Buckling Verification of Laminated Glass Elements
in Compression
Claudio Amadio1 , Chiara Bedon2
Depart ment of Civ il Engineering and Architecture, University of Trieste
Piazzale Europa 1, 34127 Trieste, Italy
1
amadio@units.it; 2 chiara.bedon@phd.units.it
Abstract- Because of the characteristic of high slenderness
ratios, monolithic and laminated glass elements are frequently
subjected to buckling phenomena. As regards laminated glass
beams and panels, in particular, the effects of possible
temperature or time-load variations represent only some
aspects that make complex their global structural response. In
this context, the paper focuses on the load-carrying behavior of
in-plane compressed laminated glass elements. In it, some
analytical formulations are presented to describe realistically
their typical behavior. As shown, the proposed formulations
are in good agreement with experimental and numerical data
available in literature. At the same time, they allow to perform
a rational buckling verification of such brittle load-bearing
elements. Finally, according to the suggestions the Eurocodes
give for the verification of traditional structural elements, a
series of buckling curves opportunely calibrated are proposed
to guarantee the requisites of resistance, serviceability, and
durability typically imposed in the design of conventional
structural systems.
Keywords- Laminated Glass; In-plane Compressed Beams
and Panels; Sandwich Elements; Equivalent Thickness;
Buckling Curves
I. INT RODUCTION
The use of monolithic or laminate glass (LG) elements in
modern and innovative architectural applications showed a
strong increase in the last years. Because of aesthetic,
lighting, and architectural advantages, glass elements are
frequently used as structural components able to sustain
loads. However, the real capabilities of such innovative
bearing co mponents are currently not well known and
several aspects related to their typical load-carry ing behavior
are very comp lex to evaluate. The load-carry ing capacity of
LG beams or panels, for examp le, strongly depends on the
degradation of the mechanical p roperties of the interlayer, as
well as on the presence and the amplitude of possible
imperfections, or the presence of additional external loads.
In this context, several authors observed that temperature
variations could strongly influence the mechanical properties
of the thermop lastic materials usually adopted to bond
together the glass panes [1, 2, and 3]. Recently, nu merous
authors focused on the buckled response of structural g lass
elements in several boundary or loading conditions. Luible
[4], for examp le, investigated the load-carry ing behavior of
LG beams or panels and performed experimental tests to
analyze their typical response. Belis [5] studied the lateral-
torsional buckling of LG beams. In [6, 7], the authors
proposed an exact analytical approach for the buckling
verification of LG beams in co mpression or in out-of-p lane
bending. The behavior of LG panels under in-p lane
JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101
- 90 -
compression or shear has been deeply examined in [8, 9, and
10].
Rationally, a realistic buckling verificat ion of LG
compressed beams and panels should be performed by
contemporarily satisfying a series of requisites concerning
the resistance, the serviceability, and the durability of such
brittle elements. At the same t ime, the effective connection
offered by the interlayer should be precisely estimated, since
strongly time-load and temperature dependent. Nevertheless,
consolidate verification criteria available in literature for the
buckling verification o f trad itional structural elements (steel,
for example [11]) cannot be directly applied to LG elements.
Because of this reasons, in the paper some analytical
formulat ions are proposed for the buckling verification of in-
plane compressed LG beams and panels. The aim is to
derive simple and consistent design rules for pane-like g lass
columns with laminated sections subjected to axial
compressive loads. As shown, these analytical models are in
good agreement with nu merical and experimental data.
In addition, according to the suggestions of Eurocodes 3,
4, 5 [11, 12, and 13], a series of buckling curves opportunely
calibrated are proposed to guarantee the requisites of
resistance, serviceability and durability typically required in
the design of conventional structural systems. As proposed
in the following sections, these buckling curves are in good
agreement with experimental data collected in literature, as
well as with nu merical results.
II. ANALYTICAL MODEL FOR COMPRESSED LAMINATED
GLASS BEAMS
The analysis of co mpressed laminated glass beams is
generally performed by using the elastic theory of sandwich
elements [14]. Recently, based on these sandwich
formulat ions, an equivalent thickness approach has been
proposed to study the behavior of a monolith ic co mpressed
element characterized by a rectangular cross section of
thickness teff [4]. As a result, the flexu ral behavior of the
composite beam is described through the classical theory of
deflection. This analytical formu lation undoubtedly
constitutes a suitable design method for compressed LG
elements, but does not allow understanding how, depending
on variations of temperature and load duration, the
mechanical properties of the interlayer can influence the
global response of the layered element.
In this last years, also Blaauwendraad [15] proposed a
simp lified formulat ion able to easily control the transition
between the layered limit (abs, absence of connection effective level of connection between the glass panes,
between the glass sheets) and the monolithic limit (full, Blaauwendraad suggests also a unity check for tensile
presence of a rigid connection between the glass sheets). By stresses. Also in this circu mstance,the formulation is simp le,
means of a dimensionless parameter representative of the but provides only approximate results.

Fig. 1 Analytical model for compressed LG beams

The alternative [6], consists in an analytical model based K EJ full , (2)

