You are on page 1of 16

Experimental and Numerical

Studies on the Progressive


Collapse Behavior of
Tensegrity Systems
B. Shekastehband1,*, K. Abedi2 and N. Dianat3
1Assistant Professor, Civil Engineering Department, Urmia University of Technology, Urmia, Iran.
b.shekastehband@uut.ac.ir
2Professor, Civil Engineering Department, Sahand University of Technology, Tabriz, Iran.
3PhD, Structural Laboratory of Sazehaye Fazaei Iran, Safira Company, Tehran, Iran.

(Submitted on 14/02/2013, Reception of revised paper 03/11/2013, Accepted on 07/01/2014)

ABSTRACT: To better understanding the response of tensegrity systems


subjected to the progressive collapse, the experimental and numerical studies
were carried out on a 3.0¥3.0¥0.7 m prototype. The primary tests on the
constituent elements were repetitive loading and unloading tests on the cable
elements as well as tensile tests on the tubes and buckling tests on the struts
(consisted of tubes and end connections). The experimental tensegrity model
was loaded up to a load level at which buckling of a strut led to successive
buckling of struts causing overall collapse. The tests results are served as a
basis for calibration of numerical models and furthermore for the development
and verification of a new approach to structural design of tensegrity systems.
The most important conclusion was that the collapse behavior of the tensegrity
model is strongly dependant on the fastener bolts failure after buckling of the
struts. Additionally, kinetic energy released during the progressive collapse in
the tensegrity model caused catastrophic results in the supporting systems.

Key Words: tensegrity systems, progressive collapse, fastener bolts failure,


residual stress, cable rupture

recognized that a number of members are critical to


Tensegrity systems are composed of any given set system integrity and serious strength reductions can be
1. INTRODUCTION

of cables connected to a set of struts in which cables produced by loss of any of them. Furthermore, when
connectivity must be able to stabilize the these members are lost suddenly, their forces are shed
configuration [1]. These systems are mainly statically in a dynamic manner into the structure, causing more
and kinematically indeterminate systems [2]. They severe damage [5].
typically contain a large number of members, and Member failure in these systems, which affects a
possess a high degree of statically indeterminacy. The small portion initially, has potential for propagating to
stability analyses performed on these systems have other parts of the system and may ultimately cause
indicated that despite of high redundancy, buckling of overall collapse of the system - so-called “progressive
a strut (or set of struts) or rupture of a cable may cause collapse phenomenon”. Progressive collapse is
a progressive collapse to occur [3–4]. It was the most dangerous form of instability and generally

*Corresponding author e-mail: b.shekastehband@uut.ac.ir

International Journal of Space Structures Vol. 29 No. 1 2014 9


Experimental and Numerical Studies on the Progressive Collapse Behavior of Tensegrity Systems

lasts for a short duration. Therefore, it is impossible to more precisely the actual collapse behavior that could
prevent progressive collapse in a structure once it otherwise be overlooked in the design process and in
occurs. This implies the importance of understanding the finite element modeling. The main aims of the
of the tensegrity systems response during struts present work are as follows:
buckling or cables rupture. In this article, the results • Comparing and adjusting the finite element
of experimental and numerical studies on the model with the experimental results in order to
progressive collapse behavior of a 3.0¥3.0¥0.7 m achieve a reliable and robust model;
prototype under load control are presented and • Evaluating the effects of residual stress of the
discussed. Numerical analyses on this tensegrity buckled struts on the progressive collapse of
model are based on the proposed procedure by Abedi these systems;
[6–7] in the braced dome structures. This procedure • Putting forward some design recommendations
consists of three main steps, namely nonlinear static for preventing the occurrence of progressive
analysis, linear eigenvalue analysis and nonlinear collapse.
dynamic analysis. At the first step, a nonlinear static
collapse analysis is carried out to find the load- 2. EXPERIMENTAL STUDIES
deflection responses of the system using the behaviors
of struts and cables obtained by buckling and tensile The experimental study was conducted on a
2.1. General

tests. If the response of the system is a “local collapse 3.0¥3.0¥0.7 m prototype, assembled from nine half-
with dynamic effects”, the subsequent steps can cuboctahedron modules of 1.0¥1.0¥0.7 m, as
be undertaken. At the second step, a linear eigenvalue illustrated in Fig. 2. Struts and cables of the tensegrity
analysis is carried out to obtain the natural model were fabricated from tube 35¥2 mm and IWRC
frequencies of the system in the strained 6¥26 (8 mm), respectively. This model contains
configuration. By doing so, the appropriate time contiguous strut configuration in which modules are
increment for the nonlinear dynamic analysis as well connected by node to node connection type. The
as suitable damping factors can be predicted. In the model possess irregular layout in which struts are
final step, a nonlinear dynamic analysis is carried out connected in a zig-zag way in plan.
to evaluate the possibility of progressive collapse of
the tensegrity model. The aforementioned procedure
can be summarized in a flowchart as indicated in
Fig. 1. It is worth noting that the numerical analyses
were carried out using ABAQUS [8], a nonlinear
finite element software package. The experimental
study on these systems are used to validate the
proposed numerical procedure and also to ascertain

(a)

(b)
Figure 1. Identification flowchart for occurrence of Figure 2. Tensegrity model with irregular layout composed of
progressive collapse in tensegrity systems. 36 struts (thick lines) and 84 cables (thin lines).
(a). Perspective view (b). Plan view.

