Pm 2R H t de t
¼ 25:230 þ 15:09 ¼ 1:84 ¼ 0:86  0:543
Jones [3] M0 t R 2R 2H 2R
¼ 150:95
rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi
(5) Wierzbicki and Pm 2R H 2R
¼ 35:22 ¼ 189:7 ¼ 1:34
Bhat [28] M0 t R t
rffiffiffiffiffiffiffi rffiffiffiffiffiffiffi
(6) Wierzbicki et al. Pm 2R H 2R
¼ 31:74 ¼ 172:4 ¼ 1:31
[30] M0 t R t

P  t  r 
(7) Pugsley [32] a -
¼ 9:097
2r0 pRt R r0
rffiffiffiffiffiffiffi
(8) Wierzbicki [33] Pm 3 2R
¼ 62:88
M0 t
rffiffiffiffiffiffiffi
(9) Abramowicz and Pm 2R
¼ A1N þ A2N
Jones [3] M0 t
rffiffiffiffiffiffiffi
(10) Singace et al. [25] Pm 2R
¼ 22:27 þ5:632 ¼ 126
M0 t
 0:32
(11) Guillow et al. [9] Pm 2R
¼ 72:3 ¼ 212
M0 t
Test exp. 156.3

Fig. 14 shows typical measured load–displacement and means load–displacement

curves. These curves were characterized by a high initial peak load followed by oscil-
lations. The high frequency content was assumed to be associated with the vibration
of the test rig. It can be seen that no significant change in the force–displacement
curve is found when increasing the mass of the impactor from 87 to 117 kg. The only
difference observed is an increased axial crushing displacement of the specimen when

Table 5
Flow stress
r0 derived by Wierzbicki and Wierzbicki and ðru þ r0:2 Þ=2
Schneider [31](MPa) Abramowicz [29] (MPa)
Engineering curves 233 245 232
True curves 243 256 242
1116 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 11. Ratio of calculated and experimental mean crushing force.

Table 6
Ratio of calculated and experimental mean crushing force

PExperimental
mean =PAnalytical
mean
r0 (MPa) SR001 SR002 SR003 SR004 SR005 SR01 SR02 Mean
233 1.039 1.169 1.071 1.169 1.071 1.071 1.104 1.099
242 1.003 1.128 1.034 1.128 1.034 1.034 1.065 1.06

Table 7
Dynamic test results
Test no. PExp
max PExp
mean
V0 M dfinal H Deformation modes
ðkNÞ ðkNÞ (m/s) (Kg) (mm) (mm)

DRA01 106 33 7.2 87 88 21.5 1 lobe (Axisymmetric)

DRA02 108 39 7.2 87 39 13.8 2 lobes (Axisymmetric)
DRA03 108 38 7.2 87 53 16.3 3 lobes (Axisymmetric)
DRA04 106 38 7.2 87 50 15.5 3 lobes (Axisymmetric)
DRA10 107 38 8.97 117 99 17.0 4 lobes (Axisymmetric); then ovalization
DRA11 120 38 7.83 117 85 17.6 4 lobes (Axisymmetric); then ovalization
Average 109 37 7.6 97 – –

increasing the mass of the impactor as a result of increased impact energy. Fig. 14
also shows that the spread in the values of the mean crushing force is less than 3%.
At the final displacement, all the impact energy was absorbed by the specimen.

6. Comparison of static and dynamic tests

A representative comparison of dynamic and static load–displacement curves is

shown in Fig. 15. As can be seen, the main difference is related to the first part of
D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1117

Fig. 12. Deformed modes (dynamic tests).

Fig. 13. Impact energy versus axial crushing displacement.

Fig. 14. Comparison between dynamic tests.

1118 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 15. Comparison between static and dynamic tests.

the impact, where the dynamic force level is significantly higher than the static one
(about 40–60%). The real responses of force–displacement were perturbed by the
vibration of the adjacent structure. The mean dynamic crushing loads are about
10% higher than the corresponding values in quasi-static tests and the obtained
modes of deformation are generally the same as in the static tests. The increased
force level at the beginning of the impact compared with a static test leads to a
decay ratio between the measured dynamic and static mean loads with respect to
the axial displacement. This is shown in Fig. 16 for one selected test: the decay
ratio falls from 1.48 at a displacement of 6 mm to 1.07 for a 100 mm displacement.
Gupta [GUP 97] and Langseth and Hopperstad [20] also showed the same obser-
vations for their experimental studies on circular and square aluminum extrusion.
Their explanation for this difference is that it is only due to the inertia effects set up