2 =
on the original elastic theory of Newmark et al. [16], EA* EJ abs
concerning the flexural behavior of 2-layer co mposite beams
with deformable connection. As shown in [6], the advantage Gint b (3)
K=
of this new formu lation consists in the possibility of taking tint
into account the real thickness of the layers constituting the ( EA1 )( EA2 ) Ebt1t 2
EA* = = (4)
LG beam. At the same time, the proposed formulat ion
EA1 + EA2 t1 + t 2
allo ws taking into account the effective level of connection
offered by the interlayer, thus it can be used to estimate the the equivalent axial stiffness,
buckled response of a co mpressed LG beam in a well-
defined condition of temperature and load duration. t t 2 t t
2
(5)
EJ full = EJ abs + Eb t1 1 + int + t 2 2 + int
Although the model applies only to 2-layer structural 2 2 2 2
systems, that is to beams consisting of two interacting glass
sheets, tied together by a shear connection able to transfer the monolithic flexural stiffness,
the horizontal shear fro m one element to the other, it
Eb 3 3 (6)
provides accurate results. EJ abs = (t1 + t 2 )
12
In this context, let us consider a 2-layer LG beam, pinned
the layered flexural stiffness.
at the ends of its buckling length L0 , subjected to an axial
compressive load N (Fig. 1). The beam, having a rectangular Assuming for the init ial sine-shape imperfection of the
cross section (width b), is assumed to be constituted by two composite beam a maximu m amp litude w0 , its total
external glass layers (thicknesses t1 , t2 , elastic Youngs maximu m deflection is:
modulus E, shear modulus G) and a middle interlayer
L
(thickness tint , shear modulus Gint ), and affected by a wmax = w0 + w 0 , (7)
sinusoidal imperfect ion of maximu m amp litude w 0 . An 2
extended experimental campaign performed by Belis et al.
and the corresponding tensile stress max can be estimated as:
[17] on 312 glass beams with variable length, height,
thickness, glass type, recently highlighted that the sinusoidal N Nwmax F . (8)
max = + +
shape describes the initial imperfect ion in monolithic or A Wy Amin
laminated glass beams with a good level of accuracy.
In Eq. (8), F is the axial load acting in each g lass pane,
The transversal displacement w(z) of the simply due to the flexural deflect ion of the beam:
supported beam [w(0)= w (L0 )= 0 and wII(0)= wII(L0 )= 0] due
to the axial co mpression N is [6]: Nwmax + max EJ abs ; (9)
F=
d
( EJ abs L + EJ full ) L Nw0 sin(z L0 )
2 2 2 2
(1)
w( z ) =
0 0
d represents the distance between the centroidal axis of each
2 EJ abs L20 ( EJ full 2 NL20 ) + EJ full 2 ( EJ abs 2 NL20 ) glass pane (Fig. 1); max is the midspan curvature of the
with: beam corresponding to the deflection wmax :

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 91 -
 2 N cr( E ) are
max = 2
wmax ; (10) the figure, also three specific values of
L0 highlighted. These crit ical loads are evaluated in presence of
Wy is the elastic resistant modulus of the total cross section short-term loads (3 seconds) and room/ med iu m temperatures
(thickness t= t1 + tint + t2 ): (T= 20C, Gint = 8.06N/ mm2 ; T= 30C, Gint = 0.971N/ mm2
[18]), or long-term loads (1 year) and h igh temperatures (T=
bt 2 ; (11) 50C, Gint = 0.052N/ mm2 [18]).
Wy =
6
2.00x103
and Amin , with tmin = min(t1 , t 2 ), is the minimu m transversal N(E)cr,full= 1956N
N
area of each glass pane: 1.75x103

Amin = bt min (12) 1.50x103

Eq. (7) represents useful informat ion, since it allo ws 1.25x103

N(E)cr [N]
describing the load N-transversal displacement w of a N
generic LG beam in co mpression, by taking into account the 1.00x103
5/1.52/5mm
b= 200mm, L0= 3000mm
effective shear stiffness Gint of the interlayer and the 2
7.50x10 N(E)cr (Eq.(13))
presence of possible sinusoidal Gint= 8.06N/mm2
imperfections w( z ) = w0 sin(z L0 ) . If the co mpression N 5.00x102 Gint= 0.971N/mm2
Gint= 0.052N/mm2
gradually increases, the loss of stability of the beam 2
2.50x10 N(E)cr,abs= 320N
typically shows in the form of an abrupt and non-
proportional increase of the corresponding displacement 0.00x100
wmax. 10-4 10-3 10-2 10-1 100 101 102 103 104
Gint [N/mm2]
Consequently, Eq.(7) can be used to express the critical
buckling load N cr( E ) of the LG beam, assumed as the Fig. 2 Effects of temperature and time-load conditions on the critical
buckling load N cr( E ) of a compressed LG beam (5/1.52/5mm, b=200mm x
asymptotical value N to wh ich the growing displacement
L0 = 3000mm) [18]
wmax tends:
2 EJ abs EJ full 2 L20 + 2 . (13) III. NUMERICAL VALIDATION
N cr( E ) =
L 2 2 EJ L2 + EJ 2
0 abs 0 full To validate the analytical approach proposed for the
buckling analysis of compressed LG beams, two different
Depending on the shear modulus Gint of the material
nonlinear finite-element (FE) models were constructed with
constituting the interlayer, the critical load N cr( E ) of a generic
the commercial nonlinear code ABA QUS [19]. In the first
LG beam is always co mprised between the well-known limit three-dimensional (3 D) FE-model, the glass panes and the
values N cr( E, )full (monolithic limit, Gint , that is ) middle PVB-film have been described by means of 3D
and N cr( E,abs
)
(layered limit, Gint 0, that is 0): eight-node elements. In similar models, it is important to
define a sufficiently accurate mesh for the elements, since
2 EJ full , the convergence of simulations as well as the accuracy of
N cr( E, )full = (14) numerical results may be seriously compro mised. Because
L20
of this reason, two elements over the depth of each g lass
2 EJ abs . sheet and the interlayer have been used. At the same time, an
N cr( E,abs
)
= (15)
opportune mesh has been applied in the width of the
L20
examined LG beams (Fig. 3).
This finding constitutes an important aspect in the
analysis of LG elements, sine the material commonly used to
bond together the glass sheets (PVB Butacite, SG , EVA,
etc.) consists in thermoplastic materials strongly temperature
and load time-dependent. PVB-films, in particular, have
good mechanical properties if subjected to room
temperatures or short-term loads, but present a strong
degradation of shear stiffness with high temperatures and
long-term loads [18].
In Fig. 2, for example, the effects of stiffness degradation
on the value of the critical buckling load N cr( E ) are proposed
for a PVB-laminated glass beam (5/1.52/5mm) having
dimensions b=200mm x L0 = 3000mm. In particular, N cr( E ) is
evaluated by means of Eq. 13 by assuming Gint a value
comprised in the range 10-4 N/ mm2 < Gint < 104 N/ mm2 . In

