Challenging Glass Conference Proceedings – Volume 9 – June 2024 – Louter, Bos & Belis (Eds.

)
International Conference on the Architectural and Structural Application of Glass
Challenging Glass Conference 9 – 19 & 20 June 2024 – TU Delft – The Netherlands
9

Design and Stability of Laminated Glass Beams and

Cantilevers with Continuous Lateral Silicone Restraint

Richard Green a, Chiara Bedon b, Laura Galoupi c, Andrew Crosby d

a Green Facades PLLC, USA, [email protected]

b University of Trieste, Italy
c University of Parma, Italy
d RJC Engineers, Canada

Abstract
The stability of monolithic glass beams is reasonably well defined; as an elastic material it behaves in
a similar manner to other elastic materials such as steel, for which there are many equations of
different forms which give similar results. Special care is required for continuous restraint to the
tension flange. Equations presented in Australian Standard AS1288 Glass in Buildings – Selection and
Installation have been used successfully for many years for monolithic fins when used with the strength
model of AS1288 but require a more comprehensive approach when using laminated fins and/or
strength models that allow higher levels of stress. A review of equations for cantilevers results in a
wider range of approaches with significant variance between the outcomes of various published steel
and glass standards. AS1288 has been used as the default standard for stability of glass fins, however
for cantilevers it appears to have a misprint which has existed for decades. This paper presents
strategies for determining the moment capacity of beams and cantilevers made of laminated glass
with continuous flexible buckling restraints, such as structural silicone, which have initial imperfections
and a known design strength capacity. Where multiple wave lengths form, the warping stiffness may
contribute and formulations for rectangles are presented. The accuracy and validity of the approach is
also assessed by means of comparisons with the outcomes of Finite Element numerical analyses.

Keywords
Glass Structures, Glass Beams, Glass Cantilevers, Laminated Glass, Continuous Elastic Restraint,
Imperfect Slender Beams

Article Information
• Digital Object Identifier (DOI): 10.47982/cgc.9.599
• Published by Challenging Glass, on behalf of the author(s), at Stichting OpenAccess.
• Published as part of the peer-reviewed Challenging Glass Conference Proceedings, Volume 9, June 2024, 10.47982/cgc.9
• Editors: Christian Louter, Freek Bos & Jan Belis
• This work is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.
• Copyright © 2024 with the author(s)

1 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

1. Introduction
The use of laminated glass fins constitutes a recently emerging trend in the façade industry. In modern
buildings, glass fins are used to support glazed façades and enhance their rigidity, and as supports for
glazed roofs. As glass is brittle and can be damaged by non-design load events (e.g. impact), lamination
is used to ensure a fail-safe response when one ply breaks. Due to their high slenderness, the design
of glass fins requires particular attention to instability phenomena.
Currently, the main complete standard in this regard is the Australian Standard AS1288 (Appendix C),
containing stability options for unrestrained monolithic beams (fins), beams with intermediate
restraints, cantilevers, and continuously restrained beams. The continuously restrained beam case is
perhaps the most common. The source of the formula for the continuously restrained case is not
recorded in AS1288 and reputedly has origins in rectangular timber joist equations, which assume a
rigid fixity to the diaphragm providing the restraint. However, when the restraint is flexible, such as in
the common case where structural silicone is used, the formula can be non-conservative due to
reduction of the elastic critical buckling load and development of secondary stresses from minor axis
bending. The magnitude of the second-order stress is a function of the level of imperfection present
in the construction. The AS1288 model generally provides adequate outcomes when used in
combination with the design strength levels in AS1288, because the design stress limit for fins is
relatively conservative, employing a glass type factor of 2.5 for fully tempered glass, instead of the
value 4.0 in ASTM E1300, where glass type factor is the ratio of design capacity relative to that of
annealed glass. If using higher stresses for design, without the ‘buffer’ of lower design stresses in
AS1288, a more accurate formula for the stability is required and second-order effects need to be
taken into consideration.
For cantilever beams, the equations for the Euler elastic critical moment was examined from multiple
standards and sources, including AS 1288 Glass in Buildings, AISC 360 Specification for Structural Steel
Buildings, NCCI SN006a-EN-EU Elastic Critical Moments of Cantilevers (reflecting Eurocode 3), AS 4100
Steel Structures, Timoshenko and Gere Theory of Elastic Stability (1956), Nethercott and Trahair (1976)
and NASA N.A.C.A. Technical Note No 601 (Dumont,·1937). From these various sources the relationship
between the ‘g factors’ as presented in AS1288 and formulas presented elsewhere is clarified, which
allows broader comparison, and a summary of the most useful sources is presented and discussed.
Kala (Kala, 2013) provides a derivation of moment capacity of beams for a given stress level and level
of imperfection, geometric parameters, and elastic buckling moment. The derivation only holds true
for the elastic range, but glass is elastic to fracture and, as rectangular fins are doubly symmetric, the
compression limits and tension limits are the same magnitude (axial loads are not considered at this
point). Finite element modelling has found the equation to have good agreement for the tension-
controlled stress failure.
Bedon et al. (2015) demonstrate that a wavelength approach can be taken to glass fins with continuous
elastic lateral restraints. The approach assumes the critical half-wavelength must be the length of
member divided by an integer and requires testing for a number of integers to find the most critical
(minimum) one. Bedon (2021) proposes separate equations for tension flange and compression flange
with flexible continuous restraint. However, by selecting a suitable sign convention, the two equations
can be merged in a unique approach, and further elaborated to capture the actual bending capacity,
compared to rigid mechanical restraints (Bedon, 2021).

2 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Laminated glass, composed of two or more glass plies bonded by polymeric interlayer(s), further adds
to the complexity of the problem. The flexural and torsional response of this kind of composite is
usually modelled by means of “effective thickness” methods, consisting in the definition of a
monolithic “equivalent” element, of appropriate thickness, having the same flexural/torsional
properties of the considered laminate. The effective thickness value depends on the mechanical and
geometrical properties of both glass plies and interlayer(s), as well as by loading and boundary
conditions. The Wölfel-Bennison effective thickness model in ASTM E1300 only correctly predicts the
stiffness of simply supported beams under uniformly distributed load, with important inputs including:
the length of the member, thickness of the interlayer and shear stiffness of the interlayer for a given
temperature and duration. However, this model, developed for laminated glass elements under
bending, significantly over-predicts the torsional stiffness of the glass fin and is non-conservative for
stability calculations. Galuppi et al. (Galuppi, et al. 2013, 2014 and/or 2020) have developed an
Enhanced Effective Thickness approach which includes additional parameters for torsional effective
thickness that are strongly influenced by the width of the fin.
The proposed method predicts a more realistic capacity of glass beams and fins utilizing a combination
of effective section properties (including torsional stiffnesses), a stability model (that accounts for
flexibility of the restraint), and a method that considers second-order stresses (due to imperfections)
in combination with the available design strength. Analytical predictions are validated towards Finite
Element (FE) numerical analyses carried out in ABAQUS (Simulia, 2022) for a selection of geometrical
and mechanical configurations.

2. Glossary and Conventions

Description Conventional US Conventional EU Conventional AU
Symbol Symbol Symbol
As used in paper (and some prior (AS 1288)
(Alt symbols not used in papers)
brackets)

Axis system
Width x y x
Height y z y
Length z x z

Breadth of beam B b

Length of beam L L, Lo L

Thickness: overall, plies h h t

3 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Effective thickness ℎ�𝑤𝑤 ℎ𝑒𝑒𝑒𝑒,𝑤𝑤 or, ℎ𝑒𝑒𝑒𝑒𝑒𝑒
Bending stiffness
ℎ�𝑖𝑖;𝜎𝜎 ℎ𝑒𝑒𝑒𝑒,𝑖𝑖,𝜎𝜎
Bending stress
Torsional stiffness
ℎ�𝑤𝑤;𝜏𝜏
Torsional stress ℎ�𝑖𝑖;𝜏𝜏
Young’s Modulus - Glass Eg E E

Shear Modulus G G
Glass Gg
Interlayer Gint
Structural Silicone Gss

Moment of Inertia
(Second moment of area)
Major axis Ix Iy Ix
Minor axis Iy Iz, Jz Iy
Per unit width – includes prime ’ I’
Effective composite 𝐼𝐼𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒

Elastic Section Modulus

Major axis Sx Wy Zx
Minor axis Sy Wz Zy

Torsional Modulus J It, Jt J

Per unit width – includes prime ’ J’
Effective composite 𝐽𝐽𝑒𝑒𝑒𝑒𝑒𝑒

Warping Constant Iw (Cw) Iw Iw

Load application point above shear ya zg yh

center

Lateral restraint spring location above ym zm

shear center

Elastic (Euler) Critical Moment Mcr Mcr Mcr

Reduced Elastic Critical Moment M*cr

Elastic Critical Buckling Moment Cb, C1, C2 C1, C2 g1, g2, g3

modification factors (meaning varies)

Elastic Critical Buckling Axial Load Ncry Pz

Spring stiffness of continuous restraint - kx ky Not used,

lateral assumes rigid

Imperfection auo avo

Separation of Centroids H, H1, H2… H

Design tensile capacity of glass F’g fg,d ΦRu

Shear coupling coefficient

Bending ηb η
Torsion ηt

Lamination geometry coefficient Κ, Κb, Κt

4 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Loading and boundary coefficient
Bending ψb ψ
Torsion ψt

Material capacity reduction factor Φ 1/γ Φ

3. Method for Analysis of Glass Beams and Cantilevers

The method considers multiple components:

• The (Euler) elastic critical moment as a function of support conditions and load profile.
• The effective thickness of a member, taking into consideration the support conditions, load profile
and buckling profile.
• The effects of continuous elastic restraint on the buckling profile and elastic critical buckling
moment.
• The effects of imperfections which are amplified by applied loads.

