You are on page 1of 9

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/37455647

Design of Glass Beams Subjected to Lateral


Torsional Buckling
Article January 2006
DOI: 10.2749/222137806796168903 Source: OAI

CITATIONS

READS

168

2 authors:
Andreas Luible

Michel Crisinel

Lucerne University of Applied Sciences and Arts

cole Polytechnique Fdrale de Lausanne

38 PUBLICATIONS 155 CITATIONS

140 PUBLICATIONS 287 CITATIONS

SEE PROFILE

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

KTI-Projekt Gebudehlle - Ein Entscheidungs- und Beurteilungsinstrument zur Erstellung


nachhaltiger Fassaden mehrgeschossiger Gebude View project

All in-text references underlined in blue are linked to publications on ResearchGate,


letting you access and read them immediately.

Available from: Andreas Luible


Retrieved on: 11 November 2016

Design of Glass Beams Subjected to Lateral Torsional Buckling


Andreas LUIBLE
Dr. sc. techn.
Schmidlin Ltd.
Facade Technology
Aesch, Switzerland

Michel CRISINEL
Civil Engineer, Lecturer
Ecole polytechnique fdrale
de Lausanne (EPFL),
Lausanne, Switzerland

Andreas Luible, born in 1971,


Received his engineering degree
from the TU Munich in 1999 and
received his Ph.D. from the Swiss
Federal Institute of Technology
Lausanne EPFL in 2004. He is
currently working as Senior Faade
Engineer for Schmidlin.

Michel Crisinel, born in 1945,


received his civil engineering degree
from EPFL in 1968. He has worked
for a consulting firm prior to his
joining the Steel Structures Lab
ICOM. His main research interest
includes steel-concrete composite
construction and glass structures.

andreas@luible.name

michel.crisinel@epfl.ch

Summary
This paper describes investigations conducted at the Steel Structures Laboratory ICOM of EPFL on
the lateral torsional buckling of single layer and laminated safety glass. The structural behaviour is
studied by means of buckling tests, analytical and numerical models. It is shown that the initial
fracture occurs always on the glass surface under tensile stress. The load carrying behaviour of
laminated glass can be simplified to a monolithic beam with an equivalent cross section. A design
method with buckling curves using a slenderness ratio based on effective tensile strength seems
applicable for the design of glass beams. As a result of numerical simulations, recommendations for
the future development of lateral torsional buckling curves for glass beams are given.
Keywords: Stability, glass, lateral torsional buckling, glass beam, structural use of glass.

1. Introduction
In the last years the material glass is applied more and more for structural elements in modern steel
glass facades e.g. beams, columns, and stiffeners. Glass is a material that is able to resist very high
compression stresses and due to their high slenderness, these elements tend to fail because of
instability. Until now no design methods are developed as for steel, e.g. buckling curves. Therefore
experimental and theoretical investigations of the fundamental stability problems (column buckling,
lateral torsional buckling, plate buckling) of single layer and laminated glass elements have been
conducted at the Steel Structures Laboratory ICOM, EPFL in [1]. The objectives were to investigate
the load carrying behaviour and to study possible design methods for stability-critical glass
elements. Summarized results concerning column buckling and plate buckling are published in
[2][3][4][5]. In this paper the main results related to the lateral torsional buckling of glass beams are
presented.

2. Analytical and numerical models


Lateral torsional buckling is a limit state of structural stability, where a beam is loaded with pure
bending. The deformation is a combination of lateral deflection and twisting. In glass structures this
type of stability failure can occur, for example in beams or fins used as stiffeners in facades.
2.1 Single layer glass
The critical torsional buckling moment of a beam with a rectangular cross section is:

M cr , D = C1

2 EI z
LD 2

GK L 2
C 2 z a + 2 D + C2 z a
EI z

(1)

where E = young modulus, Iz = moment of inertia about z-axis, G = shear modulus, K = torsion
constant, and LD = beam length. The factors Ci and za take into account different boundary
conditions, different bending moments and the distance between the centre of gravity and the point
where the load is applied [6]. Due to their rectangular cross-section, warping torsion may be
neglected in single layer and laminated safety glass beams.
LD
My

My

VIEW

end supports

My

My

TOP VIEW
initial position

SECTION

final position

Fig. 1: Lateral torsional buckling.

Similar to column buckling the lateral


torsional buckling resistance is not limited by
the critical torsional buckling moment Mcr,D.
Due to imperfections of the beam the lateral
deformation and twisting start already to
increase under very small loads and the lateral
torsional buckling resistance is reached when
the maximum stresses in the beam exceeds the
material resistance. Bifurcation buckling (e.g.
Eq.(1)) over-estimates the real lateral torsional
buckling resistance.
In order to describe the real load carrying
behaviour, analytical and numerical models
(finite element method - FEM) were
developed. It was found in [1] that due to the
slender geometries of glass beams the nonlinear numerical model is more suitable to
describe the load carrying behaviour, than
analytical models based on the linear elastic
beam theory.