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 92 -
 buckling analyses performed with the Mshell FE-model do
not require long processing time, but the accuracy of results
is co mparable to that of the 3D FE-model only in p resence
of sufficiently rig id interlayers (Gint > 10N/ mm2 ). As shown
in Fig. 4, the MShell FE-model clearly tends to
underestimate the effective critical load N cr( E ) of LG beams
if the interlayer is soft. This aspect should be taken into
account in the analysis of LG beams assembled with PVB-
films, which are strongly time and load-time dependent.
2.00x103
N(E)cr,full= 1956N
1.75x103 N

Fig. 3 3D and Mshell numerical models for LG beams in compression 1.50x103

(ABAQUS)
1.25x103
The external glass sheets and the PVB-film were

N(E)cr [N]
connected together by using the same nodes. To avoid 1.00x103
N
possible eccentricities, boundaries were applied at the
central nodes of the PVB-film, at both the ends of each 7.50x10
2 5/1.52/5mm
b= 200mm, L0= 3000mm
simp ly supported LG beam. The co mp ressive axial load was N(E)cr (Eq.(13))
introduced in the FE-model in the form of uniformly 5.00x102 ABAQUS (3D)
distributed pressure acting on the lower and upper surfaces ABAQUS (MShell)
of 3D elements. 2.50x10
2
N(E)cr,abs= 320N

The second and simp lest FE-model (MShell) consists in

0.00x100
mu ltilayer co mposite shell elements (S4R) able to describe
the real flexu ral stiffness and the effective th icknesses of the 10-4 10-3 10-2 10-1 100 101 102 103 104
2
Gint [N/mm ]
layers constituting the analyzed LG beams (Fig. 3). In this
specific circu mstance, the axial co mpression was described Fig. 4 Comparison between analytical and numerical (ABAQUS, 3D and
in terms of uniform co mpressive shell edge loads acting on Mshell FE-models) critical loads of LG beams in compression
the lower and upper edges of each LG beam. Further co mparisons were performed to validate the
Concerning the materials, in both the FE-models g lass proposed analytical procedure [6]. Results proposed in Fig. 5,
has been modeled as an isotropic, linear-elastic material for example, refer to a LG beam obtained by assembling two
characterized by Youngs modulus E= 70000N/ mm2 and 5mm thick glass sheets and 1.52mm thick PVB film.
Poissons ratio v= 0.23. Also PVB has been described as a 1.50
linear elastic material, characterized by equivalent Full
Analytical
mechanical properties able to take into account for the ABAQUS
degradation of its shear stiffness Gint due to temperature or 1.25 Abs
Analytical
load-time variat ions [6, 18]. In all the performed simulat ions, ABAQUS
Poissons ratio for PVB was fixed equal to vint = 0.498 [18]. 1.00 Temperature: 20C
Load: 3s
Firstly, parametric buckling analyses were performed Analytical
ABAQUS
with the 3D and Mshell FE-models to investigate the effects
N
R

0.75
of mechanical (PVB stiffness) or geometrical properties
(slenderness of the beam, ratio between the thicknesses of
glass and interlayer) in the buckling response of LG beams 0.50
in compression. In this simu lation phase, also the accuracy
of the simp lest Mshell FE-model was checked.
0.25
The main results are proposed in Fig. 4, in the form of N

critical loads N cr( E ) of a 5/1.52/5mm beam (b = 200, L0 =

0.00
3000mm) characterized by PVB-interlayers of various
0 50 100 150 200 250 300 350
stiffness (10-4 N/ mm2 < Gint < 104 N/ mm2 ). Analytical results = L0 / full
(Eq. (13)) are co mpared with numerical pred ictions.
Fig. 5 Comparison between analytical and numerical (ABAQUS) results for
As shown, the 3D FE-model provides results in good LG beams in compression (w0 = L0 /500).
agree with analytical calculations. Nevertheless, if the mesh
The width b of the beam was fixed equal to 20mm and
of the 3D FE-model is not sufficiently detailed, the obtained
its buckling length L0 was modified in a pre-established
critical load N cr( E ) could be overestimated, especially in range (L0 = 200, 400, 600, 800, 1000mm). The maximu m
presence of soft interlayers (Gint < 10N/ mm2 ). Contrarily, amp litude w0 = L0 /500 of the init ial imperfection was

