Engineering Structures 79 (2014) 276289

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Exploratory numerical analysis of two-way straight cable-net faades

subjected to air blast loads
Chiara Bedon , Claudio Amadio
University of Trieste, Department of Engineering and Architecture, Piazzale Europa 1, 34127 Trieste, Italy

a r t i c l e

i n f o

Article history:
Received 13 May 2014
Revised 13 August 2014
Accepted 13 August 2014

Keywords:
Cable-net faade
Air blast load
Viscoelastic devices
Frictional devices
Energy dissipation
SDOF approach
Advanced numerical modelling
Dynamic nonlinear simulations

a b s t r a c t
Based on numerical and analytical results of previous literature contributions, the paper investigates the
dynamic behaviour of a structural two-way straight cable-net faade subjected to medium-level blast
loads. Numerical studies are dedicated, specically, to a cable-net prototype already experimentally
investigated under seismic loads. Several numerical nite-element (FE) models are developed in
ABAQUS/Standard. A geometrically simplied, lumped-mass FE-model (M01), a detailed cable-glass model
(M02) and a further simplied but computationally efcient cable-glass model (M03) are presented. The
models are assessed and calibrated to test data of literature, as well as to simplied SDOF analytical formulations, in order to investigate the dynamic response of the faade under high-rate impulsive loads.
Since axial forces in the cables and maximum stresses in the glass panels abruptly increase when the
explosion occurs, specic viscoelastic and frictional devices are applied at the connection between glass
and cables, as well as at the end of the cables. The main advantage of the presented technological solution
is given by the partial dissipation of the incoming energy. In the rst case, viscoelastic connectors add
exibility to the point-supported glass elements, hence reducing the amount of incoming blast energy.
In the latter case, frictional devices dissipate through friction mechanisms the strain energy stored by
the bearing cable-net. A proper combination of these multiple devices, in conclusion, can manifest a
marked increase of structural efciency for the examined structural typology.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
The rising desire of transparency in modern buildings markedly
increased the use of glass as a construction material able to carryon loads in the form of innovative structural members, roofs, curtain walls. As a result, during the last decades traditional design
concepts rapidly evolved, in conjunction with ongoing manufacturing developments, in order to provide appropriate technological
solutions to continuous advanced architectural demands [18].
Glass faades, specically, highlighted several technique
improvements and optimizations, aiming to cover wide surface
often characterized by irregular shapes.
While the traditional unitized curtain wall or stick systems,
for example, typically consists of modular units in which the glass
panels are sustained by a metallic frame composed of aluminium
transoms and mullions rigidly connected to the structural backup
(e.g. concrete slab), cable-supported faades are especially used
when large volumes and surfaces must be enclosed. Although
rstly built about 20 years ago, due to their transparency, easy
Corresponding author. Tel.: +39 040 558 3842.
E-mail addresses: bedon@dicar.units.it, c.bedon@libero.it (C. Bedon).
http://dx.doi.org/10.1016/j.engstruct.2014.08.023
0141-0296/ 2014 Elsevier Ltd. All rights reserved.

constructability and energy savings, the latter solution rapidly

propagated in worldwide as a new kind of glazing system.
Compared to other steel-glass typologies (e.g. unitized curtain
walls), cable-net systems represented for several years a marked
evolution of the usual design concepts, since typically associated
to high transparency and extreme minimization of the steel
supporting systems. The main advantage of cable-net faades is
in fact given by the intrinsic stiffness and stability offered by
appropriately pre-tensioned stainless steel cables.
In practice, single-way or two-way plane cable nets are typically designed to sustain gravity loads and orthogonal distributed
ordinary loads (e.g. wind). In the latter case, the appropriate limitation of out-of-plane deections due to in-service pressures represents the most critical design goal for these exible structural
systems, since although subjected to often moderate pressures,
large deections could derive from their typical overall dimensions
and complex geometries (e.g. New Beijing Poly Plaza Cable-Net
Wall (Fig. 1a, [9]), Kempinski Hotel at Munich Airport (Fig. 1b,
[10]), etc.).
In this context, numerous authors investigated through numerical, experimental or analytical models the static and dynamic
response of cable-net faades in different loading conditions.

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

277

Fig. 1. Examples of two-way straight cable-net faades. (a) New Beijing Poly Plaza Cable-Net Wall [9] and (b) Kempinski Hotel at Munich Airport [10].

Careful attention was paid especially for the study of the interaction between the glass panels and metal connectors [11,12]. Feng
et al. [13,14] deeply investigated the behavioural trends of cablenet faades subjected to seismic events. In [15], they also highlighted the effects of glass panels stiffness on the global dynamic
response of steel-glass faades, providing critical discussion and
comparison of experimental and numerical predictions. Studies
presented in [16,17] have been dedicated to the nonlinear behaviour of cable-supported faades under wind loads. Main features
and comparative discussion of cable-truss glass faades are proposed in [18].
While extended rigorous investigations are available for steelglass faades under ordinary loads, further knowledge is currently
required to understand and optimize their structural behaviour
under exceptional loading conditions. The dynamic behaviour of
glazing systems under high-strain rate and impulsive loads such
as explosion, for example, still represents a research topic of great
interests and a challenge for designers. Wide series of shock tube
experiments, analytical investigations and FE numerical studies,
in this context, have been dedicated to single laminated glass
panels subjected to air blast loads [1925]. Weggel et al. [26]
investigated the dynamic behaviour of unitized curtain walls under
low-level explosions. Numerical parametric simulations have been
discussed in [27], in order to provide pressure-impulse (PI) design
curves for the blast verication of laminated glass curtain walls.
Analytical and numerical studies have been recently proposed for
the dynamic buckling verication of monolithic and laminated
glass beam-like and columns under blast loads [28].
For the specic topic of cable-supported faades, Teich et al.
[29] focused on the analysis of the dynamic response of these systems under explosive events, emphasizing the inuence of the negative phase of a blast wave pressure on their global structural
behaviour. Wellershoff et al. [30] highlighted, through application
to the case study of the World Trade Center building, the typical
behaviour of a single-way straight cable supported faade under
blast loads, suggesting possible mitigation techniques for the limitation of damages and improvement of the expected dynamic
performances.
As also highlighted in [31], the design of a blast resisting cablesupported faade strongly differs from the conventional design of a
glazing system subjected to ordinary loads. When an explosion
occurs, the structural components of a cable-supported faade
are in fact affected by exceptional stresses both compressive
and tensile, due to mainly exural deformations involved by the
acting impulsive pressures and extreme deections. While glass
presents typical high nominal compressive strength, however,
maximum tensile stresses should be properly checked and limited,
to preserve the faade integrity and avoid brittle failure mechanisms in it.
In that earlier contribution [31], appropriately designed elastoplastic devices were proposed to be installed at the ends of the

vertical bearing cables, in order to prevent their possible brittle collapse under high-strain impulsive loads. The same cable-supported
faade markedly exible due to the presence of a single layer of
pre-stressed straight cables has been successively further
improved [32] by implementation of combined multiple devices,
namely consisting in elastoplastic devices at the cables ends and
opportune viscoelastic connectors introduced at the connection
between the glass panels and the cables [33]. Certainly, under
exceptional loading scenarios, the structural dynamic performance
of these structural systems should be optimized in order to avoid,
or at least minimize, possible injuries and structural damages, hence
preserving the activities within the building.
In this paper, based on extended numerical and analytical
results of previous literature contributions [3133], the behavioural trends of a two-way straight cable-net faade subjected to
medium-level blast loads are numerically and analytically investigated. In particular, studies are dedicated to the dynamic behaviour of a cable-net faade prototype already analyzed in [15]
under seismic events.
Based on [3133], multiple typologies of dissipative devices are
introduced both at the connection between the glass panels and
the pretensioned cable-net, as well as at the cable ends.
Modal analyses and nonlinear dynamic incremental simulations
are performed on a lumped-mass model (M01), a geometrically
detailed cable-glass model (M02) and a further simplied but computationally efcient cable-glass model (M03), able to take into
account the bending stiffness contribution of glass panes in the
form of equivalent beam elements. Parametric dynamic simulations are then performed, in order to highlight the efciency and
the criticalities of each possible solution.
As shown, by means of appropriate energy dissipating components, the increase of maximum axial forces in the bearing cables
and the maximum tensile stresses in the glass panels can be markedly reduced, hence the structural stability of the faade system
can be prevented.