Fig. 16. Ratio of dynamic and static mean loads versus displacement—specimens SR01 and DRA10.
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1119

at the instant of impact due to the lateral movement of the sidewalls in order to
initiate the folding process. This is true, but it may also be due to other external
factors such as the vibrations of the plate fixed on four load sensors on which the
test specimen was positioned (Fig. 4). These vibrations appeared when the impact-
ing truck was in contact with the specimen in order to initiate the folding process.
The plate was submitted to forced vibrations through the intermediary of the speci-
men. The force applied to the plate was composed of peaks and valleys represent-
ing the stiffening and softening phases corresponding to fold formation on the
tested structure. Section 8.3 shows the numerical modelization clarifying this
phenomenon.
6.1. Scatter between the first peak loads
Our analysis in this section will be only on the spread of the first peaks. Fig. 17
shows the spread of the obtained experimental dynamic first peak load values for
tests velocities 7.2, 7.83, and 8.97 m/s, around the average first quasi-static peak
load. The ratios Pdmax =Psmax of the dynamic to quasi-static peak loads are quantified
in Table 7. It was found that the dynamic first peak load was 40–60% greater than
the quasi-static first peak load.
6.2. Scatter of the mean crushing load

Fig. 18 shows the scatter of the mean dynamic and static loads. The ratios
between them are presented in Table 8. From this table, we see that the dynamic
mean crushing load is 10% greater than the quasi-static mean crushing force. This
confirms the non-sensitivity of aluminum A6060 T5 to the strain rate effect.
6.3. Effect of the impacting mass on the mean load
To quantify the effect of the impacting mass on the value of the mean crushing
force, we refer again to the values presented in Table 7. This table shows no signifi-

Fig. 17. Spread of the dynamic first peak loads around the static average first peak load.
1120 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 18. Spread of the dynamic mean crushing loads around the static average mean load.

cant difference between the mean crushing loads for the velocity of 7.2 m/s for dif-
ferent masses of the impactor (87 and 117 kg). At this impact velocity, the differ-
ence between the average values is less than 4%. This difference, however, vanishes
for the tests at velocities 7.83 and 8.97 m/s for a constant mass (117 kg).
7. Influence of the strain rate

Fig. 19 presents the ratio of the dynamic to static specific energies of all the tests.
This ratio was plotted versus the strain rate calculated by the equation proposed
by Abramowicz and Jones [2]: for the axisymmetric or concertina crushing mode,
0:25v0
e_ ¼  rffiffiffiffiffiffiffi ð12Þ
t
R 0:86  0:618
2R
and for the non-axisymmetric or diamond crushing mode,
0:370v0
e_ ¼ ð13Þ
R

Table 8
Ratio of dynamic to static peak and mean loads
First peak load V0 (m/s) Dynamic test results M (Kg) Mean load Dynamic test results

Pdmax ðkNÞ 7.2 106 and 108 87 Pdmax ðkNÞ 33–39

Pdmax =Psmax 1.47–1.5 Pdmax =Psmax 0.9–1.1
Pdmax ðkNÞ 8.97 107 117 Pdmax ðkNÞ 38
Pdmax =Psmax 1.48 Pdmax =Psmax 1.08
Pdmax ðkNÞ 7.83 120 117 Pdmax ðkNÞ 38
Pdmax =Psmax 1.66 Pdmax =Psmax 1.08
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1121

Fig. 19. Strain rate effect.

The horizontal line in this figure defines the ratio for a strain-rate-insensitive
material. It is obvious that the experimental results are slightly above this horizon-
tal line. This is because the aluminum alloy A6060 T5 is not sensitive to the
strain rate magnitude at the tested range (91 s1 < ^e < 114 s1 ).

8. Description of finite element modeling

A non-linear finite static and dynamic finite element analysis was performed. The
numerical simulations were carried out using the explicit dynamics FE code
RADIOSS version 4.1 [22]. In the finite element model, the lower fixed end was
modeled with a rigid body connecting all lower nodes (z ¼ 18 mm). The rigid body
was completely fixed in translation and rotation.
During progressive formation of plastic folds, using self-contact interface pre-
vented interpenetration between two folds.
For all static and dynamic simulations, the aluminum alloy was modeled with an
elasto-plastic law, with isotropic hardening and Von Mises yield criterion. The true
stress/strain curve given in Fig. 1 is discretized by 18 points. The four-node shell
element [7] with five integration points through the thickness was adopted.

8.1. Simulation of static test

The explicit solution method is a true dynamic procedure originally developed to

model high-speed impact events, in which inertia plays a dominant role in the solution.
Therefore, in a quasi-static analysis, the goal is to model the process in the shortest
time period in which inertial forces remain insignificant. To simulate the actual
quasi-static experiments, the speed of the crosshead is usually set to 0.01–1 mm/s.
This range of velocity is too slow for the numerical simulation. This is due to the fact
that the explicit time integration method is only conditionally stable, and therefore, in
general, very small time increments have to be used.
1122 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 20. Quasi-static simulation—energy and crushing load response.