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 93 -
introduced in the model as an init ial sine-shape imposed
displacement. In this specific circumstance, Gint was
assumed equal to Gint = 8.06N/ mm2 , as suggested by
Bennison (T= 20C, load duration: 3 seconds [18]). At the
same time, to simulate the presence of an ext remely rigid
interlayer (monolithic limit, full) as well as a soft interlayer
(layered limit, abs) the shear moduli Gint,full = 500N/ mm2 and
Gint,abs = 0.0001N/ mm2 were considered.
Static incremental analyses were performed with 3D FE-
model to investigate the load-carrying behavior of imperfect
LG beams subjected to an increasing compression N.
Geo metrical nonlinearity was taken into account to simu late
the buckling response of the examined LG elements. As the
axial co mpression growth, the maximu m tensile stress
occurring in the glass surface was monitored. Analyses were
stopped at the reaching of a prefixed value Rk of
characteristic tensile strength for glass. In the specific
example, a characteristic strength Rk = 17 N / mm 2 was
taken into account, as suggested in [20] for the verificat ion
of float glass elements subjected to long-term loads. For
each LG beam, the value of the compressive load N *
associated to a maximu m tensile stress max ( N * ) = Rk was
collected. Analytical and numerical results are presented in
Fig. 5 in terms of reduction factor R:
N* ,
R=
Rk A
with A = bt the total cross-sectional area, and full
slenderness ratio :
L0 ,
= different conditions, respectively referred to requirements of
full
with full the radius of gyration of the total cross section
(hypotheses of rigid connection between the glass panes):
J ,
full =
A
and J = bt 3 12 the mo ment of inertia.
Also in this circumstance, analytical and numerical
results are in good agreement.
IV. ADDITIONAL LOADS
Clearly, the proposed analytical model is able to provide
accurate results, agreeing well with the nonlinear numerical
results. Moreover, although applicable only to 2-layer
simp le structural systems, the same formu lation can be
easily applied to the analysis of LG beams subjected to
compressive loads N and to simu ltaneous transversal loads q.
As discussed in [6], in fact, the maximu m deflection of the
LG beam can be estimated as:
( 2 EJ abs L20 + EJ full 2 )( L20 q + 8 Nw0 )
wmax =
8[ EJ abs L ( EJ full NL ) + EJ full ( EJ abs NL )]
2 2
0
2 2
0
2 2
in which q is the maximu m amp litude of the transversal
sinusoidal load affecting the beam; w0 represents the
JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101
maximu m amp litude of the sine-shape imperfection; , EJfull ,
EJabs are given by Eqs. 2, 5, and 6.
In this specific circu mstance, the maximu m tensile
stresses associate to the deflection wmax is:
N Nwmax qL20 F , (20)
max = + +
A Wy 8W y Amin
with:
( Nwmax + qL20 ) + max EJ abs , (21)
F=
d
whereas the critical buckling load N cr( E ) of the LG beam can
be still evaluated by means of Eq. 13.
V. DESIGN CRIT ERIA FOR COMPRESSED LAMINATED GLASS
BEAMS
In the previous sections, it was shown that the maximu m
transversal displacement wmax of a compressed LG beam
caused by an axial load N can be evaluated by means of Eq.
(7). At the same time, the corresponding maximu m tensile
stress can be estimated by means of Eq. (8). If additional
external loads perpendicular to the plane of the beam are
present (sinusoidal loads of maximu m amp litude q or,
equivalently, uniformly distributed loads of amplitude q),
(16) the corresponding deflection and mid-span tensile stress are
given by Eqs. (19) and (20).
In this context, in accordance with the Limit State
approach, the buckling verification of co mpressed LG beams
should be developed by contemporarily satisfying three
(17)
structural resistance, serviceability, and durability. Moreover,
to perform a reasonable verificat ion, the presence of an
initial sinusoidal imperfect ion of maximu m amplitude w0
should always be taken into account, to represent the
possible effects of geometrical deformat ions (residual
(18)
stresses, geometrical imperfections due to fabrication) or
eccentricities (of load or boundary, or a co mbination of them)
in the beam. Rationally, as suggested by Belis et al. [17], the
amp litude of the init ial sine-shape imperfect ion w0 should be
at least assumed equal to 1/400 of the structural span.
In this context, the maximu m tensile stress Ed in each
cross section (Eq. (8) or Eq. (20)), if additional transversal
loads are present) should be compared with the design
tensile resistance of glass Rd :
Ed Rd (22)
At the same time, the maximu m deflection max of the
LG beam (Eq. (7) or Eq. (19)) might not exceed a specific
value, defined as a ratio of its buckling length L0 , as for
example:
L0 (23)
max
+ w0 k
2
0 with k= 120.
(19)
Finally, the colu mn buckling verification should require
the comparison of the design compressive load N Ed and the
buckling resistance Nb,Rd of the compressed LG beam:
- 94 -
 N cr( E ) , (24) 1 1 1
N Ed N b , Rd = 2 + 1 + = 0, (30)
M1
2

2

2

with N given by Eq. (13) and M1 = 1.40, for examp le, a

(E)
cr
where:
buckling safety factor. A Rk is a normalized slenderness and (31)
=
N cr( E )
A. Alternative Verification Criteria for Compressed
Laminated Glass Beams N is a buckling reduction factor. (32)
=
A Rk
For the buckling verification of tradit ional structural
elements, constructed with conventional materials as steel or The main advantage of the analytical approach proposed
concrete, consolidate verificat ion criteria are availab le in by Ayrton and Perry consists in the definition of an
literature. The Ayrton-Perry formu lation [21], fo r examp le, equivalent init ial sine-shape imperfection representative of
was originally formu lated for the analysis of geometrically geometrical imperfect ions, as well as residual stresses or
imperfect colu mns loaded by uniform co mpressive loads. In possible eccentricit ies. Because of this reason, it actually
accordance with this analytical approach, the initial constitutes the theoretical background of the colu mn design
imperfections, as well as other effects (residual stresses, curves proposed by several codes for the verificat ion of
possible eccentricities) can be efficiently described through compressed steel members.
a generalized imperfection factor.
With reference to Eq. (30), for examp le, the Eurocode 3
Specifically, the maximu m second order lateral [11] estimates the buckling resistance of co mpressed steel
displacement of a co mp ressed LG beam affected by a members by taking into account a series of appropriate
sinusoidal imperfection of maximu m amp litude w0 can be buckling curves, opportunely calibrates to take into account
expressed as: the effects in their buckled response of possible initial
w0 (25) imperfections of different amplitude, as well as the different
wmax =
N cross-section class. Nevertheless, this method cannot be
1 (E)
N cr directly applied to LG beams.
with N cr( E ) the elastic critical load (Eq. (13)) and N the In this context, the buckling verification o f a co mpressed
applied co mpression. LG beam can be still performed by satisfying the condition
given by Eq. (24), in wh ich the design buckling resistance
If Rk is the characteristic tensile strength of glass, the Nb,Rd of the layered beam is:
failure of the beam occurs when:
N b , Rd = A Rd , (33)
N Nwmax
+ = Rk , (26)
A Wy with an opportune reduction factor.

with A the cross-sectional area and Wy the elastic resistant In Eq. (33), the reduction factor can be estimated by
modulus (Eq. (11)). means of the expression obtained by Eq. (30) and suggested
The substitution of Eq.(25) into Eq. (26) p rovides the by the Eurocode 3 [11]:
well-known exp ression [21]: 1 , with 1 (34)
=
2
( Rk max ) ( cr( E ) max ) = max cr( E ) , (27) + 2

where: where:
2
max is the maximu m tensile stress due to the applied = 0.5 [1 + imp ( 0 ) + ] , (35)
compression N ,
is the slenderness of Eq. (31),
N (E) (28)
(E)
cr = cr imp = 0.71,
A
is the critical stress of the beam, with N cr( E ) given by Eq.(13), 0 = 0.60 .
and
A In this case, the imperfect ion coefficients imp and 0
=w (29)
0
Wy have been opportunely calibrated on the basis of numerical
and experimental data available in literature for co mpressed
is a generalized non-dimensional imperfection factor. monolithic or laminated glass beams [4, 22, 23]. In the
Eq. (25) is the analytical expression able to represent the specific, the coefficient 0 indiv iduates the values of
relationship between the applied co mpressive load N, the slenderness associated to a reduction factor equal to
Eu ler critical load N cr( E ) and the equivalent in itial deflection = 1 . The value of the coefficient imp individuates the
of maximu m amp litude w0 for the comp ressed LG beam. Eq. maximu m allowable imperfect ion for the co mpressed
(25) can also be written in the standard form: member.