10 International Journal of Space Structures Vol. 29 No. 1 2014


B. Shekastehband, K. Abedi and N. Dianat

The studied tensegrity model consists of 40 nodes, 36


2.2. Construction and self-stress design

struts and 84 cables. For all the struts of this model, a


uniform self-stress distribution is chosen except struts
of central module whose coefficients is double of the
others. Based on the numerical analysis, these
coefficients were chosen in a way that the model
shows local collapse due to snap-through buckling of
struts under external loading. The specifications of
cable elements and struts of the studied tensegrity
model are given in Table 1. The comparison between Figure 4. Applying axial forces by tension-meter.
numerical target self-stress state and the actual
measured state on the experimental model in the free-
standing position was shown in Fig. 3. There are
relatively small errors between numerical and
corresponding experimental values [9].
It is worth noting that forces of cable elements are
applied by a tension-meter (see Fig. 4). By installing
this device on a cable element, the amount of tension
force is measured. The adjustable end of the cable is
screwed and the amount of tension force is changed.
The model, made up 9 modules, is assembled on a
platform shown in Fig. 6. Individual modules are built
on a jig as shown in Fig. 5. At First, modules are fixed
at four corners by connecting the lower joints of the

Table 1. Specifications of the tensegrity model


Figure 5. Construction of modules in the jig.
Length Cross
Elements Number (mm) section modules to the platform (see Fig. 6(a)). Then, for
Struts 1 to 36 1319 ∅35¥2 mm building the edges of model, the aforementioned jig
Lower Cables 37 to 60 1000 IWRC are used again (see Fig. 6(b)). Finally, the central
6¥26 (8 mm) module of the model is constructed using the jig (see
Upper Cables 61 to 96 707.1 IWRC Fig. 6(c)). In this stage, all joints are in their position.
6¥26 (8 mm) However, by restraining the upper joints in the jig, the
Bracing Cables 97 to 120 860.2 IWRC accuracy of geometrical position of joints is
6¥26 (8 mm) controlled and if needed, their position is adjusted.
Note that all lower joints of the assembled model are
connected to platform. This platform was fixed to the
strong floor.
The joint systems used in the tensegrity model
were indicated in Fig. 7 which consists of spherical
forged solid steel ball joint, struts and cables. The
struts, consisted of the tubes and the end cones with
bolts and sleeves, are connected to the nodes
by screwing the bolts (see Fig. 8 (a)) [10]. The
diameter of high strength fastener bolts of the struts
was 12 mm. The cable elements consisted of wire
Figure 3. Numerical target and actual measured self-stress rope, right-hand threaded bolt, left-hand threaded
states on the experimental tensegrity model in the free- bolt and right-left hand bolt, were also connected
standing position. to the joints by screwing the end connections

International Journal of Space Structures Vol. 29 No. 1 2014 11


Experimental and Numerical Studies on the Progressive Collapse Behavior of Tensegrity Systems

(a)

(a) (b)

Figure 8. (a). Strut consists of tube, end cone, fastener bolt,


and sleeve [10] (b). Cable element; 1-wire rope, 2-right-hand
threaded bolt, 3-left-hand threaded bolt, 4-right-left hand bolt.

(see Fig. 8(b)). It is worth noting that diameter and


length of the hollow bolts used in the end
connections (parts 2 and 3 in the Fig. 8(b)) were
20 mm and 60 mm, respectively. The ball joints
were drilled so that the angular errors corresponding
to connection were between 1º and 2º.
(b)

Experimental tests on the component parts of the


2.3. Tests on the constituent elements

tensegrity model were carried out to determine their


behavior and influences on the collapse behavior of
the system, as well as to establish the considerations of
member modeling to be used in the finite element
modeling. Three tests were carried out in the different
parts of the model.
• Tensile test of the cable elements;
• Tensile test of the tubes used in the struts;
• Buckling test of the struts.
(c)

The tensile tests on the cable elements were conducted


2.3.1. Tensile test of the cable elements
Figure 6. The sequence of construction of the tensegrity
model. (a). Fixing four corner modules to the platform by means of a universal electro-mechanical testing
(b). Construction of edges of the model on platform using jig
(c). Construction of central module of the model using jig.
machine (see Fig. 9). Fig. 10 shows the axial stress-
axial strain response of the three types of cable
elements. Cable element types 1, 2 and 3 were loaded
and unloaded five times up to tensile force 5 kN, 15 kN
and 25 kN, respectively. After fifth loading cycle, the
elements were loaded up to rupture and stress-strain
data were then recorded. Table 2 gives the summary of
the experimental results for the mechanical properties
of three types of cable elements. It should be realized
here that the module of elasticity depends strongly on
the level of cyclic loading. These tests were performed
under displacement control at room temperature and
the strain rate was 5.0 mm per minute. In the elements,
the tensile rupture position of wire ropes was away
Figure 7. The joint detail. from the end connection (see Fig. 9) and no slippage

