

 

1 Structural assessment and lateral-torsional buckling design of glass beams

2 restrained by continuous sealant joints

3 Chiara Bedon1*, Jan Belis2, Claudio Amadio3

5 Abstract

6 Glass is largely used in practice as a structural material, e.g. as beam and plate elements able to carry loads.

7 Their structural interaction is often provided by mechanical connections, although recent trends are moving

8 towards the minimization of metal components and the primary involvement of adhesives or silicone

9 structural joints working as partially rigid continuous restraints.

10 In this work, the lateral-torsional buckling (LTB) behavior of glass beams laterally restrained by continuous

11 silicone joints is assessed. Based on earlier contributions of literature and extended parametric Finite-

12 Element (FE) numerical investigations, closed-form solutions are suggested for the estimation of their

13

14 incremental nonlinear analyses, their global LTB response is also investigated, to assess their sensitivity to

15 initial geometrical imperfections as well as their prevalent LTB failure mechanism. In conclusion, a

16 generalized buckling design curve able to account for the structural contribution provided by structural

17 silicone joints is proposed for a rational and conservative LTB verification.

18

19 Keywords: lateral torsional buckling (LTB); glass beams; analytical models; finite-element modeling;

20 structural silicone joints; composite sections; incremental buckling analysis; imperfections; buckling design

21 methods; buckling curve.

22

23

1
University of Trieste, Department of Engineering and Architecture, Piazzale Europa 1, 34127 Trieste, Italy. (*)
Corresponding author. Email: [email protected].
2
Ghent University, Laboratory for research on structural models LMO, Technologiepark-Zwijnaarde 904, B-9052
Ghent, Belgium.
3
University of Trieste, Department of Engineering and Architecture, Piazzale Europa 1, 34127 Trieste, Italy.
1
 1 1. Introduction

2 Glass is largely used in practice as a structural material, e.g. as beam or plate elements able to carry the loads

3 deriving from other structural components or external forces. Especially in the field of façades and building

4 envelopes, the use of glass panels combined with steel frames, aluminum supporting bracings or cable-nets

5 resulted in an extremely wide variety of case studies. Typical applications of glass assemblies are often

6 derived and properly modified, to account for the tensile brittle behavior of glass from practice of

7 traditional construction materials (e.g. steel structures, sandwich structures, etc.), and take the form of

8 properly designed and well-calibrated mechanical or chemical connections (e.g. steel fasteners, silicone

9 sealant joints, adhesives, etc.) able to offer a certain structural interaction among multiple glass components.

10 Recent design trends, however, are often oriented towards the minimization of metal joints and mechanical

11 connectors. Typical examples consist in fact in frameless glazing systems, in which glass to glass interaction

12 is provided by sealant joints or adhesives only (Fig.1). This is the case of beam-like glass elements used in

13 practice as stiffeners for façade or roof plates, where the coupling between them is often provided by

14 continuous silicone joints. From a structural point of view, the effect of silicone joints can be compared to a

15 partially rigid shear connection, of which the effectiveness should be properly taken into account.

16 In [5], for example, results of a recent research study carried out on laterally restrained (LR) glass beams in

17 LTB have been presented. Assessment of existing analytical models available in literature for the prediction

18 LR beams under constant bending moment, accounting for the

19 shear stiffness provided by continuous, partially rigid lateral restraints, has been presented. Based on

20 extended finite-element (FE) linear bifurcation analyses (lba), the effects of various loading conditions of

21 practical interest (e.g. mid-span concentrated F or uniformly distributed loads q, applied both at the top or

22 bottom edge of the laterally restrained beams) have then been also emphasized. The final result consisted in

23

24 in LTB, thus in correction factors numerically calibrated to properly take into account the effects of silicone

25 joints, beam-to-joint stiffness ratios, loading condition.

26 In this work, based on earlier contributions [5, 6, 7], the LTB response of glass beams laterally restrained by

27 means of structural silicone joints is further assessed by means of incremental buckling, Finite-Element (FE)

2
 1 numerical simulations. While several studies have been dedicated to the assessment of the LTB structural

2 behaviour of laterally unrestrained (LU) glass beams, and buckling design methods have also been proposed

3 (e.g. [8-11]), the extension of the same simplified analytical methods must be checked. At the same time,

4 although practical formulations are provided in [5]

5 glass beams under various loading conditions, it is well known that a proper assessment of the buckling

6 resistance of a given structural system should be carried out by means of more refined analyses able to

7 account for several mechanical and geometrical aspects.

8 For this purpose, small specimens of silicone joints are subjected to shear experiments, to properly estimate

9 their elastic stiffness, but also the ultimate resistance and deformation capacities, with respect to common

10 applications of practice. Incremental nonlinear analyses are then performed on a wide set of geometrical

11 configurations. The main advantage of these FE-investigations is given by the appropriate description of

12 effects deriving from initial curvatures (with specific shape and amplitude), as well as the detection of the

13 ultimate condition as the first attainment of tensile cracking in glass or failure of the silicone joint,

14 respectively. Based on extended parametric studies, a practical design method based on a suitable design

15 buckling curve is also proposed for the LTB verification of LR glass beams.

16

17 2. laterally restrained glass beams in LTB

18 2.1.Literature background for LR members in LTB

19 LTB of structural beams with lateral restraints has been widely investigated and assessed in the last years. In

20 [12] and [13], research studies have been dedicated to the typical LTB response of doubly-symmetric steel I-

21 beams, with careful attention for possible distortional buckling phenomena in the steel webs. Khelil & Larue

22 proposed in [14, 15] a simple analytical model for the assessment of the critical buckling moment in steel-I

23 sections with LR tensioned flanges, highlighting that the presence of rigid continuous lateral restraints in

24 steel I-beams under LTB can have a s critical buckling

25 moment. The same authors presented in [16] a further alternative, analytical approach for the LTB

26 assessment of I-beams continuously restrained along a flange by accounting for the buckling resistance of an

27 ile. The latter approach, due to its basic assumptions, typically consisted in a

3
 1 conservative analytical prediction for the LTB resistance of rigidly LR steel I-beams. Conversely, the main

2 advantage of this method consisted in the implementation of Appendix values of practical use for designers.

3 The LTB behaviour of thin-walled cold-formed steel channel members partially restrained by steel sheeting

4 has been assessed, under various boundary conditions, by Chu et al. [17], by means of an energy-based

5 analytical model. Bruins [18] numerically investigated the LTB response of steel I-section profiles under

6 various loading conditions (e.g. distributed load q, mid-span concentrated force F, constant bending moment

7 My) and laterally restrained by single, elastic, discrete connectors, highlighting through parametric FE-

8 numerical studies and earlier experiments that partial elastic restraints can have significant influence on the

9 overall LTB response. The effects deriving from initial geometrical curvatures with different shape were also

10 emphasized by means of FE simulations, while simple equations were proposed as strength design method

11 for Chu et al. [19] assessed by means of an energy-based method the

12 influence of lateral restraints on the LTB response of cold-formed steel zed-purlin beams under various

13 loading / boundary conditions, demonstrating that lateral restraints generally provide an increase of the

14 unrestrained critical load, but this improvement is largely affected by boundaries or the point of load

15 application. Further assessment of structural effects deriving from discrete rigid supports on the buckling

16 behaviour of steel beams and braced columns are discussed in [20-24].

17

18 2.2.Glass beams under constant bending moment My

19 Based on [5, 6, 7, 15], in this work the attention is focused on the global LTB behaviour and ultimate

20 buckling resistance of LR glass beams, rather than on their .