4. Elastic Critical Buckling of Unrestrained Members

Elastic Buckling Formulas for Unrestrained Beams

There are many available forms of the critical elastic buckling formula (also known as Euler Buckling
Moment). The form used by the European steel code (EN 1993-1-1) was selected as a basis because it
takes into consideration the position of load application and the moment profile and has a broad range
of supporting documents. While US engineers may be more familiar with the AISC (ANSI/AISC 360)
format with a Cb coefficient, accounting for the moment profile, the lack of a provision for load position
was considered a disadvantage for slender rectangular fins without flanges and without torsional
restraint at the lateral restraint, so it was not selected. The equation is re-written using the adopted
axis convention.

𝜋𝜋2 𝐸𝐸𝐼𝐼𝑦𝑦 𝑘𝑘 2 𝐼𝐼 (𝑘𝑘𝑘𝑘)2

𝑀𝑀𝑐𝑐𝑐𝑐 = 𝐶𝐶1 ��� � 𝑤𝑤
+ + (𝐶𝐶2 𝑦𝑦𝑎𝑎 )2 − 𝐶𝐶2 𝑦𝑦𝑎𝑎 � (1)
(𝑘𝑘𝑘𝑘)2 𝑘𝑘𝑤𝑤 𝐽𝐽 𝜋𝜋2 𝐸𝐸𝐼𝐼𝑦𝑦

where:
C1 is a coefficient due to the moment profile (Table 1) (Note: C1 is similar to Cb coefficient (+/-
16% for cases in Table 1) as defined by the AISC standard (ANSI/AISC 360);
C2 is a coefficient due to the moment profile (Table 1);
Eg is the Young’s Modulus of glass;
Gg is the Shear Modulus of glass;
Iy is the second moment of area about the minor axis;
L is the distance between points of bracing of lateral torsional buckling;
J is the Saint Venant’s Torsional Stiffness, evaluated as
J = (B.h3/3) (1-0.63 h/B) (The latter term allowing for beams that are not thin (AS 1288, 2021),

5 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 where
B is the width of the beam,
h is the thickness of the beam,
ya is the position of the load above the shear center when the load is acting toward the shear
center, i.e. ya is positive when the load is applied on the compression side of the beam.
k and kw are effective length factors. (The factor k refers to end rotation on plan. It is analogous to the
ratio of the buckling length to the system length for a compression member. k should be taken as not
less than 1 unless less than 1 can be justified. The factor kw refers to end warping. Unless special
provision for warping fixity is made, kw should be taken as 1. (NCCI SN-003-EN-EU)

Table 1: Reference values for C1 and C2.

Bending Moment C1 C2

Uniform (Constant) 1 0

Linear (zero at midspan) 2.7 0

Parabolic (simply supported, zero at both extremities) 1.127 0.454

Parabolic (fixed end) 2.578 1.554

Triangular (central point load, simply supported) 1.348 0.630

Triangular (central point load, fixed end) 1,683 1.645

For additional coefficients, including beam segments or half-wave moment profile,

reference can be made to NCCI SN-003-EN-EU.

4.1.1. Elastic Critical Buckling Moment for Monolithic Beams

As glass beams are typically un-flanged and the buckling length is “long”, then the warping term is
small and generally ignored, reducing the equation to the form shown in equation (2). (For
continuously restrained beams with short half-wave lengths the warping stiffness can become
appreciable, and the warping component is further discussed in that section.)
For monolithic glass beams, the critical buckling moment may be evaluated as

�π2 𝐸𝐸g 𝐼𝐼y � �𝐺𝐺𝑔𝑔 𝐽𝐽𝐿𝐿2 �

MCR = C1 ∙ ∙ ��(𝐶𝐶2 𝑦𝑦𝑎𝑎 )2 + − 𝐶𝐶2 𝑦𝑦𝑎𝑎 � (2)
𝐿𝐿2 𝜋𝜋2 𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦

4.1.2. Elastic Critical Buckling Moment for Laminated Beams

For laminated glass beams, equation (2) takes the form:

𝜋𝜋2 𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦 𝐺𝐺𝑔𝑔 𝐽𝐽𝑒𝑒𝑒𝑒𝑒𝑒 𝐿𝐿2

𝑒𝑒𝑒𝑒𝑒𝑒
𝑀𝑀𝑐𝑐𝑐𝑐 = 𝐶𝐶1 ��(𝐶𝐶2 𝑦𝑦𝑎𝑎 )2 + − 𝐶𝐶2 𝑦𝑦𝑎𝑎 � (3)
𝐿𝐿2 𝜋𝜋2 𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦
𝑒𝑒𝑒𝑒𝑒𝑒

where the subscript “eff” refers to the effective stiffness.

The Enhanced Effective Thickness (EET) method proposed by Galuppi et al. can be utilized to calculate
the laminated stiffness for both lateral stiffness (Galuppi et al., 2014 and Galuppi et al. 2013), torsional
stiffness (Galuppi et al., 2020), and elements undergoing compressive buckling (D’Ambrosio et al.,

6 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 2020). The EET equations are here reformatted, showing that they can take a common form for 1-
dimensional (beam) bending and torsion. This may be done by introducing a term Κ, defined in this
paper for each of the characteristic stiffnesses being analyzed. Some formulas and derivations from
the papers mentioned above for bending are included for completeness.

Elastic Buckling Formulas for Unrestrained Cantilevers

The most commonly applied formula for cantilever glass fins is from AS1288 Appendix C, which is of
the form:

1
𝑦𝑦ℎ
𝑀𝑀𝐶𝐶𝐶𝐶 = (𝑔𝑔2 /𝐿𝐿𝑎𝑎𝑎𝑎) �(𝐸𝐸𝐸𝐸 )𝑦𝑦 (𝐺𝐺𝐺𝐺)�2 �1 − 𝑔𝑔3 ( )[(𝐸𝐸𝐸𝐸 )𝑦𝑦 /(𝐺𝐺𝐺𝐺)]1/2 � (4)
𝐿𝐿𝑎𝑎𝑎𝑎

(Note that the division sign is missing between the EI/GJ terms in the 2021 printing of AS 1288)

Table 2: Comparison of stability factors.

Loading Slenderness factors (AS1288) Slenderness factors (converted EU)

g2 g3 C1 C2

2.0
for a point load on the cantilever 4.0 1.25 0.318
(Corrected 1.0)

For a distributed load on a cantilever 6.4 2.0 2.037* 0.637

*AS4100 suggests 2.25 for loads applied at the centroid.

However, the line for the point load on a cantilever is contradicted by Timoshenko and Gere which
states:

4.013�𝐸𝐸𝐼𝐼𝜂𝜂 𝐶𝐶 𝐸𝐸𝐼𝐼𝜂𝜂
𝑃𝑃𝐶𝐶𝐶𝐶 = �1 − 𝑎𝑎𝑙𝑙� � (5)
𝑙𝑙 2 𝐶𝐶

Where C = GJ and a is the height of the point load above the centroid.
Noting that 𝑀𝑀𝐶𝐶𝐶𝐶 = 𝑃𝑃𝐶𝐶𝐶𝐶 ∗ 𝑙𝑙 , we can see that the g3 term for a point load should be 1. Finite element
modelling spot checks confirms that the Timoshenko and Gere presentation is the correct one.
For a distributed load Timoshenko and Gere only present the case for the load at the centroid.

12.85�𝐸𝐸𝐼𝐼𝜂𝜂 𝐶𝐶
(𝑞𝑞𝑞𝑞)𝐶𝐶𝐶𝐶 = (6)
𝑙𝑙 2

Noting that 𝑀𝑀𝐶𝐶𝐶𝐶 = (𝑞𝑞𝑞𝑞)𝐶𝐶𝐶𝐶 ∗ 𝑙𝑙/2, we can see a direct correlation and the g3 factor appears to behave
sufficiently accurately in spot checks.
For comparison to the equation of the format in (2), moving one of the π terms into the brackets,

�π𝐸𝐸g 𝐼𝐼y � �𝐺𝐺𝑔𝑔 𝐽𝐽𝐿𝐿2 �

MCR = C1 ∙ ∙ ��(𝐶𝐶2 𝜋𝜋𝑦𝑦𝑎𝑎 )2 + − 𝐶𝐶2 𝜋𝜋𝜋𝜋𝑎𝑎 � (7)
𝐿𝐿2 𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦

7 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 �𝐺𝐺𝑔𝑔 𝐽𝐽𝐿𝐿2 �
And noting that it can be demonstrated that (𝐶𝐶2 𝜋𝜋𝑦𝑦𝑎𝑎 )2 is small relative to , then the forms of
𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦
the stability equations can be related by g2 = C1π and g3 = C2π. As C1 also relates to the moment
modification factor relative a beam with constant moment (also referred to as Cb in AISC 360 and the
Aluminum Design Manual and αm in AS4100), the format in (2) is now more customary, it is presented
in further formulations.
NASA document N.A.C.A Technical Note No 601 (Dumont, 1937) presents the factor K which is the
same as “g2” in AS1288 and C1*π, in the customary European formatting for solid rectangular beams.
(Note K’ is a similar relation to buckling stress and is not related to g3.)