One of the ways of achieving a quasi-static process by using the explicit dynamic
procedure is to scale up the mass, while keeping velocity very low. Scaling up the
mass results in a large time step, therefore reducing the number of time step incre-
ments for such a low loading rate. In this study, initial velocities of 0.5, 1.0, and
5 m/s were applied at the top of the specimen with energy equivalent to that
absorbed by the specimens during the experimental tests. It was observed that the
total kinetic energy is very small compared to the total internal energy over the per-
iod of the crushing process, and the load–displacement responses are independent of
the velocity, which shows that inertia effects are negligible at the peak load value up
to this velocity. Fig. 20 shows quasi-static simulations with three different mass and
loading rates. It shows that the kinetic energy is very small (quasi-null) compared to
the internal energy for all of the three different mass and applied velocities. As a
result, their corresponding crushing force responses are similar to one another.
Therefore, the numerical simulation can be considered as a quasi-static analysis.
Fig. 21 shows the deformation modes obtained from the numerical simulation
with v ¼ 5 m=s and v ¼ 10 m=s, respectively; these modes show relatively good
agreement between the numerical predictions and the test results. For the run with

Fig. 21. Quasi-static numerical deformation modes.

 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1123

Fig. 22. Crushing load versus axial displacement obtained by different numerical simulations.

low impact velocity (5 m/s), three axisymmetric folds appear at the top end and the
ovalization mode of deformation installed just after. The same mode of deformation
was obtained for the test SR001. However, a complete axisymmetric mode of defor-
mation was obtained for the run with the velocity of 10 m/s as for the test SR00.
A static numerical simulation was also carried out using the general-purpose
implicit FE code ABAQUS/Standard [11]. The same FE model was used and a
uniform displacement of 10 mm for all the nodes of the uppermost section was
imposed. The eight-node shell element with reduced integration points (S8R5) was
used. The results are presented in Fig. 22, which shows an identical behavior to
that obtained, by RADIOSS with a peak load of (Pmax ¼ 73 kN). The displace-
ment on ABAQUS was limited to 10 mm in the post-failure range before the
occurrence of the first contact between the folds. A sensitivity analysis on the size
of the elements was carried out in order to minimize mesh refinement effect on the
accuracy of the numerical results.
The simulation study is extended to cover the verification of the influence of load-
ing conditions on the deformation mode and crash characteristics. Two, three and
v v v
four equidistance (h ¼ 180 , 120 and 90 , respectively) lumped masses of 60, 40 and
30 kg were applied at the uppermost section (z ¼ 200 mm). The other parameters
were kept as before (dynamic test). Table 9 compiles the most relevant numerical
results obtained from this study. Firstly, one can conclude that the deformation
mode was changed according to the mode of loading and the expected mode of
deformation was correctly achieved (Fig. 23); secondly, the obtained kinematics cor-
responded to that obtained by static test (Fig. 8) ‘‘diamond’’ on three harmonics.
For the finite element simulation of the test specimen SRTER, the 1st peak load
and the absorbed energy predictions were within acceptable limits (5% and 3%,
respectively), as shown in Fig. 24 and Table 8, but the specimen required a 3%
lower mean force to crush it than the experimental specimen. The numerical simu-
lations gave a good approximation of the extrusion’s behavior.
1124 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Table 9
Results of numerical analysis of circular tubes loaded by 0, 2, 3 and 4 concentrated mass
Point mass Numerical simulation Experimental test
Pmax Pmean E (N m) Mode of deformation Pmax Pmean E (N m)
(kN) (kN) (kN) (kN)

0 (specimen SR01) 72 34 3500 Axisymmetric 71 35 3541

2 56 38 3745 Diamonds (N ¼ 2)
3 (specimen SRTER) 73 38 4032 Diamonds (N ¼ 3) 69 39 3899
4 71 47 4594 Diamonds (N ¼ 4)

8.2. Simulation of impact tests

The simulations of the dynamic tests were done for the DRA10 and DRA11
tests by taking into account the mass and initial velocity applied during the tests.

Fig. 23. Numerical deformed modes (diamond): (a) two-point load; (b) three-point load; (c) four-point load.
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1125

Fig. 24. Comparison between dynamic numerical simulation and static test results.

