Thin-Walled Structures 90 (2015) 191–201

Contents lists available at ScienceDirect

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

Minimum mass design of thin tubular structures under eccentric

compressive loading
A. Rashedi a,1, I. Sridhar a,n, K.J. Tseng b, N. Srikanth c
a
School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
b
School of Electrical and Electronic Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
c
Energy Research Institute @ NTU (ERI@N), 1 CleanTech Loop, #06-04 CleanTech One, Singapore 637141, Singapore

art ic l e i nf o a b s t r a c t

Article history: A minimum mass design study applicable to thin circular tube is performed for various modes of
Received 7 July 2014 eccentric compressive loading. Axial crushing failure mode, frequently noticeable in uniform axial
Received in revised form compressive loading of thin circular tube, does not appear in eccentric compression. Hence, other
3 December 2014
compressive failure modes, e.g., global buckling, yield and local buckling are studied with respect to non-
Accepted 7 January 2015
dimensional load and geometric shape factors for a fixed-free condition. These modes are predominant
Available online 12 February 2015
in ductile engineering alloys. A failure mode map in terms of non-dimensional load and shape-factor for
Keywords: a given load-eccentricity are obtained and the prescription for minimum mass is given.
Thin circular tube & 2015 Elsevier Ltd. All rights reserved.
Eccentric loading
Material efficiency
Shape factor
Optimal design
Failure mode

1. Introduction structural material families, such as, for composite materials,

Material efficiency attracted enhanced interest of researchers in
recent times. A low material design ensures reduction in material
cost, embodied energy, green house gas emission and life cycle
impacts. Moreover, it serves the national interests, such as, less
extraction of raw materials, lesser dependence on imports and
increased self-reliance. Hence, scientists have explored the ways on
how to make a design with less material for long time. In this
perspective, shape efficiency based design plays a prominent role.
Higher shape factor of a structure indicates its enhanced shape
efficiency in terms of material consumption in comparison to a low
shape factor based structure under same magnitude of load.
Generally, the more slender the shape, the larger the weight
savings; but there is a limit; making a product too thin and it will
buckle—so there is a maximum shape factor for each material that
depends on its mechanical properties. Shape efficiency based
design is additionally dictated by manufacturing constraints. Many
different shapes can be produced out of metals; but for some other
n
Correspondence to: Department of Engineering, University of Cambridge,
Trumpington Street, Cambridge CB2 1PZ, United Kingdom
E-mail addresses: amma0002@ntu.edu.sg (A. Rashedi),
msridhar@ntu.edu.sg (I. Sridhar).
1
Department of Engineering, University of Cambridge, Trumpington Street,
Cambridge CB2 1PZ, United Kingdom.
manufacturing constraints play a wider role due to complexity in
their processing and manufacturing routes. Thus, composite mate-
rials cannot be designed with as high shape factor as metals.
Individual metal–matrix, polymer composite or other material also
has their own distinct maximum attainable shape factor which
ultimately determines how thin a section can be designed out of
that particular material. For example, cast iron, austenitic, BS grade
F1 material has an elastic bending shape factor of 24 which
indicates the most efficient shape made of this alloy can attain
maximum 24 times resistance to elastic bending (stiffness) deflec-
tion in comparison to a solid beam of circular cross-section of the
same area [1]. Following the conservative design guideline, a
section made with this alloy may not employ this maximum
possible shape factor of 24; but it is certain that material consump-
tion and, hence, the cost and companion life cycle impacts can be
greatly reduced with an increased shape factor while maintaining
the same stiffness as the beam of a square cross-section. Thus
increased shape factor can intensely contribute to increased mate-
rial efficiency. Hence, designers have focused on product designs
with various efficient shapes; e.g., tubular, box, I-section, etc.
Few research groups also focused on shape efficiency based non-
dimensional failure plots under various modes of failure. Weaver
and Ashby [2] demonstrated competitive failure plots based on
global buckling, yield and local buckling compressive failure modes
in pursuit of optimal shape efficiency based design. In addition to