Fig. 1: Stability coefficients for various boundary conditions.

8 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 For glass design, the more common usage is to cantilever glass fins from the ceiling with a combination
of point load at the tip and distributed load on the length. In this instance the load is restrained laterally
which strongly influences the stability. Here the work by Trahair incorporated in AS 4100 is particularly
useful and as the load mechanism provides restraint and cannot move laterally, the C2 (and hence also
g3) terms should be considered to be zero.

Table 3 Cantilever stability coefficients

Slenderness factors Slenderness factors

Loading
(AS1288) (converted EU)

g2 g3 C1 C2

For a point load on the cantilever with end restrained 5.5 - 1.75 -

For a distributed load on a cantilever 11 - 3.5 -

While spot checks for UDL of AS1288 for distributed load on cantilevers appears to have reasonable
agreement, SN006a-EN-EU suggests a more complex interaction with g3, or suggests g3 = 0 for Iww = 0.
Nethercott and Rockney also suggest g3=0. While the warping coefficient, Iww is often ignored for glass
fins due to the slenderness ratio, it is not zero. It is also worth noting for laminated glass fins the
assumption of thin-walled sections where the Saint-Venant torsion constant J = b.t3/3 can be non-
conservative. Comparison of different methods as a function of Iww is well presented in depth by Kraus,
Crisan and Wittor (2021) in Stability Study of Cantilever-Beams – Numerical Analysis and Analytical
Calculation (LTB). To make better use of the table in NCCI it is important to have accurate torsional
properties.

Fig. 2: Comparison of cantilever stability approaches, from Kraus, Crisan and Wittor (2021).

9 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Torsion Constants for Rectangular Shapes
The Saint Venant torsion constant in the most accurate form is an infinite series.

1 192 𝑎𝑎 1 𝑛𝑛𝑛𝑛𝑛𝑛
𝐽𝐽 = 𝑎𝑎3 𝑏𝑏 �1 − ∑∞
𝑛𝑛=1,3,5, tanh � (8)
3 𝜋𝜋5 𝑏𝑏 𝑛𝑛5 2𝑎𝑎

where: a is the minor dimension and b is the major dimension.

There are a variety of simplifications: the simplest and most common is “thin-walled” approximation
for a<<b

1
𝐽𝐽 = 𝑎𝑎3 𝑏𝑏 (9)
3

The version in AS 1288 is:

1 𝑎𝑎
𝐽𝐽 = 𝑎𝑎3 𝑏𝑏 �1 − 0.63 � �� (10)
3 𝑏𝑏

J/a3b vs a/b
0,35

0,3

0,25

0,2 TRUE
AS1288
0,15
Thin-walled
0,1

0,05

0
0 0,2 0,4 0,6 0,8 1 1,2

J/a3b vs a/b
0,35

0,3

0,25

0,2 TRUE
AS1288
0,15
Thin-walled
0,1

0,05

0
0 0,05 0,1 0,15

Fig. 3a & 3b: Comparison of torsion constant (J) formulations.

10 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 For monolithic fins with common aspect ratios of 20:1 (a/b = 0.05) the thin-walled is still valid, however
for laminated assemblies of greater thickness the deviation may become significant, in which case the
version in AS 1288 is adequately accurate (<0.5% deviation) for aspect ratios of 2:1 or greater.
For the warping constant, the authors were unable to find a closed form solution for rectangles of non-
negligible thickness. For I beams, where the thickness of the flange is small relative to the separation
of the flange centroid from the shear center, Iww is well defined by using the centerline approximation
where

𝐼𝐼𝑦𝑦 .𝑑𝑑2
𝐼𝐼𝑤𝑤𝑤𝑤 ≅ ∑ 𝐼𝐼𝑦𝑦.𝑓𝑓𝑓𝑓 . ŷ2𝑓𝑓𝑓𝑓 ≅ 4
𝑓𝑓𝑓𝑓
(11)

Where ŷ𝑓𝑓𝑓𝑓 is the distance from the centroid of the flange of the I-beam from the shear center and
𝑑𝑑𝑓𝑓𝑓𝑓 is the distance between the centroids of the flanges (or the height of the beam less a flange
thickness) This approach is also valid for the layered state laminated assembly with the assumption
that all plies rotate about a common shear center.
For rectangular sections the relationship between Iww is more complex and as the section becomes
more square, symmetry dictates that all corners have to have the same warping function, hence must
be zero. The authors were unable find a closed form solution for Iww, however by conducting multiple
calculations using the membrane analogy, calculation of Iww in the section property of Strand7 (Strauss
7 in Europe), a suitable curve fit can be formed.

𝐵𝐵3 𝐷𝐷3
𝐼𝐼𝑤𝑤𝑤𝑤 = [1 − (2.4649 𝑥𝑥 4 − 6.9103 𝑥𝑥 3 + 5.4827 𝑥𝑥 2 − 0.0567 𝑥𝑥)] (12)
144

where 𝑥𝑥 = 𝑎𝑎/𝑏𝑏, a is the minor dimension and b is the major dimension.

The above equation has an error less than 1% across the range.
1

0,9

0,8

0,7
Correction factor kww

0,6

0,5

0,4 y = 2,4649x4 - 6,9103x3 + 5,4827x2 - 0,0567x

R² = 1
0,3

0,2
Calculation Points
0,1
Poly. (Calculation Points)
0
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
a/b

Fig. 4: Correction Factor kww vs a/b, where Iww = 1/144 a 3b3 (1-kww).

11 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 For most practical glass members with aspect ratios of 5:1 or greater, this is approximated well (<3%
error at 5:1 and for 8:1 or greater there is < 1% error).

𝐵𝐵3 𝐷𝐷3 𝐵𝐵 2
𝐼𝐼𝑤𝑤𝑤𝑤 = �1 − 4.5 � � � (13)
144 𝐷𝐷

0,2

0,18

0,16

0,14
Correction factor kww

y = 2,4649x4 - 6,9103x3 + 5,4827x2 - 0,0567x

0,12 R² = 1

0,1

0,08

0,06

0,04 Calculation Points

Slender Approx
0,02
Poly. (Calculation Points)
0
0 0,02 0,04 0,06 0,08 0,1 0,12 0,14 0,16 0,18 0,2
a/b

Fig. 5: Simplified Iww for slender beams.

(Note that the approximation of Iww = Iyy * D2 /4 that is common for I-beams and is sometimes also
applied to rectangles equates to a correction factor of 48/144 = 0.33, so is conservative for warping
stiffness for a/b < 0.3.)

Summary of Stability Factors for Elastic Critical Buckling

Table 4: Values for C1 and C2 - Beams

Bending Moment Minor axis rotational fixity (ry) C1 C2

Free 1
Uniform (Constant) 0
Fixed 2

Free
Linear (zero at midspan) 2.7 0
Fixed

Free 1.127 0.454

Parabolic (simply supported, zero at both extremities)
Fixed 1.942 0.573

Free 1.301 1.554

Parabolic (fixed end)
Fixed 1.719 1.655

Free 1.348 0.630

Triangular (central point load, simply supported)
Fixed 2.132 0.828

Free 1,683 1.645

Triangular (central point load, fixed end)
Fixed 2.069 1.687

12 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 For additional coefficients for beams reference: NCCI SN-003-EN-EU

Table 5: Values for C1 and C2 - Cantilevers

Bending Moment Profile Free end lateral fixity (dx) C1 C2

Free 1.27 0.318*

Cantilever Point load at end
Fixed 1.75 0

Free 2.040 0.637

Cantilever uniform load
Fixed 3.500 0

For additional coefficients for cantilevers reference: NCCI SN-006-EN-EU

NOTES:

For tapered cantilevers further information is available in Timoshenko and Gere (1956)

C1 is the same as g2/π in AS1288 and C2 is the same as g3/π in AS1288.

* Corrections to coefficients in AS1288 are based on the work of Timoshenko and Gere and as tested against finite element models.

The coefficients for cantilevers with lateral end restraint is based on AS4100 reflecting the work of Trahair.

5. Enhanced Effective Thickness (EET) – Bending

Depending on the degree of shear coupling of the glass plies through the interlayer, the out-of-plane
bending response of a laminated glass element is intermediate between that of free sliding plies
(layered behavior) and that of a glass monolith (monolithic behavior) of the total thickness. In
particular, by considering a laminated glass element, of unit width, composed of N glass plies:

• For the layered behavior, the moment of inertia with respect to the minor axis is:

𝐼𝐼 ′ 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = ∑𝑁𝑁 ′
𝑖𝑖=1 𝐼𝐼 𝑖𝑖 (14)

where: I’i = hi 3/12 [mm4/mm (or in4/in)] is the moment of inertia per unit width of the i-th
glass ply [mm4/mm] (notation of Figure 1). (The “prime” mark indicates “per unit width”.)
Note that, while metric units are presented here, any consistent unit system can be used.

• For the monolithic limit, the moment of inertia is:

𝐼𝐼 ′ 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = 𝐼𝐼′𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 + 𝐼𝐼′𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (15)

Where:

𝑁𝑁
𝐼𝐼′ 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = �𝑖𝑖=𝑖𝑖 ℎ𝑖𝑖 𝑑𝑑𝑖𝑖2 (16)

and di [mm] is the distance of the centroid of the i-th ply from the centroid of the cross-section
of the laminated package (Figure 6), while hi is the area for a strip of unit width (B = 1).
Also, Ai =B hi /1 [mm2/mm] = hi [mm2/mm] is the area per unit width of the cross-section of the i-th
plate, i.e. the ply thickness of the i-th ply.