21 For this purpose, let us consider first the laterally unrestrained (LU) monolithic beam depicted in Fig. 2. The

22 beam, having a monolithic rectangular b × t cross section composed of glass (with E and G

23 modulus and shear modulus respectively), is simply supported at the ends of its buckling length L0. Fork end-

24 restraints enable the occurring of out-of-plane deflections due to the applied positive (e.g. bottom edge in

25 tension) constant bending moment My.

26 When a continuous lateral restraint is introduced along its top edge (e.g. Fig. 3a) to provide a connection

27 between the glass beam and the supported panels, the structural interaction among them can be described in

4
 1 the form of a partially rigid, continuous shear restraint with elastic stiffness per unit-of-length ky and

2 rotational stiffness per unit-of-length k (Fig. 3b).

3 From a practical point of view, as also discussed in [5], the problem described in Figs. 2-3 requires the

4 implementation of closed-form solutions of suitable use. However, the presence of continuous elastic,

5 partially rigid, lateral restraints typically leads to rather complex analytical models, often able to provide

6 closed-form solutions for simple loading / boundary conditions only, hence suggesting the implementation of

7 computationally expensive FE-models for further detailed studies. With reference to Figs. 2 and 3,

8 specifically, the elastic LTB behaviour of LR glass beams can be rationally described by means of the

9 analytical model proposed by Larue et al. in [15]. It was also shown in [5]

(E)
10 moment M cr , R can in fact be cal R

11 (with k

2 2 2 2
(E) L0 nR L0 2 L0
12 M cr , R zM k y EJ z ky GJ t z ky
M . (1)
nR L0 nR nR

13 where nR 1 is an integer representative of the number of half-sine waves able to minimize Eq.(1).

14 In the same equation, Jt is the torsional moment of inertia (Jt bt3/3 when b/t > 6); Jz = bt2/12 signifies the

15 moment of inertia about the minor z-axis; My is the applied bending moment; zM is the distance between the

16 lateral restraint and the x-axis, while v and x denote the vertical deflection of the beam (z-direction) and the

17 rotation of the cross-section about the x-axis.

18 It is thus expected, mainly based on the elastic shear stiffnesses ky provided by the adopted continuous joints,

(E)
19 that the estimated critical moment M cr , R will M cr( E ) of the same LU

20 beam geometry. In the latter case, in fact, the beam would be able to offer a maximum LTB theoretical

21 resistance equal to:

n
22 M cr( E ) M cr EJ z GJ t EJ z GJ t , (2)
L0 L0

23 with n=1 is the integer value able to minimize independently on the geometrical and mechanical properties

24 of the investigated beams the predicted critical moment M cr( E ) .

5
 1 The structural benefit provided by continuous lateral restraints, specifically, could be rationally quantified in

2 the form of a magnifying coefficient RM able to give (for 0 < ky

3 M cr( E, R) RM M cr( E ) , (3)

4 with M cr( E ) given by Eq.(2) and RM= f (ky, b, t, L0, zM, nR) accounting for the effects deriving from the joint

5 shear rigidity ky, the beam aspect ratio, glass elastic stiffnesses, as well as the position of the applied elastic

6 restraints (zM) or the number nR of half sine waves able to minimize, based on Eq.(1), the predicted critical

(E)
7 buckling moment M cr , R .

8 When zM= b/2 (Fig. 3), for example, Eq.(3) leads to:

6 bL30 k y 3b 2 L20 k y 16n R2 Ebt 3 12 L40 k y 97.4n R4 Ebt 3

9 RM , (4)
39.5 nR2 bt 3 E

10 where E= 70GPa and G 0.41E for glass [25].

(E)
11 Differing from the case of a LU beam (Eq.(2)), where the minimum critical moment M cr strictly depends

12 on n=1, it is important to notice that this is not the case of LR beams with a partially rigid connection, where

13 an integer value nR ritical one able to minimize Eq.(1), due to a combination of

14 mechanical and geometrical aspects such as the beam slenderness ratio, the joint stiffness, the beam-to-joint

15 stiffness ratio, etc. Based on the ky stiffness of the adopted lateral restraint, moreover, the critical buckling

(E)
16 moment M cr , R given by Eq.(1) for a same beam geometry can largely vary. Examples are shown in Fig.4,

17 where the amplification factor given by Eq.(4) for a monolithic glass beam (L0= 3000mm, b= 300mm, t=

18 10mm) is proposed for various shear stiffnesses ky (10-4 N/mm2 500 N/mm2). As demonstrated also in [5],

(E)
19 moment M cr , R , that is the absolute minimum RM amplification factor, is given by

20 the lower envelope of RM plots obtained by changing the number of half sine waves nR.

21 In the same figure, the typical shear stiffness contribution ky expected from silicone joints is also emphasized

22 (dotted 0.184N/mm2 ky N/mm2 per unit-of-length values correspond to the range

23 of elastic stiffnesses offered by common structural sealants available in commerce for glass [5, 27, 29, 30].

6
 1 Maximum structural benefits deriving from application of continuous silicone joints in glass beams in LTB is

2 also emphasized in Fig.5, in the form of a magnifying factor RM proposed as a function of various

3 thicknesses t for beams with b= 100mm and L0= 1000mm or L0= 5000mm, respectively. The collected RM

4 values, specifically, are derived from Eq.(4) as the minimum envelope of the analytical estimations obtained

5 for each beam geometry (t, b, L0) and joint stiffness (ky).

7 2.3.Preliminary FE-numerical validation

8 FE-models were also developed with the computer software ABAQUS/Standard [26], to verify the accuracy

9 of Eq.(5) and to investigate the global LTB behavior of glass beams laterally restrained by means of

10 continuous silicone joints. Linear bifurcation analyses (lba) were carried out on a wide range of geometrical

11 properties, with continuous lateral restraints characterized by a sufficiently extended set of shear stiffnesses

(E)
12 ky. The critical buckling moment M cr , R of each beam subjected to a constant and positive bending moment

13 My was numerically predicted and compared to the corresponding analytical estimation (Eq.(4)).

14 Accordingly, the correspondence between the numerical and analytical number of half sine-waves nR

15 associated to the lowest critical moment was checked. The typical FE-model consisted of S4R 4-node,

16 quadrilateral, stress/displacement shell elements with reduced integration and large-strain formulation (type

17 S4R of ABAQUS element library; Fig. 6). Glass was described as an isotropic, indefinitely linear elastic

18 material (E= 70GPa, = 0.23 [25]). A refined and regular mesh pattern was used, with lmesh the characteristic

19 size of quadrilateral shell elements comprised between 3mm and 15mm, depending on the b×L0 dimensions

20 of the studied beams.

21 To take into account the presence of continuous, partially rigid lateral restraints, a series of indefinitely linear

22 elastic springs directly connected to a rigid substructure and characterized each one in terms of elastic

23 stiffness Ky, was introduced along the top edge of each beam (zM= b/2, Fig. 6). The elastic stiffness Ky, being

24 dependent on lmesh, was in fact estimated as Ky= ky × lmesh (Ky= ky × 0.5 lmesh for the two springs close to the

25 beam end sections), with 10-4 N/mm2 ky 4

N/mm2 the range of elastic stiffnesses taken into account

26 throughout this exploratory parametric study. Loads and boundaries for the simply supported, fork-end

27 restrained beams in LTB were finally introduced in each FE-model by means of bending moments My
7
 1 applied at the barycentrical node of the end sections, as well as nodal translational and rotational restraints

2 for the same cross-sectional nodes. These modeling assumptions, rather in agreement with earlier studies

3 related to the LTB response of laterally restrained structural members (e.g. [19]) were then accounted for a

4 numerical and analytical study extended to a sufficiently wide rang t

5 b L0 Comparative examples are proposed in Fig.7, where the

6 rather good agreement between numerically and analytically predicted RM amplification factors is

7 emphasized for three selected geometrical configurations, by changing the joint shear stiffness ky. Extended

8 lba simulations presented in [5] highlighted that in general Eq.(4) provides fairly accurate predictions for the

9 LTB critical moment of glass beams restrained by silicone joints. Parametric calculations also resulted in an

10 average percentage discrepancy RM , with:

RM Eq .( 5)
RM ABAQUS
11 RM 100 (5)
RM ABAQUS

12 A maximum discrepancy equal to -length ky of

13 4 N/mm2 ky 136 N/mm2).