http://dx.doi.org/10.1016/j.tws.2015.01.006
0263-8231/& 2015 Elsevier Ltd. All rights reserved.
192 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201
Nomenclature
ρ density of material
σ failure strength of thin shell structure
σb maximum compressive stress due to bending moment
σc uniform compressive stress due to axial load
σ gb global buckling failure stress
σ lb local buckling failure stress
σy yield failure stress
σ ys yield strength of the material
ψ solidity ratio
υ Poisson ratio
α thin shell strength reduction factor
ϕeB ; ϕ elastic bending shape factor; also indicates ratio
between radius and thickness
ϕopt optimal shape factor
γ thin shell safety factor
c distance from bending neutral axis to outer surface
e eccentricity
e eccentricity ratio
these three failures, Meidell [3] focused on another failure mode,
i.e., axial crushing based compressive failure in his manifestation of
shape efficiency based design. The later article particularly focused
on detailed shape efficiency and optimal mass based design
procedure as varies with load and shape factor based on aluminum.
In another article, Rothwell [4] studied how the dimensions of an
offshore tubular member can be changed in pursuit of wave loading
reduction based on shape efficiency and buckling load constraints.
All these above-mentioned studies, however, only focused on uni-
form compressive loading; whereas, in real life, many engineering
structures and products frequently encounter various modes of
eccentric compressive loading rather than uniform compressive
one. Wind turbine tower is a typical example of such scenario
whereby nacelle weight remains at an eccentricity to the tower
neutral axis and this load eccentricity shifts the competitive failure
boundary regions in a typical failure plot which ultimately affects
optimal dimension based design issues. Additionally, various com-
pressive failure modes also show different patterns under eccentric
loading with reference to uniform compressive loading scenario.
However, to the best of author’s knowledge, no published literature
so far focused on any failure mode map with all eccentric com-
pressive loading together. Hence, an eccentric compression based
failure plot is necessary to find the optimal shape efficiency based
designs for similarly loaded thin shell structures. This article,
accordingly, focuses on various leading eccentric compressive fail-
ure modes to determine the shape efficiency based failure plots
based on non-dimensional factors.
2. Competing failure modes in eccentric compressive loading
Thin-walled tubular structures under compressive loading can
fail by any of the following mechanisms based on the material
properties and geometry: (i) onset of plasticity or yield failure,
(ii) global buckling, (iii) local buckling, (iv) axial crushing,
(v) inversion of shell, (vi) splitting of shell, etc [5]. Inversion of shell
or invertube failure and tube splitting occur only when there is a die
shape arrangement inside the hollow section of a thin shell structure
which is not the case under study. Hence, plastic flow of material,
global buckling, local buckling and axial crushing are the remaining
dominant compressive failure modes. Out of these, plastic failure
occurs when the structure can no longer bear any higher loading
after its critical yield strength property. Global buckling failure
m column buckling mode parameter with expression of:
m2 ¼ P=EI
mgb mass with respect to global buckling equation
mlb mass with respect to local buckling equation
mmin minimal mass
my mass with respect to yield equation
r radius of thin shell
rg radius of gyration
t=2 blade shell thickness
t thickness of thin shell
wmax maximum transverse displacement in global
buckling mode
A cross-sectional area
D diameter of thin shell
E Young’s modulus of material
L length of thin walled structure
Lef f effective column length
P axial load
P avg average crushing load
X 1 ; X 2 ; X 3 eccentricity based factors
occurs when the tubular column length exceeds a critical dimension.
Eccentricity condition does not affect the critical global buckling
load; however, the stress developed in an eccentric condition is
different than non-eccentric uniformly distributed or point load
condition. Other than this, local buckling mode is also affected by
eccentricity. Liu et al. [6] explored various eccentric conditions on
local bucking of steel tubular column and found that ultimate failure
load greatly reduced with an increasing load eccentricity. Apart from
eccentricity effect, local buckling defines a limit on how much thin a
section can be made. Usually, for a given cross-sectional area, a
higher amount of inertia can be obtained with a thin-wall section
which reduces its mass eventually. However, if the shell section is
too thin, local buckling occurs even before overall column buckling
or yielding. Axial crushing is another prominent compressive failure
mode for ductile thin wall structures which differs from usual yield
failure with accompanying distinct mode shapes (e.g., diamond
mode, ring or concertina mode, mixed mode). Since the pioneering
work of Pugsley and Macaulay [7], Alexander [8], number of
researchers and scientists focused on the study of this failure mode
with corresponding energy dissipation and crashworthiness evalua-
tion. Accordingly, axial crushing is seen to occur in circular tube [8],
square tube [9], tubular ring [10], honeycomb cells [11], multi-corner
columns [12], stepped thin-walled tubes [13], corrugated tubes [14]
and many others. Out of these, thin shell structures are commonly
used in energy and vibration absorption systems because of their
stable plastic collapse mechanism and comparatively high energy
absorbing capacity.
Within the failure modes, critical failure loads of global buck-
ling, local buckling and yield failure for a thin shell structure are
available in literature. However, axial crushing for the same is still
to be explored in case of eccentric compressive loading. Hence,
axial crushing failure is studied first to examine whether it is
possible in case of eccentric compressive loading.
2.1. Axial crushing in eccentric compressive loading
2.1.1. Application of eccentric compressive loading
Applying non-uniform loading on a tubular structure based on
a usual universal testing machine is not straightforward. Uniaxial
mechanical characterization testing frame only provides uniform
axial loading for one axis. Hence, adjustments are made by
attaching a three-point bending jig of 100 kN load bearing capacity
to a uniaxial mechanical testing machine. One of the two bottom
 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201
rollers of the bending jig is removed to apply non-eccentric
loading; this bottom roller lays upon a small solid stainless steel
metal piece which behaves like a rigid body due to its high
stiffness properties. The metal piece remains on top of the speci-
men to transfer the loading from the test frame to the specimen. A
schematic diagram of the experimental setup is highlighted
in Fig. 1.
Finite element (FE) simulations are conducted simultaneously
in commercial software ABAQUS to numerically verify whether the
experimental set-up can generate a true non-uniformly distribu-
ted loading or not. For this, a thick solid plate is placed on top of a
solid cylindrical model. Material definitions of the solid plate and
cylindrical model comprise of steel (with elastic modulus of
200 GPa, Poisson’s ratio 0.3) and aluminum (with elastic modulus
of 70 GPa, Poisson’s ratio 0.35), respectively. Thick solid plate acts
like a rigid body on top of the cylindrical model which resembles a
deformable one (i.e., elastic–plastic solid). A surface interaction is
created between the rigid and deformable bodies to ensure in-
between sliding of the two whereas the master surface remains in
the rigid body and slave surface remains in the deformable body.
A pressure loading of 10 MPa is applied upon the steel plate at
various eccentricities. Fig. 2 shows the pressure loading at eccen-
tricity ratio ðe ¼ e=rÞ 0.25 upon the FE model, where e and r
indicate eccentricity of loading axis from neutral axis and radius of
the cylindrical model, respectively. Boundary condition is applied
at the bottom of aluminum cylinder by fixing all translational
movements in vertical (y) and out-of-plane (z) direction. Meshing
is done with 3D hexagonal stress element (C3D8R); it is an 8 node
linear brick, reduced integration element. Total 4456 elements are
used in the simulation.
Various eccentricity ratios are studied based on the FE geometry
and simulation results of e ¼ 0:33; 0:66 are presented in Fig. 3.
The negative stresses in the plots indicate compressive stress varies
from one end to another diametric end of cylindrical model top
section. Based on the simulation runs, it is visualized that the setup
is capable of applying non-uniformly distributed loading.
2.1.2. Specimen preparation
Henceforward, a set of aluminum shell structures are prepared
by shear cutting out of locally available aluminum tube alloy grade
of 6063-T5. The final tube dimensions stand at 124 mm in length,
34 mm outer diameter and 1.05 mm thickness. This resulted in
Fig. 1. Schematic diagram of experimental set-up.
193
40 mm* 40 mm*
10 mm solid, rigid
block
Φ30 mm * 60 mm
deformable cylinder
Fig. 2. FE model of non-uniformly distributed loading.
1.00E+07 ē = 0.66
Non-uniformly distributed compressive
5.00E+06
ē = 0.33
0.00E+00
0 5 10 15 20 25 30 35
stress (MPa)
-5.00E+06
-1.00E+07
-1.50E+07
-2.00E+07
-2.50E+07
Distance from one end of specimen to another diametric end (mm)
Fig. 3. Non-uniformly distributed stress profile across two diametric ends of
specimen top.
almost desired length except some rough edges at end cross-
sections with respect to the specimen axis. To get further accuracy
in end face finishing, tubes are further processed in a Harrison
M300 manual lathe machine. Next, to increase precision of the
specimen in line with ASTM compressive testing standard of
metals (ASTM E9-09), 124 mm long pieces are shortened further
to almost 120 mm by reducing 2 mm from each end. This ensures
perpendicular end face with respect to specimen axis which is
important in axial crushing study. The stainless steel metal piece is
simultaneously prepared by subtractive machining using LAGUN
FU-1250 manual milling machine. Steel metal piece acts as a rigid
body and converts the line load of bending jig into a distributed
load. The bottom roller of bending jig can be translated up to few
centimeters. By translating the roller, necessary offset (i.e. distance
between centre of the specimen and loading point) is maintained
to apply both uniformly distributed and non-uniformly distributed
loading. Thus the bending jig converts the Instron test-frame’s
uniform loading into an eccentric, concentrated line loading and,
afterwards, the rigid body in contact to the bending jig converts
the same into a non-uniformly distributed loading.
2.1.3. Experimental procedure and data analysis
After machining stage, axial crushing testing is initially performed
for a uniform load case on the uniaxial test machine. Here, the
objective is to study the axial crushing failure pattern under uniform
loading with the modified arrangement. Loading was applied under
displacement control at a rate of 2 mm/min. The test machine
(Instron 5569) has an accuracy within 2% of the indicated load.
Load–displacement data is recorded using Instron test-frame’s soft-
ware and observed failure mode is shown in Fig. 4.
194 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201