2.2 Laminated safety glass


The critical lateral torsional buckling moment of laminated safety glass may be calculated using
Eq.(1), where the lateral bending stiffness EIz and the torsional stiffness GK are replaced by a
equivalent stiffness, EIz,eff and GKeff. Both stiffness are determined using sandwich theory [7] in
order to take into account the composite action of the PVB interlayer in laminated safety glass [1].
2.2.1 Equivalent stiffness for laminated safety glass with two glass layers
2 + + 1
EI z ,eff = EI s
1 + 2

(2)

with
I S = h(t1 z12 + t2 z22 ) ; =

t PVB
EI S
I1 + I 2
; =
IS
G PVB h ( z1 + z 2 ) 2 LD 2

(3)

where t1, t2, tPVB, z1, z2 see Fig. 2 and h = beam height, E = young modulus of glass, Ii = moment of
inertia of the related glass layer, GPVB = shear modulus of PVB and LD = buckling length.

GK eff = GK glass1 + GK glass 2 + GK comp

(4)

with

GK comp

h
tanh

2
= GI S 1
h

(5)

t +t
tt
I S = 4 1 2 + t PVB 1 2 h;
2
t1 + t2

GPVB t1 + t2
G t PVB t1 t2

(6)

where G = shear modulus of glass, Kglassi = torsion constant of the related glass layer, and Kcomp =
torsion constant due to composite action of the PVB.
glass
PVB

glass

t1
tPVB
t2

glass

t1
tPVB

z1

PVB

glass

z2

t2
tPVB

glass

z1
z2

t1

Fig. 2: Laminated safety glass with two glass layers (left) and three glass layers (right).
2.2.2 Equivalent stiffness for laminated safety glass with three glass layers
The following formulas apply to laminated safety glass with three glass layers whereas the external
layers have the same thickness.
2 + + 1
EI z ,eff = EI s
1 + 2

(7)

with

I S = 2h t1 z12 ; =

tPVB
EI S
2 I1 + I 2
; =
;
2
2GPVB h z1 LD 2
IS

(8)

where t1, t2, tPVB, z1, (Fig. 2), and h = beam height, E = young modulus of glass, Ii = moment of
inertia of the related glass layer, GPVB = shear modulus of PVB and LD = buckling length.
GK eff = GK glass1 + GK glass 2 + GK glass 3 + GK comp

(9)

with GKcomp according to eq.(5) whereas Is and become:


I S = 2 ( t2 + 2t PVB + t1 ) t1 h ;
2

GPVB t1 + t2
;
G tPVB t1 t2

(10)

In order to study the lateral torsional buckling behaviour of a glass beam composed of laminated
safety glass with imperfections a finite element model with [1] was developed (Fig. 3).

symmetric axis

v0

x
u

end
support

v
z

glass
x
z

restraints:
u=0, y=z=0

Fig. 3: FEM-Model laminated safety glass.

Only half of the glass beam was modelled


because of the symmetrical system. The
glass layers were modelled with shell
elements and the PVB interlayer with
volume elements. For the elastic model
elastic volume elements were used in case of
the visco-elastic model [8] visco-elastic
volume elements were used. Each simulation
was started with a modal analysis of the
system.
The
resulting
eigenvalue
corresponds to the critical buckling load of
the beam. In the next step the initial
deformation was applied as a scaled shape of
the first eigenform of the considered system.
With this imperfect system the non linear
structural calculation was carried out.

3. Experimental Investigations
3.1 Test setup
For the test setup a simply supported beam was subjected to a concentrated load at mid-span. The
main difficulty during the test was the load application that had to follow the lateral displacement of
the top edge of the glass beam. The hydraulic jack and the load introduction device were therefore
fixed on a carriage (Fig. 4). The load introduction device allowed a free rotation of the top glass
edge. A PTFE interlayer was used to reduce friction between the sliding parts.

Fig. 4: Left: Test setup lateral torsional buckling test, Right: Buckled glass beam.
In Fig. 4 a deformed single layered glass beam is shown. The lateral displacement was measured
with displacement transducers which were connected to the measure point on the glass with steel
wires. Two load cells were fixed on the load introduction device.