14 As expected, major effects of different joint stiffnesses ky manifested not only in a significant increase of the

(E)
15 RM coefficient for the examined beams, thus of their theoretical critical buckling moment M cr , R , but also in

16 a substantial modification of their reference buckling shape (e.g. critical nR value), thus in a variation of their

17 global LTB response. In this sense, a detailed incremental buckling investigation should be performed by

18 taking into account the actual critical buckling shape for each geometrical configuration, with u0,max= L0/400

19 the recommended amplitude [33].

20 Parametric studies highlighted an almost exact correlation between the analytical and numerical number of

21 half sine-waves nR associated to comparative data collected in Fig.7, especially in presence of lateral

22 restraints able to provide a buckling stren RM

23 nR -10 depending on the beam geometry). Some examples are proposed in Fig.8 for some

24 beam geometries and joint stiffnesses of practical interest for this study.

8
 1 A partial lack of correlation between analytical and numerical half sine-waves nR able to provide the lowest

2 critical buckling moment for the studied beams was found only in presence of stiff continuous lateral

3 restraints (10 N/mm2 ky 2

N/mm2 nR , in this investigation), not of primary interest for

4 this study.

7 2.4.Glass beams under uniformly distributed load or concentrated load at mid-span

8 While for LR glass beams under constant bending moment My it was shown that closed-form solutions of

9 literatur

10 critical load and fundamental buckling shape (e.g. nR)), the same approach cannot be directly applied to other

11 loading conditions. This is the case for example of beams in LTB under distributed loads q or mid-span

12 concentrated loads F. Although for LU beams in LTB it is reasonable to estimate the maximum effects

13 deriving from q or F loads applied to their barycentrical axis (e.g. point G of Fig.9) by accounting for an

14 equivalent bending moment

M y , max
15 My M y ,eq , (6)
k1

16 with:

17 k1 a correction factor depending on the distribution of the applied loads (Table 1), and

18 My,max the maximum bending moment due to the applied q or F loads respectively, that is

19 M y ,max qL20 8 or M y ,max FL0 4 ;

20 the same simplified method cannot be used for LR beams, due to the presence of partially rigid, continuous

21 lateral restraints able to modify the global LTB behavior of the examined beams. At the same time,

22 according to practical applications of glass beams and fins in roofs and façades, it is rationally expected an

23 eccentrical application of q and F loads (e.g. eload eload zM = +b/2 denoting top-edge loads (point A

24 of Fig.9) and eload zM = b/2 signifying bottom-edge loads (point B of Fig.9)) will further affect the

25 theoretical LTB resistance of the studied systems.

9
 1 Exploratory FE-studies confirmed, for example, that the mentioned loading conditions can strongly modify

2 nR not directly

3 comparable to the solution derived in presence of My moments.

4 In terms of magnifying factor RM obtained for same beam geometry, for example, interesting numerical

5 comparisons are collected in Fig.10 for the same beam geometry under (a) top-edge or (b) bottom edge

6 distributed q or mid-span F loads, respectively. In these plots, specifically, the RM amplification factor is

7 separately calculated for the q or F loading conditions in accordance with Eq.(5) as the ratio between the

8 numerical critical buckling moment Mcr,R (ABAQUS-lba) of each LR beam and the corresponding LU

9 critical moment Mcr (RM

(E) L0
10 M cr( E, R) F
Fcr , R ABAQUS eb
(7)
4

11 and

(E ) L20
12 M cr( E, R) q
q cr , R ABAQUS eb
. (8)
8

13 As also discussed in [5], the application of top-edge distributed loads (qA) typically results almost

14 in numerical predictions generally comparable to those deriving

15 from constant moments My. Major discrepancy was conversely found for the same beam geometries and joint

16 stiffnesses ky under the action of top-edge mid-span loads (FA). The latter condition generally resulted in non-

17 conservative predictions for the examined LR beams, independently on the load eccentricity eload from the

18

19 bottom-edge mid-span loads F, typically manifesting in limited number of half sine-waves nR (e.g. Fig.11).

20 From a practical point of view, approximate or closed-form solutions for the estimation of the LTB

21 theoretical buckling resistance of LR glass beams under various loading conditions (e.g. Eq.(6) and Table 1)

22 would certainly represent a suitable method for designers. However, the correction factor k1 mentioned in

23 Eq.(6) should be properly calculated, since strictly related to the joint stiffness ky, the number of half sine

24 waves nR associated to each loading condition, etc.

10
 1 In this context, it should be noticed that wide series of parametric numerical simulations (ABAQUS-lba)

2 discussed in [5] generally provided predictions well agreeing with plots collected in Fig.10. As a result,

3 numerically calibrated correction factors k1* associated to various beam geometries and loading conditions,

4 but well defined joint shear stiffness (0.184N/mm2 [7]), were estimated as:

* M cr( E, R) i
M cr( E, R) i
5 k
1 f (k y , eload ) , (9)
M cr( E, R) RM M cr( E )

(E)
6 where M cr , R i
oment derived from Eqs.(7)-(8), depending on the examined

(E)
7 loading condition, while M cr , R is obtained from Eq.(3) for a same beam geometry and joint stiffness, with

8 M cr( E ) and RM respectively defined in Eqs.(2)-(3) under constant bending moments My.

9 Results collected in Figs.12 and 13 for LR beams under top-edge or bottom-edge q or F loads, respectively,

10 t L0 , are proposed

11 in the form of k1* correction factors (Eq.(9)) as a function of the beams L0/b ratios. As expected, the so

12 obtained correction factors k1* manifested partial sensitivity to the beams geometrical properties and

13 torsional stiffness GJt (e.g. higher dispersion of k1* values for beams with small L0/b ratios). In any case, an

14 almost stable LTB behaviour was found throughout the parametric study, for each loading condition.

15 In the same Figures, for this reason, minimum values k1,min* are also highlighted, since an approximate but

16

17 performed as:

*
18 M cr( E, R) i
k1, min RM M cr( E ) , (10)

19 with:

20 i denoting the specific loading condition (top-edge or bottom-edge q or F loads),

21 M cr( E, R) i
given by Eqs.(7)-(8)

22 M cr( E ) and RM defined in Eqs.(2)-(3), and

*
23 k1 min i
signifying the corresponding minimum correction factor for each i loading condition

24 (Figs.12-13 or Table 2).

11
 1

2 3. Experimental calibration of the constitutive mechanical behavior of structural sealant joints

3 Based on the LTB background discussed in Section 2, further studies were focused on the mechanical

4 characterization of the constitutive behavior of structural sealant joints of common use in glass practice.