2. Structural behaviour of cable-net faades under blast loads

2.1. Cable-net faade layout
In this paper, the cable-net faade recently analyzed under
seismic events [15] is numerically investigated to demonstrate the
feasibility and potentiality of passive control systems discussed in
[3133] for the mitigation of blast effects on single-way cablesupported faades to high-strain impulsive loads, by application of
the same mitigation technique to a two-way planar cablesupported faade. The mentioned glazing system consists,
specically, in a net of pretensioned steel cables, a series of square
monolithic glass panels and special clamping joints able to provide
appropriate structural interaction between glass and steel

278

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

components. The faade has overall dimensions B = 11.125 m 

L = 11.125 m and is obtained by assembling 49 b = 1.5 m  h = 1.5 m
glass panels having nominal thickness t = 0.016 m. Vertical and horizontal / = 12 mm-diameter steel cables are subjected to an initial
pretension force Hv,0 = Ho,0 = H0 = 40 kN and linked together at their
intersections. To form a bearing net, the cables are spaced at intervals of 1.5 m in both the x and y directions (Fig. 2). Each glass panel is
then four-point supported and connected to the cable-net by means
of stainless steel clamping joints (Fig. 3). The faade prototype is
nally sustained by means of a steel bracing frame made of square
tubes (side length 0.15 m, thickness 0.008 m) and by three further
diagonal square tubes (side length 0.10 m, thickness 0.005 m),
hinged at the top of the steel frame to provide an appropriate outof-plane stiffness to the composite steelglass system.
2.2. Blast load denition

Fig. 3. Schematic view and cross-section of the typical stainless steel clamping joint
[15].

An explosive event typically represents an exceptional loading

condition for structures that involves extremely rapid release of
energy. This energy propagates, generally in thousandths of seconds, in the form of light, heat, sound and shock waves. From a
structural point of view, the simplest way to properly simulate
the effects of far eld explosive events on buildings and systems
takes the form of an air blast shock wave consisting in a pressuretime history having specic positive and negative phases
[35]. Both these positive and negative phases, properly calibrated
to describe the assigned design blast load, should be generally
taken into account in simulations, as also suggested by Dharani
and Wei [21] or Teich et al. [29]. The positive phase only, in particular cases, could in fact provide maximum dynamic effects markedly higher than those deriving from the total blast shock wave,
hence resulting in improper predictions.
In practice, the positive phase, of total duration t
d , usually takes
the form of the exponential function proposed by Friedlander [36]:



t
kp
t
pt p0 1  e td ;
td

0  t  td ;

being Eq. (1) representative of the pressure waveform p(t) in the

time interval comprised between t = 0 (e.g. maximum overpressure

p0) and t t
d (e.g. end of the positive phase, with p(t d ) = 0).
Three main parameters can affect the exponential pressure
function given by Eq. (1), namely (i) the distance between the
explosion source and the invested building (stand-off distance,
dTNT), (ii) the height of the explosion source from ground (hTNT)
and (iii) the quantity of explosive (equivalent mass of TNT, MTNT).
At the same time, the overpressure peak p0 is dened in accordance
with recommendations provided by the TM 5-1300 standard [37].

The shape coefcient kp is properly estimated to take into account

the velocity of decay of the overpressure peak p0 within the interval 0 6 t 6 t
d . Several methods and empirical equations have been
proposed in literature for a practical calculation of kp, based on the
assigned blast pressure function (Eq. (1)) and its impulse ratio [35].
In this work, according to [38], kp was calculated as:

kp 5:2777Z 1:1975 ;
dTNT/MTNT1/3

2
1/3

being Z =
[m/Kg ] the scaled distance.
While the positive phase namely represents the ambient pressure increase and modication due to the occurring explosion,
the negative phase of an ideal blast event is characterized by a
pressure amount which is lower than the reference ambient pressure. This negative phase is typically described in the form of a triangular shape associated to a negative under-pressure peak pneg
and a total duration ttot [36].
In this paper, a medium-level air blast load (Level C-GSA [34])
was taken into account to study the behavioural trends of the
investigated faade prototype. The main parameters of the mentioned design blast load, based on recommendations of [34] and
[37], are the impulse per unit of surface (i = 272 kPa ms), the overpressure peak (p0 = 42.9 kPa), the duration of the positive phase
(t
d 0:0165 s) and the total impulse duration (ttot = 0.0748 s).
The resulting time-varying pressure function is proposed in Fig. 4.
2.3. Dissipative devices
A structural system subjected to an explosive load should be
generally able to be as much as possible exible and dissipative,

Fig. 2. Front and side elevations of the studied cable-net faade [15].

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

Pressure [Pa]

6x10 4

BLAST LOAD
MTNT= 50Kg
dTNT= 30m

4x10 4

2x10 4

-2x10 4
0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

Time [s]
Fig. 4. Blast load time-varying pressure function (Level C-GSA [34]).

in order to decrease the incoming energy and partly dissipate it,

hence avoiding the attainment of large deformations and the possible failure mechanisms in its basic components. In the case of
structural glass systems, various technological advancements have
been proposed over the last decades to provide additional plastic
exibility after glass cracking and to ensure possible injured. Laminated glass, for example, still represents the typical blast resisting glass component, since the thermoplastic foils used to bond
together the glass panels are able to prevent the possible detachment of shards after collapse, as well as to increase the damping
capabilities of the cracked glass panels. Under extreme loading
conditions, however, the use of laminated glass alone cannot provide appropriate exibility and damping capacities to complex
structural systems such as wide glazing faades.
In this work, based on [3133], different technological solutions
are applied to the faade object of study. First, viscoleastic (VE)
connectors able to partly dissipate the incoming energy by means
of viscous phenomena are applied in the point of connection
between the glass panels and the cable-net (Fig. 5a). In them, a viscoelastic rubber layer of total thickness hVE can be used to provide
appropriate absorption of the incoming energy. Its key parameters
are in fact the elastic stiffness kd and the damping coefcient cd,
being [33]:

kd

G0 AVE
hVE

gkd
;
x1

and

cd

with x1 the fundamental pulsation of the faade, AVE the shear

resisting area of the viscoelastic layer, G and g the shear storage
and loss factor respectively of the adopted rubber compound.

279

The main advantage of these devices is given by their high

potentiality under high-impulsive as well as moderate/low
dynamic loads. The structural efciency of VE devices is in fact
strictly related to their sliding sVE . The large the VE device slides
and the high is the term of dissipated energy. Nevertheless, compared to the total thickness of the viscoelastic layer hVE, this maximum sliding should not exceed the ratio sVE
max =hVE 2  2:5.
The second typology of devices consists in frictional connectors,
having a typical rigid-plastic (RP) behaviour, introduced at the
cable ends (Fig. 5b). In this case, the input energy is partly dissipated in heat by friction mechanisms that can activate at the
attainment of maximum axial forces in the cables Hmax higher than
the sliding force Fs. As far as the axial forces Hmax in the cable-net
do not exceed Fs, the RP devices do not activate and the glazing system behaves as a traditional cable-supported faade rigidly connected to the structural background. The value of the sliding
force Fs, as well as the position of RP devices, should be consequently properly assessed in order to maximize their efciency.

