Glass Struct. Eng. (2016) 1:131–151DOI 10.1007/s40940-016-0014-5

CHALLENGING GLASS PAPER

Development and test logics for structural silicone bondingdesign and sizing

Anneliese Hagl

Received: 28 December 2015 / Accepted: 14 March 2016 / Published online: 19 April 2016© Springer International Publishing Switzerland 2016

Abstract The application of structural silicone bond-ing requires dedicated design procedures due to techni-cal challenges such as high degree of incompressibil-ity of silicone, low Young’s modulus and non-linearmaterial characteristics. Advanced bonding designsfeature point-wise or line-type bonding geometriesbeyond the application range of ETAG 002 which isthe European guideline for structural silicone glazing.No straight forward approach exists to extend the rulesof ETAG 002 accordingly. In order to fill this gap, thefollowing topics need to be addressed: (a) Siliconebonding material tests for the identification of elas-tic characteristics beyond ETAG 002, (b) Small sam-ple tests for identification and quantification of failuremechanisms in addition to ETAG 002, (c) Structuralmechanical analysis methods complementary to mate-rial and small sample tests, (d) Safety concept exploit-ing the abovementioned tests and analyses for ensuringadequate bonding performance. This paper outlines acomprehensive development approach answering thislist in detail. Common material tests will be criticallyreviewed in view of fracture behaviour as small sam-ple tests demonstrate a totally different failure mecha-nism compared to usually applied dog-bone or H-typesample tests. Complementary Finite Element Analysesbased on material test characteristics allow correlatingthe test results with limit loading levels. These values

A. Hagl (B)A. Hagl Ingenieurgesellschaft mbH, Am Steinberg 34,82237 Wörthsee, Germanye-mail: hagl@a-hagl-ingenieure.de

serve for extrapolation from small sample test results tothe envisaged full scale application. Within this extrap-olation procedure, an adequate safety concept needsto be embedded accounting for temperature, humidity,aggressive environment etc. In this paper, a potentialsafety concept is presented for advanced line-type andpoint-wise bonding geometries.

Keywords Silicone · Bonding design · StructuralGlazing · ETAG 002

1 Introduction

Nowadays, large glass elements and glass facadesattract from aesthetic point of view of clients, architectsand engineers in increasing manner. Advanced manu-facturing and construction techniques and increasingknowledge how to engineer with this material, glassappears as ideal material for protecting humans on theone hand from unfriendly environmental conditionsand for serving transparency on the other hand in orderto provide natural light to the inhabitants. In addition,glass facades are often used as stylish elements for rep-resentative buildings.

The application of glass elements immediately posesthe challenge to select an adequate connection tech-nology. For classical glass facades, a simple bondingof rectangular cross section made by silicone mate-rial was already applied quite early in order to con-nect the glass elements with the underlying framework.

123

132 A. Hagl

For more advanced designs e.g. using glass fins, firstchoice in the past was the usage of screws and boltsas structural design elements ensuring the mechanicalload path between glass elements and building. Draw-backs of this approach were:

• The point-wise introduction of loads leading tounfavourable stress concentrations

• The preparation of the glass elements with drilledholes and the hereby required accuracy (especiallyfor laminated glass)

• Thehighnumber of parts asking for related logistics

During were developed in order to replace theseconnection elements for load carrying structures byadvanced bonding designs. Figure 1 shows as exam-ple a comparison of a conventional design approachbased on bolted point supports and the realized solutionusing a U-type bonding (Hagl 2003; Schadow 2006).It is obvious that this design offers a lot of advantagessuch as:

• An optically attractive solution due to the bondinghidden behind a façade profile

• Low manufacturing costs due to avoidance ofdrilling holes

• Simple assembly due to high accuracy of pre-assembled components, low number of parts andreduced danger of glass edge damaging due to themetallic edge protection

• More efficient material usage due to avoidance ofstress concentrations leading to lower glass thick-ness

• Improved failure behaviour due to distributed sup-port of the glass component by the bonding and theprofile

Silicone proved to be a suitable bonding material forstructural engineering due to its special characteristics.Its usage in bonding is also labelled as soft bondingdue to its high flexibility evoked by low material stiff-ness, high strains and the suitability of large bondingthickness of several millimetres. Thus, it is well suitedto compensate e.g. for thermal strains and for manu-facturing tolerances compared to stiff bonding designssuch as polyurethane etc. Other characteristics of sili-cones are:

• The resistance against weathering e.g. sun light,heat, ozone, SO2

• Low tendencies of aging such as stiffening or soft-ening (Wolf and Cleland-Host 2004)

• Applicability for awide temperature range (−50 ◦Cup to +150 ◦C)

• No oxidation at high temperatures• Low absorption of humidity and related moistureexpansion (swelling)

One reason for this special behaviour is related tothe backbone of very stable Si–O chemical bond-ing compared to classical C–C backbones of usualrubber derivates (Habenicht 1997). From physicalpoint of view, silicone shows the typical behaviourof hyper-elastic materials separating it clearly fromclassical behaviour of structural engineering materi-als such as concrete, steel and glass, see also Table 1.Hyper-elasticity means large elastic strains on theone hand and almost perfect incompressibility on theother hand (Treloar 2005). Furthermore, (filled) sil-icone shows typical rubber-like features such as theMullins effect (Mullins 1948) i.e. softening after load-ing and visco-elasticity which might pose additional

Fig. 1 a and bConventional versusbonding design for glass finsupport (design studyHerz-Jesu Church, Munich)

wind load

d = 20 mm

horizontal glass beam 3 x 15 ESG

300 mm

48 mm

Fixation

(a)

38 mm

wind load

adhesive

horizontal glass beam 10/15/10 ESG

steel U

300 mm

vertical glass beam

(b)

123

Development and test logics for structural... 133

Table 1 Mechanical properties of representative structural materials

Material Young’s modulus (N/mm2) Poisson’s ratio Density (kg/m3) Thermal expansion (10−6/K)

Silicone 1−2 <0.5 1330 190

Glass 70,000 0.23 2500 9

Stainless steel 170,000 0.3 7980 12

Aluminium 70,000 0.3 2700 24

Concrete (C25/30) 31,000 0.15−0.25 2500 10

challenges in design as material models for siliconemight get quite complex (depending on the modellingobjectives).

In order to simplify the design task with such a com-plex material, guidelines exist which trade simplicityof the design procedure with a very narrow design win-dow. The ETAG 002 (EOTA 2001) is an example forsuch an approach balancing simple tests and engineer-ing methods with tight design constraints. Hereby theinnovative potential of bonding in structural engineer-ing application is significantly limited. Focus in ETAG002 is put on two-sided line type bonding designs ofrectangular shape which can be considered as standardapplication of ‘mass production’ type for structuralglazing of facades. Material performance is mainlytaken into account by exploiting H-type samples whichrepresent the bonding geometry and boundary condi-tions in small scale. Figure 2 shows the behaviour ofsilicone under tensile loading for both a conventionaldog-bone specimen and the H-type small sample testof ETAG 002. As the presented properties are the engi-neering strains and engineering stresses (hereby elim-inating different test lengths and cross section areas ofthe specimens), the results can directly be plotted in onefigure. The H-type specimen typically features higherstiffness and lower limit stresses and strains than thedog-bone specimen under tensile loads. The test curvesplotted in this figure are representative curves selectedfrom a database consisting of more than 100 samples.The test results in this figure—as well as for all theother figures if not otherwise mentioned—are obtainedat room temperature and approximately 50% relativehumidity. Furthermore, representative curves are alsoused for the other figures.