6 and (10), would require as input parameter the elastic shear stiffness ky on the adopted sealant joints, refined

7 buckling investigations should be able to properly take into account the full stress-strain behavior of joints up

8 to failure. Differing from LU glass beams, where the buckling failure mechanism could result from the

9 limited glass tensile strength only, the effective LTB resistance of LR glass beams should be in fact assessed

10 by correctly taking into account the structural contribution of silicone joints e.g. increased critical buckling

(E)
11 load M cr , R and a number of half sine-waves nR but also the possible failure mechanisms occurring in the

12 joints themselves, together with glass tensile cracking. As a result, the elastic shear stiffness ky would not

13 represent an exhaustive information for the development of more refined studies.

14

15 3.1.Specimens and test methods

16 Shear tests were performed at Ghent University [6, 7] and ten experiments, equally divided in two series (A

17 and B), were carried out on small specimens of silicone sealant joints. The chosen material was Dow

18 Corning® 895 (DC 895) [27], a one-component sealant used in practice for glass structures. Based on the

19 producer recommendations, structural sealant joints made with DC 895 should have a width wsil (e.g. the

20 joint dimension in the direction of shear loading) determined by structural and thermal calculations, but in

21 any case comprised between wsil,min= 6mm and wsil,max= 15mm.

22 Based on these assumptions, each specimen well representative of a small part of a continuous elastic

23 sealant joint - consisted of a sample with total length lsil= 100mm. The difference between specimens of

24 series A and B was then given by the wsil × hsil cross-sectional dimensions of each joint, respectively equal to

25 wsil= hsil= 6mm (series A) and wsil= hsil= 15mm (series B). An appropriate steel device was used to position

26 the specimens and connect them to the loading machine. Prior to the execution of experiments, proper and

12
 1 complete curing of the silicone was ensured systematically by means of extra specimens, prepared for this

2 purpose together with the shear test specimens.

3 The typical test setup for shear experiments is depicted in Fig.14. Displacement-controlled tests were carried

4 out at 23° C, and a constant speed deformation of 5mm/min, as recommended by ETAG 002 [26], was

5 ensured during the experiments. While the top edge of the metal device was kept fixed and rigidly connected

6 to the loading machine, the bottom end of the steel device was subjected to a linearly increasing vertical load

7 F, up to failure of the specimens. The corresponding displacement u was continuously monitored during each

8 test.

10 3.2.Test results

11 Experiments generally resulted in a uniform response of specimens and in a similar pre-destructive behavior,

12 for both the series of specimens. Exemplificative images showing the progressive damage and failure of a

13 specimen are proposed in Fig.15. Based on the obtained test results, it was also found that in general the

14 elastic load F-displacement u response of specimens under shear loads can be rationally described in the

15 form of a straight line (Fig.16).

16 Based on test measurements, the average elastic stiffness per unit-of-length ky was equal to 0.184N/mm2, as

17 also highlighted in [5]. The tested specimens also highlighted an almost stable behavior, attaining large

18 displacements before failure, with an ultimate elongation u du-wsil)/wsil

19 series A and B respectively, with du denoting the maximum attained displacement. The obtained average

20 ultimate elongation u,avg

21 sealants available in commerce, being their maximum elongation at failure typically comprised between

22 u,nom u,nom 600% [27, 29, 30]. In terms of ultimate shear/tensile stress u,avg, this parameter was

23 derived from experimental measurements as the average ratio between the failure load Fu of each series of

24 specimens and the corresponding resisting area Asil, hence resulting in u,avg= 0.94 N/mm2. Also in this latter

25 case, the calculated strength was in rather good agreement with the nominal ultimate tensile resistances of

26 common structural sealants ( u,nom= 1.06 N/mm2 [29] and u,nom= 1.2 N/mm2 [30]).

27
13
 1 4. Refined FE-parametric numerical study: incremental nonlinear analyses

2 4.1.General solving approach

3 In accordance with the mechanical properties of silicone specimens derived from experiments discussed in

4 Section 3, further extended parametric investigations were carried out by means of additional FE-models

5 derived from Section 2.

6 As known, lba simulations provide rather poor information, compared to static incremental nonlinear (inl)

7 analyses, since they only provide the theoretical buckling resistance of a given structural system. A proper

8 estimation of the effective LTB strength of the same beams discussed in Section 2, should in fact properly

9 take into account that the expected failure load M cr( E, R) could be strongly affected by several aspects, such as

10 the effects of possible initial geometrical imperfections (with specific shape and amplitude), the premature

11 tensile cracking of glass as well as possible failure mechanisms in the silicone joints, if subjected to large

12 strains and tensile stresses exceeding their ultimate values ( u,avg, u,avg). Further inl analyses were thus

13 performed (static Riks procedure) on a selected set of geometrical configurations, based on FE-models

14 derived from Section 2 (Fig.6) and properly modified.

15

16 4.2.Reference configuration and detection of the LTB failure condition

17 The axial springs representative of the sealant joints were characterized by means of the average shear

18 stiffness ky =0.184N/mm2 derived from the experiments (Section 2). The indefinitely linear elastic

19 constitutive behavior of the same axial springs was otherwise characterized in terms of ultimate elongation

20 u u,avg and ultimate tensile stress u u,avg (and u properly related, in all the FE-models, to the mesh size

21 lmesh (e.g. area of influence of each spring) adopted for each LR beam) taken from Section 3. A brittle failure

22 mechanism was also assigned to each axial spring, so that during each inl simulation, based on the assigned

23 ky, u and u mechanical parameters, the single connector could fail and be released at the first attainment of

24 the ( u, u) ultimate condition. The examined glass beams were then assumed composed of different glass

25 types (annealed (AN), heat strengthened (HS) and fully tempered (FT), respectively), characterized each one

26 by isotropic linear elastic behavior (E= 70GPa, = 0.23) and specific nominal characteristic tensile resistance

14
 1 ( Rk,AN= 45MPa, Rk,HS= 70MPa, Rk,FT= 120MPa respectively, based on [25]). Possible tensile cracking of

2 glass, in this sense, was manually checked by continuously monitoring the maximum envelope of tensile

3 stresses along the L0 × b surface of each beam. Careful attention was finally paid for possible initial

4 geometrical imperfections. Based on preliminary lba simulations carried out on the same beams, inl analyses

5 shapes having a

6 specific number of half sine-waves nR and a maximum amplitude u0,max along the beam length L0.

8 4.3.Discussion of numerical results

9 4.3.1. LR beams under constant bending moment My

10 Examples are proposed in Figs.17, 18 and 19 for the L0= 3000mm × b=300mm × t=10mm beam previously

11 investigated, subjected to positive constant bending moments My. In them, the sensitivity of numerical

12 buckling predictions to the shape (nR) and maximum amplitude u0,max of initial geometrical imperfections, as

13 well as to the effects of a possible failure in the silicone joints and a premature tensile cracking of glass

14 ( max Rk ) are highlighted.

15 In Fig.17, specifically, the maximum envelope of out-of-plane displacements umax are proposed as a function

16 of the RM amplification factor for the same LU beam or restrained by means of continuous silicone joints

17 (LR). For both the beams, the maximum amplitude of the initial geometrical imperfection is set equal to

18 u0,max= L0/400, being the corresponding lba buckling shapes obtained performed on both the FE-models (with

19 n= 1 a as conventionally done for the buckling analysis of unrestrained beams and nR= 4 for the LU and

20 LR beams respectively). As shown, preliminary neglecting possible cracking mechanisms in glass (and

21 damage in the silicone joint, when present), the LU beam would ideally carry on a maximum bending

(E)
22 moment asymptotically tending towards the theoretical critical buckling moment M cr given by Eq.(2) (e.g.