As for the static simulation, the strain rate effect was not taken into account.
Fig. 26 shows a comparison of the predicted and measured load–displacement,
mean load–displacement and energy–displacement curves for the studied extru-
sions. An exact match between the predicted and experimental curves was not
achieved, even if the specimen showed good correlation during the first part (initial
rigidity) of the deformation process. Furthermore, the first peak load was nearly
the same as that obtained in the static tests. Despite such discrepancy, the mean
load and the absorbed energy as a function of the axial displacement are predicted
with satisfactory accuracy. The average ratio of the predicted to measured values
was found to be within 6%.

8.3. Towards a better dynamic modelization

As illustrated in Fig. 25, the dynamic load was significantly higher and more per-
turbed than the corresponding static one. A limited number of data are available in
the literature regarding this phenomenon in an aluminum tube during the crash
process. Langseth and Hopperstad [20] found from experimental work on square
aluminum extrusions that this effect was only related to the inertia forces. From
our point of view, and as the aluminum alloy was recognized as strain rate insensi-
tive in the considered strain range, the observed difference was assumed to be asso-

Fig. 25. Comparison between dynamic numerical simulation, dynamic and static test results.
1126 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 26. Comparison between dynamic numerical simulation and dynamic test results.

ciated with number of external parameters that could influence the accuracy of the
measured impact force. Possible errors may result from wave propagation phenom-
enon, vibration characteristics of all masses involved and the damping properties
of the foundations. A recent numerical simulation by Karagiozova and Jones
[18,19] on aluminum alloy tubes also supports this point of view, and shows that
the stress wave propagation influences the particular pattern of buckling. Let us
now discuss this point of view in more detail. In order to give an explanation of
the differences between the measured static and dynamic crash characteristics, a
better modelization was performed taking into account the real boundaries.
To take into account the vibration of the plate on which the tube and load sen-
sors were fixed, the plate was modeled with an eight-node brick element. Load sen-
sors are represented by four fixed nodes at the lower surface of the plate and, the
specimen was modeled at the center of the plate as in the dynamic tests; only the
axial displacement of the lower end of the specimen was not restrained.

Fig. 27. Modes of deformation—modelization of the fixing plate.

 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1127

When the impacting truck came into contact with the tube, before it was cru-
shed, the load was first transmitted to the plate, thus allowing it to bend, and thus
the crushing process of the tube began. During this process, vibrations took place
in the plate and pretreated the data acquisition of the sensors.
Fig. 26 shows the results of this simulation. Here, an increase in the initial peak
load was found, which implies increased mean load and absorbed energy levels
when introducing the plate in the model and taking its vibration into account. The
predicted first peak load was within +7% of the experimental one, while the pre-
dicted mean load (calculated from load–displacement characteristics) and the
absorbed energy at a distance equal to 100 mm are within the range of 5%. Fur-
thermore, no change in the deformation mode is observed during the crash process
(Fig. 27), although a change in the deformation modes was observed during the
experimental test after the formation of the fourth fold. The same response (load–
displacement) was obtained when increasing the number of elements around the
thickness of the plate from 2 to 8. Four other identical simulations have been car-
ried out in order to verify the effect of plate thickness on the behavior of the tube.
Table 10 summarizes the results predicted from these simulations as well as a com-
parison with the experimental dynamic tests. From this table, we can see that the
first peak was influenced by the mode of vibration of the plate. A reduction of this
peak was observed by increasing the thickness of the plate.
Other factors like friction and damping can also be taken into account for more
accurate modeling of the structure. However, such factors can only be considered
at the expense of computer memory size and processing time. In other words, the
solution of the problem will be too expensive or might even be prohibitive.

8.4. Comparison of dynamic numerical analysis and tests

As in Fig. 10, the same comparison was established between the numerical
results and the results of dynamic tests. Fig. 28 shows the non-dimensional pre-
dicted mean crushing load over Mo as a function of 2R/t.
This comparison was established for 2R/t between 10 and 65, which represent
the circular sections available in the commercial markets. It can be seen that the
numerical prediction compares well with the experimental dynamic results for
2R=t ¼ 14:5; the cloud of the experimental dynamic results is within 14% of the

Table 10
Influence of plate thickness
Thickness of the Pmax Pmean Eabsorbed Psimulation
mean = Deformation mode
plate (mm) (kN) (kN) (kN) Ptest
mean

Without plate 77 39 3974 1.03 Axisymmetric (2 folds) ovalization

10 127 41 4144 1.08 Axisymmetric (4 folds) ovalization
15 118 43 4320 1.14 Axisymmetric (3 folds) ovalization
30 121 47 4696 1.24 Axisymmetric (4 folds) ovalization
150 88 44 4378 1.15 Axisymmetric (7 folds)
Test DRA11 120 38 3270 1.00 Axisymmetric (4 folds) ovalization
1128 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 28. Comparison between dynamic simulation results and experimental dynamic tests.

proposed numerical equation. It is interesting to compare our numerical dynamic

results with those of the analytical models proposed by Wierzbicki et al. [28,30], in
which the plastic hinges are allowed to move. The same figure compares these two
different results. It can be seen that the mean crushing force obtained in our
numerical analysis is close to that predicted by Wierzbicki et al.’s model [30] for
2R=t > 40, but for 2R=t < 40 the results are close to those of the analytical model
proposed in 1986. This leads us to conclude that a good degree of correlation exists
between our numerical results and those of the two analytical models proposed by
Wierzbicki et al.