12 International Journal of Space Structures Vol. 29 No. 1 2014


B. Shekastehband, K. Abedi and N. Dianat

Table 2. Summary of the mechanical properties of


three types of cable elements used in the tensegrity
model

Elastic
limit
Modulus of stress Failure
elasticity (at 0.2%) resistance
Sample (MPa) (MPa) (MPa) Elongation
Type 1 1.25E+5 1350 1740 2.65%
Type 2 1.37E+5 1380 1670 2.48%
Type 3 1.93E+5 1640 1640 1.32%

mode of the cables between the tensile test and the


tensegrity model would be different.

2.3.2. Tensile test of tubes used in the

The struts of this model were made from tube


struts

352 mm. Therefore, several tubes were tested in


tension to determine their mechanical properties and
stress-strain relationships. Metal plugs were inserted
Figure 9. Tensile test of the cable elements used in the
tensegrity model. far enough into the ends of the specimens to permit the
testing machine jaws to grip the specimens properly
[12] (see Fig. 11). The tests were carried out under
displacement control until failure and the displacement
rate was fixed at 5 mm per minute. Fig. 12 indicates
the typical stress-strain curve of the samples in
tension. According to these tests, the average modulus
of elasticity is 204 GPa, the average yield stress is
345 MPa, the average ultimate stress is 490 MPa and
the average elongation is 10%. This material property
was used in the finite element modeling of individual
struts in order to obtain their axial load- axial
displacement responses.

The pre-buckling and post-buckling behaviors of


2.3.3. Buckling test of struts

struts, used in the studied tensegrity model, were


obtained through performing compression tests on five
Figure 10. Stress-strain curve of three types of cable specimens with the same geometrical specifications.
elements used in the tensegrity model. In order to account for the rigidity effect of ball-bolt
joint system on the buckling strength and post-
buckling behavior of struts, the specimens were
occurred in the wires in the end connection satisfying connected to the supports through ball joint systems
the requirements of ASTM [11]. It is worth noting that (Fig. 13). It should be noted that the ball joints in the
the support conditions in the tensile test and in the physical model are restrained by the elements
physical model are not the same. The cable attachment connected to the joints. Therefore, it seems that the
in the model does not allow the cable to orientate itself support conditions in the buckling test and in
correctly. Therefore, it is expected that the failure the physical model illustrate the non-confirmation of

International Journal of Space Structures Vol. 29 No. 1 2014 13


Experimental and Numerical Studies on the Progressive Collapse Behavior of Tensegrity Systems

Figure 11. Sample tube 35 ¥ 2 mm under tension test.

Figure 12. Typical experimental stress-strain behavior of tubes


35 ¥ 2mm used in the fabrication of struts of the studied
tensegrity model.

the pinned-end connections. For the fastener bolts of


the struts, high strength steel (of grade 10.9) was used.
Fastener bolts of the struts did not fail in the buckling
test of struts. In addition, plastic deformation did not
occur in the sleeves of the buckled struts. The results
show good repeatability for different samples.
Buckling forces of struts are given in Table 3. The end
conditions of the tested struts do not confirm the
pinned-end connections. Buckling forces of struts
obtained from finite element method, for different
effective length factors considering initial geometrical
imperfection with maximum amplitude of e = 0.001L
at the mid-length was given in Table 4. It should be
noted that there are many sources of imperfections in
the struts of tensegrity systems, such as an initial
curvature of the members, unavoidable end
eccentricity in axial load application and residual Figure 13. Buckling test of struts.
stresses in the section of members. Modeling of all
these imperfections imposes extra cost on the analyses with the bowing of the struts is considered in the
of these systems. Also, geometric modeling of these extracting individual struts’ behavior.
imperfections is so complicated. Therefore, simplified Considering Tables 3 and 4, the average buckling
imperfection containing initial curvature associated force of struts tested is approximately equal to

14 International Journal of Space Structures Vol. 29 No. 1 2014


B. Shekastehband, K. Abedi and N. Dianat

Table 3. Experimental buckling force and stress of imperfection amplitude of e = 0.001L with corres-
tested struts ponding experimental responses. In the finite element
modeling of struts, to obtain their axial load- axial
Buckling displacement response, it is assumed the effective
Buckling force stress length factor to be k = 0.8.
Sample (kN) (MPa)
1 45.956 222.01 The positions of the supports for the tensegrity model
2.4. Loading system and test frame