23 RM for the LU beam). The laterally restrained beam (LR), otherwise, would be able to offer a

24 significantly higher buckling resistance, e. 7 times the LU geometry, almost comparable to the

25 corresponding theoretical critical buckling moment (Fig.17, M cr( E, R) 5.70M cr( E ) ).

15
 1 However, in the same Figure it is also possible to notice that for the LR beam both possible failure

2 mechanisms occurring in glass or in the silicone joint would result in marked decrease of its ideal LTB

3 resistance. By assuming in the same beam an indefinitely linear elastic mechanical behavior for glass, for

4 example, the LTB failure mechanism would be governed by the progressive collapse of few axial connectors,

5 representative of the silicon joint, typically resulting in an ultimate failure load significantly lower than the

(E)
6 theoretical M cr , R value (point A of Fig.17, RM= 4.22). Due to the separate failure of these connectors along

7 the beam buckling length L0, in conjunction with the typical elastic LTB deformed shape of the examined LR

8 beam (e.g. nR), the post-cracked LTB response would also be characterized by an unsymmetrical deformed

9 configuration (path AB of Fig.17).

10 In Fig.18, the R stress ratio is also proposed for the silicone joint of the LR beam presented in Fig.17, where

11 R denotes the ratio between the measured stress max in each axial connector and the corresponding ultimate

12 resistance u. Results are shown, along the beam buckling length x/L0 x L0), as a function of the

13 applied bending moment (e.g. specific RM loading configuration derived from Fig.17). As shown, due to the

14 assumed geometrical configuration for the examined LR beam (nR= 4), damage in the axial connectors (e.g.

15 R = ±1) first occurs where the beam undergoes the maximum out-of-plane deflections. In the same Figure it

16 is also possible to notice that according to Fig.17 that the ultimate LTB resistance of the examined LR

17 beam would be clearly affected by the limited tensile resistance of glass. Depending on the type of glass and

18 the corresponding characteristic tensile strength Rk, the LTB collapse would occur due to premature glass

19 failure (with AN, HS, FT in Figs.17-18 denoting the attainment of the tensile resistance for AN, HS and FT

20 glass types, respectively). It is interesting to notice, in this context, that almost the same buckling collapse

21 mechanism was found for all the beam geometries taken into account in this parametric investigation, and the

22 failure of the silicone joints, accordingly, typically occurred for higher bending loads only. However, a

23 detailed LTB investigation should necessarily take into account both the possible collapse mechanisms, for a

24 given LR beam, since strictly related to several mechanical and geometrical influencing parameters.

25 Further related examples are shown for the same beam geometry in Fig.19, where again the inl numerical

26 predictions are proposed as the maximum envelope of transversal displacements (umax u0,max), as a function

16
 1 of the corresponding RM coefficient. Fig.19, specifically, emphasizes that the assumption of improper

2 buckling shapes for the description of possible initial geometrical imperfections for the examined LR beams

3 would typically result in a marked overestimation of their initial stiffness against out-of-plane deformations

4 and in an unrealistic overall LTB ultimate resistance (e.g. points B and AN, HS, FT of Fig.19). In addition

5 (buckling shape at point B of Fig.19), the inaccurate description of possible initial deviations from

6 straightness would result in the progressive but inappropriate modification of the overall deformed shape

7 and in the variation due to the increase of the applied bending moments My of the number of half-sine

8 waves nR leading the beam to failure.

9 Certainly, apart from the shape assumed for the description of initial curvatures the maximum amplitude

10 u0,max of geometrical deformations would also significantly affect the predicted buckling failure loads, as for

11 example shown in Fig.20. The difference between the proposed curves is given, for a same beam geometry,

12 joint shear stiffness ky and loading condition (My), by the amplitude u0,max

13 configuration (e.g. nR= 4). In the same Figure, the points A, B, C

14 or to the attainment of maximum

15 tensile stresses exceeding the corresponding resistance values, respectively. In this work, according to [31],

16 the reference maximum amplitude u0,max= L0/400 was kept reasonably constant throughout the parametric

17 FE-numerical study.

18

19 4.3.2. LR beams under various loading conditions

20 Application to the examined LR beams of different loading distributions (e.g. top-edge or bottom-edge q and

21 F loads) also manifested, as expected from lba exploratory investigations carried out in Section 2, in further

22 modification of the predicted overall LTB response of the same LR beams.

23 An example is shown in Fig.21 for a fixed LR geometry under bottom-edge (FB) mid-span loads. Due to the

24 specific loading case, and based on a preliminary ABAQUS-lba simulation, the inl analysis is carried out on

25 a FE-model affected by an initial curvature shape well agreeing with Fig.11, with u0,max= L0/400. Results are

26 proposed in the form of the RM amplification factor versus the maximum relative out-of-plane deflection

27 (umax u0,max). In the same Figure, numerical predictions derived from Fig.17 for the same LR beam

17
 1 geometry under constant bending moment My are also recalled, to emphasize the effect of various loading

2 distributions. As shown, the implemented load eccentricity eload (with eload= zM for the FB case), typically

3 resulted in an appreciable LTB strengthening contribution (point B of Fig.21), compared to the same beam

4 geometry under constant moment My. However, also in this case glass tensile cracking would represent the

5 first condition of failure (points AN, HS, FT of Fig.21). Major effects deriving from the specific application

6 of loads were found not only in terms of maximum RM amplification factors expected from each

7 configuration, but also in modification of the initial elastic stiffness for the investigated beams, as well as in

8 partially different distribution and increase of maximum tensile stresses and out-of-plane deformations in the

9 LR glass beams.

10

11 5. Standardized LTB design method for glass beams with partially rigid lateral restraint

12 5.1.General approach

13 According to analytical, numerical and experimental results presented in this paper, it was shown that

14 structural silicone joints can generally strongly modify the LTB response of glass beams, both in terms of

15 ling shape, as well as in terms of overall load

16 vs. out-of-plane displacement response.

17 For practical purposes, in this context, a suitable design buckling approach for LR glass beams in LTB

18 should be able to correctly take into account the beam slenderness ratio and sensitivity to initial

19 imperfections, as well as the beam-to-joint stiffness ratio and its effects on the overall buckling response. At

20 the same time, a suitable design method should be able to properly take into account in the estimation of the

21 design buckling resistance, the effects deriving from both the brittle failure of glass in tension, as well as

22 possible collapse in the structural silicone joints. In this sense, parametric studies discussed in Section 4

23 highlighted for the range of silicone joint and glass beam mechanical and geometrical properties examined

24 in this study that the LTB collapse of continuously laterally restrained beams is strictly related to tensile

25 cracking of glass, rather than to possible failure in the silicone joints. However, the effects of these structural

26 joints on the expected overall buckling response should be carefully estimated, for design purposes.

18
 1 For the verification of LU glass beams in LTB, practical design buckling curves developed in agreement

2 with the Eurocode standard for steel structures [32] were for example proposed in [33]. For a given glass

3 beam geometry (L0, b, t), glass type (e.g. Rk) with continuous silicone joints (zM= b/2, ky) a suitable design

4 procedure could require for example the satisfaction of the condition:

5 M Ed M b , Rd , (11)

6 with M Ed M y signifying the design equivalent bending moment and Mb,Rd denoting the design buckling

7 resistance. In [33], for example, it was proposed to calculate Mb,Rd as:

8 M b, Rd LT M Rd , (12)

9 with

10 M Rd Wy Rd (13)

11 Wy b2t / 6 (14)

12 the elastic resistant modulus; Rd= Rk/ M the design tensile resistance of glass and M a partial safety factor

13 (the value M= 1.4 was suggested in [33]).