13 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Fig. 6: Geometry of a laminated glass element (cross-sectional view).

In Figure 6, hi (i=1..N) is the thickness of the i-th ply of glass, hi (i=1..N) is the thickness of the i-th
interlayer, yi is the position of the centroid of the i-th ply, Yc is the position of the centroid of the glass
plies and di is the distance (with sign) from the overall centroid (Yc) to the centroid of the i-th ply. B is
the width of the assembly.
The distances di may be evaluated as:

∑𝑁𝑁
𝑖𝑖=1 𝑦𝑦𝑖𝑖 ℎ𝑖𝑖
𝑑𝑑𝑖𝑖 = 𝑦𝑦𝑖𝑖 − 𝑌𝑌𝑐𝑐 , where 𝑌𝑌𝑐𝑐 = ∑𝑁𝑁
(17)
𝑖𝑖=1 ℎ𝑖𝑖

For laminated glass beams, the EET method defines the equivalent moment of inertia as the harmonic
mean of the moment of inertia of the cross-section at the monolithic and layered limit, with the
harmonic mean weighted using a shear transfer coefficient, accounting for the degree of coupling
between glass plies due to the presence of the interlayer: For laminated glass beams, the EET method
defines the equivalent moment of inertia as the harmonic mean of the moment of inertia of the cross-
section at the monolithic and layered limit, with the harmonic mean weighted using a shear transfer
coefficient, accounting for the degree of coupling between glass plies due to the presence of the
interlayer: 𝐼𝐼𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 is the effective moment of inertia of the laminated package, assuming an intermediate
value between Itotal and Iplies.
The effective moment of inertia may be written equivalently as:

1 𝜂𝜂𝑏𝑏 1−𝜂𝜂𝑏𝑏
= + (18)
𝐼𝐼′ 𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 𝐼𝐼 ′ 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝐼𝐼 ′ 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

1
𝐼𝐼 ′ 𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 = 𝜂𝜂𝑏𝑏 1−𝜂𝜂𝑏𝑏
(19)
�′ +′ �
𝐼𝐼 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝐼𝐼 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

where 𝜂𝜂𝑏𝑏 is a non-dimensional coefficient (suffix b refers to “bending”) depending on the

geometry of the beam, on the loading and boundary conditions, and on the mechanical
properties of glass and interlayer. The value of 𝜂𝜂𝑏𝑏 ranges from 0 (layered limit) and 1
(monolithic limit).

14 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

5.1.1. Shear Coupling Coefficient for Beams
The shear coupling coefficient ηb is a function of the geometry of the laminate and the properties of
the interlayer (Galuppi et al. 2014)]

1
𝜂𝜂𝑏𝑏 = 𝐸𝐸𝑔𝑔 .𝐼𝐼′ 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 .𝐼𝐼′ 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
(20)
1+ 𝛹𝛹
𝐺𝐺int �𝐼𝐼′ total �.𝛫𝛫𝑏𝑏

where:
Eg is the Young’s modulus of glass;
Gint is the relaxation shear modulus of the interlayer for the applicable temperature and
duration;
Κb is a function of the glass and interlayer geometry defined as:

(𝑦𝑦𝑖𝑖+1 − 𝑦𝑦𝑖𝑖 )2 𝐻𝐻𝑖𝑖 2�

𝛫𝛫𝑏𝑏 = ∑𝑁𝑁−1
𝑖𝑖=1 �ℎ = ∑𝑁𝑁−1
𝑖𝑖=1 ℎ𝑖𝑖𝑖𝑖𝑖𝑖,𝑖𝑖 (21)
𝑖𝑖𝑖𝑖𝑖𝑖,𝑖𝑖

The coefficient ψ , appearing in eq. 20, is a function of the beam length L, the loading profile and the
boundary and loading, tabulated in Table 6 for the case of interest. For lateral-torsional buckling** 1
use π2/L2 , as in Nizich (2022).

Table 6: Laminated Beams Boundary Conditions: values of coefficient Ψ.

Loading and boundary conditions ψ

168
17 𝐿𝐿2

𝜋𝜋 2
𝐿𝐿2

𝑛𝑛𝑅𝑅 2 𝜋𝜋 2
𝐿𝐿2

NOTE: Refer D’Ambrosio et al. (2020) and Galuppi et al. (2013) for different loading and boundary conditions.

*1
Note for systems with multiple half-wavelengths, use the distance between points of contraflexure as “L”. For not-
restrained laminated glass fins subjected to lateral-torsional buckling, the effective bending inertia 𝐼𝐼𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 may be evaluated
by using coefficient ψ=π2⁄a2, where a is the half wave-length, as per Table 6. As ψ will affect the effective thickness and may
change nR, the solution may be iterative.

15 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

5.1.2. Deflection - Effective Thickness for Beams in Bending
The deflection-effective thickness is defined as the thickness of the monolithic glass exhibiting the same
maximum deflection of the considered laminated glass element under bending. Obviously, the
deflection is inversely proportional to the moment of inertia and, therefore, to the cube of the
thickness of the equivalent monolith. The deflection-effective thickness is then equal to:

ℎ�𝑤𝑤 = �12. 𝐼𝐼 ′ 𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒

3
(22)

5.1.3. Stress - Effective Thickness for Beams in Bending

The stress-effective thickness is defined as the thickness of monolithic glass exhibiting the same
maximum stress in one of the plies of the considered laminated glass element. In general, one defines
a stress-effective thickness for each glass ply. The stress-effective thickness of the i-th plate is given
by:

6
ℎ�𝑖𝑖;𝜎𝜎 = (23)
� 𝜂𝜂′𝑏𝑏�𝑑𝑑𝑖𝑖 � + ℎ𝑖𝑖
𝐼𝐼 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 2.𝐼𝐼′ 𝑦𝑦
𝑒𝑒𝑒𝑒𝑒𝑒

Enhanced Effective Thickness – Torsion*

5.2.1. Effective Torsional Stiffness

Depending on the degree of shear coupling of the glass plies through the interlayer, the torsional
response of a laminated glass element is intermediate between that of free sliding plies (layered
behavior) and that of a monolith (monolithic behavior) of the total thickness. In particular, for a
laminated glass element with unit width, composed of N plies of glass:

• For the layered limit:

1 𝑁𝑁
𝐽𝐽′ 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = ∑𝑁𝑁
𝑖𝑖=1 𝐽𝐽′𝑖𝑖 = �𝑖𝑖=1 ℎ𝑖𝑖3 (24)
3

where J’i = hi 3/3 [mm4/mm] is the torsional moment of inertia of the i-th glass ply per unit width
[mm4/mm] and B is the width of the ply (notation of Figure 1).

• For the monolithic limit

𝐽𝐽′ 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = 𝐽𝐽′ 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 + 4 𝐼𝐼 ′ 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (25)

𝑁𝑁
1 𝑁𝑁 ℎ𝑖𝑖3 𝑁𝑁
𝐽𝐽′ 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 = �𝑖𝑖=1�ℎ𝑖𝑖3 + 12𝑑𝑑𝑖𝑖2 ℎ𝑖𝑖 � = � � � + 4 �𝑖𝑖=1 𝑑𝑑𝑖𝑖2 ℎ𝑖𝑖 (26)
3 𝑖𝑖=1 3

According to the Enhanced Effective Thickness approach for torsion (Galuppi et al., 2020), the effective
torsional stiffness of the laminated element may be evaluated by considering:

1 𝜂𝜂𝑡𝑡 1−𝜂𝜂𝑡𝑡
= + , (27)
𝐽𝐽′ 𝑒𝑒𝑒𝑒𝑒𝑒 𝐽𝐽′ 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝐽𝐽′ 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

16 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 1
𝐽𝐽′ 𝑒𝑒𝑒𝑒𝑒𝑒 = (28)
𝜂𝜂𝑡𝑡 1−𝜂𝜂𝑡𝑡
� ′ + ′ �
𝐽𝐽 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝐽𝐽 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

where J’total is the Saint Venant (uniform) torsional stiffness per unit width at the monolithic limit
(defined by eq. (25); and where J’plies is the Saint Venant (uniform) torsional stiffness per unit width for
the glass plies (defined by eq. (24)). The torsional stiffness of the beam is

Jeff = J’eff * B (29)

where B is the breadth.