The difference in behaviour between the two testarticles can be easily explained referring to Table 1.Typically the attachment plates of the H-type speci-men are made of aluminium, steel or glass showing

0

0.5

1

1.5

2

2.5

0 0.5 1 1.5 2

Engineering Strain [-]En

gine

erin

g St

ress

[N/m

m²]

ETAG H-Specimen

Dog-bone

Fig. 2 Silicone tensile tests: Dog-bone versus ETAG002 H-typespecimen

Young’s moduli of several orders of magnitude higherthan silicone. Thus, the silicone bonding in the H-typespecimen can be considered as almost rigidly attachedleading to different kinematics. While this very stiffattachment leads to significant suppression of lateralcontraction of the silicone close to the interfaces, thedog-bone specimen allows the silicone to freely con-tract in lateral direction. In combinationwith the almostperfect incompressibility of silicone, the constraintsin lateral contraction of the H-type specimen leads tohigher effective stiffness as the H-type specimen cannot trade elongation versus lateral contraction in thesame manner as the dog-bone specimen consideringthe boundary condition of nearly constant volume.

On the other hand, the suppression of lateral contrac-tion of theH-type specimen leads to non-uniform stressdistributions and to stress concentrations allowing toexplain the earlier failure with respect to engineeringstrains. In combination with higher stiffness, this prop-erty translates also into lower engineering stress val-ues to be achieved for the H-type specimen. Conclud-ing, the combination of typical structural engineeringmaterialswith silicone—asdemonstratedby theH-typespecimen by the attachments typically built of met-als and/or glass—offers the possibility to significantly

123

134 A. Hagl

Fig. 3 Comparison ofmechanical characteristicsfor different siliconeapplications loaded bynominal stressσN = 1N/mm2

Specimen Type

Dog-Bone Specimen

ETAG H-type Specimen

Planar Round Point Support

U-type Point Support

Stress and strain field Uni-axial (1D) Complex (3D) Complex (3D) Complex (3D)

Strain εN=Δl / l0* 1,19 0,82 0,08 0,06

Stiffness σN / εN 0,84 1,21 12,9 16,9

Stiffness related to dog-bone 1 1,44 15,4 20,1

* Nominal stress 1 N/mm2* Nominal stress 1 N/mm2 = σN

influence the properties of the bonding by adequatedesign of the bonding geometry. This interesting fea-ture allows tailoring mechanical characteristics of thebonding such as stiffness and even failure mechanismsto a large extent. This statement is underlined by theexamples given in Fig. 3 in view of flexibility.

In Fig. 3, strains of different silicone applications arecompared for a nominal loading of 1N/mm2 for a rep-resentative dual component structural glazing silicone.Based on the strain values, a secant stiffness is derivedfor a dog-bone specimen, for the H-type specimen ofETAG 002, for a planar point support and for a U-typepoint support. The stiffness properties of the latter threeassemblies is referred to the dog-bone one showing thatfor these applications, the stiffness of the bonding canbe easily varied beyond one order ofmagnitude.Aswillbe shown later, also the failure mechanisms of theseapplications differ e.g. due to dual load paths in caseof the U-type point support. The flange length is herethe key parameter for partitioning into a tension loadpath and a shear load path. It is obvious that for theadequate tailoring of the bonding properties, a moredetailed knowledge of the silicone material is requiredcompared to the more pragmatic approach followed byETAG 002. The paper addresses this challenge by thefollowing sections:

• Siliconebondingmaterial tests for the identificationof elastic characteristics beyond ETAG 002

• Small sample tests for identification and quantifi-cation of failure mechanisms

• Structural mechanic analysis methods for siliconebondings

• Safety concept for ensuring adequate bonding per-formance

2 Silicone bonding material tests for theidentification of elastic characteristics beyondETAG 002

In order to describe the siliconematerial characteristicsin bonding designs appropriately, two pre-requisitesneed to be fulfilled in advance. First, material laws needto be set up which describe the structural behaviour ofthe silicone material to an adequate extent from engi-neering point of view. Second, the parameters of thesematerial laws need to be determined by results of ade-quate material experiments. In this section, focus is putonmechanical properties such as elasticity and fracturebehaviour due to the special characteristics of the sili-cone material. Other physical properties such as ther-mal or humidity based extension of the test materialmight be identified by conventional tests as for othermaterials.

Regarding the elastic properties,material lawsmightdescribe time independent or time dependent behav-iour (e.g. depending on strain rates). It is obvious thattime independent behaviour is simpler to manage fromboth descriptive and experimental point of view as oneexternal input i.e. the time dependent load history canbe ignored. In view ofmaterial law application, it needsto be checked for this approach whether the sizing ofthe bonding is significantly affected by strain rates ornot. For conventional loading of the silicone bonding

123

Development and test logics for structural... 135

e.g. by dead loads, thermal loads or wind loads theneglecting of load rates can be assumed for sizing pur-poses. For bomb blast linked to high strain rates this issurely not the case.

With respect to time independent material laws, lin-ear elastic or hyper-elastic material laws might be suit-able depending on the required strain amplitudes. Apeculiarity of filled rubbers and silicones is theMullinseffect describing softening of the material characteris-tics after pre-loading. This behaviour can be interpretedas a kindof damaging, andmodifiedhyper-elasticmate-rial lawswere developed in order to describe this behav-iour e.g. using a damage parameter. The Mullins effectshould not be mixed up with visco-plastic behaviourwhere plastic strains remain after unloading of thematerial. For the description of the time dependentbehaviour, visco-elastic properties need to be identifiedby time dependent loading e.g. by cyclic time historiesof varying amplitudes and rates.

Concerning fracture behaviour, fatigue needs to bedifferentiated from typical quasi-static behaviour. Asalready outlined in Sect. 1, fracture behaviour of sili-cone might differ comparing pure material tests (e.g.dog-bone tests) and small sample tests (such as H-typespecimens) e.g. evoked by non-uniform stress distrib-utions. Figure 4 demonstrates the impact of bondingstress singularities at the edges of the bonding by afailure starting in these areas with high stress values.Thus the failure is triggered by the special design ofthe specimen not allowing any conclusions on the purematerial behaviour itself by simple analysis means. Inany case in this paper, it is assumed that silicone bond-ings mechanically degrade due to cohesive failure andnot due to adhesive failure. This failure mechanism cantypically be ensured by adequate surface treatment e.g.by the use of primers etc. in accordance tomanufacturerinstructions and guidelines.

For most of today’s structural engineering appli-cations, silicone behaviour can be approximated by atime independent description taking into account limitloads i.e. limit stresses or stiffness degradation limitsfor bonding sizing purposes. Thus, material tests arerequired to provide related material properties as sil-icone characteristics can significantly vary dependingon the chemical ingredients and the production process.Furthermore, curing might be an important parame-ter especially for one component silicones curing byhumidity.

The most basic description of isotropic elasticmaterial—i.e. independent from load orientation—isgiven by three parameters of which two are indepen-dent and the third can be derived. These parametersare:

• The Young’s modulus E describing the change ofnormal stress σ due to change of normal strain ε inlaterally unconstraint conditions: σ = E ε

• The shearmodulusGdescribing the changeof shearstress τ due to change of shear strain γ : τ = G γ

• The Poisson’s ratio ν describing the lateral con-traction εq due to longitudinal extension εl : ν =− εq / εl.

The material properties E, G and ν are linked by theequation E = 2G(1+ν). It should be highlighted here,that hyper-elastic material such as rubber and siliconecan also be described by these linear elastic materiallaws for small strains. Thus, it can be directly con-cluded that at least two different kind of material testsare required to describe the elastic behaviour of sili-cones. Exploiting the assumption of incompressibility,one set of tests can be replaced by setting ν = 0.5.

Typical test set-ups for this purpose are tension andshear tests. For tension tests, dog-bone specimens areproven test geometries in order to identify materialproperties under uniaxial strain, see Fig. 5 left. Regard-ingpost-processingof the experiment it has to bekept inmind that the displacement travel of the testingmachineis not suitable as only the middle part of the speci-men satisfies almost perfect uniaxial material loading.Thus, extrameans are required tomeasure the displace-ments of the test range. As silicone material is verysoft, optical techniques such as video extension-metersare applied in order not to affect the test results bymechanical devices leading tomechanical interactions.As already mentioned in Sect. 1, H-type specimens are

Fracture initialisation

Fig. 4 H-type specimen starting to fail at bonding edge

123

136 A. Hagl

clamped in tension testing machine

H-specimen is not useful for material testing under tension loading!