14 In Eq.(12), moreover:

1
15 LT , for LT 1 (15)
2 2
LT LT LT

16 is a buckling reduction factor for beams in LTB, where:

2
17 LT 0.5 1 imp LT 0 LT (16)

18 and

19 LT Wy Rk M cr(E ) (17)

20 The coefficients imp and 0 mentioned in Eq.(16) are representative of possible initial imperfections. Based

21 on the assessment to experimental LTB results available in literature as well as on FE-numerical

22 investigations, the values imp= 0.45 and 0= 0.20 were proposed in [33] for the verification of LU glass

19
 1 beams affected by initial sine-shaped imperfections with maximum amplitude u0,max= L0/400 and subjected to

2 constant bending moments My, mid-span concentrated loads F or uniformly distributed loads q.

3 In Eq.(17), moreover, Mcr(E) given by Eq.(3) for LU beams.

5 5.2.LTB design procedure for LR glass beams under constant bending moment My

6 In this work, to assess the possible extension of the LTB design curve proposed in [33] for LU beams, further

7 extended FE-investigations were carried out, as partly discussed in Section 4.

8 Monolithic glass beams with various geometrical properties (buckling length L0= 1000mm-5000mm, with

9 step increment of 500mm between each series of beams; height b= 100m, 200mm, 300mm; nominal

10 thickness t= 6mm, 8mm, 10mm, 12mm, 15mm, 19mm) and glass types (AN, HS, FT) opportunely

11 combined with each other were analyzed. Each beam, subjected to a constant, positive bending moment

12 MEd My, was assumed affected by an initial geometrical imperfection of maximum amplitude u0,max= L0/400,

13 obtained as the scaled critical buckling shape for each configuration (ABAQUS-lba). Continuous silicone

14 joints were also characterized as discussed in Section 3, and shear stiffness values ky were assumed

15 comprised between ky= 0.184 N/mm2 (average experimental value, Section 3) and ky= 0.6136 N/mm2

16 (maximum nominal value derived from [29, 30]).

17 As partially discussed in Section 4, for all the examined beams the failure condition was identified as the

18 lower condition due to (i) glass tensile failure and (ii) collapse of the silicone joint, thus the failure bending

19 moments Mu* where separately collected for each FE-simulation. Numerical predictions are collected in

* *
20 Fig.22, where normalized FE-results are expressed in the form ( LT , LT ) and compared to the design LTB

21 curve proposed in [33] ( imp= 0.45, 0= 0.20), being:

* M u*
22 LT , (18)
W y Rk

* (E)
23 LT Wy Rk M cr , R , (19)

(E)
24 with M cr , R the minimum critical buckling moment obtained from Eq.(1).

20
 1 As shown, an interesting agreement between the LTB design curve calibrated in [33] and the ultimate

2 buckling resistance of laterally restrained beams was found. Compared to LU beams, the primary effect of

3 additional continuous lateral restraints, due to increased stiffness and overall resistance, typically results in a

4 decrease of the normalized slenderness ratio LT (Eq.(19)) and an increase of the maximum load carrying

5 capacity, hence providing an increase of the buckling coefficient LT . Numerical predictions collected in

6 Fig.22, specifically, have limited slenderness ratios LT LT

7 joints. The same LU beam geometries, otherwise, would be characterized by higher slenderness values,

8 typically up to LT -2.5. Maximum structural benefits deriving from continuous silicone joints, in this

9 context, were found for beams with small b/L0 ratios. However, appreciable structural efficiency was

10 generally obtained for all the examined beams, as for example shown in Fig.23 (L0= 3000mm, b= 300mm, t=

11 10mm; AN glass; ky= 0.184 N/mm2). In conclusion, based on extended assessment and validation of methods

12 discussed in this paper, it is expected that the LTB verification of glass beams subjected to constant bending

13 moments My and laterally restrained by means of continuous structural silicone joints could be performed by

14 means of Eq.(11), with LT given by Eq.(15) and LT given by Eq.(19).

15

16 5.3.LTB design procedure for LR glass beams under uniformly distributed q or mid-span F loads

17 In agreement with the extended studies partly discussed in previous Sections for the LTB response of LR

18 glass beams under uniformly distributed loads q or mid-span concentrated loads F, it is expected that the

19 same design buckling curve presented in Section 5.2 for beams under constant bending moments My could

20 provide rational predictions.

21 Examples are shown in Fig.24 for a beam under top-edge mid-span loads FA (L0= 3000mm, b= 300mm, t=

22 10mm; AN glass; ky= 0.184 N/mm2). Data are collected in the form ( LT , LT ) for the same geometry

23 laterally unrestrained (LU) or restrained by means of a continuous silicone joint (LR). In the latter case, the

24 non-dimensional parameters ( LT , LT ) are calculated according to Eqs.(18)-(19), based on ABAQUS-lba

25 -inl (e.g. equivalent ultimate bending moment Mu)

21
 1 predictions. From a practical point of view, based on the proposed comparisons, it is clear that the presented

2 LTB design buckling curve could be used for the verification of LR beams under various loading conditions.

3 The LTB verification, specifically, could be in fact carried out by means of Eq.(11), by simply calculating

4 the corresponding slenderness ratio LT

(E)
5 M cr , R obtained from Eq.(10).

6 In Fig.24, specifically, it can be seen both for a LU and LR beam geometry that a full FE-numerical

7 calculation of the corresponding ( LT , LT ) parameters would provide close agreement with the

8 corresponding analytical predictions. The LU and LR dots, specifically, are in fact calculated in accordance

9 with Eqs.(18) and (19), that is by means of preliminary buckling analysis (for the estimation of the

(E)
10 M cr , R of Eq.(19)) and successive inl-analyses (for the calculation of

11 the equivalent failure moment Mu*, as specified in Eq.(18)). The LU and LR dots are compared in Fig.24

12 with LU* and LR* points, respectively. In the case of the unrestrained beam (LU *), the corresponding ( LT ,

(E)
13 LT ) parameters are calculated by means of Eqs.(15) and (19), with M cr , R derived from Eq.(10), by

14 assuming RM= 1 and k1= 1.32 (Table 1). For the LR* point, conversely, the buckling coefficient LT is given

15 again by Eq.(15), while the corresponding normalized slenderness ratio LT (Eq.(19)) is analytically

(E )
16 M cr , R provided by Eqs.(10), with

17 k1,min*= 0.75 (Table 2) and RM derived from Eq.(4).

18 As shown, the proposed method and assumptions for LR beams under various loading conditions (e.g.

19 correction factors listed in Table 2), would result in a conservative but rather acceptable prediction for the

20 examined geometry, compared to a full FE-numerical calculation, hence confirming the general validity of

21 the proposed approach.

22

23

24

22
 1 6. Summary and conclusions

2 In this paper, results of a recent research activity on the LTB behaviour of beam-like glass elements with

3 continuous lateral restraints have been discussed. Depending on combinations of silicone joint stiffnesses

4 and beam geometrical properties, analytical and FE-numerical calculations highlighted that their critical

5 buckling moment can be strongly increased, especially in the case of slender beams.

6 ent yields poor information only on the actual LTB ultimate resistance of

7 laterally restrained (LR) glass beams, and detailed incremental analyses should be generally carried out to

8 properly assess the effects of multiple mechanical and geometrical aspects (e.g. beam-to-joint stiffness ratio,

9 failure mechanisms in glass or joints, initial geometrical imperfections, loading condition, etc.).