8.5. Effect of mass and impact initial velocity

8.5.1. Review of previous studies
The following section summarizes the available literature regarding the inertia
effects set up in an aluminum tube during dynamic loading. A few publications
were found treating this problem numerically. An experimental study of plate
structures subjected to an in-plane mass impact was reported by Tam and Call-
adine [27]. They showed that, even within the range of law-velocity impacts, the
structure exhibited inertia and strain rate effects and it was both velocity and
impact mass sensitive. Later, Su et al. [26] examined this process theoretically by
studying a Type II structure (a column with two separate bars fixed together at
clamped supports), assuming deformation to take place at plastic hinges. An elastic
perfectly plastic constitutive relation for the material was assumed and the main
variables were the mass ratio of the projectile and the bars (1:25 <
mass ratio < 150), the impact velocity of the impactor and the initial geometry.
They found that the main parameter controlling the dynamic response was the
effective mass ratio of the projectile and bars and that this kind of structure should
be defined as an ‘‘inertia sensitive’’ rather than a ‘‘velocity sensitive’’ one. Further-
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1129

more, a reduction of the inertia effects was found by introducing initial geometrical
imperfections. A similar theoretical study on a simply supported rod using a dis-
crete model for elasto-plastic buckling retaining the axial and lateral inertia forces
was carried out by Karagiozova and Jones [17]. The mass ratio in this case was
0:8 < mass ratio < 100. They found that the inertia of the striking mass plays an
important role in the development of the buckling shape of the rod and that quasi-
static methods of analyses can be applied correctly when the striking mass is much
larger than the mass of the rod.
As the response of circular tubes was also characterized by a peak load followed
by a decreasing load, or even by a number of such phases, it can be expected that
there are strong similarities between the Type II structure and the present alumi-
num tube problem. However, in order to extend the parametric range of the
experimental test program and to see if the physical observations for more simple
structures (e.g. Type II structure) are representative of the present tube problem, a
numerical study was performed to investigate the effect of impact velocity and
mass ratio on the tube response.
Table 11 shows the numerical test program set up to study the response of the
tube as a function of mass and initial velocity. The mass of the impactor (rigid
wall) was varied to take the variation of this quantity into account. The tube was
modeled as described in Section 8. At the uppermost section, rotational degrees of
freedom were prevented in order to obtain progressive folding for all combinations
of mass and impact velocities and to avoid unrealistic deformation modes.
Series D1 was performed to study the effect of keeping the impact velocity con-
stant and varying the mass of the projectile. In series D2, the impacting initial velo-
city was varied, keeping the mass of the impactor constant. Finally, series D3 was

Table 11
Effect of mass and initial impact velocity
Series Model parameters
Constant Variable
D1 Impact initial velocity vi ¼ 8:97 m=s Mimpactor ¼ 25 kg
50 kg
75 kg
100 kg
117 kg
D2 Impacting mass Mimpactor ¼ 50 kg vi ¼ 5:0 m=s
10 m/s
15 m/s
20 m/s
25 m/s
D3 Constant impact energy ð1=2ÞMv2i ¼ 5750 Nm vi ¼ 4:796 m=s and Mimpactor ¼ 500 kg
vi ¼ 10:724 m=s and Mimpactor ¼ 100 kg
vi ¼ 15:166 m=s and Mimpactor ¼ 50 kg
vi ¼ 21:448 m=s and Mimpactor ¼ 25 kg

Mass of extrusion, m ¼ 0:196 kg.

1130 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 29. Series D1: constant initial velocity ¼ 9 m=s.

included to study the effect of keeping the impact energy constant. The reference
kinetic energy in this case was 5750 N m.