Shear Poisson’s Ratio Tension

Fig. 5 Typical test specimens used for silicone materials

not an alternative to these tests; they might be con-sidered as complementary tests depending on the testobjectives.

The samples discussed here are based on a standard2-component structural glazing silicone of Dow Corn-ingwhich is certified according toETAG002.The othertests in this paper refer to the same material.

Regarding shear tests, it is in general more difficultto achieve a pure shear state as a pure shear state ischaracterised by shear loads acting on all four inter-faces in the shear plane. Typically a simple shear testis used instead as shown in Fig. 5 with respect to thethree specimens in the middle. Simple shear tests aredesigned by two attachment plates which introduce rel-ative displacements. Free warping as a kind of imper-fection appears at the free surfaces at the ends of thespecimen. In case the specimen is long compared tothick the area with the free warping can be neglecteddue to low impact. Nevertheless, rotation of the attach-ment plates might be a more important issue affectingthe test results.Multiple specimensmight eliminate thisissue by trading the improved load introduction versusloss of quantitative insight during failure as themultiplebodies might fail differently. Figure 6 underlines thisstatement by showing the different pre- and post-failurecharacteristics of the specimens.

Specimen 2xS shows significant rotations whichmight also explain the different behaviour in the pre-failure domain.

Silicone material shows an almost perfect incom-pressibility as typical characteristics of hyper elasticmaterial behaviour. The level of incompressibility canbe assessed by comparing the bulk modulus with theshear modulus. This behaviour is also described by thePoisson’s ratio ν. If the Poisson’s ratio is approaching0.5 from below it can be shown by simple analyticalmeans that the material tends towards incompressible

0

200

400

600

0 5 10 15 20 25

Displacement [mm]

Load

[N]

1 x S w fixing 4 x S 2 x S

curves nearly parallel = same G

Fig. 6 Comparison of shear test specimens

0

2,5

5

7,5

10

0 0,1 0,2 0,3 0,4 0,5

Poisson's ratio

late

ral c

ontr

actio

n im

pact

on s

tiffn

ess

1

Glass νν = 0,23

Steel νν = 0,3

Silicone νν ≈≈ 0,498

Fig. 7 Impact of Poisson’s ratio on stiffness in case of perfectsuppression of lateral contraction

behaviour. From a theoretical point of view, incom-pressibility leads to a decoupling of the hydrostaticpressure (or hydrostatic tension) from the displacementfield meaning that hydrostatic stresses are not affectedby deformations but by boundary conditions i.e. exter-nal loads. On the other hand, perfect blocking of lat-eral contraction leads to infinite stiffness for the incom-pressibility case as shown in Fig. 7. In this figure, theratio of constraint to unconstraint stiffness is plottedversus Poisson’s ratio.

An evaluation of the Poisson’s ratio for a representa-tive two-component silicone by optical means (bi-axialvideo extensometer measurements) by using the speci-men geometry in Fig. 5 right, is demonstrated in Figs. 8and 9. Figure 8 shows the lateral strains εq plotted ver-sus the longitudinal strains εl. In theory, the negativeslope of the regression line provides the Poisson’s ratio—here 0.4979—and the line should intersect the origin.This figure demonstrates very good agreement betweentheory and experimental data. On Fig. 9, the obtainedPoisson’s ratio is directly plotted versus longitudinalstrain.

Perfect or almost perfect incompressibility has alsoother consequences with respect to constitutive mater-ial laws and several test approaches are no longer inde-pendent. Typical material testing scenarios are:

• Simple shear tests, see Fig. 10

123

Development and test logics for structural... 137

y = -0,4979x + 0,0013

-0,4

-0,3

-0,2

-0,1

0,00,0 0,2 0,4 0,6 0,8

l) [-]q)

[-]

test

Linear ( test)

Fig. 8 Regression of lateral strains versus longitudinal strains

0,00

0,25

0,50

0,75

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

l) [-]Poiss

on's

Ra�o

[-]

Fig. 9 Poisson’s ratio versus longitudinal strain

• Compression tests, see Fig. 11• One-axial tension or compression tests, see Fig. 12• Bi-axial tension (or compression) tests, see Fig. 13• Plain tension tests (also labelled as pure shear tests),see Fig. 14

Simple shear tests are typically characterised by largedisplacements and low stiffness for hyper-elastic mate-rial such as silicone as the molecular structure and therelated properties of this kind of polymers allows highmobility of the chain molecules and thus low resis-tance against external deformations. As already men-tioned, care has to be given to specimen design in orderto reduce warping phenomena violating perfect shearconditions as much as possible.

Different to shear tests, compression tests typicallyreveal orders of magnitude larger stiffness values com-pared to the shear tests. The molecular explanation forthis behaviour is seen in the high energy required inorder to change the length of the molecule chains. Itshould be added that it is quite difficult from practi-cal point of view to design adequate test set-ups forpure compressive (or tension) loading. Thus other testswere defined in order to quantify the Poisson’s ratioand thus compressibility. One approach is shown inFig. 5 right where the Poisson’s ratio is determined bysensing the lateral contraction by optical means such

Fig. 10 Simple shear test

Fig. 11 Compression test

as video extensometers. Measurement accuracy mightbe an issue here as the fundamental change of mate-rial behaviour approaching ν = 0.5 is not adequatelyscaled by measuring geometric parameters. An alter-nate approach is described in (Wolf and Descamps2002) exploiting the characteristics of wave propaga-tion in the silicone material.

In Fig. 12, a tension test set-up is shown. For per-fect incompressibility, same test results are achieved bystretching the specimen uniaxial as shown on the left in

123

138 A. Hagl

uni-axial tension test

kinematic equivalence λ2 = λ3 = λ1

-0.5

equibiaxial compression test

1

2 3

Fig. 12 Tension test

direction 1 or by compressing lateral directions 2 and 3as shown on the right. The stretch λ is hereby definedby the deformed specimen length divided by the unde-formed length. The difference from material point ofview is only the hydrostatic pressure while the defor-mation patterns are the same. Based on this outcome, a

equibiaxial tension test

uni-axial compression test

kinematic equivalence λ1 = λ2, λ3 = λ1

-2 = λ2-2

3

1

2

Fig. 13 Bi-axial tension test

unique manifold of loading schemes exist to generatethe motion patterns in the figure as load combinationsof axial loading and lateral compression can be usedfor the same result. Of course, the statement is valid aswell for uniaxial compression and lateral tension.

In a complementary manner to Fig. 12, duality isshown for bi-axial tension tests in directions 1 and 2 anduniaxial compression tests in direction 3 in Fig. 13. Thisduality is interesting in the sense that uniaxial compres-sion tests are typically difficult to design and to performdue to potential buckling of the specimen leading tocompact specimen designs and due to friction issuesat the interfaces if placed between interfacing plates ofthe testing machine. Nevertheless, the adequate intro-duction of biaxial loading schemes into the specimenby mechanical components challenges the test set-upas well. Due to buckling issues, a biaxial compressiontest might be less practical. Here, a uniaxial tension testmight be favoured instead.

If displacements of the specimen can be perfectlyconstraint in one direction, plain tension or compres-sion tests can be performed, see Fig. 14. The interestingpoint of these tests is that due to incompressibility con-ditions, tension strains in one direction are coupled tocompression strains in the other direction in such a waythat pure shear motions are achieved under 45 deg ori-entation angle for small motions. Thus perfect shearconditions can be achieved by this test approach. Asalready mentioned, the challenge in designing the testset-up is the avoidance of motions and strains in thelateral direction 2.