10 For this purpose, refined incremental nonlinear FE-analyses were performed on a large number of glass

11 beams, to properly assess their global LTB response up to failure. Specifically, parametric analyses were

12 carried out to take into account the effects of possible geometrical imperfections both in terms of maximum

13 amplitude and reference shape as well as the premature buckling failure deriving from glass cracking in

14 tension, or the occurring of possible failure mechanisms in the structural silicone joints. Simulations

15 generally confirmed the appreciable efficiency of structural silicone joints, compared to laterally unrestrained

16 (LU) beams, and highlighted although in presence of continuous lateral supports that their ultimate

17 buckling resistance is strictly related to failure of glass in tension. Further studies were then extended to LR

18 glass beams in LTB, by taking into account the effects of various loading conditions (e.g. top-edge or

19 bottom-edge uniformly distributed q or mid-span concentrated loads F). Based on earlier contributions of

20 literature, finally, a design LTB curve recently calibrated for the verification of LU glass beams has been

21 recalled and applied to LR beams.

22 As shown, a rather good agreement was found, hence suggesting its possible extension for the design and

23 verification of the studied structural typology. Certainly, further improvements of the current method could

24 be derived from experimental and refined FE-investigations able to account for the structural interaction

25 between the studied glass beams and the supported glass roof/plates. However, it is expected that discussed

26 comparisons and methods could provide useful background and suitable tools for practical applications.

27

23
 1 References

2 [1] www.dupont.com

3 [2] http://www.spgintercitynews.ch/rachat-du-batiment-merck-serono-une-belle-victoire-pour-spg-intercity/

4 [3] http://natralight.co.uk

5 [4] http://malishevwilson.com

6 [5] Bedon C, Amadio C (2015). Analytical and numerical assessment of the strengthening effect of

7 structural sealant joints for the prediction of the LTB critical moment in laterally restrained glass beams.

8 Materials & Structures, under review.

9 [6] Verhoeven E (2008). Effect van constructieve kitvoegen op de stabiliteit van glazen liggers (in Dutch).

10 Master thesis, LMO Laboratory for Research on Structural Models, Ghent University.

11 [7] Belis J, Bedon C (2014). Strengthening effect of structural sealants on the LTB behaviour of glass

12 beams. Proceedings of Challenging Glass 4 & COST Action TU0905 Final Conference, EPFL Lausanne,

13 6&7 February.

14 [8] Belis J, Van Impe R, Lagae G, Vanlaere W (2003). Enhancement of the buckling strength of glass beams

15 by means of lateral restraints. Structural Engineering and Mechanics, 15(5): 495-511.

16 [9] Luible A (2004). Stabilität von Tragelementen aus Glas. Dissertation, EPFL Lausanne, Thése 3014.

17 [10] Belis J (2005). Kipsterkte van monolithische en gelamineerde glazen liggers. Ghent: Ghent

18 University.

19 [11] Belis J, Bedon C, Louter C, Amadio C, Van Impe R (2013). Experimental and analytical assessment

20 of lateral torsional buckling of laminated glass beams. Engineering Structures, 51: 295-305.

21 [12] Kalkan I, Buyukkaragoz A (2012). A numerical and analytical study on distortional buckling of

22 doubly-symmetric steel I-beams. Journal of Constructional Steel Research 70: 289-297.

23 [13] Vrcelj Z, Bradford MA (2006). Elastic distortional buckling of continuously restrained I-section

24 beam-columns. Journal of Constructional Steel Research 62: 223-230.

25 [14] Khelil A, Larue B (2008). Simple solutions for the flexural-torsional buckling of laterally restrained

26 I-beams. Engineering Structures 30: 2923-2934.

24
 1 [15] Larue B, Khelil A, Gueury M (2006). Elastic flexural-torsional buckling of steel beams with rigid

2 and continuous lateral restraints. Journal of Constructional Steel Research 63: 692-708.

3 [16] Larue B, Khelil A, Gueury M (2007). Evaluation of the lateral-torsional buckling of an I beam

4 section continuously restrained along a flange by studying the buckling of an isolated equivalent profile.

5 Thin-Walled Structures 45: 77-95.

6 [17] Chu X-T, Kettle R, Li L-Y (2004). Lateral-torsional buckling analysis of partial-laterally restrained

7 thin-walled channel-section beams. Journal of Constructional Steel Research 60: 1159-1175.

8 [18] Bruins RHJ (2007). Lateral-torsional buckling of laterally restrained steel beams. Report nr A-2007-

9 7, Technische Universiteit Eindhoven.

10 [19] Chu X-T, Rickard J, Li L-Y (2005). Influence of lateral restraints on lateral-torsional buckling of

11 cold-formed steel purlins. Thin-Walled Structures 43: 800-810.

12 [20] Taras A, Greiner R (2008). Torsional and flexural torsional buckling A study on laterally restrained

13 I-sections. Journal of Constructional Steel Research 64: 725-731.

14 [21] Bradford MA (2000). Strength of compact steelbeams with partial restraint. Journal of

15 Constructional Steel Research 53: 183-200.

16 [22] Bradford MA (1998). Inelastic buckling of I-beams with continuous elastic tension flange restraint.

17 Journal of Constructional Steel Research 48: 63-77.

18 [23] Zhang L, Tong GS (2011). Lateral buckling of eccentrically braced RHS columns. Thin-Walled

19 Structures 49: 1452-1459.

20 [24] Nguyen CT, Joo HS, Moon J, Lee HE (2012). Flexural-torsional buckling strenght of I-girders with

21 discrete torsional braces under various loading conditions. Engineering Structures 36: 337-350.

22 [25] EN 572-2:2004. Glass in buildings Basic soda lime silicate glass products. CEN.

23 [26] Simulia (2009), ABAQUS v.6.9 Computer Software and Online Documentation, Dassault Systèmes.

24 [27] Dow Corning (2011). Product Information Dow Corning 895, Structural Glazing Sealant, one-part

25 silicone rubber.

26 [28] EOTA (1999). ETAG 002, Guideline for European technical approval for structural sealant glazing

27 systems (SSGS) part 1: Supported and unsupported systems.

25
1 [29] Henkel (2012). Pattex SL 690 Solyplast - Structural Glass.

2 [30] Bostik (2008). V-70 High Strenght Structural Glazing Silicone Adhesive Sealant Technical Data

3 Sheet.

4 [31] Belis J, Mocibob D, Luible A, Vandebroek M (2011). On the size and shape of initial out-of-plane

5 curvatures in structural glass components. Constr Build Mater;25(5):2700 13.

6 [32] UNI-EN 1993-1-1: 2005. Eurocode 3 design of steel structures Part 1-1: general rules and rules

7 for buildings; May 2005.

8 [33] Amadio C, Bedon C (2013). A buckling verification approach for monolithic and laminated glass

9 elements under combined in-plane compression and bending. Engineering Structures 52: 220-229.

26



List of Tables

Table 1

Table 1. Correction factor k1 for LU beams in LTB [34].

Loading case Loading condition Moment distribution Correction factor k1

(a) Constant 1

(b) Parabolic 1.13

(c) Triangular 1.32

1
Table 2

Table 2. Correction factors k1,min* for LR glass beams in LTB (with ky= 0.184N/mm2 the joint stiffness) [5].

Loading case Loading condition Load application Correction factor k1,min*

(a) Top-edge 0.95

(b) Bottom-edge 0.95

(c) Top-edge 0.75

(d) Bottom-edge 0.85

2



List of Figures

Figure 1

(b)

(a) (c) (d)

Fig.1. Example of application of glass fins as support for façade panels and roofs.
(a) [1]; (b) [2]; (c) [3]; (d) [4].