This provides the effective section properties to calculate the critical elastic buckling moment for
laminated fins. Notice that the EET method for torsion presented by Galuppi et al. (2020) is based on
𝐵𝐵ℎ𝑖𝑖 3
the thin-walled approximation of 𝐽𝐽𝑝𝑝𝑝𝑝𝑝𝑝 = .. However, AS1288 presents a more conservative and
3
accurate version of the formula, that is:

𝐵𝐵ℎ𝑖𝑖 3 𝐵𝐵
𝐽𝐽𝑝𝑝𝑝𝑝𝑝𝑝 = �1 − 0.63 � (30)
3 ℎ𝑖𝑖

Thus, the equivalent term that could be substituted in equation (25) is:

1 𝑁𝑁 𝐵𝐵
𝐽𝐽′ 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 = �𝑖𝑖=1 ℎ𝑖𝑖3 �1 − 0.63 � (31)
3 ℎ𝑖𝑖

If this format is used, it should also be used consistently when back-solving the effective thickness for
torsion (35), but that is not presented here because only the torsional stiffness Jef is used in subsequent
equations.
The shear coupling function previously presented by Galuppi et al. (2020) can be rewritten as

1
𝜂𝜂𝑡𝑡 = 𝐸𝐸𝑔𝑔 𝐽𝐽′ 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 . 𝐼𝐼′ 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐
(32)
1+ 𝛹𝛹𝑡𝑡
�1−𝜈𝜈2 �𝐺𝐺 𝑖𝑖𝑖𝑖𝑖𝑖 𝐽𝐽′ 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡. Κ𝑡𝑡

where:

𝑁𝑁−1
𝛫𝛫𝑡𝑡 = 𝛫𝛫𝑏𝑏 = �𝑖𝑖=1 𝐻𝐻𝑖𝑖2 /ℎ𝑖𝑖𝑖𝑖𝑖𝑖,𝑖𝑖 (33)

and

𝐿𝐿2 +𝐵𝐵2
𝛹𝛹𝑡𝑡 = 6(1 − 𝜈𝜈) (34)
𝐿𝐿2 𝐵𝐵2

where B is the width; N is the number of glass plies, and N-1 is the number of interlayers.
Notice that the expression (33) for the parameter Kt coincides with that obtained for the bending
problem (eq. 21). This allows for a comprehensive formulation of the EET model, for both bending and
torsion, that is proposed here for the first time.
NOTE: This section for effective torsional stiffness is included for stability calculations. For applied torsional
loads, the use of layered finite element methods or similar is recommended to capture longitudinal stresses
due to warping of each ply, which are not calculated in this method.

17 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

5.2.2. Effective Thickness for Torsion Deflection
The torsional EET ℎ�𝑤𝑤;𝜏𝜏 defined as the thickness of a monolithic glass prism of width B and length L,
having the same geometric torsional constant of the laminated glass element, is given by formula:

ℎ�𝑤𝑤;𝜏𝜏 = 3�3 𝐽𝐽′ 𝑒𝑒𝑒𝑒 (35)

If the thick-ply formula from AS1288 presented at equation (31) is substituted at equation (25), then
the same equation should be used to iteratively back-calculate the effective thickness for torsion
deflection.

5.2.3. Effective Thickness for Torsional Stress

The thickness of a monolith for which the maximum shear stress in the i-th glass ply equals the
maximum shear stress in glass for the laminated package is:

3
�𝑤𝑤
ℎ
ℎ�𝑖𝑖;𝜏𝜏 = � (36)
ℎ𝑖𝑖 +α|𝑑𝑑𝑖𝑖 |

2 2
α= 𝑁𝑁 = 𝐸𝐸 𝐼𝐼𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 (37)
� ℎ 𝑑𝑑2 1+ 𝛹𝛹𝑡𝑡
𝐸𝐸 𝑖𝑖=1 𝑖𝑖 𝑖𝑖 �1−𝜈𝜈2 �𝐺𝐺𝑖𝑖𝑖𝑖𝑖𝑖 𝛫𝛫𝑡𝑡
1+ 𝛹𝛹𝑡𝑡
�1−𝜈𝜈2 �𝐺𝐺𝑖𝑖𝑖𝑖𝑖𝑖 𝑁𝑁−1
� 𝐻𝐻2 /ℎ
𝑖𝑖=1 𝑖𝑖 𝑖𝑖𝑖𝑖𝑖𝑖,𝑖𝑖

Continuously Elastically Restrained Fins

Bedon et al. (2015) proposes the equations for fins with continuous elastic lateral restraint under
uniform moment in the following form: (notation altered for US customary axis system, where “z” is
the longitudinal axis - see Glossary.)

5.3.1. Compression Flange Continuously Restrained by a Spring

For fins continuously restrained on the compression flange by a (distributed) spring (Bedon et al., 2015),
as shown in Figure 3, the critical buckling moment can be evaluated as:

𝐿𝐿 2 𝑛𝑛𝑅𝑅 𝜋𝜋 2 𝐿𝐿 2 𝐿𝐿 2
2
𝑀𝑀𝑐𝑐𝑐𝑐,𝑅𝑅 = 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � + ��𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦 � � + 𝑘𝑘𝑥𝑥 � � � �𝐺𝐺𝑔𝑔 𝐽𝐽 + 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � � (38)
𝑛𝑛𝑅𝑅 𝜋𝜋 𝐿𝐿 𝑛𝑛𝑅𝑅 𝜋𝜋 𝑛𝑛𝑅𝑅 𝜋𝜋

where

• ym is the height of the spring above the shear center (positive on the compression flange
side);
• kx is the stiffness of the spring per unit length;
• nR is the number of half wavelengths of the buckled mode; nR is an integer, try different
values for nR and use the minimum value for Mcr,R;
• when calculating the effective properties for laminated glass, the composite action
development length shall not exceed the half wave-length ‘a’ = L/nR.

18 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Fig. 7: Typical façade glass to beam connection by means of continuous silicone joints. (a) Overview and (b) analytical model
(cross-section). Figure adapted from Bedon et al. (2015) and revised for US sign conventions.

Notice that only the lateral stiffness of the silicone between the beam and façade glass is included in
the derivation, the torsional resistance is not included, and the weather seal between the façade glass
is not assumed to participate. If there is more than one piece of façade glass along the length of the
beam, the stiffness of the façade glass to restrain the beam should be assessed separately, similar to
springs in series.
Notice that the C2 terms, appearing in eq.s (1) and (2) for the case of unrestrained beam, related to
load position, does not appear in this formulation. This is because the load is restrained from moving
laterally.

5.3.2. Tension Flange Continuously Restrained by a Spring

Bedon et al. (2015) also develops a form for “reverse moment” with continuous elastic restraint of the
tension flange:

𝐿𝐿 2 𝑛𝑛𝑅𝑅 𝜋𝜋 2 𝐿𝐿 2 𝐿𝐿 2
2
𝑀𝑀𝑐𝑐𝑐𝑐,𝑅𝑅 = 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � − ��𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦 � � + 𝑘𝑘𝑥𝑥 � � � �𝐺𝐺𝑔𝑔 𝐽𝐽 + 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � � (39)
𝑛𝑛𝑅𝑅 𝜋𝜋 𝐿𝐿 𝑛𝑛𝑅𝑅 𝜋𝜋 𝑛𝑛𝑅𝑅 𝜋𝜋

where the input terms are as defined above.

19 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

5.3.3. Combined Equation
The above equations (27) and (28) can be efficiently consolidated into a single equation:

𝑛𝑛𝑅𝑅 𝜋𝜋 2 𝐿𝐿 2
2 𝐿𝐿 2 𝐿𝐿 2
𝑀𝑀𝑐𝑐𝑐𝑐,𝑅𝑅 = ��𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦 � � + 𝑘𝑘𝑥𝑥 � � � �𝐺𝐺𝑔𝑔 𝐽𝐽 + 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � � + 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � (40)
𝐿𝐿 𝑛𝑛𝑅𝑅 𝜋𝜋 𝑛𝑛𝑅𝑅 𝜋𝜋 𝑛𝑛𝑅𝑅 𝜋𝜋

where the term ym is positive for restraint on the compression side of the beam and ym is negative for
the reverse moment case or where the restraint is on the tensile side of the beam.
Note that substituting in nR = 1 and ym = 0, Eq. (40) reduces to the same classical equation of critical
buckling moment of an unrestrained beam (1 half wavelength) under uniform moment. When
substituting a large value for kx, Eq. (40) also approaches the results for continuously restrained fins in
AS1288, which is believed to have originated with timber beams nailed to sheathing, i.e. it ignores the
flexibility of the silicone.

In Bedon’s paper, it is suggested that capacity modification factor factors can be applied for non-uniform moment profiles
in the same manner as an unrestrained beam, however in benchmarking it is the moment profile over the critical half-
wavelength (similar to local buckling) that is relevant for accurate prediction of the elastic buckling moment. For simply
supported beams under uniform load, for nR ≥3 the moment over the critical segment is sufficiently uniform for it to be of
limited benefit. Using the formula for Cb in AISC 360 and parabolic moment profile yields the correction factors in Table 7,
which can be used in place of C1 in equations (41) and (42).

Table 7: Capacity modification factor, Cb, as a function of number of half-wave lengths

for a simply supported beam with uniformly distributed load.

nR 1 2 3 4 5 6

Cb 1.136 1.299 1.014 1.061 1.005 1.026

Laminated Beams Continuously Restrained by a Spring

For laminated beams with non-uniform moment this becomes:

𝑛𝑛𝑅𝑅 𝜋𝜋 2 𝐿𝐿 2
2 𝐿𝐿 2 𝐿𝐿 2
𝑀𝑀𝑐𝑐𝑐𝑐𝑅𝑅,𝑙𝑙𝑙𝑙𝑙𝑙 = 𝐶𝐶1 ��𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 � � + 𝑘𝑘𝑥𝑥 � � � �𝐺𝐺𝑔𝑔 𝐽𝐽𝑒𝑒𝑒𝑒𝑒𝑒 + 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � � + 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � (41)
𝐿𝐿 𝑛𝑛𝑅𝑅 𝜋𝜋 𝑛𝑛𝑅𝑅 𝜋𝜋 𝑛𝑛𝑅𝑅 𝜋𝜋

Where, again, ym >0 for restraint on the compression side, and ym <0 for the “reverse moment” case
or where the restraint is on the tensile side of the beam and C1 is a function of the moment profile in
the critical half-wave. (See also table 7)
Because the differential shear between the plies that mobilizes the interlayer under curvature reverses
direction at the point of contra-flexure of the beam, the length parameter must be taken as the half-
wavelength, not the full length of the beam when calculating the enhanced effective thickness
properties. This means that 𝐼𝐼𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 is evaluated, by means of the EET method, by using coefficient 𝜓𝜓 =
𝜋𝜋 2� , where 𝑎𝑎 is the half wave-length, as per Table 2, and 𝑎𝑎 = 𝐿𝐿� .
𝑎𝑎2 𝑛𝑛𝑅𝑅
Using the same spreadsheet for both the effective section properties and the critical moment allows
rapid testing of different nR values in order to find the respective effective section properties and the
critical elastic buckling moment.