Regarding applicability and validity of the materialtests, focus should be put on stress and/or strain rangesassumed to be relevant for the bonding designs. Fig-ure 15 shows the load—displacement or respectivelystress—strain curves for tension tests and shear tests.The tension test curve reveals a significant non-linear—regressive—behaviour before failure while the shear

Fig. 14 Plain tension tests

plane tension test

kinematic equivalence λ2 = 0, λ3 = λ1

-1

plane compression test 3

1

123

Development and test logics for structural... 139

Fig. 15 Tension and shearload displacement curves

Tension Shear

0

100

200

300

400

500

0 5 10 15 20

Displacement [mm]

Load

[N

]

0

15

30

45

60

0 10 20 30 40 50

Displacement [mm]

Load

[N

]

tests behave close to linear (or even slightly progres-sive) in the non-damaged region. Curvatures close tothe origin need to be assessed in view of test imper-fections as it typically proves to be quite difficult toperfectly insert a highly flexible specimen into the test-ing machine. It is obvious that non-linearity e.g. in thecase of the tension curve can be easily considered inthe tested range but extrapolation of the elastic mater-ial law is only adequately possible if based on physicalfirst principles e.g. provided bymaterial lawsmotivatedby physical considerations.

It should be added here that the load rate has animpact on test results leading to stiffening and higherfracture loads for high rates which is e.g. of interestfor bomb blast cases. The tests and test results dis-cussed in this paper refer to quasi-static conditions andare extracted from material test standards, see (DIN53 504 2009). Displacement rates between 50 mm/minand 500 mm/min do not show significant impact onelastic behaviour for the related dog-bone specimens.

Of course, the stress—strain relationships obtainedby the material tests which build the basis for derivingmaterial laws depend on the environmental conditions.Especially in view of sustainability aspects, a varietyof operating conditions might be of interest such as:

• Temperature variations• Creep loading e.g. in view of dead loads• Aged versus unaged conditions.

An example for the complex behaviour of siliconematerial is presented below in Fig. 16 plotting the char-acteristics under repeated loading of varying ampli-tude. The Mullins effect showing hysteresis charac-teristics is clearly visible. Interestingly, the materialproperties seem to be recovered if the amplitudes ofthe former load cycles are exceeded. This behaviouris also used in view of sizing as limit loads are linked

0

20

40

60

80

100

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

Displa cement [mm]Lo

ad [N

]

Path of further loading

Path of first loading

Fig. 16 Loading histories on material behaviour

to non-degraded material properties assuming a dom-inant i.e. critical load regime. On the other hand, carehas to be given for the pre-treatment of specimens asin case of miss-treatment, in-voluntary softening of thespecimen might occur as already mentioned above.

3 Small sample tests for identification andquantification of failure mechanisms in additionto ETAG 002

In this paper the topic “small sample tests” covers testsegments of line-type bonding designs on the one handand single point supports on the other hand. The mostpopular non-trivial line-type bonding design consistsin a U-type bonding geometry built by attaching a(PFC) U-type element to the edge of a glass element.Other designs might be L-type bonding geometries inwhich only one of the flanges exist or T-type bondinggeometries for which the inner glass panes need to bestaggered. Concerning point supports, the most naturalchoice is to consider circular geometries due to rota-tional shapes with planar surfaces of the point supportsleading to disk type bonding geometries (Wolf 2011).

Typically, the behaviour in shear for such kind ofbondings can be described quite easily by mapping

123

140 A. Hagl

material test results of shear samples to the bondinggeometry without experiencing too high deviations.In contrast, tension loading leads to a quite complexbehaviour e.g. due to suppression of lateral contractionwhich does not occur for shear deformations. Goinginto the other direction, it is widely accepted based onengineering practice that elastomerics do not fail underhigh compressive loads in view of conventional struc-tural engineering applications. Thus, focus is put onbonding behaviour under tension loading in this sec-tion.

3.1 Line type bonding designs

For line type bonding samples, the question ariseswhich sample width to be used. From practical pointof view, the testing machine and the implemented loadcell capacity pose constraints towards the upper boundsof the small sample widths. On the other hand, thealmost perfect incompressibility of silicone requestslarge widths in order to be representative especiallyfor those bonding geometries featuring large constraintinterfaces and low amount of free surfaces. In this case,the incompressibility leads to significant lateral con-traction in case of low small sample widths herebysignificantly altering mechanical performance of thebonding. Thus, a lower bound is given by representa-tiveness of the line-type bonding with respect to limitloads and other characteristics. Specimen widths typ-ically chosen in practice are in the order of 50 mm inorder to use testing machines typically available in lab-oratories. A typical result for such kind of specimen isgiven in Fig. 17 in terms of a load curve plotted ver-sus displacements for a representative two componentstructural glazing silicone. It is obvious that the patternof this curve is quite different from the results obtained

by a simple dog-bone specimen where a monotonousincrease is terminated by a sudden collapse.

The behaviour of this complex bonding allows toexploit silicone bondings in a favourable way as thesudden failure event of dog-bone specimens poses seri-ous engineering challenges by its aggressive failurepattern due to unstable crack growth (as it is e.g. thecase for brittle ceramics) while this “yield-like” behav-iour of the U-type bonding leads to good-natured fail-ure behaviour. Regarding the U-type bonding, differ-ent segments of behaviour are visible which mightbe attributed to different mechanisms in the specimen(Hagl 2008a, b).

Initially, the bonding is fully operational and dueto the high degree of encapsulation of the bonding incombination with the high level of incompressibility,the bonding reacts quite stiff which is in clear contrastto the low stiffness obtained by dog-bone specimens.This phenomenon can be localized at the front region ofthe U-type bonding geometry featuring dominant ten-sile loads. With increasing displacements, the slope ofthe load curve reduces indicating a change in bondingbehaviour. In Fig. 17 this behaviour starts at approxi-mately 2 mm displacement. The optical observation ofspecimen during testing and numerical analyses showthat due to the high stiffness of the front region com-pared to the side regions bonding stresses increase quitesignificantly in this area initiating a material degrada-tion at the front region while the side regions are stilloperative. At approximately 8.5 mm a sudden drop ofloads is noted for the specimen presented in Fig. 17.This event is linked to the total failure of the frontregion combined with a redistribution of the main loadpath towards the side regions. The side regions showthe characteristic motion pattern of shear which is typi-cally related to soft behaviour and large displacementsbefore break of the silicone. Thus, loads can still be

Fig. 17 Loads versusdisplacements for a U-typebonding under tensionloading

0

1000

2000

3000

4000

5000

0 2 4 6 8 10 12 14 16 18

Displacement [mm]

Load

[N]

Shift of load path (shear) to side regions Beginning failure

of front region

Total failure of front region

Fully operational

bonding

Representative load level:

max. service load

3 x 12 mm length 50 mm

123

Development and test logics for structural... 141

transmitted by this second load path of the side regionsindicated also by a positive slope between 10.5 and14.5 mm. Beyond 14.5 mm, total failure appears withdegradation of the side regions.

This hypothesis was validated testing degraded U-type bondings in comparison to non-degraded (base-line) samples. Specimens were manufactured usingPTFE foils applied on either the front region or the sideregion in order to disable load transfers in the relatedregions, see (Hagl 2008a). It could be shown that thedominant role of the front region for small displace-ments is clearly visible comparing the baseline bondingwith the bonding with side regions disabled both fea-turing an operational front region in the beginning andhigh initial stiffness. On the other hand the behaviourat large displacements is easily identified as dominatedby shear comparing the baseline bonding with the onewith the front region disabled.

Similar results are obtained for T-type and L-typebonding geometries as presented and discussed in thesame reference. All these bonding geometries in com-mon are front and side regions reacting by tension andshear loads and the activation of front region(s) stiff-ening by suppression of lateral contraction.