The highlighted failure mode represents a true axial crushing arrangement. Henceforward, non-uniformly distributed loading is
failure (non-axysymmetric or diamond mode) for aluminum alloy applied from the same set-up by moving the loading roller to off-
specimen, as observed in literature [15]. This ensures the load- axis, with eccentricity (e) of 2 mm and 8 mm.
frame’s capability in applying the uniform loading under modified

Fig. 4. Axial crushing of Al 6063-T5 specimen under uniformly distributed loading with distinct non-axysymmetric mode shape—(a) top view; (b) front or side view.

Fig. 5. Top view of failed samples—with eccentricity of 2 mm (a and b), (b) with eccentricity of 8 mm (c and d).
 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201 195

Fig. 6. Front view of failed samples—with eccentricity of 2 mm (a and b), (b) with eccentricity of 8 mm (c and d).

Corresponding failure modes with varying eccentricities are
shown in Figs. 5 and 6 which resemble partial end folding in highly
loaded sections of all tubular specimens. Specimens are tested with
2 mm/min displacement loading rate and all the specimens showed
similar mode of failure. Associated load versus displacement curves
with 2 mm and 8 mm eccentricity are presented in Fig. 7 which are
linear up to a peak load (point A), then load decreases with
increasing deformation until point B; then due to interaction effect
between deformed zones, load again increases until point C. 2 mm
and 8 mm eccentricity based plots represent eccentricity ratios
ðe ¼ e=rÞ of 11.76% and 47.06%, respectively.
It is evident from failure modes that these do not represent any
axysymmteric (ring) or non-axysymmteric (diamond) axial crush-
ing failure. Hence, failure profiles are further analyzed by compar-
ing the yield strength of the material with experimental failure
stress associated with peak load of Fig. 7. The critical axial
196 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201

crushing failure load for a thin shell structure can be represented,
empirically, as [15]:
P crush ¼ 4π rt σ ys  ψ 0:7
where σ ys indicates the yield strength of the material, t indicates