2 40.777 196.99 were illustrated in Fig. 15. All eight supports were
3 38.467 185.831 constrained against translation at three principal axes.
4 41.976 202.783 The experimental investigation was carried out under
5 40.738 196.802 load control, and consequently, there was no control
Average 41.5828 200.883 on the displacements. The load was only applied to the
node 21 of the model. A metal basket coated using
plastic with a capacity of 1250 liters was connected to
the node 21 of the tensegrity model using a frame
Table 4. Numerical buckling forces and stresses of constructed of two vertical and two horizontal solid
struts for different effective length factor steel bars. The basket, connected to the model, was
loaded by carefully pouring sand into the tank, at a rate
Effective Buckling Buckling of approximately 0.1 kN per minute. A load cell with
length force stress an operating capacity of 100 kN was used to measure
Tube Imperfection factor (kN) (MPa) the values of the loads applied to the tensegrity model.
35¥2 mm e = 0.001L k = 1.0 29.511 142.40 This load cell was located on the top of node 21.
k = 0.9 35.185 169.78 The test frame, consisting of four main columns,
k = 0.8 41.174 198.68 four auxiliary columns and two supporting beams, was
k = 0.7 49.31 237.94 used for the experimental investigation. The main
columns were then braced laterally using bracing
arms. The ball joints at support positions were
connected to base plates through bolts. These base
plates were in turn connected to supporting beams
resting on auxiliary columns. The schematic and actual
views of the loading systems and test frame of the
experimental model were shown in Figs. 16 (a) and
(b), respectively.

Figure 14. Comparison between experimental axial load-axial


used in the tensegrity model displacement responses and
corresponding numerical response assuming effective length
factor of k = 0.8 and imperfection amplitude of e = 0.001L for
the struts.

corresponding numerical value with effective length


factor of k = 0.8. Fig. 14 illustrates comparison of
numerical axial load-axial displacements response of Figure 15. Boundary conditions and position of applied load
struts assuming effective length factor of k = 0.8 and on the tensegrity model.

International Journal of Space Structures Vol. 29 No. 1 2014 15


Experimental and Numerical Studies on the Progressive Collapse Behavior of Tensegrity Systems

(a)

Figure 17. Position of LVDT transducers and strain gauges in


the tensegrity model.

(b)
Figure 16. Test frame and loading system of the tensegrity
signals emanating from the strain-gauges, LVDT
model. (a) Schematic view (b) Actual view.
transducers and the load cell were initialized to zero.
At first, the loading basket was loaded with sand at an
approximate rate of 0.1 kN per minute. However, by
reaching the critical load level, the rate of loading was
decreased to 0.05 kN per minute. The test was finished
2.5. Displacement and strain

The displacements and strains were measured using a when after buckling of strut 16, the progressive
measurement

data-logger with 20 measurements being made in every collapse occurred in the model.
second. The vertical displacements of 5 nodes of the
tensegrity model were measured using five linear
variable differential transformer (LVDT) transducers In the finite element modeling of the studied model,
2.7. Results

with working range of 10 mm, 25 mm, 100 mm, 100 mm axial forces corresponding to the actual measured
and 400 mm, respectively. These transducers were held self-stress state was taken into account for the self-
by a frame constructed from steel angle section bars and stress analysis. Also, axial stress- axial strain
connected to the main columns of test frame. Precision response of the struts was obtained using Arc-
electrical resistance strain gauges were used to measure Length method considering the effective length
axial strains of twelve struts in the model. All of the factor k = 0.8 and material behavior of tube
strain gauges were fixed at the mid-length of the struts. according to the tensile test result. This stress-strain
The position of LVDT transducers and strain gauges response was used as the constitutive relationship of
was illustrated in Fig. 17. struts in the finite element modeling of the studied
tensegrity model.
The self-stress level was chosen in such a way that
Before the main test was undertaken, the model was axial forces induced in the struts to be approximately
2.6. Test procedure

loaded several times up to 3 kN loads and then 50% of their critical load [13]. Table 5 gives the
unloaded to check the operation of the test equipment. percentage of axial forces induced due to self-stressing
In addition, the reading from the strain gauges, load in the struts of the model considering two different
cell and LVDT transducers were also checked to initial imperfections e = 0.001L and e = 0.002L for the
ensure that the model was deforming linearly. After struts. With this level of self-stress, firstly, adequate
checking the operation of the test equipment, the margin against premature collapse of the model was

16 International Journal of Space Structures Vol. 29 No. 1 2014


B. Shekastehband, K. Abedi and N. Dianat

Table 5. The percentage of axial forces induced


due to self-stressing in the struts of the
tensegrity model

Percentage Percentage
axial axial
force force
induced induced
Effective Initial in the struts in the struts
length imperfection of central of other
factor value module module
k = 0.8 e = 0.001L 47% 23.50%
e = 0.002L 53% 26.5%