This knowledge can be used to tailor the failuremechanisms of the U-type bonding. Small bondingthickness leads to stiffer behaviour for the front region(subjected to tension loading) and the side region(experiencing shear loading) as strains are higher forsame displacements. Thus the slopes of the load curvecan be modified by the adequate choice of bondingthickness for front and side regions. The thicknessof the shear loads is obtained by the geometric com-bination of the PFC (parallel flange channels) widthand the glass thickness while the thickness of thefront region can be adjusted using spacers or similardevices. Nevertheless, one is not totally free in select-

ing these bonding properties as manufacturer recom-mendations need to be followed asking for minimumand maximum bonding thickness e.g. in view of curingprocesses.

Load carrying capacities are affected by the bondinginterface areas at front and side region as larger areaslead to higher limit loads. As the front region geometryis given by the thickness of the glass the related areaneeds to be considered as externally defined. On theother hand, the side region area can be designed bythe length of the flanges of the PFC. Thus failure ofthe second load path i.e. at 14.5 mm in Fig. 17 can bequantitatively modified by this design means.

The test presented in Fig. 17 is performed by intro-ducing the loading of the testing machine by brack-ets acting on the edges of the PFC. On the one handthis approach guarantees a quite homogeneous loadintroduction; on the other hand designs in structuralengineering applications will be different using dis-crete anchor points such as bolts or similar. Typically,load introduction means are applied to the front side ofthe PFC/bonding thus potentially affecting mechanicalperformance of the front region by more inhomoge-neous loading patterns.

3.2 Point support bonding designs

Point supports typically differ from the line type bond-ings discussed in the section before that only a “frontregion“ of the bonding exists. This difference is alsovisible in the load curve, see Fig. 18.

In this figure, three regions with different mechan-ical characteristics can be identified. The first regionranging from zero to approximately 0.3 mm is charac-terized by large slope i.e. high stiffness. This behav-iour is comparable to the results obtained for the U-

Fig. 18 Loads versusdisplacements for a planarpoint support under tensionloading

0

1000

2000

3000

0 1 2 3 4 5

Load

[N]

micro damages in the bulk of adhesive

macro cracks occur and lead to total failure

fully operational bonding

different final failure behavior due to flaws in adhesive

Begin of stiffness degradation

Displacement [mm]

123

142 A. Hagl

Fig. 19 Beginning fracture pattern for planar point supportunder tension loading

type bondings, see Fig. 17. The second region rang-ing from 0.3 to 3 or 5 mm respectively is character-ized by a very low slope with almost no load increase.Finally, total collapse of the specimen is noted. Inthe figure, the behaviour of two different samples isshown being identical up to displacements of 2 mm.As identified from the fracture surface after inspection,flaws are assumed to favour early failure for one of thesamples.

In order to investigate the behaviour in region 2 inmore detail, load histories of specimenswere stopped atdifferent levels and the specimens cut for investigation.An example for the beginning failure pattern is shownin Fig. 19. Interestingly, the systematic analysis of thesepatterns leads to the result that the macroscopic bond-ing failure does not start outboard or in the middle butin radial locations between one third and two thirds ofthe specimen. With increasing displacements, the frac-ture area propagates in inboard and outboard directions.Typically, cracks get visible at the outer surface at theend of region 2/start of region 3.

In order to identify the beginning degradation of thepoint support which is relevant for sizing—e.g. in viewof sustaining service loads—point support stiffness isderived by numerically differentiating the load curveobtained by the testing machine, see Fig. 20.

Consistent to Fig. 18, high stiffness is obtained forregion 1 i.e. for lower displacements than 0.3 mm anda sudden reduction is visible for larger displacements.Thus a bonding degradation on microscopic level ispostulated for approximately 0.3 mm. For the planarpoint supports of 50 mm diameter and 5 mm bondingthickness the degradation is linked to a level of approx-imately 1700 N for the investigated two-componentstructural glazing silicone. As expected, larger pointsupport diameters lead to higher loads as for 70 mmdiameter, load levels of 4100 N are obtained for the

0

2000

4000

6000

8000

0 0.2 0.4 0.6 0.8 1

Displacement [mm]

Stiff

ness

[N/m

m] 0,28 mm

Fig. 20 Stiffness characteristics of planar point supports undertension loading

0

500

1000

1500

2000

2500

0 0.5 1 1.5 2

Displacement - mm

Load

- N

Fig. 21 Tension loading of planar point supports

same condition. For comparison, limit loads i.e. high-est loads are noted as 2500 N for 50 mm and 5100 Nfor 70 mm, see (Hagl 2008b).

Different attempts were taken to optimize themechanical behaviour of point supports under tensionloading. Convex and concave point support fittingswere designed and tested, see also (Hagl 2007). Theidea was to vary bonding thickness versus radius and toanalyse the resultingmechanical behaviour. The exper-imental results show that concave point support shapeslead to slightly higher loads but less displacementsin view of the total collapse. Thus potential for opti-mization seems limited in view of these tests. Otherapproaches are related to sunk point support specimens,see (Hagl 2008b). Here, dedicated pre-treatment of theglass units is required—i.e. drilling holes into the glasspanes – which needs to be justified by improved perfor-mance (e.g. leading to lower number of point supports)or other cost saving issues.

The planar point supports were also tested undercyclic loading in order to check for potential fatigueissues. Swelling tension loads were applied by pre-scribed displacements, see Hagl (2010). In Fig. 21,maximum displacements for the load cycles weredefined by 1.5 mmwhile minimum reversal conditions

123

Development and test logics for structural... 143

0

500

1000

1500

2000

2500

0 0.5 1 1.5 2

Displacement - mm

Load

- N

0.25 0.50 0.75 1.50

Fig. 22 Varying amplitude of tension cyclic loading for planarpoint supports

were given by a small tension load in order to avoidmixing up tension and compression load schemes. TheMullins effect is clearly visible by the first cycle show-ing high loads for increasing displacementswhile lowerloads are obtained for decreasing displacements of thefirst cycle and for all subsequent cycles. Two lines areadded to the figure in order to characterize this behav-iour under cyclic loading using a kind of representativeslopes. The red line connects the upper reversal pointof the first cycle with the lower reversal point betweenfirst and second cycle while the blue line connects therelated reversal points of the last tested cycle beforebreaking the specimen. For practical reasons i.e. test-ing time, the number of cycles was limited to 100 andrelated displacement rate was set to 1 mm/s. On theone hand, the number of cycles seem to be quite lowwith respect to conclusive statements in view of fatiguebehaviour and resistance. On the other hand, changesare quite small approaching the last cycles.

Next, different displacement amplitudeswere tested,see Fig. 22. The different amplitudes are representedhere by the “slopes” of the first and last cycle. Theincreasing degradation is visible by comparing the lastcycle slope with the first cycle slope showing largerslope reduction with increasing cycle amplitudes. Theoutcome of these tests indicate that a loading below 0.3mm i.e. below stiffness degradation under quasi-staticloading (Fig. 20) is also acceptable from fatigue pointof view—at least with respect to the number of cyclesperformed during testing as results for 0.25 and 0.5mmamplitudes indicate.

4 Structural mechanical analysis methods forsilicone bondings

In order to improve physical insight in the structuralmechanics of silicone bondings and in order to com-

plement the experimental activities to understand thespecimen behaviour in more detail, numerical stud-ies are applied typically based on finite element meth-ods (Gent 2001). For the adequate application of finiteelement analysis (FEA) a description of the materialbehaviour is required in terms of a material law foreach part in the model. Material laws of interest forsilicone might be:

1. Isotopic linear material description given by twomaterial parameters (e.g. Young’s modulus, Pois-son’s ratio)

This description is applicable for small andquasi-steady strains where non-linear behav-iour and rate dependency can be neglected.

2. Hyper-elastic material laws based on parametricstrain energy definitions

This description is applicable for large quasi-steady strains where repeated loading and ratedependency can be neglected.