3. Preliminary analytical assessment and estimation of

maximum blast effects
The assessment of the structural dynamic behaviour of a complex system such as a cable-supported faade under impulsive
loads should be generally performed by means of advanced numerical studies and appropriate FE simulations. From a practical point
of view, however, simple preliminary estimations could be rationally obtained by means of approximate analytical formulations.
In [31], for example, an analytical model has been proposed for
a rst estimation of blast effects in glazing faades sustained by a
series of vertical pre-stressed straight cables. In that specic circumstance, the typical multi-degree-of-freedom (MDOF), singleway cable-supported faade was approximated with an equivalent
single-degree-of-freedom (SDOF) system subjected to an external
total impulse I representative of the positive phase of the assigned
blast load.
The essential dynamic parameters of the equivalent SDOF were
estimated referring to Ritzs method, by using the classical energy
approach, and solving the original problem in a static way. As
shown in [31], the mentioned approach can be used for a rst estimation of maximum deection, velocity and pre-stressing force in
the cables for single-way planar cable-net faades under impulsive
loads. Although it does not provide information about the stress
state in the glass panels fully neglecting their bending stiffening
contribution in the global dynamic behaviour of the whole system
the approach is in fact suitable for a preliminary assessment of
the expected dynamic effects of a design explosive load.

Fig. 5. (a) Example of viscoelastic (VE) spider connector, cross-section; (b) example of frictional (RP) device, rendering.

280

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

The same analytical approach, based on analogous energy considerations, is extended in this work to steel-glass faades sustained by a planar net of pretensioned straight cables (Fig. 6).
The MDOF cable-net object of study, specically, is described in
the form of an equivalent SDOF system whose dynamic behaviour
is characterized in terms of equivalent mass m* and the equivalent
stiffness k*, being rationally negligible, under high-rate impulsive
loads such as explosions, possible structural and aerolastic damping contributions. Differing from [31], moreover, the presence of
a two-way cable-net rather than a single-way bracing system is
properly taken into account. As a result, the equivalent mass of
the cable-net faade is calculated as [40]:

M
m
4


with M signifying the total mass of the glass panels.

At the same time, the equivalent stiffness of the pretensioned
net is estimated as [40]:



2
4Htot
p
2 sin
lcable
ncables 1

being lcable the length of cables and ncables signifying the number of
cables arranged in each main direction of the net (or the average
number of vertical ncables,v and horizontal ncables,h cables, when
ncables,v ncables,h). In Eq. (6), nally, Htot represents the total initial
pretension applied to the bearing system and should be calculated
as:



Htot H0 ncables;v ncables;h

that is Htot = 2H0ncables.

In these hypotheses, the fundamental period of vibration can be
estimated as:

r
m
T 1 2p

k
T 1

being
strictly related to the level of pretension force in the cablenet.
While Eq. (8) provides the fundamental period of vibration of
the faade affected by the initial pretension force Htot  H0,tot only,

it should not be neglected that the design blast load involves an

abrupt and marked increase of axial forces in the cables, thus a signicant increase of the equivalent stiffness k* and a corresponding
decrease of the associated period of vibration T 1 .
Once the equivalent dynamic parameters M* and k* are known,
the maximum displacement umax and velocity vmax of the cable-net
due to the design blast load can be respectively calculated as:

umax

s
r
I2
1

I


Mk
Mk

and

v max

s
I2
2I

Mm M

10

being I = i  Sglass the incoming impulse, with Sglass the surface of

faade exposed to the assigned blast wave.
While the maximum velocity vmax can be directly estimated
from Eq. (10), the corresponding deection umax should be properly
predicted by means of an iterative procedure [31]. An accurate

evaluation of the equivalent stiffness k (Eq. (6)), and therefore of
the maximum displacement umax (Eq. (9)), should in fact be developed taking into account that, due to the assigned explosion, the
initial pretension force H0 affecting each cable modies abruptly.
In this context, it could be convenient to describe the nal conguration of the deformed cables constituting the cable-net in the
form of a parabolic shape function [31]. Certainly, a similar
assumption would be able only to approximately identify the global deformed shape of the studied cable-net. However, it is interesting to notice that the total increase Hblast,tot of the total initial
pretension H0,tot in the cable-net could be estimated as:

Hblast;tot

8 Ecable Acable 2
umax
2
3
lcable

11

with Ecable the Youngs modulus of steel constituting the cables and
umax the maximum deection given by Eq. (9). As a result, the maximum total pretension force affecting the cable-net after the explosion results:

Htot Hmax;tot H0;tot Hblast;tot

Fig. 6. MDOF cable-net faade and equivalent SDOF system.

12

281

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

In this context, an iterative process should be performed by

replacing Eq. (12) in Eq. (6), until the obtained value umax (Eq.
(9)) remains constant. It is clear that once the iterative procedure
is converged and umax is known, the maximum axial force occurring in each cable due to the design explosion can be rationally
estimated as:

Hmax

H0;tot Hblast;tot
Hmax;tot

ncables;v ncables;h ncables;v ncables;h

13

By replacing Hmax in Eq. (6), the corresponding increased equivalent stiffness k*blast can be consequently calculated and taken into
account in Eq. (8), hence resulting in a modied fundamental per*
iod of vibration T1,blast
markedly lower than T1*.
Main results obtained by application of the discussed analytical
procedure to the studied faade subjected to a Level C air blast load
are collected in Table 1.
The assigned design explosion involves large deections (1/20
the total span), as well as high maximum velocities (15.4 m/s)
and axial forces in the cables (1.02 HR, with HR = 160 kN the characteristic strength of the adopted cables).
In the same table, additional analytical calculations are proposed for the same faade under a Level B-GSA (low-level explosion) and a Level D-GSA blast load (high-level explosion),
respectively. As shown, independently on the intensity of the
assigned explosive loads, due to their impulsive and high-rate nature compared to ordinary loads extreme deformations and
forces are exerted on the examined structural system. Further
advanced investigations are consequently required.

4. FE-modelling of the cable-net faade and preliminary

numerical investigations
Several numerical models were successively developed in ABAQUS, in order to investigate in detail the dynamic behaviour of the
studied faade under medium-level blast loads but preserving the
computational cost of dynamic nonlinear analyses. Dynamic simulations, in particular, were performed by means of the implicit
ABAQUS/Standard solver (Section 4.3).
Prior to the execution of parametric dynamic studies, however,
the FE-models were properly assessed through linear modal
analyses to experimental results of free vibration tests presented
in [15] for the same glazing system. Detailed description of
FE-models properties is discussed in Section 4.1.
4.1. FE-model properties
4.1.1. Lumped-mass model (M01)
The rst geometrically simplied FE-model (M01) consisted in a
net of pretensioned cables (/ = 12 mm) and a series of lumpedmasses (mi = 89.64 kg) applied at the interceptions between the
cables representative of the self-weight of 49 monolithic
1.5  1.5  0.016 m3 glass panels (MDOF system of Fig. 6).

In it, the cables were described in the form of truss elements

(T3D2 elements) composed of stainless steel, being the material
characterized in the form of an indenitely linear elastic mechanical behaviour, with Esteel = 1.03  1011 Pa the Youngs modulus,
qsteel = 7300 kg/m3 the density and msteel = 0.3 the Poissons ratio.
Compared to M02 and M03 FE-models (Sections 4.1.2 and 4.1.3),
the main feature of M01 FE-model was given by its marked computational efciency, since consisting in 800 degrees of freedom
(DOFs). Nevertheless, its main limit was represented by fully
neglecting the stiffening contribution of glass panels.
In it, specically, the initial pretension force (H0 = 40 kN) was
applied both to vertical and horizontal cables in the form of equivalent imposed displacements. In the case of the vertical cables,
pinned at the lower end (Ux = Uy = Uz = 0), a vertical displacement
(y-direction) was applied at their top end. Similarly, for the horizontal
cables pinned at their left end (Ux = Uy = Uz = 0), an horizontal displacement (x-direction) was imposed at their right end.
The effect of clamping joints (Fig. 3) was then taken into account
in the form of a series of translator connectors able to link together
the nodes of vertical and horizontal cables at their interceptions. In
this manner, relative rotations were prevented, so that the coupled
nodes of the cable-net could undergo the same out-of-plane deection when subjected to orthogonal loads, like for example external
pressures acting on the glass panels surface. A further implicit
advantage deriving from this modelling assumption, consisted in
the separate and independent application of the initial pretension
force to vertical and horizontal cables respectively.
In accordance with the experimental measurements presented in
[15] for the cable-net without glass panels (nexp  0.1%), structural
and aerolastic damping contributions were preliminary neglected.