thickness of specimen and ψ indicates solidity ratio with ψ ¼ 2t=r .
On the other hand, based on combined bending and compressive
failure condition of tubular specimen, the critical failure stress and
failure load can be represented as
P Mc 2π r 2 t σ ys
σ ys ¼ þ )P¼
A I r þ 2e
The yield strength ðσ ys Þ of the material is measured by uniaxial
tensile testing of aluminum specimens of same grade material.
Curved tensile test specimens of 50 mm gage length (extensometer
gage length), 12.8 mm width and 1.05 mm thickness are machined
using electrical discharge machining (EDM) from 300 mm long
tubular pieces according to ASTM E8/E8M standard [16]. Load
control rate of 1.15 MPa/s is applied in Instron 5569 machine and
an averaging type extensometer is used for strain measurement
according to aforesaid standard. The test machine has an accuracy
within 2% of indicate load. The failed samples are shown in Fig. 8
and associated stress versus strain responses are plotted in Fig. 9
which show typical elastic–plastic behavior.
The stress–strain diagrams indicate average yield strength value
ð1Þ
of around 190 MPa. Comparing this strength value with the stress
expression of Eqs. (1) and (2) shows that combined bending and
compressive failure occurs in the specimens rather than any axial
crushing failure. This also indicates no axial crushing failure can
occur in case of non-uniformly distributed loading.
2.2. Failure mode map under eccentric compressive loading
ð2Þ
Henceforward, the study concentrates on developing a failure map
with respect to non-dimensional load and shape factors based on
remaining eccentric compressive failure modes. A companion flow-
chart is presented in Fig. 10 outlining the corresponding analysis
method.
2.2.1. Failure stresses for eccentric compressive condition
A thin shell structure in axial compression is hereby considered
to fail either by column buckling (buckling in a wavelength related
to overall length) mode or by local or shell buckling (buckling in a
wavelength related to small section of the overall column) mode
or by onset of plastic failure (influenced by the yield strength of

Specimen with 2 mm eccentricity

Specimen with 8 mm eccentricity Specimen 1 Specimen 2
250
20 A
18
Compressive load kN)

16 200
Tensile stress (MPa)

14
12 A 150
10
8 C 100
6 B
4
B C
50
2
0
0 1 2 3 4 5 6 7 8 0
Compressive displacement (mm) 0 2 4 6 8 10 12 14
Tensile strain (%)
Fig. 7. Load–displacement plot from non-uniformly distributed compressive test-
ing of shell structures. Fig. 9. Stress–strain plot from tensile test experiment for Al 6063-T5 specimen.

Fig. 8. Yield strength measurement of shell specimen—(a) Setup with averaging extensometer of 50 mm gage length, (b) initial and failed condition of Al 6063-T5 sample.
 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201 197

Individual failure stress [Eqs. collapse mode failure. Local buckling, the third failure mode, is
(5)-(6), (12)] mostly affected by shape imperfection issues and this can be
expressed as [18]
α E 1
Boundary line equation σ lb ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð7Þ
between different failure modes γ 31  υ2  ϕ
[Eqs. (13)-(15)]
where E represents material’s Young’s modulus, υ represents
Poisson ratio, α indicates shell strength reduction factor and γ
Determination of which failure represents additional safety factor. Shell strength reduction factor
modes will dominate to the left of ðαÞ considers the effect of buckling load reduction due to impend-
each boundary segment [Eq. (20)] ing imperfection of support structure upon the foundation. Due to
post-buckling behavior, cylindrical shells are also subjected to
meridional compressive stresses; this is covered by the additional
Corresponding load and shape
safety factor of γ ¼ 1:33. In the presence of bending and buckling,
factor equations
the shell strength reduction factor ðαÞ can be calculated as [18]
[Eq. (33)]
αc σ c þ αb σ b
α¼ ð8Þ
σc þ σb
Mass expression derived from
shape factor [Eq. (37)] where σ c ¼ P=A ¼ P=π Dt; σ b ¼ Mc=I ¼ 4Pe=π D2 t and compressive
reduction factor, αc ; and bending reduction factor, αb ; indicate
Fig. 10. Flow chart of failure mode map and mass minimal design under eccentric uniform compressive stress due to axial design load and maximum
loading.
compressive stress due to design bending moment, respectively,
with the following expressions [18]:
the material rather than any axial crushing mode). Out of this,
0:83
column buckling mode is greatly influenced by end condition αc ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð9Þ
1 þ 0:01ϕ
which can be expressed in terms of effective length. Column
straightness also plays a significant role in global buckling. Max- αb ¼ 0:1887 þ0:8113αc ð10Þ
imum stress due to eccentric global bucking develops at the
outermost surface of the column which can be represented as [17], For a ϕ ratio of 15–40; which represents a typical thin shell
structure, compressive reduction factor αc varies from 0.63 to 0.73.
P M max c P Pðwmax þ eÞc P Pðwmax þ eÞr
σ gb ¼ þ ¼ þ ¼ þ ð3Þ An αc value of 0.68 is assumed hereby for next steps:
A I A I A Ar 2g
0:68PD þ 2:96Pe 1 þ 2:18e=r 1 þ 2:18e
where axial load, P, is applied at an eccentricity, e; from the neutral
α¼ ¼ 0:68 ¼ 0:68 ð11Þ
PD þ 4Pe 1 þ 2 e=r 1 þ 2e
axis of a simply-supported column of length, L: Normal stress in the
Based on different input values of P; D or e; the value of shell
column develops from both of axial load, P, and bending moment,
strength reduction factor ðαÞ can be approximated as 0.7. Even-
M: Herein, wmax represents maximum transverse displacement, A is
tually, based on additional reduction and partial safety factors,
the cross-sectional area, I is the moment of inertia of the cross section,
local/shell buckling stress ðσ lb Þ can be expressed as
c is the distance from thep neutral
ffiffiffiffiffiffiffi axis to the concave extreme side of
the column and r g ðr g ¼ I=AÞ represents the radius of gyration.
Maximum transverse displacement of the column occurs at 1 þ 2:18e E 1
σ lb ¼ 0:295 pffiffiffiffiffiffiffiffiffiffiffiffiffi ð12Þ
mid-sectionh of thepcolumn
ffiffiffi pffiffiffiffiffilength. i Substituting this expression of 1 þ 2e 1  υ2 ϕ
wmax ¼ e sec Lef f P =2 EI 1 in Eq. (3), buckling stress can be
2.2.2. Determination of boundary line for different eccentricity
given by
( To represent failure mode chart in a uniform scale, it is
pffiffiffi!) ( pffiffiffi!)
P 2ec L f P P 2e Lef f P necessary to define the boundary zones of each failure mode in
σ gb ¼ 1 þ 2 sec efp ffiffiffiffiffi ¼ 1 þ sec pffiffiffiffiffi ð4Þ
respect to structural and loading parameters. Accordingly, the
A r 2 EI A r 2 EI
boundary zones of three aforementioned failure modes are deter-
Utilizing the trigonometric expression of secant formula, σ gb mined by equating all above-mentioned failure stresses of Eqs. (5),
can be approximated as (6) and (12). Corresponding boundary line equations are then
( !) sffiffiffiffiffiffiffiffiffiffiffiffiffi expressed in terms of two non-dimensional factors: structural
 