Figure 18. Three types of cable elements used in the


tensegrity model.
achieved. Secondly, it was expected that the model
would demonstrate partial progressive collapse.
Finally, the dynamic jump would be occurred below
the load of 30 kN. It is worth noting that special type
of collapse in which buckling of a strut at the snap load
level causes the buckling of the other struts, is called
“partial progressive collapse”. In this case, the system
experiences very large deflection at the end of
dynamic jump during snap-through and continue to
carry load after damping out at the end of snap-
through.
During the assembly process of the experimental
model, because of the problems encountered during (a)
the erection process, it has to be disassembled. Then
the model was assembled again and target self-stress
level was applied through cable elements. This
repeated process led to changes in modulus of
elasticity of the cables. Therefore, in the finite
element modeling, cable elements were divided into
three types (see Fig. 18). The stress-strain responses
of these three types of cables were extracted by
means of tensile test on cable elements as illustrated
in Fig. 10.
According to the numerical analysis, it was
expected that irrespective of effective length factor of (b)
struts, the studied model demonstrate partial
progressive collapse. Fig. 19 illustrates that in Figure 19. Numerical load-deflection responses of the
tensegrity model for different effective length factor
tensegrity model for different effective length factors considering imperfections of e = 0.001L and 0.002L. (a)
of struts considering initial imperfections of e = Responses for imperfection of e = 0.001L (b) Responses for
0.001L and e = 0.002L, partial progressive collapse imperfection of e = 0.002L
was occurred. However, in the collapse test conducted
on the model, it was observed that the buckling of the fastener bolts were failed suddenly. It is worth
strut 16 occurring at load level of PCr = 18838 N noting that before buckling of strut 16, it was
caused the overall collapse of the structure. In strut 16 observed that the cables 66, 79, 83 and 92 were
and other buckled struts during progressive collapse, slackened.

International Journal of Space Structures Vol. 29 No. 1 2014 17


Experimental and Numerical Studies on the Progressive Collapse Behavior of Tensegrity Systems

k = 0.8 and initial imperfection of e = 0.002L for struts.


The following points can be concluded:
2.7.1. Before the occurrence of

Fig. 20 shows the numerical static load-displacement • Before buckling of struts 16, cables 92, 83, 79 and
snap-through

responses of the studied model. These responses have 66 were slackened at the load levels of 10378 N,
been obtained assuming the effective length factor of 11642 N, 16012 N and 17706 N, respectively.
k = 0.8 and considering four different initial • At the load level of PS1 = 18953 N, the model
imperfections of e = 0.001L, e = 0.002L, e = 0.003L experiences coupled nodal and member snap due
and e = 0.005L for struts. A comparison between to buckling of strut 16. The load-shedding
numerical snap loads corresponding to different initial characteristic of the strut 16 is so abrupt and during
imperfections with obtained experimental critical load the redistribution process, the other members can
was given in Table 6. not absorb the redistribution loads quickly enough
Looking at Table 6, it is observed that numerical snap and instability occurs in the tensegrity model. The
load considering initial imperfection of e = 0.002L for instability proceeds until the strut 16 reaches its
struts is a very close to the experimental critical load. post-buckling residual strength, whereupon the
Therefore, in the finite element modeling of the studied situation is restabilised and the other members
tensegrity model, it was assumed that effective length then continue to carry load until the strut 31
factor and initial imperfection of struts to be k = 0.8 and e buckles at the load level of Pm = 17496 N. The
= 0.002L, respectively. This is quite logical because the aforementioned process is repeated again until
buckling tests conducted on the individual struts implied buckling of strut 15 at the load level of PS2 =
to effective length factor of k = 0.8. In addition, there are 21997 N. Instability occurs again in the model.
geometrical imperfections in the holes of ball joints which Under displacement controlled loading, the
would be added to the initial imperfection of the struts. deformation of the grid continues to increase as the
Fig. 21 illustrates the schematic representation of applied load decreases. After reaching strut 15 to
the numerical load-deflection response of the its post-buckling residual strength, the situation is
tensegrity model considering effective length factor of restabilised until the cable 55 is ruptured at the

Figure 20. Numerical load- deflection response of the Figure 21. The schematic representation of numerical
tensegrity model for different possible imperfections load-deflection response of the tensegrity model considering
considering effective length factor of k = 0.8. k = 0.8 and e = 0.002L.

Table 6. A comparison between numerical snap load and experimental critical load of the studied model

Initial imperfection Numerical snap Experimental critical


of struts load (N) load (N) Discrepancy (N) Percentage error
e = 0.001L 21679.6 18838
e = 0.002L 18948.7
-2841.6 -15%

e = 0.003L 17135.3 1702.7 9%


-110.7 -0.5%

e = 0.005L 14349.9 4488.1 23.80%

18 International Journal of Space Structures Vol. 29 No. 1 2014


B. Shekastehband, K. Abedi and N. Dianat

load level of Pu = 43229 N and overall collapse is


occurred in the model. It is worth noting that when
strut 16 buckles, the model will be compelled to
seek stable equilibrium state at this load. The move
to such an equilibrium state involves a large
change in deformation or a dynamic jump.
Figs. 22 (a) and (b) show the experimental and
numerical (static procedure) load-deflection responses (a) (b)
of the tensegrity model at the nodes 21 and 36,
Figure 23. Slackening of cables before buckling of strut 16 in
respectively. As the experimental investigation was the model. (a). Slackening of cables 92 and 79 (b). Slackening
undertaken under load control, the unstable part of the of cable 83.
equilibrium path could not be traced. Therefore, the not show the unstable parts of the equilibrium path.
experimental load-deflection responses of the model do However, dashed lines in these figures are schematic
representation of unstable part in the experimental
model. Slackening of the cables 92, 79, 83 and 66 was
also shown on the experimental load-deflection
responses of the model. The experimental and
numerical load-strain responses of the struts 23 and 32
in the tensegrity model are illustrated in Figs. 22(c) and
(d), respectively. Tables 7 and 8 give both the numerical
and experimentally obtained values of the vertical
displacements and axial strains at the critical load level.
(a)