3.2 Main results


Seventy-nine lateral torsional buckling tests on single monolithic and laminated safety glass were
carried out. All tested beams were either of heat strengthened or of toughened glass.
3.2.1 Single layer glass
The results of the tests showed a good agreement with the numerical simulation. Similar to column
buckling, the maximum force in the test approaches the critical buckling load Fcr,D calculated with
the numerical model [1]. During the test the stress distribution on the glass surface was controlled
with strain gauges and compared with
F [kN]
the numerical simulation. It could be
50
seen that the stress distribution is non
L D = 1300 mm
linear over the beam height h. The more
h = 200mm
the lateral displacement increased this 40
t = 10/1.52/10 mm
non linear effect became stronger. The
breakage pattern showed that the initial
fracture occurs always on the glass 30
test
breakage of
surface under tensile stress.
both glasses
numerical simulation

3.2.2 Laminated safety glass


The buckling tests Fig. 5 confirmed that
the load carrying behaviour of laminated
safety glass is characterised by the viscoelastic behaviour of the PVB interlayer.
The temperature and the load duration
therefore have an important influence on
the lateral torsional buckling resistance
of a laminated glass beam. The
developed visco-elastic numerical model
showed a good agreement with the test
data.

20
vsup

10

x
y

LD

0
0

10

20
30
v sup [mm]

40

50

Fig. 5: Test result of laminated safety glass.

4. Most significant design parameters


The study of the load carrying behaviour showed that the dispersion of the glass thickness t, the
initial deformation of a glass beam, the composite action due to the PVB interlayer and the tensile
strength of the glass close to the edge have the most significant influence on the lateral torsional
buckling resistance of a glass beam.
4.1 Glass thickness
The dispersion of the glass thickness was measured on more than 200 test specimen from two
different glass manufacturers.
The thickness t of annealed flat glass panels differs from the nominal value because glass
manufacturers try to save material in making the most use of the thickness tolerances specified by
the codes. The real glass thickness is often less than the nominal value, therefore reducing the
moment of inertia of the cross section and, thus the buckling strength. The aforementioned
measurements confirmed that the values follow a normal distribution. The 5% percentile value is
97.61% of the nominal glass thickness.
4.2 Initial deformation
The initial geometric deformation v0 of flat glass is mainly caused by the tempering process. The
geometric initial deformation was measured with a taut steel wire on more than 200 specimens. The

results confirmed that non-tempered annealed flat glass has a very low initial deformation (< 1/2500)
while heat-strengthened and fully toughened glass can have a sinusoidal initial deformation up to
1/300 of the length L. The nominal thickness of the glasses has an influence on the statistical
distribution but for design in practice the simplified assumption of one single distribution might be
sufficient. Laminated safety glass showed the same results. The measured values followed a normal
distribution with a 95% percentile value of 1/386. However maximum initial deformations depend
strongly on the quality of the furnace and can therefore vary between different glass manufacturers.
4.3 Tensile strength of the glass edge
Due to the high compressive strength of glass the tensile strength of the glass surface is always
determinant for the buckling resistance of glass beams with dimensions used for building
application. For the determination of the buckling resistance the tensile strength near the glass edge
under tension has to be known. Most of existing design codes for glass give only a tensile strength
for the centre of the glass pane which is not valid for the glass edge. The tensile strength of the glass
edge depends on the tensile strength of the material, which is influenced by critical damages and
initial flaws, and the distribution of residual stress due to tempering. The latter is not a constant
value around the edge [9][10]. In a certain distance from the glass edge and in the middle of the
glass edge the residual stress becomes minimal and although there are no critical flaws it may be
critical for the breakage.
4.4 PVB interlayer
In practice, the visco-elastic behaviour of the PVB interlayer can be simplified by an elastic
interlayer with equivalent shear modulus, GPVB. It was found, that a shear modulus of the interlayer
higher than 300 N/mm2 is necessary in order to assume a monolithic behaviour of the beam.
Realistic values of GPVB for PVB are < 5 N/mm2. Even new materials [1] which are stiffer than PVB
are not able to create a load carrying behaviour similar to a monolithic cross section. A significant
composite action due to the shear interlayer may taken into account for soft interlayer materials
such as PVB only for short term loads (e.g. wind loads).

5. Design by buckling curves


The lateral torsional buckling resistance of a glass beam may either be determined with an
appropriate FEM model or with buckling curves.
In [1] it is shown that FEM models are suitable to describe the load carrying behaviour of glass
beams. Nonlinear effects, different initial imperfection as well as different boundary conditions may
easily be taken into account. Nevertheless FEM models are quite fastidious and too complicated for
a simple and fast design.
In order to simplify the design process the development of buckling curves was investigated in [1]
by means of the developed FEM models. For the buckling curves a slenderness ratio D and a
reduction factor D similar to the lateral torsional buckling design of steel beams were defined. In
contrast to steel, both are based on the tensile strength, since the compressive strength does not limit
the buckling strength of glass:
D =

p ,t
2 p ,t I y
=
M cr , D h
cr , D

(11)

where p,t = tensile strength of the glass and cr,D = critical lateral torsional buckling stress.
The critical lateral torsional buckling moment Mcr,D may be calculated with Eq. (1). For the design
of a laminated safety glass the equivalent lateral bending stiffness EIz,eff (Eq. (2) and (7)) and the

equivalent torsional stiffness GKeff (Eq. (4) and (9)) may be used. The reduction factor D in the
buckling diagram is a function of the slenderness ratio D :