20 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 For systems which have a relatively short half-wave length and high nR, the effective St Venant
torsional stiffness can become small and the warping stiffness can become appreciable. In this case,
an expanded form of the equation including the warping constant is proposed:

𝑛𝑛𝑅𝑅 𝜋𝜋 2 𝐿𝐿 2 𝐿𝐿 2
2 𝐿𝐿 2
𝑀𝑀𝑐𝑐𝑐𝑐𝑅𝑅,𝑙𝑙𝑙𝑙𝑙𝑙,𝐼𝐼𝐼𝐼 = 𝐶𝐶1 ��𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 � � + 𝑘𝑘𝑥𝑥 � � � �𝐸𝐸𝑔𝑔 𝐼𝐼𝑤𝑤 � � + 𝐺𝐺𝑔𝑔 𝐽𝐽𝑒𝑒𝑒𝑒𝑒𝑒 + 𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � � +
𝐿𝐿 𝑛𝑛𝑅𝑅 𝜋𝜋 𝑛𝑛𝑅𝑅 𝜋𝜋 𝑛𝑛𝑅𝑅 𝜋𝜋
𝐿𝐿 2
𝑦𝑦𝑚𝑚 𝑘𝑘𝑥𝑥 � � (42)
𝑛𝑛𝑅𝑅 𝜋𝜋

This case then raises the question of how to calculate the warping constant Iw for a laminated fin. While
solid rectangles have a non-zero warping constant, the authors are not aware of a formula for their
calculations for laminated elements. For the fully composite state, finite element solutions, such as the
beam section generator in Strand7 (Strauss7 in Europe) may be used.
The warping constant for the layered state is well approximated by the warping function of the
centerlines of the plies. Similar to the warping constant of an I-beam, the warping constant of
symmetric layered fins is:

ℎ𝑖𝑖 𝐵𝐵3
𝐼𝐼𝑤𝑤;𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 ≅ ∑𝑛𝑛𝑖𝑖=1 𝑑𝑑𝑖𝑖 2 (43)
12

For non-symmetric laminates it is required to solve the fundamental equations with the net axial thrust
of each ply being zero, to give the additional equations to be able to solve the matrix for the shear
center. That is not covered in this paper and evaluation by numerical methods is a more accurate
alternative.
For the cases tested at the time of writing, the interpolation function has not been fully developed,
however noting that:
• both the layered and the solid have warping constants of similar magnitude;
• both Saint-Venant torsion and warping constants are dominated by shear across the width of the
section;
• the effect of the warping constant is relatively small, so some inaccuracy will have a small influence
in the overall result;
• it is suggested that using the interpolation function for Saint Venant torsion may be sufficiently
accurate until a better option is developed. (Note for very critical cases, confirmation by testing and
layered finite element models is recommended.)

1
𝐼𝐼𝑤𝑤 𝑒𝑒𝑒𝑒 ≅ (44)
𝜂𝜂𝑡𝑡 1−𝜂𝜂𝑡𝑡
� + �
𝐼𝐼𝑤𝑤;𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝐼𝐼𝑤𝑤;𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

21 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Spring Stiffness
Typically, the spring is structural silicone to the façade glass. Where the façade glass is a single piece
of glass that is fixed laterally, the spring is well approximated by the stiffness of the silicone. For shear
stiffness, the constitutive response is approximately linear.
The local spring stiffness per unit length can be calculated as:

𝑤𝑤𝑠𝑠𝑠𝑠𝑠𝑠
𝑘𝑘𝑥𝑥 = 𝐺𝐺𝑠𝑠𝑠𝑠 . ∑ �ℎ (45)
𝑠𝑠𝑠𝑠𝑠𝑠

where:
Gss is the shear stiffness of the structural silicone;
∑ 𝑤𝑤𝑠𝑠𝑠𝑠𝑠𝑠 is the total bite of the silicone;
hsil is the glue-line thickness of the silicone (See also Figure 7).

5.5.1. Multiple Facade Glass Panels

The above formula is for the spring stiffness between the fin and facade glass that braces it. Where
there are multiple pieces of façade glass along the length, the flexibility of the support from the façade
glass needs to be considered in the spring constant. (Similar to springs in series reducing the apparent
stiffness.)

5.5.2. Spring Stiffness of a Silicone Bite in Tension/Compression

In this configuration, the stiffness is non-linear. It is suggested to consult with the manufacturer for
suitable values. However, it is conservative to use a lower bound approximation of the stiffness.

6. Imperfect Beams
With the effective section properties and critical elastic buckling moment, the capacity can be
calculated by considering the initial imperfections. As the critical buckling moment is approached, the
second-order effects become significant in calculating the total tensile stress that limits the capacity
of the beam.
The Kala equation for imperfect beams (Kala, 2013) provides a method for including the second order
effects as a function of target stress level, the level of imperfection, section properties, and the elastic
critical buckling moment.

∗ 𝐷𝐷 )
�4𝐷𝐷12 +(𝐷𝐷4 +𝐷𝐷5 )2 +4𝐷𝐷1 (𝐷𝐷4 −2𝑀𝑀𝑐𝑐𝑐𝑐 3 2𝐷𝐷1 +𝐷𝐷4 +𝐷𝐷5
𝑀𝑀𝑛𝑛 = − ∗ 𝑆𝑆 + ∗ 𝑆𝑆 (46)
4𝑀𝑀𝑐𝑐𝑐𝑐 𝑦𝑦 4𝑀𝑀𝑐𝑐𝑐𝑐 𝑦𝑦

with:
𝐷𝐷1 = 𝑓𝑓 ′ 𝑔𝑔 𝑀𝑀𝑐𝑐𝑐𝑐
∗
𝑆𝑆𝑥𝑥 𝑆𝑆𝑦𝑦
∗
𝐷𝐷2 = 𝑀𝑀𝑐𝑐𝑐𝑐 𝑆𝑆𝑦𝑦 + 𝑁𝑁𝑐𝑐𝑐𝑐𝑐𝑐 |𝑎𝑎𝑢𝑢0 |𝑆𝑆𝑥𝑥
∗
𝐷𝐷3 = 𝑀𝑀𝑐𝑐𝑐𝑐 𝑆𝑆𝑦𝑦 − 𝑁𝑁𝑐𝑐𝑐𝑐𝑐𝑐 |𝑎𝑎𝑢𝑢0 |𝑆𝑆𝑥𝑥
2
𝐷𝐷4 = 2𝑁𝑁𝑐𝑐𝑐𝑐𝑐𝑐 𝐼𝐼𝑥𝑥 |𝑎𝑎𝑢𝑢0 |

22 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 ∗
𝐷𝐷5 = 2𝑀𝑀𝑐𝑐𝑐𝑐 𝐷𝐷2
𝐸𝐸𝑔𝑔 𝐼𝐼𝑦𝑦
𝑁𝑁𝑐𝑐𝑐𝑐𝑐𝑐 = 𝜋𝜋 2
𝐿𝐿2
where:
Ix is the second moment of area about the X axis (strong axis);
Iy (𝐼𝐼𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 for laminated sections) is the (effective) second moment of area about the Y axis
(weak axis);
Sx is the elastic section modulus about the X axis (strong axis);
Sy is the (effective) elastic section modulus about the Y axis (weak axis), for laminated glass
use

2
ℎ�𝑖𝑖;𝜎𝜎;𝑚𝑚𝑚𝑚𝑚𝑚 �
𝑆𝑆𝑦𝑦𝑒𝑒𝑒𝑒𝑒𝑒 = 6 (47)

where:

ℎ�𝑖𝑖;𝜎𝜎;𝑚𝑚𝑚𝑚𝑚𝑚 is the minimum stress-effective thickness, evaluated according to eq. (23).

𝐹𝐹 ′𝑔𝑔 is the glass design tensile strength (i.e. the target stress for a given level of reliability);

For the purpose of 𝑁𝑁𝑐𝑐𝑐𝑐𝑐𝑐 (appearing in the definition of D2, D3 and D4), L is the overall length of the
beam, not the half-wave length, because the unrestrained flange will buckle in a single half-wave (L)
in the critical case (assuming no torsional restraint from the silicone). For laminated glass beams, Ncry
𝜋𝜋2
may be evaluated by adopting the EET approach (see Sect. 2.2), with 𝜓𝜓 = , or by adopting other
𝐿𝐿2
168
models proposed by the literature [15]. Note that 𝜓𝜓 for weak axis beam bending stiffness is =
17𝐿𝐿2
9.882 𝜋𝜋2 9.870
, is only a 1.3% difference to 𝜓𝜓 = = for axial buckling. The difference in effective stiffness
𝐿𝐿2 𝐿𝐿2 𝐿𝐿2
is even smaller and, for simplicity and allowing for the non-uniform compression along the beam, the
more conservative beam value was used in the numerical comparison in section 4 below.
As the imperfection parameter auo needs to capture the maximum of the torsional imperfection, lateral
imperfection or combination thereof, the following definitions are proposed to capture the worst case.
The maximum assumed design imperfection, auo, is measured perpendicular to the Y axis as defined
below in figure 3:

23 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Fig. 8: Imperfect Beams’ Characteristic Dimension, auo.