Considering both the numerical and experimental


load-deflection and load- strain responses, and Tables
7 and 8, the following points can be concluded:
• Before the occurrence of the buckling in strut 16,
there is relatively good agreement between the
numerical and experimental responses (approxi-
mately 0.5% for the load, 15.7% for the average
displacements and 25% for the average axial strains
(b)
at the critical load level). This conformity verifies the
finite element modeling of the model.
• As shown in Fig. 23, cables 92, 79, 83 and 66
were slackened before buckling of strut 16. This
is completely consistent with the results of finite
element modeling. Even though due to placement
of camera, it was not possible to record the
slackening of cable 66, but it was confirmed with
(c) view observations.

Thenumericalprocedureoutlinedintheintroduction,has
2.7.2. After the occurrence of snap-through

been used to evaluate the dynamic behavior of the studied


model after the occurrence of the snap-through
phenomenon. The frequency analysis was undertaken to
determinetheappropriatetimeincrement(0.0008sec)for
the nonlinear dynamic analysis as well as suitable
damping factors according to the Rayleigh damping
(d)
Figure 22. Numerical and experimental load-deflection and load- method. The damping ratios were assumed to be 1% and
strain responses of the studied model before the occurrence of
snap-through. (a). Load-deflection responces at node 21
1.5% [14]. A nonlinear dynamic analysis was then
(b). Load-deflection responces at node 36 (c). Load-strain undertaken to follow the behavior of the model after the
responces at strut 23 (d). Load-strain responces at strut 32. dynamic jump, predicted by the nonlinear static analysis.

International Journal of Space Structures Vol. 29 No. 1 2014 19


Experimental and Numerical Studies on the Progressive Collapse Behavior of Tensegrity Systems

Table 7. Experimental and numerical deflections of the nodes 2, 3, 21, 36 and 37


at the critical load level

Procedure Node 2 Node 3 Node 21 Node 36 Node 37


Static Analyses 25.10 5.55 51.17 39.80 29.40
Test 21.33 7.91 47.85 37.34 35.13
Discrepancy 3.77 3.32 2.46
Percentage Error 18% 7% 7%
-2.36 -5.73
-30% -16.30%

Table 8. Experimental and numerical strain of the some struts at the critical load level

Procedure Strut 11 Strut 12 Strut 14 Strut 15 Strut 16 Strut 22 Strut 23


Static Analyses –0.000071 –0.000289 –0.000125 0.000421 –0.000653 0.000028 –0.000207
Test –0.000059 –0.000529 –0.000127 0.000573 –0.000525 0.000030 –0.000312
Discrepancy –0.000012 0.000240 0.000002 -0.000152 –0.000128 –0.000002 0.000105
Percentage Error 20% –45% –2% –27% 24% –6.10% –33.70%

Procedure Strut 29 Strut 30 Strut 31 Strut 32 Strut 36


Static Analyses 0.000363 –0.000044 0.000420 0.000315 0.000230
Test 0.000650 –0.000038 0.000547 0.000278 0.000158
Discrepancy –0.000287 –0.000006 –0.000127 0.000037 0.000072
Percentage Error –44.20% 16.30% 23.20% 13% 46%

Fig. 24 illustrates the numerical time- displacement load-deflection response approaches the corresponding
responses of nodes 2, 21 and 37 during the snap-through. staticresponseandcontinuesalongthatpath.However,in
Taking into account the dynamic effect of snap-through the experimental model, it was observed that local
led to buckling of strut 15 which expected to be stable collapseduetobucklingofstrut16waspropagateoverthe
based on the static analysis. Static and dynamic load- structure and led to overall collapse. This fundamental
deflection responses of the model at the node 21 were discrepancybetweenexperimentalandnumericalmodels
illustrated in Fig. 25. It can be seen that at the end of after occurrence of the snap-through is referred to the
dynamic jump, the further motion is oscillatory in sudden failure of fastener bolts of the strut 16 after
character (solid red line), and eventually will be damped buckling.Failureoffastenerboltscausedthecontribution
out. When the applied load is subsequently increased, the ofthesesstrutstothestiffnessmatrix(K)tobedisappeared

Figure 24. Numerical time- deflection responses of the Figure 25. Comparison between numerical static and dynamic
tensegrity model at the nodes 2, 21 and 37. load-deflection responses of the tensegrity model.