8.5.2. Numerical results

One of the quantities of particular interest in the dynamic analysis of tubes is the
interface force between the impactor and the top of the specimen as a function of
axial displacement. In the following, these curves will be used together with the
corresponding mean force as a function of the axial displacement to interpret the
parametric study. For simplicity and clarity, only three load–displacement curves
are shown, while a complete set of curves is shown for the mean load.
Fig. 29 shows the results from series D1, where the impact velocity was kept
constant at 9 m/s. It can be seen that no significant change in the force displace-
ment curve is found when increasing the mass of the impactor from 25 to 117 kg.
The only difference observed is an increased axial deformation of the tube when
increasing the mass of the impactor, as a result of increased impact energy. This
means that the extrusion behaves as a non-linear spring and the stiffness of the
spring is independent of the mass of the impactor. Furthermore, it was observed
that the initiation of the folds starting from the lower, clamped end.
Fig. 30 shows the results from series D2. In this case, the mass was kept constant
at 50 kg, while the impact velocity was varied in the range 5–25 m/s. It was

Fig. 30. Series D2: constant impactor mass, M ¼ 50 kg.

 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1131

Fig. 31. Series D3: constant impact energy ¼ 5750 N m.

observed that when increasing the impact initial velocity, the initial peak load was
increased, which implies an increased mean load.
Fig. 31 shows the results from the analysis of series D3, where a constant energy
input was used. From the mean load curves, it can easily be seen that an increased
impact velocity gives an increased mean load level for the same displacement.
From analysis series D2 and D3, the main conclusion that can be drawn is that
an increased impact velocity increases the initial peak (buckling) load. Numerical
simulations have shown that the force at the impacted end is not equal to that at
the clamped end; this reinforces our point of view (Section 8.3) that it is not only
the lateral inertia effects in the tube that control the dynamic behavior but all para-
meters discussed in Section 8.3.

8.6. Influence of initial imperfections

The process of manufacturing by the extrusion of liquid aluminum systemati-

cally leads to some geometric imperfections. Even if the magnitude of these imper-
fections is very small, one can expect that they have an effect on the deformation
mode or on the mean crushing load.
The quasi-static experimental test showed that mode 2 was likely to appear dur-
ing the formation of every fold, leading to ovalization of the right section. Further-
more, the analysis of initial defects in Section 4.1.3 shows a predominance of the
amplitude of the corresponding defect to the second harmonic. It will be judicious
to judge this type of defect numerically. The amplitude a/t was varied between 5%,
10%, and 100% to initiate this mode.
From Fig. 32, where the mean dynamic crushing forces are presented, we can see
that the mean crushing load or the dissipated energy is 8–11% smaller than the cor-
responding one without any initial defect. This means that the imperfections in
mode 2 influence the capacity for dissipation of energy. In addition, it was also
observed that the amplitude of the defect has no influence on the value of the first
peak, which confirms the experimental observations (see Table 3). To further study
the behavior of the tubes under axial deformation, plots of the deformation pro-
files are shown in Fig. 33. Two different amplitudes are shown, namely 0.1 and
1132 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 32. Mean load–displacement curves; effect of circumferential imperfections in mode 2.

2.0 mm. For the three amplitudes, the initiation of the first fold was axisymme-
trically switched to mode 2. Hence, the deformation shape at an axial displacement
of 115 mm is approximately the same. The main conclusion that can be drawn is
that the dynamic simulation of the initial imperfections provided a good prediction
of the results obtained in the experimental program. The mode of deformation was
well predicted as well as the mean load–displacement and dissipated energy–dis-
placement curves.
Numerical simulations have been carried out when representing the geometrical
imperfections by three and four sine waves in the circumferential direction of the
extrusions, over all the tube length. In this case, the number of sine waves con-
trolled the collapse mode developed and therefore had an influence on the energy

Fig. 33. Deformation shape.

 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1133

Table 12
Numerical simulation of circular tube with circumferential imperfections in mode 3 and 4
Mode N¼0 Ncircumferential ¼ 3 Ncircumferential ¼ 4
a/t 0 5% 10% 50% 100% 200% 5% 10% 50% 100%
A (mm) 0 0.1 0.2 1 2 4 0.1 0.2 1 2
Pmax (kN) 74 77 77 77 77 79 75 75 76 78
Pmean (kN) 46 41 39 37 37 37 43 40 48 35
Eabsorbed 4564 4146 3904 3707 3724 3684 4390 4016 4780 3535
(N m)
SD 0 9% 14% 19% 18% 19% 4% 12% +5% 22%
dfinal (mm) 100 111 115 118 119 119 107 113 96 120

absorption. Five amplitudes, for each mode, were tested, showing that an increased
value of amplitude of the imperfection will decrease the energy absorption (see
Table 12). Furthermore, the reduction in the energy-absorption properties for
mode 3 are about 9% for a tube with an imperfection amplitude (a/t) of 5% and
14% for a tube exhibiting an imperfection amplitude (a/t) of 10%, and about 19%
for the other amplitudes (Fig. 34) for the case of mode 3.
It was observed for an imperfection (a/t) of 5% that the first three folds formed
progressively in mode 0, whereas the following one corresponded to mode 3. The
formation of the other folds will be influenced by the amplitude of the initial
imperfections. Mode 3 will therefore be likely to appear during the formation of
the next folds, leading to a final mode of deformation of type diamond with three
corners.
To further study the behavior of the tubes under axial impact, the plot of the
deformation profiles for different values of a/t, namely 5%, 10%, 50%, and 100%,
are shown in Fig. 35. All simulations were stopped at a time equal to 22 ms, and
the final crushing displacement is shown in Table 12. A noticeable difference in the

Fig. 34. Energy–displacement curves; effect of circumferential imperfections in mode 3.