4.1.2. Full cable-glass model (M02)

In the case of the full cable-glass model (M02), the main components of the faade were described with careful attention, in
order to appropriately simulate their structural interaction under
high-strain impulsive loads such as explosions.
The prestressed cable-net alone, as well as the mechanical
interaction between vertical and horizontal cables, was described
as for the model M01 (Section 4.1.1). In addition, clamped joints
and glass panels were also properly taken into account. All the considered modelling assumptions typically resulted in detailed geometrical and mechanical description of the studied faade, but
also in marked increase of the involved DOFs (70,000), compared
to M01 FE-model. Specically, three-dimensional four-node homogeneous shell elements with reduced integration formulation (S4R
elements available in the ABAQUS library [39]) were used to
describe the monolithic glass panels, with t = 0.016 m the total
nominal thickness. Default parameters were taken into account
for the hourglass control. Based on preliminary studies not
included in this paper, an almost regular and structured mesh pattern was used, with lmesh = 0.1 m the average size for each
1.5  1.5 m glass sheet (Fig. 7). At the corners of each panel close
to the clamping joints and along the panel edges, a rened mesh

Table 1
Analytical estimation of maximum dynamic effects on the studied cable-net faade subjected to air blast loads. i: impulse per unit of surface; p0: overpressure peak.
Parameter

Description

Level B (low)
i = 168 kPa ms
p0 = 30.4 kPa

Level C (medium)
i = 272 kPa ms
p0 = 42.9 kPa

Level D (high)
i = 439 kPa ms
p0 = 62.6 kPa

T1 [s]
umax [m]
vmax [m/s]
Hmax [kN]
T1,blast [s]

Fundamental period, Eq. (8)

Maximum displacement, Eq. (9)
Maximum velocity, Eq. (10)
Maximum total pretension in the cables, Eq. (13)
Expected fundamental period after the explosion
has been occurred, Eq. (8) with H0  Hmax

0.578
0.526
9.47
109
0.350

0.578
0.699
15.33
163
0.286

0.578
0.912
24.74
249
0.231

282

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

Fig. 7. Detail of glass-to-cables connection and adopted mesh pattern (ABAQUS/

Standard, M02 FE-model).

pattern and a free meshing technique were used. In this case, the
minimum element size was set equal to lmesh  0.003 m (Fig. 7).
The nal choice for the adopted mesh pattern globally derived
from an appropriate balance between mesh renement, corresponding predictions in the glass panels and main purpose of this
research work. The investigation of possible damage mechanisms
and post-cracked behavioural trends for the same faade under
blast loads, hence requiring the implementation of specic damage
mechanical models for glass (e.g. [2327,41]), would in fact require
an extremely rened mesh pattern, being those material models
(e.g. the brittle cracking ABAQUS option [39]) generally strictly
dependent on the characteristic size of elements [42].
Concerning the mechanical characterization of toughened glass,
in accordance with [3133], an isotropic, linear elastic characteristic behaviour was in fact dened, with Eglass = 70 GPa the Youngs
modulus, qglass = 2490 kg/m3 the density and mglass = 0.23 the Poissons ratio. As a result, possible brittle cracking of glass panes
was fully neglected. From a numerical point of view, this assumption directly manifested in geometrical and mechanical simplied
modelling hypotheses, but anyway in reasonable FE-predictions for
the studied faade, as discussed in Sections 45. Glass is in fact a
material characterized by typical high resistance in compression
and brittle behaviour in tension, being its nominal quasi-static
characteristic strength generally comprised between 45 MPa
(annealed glass) and 120 MPa (fully tempered glass) often representative of the main inuencing parameter for designers [43].
Under high-strain loads, however, conventional quasi-static tensile
strengths are usually replaced with magnied corresponding values able to take into account strain rate effects. Under explosions,
having a typical duration of 0.0050.025 s, the magnied tensile
strength recommended by design manuals and standards for
toughened or thermally tempered glass as for the studied faade
is 160 MPa [44]. Further references are available for glass under
blast loads, but certainly the estimation of the actual tensile
strength under high-strain rates is rather difcult to exactly predict. Blast design strength values for tempered glass can be in fact
found comprised between 145 MPa and 209 MPa [45,46], thus a
magnifying dynamic coefcient at least equal to 1.2 is generally
accepted for FT glass. Comparative discussion of technical documents recommendations and assumptions is provided in [47].

Experimental studies have been performed also in [48], to investigate the dynamic increase of both compressive and tensile
strengths for glass under high-strain loads. Data measurements
and high-speed camera acquisitions obtained from tests performed
on small cylindrical glass specimens in the range from 103 to
103 s1 highlighted that the compressive strength and Young modulus of glass are slightly sensitive to strain-rate effects. Conversely,
the quasi-static tensile strength manifested marked dependency
on strain-rate effects and an appreciably increased dynamic
resistance.
In these hypotheses, the occurring of possible failure mechanisms in the M02 FE-model was indirectly prevented by continuously monitoring the maximum tensile stresses in glass, hence
ensuring the exceeding of the reference magnied resistance (Sections 45).
Careful consideration was then dedicated to the modelling of
the clamping joints, since constituting the key component of
numerical simulations. The typical metallic joint consisted of four
rigidly connected stainless steel beams (B31 elements), representative of the nominal geometry of the metal fasteners. A weld connector, able to provide a fully bonded connection between the
relative displacements (ux, uy, uz) and rotations (rx, ry, rz) of these
four beams was used, hence ensuring the desired restraining effect
between them. Concerning the glass panel-to-joint and joint-tocable structural interaction, each clamping joints was attached to
the adjacent glass panels by means of appropriate join and slideplane connectors (Fig. 8). The rst typology of connectors, able to
prohibit possible relative displacements (ux, uy, uz) in the interested
nodes, was inserted at the top-left corner of each glass panel. Slideplane connectors able to guarantee small relative displacements in
the plane of glass sheets were used, conversely, for the three other
corners of each panel. In this manner, the attainment of peaks of
stress at the connection points of the glass panels was prevented,
especially when applying the initial pretension force in the cables.
Finally, the structural interaction between the so modelled
clamped joints and the pretensioned cable-net was guaranteed by
means of a series of two additional connectors (Fig. 8, Connector
1 and Connector 2), able to provide in presence of external
pressures orthogonal to the surface the faade the same out-ofplane deection for the glass panels and the cable-net in the interested nodes.
As also discussed in [15,31,32], the mass of clamping joints was
neglected. At the same time, since characterized by a negligible
stiffness [26], the layers of structural silicone sealant interposed
between the adjacent glass panels were not taken into account.
With reference to possible structural and aerolastic damping
effects [15], shock excitation experiments performed on the
cable-net prototype with the glass panels highlighted as expected
a marked increase of dissipative capabilities for the studied system (nexp  3.4%), compared to the cable-net alone.
In this context, it should be noticed that the obtained experimental value is rather in agreement with simple analytical calculations derived from literature. Assuming the total damping ratio
nTOT = nstructural + naerolastic  3.4% as the sum of structural and aerolastic damping terms, the test result can be in fact rationally
assessed, being nstructural  1% [49,50] the conventional structural
contribution and naerolastic  2.43% the aerolastic term, with:

naerolastic

cd qair Sglass v max

:
2Mtot x1

14

In Eq. (14), cd = 1 is the drag coefcient [23], qair = 1.225 kg/m3

represents the density of air, Sglass = 110.25 m2 is the faade surface
exposed to blast, while Mtot = 4392 kg is the total mass of glass
panels. Moreover, vmax represents the velocity attained by the moving faade due to the design explosion, whereas x1 is its fundamental frequency of vibration. Based on Eqs. (10) and (8),

283

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

Fig. 8. Detail of a clamping joint (ABAQUS/Standard, M02 FE-model).

vmax = 15.33 m/s

and x1 = 2p/T1 = 10.88 rad/s respectively, hence

resulting in a total damping ratio nTOT  3.43% which well agrees
with shock excitation predictions [15].
The so calculated damping contribution was described in the
M02 FE-model as a Rayleigh mass proportional term. Preliminary
numerical studies not included in this paper conrmed the negligible effects of a similar damping term on the global dynamic
response of the studied faade under high-rate impulsive loads.
As a result, damping was successively fully neglected, as in the case
of the cable-net alone (M01 FE-model).

the corresponding equivalent diagonal beam FE-model resulted

in a very good agreement in terms of maximum deections (3.8%
of discrepancy), maximum tensile stresses close to the point supports (4.8% of discrepancy) and maximum reaction forces (almost
coincident) transmitted to the structural background.
At the same time, further modal analyses performed on both
these FE-models manifested an almost identical fundamental period of vibration (0.28% of discrepancy, Fig. 9). It was thus expected,
that the FE-model M03 could provide appropriate dynamic predictions for the studied faade.