P e  
2
P 2e 1 Lef f P PEπϕ
σ gb ¼ 1 þ 1þ ¼ 1 þ4:47 ¼ ð1 þ 4:47eÞ loading coefficient or load factor P=σ ys L2 , and shape factor ϕ .
A r 8 EI A r 4L2ef f Thus, the boundary line equations between global buckling and
ð5Þ yield, global buckling and local buckling, yield and local buckling
can be expressed, respectively, as:
here, ϕ indicates a ratio between radius (r) and thickness (t) of
shell structure; eventually, it matches with the elastic bending P 4 σ ys 1 1 2
¼ ð13Þ
shape factor ðϕB Þ of the structure which signifies the stiffness ratio
e
σ ys L2ef f π E ϕ 1 þ4:47e
between a shaped section and a solid circular section of the same
material, length and cross-sectional area. In another notation, P 4 E 1 1 0:295ð1 þ 2:18eÞ 2
e ¼ e=r expresses eccentricity ratio whose value varies from 0 to ¼ ð14Þ
σ 2
ys Lef f π σ ys 1  υ2 ϕ3 ð1 þ 4:47eÞð1 þ 2eÞ
1 depending on varying eccentricity e.
Next, onset of plasticity can be expressed by material’s yield
E 0:295 1 þ 2:18e
strength ðσ ys Þ as ϕ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ð15Þ
σ ys 1  υ2 1 þ 2e
σ y ¼ σ ys ð6Þ
Eqs. (13)-(15) are applicable for any kind of column irrespective of
Hereby ‘onset of plasticity’ simply refers to the maximum stress its end support condition. Most single-body column structures can,
in the tube reaching the yield strength rather than any other however, be classified as fixed-free which means the effective length
198 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201