NOTE: At the time of writing it has been noted that due to the high slenderness of glass fins, with large
differences between Ix and Iy, that for unrestrained beams it is important to consider application of
loads to the principal axies with the associated minor axis bending stress. However in the context of a
continuously restrained beam, the formulation previously stated by Kala is acceptable with suitable
reductions as suggested below.

Parametric Analysis
For this case of a theoretical 6 m beam with breadth of 450 mm, and a construction of 9.02 mm+1.52
mm+9.02 mm laminated glass (ASTM minimum thickness for 3/8” (10mm) glass), the interlayer having
a stiffness of 3 MPa, with 2 structural silicone bites of 6 mm and an 8 mm glue-line which have a shear
modulus of 0.3 MPa, and an initial imperfection of 12 mm (L/500), figure 4 shows the nominal moment
design capacity, for both tension and compression edges, as a function of the glass design stress.

Fig. 9: Nominal Moment Capacity vs Glass Design Capacity (F'g) for beams with continuous elastic restraint to compression
edge (Mn.comp) and tension edge (Mn.tens).

24 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Figures 10a-f are for the same scenario described above for figure 9 with nominated input parameters
varied.

Mn.compression v F'g , Various imperfections

Nominal Moment Capacity (Nmm)

8,00E+07
7,00E+07
6,00E+07
5,00E+07
4,00E+07
3,00E+07
2,00E+07
1,00E+07
0,00E+00
0 50 100 150 200 250 300
Glass Design Capacity (MPa)
Nil imperfection L/700 L/600 L/500 L/400 L/300 L/200 L/100

Fig. 10a. Nominal Moment Capacity (Nmm) with compression edge restrained vs Glass Design Capacity (F'g) (MPa),
for different values of the initial imperfection.

Mn.comp v F'g , Various imperfections

Nominal Moment Capacity (Nmm)

8,00E+07
7,00E+07
6,00E+07
5,00E+07
4,00E+07
3,00E+07
2,00E+07
1,00E+07
0,00E+00
0 50 100 150 200 250 300
Glass Design Capacity (MPa)
Nil imperfection L/700 L/600 L/500 L/400 L/300 L/200 L/100

Fig. 10b: Nominal Moment Capacity (Nmm) with tension edge restrained vs Glass Design Capacity (F'g) (MPa),
for different values of the initial imperfection.

25 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Mn.compression v F'g, Various silicone bites (mm)
1,20E+08
Nominal Moment Capacity (Nmm)
1,00E+08

8,00E+07

6,00E+07

4,00E+07

2,00E+07

0,00E+00
0 50 100 150 200 250 300
Glass Design Capacity (MPa)

0 6 8 10 12 16 20 24

Fig. 10c: Nominal Moment Capacity (Nmm) with compression edge restrained vs Glass Design Capacity (F'g) (MPa),
for different thickness of the silicone bite (mm).

6,00E+07 Mn.tension v F'g, Various silicone bites

Nominal Moment Capacity (Nmm)

5,00E+07

4,00E+07

3,00E+07

2,00E+07

1,00E+07

0,00E+00
0 50 100 150 200 250
Glass Design Capacity (MPa)
0 6 8 10 12 16 20 24

Fig. 10d: Nominal Moment Capacity (Nmm) with tension edge restrained vs Glass Design Capacity (F'g) (MPa),
for different thickness of the silicone bite (mm).

26 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Mn.compression v F'g, Various Gint (MPa)
Nominal Moment Capacity (Nmm) 1,40E+08
1,20E+08
1,00E+08
8,00E+07
6,00E+07
4,00E+07
2,00E+07
0,00E+00
0 50 100 150 200 250 300
Glass Design Capacity (MPa)

0.00 0.1 0.2 0.3 0.5 1 2

3 5 10 20 30 50 100

Fig. 10e: Nominal Moment Capacity (Nmm)with compression edge restrained vs Glass Design Capacity (F'g) (MPa),
for different shear moduli of the interlayer (MPa).

Mn.tension v F'g, Various Gint (MPa)

1,40E+08
Nominal Moment Capacity (Nmm)

1,20E+08
1,00E+08
8,00E+07
6,00E+07
4,00E+07
2,00E+07
0,00E+00
0 50 100 150 200 250 300
Glass Design Capacity (MPa)

0.00 0.1 0.2 0.3 0.5 1 2

5 10 20 30 50 100

Fig. 10f: Nominal Moment Capacity (Nmm) with tension edge restrained vs Glass Design Capacity (F’g) (MPa),
for different shear moduli of the interlayer (MPa).

From Figures 10a and 10b, it can be observed that imperfections can significantly increase the stress
for a given moment capacity approaching the buckling load, a phenomenon not captured in the AS1288
method.
From Figures 10c and 10d it can be observed that the introduction of any silicone bite has a significant
effect and then added stiffness a lessor effect.
From Figures 10e and 10f, it can be observed that the shear stiffness of the interlayer has a significant
impact on the moment capacity of the beam.

27 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 It should be noted that in Kala’s derivation (Kala, 2013), initially both the lateral imperfection and the
torsional imperfection are defined separately. In the final version of the formula, the torsional
imperfection is assumed to be a function of the lateral imperfection. In benchmarking, the
approximation works well for imperfections that resemble the primary buckling mode, including the
number of wavelengths where appropriate, but can be somewhat conservative or non-conservative
for other imperfections. To compensate, it is suggested to use a reduced Mcr to account for potential
variations in the actual imperfection shape. A 15% reduction is proposed and has been found to be
adequate for a limited number of checks for imperfections of practical magnitude but has not been
tested exhaustively. Below, in figures 11b, 11c and 11d are examples of the comparison of the analytic
method proposed here with finite element using layered brick models of the laminated glass fin,
prepared using Strand7/Strauss7.

Fig. 11a: The seed imperfection was the primary buckling mode of the unbraced beam under negative load scaled to 12mm
for all cases.

Fig. 11b: Comparison of finite element methods (FEM) and proposed methods for Nominal Moment Capacity vs Glass
Design Capacity (F'g), for elastic restraint to compression edge, Gint = 0.97 MPa; kx = 0.45 N/mm/mm.

28 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Fig. 11c: Comparison of finite element methods (FEM) and proposed methods for Nominal Moment Capacity vs Glass
Design Capacity (F'g), for elastic restraint to tension edge, Gint = 0.97 MPa; kx = 0.45 N/mm/mm.

Fig. 11d: Comparison of finite element methods (FEM) and proposed methods for Nominal Moment Capacity vs Glass
Design Capacity (F'g), for elastic restraint to compression edge, Gint = 140 MPa; kx = 0.45 N/mm/mm.

In figure 6b we see that because the seed imperfection poor, with widest imperfection at the tension
edge and has a single half wave (nR = 1) whereas the critical elastic restrained mode has five half waves
(nR = 5), there is a ‘snap’ from amplification of the initial imperfection to critical mode. Discrepancies
between the theory and modelling in figures 6c and 6d also reflect the limitations of simple seed
imperfections with lack of agreement to the critical case. See also section 6.
It is suggested for design to use a reduced elastic critical buckling moment in the Kala equation, M*cr,
where M*cr = 0.85 Mcr.
Using M*cr with a 15% reduction also means that the Kala equation asymptotes to a capacity 15% below
the elastic critical moment, hence no further reduction is required when design is performed to an
LRFD (Load and Resistance Factor Design, i.e. limit state) level of load and stress capacity.
For LRFD strength capacity consider referencing EN 16612 or CEN/TS-19100. Note that the stress
values in the appendix of ASTM E1300 are for statistically acceptable use in windows with limited
consequence in the event of failure and are not indexed to an appropriate level of reliability for
structural applications. Design to ASD (Allowable Stress Design) would require additional checks for
buckling limits and design factor relative to wind load variation; alternatively the formula above can
be used with Mallow,ASD = Mn / 1.6. (or appropriate weighted load factor).

29 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

7. Numerical Validation
In the comparisons below, it can be seen that the Kala equation (46) follows benchmark finite element
models well for practical ranges of imperfections and moments less than 0.85 Mcr.. Above 0.85 Mcr
the finite element models will capture post-buckling behavior, whereas the Kala equation will
asymptote to Mcr. Thus, the Kala equation has good accuracy at typical design levels of moment and is
typically conservative where it deviates from actual.
To verify the accuracy of the analytical approach, a set of numerical simulations was carried out in
ABAQUS software, by taking into account various geometrical and mechanical configurations of
technical interest. The reference finite element (FE) model was described in ABAQUS according to
Bedon et al. (2015) that is in the form of multilayer composite shell elements for the laminated glass
section, with a set of springs to reproduce the adhesive joint. For the purpose of present comparison,
both glass and interlayer materials were mechanically modelled as linear elastic. For a given
geometrical and mechanical configuration, the typical analysis consisted in a nonlinear simulation with
imposed initial imperfection and monotonically increasing bending moment for each beam. To account
for the initial imperfection, the corresponding shape was Imported from a preliminary buckling analysis
(Bedon et al., 2015). The subsequent nonlinear incremental analysis was thus carried out by monitoring
the evolution of tensile stress peaks in glass, in order to capture the expected failure bending moment.
Note that the initial imperfection shape was selected based on the critical buckling moment and on
the corresponding number of half sine waves for each examined configuration.
Typical comparisons were carried out by monitoring the load-stress trend for selected configurations,
as well as the failure bending moment for a given tensile strength. An example can be seen in Figure
7, where failure bending moments are compared for a multitude of input configurations in terms of
analytical of FE numerical estimates for the selected laminated glass section (with 8 mm the thickness
of three glass layers, 1.52 mm the thickness of interposed bonding films, with 0.3 m the height and
1/300 the initial imperfection). Different colors denote a variation in the stiffness of interlayers (Gint=
3 MPa or 7 MPa respectively).
In general, as in Figure 7, the percentage scatter of analytical and numerical predictions was measured
in an average of less than 5%, and up to 8-9% for some cases. The comparative analysis showed that
the softer is the interlayer and the higher is the calculated percentage scatter. In any case, the accuracy
and robustness of the proposed analytical approach was emphasized by the measured limited scatter.