1
Figure 2

(a) (b)
Fig. 2. (a) LTB of a laterally unrestrained monolithic beam under constant bending moment My; (b) cross-
section.

Figure 3

(a) (b)
Fig. 3. Typical glass beam-to-roof connection by means continuous silicone joints.
(a) Overview and (b) Analytical model (cross-section).

2
Figure 4

29 L0=3000mm, b=300mm 3
t=10mm
25

21
2

17
RM

nR=1
13
10
9

1
-4 -2 0 2
10 10 10 10
ky [N/mm2]

Fig.4. Analytical estimation of the non-dimensional RM coefficient (Eq.(4)) for a monolithic glass beam
under positive, constant bending moment My.

3
Figure 5

29
L 0=1000mm
25 ky=0.6136N/mm2 L 0=5000mm

21

17
RM

13
ky=0.184N/mm2
9
ky=0.6136N/mm2
5 ky=0.184N/mm 2

1
6 7 8 9 10 11 12 13 14 15 16 17 18 19
t [mm]

Fig.5. RM coefficient (Eq.(4)) for LR monolithic glass beams (with 0.184N/mm2 ky 0.6136N/mm2 derived
from [5, 27, 29, 30], k = 0 and b= 100mm).

4
Figure 6

Fig.6. Preliminary FE-numerical study. Detail of the ABAQUS/Standard FE-model.

5
Figure 7

29 L0=3000mm, b=300mm
t=6mm (Eq.(4))
25 t=6mm, ABAQUS-lba
t=10mm (Eq.(4))
21 t=10mm, ABAQUS-lba
t=19mm (Eq.(4))
17 t=19mm, ABAQUS-lba
RM

13

1
10-4 10-2 100 102
ky [N/mm2]

Fig.7. Analytical (Eq.(4)) and numerical (ABAQUS-lba) comparison of RM amplification factors (k = 0).

6
Figure 8

(a) nR,crit= 1 (ky= 0) (b) nR,crit= 4 (ky= 0.184N/mm2 [5])

(c) nR,crit= 2 (ky= 0.184N/mm2 [5])

Fig.8. Critical buckling shapes of (a) LU glass beams, compared to (b), (c) LR glass beams.
(a) and (b) L0= 3000mm, b= 300mm, t= 10mm; (c) L0= 3000mm, b= 400mm, t= 25mm.
ABAQUS/Standard, white-to-black contour plot.

7
Figure 9

Fig.9. Reference analytical model for the LTB analysis of LR beams under uniformly distributed loads q or
concentrated loads F at mid-span.

8
Figure 10

10 10
L0=3000mm, b=300mm, t=10mm L0=3000mm, b=300mm, t=10mm
9 My 9 My
8 qA 8 qB
FA FB
7 7
6 6
RM

RM
5 5
4 4

3 3
2 2
1 1
-4 -2 0
10 10 10 10-4 10-2 100
2 2
ky [N/mm ] ky [N/mm ]
(a) (b)
Fig.10. Numerically estimated amplification factor RM (ABAQUS-lba) for beams under constant bending
moment My, distributed load q or mid-span concentrated load F. (a) top-edge loads; (b) bottom-edge loads.

Figure 11

Fig.11. Critical buckling shape of a LR glass beams (L0= 3000mm, b= 300mm, t= 10mm, with ky=
0.184N/mm2 [7]) under top-edge mid-span loads FA. ABAQUS/Standard, white-to-black contour plot.
9
Figure 12

1.2 1.2
ky=0.184N/mm2
FA (ABAQUS-lba)
1.1 1.1
k1,min*

1.0 1.0
k 1*

k 1*
0.9 0.9

0.8 0.8

ky=0.184N/mm2
0.7 0.7
qA (ABAQUS-lba)
k1,min*
0.6 0.6
7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30
L0/b [-] L0/b [-]
(a) (b)
Fig.12. Numerically estimated k1* correction factors for LR beams in LTB, with ky= 0.184N/mm2 the joint
stiffness (1000mm L0 5000mm; 8mm t 25mm).
(a) distributed loads qA; (b) mid-span concentrated loads FA.
Loads applied at the top-edge of the beams.

10
Figure 13

1.2 1.2
ky=0.184N/mm2
1.1 1.1 FB (ABAQUS-lba)
k1,min *

1.0 1.0
k 1*

k 1*
0.9 0.9

0.8 0.8

ky=0.184N/mm2
0.7 qB (ABAQUS-lba) 0.7
k1,min*
0.6 0.6
7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30
L0/b [-] L0/b [-]
(a) (b)
Fig.13. Numerically estimated k1* correction factors for LR beams in LTB, with ky= 0.184N/mm2 the joint
stiffness (1000mm L0 5000mm; 8mm t 25mm).
(a) distributed loads qB; (b) mid-span concentrated loads FB.
Loads applied at the bottom-edge of the beams.

11
Figure 14

Fig.14. Test setup for shear experiments on silicone sealant joints [6, 7].

Figure 15

Fig.15. Progressive shear failure of a structural silicone specimen [6, 7].

12
Figure 16

600

500

400
F [N]

300

200

100
wsil= hsil= 6mm, lsil=100mm
0
0 5 10 15 20 25 30 35
d [mm]

Fig.16. Typical load F – displacement u curve obtained from experiments [6].

Figure 17

Fig.17. LTB response of a monolithic glass beam laterally unrestrained (LU, n= 1) or restrained (LR, nR= 4).
Effects of continuous lateral restraints (ABAQUS-inl).

13
Figure 18

1.0
0.8
RM= 3.78 (FT)
0.6
0.4 x/L0=0 RM= 2.81 (HS)

0.2 RM= 1.94 (AN)

x/L0=1
0.0
-0.2
-0.4
-0.6
L0=3000mm, b=300mm, t=10mm
-0.8 RM= 4.22 (joint)
ky= 0.184N/mm2, nR= 4, u0,max= L0/400
-1.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
x / L0

Fig.18. Stress ratio R evolution in the silicone joint, as a function of the applied bending moment
(ABAQUS-inl).

Figure 19

Fig.19. LTB response of a LR monolithic glass beam.

Effects of imperfection shape and glass type (ABAQUS-inl).
14
Figure 20

6 Mcr,R(E) 5.70Mcr(E)

5 B
A
4

3 C L0=3000mm, b=300mm, t=10mm

nR= 4, ky= 0.184N/mm2
2 FT
u0,max= L0/1000
HS
1 AN u0,max= L0/400
Imperfection shape u0,max= L0/100
0
0.00 0.01 0.02 0.03 0.04 0.05 0.06
(umax - u0,max) [m]

Fig.20. LTB response of a LR monolithic glass beam.

Effects of imperfection amplitude and glass type (ABAQUS-inl).

Figure 21

Fig.21. LTB response of a LR monolithic glass beam.

Effects of loading condition, with FB denoting mid-span bottom-edge concentrated loads (ABAQUS-inl).
15
Figure 22

Fig.22. Design buckling curve for LR glass beams in LTB under constant bending moment My.

Figure 23

Fig.23. Design buckling curve for glass beams in LTB under constant bending moment My. Calculation
example for a same beam geometry laterally unrestrained (LU) or continuously restrained (LR) by means of
silicone joints.
16
Figure 24

Fig.24. Design buckling curve for glass beams in LTB. Calculation example for a same beam geometry
laterally unrestrained (LU) or continuously restrained (LR) by means of silicone joints, subjected to top-edge
mid-span concentrated loads FA.

17