  8 9
Lef f will be about twice of the actual length (L). Accordingly, for a < y ¼  2:46  x
> >
=
fixed-free column, the boundary line equations between global buck- y ¼ 1:57  3x j for e ¼ 0:25 ð25Þ
ling and yield, global and local buckling, yield failure and local buckling >
: >
;
x ¼ 2:014
stand, respectively, as:
8 9
P 16 σ ys 1 1 2
16 σ ys 1 < y ¼  2:83  x
> >
=
¼ ¼ X ð16Þ y ¼ 1:21  3x j for e ¼ 0:50 ð26Þ
σ ys L2 π E ϕ 1 þ4:47e π E ϕ 1 >
: >
;
x ¼ 2:02
2
P 16 E 1 1 0:295ð1 þ 2:18eÞ 16 E 1 1 8 9
¼ ¼
< y ¼  3:08  x
X2 > >
σ ys L2 π σ ys 1  υ2 ϕ3 ð1 þ 4:47eÞð1 þ 2eÞ π σ ys 1  υ2 ϕ3 =
y ¼ 0:96  3x j for e ¼ 0:75 ð27Þ
ð17Þ >
: >
;
x ¼ 2:024
E 1 0:295ð1 þ2:18eÞ E 1
ϕ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiX ð18Þ Accordingly, based on Eqs. (24)–(27), failure modes are plotted
σ ys 1  υ2 1 þ 2e σ ys 1  υ2 3 in Fig. 11 for different eccentricity ratios. Each dominant failure
where X 1 ; X 2 , X 3 are functions of eccentricity, given by region is highlighted herein with respect to both load and shape
factors. It appears that with increasing eccentricity, boundary
2 2
1 0:295ð1 þ 2:18eÞ line between global buckling and yield failure shifts towards lo
X1 ¼ ; X2 ¼ and
1 þ4:47e ð1 þ 4:47eÞð1 þ 2eÞ wer load factor values. This indicates the global buckling failure
0:295ð1 þ 2:18eÞ dominating zone decreases and yield failure zone increases with
X3 ¼ ð19Þ
1 þ 2e eccentric loading. The higher value the eccentricity carries, the
narrower the global buckling failure zone turns out. In another
Sequence of yield, global and local buckling failure modes need
perspective, yield failure occurs at lower load factor with increas-
to be established to plot all failure modes in a single chart with
ing eccentricity.
respect to varying shape factors. This is determined by applying
Yield and local buckling failure boundary line simultaneously
ϕ-1 criterion to all stress ratio expressions. ϕ-1 defines
shifts rightwards. This indicates yield failure can now occur at
which failure modes will stay to the right of each boundary
higher shape factor with increasing eccentricity, i.e., more thinner
segment shared by each pair of two individual failure modes, as
tubes can now fail by yield for which local buckling was the earlier
follows. Based on Eqs. (5) and (6),
sffiffiffiffiffiffiffiffiffiffiffiffiffi usual failure mode. Due to shift in global and local buckling
σ gb PEπϕð1 þ4:47eÞ boundary line, global buckling failure zone also decreases and
¼ ; if ϕ-1; σ gb 4 σ y ð20Þ
σy 16L2 σ ys local buckling failure zone consequently increases with higher
eccentricity.
Based on Eqs. (5) and (12),
sffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi 2.3. Optimal column design
σ gb PEπϕð1 þ4:47eÞð1 þ 2eÞ 1  υ2 ϕ
¼ ; if ϕ-1; σ gb 4 σ lb
σ lb 16L2 1 þ 2:18e E 0:295
Mass of a tubular column can be expressed in terms of its
ð21Þ length, cross-sectional area and material density as:
Based on Eqs. (6) and (12), m ¼ AρL ð28Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffi
σy ð1 þ 2eÞ ϕ 1  υ2 where A stands for cross-sectional area and ρ indicates density.
¼ σ ys ; if ϕ-1; σ y 4 σ lb ð22Þ
σ lb 1 þ 2:18e 0:295E Considering A as a free variable, for a given length, L, and load,
P, mass equations can be defined based on individual stress
Eventually, overall failure mode sequence can be established
equations, as:
from Eqs. (20)–(22), as such:
P
σ gb 4 σ y 4 σ lb ð23Þ m ¼ AρL ¼ ρL ð29Þ
σi
where σ i stands for global buckling, yield or local bucking stress. σ i
2.2.3. Failure mode map for a cast iron alloy with varying is replaced with σ gb , σ y and σ lb of Eqs. (5), (6) and (12),
eccentricity respectively, to get each individual mass equation, as follows:
A failure mode plot can be represented based on the sequence pffiffiffi
P 4ρ P L2
of Eq. (23). Accordingly, a high strength cast iron alloy of BS 900/2 mgb ¼ ρL ¼ pffiffiffiffiffiffiffiffiffiffi ð30Þ
grade (cast and normalized) is selected arbitrarily to exhibit a
σ gb Eπϕð1 þ 4:47eÞ
material-specific failure-mode plot. Properties of this cast iron
P ρPL
alloy are collected from commercial material selection database my ¼ ρL ¼ ð31Þ
σy σ ys
Cambridge Engineering Selector (CES) [1]. Two logarithmic scales
are used in x; y axes with the axis notation x ¼ log 10 ϕ and pffiffiffiffiffiffiffiffiffiffiffiffiffi
P ρϕPL 1  υ2 1 þ 2e
y ¼ log 10 P=σ ys L2 . mlb ¼ ρL ¼ ð32Þ
Using different values of eccentricity ratio ðeÞ, the eccentricity
σ lb 0:295 E 1 þ 2:18e
factors X 1 ; X 2; X 3 are calculated from Eq. (19). Simultaneously, where mgb ; my ; mlb indicate mass related to global buckling, yield
boundary segment equations of cast iron, BS 900/2, cast and and local buckling failure, in Eqs. (30)–(32), respectively. It is realized
normalized material are determined by generic boundary line from Eq. (30) that, in case of global buckling, mass is inversely
equations, as represented in Eqs. (16)–(18). Final, cast iron alloy proportional to the square-root of shape factor; whereas for yield
specific boundary line equations can be expressed as: failure, it is independent of shape factor, as evident from Eq. (31);
8 9 and, for local buckling, mass increases proportionately with the
< y ¼  1:81 x
> >
= increase in shape factor, as represented in Eq. (32). Based on
y ¼ 2:2  3x j for e ¼ 0 ð24Þ
>
: >
; Eqs. (16)–(18), corresponding load factor equations can be expressed
x ¼ 2:001 in terms of material property, eccentricity factor and limiting shape
 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201 199

Fig. 11. Dominant failure modes with respective boundary zones in varying eccentricity ratio ðeÞ.

factor, as follows: 10 5
8 16 σ ys 1
< π E ϕX 1 ; if 1 o ϕ r σEys pffiffiffiffiffiffiffiffiffi
1 ffi
2 3
X ð33aÞ
P 1υ
¼ ð33a; bÞ
16 E 1
σ ys L2 : π σys 1  υ2
1
X ;
ϕ3 2
if E pffiffiffiffiffiffiffiffiffi
1 ffi
σ ys 1  υ2 X 3 oϕ ð33bÞ
Load factor (P/σysL2)