1134 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

Fig. 35. Deformation mode of circular tubes with circumferential imperfections in mode 3. (a)
A=t ¼ 5%, (b) A=t ¼ 10%, (c) A=t ¼ 50%, (d) A=t ¼ 100%.

Fig. 36. Deformation mode of circular tubes with circumferential imperfections in mode 4. (a)
A=t ¼ 5%, (b) A=t ¼ 10%, (c) A=t ¼ 50%, (d) A=t ¼ 100%.
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1135

nature of deformation mode is observed. For an imperfection amplitude of

a=t ¼ 5%, the diamond collapse mode with three circumferential lobes appears
after the initiation of three axisymmetric folds, while increasing the amplitude to
10% reduces the number of axisymmetric folds to one and the diamond mode is
seen. However, increasing the amplitude of the imperfection to a=t ¼ 100%, a
change in the mode is observed: the axisymmetric folds disappear and the diamond
collapse mode with three circumferential lobes is seen.
Fig. 35 shows the final deformed profiles for tubes with an introduced geometri-
cal imperfection in four circumferential lobes (corners). It was observed that for an
amplitude of A=t ¼ 100%, there was a rotation of the tube around the longitudinal
axis (direction of crushing), which implies a reduction in the energy absorbed by
this tube (about 22%). For an amplitude of a=t ¼ 50%, the initiation of the lobes
and the formation of a complete circumferential lobe are progressive in diamond
mode with four circumferential corners (Fig. 36(c)). The comparison of the ener-
gies shows that the diamond mode absorbs 5% more energy (E ¼ 4780 N m) than
the axisymmetric mode (E ¼ 4564 N m).

9. Conclusions

Experimental static and dynamic results of tests on circular tubes with D=t ¼ 14
of alloy A6060 Temper T5, are presented in this paper. Then, the numerical simu-
lations achieved on the same tubes are compared using these experimental data-
bases in order to verify and validate the proposed numerical model. Using this
validated model, parametric studies have been carried out in order to: (1) evaluate
the behavior of these tubes while varying the mass and the velocity of the impac-
tor; (2) study the effects of geometric imperfection, support conditions and type of
loading on the static and dynamic behavior of these circular tubes. The main con-
clusions from this study can be summarized as follows:

. The numerical model developed in this work correctly follows the experimental
test predictions. The simulation indicates that the crushing process follows the
actual test. The numerical model well predicted the 1st peak load, the mean
crushing force and the number of folds within 1% for the 1st peak load and 4%
for the mean crushing force.
. The comparison of the mean crushing force from different analytical solutions
and our numerical predictions for 10 < D=t < 65 shows good concordance
between them. Therefore, with the ability of numerical simulation to predict the
crash behavior of thin-walled tubes, this type of analysis is expected to be highly
useful in the design and analysis of circular aluminum tubes before actual
destructive tests are to be realized.

References
[1] Abramowicz W. Effective crushing distance in axially compressed thin-walled metal columns. Int J
Impact Eng 1983;1:309–17.
1136 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137