3. Hyper-elasticmaterial laws includingdamagemod-els

This description is applicable for large quasi-steady strains under repeated loadingwhere ratedependency can be neglected. Damage parame-ters are included to represent the Mullins phe-nomenon.

4. Visco-elastic material laws

This description is applicable for time depen-dent analyses in order to cover rate dependencyof the silicone material characteristics.

The parameters of the chosen material laws need to beidentified by material tests. It is important to note thatthe material model validity is only ensured to strainlevels as considered for material parameter identifica-tion. Validity and accuracy of the material laws regard-ing extrapolation in terms of strains, strain rates andtemperature is open and more or less directly linkedto the nature of the material law, e.g. only descrip-tive/empirical suitable with focus on interpolation orbased on physical principles allowing some kind ofextrapolation.

Commercial FE packages provide automatic proce-dures to adapt the parameters of the selected mater-ial model based on provided specimen test results in

123

144 A. Hagl

a black box approach. Regarding this approach, valid-ity of specimen test results has to be guaranteed, e.g.in view of the load curve origin in case of non-linearmaterial behaviour, and typically noweighting betweentest points can be performed which might be useful toadjust the model to the load regimes of major interest.The automatic determination of the material constantsis typically based on the following tests for distortionaldeformation constants:

• Simple tension/compression tests• Equi-biaxial tension tests• Simple shear tests• Pure shear tests

For volumetric deformation constants, pure volumetriccompression test results are required in principal.

In order to increase flexibility and to get more physi-cal insight, amanual procedure is presentedbelowstart-ing with the exploitation of shear tests. The advantagesand disadvantages of these two different methods canbe concluded as follows:

• Automatic determination by related pre-processorsof finite element packages:

– High user friendliness especially for novices– Implementation errors reduced by automatictransfer of results

– Nomathematical background required for opti-mization

• Manual determination as presented in this section

– High flexibility including and combining dif-ferent experimental results

– Total control of optimization procedure inclu-sive weighting for different test points

– Consideration of imperfect test conditions byrelated parametric FE studies (taking intoaccount imperfections, see e.g. Fig. 23)

Silicone behaves quite linear in a large range of shearoffering the advantage of being quite insensitive inview of origin point issues and amplitudes of the loadcurves, see Fig. 24. The slope of the shear stress ver-sus strain curve obtained by relating shear loads andshear displacements to the specimen geometry pro-vides the shear modulus G. In combination with thePoisson’s ratio ν this information is sufficient to definethe isotropic linear material law. Young’s moduluscan be derived assuming isotropic linear behaviour byE = 2G(1 + ν).

2.

1.75

1.5

1.25

1.

0.75

0.5

0.25

0.

Fig. 23 FE analysis of ETAG 002 specimen under shear loading

0

100

200

300

400

500

0 5 10 15 20 25

Displacement [mm]

Load

[N]

test FEA

Fig. 24 Shear test results and comparison with FEA

In case the FEA supports perfectly incompressiblematerial laws by special finite element formulations thePoisson’s ratio can be set to 0.5 at first glance assum-ing perfect incompressibility of silicone. Alternatively,values from literature (Wolf and Descamps 2002) orderived by geometrical measurements as shown inFig. 5 right can be applied. Parameter studies relatedto the Poisson’s ratio might help to determine whetheradditional effort has to be spent on a more accuratePoisson’s ratio definition. Furthermore, the applicationof Poisson’s ratio values might be limited by numeri-cal conditioning in case the formulation of the appliedfinite elements is not especially tailored towards thesimulation of almost incompressible behaviour.

In order to check the correct implementation of thematerial law, it is recommended to model the speci-men test also by FEA and to compare results as shownin Fig. 24. In this figure, experimental issues close tothe origin are visible which are caused by imperfectplacing the specimen in the testing machine. In case ofinadequate manipulation of the specimen, the Mullin’seffect might by visible for small loads i.e. close tothe origin. The explicit exploitation of almost perfect

123

Development and test logics for structural... 145

linearity in shear compensates this issue. In additionthe comparison allows to check for test imperfectionssuch as warping on free surfaces as the FEA will fea-ture this imperfections as well if adequately set up, seeFig. 23.

Typically tension tests by dog-bone specimens areavailable for the material descriptions. The load curvesshow declining characteristics at least for lower strains.Thus, these test results allow an adjustment of hyper-elastic material laws. In this demonstration, a polyno-mial description for the strain energy density of hyper-elastic behaviour is applied (Cij, Dk material coeffi-cients, I1, I2, J invariants):

W =N∑

i+ j

Ci j(I1 − 3

)i (I2 − 3

) j +N∑

K=1

1

Dk(J −1)2k .

In case of a low order material model featuring onlytwo coefficients C10 and C01, the model is labelledMooney-Rivlin. It can be shown by analytical meansthat for polynomial descriptions, the following state-ment used as boundary condition for parameter iden-tification is valid for small strains: C10 + C01 = G/2.Thus, the objective to match the experimental resultscan be mapped to an adjustment of the two coeffi-cients using “test” functions f10 and f01 which wereobtained for C10 = G/2,C01 = 0 and C10 = 0,C01 = G/2. For this purpose a linear combination yfea

of the “test” curves is set up using the scales α and 1−α

for the two test curves. Here, the special approach of thestrain energy function consisting of a sum of terms isexploited for a given deformation pattern by adjustingthe contributions of the various terms. This procedure issketched in Fig. 25 by showing the test curve and thesetwo numerical functions. A classical approach for thedetermination of the unknown coefficients consists inbuilding the squared sum of differences between testand analysis and to minimize this sum using non-linearsolvers capabilities in spreadsheet programs by askingfor a minimum varying the coefficients C10 and C01 bythe scale α.

This approach is formalized by the following steps,see also Fig. 26:

1. Calculation of difference between numerical andexperimental forces/stresses of the collocationpoints for initial selection of C10 and C01:

�i = yexpi − yfeai

0

10

20

30

40

50

0 10 20 30 40 50 60 70 80

Displacement [mm]

Load

[N]

tests C01 = 0 C10 = 0

Fig. 25 Tension test curve and comparison and “test” curves f10and f01

2. Weighting of test points if required (e.g. lowweights for test points not trust worthy – for exam-ple Mullins effect on low strain area – or highweights for test points comparable to target loadingasking for high accuracy):

�wi = wi�i

3. Summing up of the square of all differences:

� = �(�wi )2

4. Minimize sum � by adjusting coefficients C10 andC01: In case of Mooney-Rivlin the optimal value α

leads to C10 = α /2G and C01 = (1 − α)/2G.

The following remarks are added here for complete-ness:

• The Neo-Hook material law is based on C10 onlywhich candirectly bederived from the shear tests byC10 = G/2. It should be noted that the Neo-Hooklaw might be applied for higher strain levels thanthe isotropic linear elastic material law but lowerstrain levels compared to Mooney–Rivlin.

• By enriching the test functions with higher orderterms, hyper-elastic strain functions of higher ordercan be derived by the same approach e.g. modellingalso turning points if required for high strain levels.Nevertheless, it should be highlighted that materialmodels should be on the one hand as accurate asrequired in view of the applications but on the otherhand as simple as possible in view of numericalstability. The most complex material model is notalways the best.

• Boundary conditions might be added for the opti-mal coefficients Cij e.g. by asking for Cij ≥ 0.This condition guarantees stability of the mater-ial law by relating higher strains to higher strainenergies as it is known for elastic material as basicprinciple.