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 95 -
 1.4
In Fig. 6, several nu merical and experimental data are
Euler
presented to validate the proposed verification approach. In EC Curve, e= 0 (imp= 0.71, 0= 0.60)
particular, experimental data refer to buckling tests recently 1.2
EC Curve, e= t/6 (imp= 1.80, 0= 0.40)
performed by Luible [4] and Aiello et al. [22] on monolithic ABAQUS (w0= L0/400, e= t/6)
or laminated glass beams in co mpression. Additional 1.0
experimental data are collected in [23]. Glass beams present
a typical brittle-elastic behavior, and due to the absence of 0.8
post-critical resistance in them, an abrupt failure generally

occurs. Because of this reason, the safety factors should be
0.6
carefully calibrated.
1.4
Euler 0.4
EC Curve (imp= 0.71, 0= 0.60)
1.2 M1= 1.40
0.2
Test [4]
Test [22]
1.0
Test [23] 0.0
ABAQUS (w0= L0/400)
0.0 0.5 1.0 1.5 2.0 2.5
_
3.0 3.5 4.0 4.5 5.0
0.8

0.6 Fig. 7 Buckling curves for the verification of LG beams in compression

(e= t/6), comparisons with numerical results (ABAQUS)
0.4
VI. ANALYTICAL MODEL FOR IN-PLANE COMPRESSED
LAMINATED GLASS PANELS
0.2

Laminated glass panels, as well as beams, are largely

0.0 used in the realizat ion of faades, roofs, stiffeners, etc. The
0.0 0.5 1.0 1.5 2.0 2.5
_
3.0 3.5 4.0 4.5 5.0 LG panels are main ly associated to the realizat ion of
futuristic and innovative architectures in modern buildings.
However, due to their typical slenderness, they can be
Fig. 6 Buckling curves for the verification of LG beams in compression (e=
0), comparisons with numerical (ABAQUS) and experimental results [4, 22, affected by buckling problems. Consequently an appropriate
23] verification criterion should be adopted to prevent possible
failure mechanis ms.
In the same figure, also numerical results are reported,
obtained by performing in A BAQUS (3D FE-model) a series Recently, nu merous authors focused on the buckled
of static incremental analysis on compressed glass beams response of glass panels subjected to in-plane compression
characterized by different geometrical propert ies. As [4, 8] or in-p lane shear [9], providing useful experimental
suggested by Belis et al. [17], in this specific circu mstance data, sophisticated numerical simulat ions, and interesting
an init ial sine-shape imperfect ion of maximu m amplitude analytical considerations. In [10], an equivalent thickness
w0 = L0 /400 was taken into account. As shown in Fig.6, the approach has been proposed to study the buckling response
buckling curve proposed for the verification of co mpressed of LG panels in several boundary conditions, subjected to in-
beams is in good agreement with nu merical and plane compression or in-plane shear. Nevertheless, the
experimental data. Addit ional analytical co mparisons knowledge on compressed LG panels behavior is still
allo wed to notice that a similar limitation (Eq. (34), with limited and with constrained applications.
imp = 0.71 and 0 = 0.60 ) appro ximately coincides, for Let us consider, for examp le, a LG panel simp ly
> 1.10 , with the assumption of k = 120 in Eq.(23) or supported on the four edges (length a, width b), obtained by
assembling two monolithic g lass sheets and a middle
M 1 = 1.40 in Eq.(24).
interlayer (Fig. 8), subjected to in-plane co mpression
In Fig. 7, additional numerical results are proposed for (pressure force per unit length Ny, in y-direct ion).
compressed LG beams affected by an initial sine-shape
imperfection of maximu m amp litude w0 = L0 /400 and an
accidental eccentricity e= t/6, with t the total thickness of the
beam. In this circu mstance, is interesting to notice that
although the applied eccentricity has negligib le amp litude, it
strongly reduces the buckling resistance of the examined
beams. As a result, the optimal buckling curve (wh ich can be
called EC curve), to be taken into account for their
verification is characterized by imperfection factors equal to
imp = 1.80 and 0 = 0.40 (Fig. 7). Undoubtedly, a similar
verification approach can be applied only to purely
compressed glass beams, but it could constitute a useful
design method.

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 96 -
 panel and to define its ultimate strength. The panel, due to
the membrane effects that typically characterize its behavior
in the post-buckled reg ime, is in fact able to sustain greater
loads than N y( E,cr) ,lam . In addition, no sandwich-based
formulat ions are available to describe the axial co mpressive
load N-transversal displacement w of layered panels.
Therefore, sophisticated numerical simu lations should be
performed to realistically investigate their typical behavior.

Fig. 8 Simply supported LG panel subjected to in-plane compression: Because of this reason, in [10] an equivalent thickness
geometry and cross section approach has been presented to precisely investigate the
buckled response of in -plane co mpressed LG panels. In
The crit ical buckling load N y( E,cr) ,lam of the LG panel is accordance with this simplified but accurate formulat ion, the
commonly estimated by means of the linear elastic theory of behavior of the LG panel can be described by means of the
sandwich elements. In accordance with Zenkerts classical theory of monolithic plates, by assuming:
formulat ion, in particular, the crit ical load N y( E,cr) ,lam is given
t eq , w = 3 t13 + t 23 + 12J s , (44)
by [24]:
with:
D1 + D2 mb
2
As b 2 (45)
+ 1 + , (36) =
1
a Ds D
2 2
mb a D 2

N =
(E)
+ 1 + 9.6
EJ s tint
a mb mb 2
y , cr , lam
As b2 Gint t s
2 2

+ 1 + 2
a Ds
= min(a, b) , (46)
with:
D = D1 + Ds + D2 , (37) J s = t1t s22 + t 2 t s21 , (47)

Eti3 , (38) t s t1 , (48)

Di = t s1 =
12(1 2 ) (t1 + t 2 )

( Et1 z12 + Et 2 z 22 ) , (39) tst2 , (49)

Ds = t s2 =
(1 2 ) (t1 + t 2 )

Gint ( z1 + z 2 ) 2 , (40) t s = 0.5(t1 + t 2 ) + t int , (50)

As =
t int
and a coefficient defined in function of the boundary and
zi = 0.5(t i + t int ) . (41) loading conditions of the panel.