20 International Journal of Space Structures Vol. 29 No. 1 2014


B. Shekastehband, K. Abedi and N. Dianat

and its internal force to be transmitted to the remaining buckling of the struts leading to the overall collapse of
structure immediately. Figs. 26 (a) and (b) illustrate the tensegrity system.
schematicrepresentationofpost-bucklingcurvesofstruts In this case, in which overall collapse has occurred
inthenormalperformanceandinthecaseofsuddenfailure in the system, the static analysis is sufficient and there
of fastener bolts, respectively. In these figures, Pb and Pr is no need to perform dynamic analysis. However, to
represent buckling load and post-buckling residual assess the dynamic effect of sudden failure of fastener
strength, respectively. In the normal performance, the bolts on the collapse behavior of the studied model and
magnitude of post-buckling residual strength is sequence of buckling of struts and rupturing of the
approximately 0.2 of buckling load (Pr = 0.2Pb). cables, a nonlinear dynamic analysis was undertaken.
However,inthecaseofsuddenfailureoffastenerbolts,the Due to the buckling of strut 16 and failure of its
residual strength is vanished immediately (Pr = 0). fastener bolt, a successive buckling of struts and
The ultimate strain corresponding to the failure of rupture of cables were occurred leading to the overall
fastener bolts was measured by means of a strain collapse. The sequence of buckled struts and ruptured
gauge installed on the strut 16. Now in the finite cables during the collapse propagation was illustrated
element modeling of the studied tensegrity model, the in Fig. 28. As indicated in this figure, buckling of nine
stress-strain relation of the struts is considered to be as struts and rupturing of three cables were occurred in
indicated in Fig. 26 (b) in which ultimate strain of strut only 1 second. The numerical analysis implied that due
16 is 0.0040. The static load-deflection response of the to propagation of collapse in the model, a considerable
model was shown for nodes 21 and 37 in Fig. 27. The large tensile force was created in the strut 35.
positions at witch struts 16, 31 and 15 have buckled Therefore, due to weakness of fastener bolts, it is
were also shown in this figure. It is observed that expected that tensile failure to be occurred in these
considering the failure of fastener bolts of the struts in bolts. Figs. 29 and 30 show the configuration of
the finite element modeling caused progressive the tensegrity model before and after the progressive

Figure 26. Schematic representation of post-buckling curves


in the normal performance and in the case of sudden failure
of fastener bolts. Figure 28. The sequence of buckled struts and ruptured cables
during the collapse propagation of the studied model.

Figure 27. Static load-deflection responses of the tensegrity


model at the nodes 21 and 37 corresponding to the sudden Figure 29. Configuration of the model before progressive
failure of the fastener bolts of the struts after buckling. collapse.

International Journal of Space Structures Vol. 29 No. 1 2014 21


Experimental and Numerical Studies on the Progressive Collapse Behavior of Tensegrity Systems

Figure 31. A typical buckled strut and its fastener bolts which
was failed from pin holes.

(a)

(b)
Figure 30. Plan and elevation views of the model after Figure 32. Schematic view of a ball-bolt joint system in
progressive collapse. (a). Plan view (b). Elevation view. which nut system was used instead of sleeve- pin system [13].

collapse, respectively. Dropped struts due to failure of struts in both numerical and experimental models were
fastener bolts were also illustrated in Fig. 30(a). As highlighted. There are some discrepancies between
expected in the numerical analysis, fastener bolt of numerical and experimental models. Some of the
strut 35 due to large tensile force was failed (see Fig. struts expected to buckle in the finite element model;
30(b)). Fig. 31 illustrates a typical strut in which however, they have not buckled in the experimental
fastener bolts were failed from pin holes. In fact, the model and vise versa. In addition, the cables ruptured
induced holes in the bolts caused the effective cross in the numerical and experimental models are
section area of bolts to be decreased, leading to a completely different. These differences can
weakness in this position. It seems that this problem be attributed to the kinetic energy released during the
can be avoided by using other types of connections, in progressive collapse (after buckling of struts 16, 31
which instead of sleeve-pin system, other systems and 15). This kinetic energy was so considerable that
such as nut system were used (see Fig. 32). it resulted in the torsional buckling of the supporting
Table 9 gives the list of buckled struts and ruptured beams (see Fig. 33).
cables during progressive collapse. The same buckled

Table 9. List of buckled struts and ruptured cables during progressive collapse in the studied
tensegrity model

Buckled Struts Experimental Model 2 5 11 13 14 15 16 24 25 29 34 36


Numerical Model 4 10 15 16 17 23 24 31 34 36
Ruptured Connection
Bolts Experimental Model 11 13 14 15 16 29 35a
Ruptured Cable Experimental Model 53 78 79
Elements Numerical Model 39 48 55
aStrut 35 was in Tension