[2] Abramowicz W, Jones N. Dynamic axial crashing of circular tubes. Int J Impact Eng 1984;2:
263–281.
[3] Abramowicz W, Jones N. Dynamic progressive buckling of circular and square tubes. Int J Impact
Eng 1986;4:234–70.
[4] Abramowicz W, Jones N. Transition from initial global bending to progressive buckling of tubes
loaded statically and dynamically. Int J Impact Eng 1997;9:415–37.
[5] Alexander JM. An approximate analysis of the collapse of thin cylindrical shells under axial load-
ing. Q J Mech Appl Maths 1960;13:1–9.
[6] Andrews KRF, England GL, Ohani E. Classification of the axial collapse of cylindrical tubes under
quasi-static loading. Int J Mech Sci 1983;25:687–96.
[7] Belytschko T, Lin JI, Tsay C-S. Explicit algorithms for the non linear dynamics of shells. Comp
Methods Appl Mech Eng 1984;43:251.
[8] Gupta NK, Velmurugan R. Consideration of internal folding and non-symmetric fold formation
axisymmetric axial collapse round tubes. Int J Solid Struct 1997;34:2611–30.
[9] Guillow SR, Lu G, Grzebieta RH. Quasi static axial compression of thin-walled circular aluminum
tubes. Int J Mech Sci 2001;43:2103–23.
[10] Gupta NK, Gupta SK. Effect of annealing, size and cutouts on axial collapse behavior of circular
tubes. Int J Mech Sci 1993;35:519–613.
[11] Hibbit, Karlsson and Sorensen Inc. ABAQUS 6.1 theory and user’s manual. Providence: Hibbit
Karlsson and Sornesen Inc., 2000.
[12] Johnson RG, Cook WH. A constitutive model and data for metals subjected to large strains, high
strain rates and high temperature. Mech Eng 1983.
[13] Johnson W, Mamalis AG. Crashworthiness of vehicles. London: Mech. Eng. Publications Ltd;
1978.
[14] Johnson W, Reid SR. Metallic energy dissipating systems. Appl Mech Rev 1978;31(3):277–88.
[15] Jones N, Abramowicz W. Static and dynamic axial crushing of circular and square tubes. In: Reid
SR, editor. Metal forming and impact mechanics. Oxford: Pergamon Press; 1985, p. 225.
[16] Jones N. Structural impact. Cambridge: Cambridge University Press; 1989, p. 575.
[17] Karagiozova D, Jones N. Dynamic elastic-plastic buckling phenomena in a rod due to axial impact.
Int J Impact Eng 1996;18(7–8):919–47.
[18] Karagiozova D, Jones N. Stress wave effects on the dynamic axial buckling of cylindrical shells
under impact. In: Jones N, et al., editors. Proceedings of the Fifth International Conference on
Structures under Shock and Impact, Thessaloniki, Greece. UK and Boston, USA: Computational
Mechanics Publications; 1998, p. 201–11.
[19] Karagiozova D, Jones N. Dynamic elastic-plastic buckling of circular cylindrical shells under axial
impact. Int J Solids Struct 2000;37(14):2005–34.
[20] Langseth M, Hopperstad OS. Static and dynamic crushing of square thin-walled aluminum extru-
sions. Int J Impact Eng 1996;18(7–8):949–68.
[21] Mamalis AG, Johnson W. The quasi-static crumpling of thin walled circular cylinders and frusta
under axial compression. Int J Mech Sci 1983;25:713–32.
[22] RADIOSS CRASH, Guidelines, MECALOG, 06903 Sophia Antipolis Cedex, version 4.1, 1997.
[23] McGregor IJ, Meadows DJ, Scott CE, Seeds AD. In: Jones N, Wierzbicki T, editors. Structural
crashworthiness and failure impact performance of aluminum structure. London: Elsevier Applied
Science; 1993 [Chapter 10].
[24] Pugsley AG, Macaulay MR. Large scale crumpling of thin cylindrical columns. Q J Mech Appl
Maths 1960;13:1–9.
[25] Singac AA, Elsobky H. Further experimental investigation on the eccentricity factor in the pro-
gressive crushing of tubes. Int J Solids Struct 1996;32:3589–602.
[26] Su XY, Yu TX, Reid SR. Inertia-sensitive impact energy-absorbing structures. Part I: effects of
inertia and elasticity. Int J Impact Eng 1995;16:651–72.
[27] Tam LL, Calladine CR. Inertia and strain-rate effects in a simple plate-structure under impact load-
ing. Int J Impact Eng 1991;11(3):349–77.
 D. Al Galib, A. Limam / Thin-Walled Structures 42 (2004) 1103–1137 1137

[28] Wierzbicki T, Bhat SU. A moving hinge solution for axisymmetric crushing of tubes. Int J Mech
Sci 1986;28(3):135–51.
[29] Wierzbicki T, Abramowicz W. The mechanics of deep plastic collapse of thin walled structures. In:
Jones N, Wierzbicki T, editors. Structural failure. New York: Wiley; 1989.
[30] Wierzbicki T, Bhat T, Abramowicz W, Brodkin D. Alexander revisited—a two folding elements
model of progressive crushing of tubes. Int J Solids Struct 1992;29(24):3269–88.
[31] Wierzbicki T, Schneider F. Energy equivalent flow stress in crash calculations. Report No:15,
Impact and Crashworthiness Laboratory, Massachusetts Institute of Technology, 1999.
[32] Pugsley AG. On the crumpling of thin tubular structures. Quart J Mechs and Applied Maths
1979;32:1–7.
[33] Wierzbicki T, Abramowicz W. On the crushing mechanics of thin-walled structures. J of Applied
Mechanics 1983;50:727–34.