Nu merical and analytical co mparisons performed by In [10], it was shown that the coefficient should be
Lu ible [4] demonstrated that Eq. (36) predicts with a good assumed equal to:
level of accuracy the bifurcation load N y( E,cr) ,lam of simp ly 1.09
= + 1.09 . (51)
supported LG panels, for well-defined values of Gint . 2
In particular, Luible showed that the mean ratio between In this manner, the critical buckling load N y( E,cr) ,lam can
analytical and numerical critical loads of 200 LG panels
characterized by various geometrical (aspect ratio = a/b, be estimated as:

mb a Deq 2 Deq ,
thicknesses of glass sheets and interlayer) and mechanical 2 2

N y( E,cr) ,lam = + = k (52)

properties (shear modulus Gint of the interlayer) resulted
a mb b
2
b2
equal to 1.05. Moreover, the so obtained value of critical
load is always comprised between the two well-known where:
monolithic (Gint , full) and layered (Gint 0, abs) limit
values: Et eq3 (53)
Deq =
12(1 2 )
mb a D ,
2 2

N y( E,cr) , full = + (42)

a mb b
2
is the equivalent flexu ral stiffness of the composite panel.

mb a ( D1 + D2 ) .
2 2 Due to the correction factor (Eq. (51)) the critical
N y( E,cr) ,abs = + (43)
buckling load given by Eq. (52) co incides with the solution
a mb b2
of Eq. (36).
Nevertheless, the estimation of the critical buckling load
(E)
At the same time, the load N -total transversal
N does not constitute a useful criterion to study in a
y ,cr ,lam displacement w relat ionship of the in-p lane co mpressed LG
realistic manner the stability problem o f a co mpressed LG panel can be described as [10]:

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 97 -
=
N N=
yb the model in the form o f uniformly d istributed pressure
acting on the upper and lower surfaces of 3D elements. To
t 2 w w
2 2

2
2 w + 3 w 0 + 2 0 w avoid possible eccentricities in the model, boundaries were
Eb eq , w +
a 3(1 ) 8
2
t
teq , w teq , w teq , w w + w0
eq , w applied to the central nodes of the PVB-interlayer.
The second Mshell FE-model, as discussed for LG
(54) beams, consists in co mposite shell elements (S4R) able to
that is: take into account the effective thicknesses of each layer
3
2 p1 + 1
constituting the LG panel. The third Teq FE-model, finally,
w= w0
, (55)
was constructed with monolithic glass shell elements (S4R)
p 2 ( p 3 + 4 p13 + p 32 ) 3 having an equivalent thickness given by Eq. (44). In all these
with: FE-models, g lass was described as an isotropic, linear elastic
p1 = material (E= 70000N/ mm2 , = 0.23). Similarly, PVB was
considered to behave linear-elastically (int = 0.498). Since
9 3 2bE 2 ( 2 1) 24a 2 N ( 2 1) 8bE 2teq2 , w 3bE 2 w02 ( 2 1) the aim of nu merical simulat ions consisted in validating the
(56) equivalent thickness analytical approach, a series of
buckling analysis were performed in ABA QUS. The crit ical
p 2 = 9bE 2 ( 2 1) , (57)
buckling load N y( E,cr) ,lam was predicted for the examined LG
p 3 = 1944(bE 2 ) 3 t eq2 , w w0 ( 2 1) 2 , (58) panel by assuming in each analysis a d ifferent value of shear
modulus Gint (10-4 N/ mm2 < Gint < 104 N/ mm2 ).
and w0 the maximu m amp litude of the in itial imperfection of
the LG panel. The results proposed in Fig. 10 concern the critical
load N y( E,cr) ,lam , numerically and analytically evaluated,
VII. NUMERICAL AND EXPERIMENTAL VALIDATION
compared as a function of Gint . In the specific, five different
To check the accuracy of the proposed analytical solutions are compared, that is the analytical results of the
approach for compressed LG panels, also in this sandwich-based classical theory (Eq . (36)), the analyt ical
circu mstance a series of FE-models were developed with the results given by the proposed equivalent thickness approach
code ABAQUS [19]. (Eq. (52)) and the numerical results obtained by the buckling
Firstly, an accurate three-dimensional numerical FE analyses performed in ABA QUS with the 3D+shell, Mshell
and Teq FE-models. As shown, due to the calibrated
model was developed. The 3D+shell FE model consists in
glass sheets (thicknesses t 1 = t2 = 8mm) described by means correction factor (Eq. (51)), the analytical results coincide
of shell elements (S4R) and in a PVB-interlayer (th ickness with each other.
tint = 1.52mm) modeled in the form of 3D-8 node elements At the same time, they are in good agreement with
(C3D8H, hybrid formulation, incompatible modes). The numerical data obtained by ABAQUS with the 3D+shell
examined LG panel, simp ly supported along the four edges, FE-model. Clearly, the Mshell FE-model does not require
was assumed to have d imensions a= 1m x b = 1m. To ensure long processing time to perform buckling analyses but does
the accuracy of nu merical results, an accurate mesh was not agree with the analytical pred ictions, thus it should be
used in the model (20 x 20mm module). Moreover, over the used with attention. In contrary, the Teq FE-model rapidly
depth of the PVB-film, t wo 3D elements have been used converges and provides accurate results, thus it could be
(Fig. 9). used in practice to avoid 3D sophisticated simulations.
1500
8/1.52/8mm
a= 1m x b= 1m

1250 Full (Eq.(52))

Zenkert (Eq.(36)); teq,w (Eq.(52)
ABAQUS (3D+shell)
1000 ABAQUS (MShell)
ABAQUS (Teq)
N(E)y,cr,lam [kN]

Abs (Eq.(52))