4.1.3. Equivalent cable-diagonal beam model (M03)

In the third numerical model (M03), geometrical simplications
were taken into account for the description of the glass panels, in
order to reduce the computational weight of model M02 but to
provide, at the same time, appropriate dynamic predictions for
the studied faade.
Specically, the pretensioned cable-net and the clamped joints
were described as for the model M02. Conversely, in accordance
with the Grashofs method [51], the monolithic glass panels were
schematized in the form of two diagonal beam elements (B31)
composed of glass, having total thickness t = 0.016 m and equivalent width beq.
This modelling assumption allowed to preserve the description
of each faade component (e.g. cables, connectors and glass panels)
and the physical interaction between them, but at the same time to
obtain a computationally efcient FE-model (7000 DOFs, that is
1/10 than the corresponding M02 FE-model).
In doing so, the equivalent width beq was calculated to guarantee the same bending stiffness between each nominal glass panel
and the coupled diagonal beams. By equalling the maximum
deection of a point-supported panel under uniformly distributed
load q and a simply supported beam subjected to the resulting
mid-span force F = ql2/2, with l = b = h = 1.5 m the size of each
square panel, beq was in fact calculated as:

4.2. Assessment of FE-models and dynamic properties

p 3
2l

  1:244l
beq
2
2:4 1  m2glass l

Linear modal analyses were performed in ABAQUS/Standard

using the models M01, M02 and M03 respectively, in order to compare the obtained fundamental periods of vibrations and the corresponding modal shapes to experimental results obtained by Feng
et al. from free vibration tests performed on their scaled prototype
[15].
Compared to experiments, the obtained fundamental periods of
vibration highlighted as expected that the bending stiffness of
the glass panels slightly affects the dynamic behaviour of the studied glazing system (3% of discrepancy). First periods of vibration
resulted respectively equal to T1,M01 = 0.581 s, T1,M02 = 0.569 s and

15

p
The t  beq  2l diagonal beams were then linked together at
their mid-span node by means of a tie constraint, able to provide
full coupling and interaction between them. At the same time,
the self-weight of glass was described in the form of lumped
masses mi.
Although the Grashofs method represents an approximate
approach, preliminary static, nonlinear geometry numerical analyses performed on a single point-supported shell glass panel and

Tshell = 0.07916 s

(a)

Tbeam = 0.07938 s

(b)

Fig. 9. Fundamental modal shape of a point-supported glass panel. (a) shell model
and (b) equivalent diagonal beam model (ABAQUS/Standard).

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

T1,M03 = 0.567 s, thus rather in agreement with the reference test

prediction (T1,test = 0.580 s [15]), as well as with the analytical
model presented in Section 2 (Eq. (8) and Table 1, T1* = 0.578 s).
Concerning the fundamental modal shape of the examined
faade, moreover, calculations performed on the amplitudes of the
modal displacements recorded at 6 points of measure (Fig. 10a),
compared to the corresponding free vibration measurements [15],
highlighted that the simplied model M01 generally tends to overestimate the actual exibility of the faade (+5.8% average discrepancy), being not able to take into account the possible bending
stiffening contribution of the glass panels (Fig. 11). At the same
time, the model M02 more geometrically accurate provides a
very good agreement with experimental measurements (+0.17%
average discrepancy). The model M03, nally, although geometrically simplied but markedly computationally efcient, compared
to the corresponding M02 FE-model, is anyway able to rationally
reproduce the expected dynamic behaviour for the studied system
(+1.07% average discrepancy). Good level of accuracy was in fact
found between M03 predictions and the reference test measurements also for higher modes, in terms of vibration periods (e.g.
T1,M03 = 0.361 s for the third vibration mode, with T3,test = 0.354 s
the corresponding experimental value [15]).
Based on this preliminary assessment, further parametric
dynamic nonlinear investigations were performed on the M03
FE-model, as discussed in Section 4.3.

4.3. Dynamic behaviour of the glazing system under blast loads

Incremental dynamic analyses characterized by a total duration
of 32 s were performed on M01, M02 and M03 FE-models. As highlighted in [31], an explosive event involves in cable-supported
faades an abrupt increase of axial forces in the bearing cables.
While single-way cable supported faades (e.g. [31,32]) are
typically characterized by high exibility, however, two-way
cable-net faades are generally stiffer. Their structural behaviour
under high-strain impulsive loads could consequently involve
extreme damage scenarios and peaks of stress in their basic
components that should be properly predicted.
In this work, nonlinear dynamic simulations have been performed in ABAQUS/Standard. The typical simulation consisted in
three separate steps. In the two rst steps, having both a total
duration of 1 s, the initial pretension force was separately assigned
to horizontal and vertical cables. Once pretensioned the bearing
cable-net, during the third step (duration 30 s) the glazing system
was subjected to an explosive event described in the form of an

(a) TM01= 0.581 s

1.0

Modal amplitude

284

0.8
M01

0.6

M02
M03

0.4

TEST
1

Point of measure
Fig. 11. Assessment of FE-models M01, M02 and M03 (ABAQUS/Standard) to
experimental modal amplitudes [15] (rst vibration mode).

impulsive load agreeing with the pressuretime function proposed

in Fig. 4. Maximum time increment was set equal to 0.005 s.
The assigned blast load, specically, was described respectively
in the form of nodal forces Fi applied to each lumped-mass mi
(model M01), distributed pressures q acting on the surface of glass
panels (model M02) and equivalent nodal forces Fi applied to the
mid-span node of each diagonal glass beam (model M03).
4.3.1. Simplied M01 model
The geometrically simplied M01 model was preliminary investigated due to extreme computational efciency in order to
assess the accuracy of the analytical procedure presented in Section 3, as well as to highlight the possible effects of the assigned
blast load.
In a preliminary phase of this work, for example, it was calculated that the maximum deection of the studied faade affected
by ordinary wind loads (qwind = 0.75 kN/m2) is uM01
max;wind 0:025 m
(1/450 of the structural span). The obtained average velocity,
for the same loading condition, is moderate (v M01
max;wind 0:14 m=s),
compared to maximum effects typically associated to extreme
loading scenarios (e.g. Table 1). Similarly, the increase of maximum
axial forces in the bearing cables is limited, compared to the initial
M01
pretension level (HM01
max;wind 45 kN; that is Hmax;wind  1:1H 0 ).
Conversely, explosions represent an exceptional loading condition for buildings, and a prexed level of damage should be
expected. The studied cable-net system, for example, should be
able to implicitly provide an efcient support to the curtain wall.
Since the ultimate strength for the used cables is HR = 160 kN, however, the occurring of elevated axial forces should be avoided. Steel
cables have in fact typical brittle-elastic behaviour and their

(b) TM02= 0.569 s

(c) TM03= 0.567 s

Fig. 10. Fundamental modal shapes for FE-models (a) M01 (800 DOFs), (b) M02 (70,000 DOFs) and (c) M03 (7000 DOFs). ABAQUS/Standard.