10 0
Eq. (33a) indicates boundary line between global buckling and Yield
yield failure based on Eq. (16); this boundary line equation
remains valid until a specific value of shape factor is reached, as
expressed in Eq. (18). For higher shape factor, the load factor
Local
buckling
equation can be expressed by Eq. (33b) which is the boundary line 10 -5
between global buckling and local buckling. Fig. 12 is plotted based
on Eqs. (33a,b) which highlights different failure zones based on Global
different boundary lines. Based on associated load factor expres- buckling
sions of Eqs. (33a,b), the optimum shape factor and minimal mass 10 -10
equations can be defined. For this perspective, limiting load factors 10 0 10 1 10 2 10 3 10 4
are determined for different limiting shape factors of Eqs. (33a,b). Shape factor (φ)
Accordingly, if ϕ ¼ 1, which is a limiting shape factor limit in Eqs.
(33a,b), limiting load factor can be expressed as: Fig. 12. Dominant failure modes with respect to non-dimensional load and shape
factor.
P 16 σ ys
¼ X ð34Þ
σ ys L2 π E 1
If ϕ ¼ σEys pffiffiffiffiffiffiffiffiffi
1 ffi
2 3
X ; which is another limiting shape factor limit in optimum shape factor equations are expressed by inversing the
1υ
load factor expression and limiting load factors of Eqs. (33a,b)-(35),
Eqs. (33a,b), corresponding limiting load factor can be expressed
as follows:
as:
8 σ ys
16σ ys 2 X 1 pffiffiffiffiffiffiffiffiffiffiffiffi2ffi π E X 1 r σ ys L2
if 16 P
P >
> 1;
¼ 1υ ð35Þ >
>
>
< 16 σ 2ys L2 σ ys 2 X pffiffiffiffiffiffiffiffiffiffiffiffiffi
σ ys L 2 π E X3 16 σ ys
ϕ ¼ ϕopt ¼ π E P X 1 ; if 16 X 3 1  υ r σ ys L2 r π E X 1
1 2 P
π E
>
> qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
>
>   pffiffiffiffiffiffiffiffiffiffiffiffiffi
16 σ ys 2 X 1
>
X3 1  υ
Limiting load factor expressions are represented in Eqs. (34) and E L2
π 1  υ2 P X 2 ; if σ L2 r π E
3
: 16 P 2
ys
(35). Optimal shape factor based design is the one which, for a given
material and load factor, uses the least amount of material. The ð36Þ
200 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201

Based on different eccentric factor parameters ðX 1 ; X 2 ; X 3 Þ, 3. Conclusions

optimal shape factor will vary for fixed-free column condition. By
replacing respective optimal shape factors ðϕopt Þ of Eq. (36) into This article highlights failure mode map of tubular column
mass equations of Eqs. (30)–(32), mass minimal equations can be structure under various forms of eccentric compressive loading. To
represented as this objective, it initially evaluates axial crushing based compres-
8 σ ys 2 X 1 pffiffiffiffiffiffiffiffiffiffiffiffiffi sive failure under eccentric compressive loading condition. Several
> ρPL
< σ ys ; if 16 X 3 1  υ r σ ys L2
P
> π E
2
experiential failure points exhibit that plastic flow of material
mmin ¼ 1:72P 2=3 L5=3 ρð1  υ2 Þ 1=6 σ ys 2 X pffiffiffiffiffiffiffiffiffiffiffiffiffi ð37Þ closely matches with yield failure rather than any distinct axial
>
X3 1  υ
>
: 2=3 1=6 ; if σ PL2 r 16 π E
1 2
E X 2 ð1 þ 4:47eÞ ys crushing mode. Thus it is visualized that no ring or diamond mode
failure occurs for non-uniformly distributed loading. The study,
By changing the load factor limits, Eq. (37) can be represented henceforward, presents a compressive failure plot based on major
as: compressive failure modes (global buckling, local bucking and
8 σ ys 2 pffiffiffiffiffiffiffiffiffiffiffiffiffi
> ρPL π X3 yield failure) under varying eccentricity ratio. With increasing
< σ ys ;
> if E 1  υ2 r σ PL2 16 X1
ys
eccentricity, global buckling failure zone decreases and, simulta-
mmin ¼ 1:72P 2=3 L5=3 ρð1  υ2 Þ
1=6 σ ys 2 pffiffiffiffiffiffiffiffiffiffiffiffiffi ð38Þ
>
> P π X3
1  υ2 neously, yield failure zone increases in comparison to non-
: E2=3 X 1=6 ð1 þ 4:47eÞ ; if σ L2 16 X 1 r E
ys
2 eccentric load condition. This means some structures that are
Eq. (38) expresses mass minimal design equations for fixed- understood to fail in a global buckling mode in non-eccentric
free column condition based on global buckling, yield and local condition will eventually fail in yield mode with eccentric com-
buckling failure criteria. By introducing different eccentricity pressive condition. Yield vs. local buckling failure zone also shifts
values and material properties, individual material-specific failure to a higher shape factor; this indicates yield failure can now occur
plots can be demonstrated henceforward. A generic failure plot with comparatively higher radius or lesser thickness of tube for
diagram is accordingly shown in Fig. 12 based on non-dimensional which local buckling was the earlier usual failure mode. Due to
load and shape factor. shift in global and local buckling boundary line, global buckling
As an example, let us consider a thin shell structure made of failure zone also decreases and local buckling failure zone conse-
AISI 9255 normalized steel material for mass minimal case study. quently increases with increased eccentricity.
The properties of this material are taken from CES database [1], A mass minimal tubular column design guideline is proposed
which are as follows: ρ ¼ 7850 kg/m3, σ ys ¼577 MPa, E ¼211 GPa, consequently based on corresponding equations of above-mentioned
υ ¼0.29. For e ¼ 0:25, following mass minimum equation can be failure plots. A mass minimization case-study of a 30 m long tube
deduced from Eq. (38): made of AISI 9255 normalized steel is presented which shows the
8 tube only weighs  790 kg under 100 kN compressive loading
ρPL σ 2 pffiffiffiffiffiffiffiffiffiffiffiffiffi 0:27P
> 1  υ2 r σ L2
< σ ys ;
> if Eys
ys
condition with a shape factor of 43, which is the limiting shape
mmin ¼ 1:55P 2=3 L5=3 ρ 1  υ2 1=6   pffiffiffiffiffiffiffiffiffiffiffiffiffi ð39Þ factor of the particular material. Decreasing shape factor, however,
> ð Þ σ 2
>
: E2=3
; if 0:27P
σ ys L2
r E ys
1  υ2 increases the weight of the designed component.