750

500

250
Fig. 9 3D+shell numerical model for simply supported in-plane compressed
LG panels detail (ABAQUS)
0
3D elements and shell elements were connected together
10-4 10-3 10-2 10-1 100 101 102 103 104
by using the same nodes. In addition, to describe the Gint [N/mm2]
effective geo metry of the LG panel, a section offset t offset =
4mm fro m the centroidal axis of each glass pane was applied Fig. 10 Critical buckling load N y( E,cr) ,lam for in-plane compressed LG simply
to shell elements. In-plane co mp ression was introduced in supported along the edges: analytical and numerical comparisons

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 98 -
 The second proposed comparison concerns the typical Rationally, a reasonable check of the maximu m
load N-transversal displacement w relat ionship deformation should be carried out taking into account an
characterizing the buckled response of in-plane co mpressed initial sinusoidal imperfection, proportional to the first
LG panels. For this purpose, some experimental data modal shape of the panel, of maximu m amplitude w 0 . In this
available in literature for out-of-p lane displacements of context, Eng lhardt [7] experimentally investigated
simp ly supported LG panels under comp ression were taken monolithic float and toughened glass panels having a
into account [4]. Specifically, the experimental data maximu m amp litude of geometrical imperfection equal to
proposed in Fig.11 and su mmarized in Table 1 refer to three w0 = a/2000. Tests performed by Lu ible and Crisinel [25]
squared 8/1.52/8mm LG panels (a= 1m x b = 1m) tested by highlighted that initial imperfect ions in non-tempered
Lu ible. Analytical results, in particular, have been defined annealed flat glass panels have small amplitude (w0 <
for each LG panel by means of Eq. (55), having estimated
a/2500), whereas heat-strengthened and fully toughened
the corresponding equivalent thickness teq,w (Eq.(44)). As
glass panels can have sinusoidal imperfections up to w0 =
shown, experimental and analytical data are in good
agreement. a/300. Reasonably, in the buckling verificat ion of LG panels
a minimu m amp litude of imperfection w0 = a/1000 should be
700
8/1.52/8mm
considered. Moreover, Eng lhardt suggests for the coefficient
a= 1m x b= 1m k a value equal to k= 300 [7].
600
At the same time, the design compressive load Ny,Ed
500 should be compared with the design buckling resistance
Ny,b,Rd of the panel:
400
N [kN]

N y( E,cr) ,lam , (60)

N y , Ed N y ,b , Rd =
M1
300

where N y( E,cr) ,lam is given by Eq. (52). In this context,

200 Test-1
Test-2 Englhardt suggests for the buckling safety factor M1 a value
Test-3
100 Analytical-1 of 1.40 [7].
Analytical-2
Analytical-3 Also in this circumstance, an alternative verificat ion
0 criterion for in-p lane co mpressed LG panels can be derived
0 5 10 15 20 fro m buckling curves suggested by the Eurocode 3 for the
w [mm]
verification of traditional structural elements [11]. In the
Fig. 11 Load N-transversal displacement w relationship for in-plane specific case of co mpressed panels, the buckling verificat ion
compressed LG panels. Experimental [4] and analytical results (Eq. (44))
could be still performed by satisfying the condition given by
TABLE I EXPERIMENTAL DATA AND CORRESPONDING EQUIVALENT Eq. (60), in wh ich the design buckling resistance Ny,b,Rd of
THICKNESSES FOR IN -PLANE COMPRESSED LAMINATED GLASS PANELS
the LG panel can be estimated as:
a= 1m x b= 1m T [4] Gint [4] teq,w (Eq.(44))
N y ,b , Rd = A Rd , (61)
2
8/1.52/8mm [C] [N/mm ] [mm]
with an opportune reduction factor and A the total cross-
Test-1 20.5 0.81 11.14 section area.

Test-2 21.0 0.76 11.09 In Eq. (61), the reduction factor can be estimated by
means of Eq. (34), where is g iven by Eq. (35), with
Test-3 20.0 0.84 11.18 imp = 0.49 and 0 = 0.60 . In addition, the non-dimensional
slenderness (Eq. (31)) should be expressed as a function
VIII. DESIGN CRIT ERIA FOR IN-PLANE COMPRESSED of the critical buckling load N y( E,cr) ,lam of the examined LG
LAMINATED GLASS PANELS
panel, given by Eq. (53).
As proposed for LG beams in compression, the buckling
Also in this circu mstance, the imperfect ion factors
verification of simp ly supported in-plane co mpressed LG
panels should be performed by contemporarily satisfying imp = 0.49 and 0 = 0.60 have been opportunely calibrated
requisites of deformability and durability. In accordance on the basis of experimental and numerical data available in
with the Limit State approach, the maximu m transversal literature. As proposed in Fig. 12, for examp le, the limitation
displacement wmax of the panel, g iven by Eq. (55), should be provided by the suggested EC curve ( imp = 0.49 and
opportunely limited, as a function of the length a, by posing
0 = 0.60 ) is appro ximately equal to the conditions
for examp le the condition:
a expressed by Eq. (59), with k= 300 [7], and by Eq.(60), with
wmax . (59)
M 1 = 1.40 [7].
k

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 99 -
 1.4
classical theory of sandwich elements under compression are
Euler
EC Curve (imp= 0.49, 0= 0.60)
available, but often they can be used only to predict the
1.2
k= 300 value of their critical buckling load. In the paper, exact
M1= 1.30 analytical formu lations are proposed for the buckling
1.0 verification of co mpressed LG beams and panels. Fo r LG
beams, a model developed on the basis of Newmarks theory
0.8 of co mposite beams with part ial interaction is proposed. At
the same time, a simp lified but accurate equivalent thickness

0.6
formulat ion is proposed for the verificat ion of in-p lane
compressed LG panels. Co mparisons with numerical and
experimental data are presented to validate the accuracy of
0.4
analytical formu lations. In both the circu mstances, the
presented models allo w to take into account the effective
0.2
level of connection offered by the adopted interlayer. At the
same time, a criterion based on the buckling curves of
0.0 Eurocodes is suggested for a rational buckling verification of
0 1 2 _ 3 4 5 compressed LG elements.