123

146 A. Hagl

Fig. 26 Spreadsheet screenshot for determination ofcoefficients forMooney–Rivlin

Test curve

C10 = 0.316 C10 = 0Displacement

[mm]Load [N]

C01 = 0 C01 = 0.316

0 0 0 0 0 05 10.6570 7.6378 6.3729 7.1586 12.2385

10 14.8104 13.4280 9.6148 11.9835 7.991715 17.9409 18.2288 11.4348 15.6551 5.225020 20.5691 22.4482 12.5338 18.6924 3.522025 22.8966 26.3006 13.2356 21.3513 2.388030 25.0269 29.9526 13.7311 23.8076 1.486835 27.0214 33.3696 14.0455 26.0492 0.945040 28.9200 36.6581 14.2708 28.1773 0.551645 30.7507 39.8627 14.4363 30.2307 0.270450 32.5336 42.9904 14.5601 32.2204 0.098155 34.2841 46.0543 14.6545 34.1594 0.015560 36.0140 49.0638 14.7274 36.0565 0.001865 37.7327 52.0256 14.7844 37.9179 0.034370 39.4480 54.9447 14.8295 39.7483 0.090275 41.1660 57.8248 14.8655 41.5510 0.148280 42.8922 60.6687 14.8944 43.3285 0.190385 44.6309 63.4786 14.9177 45.0828 0.204290 46.3860 66.2562 14.9367 46.8153 0.184395 48.1607 69.0030 14.9522 48.5275 0.1345

100 49.9580 71.7199 14.9648 50.2199 0.0686

parameters out of MS solver 9855.518873.02126.0

C10 = 0.1963 0.1197 = C01

summed squared differences

FE Analysis Dataparameterised

values out of FE analysis

suared difference between

messurements and parameterised FE

results

Coefficients for linear combination

Material constants

Numerical forces test function 1

Numerical forces test function 2

Linear combinations of functions 1 and 2

• This approach can also be applied in modifiedform for damage models in order to consider theMullins phenomenon and for visco-elastic modelsin view of strain rate dependencies. Nevertheless,the underlying experimental data shall allow for theidentification of these phenomena.

The identified material model can now be applied tothe analysis of complex bonding designs such as thebonded point support in Fig. 27. The obtained strain andstress distributions within the silicone bonding allowthe assessment of the loading of the bonding in view ofservice and limit load levels.

5 Safety concept exploiting the above mentionedtests and analyses for ensuring adequatebonding performance

The safety concept of the guideline ETAG 002 is basedon quite high safety factors in order to compensate for

simplifications see also (Pröbster 2013). In view of themost relevant load cases the safety factors are definedas follows:

• Short duration tension loading e.g. wind loads: Theallowable stress level is set to 1/6 of the 5% fractilevalue of the fracture strength RU,5/6.

• Short duration shear loading e.g. thermal loads: Theallowable strain level is set to 1/8 of the maximumfracture strains.

• Long duration shear loading e.g. dead loads: Theallowable strain level is set to 1/10 of the short dura-tion value as defined one step before.

The related experimental database refers to H-typespecimens showing differences to dog-bone materialtests with respect to tension loading as discussed inSect. 2. Thus, this approach is affected by the givengeometry of the specimen (e.g. width and length)as failure is triggered by stress concentrations at thespecimen edges, see Fig. 28. For representative two-

123

Development and test logics for structural... 147

Fig. 27 FE Analysis ofbonded point support

0.383

0.355

0.326

0.298

0.269

0.241

0.212

0.184

0.156

0.127

0.0987

0.0703

middle0.189

corner edge 0.296 0.282

Fig. 28 Halfmodel ofETAG002H-type specimenunder tensionloading, max. principal stress distribution (N/mm2)

component structural glazing silicones these require-ments translate to the following numbers for the designstress levels:

• Short duration tension loading: σdes = 0.14 MPa= 0.14 N/mm2

• Short duration shear loading: τdes = 0.11 MPa =0.11 N/mm2

• Long duration shear loading: τ∞ = 0.011 MPa =0.011 N/mm2

For bondings in general, several load casesmight be rel-evant for sizing. In principle, loads can be introducedat discrete attachment points (e.g. point support unitsor bolt connections of PFC) by forces or moments inall three spatial directions as shown in Fig. 29. Typi-cally attachment design is performed in such a way thatmoments e.g. due to force offsets are minimized lead-ing to forces as relevant design parameters. In additionthermal strains shall not be neglected as the differentconstruction materials typically feature different ther-mal expansion. Regarding Fig. 29 tension, compres-sion, lateral and longitudinal shear need to be consid-ered. For a planar point support, this list is reduced to

Tension / Compression

SiliconeAdhesive

PFC (steel)

Side Region of Bonding

Longitudinal Shear

Lateral Shear

Front Region of Bonding

Glass

Fig. 29 U-type bonding and load scenarios

tension, compression and (longitudinal) shear due toaxi-symmetric conditions.

As already mentioned, tension loading of siliconebondings is typically the most critical loading of thebonding material. In contrast, compression loading isnot sizing for the bonding design as the silicone mate-rial is quite robust in view of loading in this direction.The lateral shear is assumed to be stabilised by theregion loaded by compression i.e. by one of the sideregions. Of course, sizing of the PFC needs to with-stand this load case. Regarding longitudinal shear, thebonding is loaded by shear only and material charac-teristics obtained by shear tests can directly be appliede.g. in accordance to the approach given by ETAG 002.This kind of loading leads to a quite soft reaction of thesilicone bonding which is typically desired as thermalstrains can be covered by shear. It should be highlightedin this context that in ETAG 002, bonding thicknessis sized by allowable shear strains evoked by thermaldisplacements. This approach can also be mapped tocomplex bondings which react soft in the related direc-tion. Thus, it is the tension load scheme which asks forspecial treatment for complex bonding geometries.

123

148 A. Hagl

Fig. 30 Maximumprincipal stress distributionsfor point supports at beginof degradation

2.

1.8

1.6

1.4

1.2

1.

0.8

0.6

0.4

0.2

0.

d = 50 mm

d = 70 mm

2.00 N/mm²

1.97 N/mm²

0

0,5

1

1,5

2

2,5

0 5 10 15 20 25 30 35 40

radial coordinate [mm]

stre

ss [N

/mm

²]

d = 70 mm d = 50 mm

2.

1.8

1.6

1.4

1.2

1.

0.8

0.6

0.4

0.2

0.

symmetryaxes

Fig. 31 Maximumprincipal stress distributions for small sampleU-type specimen at begin of degradation

In view of complex bondings such as point supportsand U-type bondings, small sample tension tests pro-vide insight into the load carrying capacities of thesebonding geometries. Figures 30 and 31 show that at thebeginning of degradation, maximum principal stressesof approximately 2 N/mm2 are obtained for the two-component structural glazing silicone. These maxi-mum principal stresses are located in the middle ofthe specimens where encapsulation of the material ismost effective leading to stress maxima in combinationwith the suppression of lateral contraction. It should beadded that the usage of engineering strains is not appro-priate in this case e.g. due to non-uniformity of stressdistributions.

As the small sample specimen tests are typically per-formed under laboratory conditions (i.e. room temper-ature, humidity), additional knowledge with respect tomaterial behaviour is required for safe design. Thus,

knowledgeof the siliconebondingbyconventional test-ing is limited to:

• Temperature: typically room temperature• Aging: typically unaged specimens• Creep: typically short term load histories

Therefore additional tension and shear tests of the sil-icone material were performed in order to investigatethe impact of these environmental conditions on sili-cone performance, see (FHM 2007). Artificial aging isrelated to the following procedure:

• 3 days @ 80 ◦ temperature• 10 days @ 45 ◦C temperature in de-mineralizedwater with cleaning agent (5%)

• 3 days @ 80 ◦ temperature• 10 days @ 45 ◦C temperature in de-mineralizedwater with cleaning agent (5%) plus UV radiation50+/−5W/mm2

• 1 day @ 23 ◦ temperature• 8 days @ 45 ◦C temperature in salted water (50g/lsalt) plus UV radiation 50+/−5W/mm2

• 2 days @ −30 ◦C temperature• 1 day @ 23 ◦C temperature → test performance

Specimens dedicated to creep were conditioned witha tension loading equivalent to 20% maximum strainfor duration of 105 days. Figure 32 shows the depen-dency of fracture loads to temperature, aging and creepof the specimens. The bars plot the mean values whilethe error bars show the variations of the individual sam-ples. The following conclusions can be drawn from thisstudy, see also (Hagl 2008a):

• Higher temperature shows slightly lower limit loadswhile lower temperature leads to an increase ofthese loads.