Fig. 12: Comparison of analytical and numerical failure bending moments for a selection of geometrical and mechanical
configurations. In evidence, the effect of interlayer stiffness (with 1/300 the amplitude of initial imperfection).

30 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 The effect of initial imperfection, as previously discussed, has a critical role in load-bearing
performance and bending capacity assessment. In this regard, Figure 8 shows that the selected
numerical configurations – when affected by critical imperfection shape with a variable initial
amplitude – are generally in good agreement with the corresponding analytical estimates. Especially
for small initial imperfections (e.g., less than 1/300), the comparative results in Figure 8 have an
associated scatter of less than 1%.

Fig. 13: Comparison of analytical and numerical failure bending moments for a selection of geometrical and mechanical
configurations. In evidence, the effect of maximum amplitude for the initial imperfection (with G= 7 MPa for the interlayer).

Finally, in accordance with Bedon et al., 2015, the shape of initial imperfection should be always
associated to the critical number of sine waves to calculate conservatively the bending capacity of a
given laminated glass beam. The use of first deformed shape (nR=1) for all configurations, would result
in minimum 10% of overestimation of bending capacity relative to a matching seed imperfection or
analytical approach (see for example Figures 6b, 6c, 6d and 9).

Fig. 14: Comparison of analytical and numerical failure bending moments for a selection of geometrical and mechanical
configurations. In evidence, the effect of initial imperfection shape and sine waves nR (with G= 7 MPa for the interlayer and
1/300 the amplitude of initial imperfection).

31 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

8. Summary
The stability equations in AS1288 for continuous restraint have been found to be non-conservative for
some circumstances such as continuous elastic restraint by structural silicone. Glass is brittle and
laminated glass is a popular method to add robustness and control the behaviour of glass beams if
they are damaged by non-design loads. The Wolfel-Bennison effective thickness model in ASTM E-1300
is non-conservative for torsional stiffness (if the beam length is used for ‘a’) and hence is non-
conservative for stability calculations of laminated glass fins. As imperfect beams approach the critical
elastic buckling moment, secondary stresses may develop that are significant relative to the tensile
capacity assumed for the glass. AS1288 has been used successfully for monolithic fins because the
strength model for structural fins in that standard is more conservative than others; the conservative
strength model makes allowances for the other inaccuracies.
The method presented includes treatment of glass fins that are monolithic or laminated, unrestrained
or with continuous elastic lateral restraint, taking into account the effect of imperfections in a system
and the increase in stress they cause as loads approach the elastic critical buckling moment.
The Enhanced Effective Thickness method by Galuppi et al. has been presented with an alternate
formulation that unifies the form of the equation for each of the target stiffnesses being considered,
introducing a new term, and allowing the rest of the equation to be standardized. This also allows
unification of the 2-ply, 3-ply and N-ply formulations previously presented by Galuppi into a single
equation.
The formulas for elastic critical buckling moment for beams with continuous elastic restraint on the
compression and tension flange by Bedon et al. have been unified into a single equation. The critical
load profile for beams with non-uniform moment has been identified as the moment on the half
wavelength rather than the overall beam.
When calculating the effective section stiffnesses for laminated beams, the characteristic length
parameter for buckling is the distance between the points of contra-flexure, i.e. the half-wavelength,
not the overall beam dimensions.
The Kala formula for imperfect beams has been introduced as a tool for calculating the moment
capacity of glass beams given a target stress, critical elastic buckling moment, and initial imperfection.
The method of analysis has been found to be favourably accurate for the practical design ranges of
moment, less than 0.85 Mcr. Above this moment the comparison with finite element analysis is highly
subject to the shape of the initial imperfection modelled, which in reality will also be unknown; thus,
a reduction factor on Mcr is proposed to allow for practical design.
The method increases the accuracy of fin and beam moment capacity, taking into account lamination,
continuous elastic lateral restraints, stress capacity, and initial imperfections.

Declaration
On behalf of all authors, the corresponding author states that there is no conflict of interest. No
funding was received to assist with the preparation of this manuscript.

32 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 References
ANSI/AISC 360 Specification for Structural Steel Buildings. American Institute of Steel Construction
ASTM E1300 (2016)- Standard Practice for Determining Load Resistance of Glass in Buildings, ASTM International, West
Conshohocken, PA, USA
Australian Standard AS1288 (2021)Glass in Buildings – Selection and Installation, Standards Australia, GPO Box 476 Sydney,
NSW 2001, Australia
Australian Standard AS4100 Steel Structures, Standards Australia, GPO Box 476 Sydney, NSW 2001, Australia
Bedon, Chiara, Belis, Jan, and Amadio, Claudio (2015). Structural assessment and lateral-torsional buckling design of glass
beams restrained by continuous sealant joints. Engineering Structures, 102 (11): 214-229.
Bedon, Chiara (2021). Lateral-torsional buckling (LTB) method for the design of glass fins with continuous lateral restraints at
the tensioned edge. Composite Structures, 266: 113790.
CEN/TS 19100-1 Design of glass structures–- Part 1: Basis of design and materials
D’Ambrosio, Gianmaria, and Galuppi, Laura (2020). Enhanced effective thickness model for buckling of LG beams with
different boundary conditions. Glass Structures & Engineering, 5(2): 205-210
Dumont C,·(1937) NASA N.A.C.A. Technical Note No 601 The Lateral Instability of Deep Rectangular Beams
https://ntrs.nasa.gov/citations/19930081361
EN 16612: Glass in building. Determination of the lateral load resistance of glass panes by calculation
EN 1993-1-1: Eurocode 3: Design of steel structures–- Part 1-1: General rules and rules for buildings
Galuppi, Laura, Manara, Giampiero, and Royer-Carfagni, Gianni (2013). Practical expressions for the design of laminated glass.
Composites: Part B, 45: 1677–1688
Galuppi, Laura, and Royer-Carfagni, Gianni (2014). Enhanced Effective Thickness of multi-layered laminated glass,
Composites: Part B, 64: 202–213
Galuppi, Laura, and Royer-Carfagni, Gianni (2020). Enhanced Effective Thickness for laminated glass beams and plates under
torsion. Engineering Structures, 206: 110077.
Green, R., Bedon, C., Galuppi, L., (2023) Design and stability of laminated glass fins with continuous lateral silicone restraint
2023 Glass Structures and Engineering 8(3):1-20DOI: http://dx.doi.org/10.1007/s40940-023-00224-1
Kala, Zdeněk (2013). Elastic Lateral-Torsional Buckling of Simply Supported Hot-Rolled Steel I-Beams with Random
Imperfections. Procedia Engineering 57: 504-514.
Kraus, Crisan and Wittor (2021) Stability Study of Cantilever-Beams – Numerical Analysis and Analytical Calculation (LTB).
CE/Papers https://d-nb.info/125069664X/34
NCCI SN-003-EN-EUNCCI: Elastic critical moment for lateral torsional buckling, by Access Steel.
https://eurocodes.jrc.ec.europa.eu/doc/WS2008/SN003a-EN-EU.pdf
Nethercott D.A., Rockey K.C.(1971), “A unified approach to the elastic lateral buckling of beams”, The Structural Engineer,
49:321–30
Nethercott, D.A., and Trahair, N.S. (1976). "Inelastic lateral buckling of determinate beams",. Journal of the Structural Division,
ASCE, 102 (ST4)
Nizich, Adam J, La GReca, Andrea M and Galuppi, Laura. Advances in Effective Thickness For Laminated Glass Structural Design,
Facade Tectonics Institute's biennial World Congress, 2022
Simulia (2022), ABAQUS computer software, v.6.14, Providence, RI, USA
SN006a-EN-EU NCCI: Elastic critical moment of cantilevers. Access Steel
Strand7 finite element software, R3.1.1, Sydney, NSW, Australia. https://www.strand7.com/
Timoshenko and Gere Theory of Elastic Stability (1956)

33 / 33 Article 10.47982/cgc.9.599 Challenging Glass Conference Proceedings – Volume 9 – June 2024

 Challenging Glass Conference Proceedings – Volume 9 – June 2024 – Louter, Bos & Belis (Eds.)
International Conference on the Architectural and Structural Application of Glass
Challenging Glass Conference 9 – 19 & 20 June 2024 – TU Delft – The Netherlands
9

Platinum Sponsor

Gold Sponsors

Silver Sponsors

Organising Partners