REFERENCES
Fig. 12 Buckling curve for the verification of LG panels under in
compression: comparisons with analytical results [1] M . M eissner, and V. Sackmann, On the effect of artificial
weathering on the shear bond and the tear strength of two
In Fig. 13, so me experimental and numerical data are
different interlayers of laminated glass, in Porc. ISAAG, 2006.
also presented to validate the proposed verificat ion
[2] M . Z. Aik, and S. Tezcan, Laminated glass beams: strength
procedure. The experimental data, in particu lar, refer to
factor and temperature effect, Comp. Struct., vol. 83, pp.
buckling tests performed by Luib le [4] on in-p lane 1742-1753, 2005.
compressed LG panels simp ly supported along the edges,
[3] B. Weller, J. Wnsch, and K. Hrth, Experimental study on
having different geo metrical p roperties. As shown in Figs. different interlayer materials for laminated glass, in Proc.
11 and 13, a significant post-critical resistance can be Glass Processing Days, 2005, pp. 120-123.
observed in the typical behavior of LG panels under [4] A. Luible, Stabilitt von Tragelementen aus Glas, Ecole
compression. A similar behavior is also confirmed by Polytechnique Fdrale de Lausanne, Thse EPFL 3014, 2004
numerical and experimental results obtained by Englhardt [7] (available: http://icom.epfl.ch/publications).
and Luible [4]. Ho wever, it is possible to notice that, [5] J. Belis, Kipsterkte van monolitische en gelamineerde glazen
ignoring the post-critical resistance of the examined LG liggers, Universiteit Gent, Laboratorium voor
panels, the proposed solution provides adequate safety M odelonderzoek, Doctoraatsthesis, ISBN: 90-8578-034-9,
requirements. 2006 (available: http://lib.ugent.be).
[6] C. Amadio, and C. Bedon, Buckling of laminated glass
1.4
FE-Englhardt [7]
elements in compression, J. Struct. Eng., vol. 137(8), 2011.
FE, w0= a/1000-Luible [4] [7] C. Amadio, and C. Bedon, Buckling of laminated glass
1.2 FE, w0= a/500-Luible [4] elements in out-of-plane bending, Eng. Struct., vol. 32, pp.
Test-Luible [4] 3780-3788, 2010.
1.0 [8] O. Englhardt, Flchentragwerke aus glas-Tragverhalten und
Stabilitt, Universitt fr Bodenkultur Wien, Dissertation zur
0.8 Erlangugn des Doktorgrades, 2007.
[9] D. M ocibob, Glass panel under shear loading-Use of glass

envelopes in building stabilization, Ecole Polytechnique

0.6
Fdrale de Lausanne, Thse EPFL 4185, 2008 (available:
http://icom.epfl.ch./publications).
0.4
[10] C. Bedon, and C. Amadio, Buckling of flat laminated glass
panels under in-plane compression or shear, Eng. Struct., vol.
0.2 36, pp. 185-197, 2012.
Euler
[11] UNI EN 1993-1-1: 2005. Eurocode 3-Design of Steel
EC Curve (imp= 0.49, 0= 0.60)
0.0 Structures Part 1-1: General rules and rules for buildings,
0 1 2 _ 3 4 5 M ay 2005.
[12] UNI EN 1994-1-1: 2005. Eurocode 4-Design of Composite
Fig. 13 Buckling curve for the verification of LG panels under in-plane Steel and Concrete Structures Part 1-1: General rules and
compression: comparisons with numerical [4, 7] and experimental results [4] rules for buildings, M ay 2005.
[13] UNI EN 1995-1-1: 2005. Eurocode 5-Design of Timber
IX. CONCLUSIONS Structures Part 1-1: General rules and rules for buildings,
M ay 2005.
Laminated glass beams and panels subjected axial [14] K. Stamm, and H. Witte, Sandwichkonstructionen-
compression are frequently subjected to stability problems. Berechnung, fertigung, ausfhrung, Springer Verlag,
In literature, several analytical formu lations derived fro m the Germany, 1974.

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 100 -
[15] J. Blaauwendraad, Buckling of laminated glass columns,
Heron, 52(1-2).
[16] N. M . Newmark, C. P. Siess, and I. M . Viest, Test and
analysis of composite beams with incomplete interaction,
Proc. Soc. Exp. Stress Anal., 9(1), pp. 75-92.
[17] J. Belis, D. M ocibob, A. Luible, and M . Vanderbroek, On the
size and shape of initial out-of-plane curvatures in structural
glass components, Const. Build. Mater, vol. 25, pp. 2700-
2712, 2011.
[18] S. J. Bennison, a. Jagota, and C. A. Smith, Fracture of
glass/poly (vinyl butyral) (Butacite) laminates in biaxial
flexure, J. Am. Ceram. Soc., vol. 82/7), pp. 1761-1770, 1999.
[19] ABAQUS version 6.9, Simulia, Pawtucket, R.I. 02860 USA,
2009.
[20] European Committee for Standardisation (CEN), Glass in
building Design of glass panes - Part 2: Design for
uniformly distributed loads, prEN 13474-2, Brussels,
Belgium, 2000.
[21] W.E. Ayrton, and J. Perry, On Struts, The Engineer, pp.
137-153, 1886.
[22] S. Aiello, G. M inaf, and N. Scibilia, Compressive behaviour
of laminated structural glass members, Eng. Struct., vol. 33,
pp. 3402-3408, 2011.
[23] M . Feldmann, and K. Langosch, Knickfestigkeit und
einheitliche Knickkurven fr scheibenfrmige Glassttzen mit
M onoglasquerschnitt aus TVG und ESG, Ernst & Sohn
Verlag fr Architektur und technische Wissenschaften GM bH
& Co. KG, Berlin Stahlbau Spezial 2010-Konstruktiver
Glasbau, 2010.
[24] D. Zenkert, The handbook of sandwich construction, United
Kingdom, Engineering M aterials Advisory Service Ltd., 1997.
[25] A. Luible, and M . Crisinel, Plate buckling of glass panels, in
Proc. Glass Processing Days, 2005.

JCES Vo lu me 1, Issue 3 September 2012 PP. 90-101

- 101 -