( )

D = f D

(12)

Hence the maximum bending moment M is:


M =

2I y
z

= D p ,t

2I y

(13)

For different types of loading, glass geometries, shear modulus of the PVB interlayer, and initial
deformation, v0, reduction factors were generated and plotted in several buckling diagrams (Fig. 6).
These diagrams may serve as a orientation for a future definition of lateral torsional buckling curves
for glass. The main results are:

1.2

It is possible to define lateral


bifurcation buckling
torsional buckling curves for glass
numerical simulation
1.0
based on the tensile strength.
buckling test
Buckling curves for example for steel
EC3 (a)
structures can not be transferred to
0.8
EC3 (c)
glass.
It might be useful to determine
0.6
several buckling curves, depending
F
on the composition of the glass
0.4
(single layer, laminated safety glass),
the type of loading and the initial
v 0 = L D /270
0.2
deformation.
Buckling curve (c) in Eurocode 3
(EN 1993-1-1:1993) [11] may be
0.0
used as a conservative approach for
0.0
0.5
1.0
1.5
2.0
2.5
slenderness D
design of glass elements since all
simulations results lay above this
Fig. 6: Simulated reduction factors for a
curve.
concentrated load at mid span compared to buckling
Further studies with additional
test results.
structural systems are necessary (i.e.
systems with intermediate lateral support or systems that are able to take into account of the
partial restraint of the beam by structural silicon joints).

6. Conclusions
The lateral torsional buckling of glass beams was studied with lateral torsional buckling tests and
numerical simulations. Due to the geometry of glass beams it was shown that analytical models
based on the linear elastic beam theory are not able to describe the real load carrying behaviour,
especially for high h/LD ratios. In order to determine the lateral torsional buckling resistance the
tensile strength of the glass close to the edge has to be known. A suitable design method for lateral
torsional buckling of glass beams might be a numerical calculation (FEM) of the buckling
behaviour and the maximum tensile stress distribution on the glass surface. In order to simplify the
design process buckling curves are more suitable. In the research work it was demonstrated how
these curves might be established. For elementary load cases and structural systems reduction
factors were simulated with the developed model. These simulation results may be used for a future
determination of lateral torsional buckling curves for glass.

7. Acknowledgment
The research work presented in this paper was primarily conducted with the support of the Swiss
National Science Foundation (SNF) and the industry partners Glas Trsch (Btzberg, Switzerland)
and Verre Industriels SA (Moutier, Switzerland).

8. References
[1]

LUIBLE, A., Stabilitt von Tragelementen aus Glas, Thse EPFL 3014, Ecole polytechnique
fdrale de Lausanne (EPFL) (free download: http://icom.epfl.ch/publications
/pubinfo.php?pubid=499), Lausanne 2004.

[2]

LUIBLE, A. & CRISINEL, M., Buckling Strength of Glass Elements in Compression,


Structural Engineering International, Vol. 14, No. 2, 2004.

[3]

LUIBLE, A. & CRISINEL, M., Plate buckling of glass panels, Glass processing days 2005,
Proceedings of the conference, June 2005, Tampere.

[4]

LUIBLE, A. & CRISINEL, M., Stability of Load Carrying Elements of Glass, Final report of
the COST action C13, Bruxelles.

[5]

LUIBLE, A. & CRISINEL, M., Stability of Load Carrying Elements of Glass, 4th European
Conference on Steel Structures, 8-10 June 2005, Maastricht, Netherlands.

[6]

HIRT, M. A. & BEZ, R., Stahlbau: Grundbegriffe und Bemessungsverfahren, Ernst & Sohn,
Berlin, 1998.

[7]

STAMM, K. & WITTE, H., Sandwichkonstruktionen - Berechnung, Fertigung, Ausfhrung,


Springer Verlag, Wien 1974.

[8]

VAN DUSER, A. & JAGOTA, A. & BENNISON, S., J., Analysis of Glass/Polyvinyl Butyral
Laminates Subjected to Uniform Pressure, Journal of engineering mechanics, Vol. 125, 1999,
pp. 435-442.

[9]

LAUFS, W., Ein Bemessungskonzept zur Festigkeit thermisch vorgespannter Glser,


Schriftenreihe Stahlbau RWTH Aachen, Shaker Verlag, Aachen, 2000.

[10] DAUDEVILLE, L. & BERNARD, F. & GY, R. Residual Stresses Near Holes In Tempered
Glass Plates, Materials Science Forum, Trans. Tech. Publications, Switzerland, Vol. 404-407,
2002, pp. 43-48.
[11] EC3: ENV 1993-1-1: Eurocode 3: Design of Steel Structures Part 1.1: General rules and rules
for buildings, January 1993.