For a given P ¼100 kN and L ¼30 m,

σ 2 pffiffiffiffiffiffiffiffiffiffiffiffiffi 0:27P Acknowledgements
1  υ2 4
ys
E σ ys L2
A. Rashedi thanks the Energy Research Institute at Nanyang
So from Eq. (39), Technological University (ERI@N) for the financial support in the
mmin ¼ ð1:55P 2=3 L5=3 ρð1  υ2 Þ1=6 =E2=3 Þ ¼ 203:47 kg form of a Research Scholarship. Additional financial support from
Economic Development Board (EDB) of Singapore through the grant
From Eq. (36), number NRF2013EWT-EIRP003-032 is gratefully acknowledged.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   
ϕopt ¼ 3 16=π E=1  υ2 L2 =P X 2 ¼ 604
References
According to CES, the maximum possible shape factor for AISI
9255 steel is 43 [1]. Hence, the shape factor needs to be reduced [1] Cambridge Engineering Selector. Material and process selection tool. Available
from ϕ ¼ 604 to ϕ ¼ 43 which moves the failure of the thin shell to a from 〈http://www.grantadesign.com/〉.
[2] Weaver PM, Ashby MF. Material limits for shape efficiency. Prog Mater Sci
little left in global buckling zone (see Fig. 12). Hence, corresponding
1997;41(1-2):61–128.
optimal mass can be achieved from global bucking zone mass [3] Meidell A. Computer aided material selection for circular tubes designed to
expression, as presented in Eq. (30): resist axial crushing. Thin Walled Struct 2009;47(8-9):962–9.
pffiffiffi [4] Rothwell A. On the estimation of the compressive design stress in circular-
P 4ρ P L2 section tubular steel members. Appl Ocean Res 1982;4(2):120–3.
m¼ ρL ¼ pffiffiffiffiffiffiffiffiffiffi ¼ 789:56 kg
σ gb Eπϕð1 þ 4:47eÞ
[5] Alghamdi AA. Collapsible impact energy absorbers: an overview. Thin Walled
Struct 2001;39:189–213.
[6] Liu D. Behaviour of eccentrically loaded high-strength rectangular concrete-
Above calculations illustrate that the shape factor should filled steel tubular columns. J Constr Steel Res 2006;62(8):839–46.
always be the minimum one between ϕopt and maximum possible [7] Pugsley AG, Macaulay M. The large-scale crumpling of thin cylindrical
shape factor of that particular material. Mass minimal design can columns. Q J Mech Appl Math 1960;13(1):1–9.
[8] Alexander JM. An approximate analysis of the collapse of thin cylindrical shells
be achieved likewise for each individual material. Based on above- under axial loading. Q J Mech Appl Math 1960;13(1):10–5.
mentioned analysis, if a column is designed as a general-purpose [9] Nannucci PR, Marshall NS, Nurick GN. A computational investigation of the
component, it is better to design such that the yield failure progressive buckling of square tubes with geometric imperfections. In:
Proceedings of the third Asia–Pacific conference on shock and impact loads
strength remains the lowest of all three failure stresses; this will
on structures, Singapore; November 24–26, 1999. p. 335–42.
ensure the final failure to be a slowly progressive one rather than a [10] Reid SR, Austin CD, Smith R. Structural impact and crashworthiness. New York:
sudden catastrophic mode. In other cases, specially, if the load Elsevier; 1984.
factor is pre-assigned, the optimal shape factor can be readily [11] Wierzbicki T. Crushing analysis of metal honeycombs. Int J Impact Eng 1983;1
(2):157–74.
achieved based on aforementioned mathematical representations [12] Wierzbicki T, Abramowicz W. On the crushing mechanics of thin-walled
and corresponding competitive failure zone plot. structures. J Appl Mech 1983;50(4):727–34.
 A. Rashedi et al. / Thin-Walled Structures 90 (2015) 191–201 201

[13] Stangl PK, Meguid SA. Experimental and theoretical evaluation of a novel [17] Bulson PS, Allen HG. Background to buckling. London: McGraw Hill; 1980.
shock absorber for an electrically powered vehicle. Int J Impact Eng 1991;11 [18] Technical Committee. Buckling of steel shells European design recommenda-
(1):41–59. tions, Technical Committee 8, structural stability. In: Technical working group
[14] Singace AA, El-Sobky H. Behaviour of axially crushed corrugated tubes. Int J 8.4, stability of shells, 5th ed.; 2008.
Mech Sci 1997;39(3):249–68.
[15] Jones N. Structural impact. Cambridge: Cambridge University Press; 1989.
[16] ASTM. Standard test methods for tension testing of metallic materials. In:
ASTM E8/E8M standard; 2010.