22 International Journal of Space Structures Vol. 29 No. 1 2014


B. Shekastehband, K. Abedi and N. Dianat

make a hole in the fastener bolt. Therefore, with


eliminating this weakness in the fastener bolt, the
sudden failure of the bolts can be avoided.
• The kinetic energy released during the
progressive collapse may lead to catastrophic
events. Therefore, this type of collapse must be
avoided in the systems. For this reason, it is
rational to avoid the occurrence of even partial
progressive collapse. As illustrated in the results,
a defect as failure of fastener bolt may cause
Figure 33. Torsional buckling in the supporting beams of the partial progressive collapse to change to the full
studied tensegrity model. progressive collapses.
• In analysis of tensegrity systems, conventional
design and analysis processes based on the
assumption that the ball-bolt connection behaves
The experimental and numerical studies have been as a pinned joint, are not valid. A ball joint system
3. DISCUSSION AND CONCLUSIONS

carried out to examine the collapse behavior of the with partial bend bearing capacity is a typical
tensegrity systems due to buckling of a strut at a semi-rigid joint system. Due to the dominant
critical load level. In order to obtain material axial behavior of the elements in tensegrity
properties and influences of the constituent elements systems, it is a common method to determine the
on the collapse behavior of the model, some effective length factor of the struts in order to
experimental tests on these elements were also carried obtain their axial force-axial displacement
out. It is expected the hypothesis of pinned joints in the responses. Then as illustrated in the numerical
ball joint system would be no longer valid and stress- procedure, these responses are used as the
strain responses of the cable elements are strongly constitutive behavior of the struts in the finite
dependant on the level of cyclic loading. element modeling of the tensegrity systems.
Given the numerical results, it was expected that the • The effective length factor of k = 0.8 was obtained
studied tensegrity model at the critical load level for the struts considering the effect of ball joint
demonstrate partial progressive collapse. However, the systems, however, the amount of effective length
full progressive collapse was observed in the factor is dependent on the geometrical dimensions
experimental model. This discrepancy was due to sudden of the fastener bolts, end cones, tube, sleeve and
failure of fastener bolts of the struts after buckling. In the ball joint. Therefore, performing buckling tests on
normal performance, the magnitude of post-buckling the struts used in the system, is necessary and of
residual strength is approximately 0.2 of buckling load. crucial importance.
However, in the case of sudden failure of fastener bolts, • Loading imperfections may be arisen from the
the residual strength is vanished immediately. Therefore, angular errors of the threaded holes of the ball
the finite element modeling of the tensegrity model was joints. These imperfections are important as
modified considering sudden failure of the fastener bolts; geometrical imperfections. Therefore, in the design
and therefore reliable numerical results comparable to the process, different appropriate initial imperfection
experimental responses were obtained. The kinetic amplitudes for struts should be considered.
energy released due to the progressive collapse was so
large, led to the torsional buckling of the supporting
beams. The fastener bolts of the struts were of 12 mm
high strength steel (of grade 10.9) and failed from the pin
REFERENCES
[1] Skelton, R. E. and Oliveira, M. C. de., Tensegrity
Systems, Springer, New York, 2009.
hole. Based on the obtained results, the following design
[2] Motro, R., Tensegrity: Structural Systems for the Future,
recommendations are put forward: Kogan Page Sci., London, 2003.
• Failure of the fastener bolts caused the partial [3] Hanaor, A., Double-Layer tensegrity grids: static load
progressive collapse to change to overall collapse. response. II: Experimental study, Journal of Structural
Therefore, it is seriously necessary to pay special Engineering, 1991, 117, 1675–1684.
[4] Abedi, K. and Shekastehband, B., Static stability
attention in the design of the fastener bolts. In behaviour of plane double-layer tensegrity structures,
addition, it is advised to use nut system instead of International Journal of Space Structures, 2008, 23 (2),
sleeve-pin system. Therefore, it is not needed to 89–102.

International Journal of Space Structures Vol. 29 No. 1 2014 23


Experimental and Numerical Studies on the Progressive Collapse Behavior of Tensegrity Systems

[5] Shekastehband, B., Abedi, K. and Chenaghlou, M.R., [10] Caglayan, O. and Yuksel, E., Experimental and finite
Sensitivity analysis of tensegrity systems due to member element investigations on the collapse of a Mero space
loss, Journal of Construction Steel Research, 2011, truss roof structure–A case study, Engineering Failure
67 (9), 1325–1340. Analysis, 2008, 15, 458–470.
[6] Abedi, K., Propagation of local instabilities in braced [11] ASTM, Standard test method for tension testing of wire
domes, Ph.D. Thesis, University of Surrey, 1997. ropes and strand, 01.03, A 931–96, 2002.
[7] Abedi, K. and Parke, G.A.R., Experimental study of [12] ASTM, Standard test methods for tension testing of
dynamic propagation of local snap-through in single- metallic materials, 03.0.1, E 8–04, 2002.
layer braced domes, International Journal of Space [13] Quirant, J., Kazi-Aoual, M. N. and Motro, R., Designing
Structures, 2001, 16 (2), 125–136. tensegrity systems: the case of a double layer grid,
[8] Hibbit, Karlsson, Sorensen, ABAQUS/Standard, User’s Engineering Structures, 2003, 25, 1121–1130.
manual, Providence, Rhode Island, USA, 2002. [14] Tatemichi, I., Hatato, T., Anma, Y. and Fujiwara, S.,
[9] Shekastehband, B., Abedi, K. and Dianat, N., Vibration tests on a full-size suspen-dome structure.
Experimental and numerical study on the self-stress International Journal of Space Structures, 1997,
design of tensegrity systems, Meccanica, 2013, 48(10). 12 (3&4), 217–224.

24 International Journal of Space Structures Vol. 29 No. 1 2014

You might also like