285

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

20
16
12
8

Velocity [m/s]

rupture would cause the collapse of the entire suspended curtainwall. At the same time, the possible cracking of glass panels should
be possibly prevented.
Main results obtained from dynamic simulations performed on
the M01 model are proposed in Figs. 1214 respectively, in the
form of maximum deections, maximum velocities and maximum
axial forces in the bearing cables.
As shown, the positive phase of blast alone involves maximum
effects that are markedly onerous for the studied system. The maximum deection (uM01
max 0:78 m;  1/15 the total span), velocity

0
-4
-8

No devices
Pos.

-12

(HM01
max

16:3 m=s) and forces in the cables

196 kN) are
(v
rather in agreement with the corresponding analytical predictions
collected in Table 1 (Level C blast load, Hmax = 163 kN). At the same
time, it is interesting to notice that the obtained value for the maxM01
max

-16

Pos. and neg.

-20
0

Time [s]

imum pretension force HM01

max 196 kN would provide, by means of

being
HM01
av g

200
175

No devices
Pos.

HR

Pos. and neg.

150
125
100
75

H0

50
25
0
0

10 12 14 16 18 20 22 24 26 28 30 32

Time [s]
Fig. 14. Maximum axial force in the cables (ABAQUS/Standard, M01 FE-model, no
devices).

and HM01
max 110 kN respectively (with

v M01
max 8:2 m=s

66 kN the average force after the explosion).

4.3.2. M02 and M03 models

Final assessment of numerical models under medium-level
blast loads was performed on M02 and M03 glazing systems subjected to the total design blast load dened in Fig. 4. As shown in
Fig. 15, an acceptable agreement was found between them. Due
to the presence of glass panels having stiffening and stabilizing
effects for the entire cable-supported system the maximum
deection in the cable-net resulted equal to 1/28 the total span

1.2
0.9
0.6

Displacement [m]

Fig. 13. Maximum velocity in the cable-net (ABAQUS/Standard, M01 FE-model, no

devices).

Axial force [kN]

M01
Eq. (8), a vibration period T M01
well agreeing
1;blast 0:261 s  0:45T 1

with the estimated nal vibration period T 1;blast = 0.286 s proposed
in Table 1.
Although the numerical results collected in Figs. 1214 conrm
the effectiveness of the SDOF analytical procedure for the preliminary assessment of the cable-net faade dynamic response under
the positive phase of blast only, it should be anyway highlighted
that both the SDOF procedure and the M01 model do not take into
account the possible bending stiffening contribution of the glass
panels, as well the possible benecial effects due to the negative
phase of the assigned blast load.
The effect of the total impulsive pressure function depicted in
Fig. 4 can be easily noticed in the same Figs. 1214. For the studied
faade, as expected, the negative phase of the design explosion
reduces in fact up to 30% the maximum effects of the positive
phase alone. As a result, the maximum deection is uM01
max 0:53 m,
that is 1/20 the total span of the faade. The maximum velocity
of the oscillating glazing system and the corresponding maximum
axial forces in the cables are also markedly reduced, compared
to the same faade under the positive phase only of blast,

10 12 14 16 18 20 22 24 26 28 30 32

0.3
0.0
-0.3
-0.6

No devices
Pos.

-0.9

Pos. and neg.

of the faade, thus sensibly lower than the corresponding M01 predictions (Fig. 15a).
In terms of maximum tensile stresses in the glass panels, simulations performed on M02 and M03 models conrmed that maximum peaks are located near the point-supports. The typical
distribution of maximum in-plane stresses obtained from the M02
and M03 FE-models is proposed in Fig. 15b for the central panel of
the faade (e.g. the panel subjected to maximum overall deections,
due to blast pressure), in the form of blue (min)-to-red (max) contour plot and vectorial representation, respectively. Despite the
approximate modelling assumptions of M03 model, acceptable correlations were generally found with the corresponding M02 model.
In Fig. 15c, for example, tensile stresses attained close to the corner
of the same glass panel depicted in Fig. 15b are shown.
At the same time, numerical predictions obtained from both
the FE-models conrmed the assumption of an indenitely linear
elastic material for glass, being the obtained maximum tensile
stresses markedly lower than the reference magnied resistance
(Section 4.1.2).
Based on these statements and comparisons proposed in Fig. 15,
parametric numerical simulations discussed in Section 5 were performed on further M03 models only.

-1.2
0

10 12 14 16 18 20 22 24 26 28 30 32

Time [s]
Fig. 12. Maximum deection in the cable-net (ABAQUS/Standard, M01 FE-model,
no devices).

5. FE-modelling of the faade equipped by dissipative devices

In a subsequent numerical investigation phase, different typologies of dissipative devices were introduced in the cable-supported

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

0.6

60

0.4

50

Tensile stress [MPa]

Displacement [m]

286

0.2

0.0

-0.2

M01
M02
M03

-0.4

M02
M03

40

30

20

10

-0.6
2.00

2.25

2.50

2.75

3.00

3.25

3.50

2.00

2.25

2.50

2.75

Time [s]

Time [s]

(a)

(c)

M02 FE-model

(b)

3.00

3.25

3.50

M03 FE-model

Fig. 15. (a) Maximum deections and (c) tensile stresses in the central glass panel (ABAQUS/Standard, M02 and M03 FE-models, no devices). (b) Typical distribution of
maximum in-plane stresses due to the applied blast load (ABAQUS/Standard; M02 (blue (min)-to-red (max) contour plot) and M03 (vectorial representation) FE-models).
Example for the central glass panel. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

faade, both at the ends of the bearing cables and at the pointsupports of the glass panels, in order to analyze their effectiveness
and to identify the most advantageous solution.
Studies were dedicated to faades equipped by (i) viscoelastic
devices only, (ii) frictional devices at the cable ends, (iii) combined
viscoelastic and frictional devices.
Based on previous works [3133], the reference M03 model
was properly modied. Viscoelastic connectors, specically, were
described in the form of equivalent axial springs having elastic
stiffness kd and damping ratio cd, and directly connected to the
glass corners. Frictional devices, indeed, were applied at the cable
ends in the form of rigid-plastic springs with sliding force Fs. Parametric studies, partly discussed in Section 5, were then carried-out
in order to assess the possible structural efciency of the proposed
devices for the investigated cable-net faade.
5.1. Numerical assessment of VE and RP devices structural capabilities
5.1.1. Viscoelastic devices
In general, comparative calculations highlighted that VE devices
can have important and efcient mitigating role in the global
dynamic behaviour of the studied faade, if properly calibrated.
VE curves collected in Fig. 16, for example, are proposed for a

faade equipped by viscoelastic connectors having an elastic

stiffness kd = kcable, with kcable the axial stiffness of the bearing
cables. Based on a preliminary estimated stiffness kd and on
Eq. (4), the corresponding damping ratio cd was calculated by
taking into account the fundamental pulsation of the studied
faade, with g = 0.6 for the rubber compound [32].
In this context, it should be noticed that the optimal ratio
between the device stiffness and the cable stiffness was found to
be close to the unit. In the case of the single-way supported faade
investigated in [32], for example, the optimal kd/kcable ratio was
40, due to the typically different dynamic behaviour of that
glazing system.
Certainly, VE devices add high damping capabilities in the
faade, hence resulting in markedly reduced tensile stresses in the
glass panels (e.g. Fig. 16c, where the tensile stresses at the point
support of the central glass panel are proposed) and consequently
lower axial forces in the cables (Fig. 16b).
The damping contribution offered by the sliding mechanism of
VE layers strongly modies the typical energy balance of the traditional, conventionally restrained glazing system (Fig. 17). Since the
glass panels are the rst component of the faade that is exposed to
the incoming explosion, the viscous dissipation offered by VE
devices (Fig. 17a) primarily manifests in a reduced kinetic energy

287

0.6

120

0.4

100

0.2

80

Axial force [kN]

Displacement [m]

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

0.0
-0.2

No devices
VE
HRP
VE+HRP

-0.4

60
40

No devices
VE
HRP
VE+HRP

H0

20
0

-0.6
2.00

2.25

2.50

3.00

2.75

3.25

3.50

2.00

2.25

2.50

2.75

Time [s]

Time [s]

(a)

(b)

3.00

3.25

3.50

50

No devices
VE
HRP
VE+HRP

Tensile stress [MPa]

40

30

20

10

0
2.00

2.25

2.50

2.75

3.00

3.25

3.50

Time [s]

(c)
Fig. 16. (a) Maximum deections, (b) axial forces in the cables and (c) tensile stresses at the corner of the central glass panel (ABAQUS/Standard, M03 FE-model).

stored by the glass panels (Fig. 17b), and consequently in a limited

strain energy transferred to the cables (Fig. 17c). This energy balance improvement results in 50% decrease of maximum stresses
in glass (Fig. 16c) and 20% reduction of maximum axial forces
transmitted to the cable-net (Fig. 16b).
5.1.2. Frictional devices
A further solution was successively numerically investigated
and frictional dissipative devices were applied at the end of the
horizontal cables only (HRP). Based on preliminary analytical estimations (Table 1), the sliding force of frictional connectors was set
to Fs = 80 kN, that is two times the initial pretension force and half
their characteristic tensile strength HR.
As expected, simulations highlighted that the introduction of
mono-directional (HRP) frictional devices strongly modies the
global dynamic response of the studied glazing system. The
cable-net is in fact partly deprived by the horizontal bearing cables,
and after the activation of frictional devices the vertical cables
must sustain the faade.
As shown in Figs. 16 and 17, however, this technical solution
does not provide efcient improvement in the dynamic response
of the traditional faade. Since HRP devices mainly act at the level
of the cable-net only, as expected, maximum tensile forces in the
glass panels are not positively affected by the presence of the HRP
mechanism (Fig. 16c; 1% decrease of maximum tensile stresses).
In terms of maximum forces in the cable-net, the increase of pretension in the vertical cables is partly reduced. In this case, due to

activation of HRP devices, its nal value is in fact almost equal to

the pre-established sliding value Fs (Fig. 16b; 23% decrease of
maximum axial forces). At the same time, the partial release of
the horizontal cables due to the activation of HRP devices allows
to further increase the exibility of the glazing system, as it can
be shown from Fig. 16a and b (17% increase of fundamental period of vibration). Detailed investigation of numerical predictions
highlighted that frictional devices activate almost simultaneously,
when the explosion invests the faade, and their measured average sliding, although moderate (0.003 m for the devices at the
top and at the base of the faade, 0.01 m for the central cables
subjected to maximum overall out-of-plane deections), allows
to dissipate by friction up to half the total incoming energy
(Fig. 17a).
5.1.3. Combined viscoelastic and frictional devices
In conclusion, both VE and HRP were applied to the examined
glazing system. Based on previous investigations performed on
the faade equipped by VE devices or HRP devices only, the main
input parameters of both the mechanisms were set respectively
equal to kd = kcable and Fs = 2H0.
As highlighted from Figs. 16 and 17, the efciency of both the
devices is maximized, being part of the incoming blast energy dissipated both by viscous damping and friction phenomena.
The interaction of multiple devices, consequently, manifests in
marked reduction of tensile stresses in the glass panels (50%),
as well as in mitigation of blast effects in the bearing cables

288

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

1.2x105

1.2x10 5
Total incoming energy
VE, viscous energy
HRP, friction energy
VE+HRP, viscous energy
VE+HRP, friction energy

8.0x10 4

Total incoming energy

No devices
VE
HRP
VE+HRP

1.0x105

Kinetic energy [J]

Dissipated energy [J]

1.0x10 5

6.0x10 4
4.0x10 4
2.0x10 4

8.0x104
6.0x104
4.0x104
2.0x104

0.0x10 0

0.0x100
2.00

2.25

2.50

3.00

2.75

3.25

3.5

2.00

2.25

2.50

2.75

Time [s]

Time [s]

(a)

(b)

3.00

3.25

3.5

5.0x10 4

Strain energy [J]

4.0x10

No devices
VE
HRP
VE+HRP

3.0x10 4

2.0x10 4

1.0x10 4

0.0x10 0
2.00

2.25

2.50

2.75

3.00

3.25

3.50

Time [s]

(c)
Fig. 17. (a) Dissipated energy; (b) kinetic energy in the faade and (c) strain energy in the cable-net (ABAQUS/Standard, M03 FE-model).

(30%), being the faade subjected due to VE devices to a

reduced impulsive load.
Certainly, a proper calibration of combined VE and RP devices,
as well as a correlation of their input parameters to the faade
mechanical and geometrical properties, is fundamental for the
optimization of their structural effectiveness. In any case, the discussed results highlighted through application to a case study
the potentiality of the examined solutions. It is thus expected that
critical discussion of comparative simulations could provide an
appropriate background for further extended developments and
applications.

6. Summary and conclusions

In the paper, the structural behaviour of a glazing faade supported by a two-way straight pretensioned cable-net has been
investigated through advanced numerical models. The faade
object of study, experimentally analyzed under seismic loads in
earlier contributions, has been in this work subjected to high-rate
explosive loads. As known, impulsive pressures due to explosions
typically involve in the invested structural systems extreme deformations and stresses, being these glazing faades generally
designed to resist ordinary loads only.
Prior to development of advanced nite-element (FE) numerical
models, simplied analytical SDOF formulations based on energy
considerations have been proposed and discussed, in order to per-

form a preliminary estimation of maximum expected effects in the

studied system.
Dynamic nonlinear simulations have then been performed on
appropriate nite-element (FE) numerical models calibrated to
experimental dynamic parameters recalled for the same faade
from earlier contributions of literature.
Careful consideration has been paid, during the numerical modeling phase, to the assessment of possible effects due to interaction
between glass panels and cables, when subjected to dynamic
impulsive loads. For this purpose, three different FE-models have
been presented and compared. The rst one (M01), being a geometrically simplied lumped-mass model, provided only approximate estimations of the expected maximum effects in the cables,
due to the lack of glass panes. In any case, comparison of M01 predictions to SDOF analytical calculations generally manifested
rather good agreement between them, hence conrming the validity of the approach. Further FE models able to include the bending
stiffening contribution of the glass panels have been successively
implemented. In the case of the M02 model, a geometrically
detailed description of cables, glass panels and connectors has
been provided. Successively, in order to reduce the computational
weight of model M02 but preserve its accuracy, a third FE-model
(M03) has been developed. In this latter case, due to the approximate description of glass panels in the form of well-calibrated,
equivalent strips, a marked increase of computational efciency
was obtained, as well as a rather appreciable agreement between
M02 and M03 dynamic predictions of maximum effects of blast
in the main faade components. Based on preliminary investiga-

C. Bedon, C. Amadio / Engineering Structures 79 (2014) 276289

tions, extended parametric dynamic analyses have been consequently performed on M03 FE-models.
Three typologies of devices have been introduced in the investigated faade, namely at the connection between glass panels
and cables, as well as at the cable ends.
The structural efciency under explosive loads of possible (i)
viscoelastic connectors, (ii) frictional devices at the cable ends
and (iii) combined multiple (i) and (ii) devices has been highlighted by means of advanced parametric dynamic nonlinear analyses. Although further extended studies are required, exploratory
investigations emphasized the high potentiality of the proposed
dissipative systems, hence suggesting possible successful applications in glazing faades sustained by two-way straight cable-nets
and subjected to high-strain blast pressures.
References
[1] Br L, Geiler B. MULTITERM-Fassade ein modulares Stahlfassadensystem
fr kombinierte Anwendungsflle. Stahlbau 2003;72(12):8338.
[2] SCI Steel Knowledge P396. New Beijing Poly Plaza Cable-Net Wall Structural
Stainless Steel Case Study 0.9; 2011 <www.steel-sci.org>.
[3] Sobek W, Straub W, Schmid A. Horizon Serono Konstruktion des weltweit
grten zu ffnenden Glasdaches und der darunterliegenden Forumfassade.
Stahlbau 2009;78(1):110.
[4] Kirchhoff J, Vaglio J, Patterson M. Double-Skin cable-net faade case study:
loyola information commons. In: Proceedings of BESS2010: high performance
building enclosures practical sustainability symposium at cal-poly Pomona;
2010.
[5] Sobek W, Weiss I. Eine Stahl-Glashalle als neuer zentraler Eingangsbereich der
Universitt Bremen. Stahlbau 2001;70(10):30913.
[6] Eekhout M, van de Rotten P. Cable stayed glass faade in the market hall
Rotterdam. In: Proceedings of challenging glass 4 & COST action TU0905 nal
conference, Lausanne, Switzerland; 2014.
[7] Schlaich J, Schober H, Moschner T. Prestressed cable net faades. Struct Eng Int
2005;15(1):369.
[8] Schobe H, Scheider J. Developments in structural glass and glass structures.
Struct Eng Int 2004;14(2):846.
[9] SOM-Skidmore Owings & Merrill (www.som.com). Rocker Faade Support
System, Poly Corporation Headquarters, Beijing. <http://www.ctbuh.org/
TallBuildings/FeaturedTallBuildings/FeaturedTallBuildingArchive2013>.
[10] Schlaich Bergermann und Partner, Screenwall Hotel Kempinski. <http://
www.sbp.de/en/build/show/9-Screenwall_Hotel_Kempinski>.
[11] Vyzantiadou MA, Avdelas AV. Point xed glazing systems: technological and
morphological aspects. J Constr Steel Res 2004;60(8):122740.
[12] Saitoh M, Okada A, Imamura R. Study on glass supporting system pinched at
corner structural characteristics and structural design method. In: Proceedings
of international symposium on theory design and realization of shell and
spatial structures; 2001.
[13] Feng RQ, Wu Y, Shen SZ. Working mechanism of single-layer cable net
supported glass curtain walls. Adv Struct Eng 2007;10(2):18395.
[14] Feng RQ, Ye JH, Wu Y, Shen SZ. Nonlinear response spectra of cable net faades.
Soil Dyn Earthquake Eng 2011;32:7186.
[15] Feng RQ, Zhang LL, Wu Y, Shen SZ. Dynamic performance of cable net facades. J
Constr Steel Res 2009;65(12):221727.
[16] Chan SL, Liu Y, Lee A. Nonlinear analysis of pre-tensioned glass wall faade by
stability function with initial imperfection. Front Arch Eng China
2010;4(3):37682.
[17] Weng ZJ, Yin LF, Shan J, Tang G, Huang W. Static analytical method of singlelayer plane canle net structures under bidirectional uniform load: basic
control equations and logarithmic solution. Eng Mech 2012;29(12):721.
[18] Mesda Y. Analytical study of the cable-truss systems on the glass curtain walls
with vertical uses. Engineering 2013;5:81926.
[19] Wedding WC, Lusk BT. Novel method to determine blast resistant glazing
system response to explosive loading. Measurement 2012;45:14719.
[20] Zhao S, Dharani LR, Liang X. Analysis of damage in laminated architectural
glazing subjected to blast loading. Adv Struct Eng 2008;11:12934.
[21] Dharani LR, Wei J. Dynamic response of laminated glass panels subjected to
blast loading. Structures under shock and impact VIII, Crete. Greece
2004;15:18190.

289

[22] Morison C. The resistance of laminated glass to blast pressure loading and the
coefcients for single degree of freedom analysis of laminated glass. PhD
Thesis, Craneld University; 2007.
[23] Larcher M, Solomos G. Laminated glass loaded by air blast waves
experiments and numerical simulations. Joint Research Centre Technical
Note, Pubsy JRC57559; 2010.
[24] Larcher M, Solomos G, Casadei F, Gebbeken N. Experimental and numerical
investigations of laminated glass subjected to blast loading. Int J Impact Eng
2012;39:4250.
[25] Hooper PA, Sukhram RAM, Blackman BRK, Dear JP. On the blast resistance of
laminated glass. Int J Solids Struct 2012;49:899918.
[26] Weggel DC, Zapata BJ. Laminated glass curtain walls and laminated glass plies
subjected to low-level blast loading. J Struct Eng 2008;134:46677.
[27] Zhang X, Hao H, Ma G. Parametric study of laminated glass window response
to blast loads. Eng Struct 2013;56:170717.
[28] Bedon C, Amadio C, Sinico A. Structural stability of compressed monolithic and
laminated glass elements under blast loads. In: Proceedings of glasscon global
conference innovation in glass technology, 710 July 2014, Philadelphia,
Pennsylvania Convention Center, USA, pp. 3452 (USB Flash Drive); 2014.
[29] Teich M, Warnstedt P, Gebbeken N. The inuence of negative phase loading on
cable net faade response. J Arch Eng 2011. http://dx.doi.org/10.1061/
(ASCE)AE.1943-5568.0000083 (in print).
[30] Wellershoff F, Lori G, Zobec M, Osterland K. Structural design of blast enhanced
cable net facades. In: Proceedings of COST Action TU0905 Mid-term
Conference on Structural Glass, Porec, Croatia; 2013.
[31] Amadio C, Bedon C. Elastoplastic dissipative devices for the mitigation of blast
resisting cable-supported faades. Eng Struct 2012;39:10315.
[32] Amadio C, Bedon C. Viscoelastic spider connectors for the mitigation of cablesupported faades subjected to air blast loads. Eng Struct 2012;42:190200.
[33] Amadio C, Bedon C. Multiple dissipative devices for blast-resisting cablesupported glazing faades. Model Simul Eng, vol. 2013, article ID 964910;
2013.
[34] GSA-TSO1-2003. Standard test method for glazing and window systems
subject to dynamic overpressure loadings. U.S. General Service
Administration; 2003.
[35] Goel MD, Matsagar VA, Gupta AK, Marburg S. An abridged review of blast wave
parameters. Defence Sci J 2012;62(5):3006.
[36] Bulson P. Explosive loading of engineering structures. A history of research and
a review of recent developments. London: E&F Spon; 1997.
[37] TM 5-1300. The design of structures to resist the effects of accidental
explosions. U.S. Dept. of the Army, the Navy and the Air Force, Technical
Manual, Washington DC; 1990.
[38] Larcher M, Hermann N, Stempniewsky L. Explosionssimulation leichter
Hallenhllkonstruktionen. Bauingenieur 2006;81(6):2717.
[39] Dassault Systmes Simulia Corp. ABAQUS ver. 6.9 Users Manual; Providence,
RI, USA; 2009.
[40] Schleyer FK. Tensile structures analysis of cables, cable nets and cable
structures. MIT Press; 1969.
[41] Wang Z, Gong G, Zhang Y, Bai C. Simulation of the response of glass window
under blast load. Transactions of Tianjin University, 14: 50410. Tianjin
University and Springer-Verlag; 2008.
[42] Bedon C, Louter C. Exploratory numerical analysis of SG-laminated reinforced
glass beam experiments. Eng Struct 2014;75(9):45768.
[43] EN 572-2. Glass in buildings basic soda lime silicate glass products. CEN;
2004.
[44] MoD/DE Functional Standard Design and Maintenance Guide 02. Glazing
standards for MoD buildings subject to terrorist threat, Ministry of Defence;
1996.
[45] Meyers GE. BLASTOP version 1 4, US Department of Energy; 1994.
[46] Mayor RP, Flanders R. Technical manual simplied computer model of air blast
effects on building walls. US Department of State, Ofce of Diplomatic
Security, Washington DC; 1990.
[47] Morison C. The resistance of laminated glass to blast pressure loading and the
coefcients for single degree of freedom analysis of laminated glass. PhD
Dissertation, Craneld University; 2007.
[48] Peroni M, Solomos G, Pizzinato V, Larcher M. Experimental investigation of
high strain-rate behaviour of glass. Appl Mech Mater 2011;82:638.
[49] Bachmann H, Ammann WJ, Florian D. Vibration problems in structures:
practical guidelines. Birkhuser; 1995.
[50] Buchholdt HA. An introduction to cable roof structures. 2nd ed. Thomas
Telford; 1999.
[51] Timoshenko S, Woinowsky-Krieger S. Theory of plates and shells. McGraw-Hill
Book Company; 1959.