123

Development and test logics for structural... 149

0

10

20

30

40

50

60

70

- 20°C + 23°C + 80°C afteraging

aftercreep

Tension Tests

Load

[N]

average max / min

Fig. 32 Dog-bone tension tests under varying conditions

0

100

200

300

400

500

600

700

- 20°C + 23°C + 80°C afteraging

aftercreep

Shear Tests

Load

[N]

average max / min

Fig. 33 H-type shear tests under varying conditions

• The results after aging do not significantly differfrom unaged specimens for same temperature i.e.room temperature.

• Creep does not degrade the performance comparingthe results at room temperature.

• The scatter of results i.e. the bandwidth of loads forall investigated conditions is of comparable size.

Figure 33 displays the related results for shear load-ing. The results are consistent to those obtained by ten-sion tests. Thus, the mechanisms acting on tension andshear loading are similar and can be extrapolated tomore complex load schemes. As only strength valuesare shown here, it should be added that concerning stiff-ness, high temperature and aging lead to lower stiffnessvalues according to the load curves of these tests. Thusit is summarized that in view of sizing:

• High temperature is critical in case high strength isrequired

• Ageing is critical in case high stiffness is required

Regarding aggressive environmental conditions actingon the silicone bonding the design of the bondingmightcontribute to high durability in addition e.g. by encap-sulating the bonding by surrounding PFCs and glassedgeswith forU-type designs. By this approach, effectsof moisture and solar radiation are reduced to a highextent postulating significant benefits in view of dura-bility. In addition, bonding designs shall be favoured ofwhich the free surfaces of the bonding are lowly loaded.The length of the PFC sides might be impacted by thisdesign approach for U-type or L-type bonding designs.

Concluding the statements above, the followingapproach is suggested in view of complex siliconebondings:

• Establishment of a hyper-elasticmaterial law on thebasis of material tests

• Hypothesis for beginning damage indicating ser-vice load limitations

• Increased safety factor in regions of relevant envi-ronmental impact

As example, point supports might be designed bythe following approach for the maximum principalstresses using the properties of the above mentionedtwo-component structural glazing silicone:

• In general assuming a safety factor of 6 in accor-dance to the ETAG 002: 2.0 N/mm2/6 = 0.33N/mm2

• In the highly encapsulated middle area the safetyfactor might be lowered to 5: 2.0 N/mm2/5 =0.40 N/mm2

• In the outboard areas potentially impacted by theaggressive environment, the safety factor might beincreased e.g. to 7: 2.0 N/mm2/7 = 0.29 N/mm2

6 Conclusions and outlook

Compared to understanding, testing and modelling ofconventional material used for structural glazing suchas steel and aluminium, no similar basis for silicone asclassical bonding material in such kind of applicationsexist. In order to make the silicone bonding applicableto structural glazing, guidelines were developed suchas ETAG 002 placing the design space of structuralglazing silicone bonding in a quite tight corset. Thiscorset denies advanced bonding designs to be coveredby the guideline and other methodological approachesare required instead. The baseline idea in order to leave

123

150 A. Hagl

Fig. 34 Development andtest logics based on materialand small sample tests

the constraints of guidelines such as ETAG 002 is todevelop a comprehensive silicone material model asit exists already for other materials such as steel. It iswell understood that such a model is highly complex asdifferent phenomena such as hyper-elasticity, Mullinsphenomena and visco-elasticity need to be treated.

In order to step forward in this direction, this paperaddresses the full range from silicone material testscharacterizing the physical properties of silicone, smallsample tests highlighting the impact of boundary condi-tions on failure mechanisms and complementary struc-tural mechanic analysis to a safety concept for ensuringadequate bonding performance.

Potential pitfalls depicted by this test and develop-ment logic exist along several axes:

• In misunderstanding of the behaviour of siliconebondings e.g. the significant stiffening under ten-sion loading evoked by incompressibility in com-bination with lateral contraction

• In material tests not properly performed due toinadequate clamping of specimens into the testingmachine—see Mullin’s effect close to load curveorigin

• In different failure mechanisms of pure siliconespecimens compared to small sample specimenswith adequate boundary conditions

• In material laws derived by simple black boxapproaches of commercial analysis packages andapplied beyond the validation ranges.

The paper demonstrates the added value of finite ele-ment analysis in view of the small sample tests forunderstanding the complex behaviour of the bonding

especially in viewof encapsulation of the siliconemate-rial by metal attachments and glass. Obviously limitstress levels exist under tension loading which arealmost independent from application design (U-typebonding, point support bonding) for the investigatedcases.

Figure 34 presents related development and test log-ics which can be used for complex bonding geometrieswith today’s knowledge.The assessment of the bondingdesign based on a synthesis of tests and finite elementanalyses might indicate that either the silicone materialor the bonding design itself is improper asking for aniteration of the development cycle by selecting anothermaterial or modifying the design. This loop structureis indicated b the dotted arrows.

Of course, significant efforts still need to be spenttowards a comprehensive material model for siliconein view of generality which would allow at least a sig-nificant reduction of test efforts. The future roadmapis seen in a safety concept based on partial factorsaccording to EC0 (DIN EN 1990 2002) with specialconsideration of the hyper-elastic properties of siliconon bondings—work is in progress.

Acknowledgments The author would like to thank DowCorn-ing GmbH, Germany, and especially Mr. Sigurd Sitte, for com-prehensive technical support with respect to the manufacturingof all needed specimens.

Compliance with ethical standards

Conflict of interest On behalf of all authors, the correspondingauthor states that there is no conflict of interest.

123

Development and test logics for structural... 151

References

DIN 53 504: Prüfung von Kautschuk und Elastomeren—Bestimmung von Reißfestigkeit, Zugfestigkeit, Reißdeh-nung und Spannungswerten im Zugversuch (Translation:Testing of rubber—determination of tensile strength atbreak, tensile stress at yield, elongation at break and stressvalues in a tensile test) (2009)

DINEN1990: EC0 -Basis of structural design. Chapter 6: Verifi-cation by the partial factor method: Creation of Load Com-binations (2002)

EOTA: ETAG 002—Guideline for European Technical approvalfor Structural Sealant Glazing systems (SSGS) – Part 1 Sup-ported and unsupported systems (2001)

FachhochschuleMünchen (FHM)FB02:GeklebteVerbindungenim Konstruktiven Glasbau (Translation: Structural bond-ing in glass constructions) Research Report. BMBF (Fed-eral Ministry of Education and Research) Project, AIF-Nr.:1755X04 (2007)

Gent, A.N.: Engineering with Rubber—How to Design RubberComponents, 2nd edn. Carl Hanser, München (2001)

Habenicht, G.: Bonding: Basics, Technology, Applications (inGerman: Original title: Kleben: Grundlagen TechnologieAnwendungen ). Springer, Berlin (1997)

Hagl, A.: Ästhetisch, konstengünstig und innovative – NeueVerbindungstechnik für ganz neue tragende Glasstrukturen(Translation: Aesthetic, cost-efficient and innovative – newconnection method for brand-new load carrying glass struc-tures). Deutsches Ingenieureblatt (2003)

Hagl,A.: Bemessung von strukturellen Silikonklebungen (Trans-lation: Design of structural silicone bondings). StahlbauVol. 76, Issue 8 (2007)

Hagl, A.: Durability by design—new results on load carrying sil-icone bonding. In: 3rd Symposium on durability of build-ing and construction sealants and adhesives, Denver, CO(2008a)

Hagl, A.: Punktuelles Kleben mit Silikonen (Translation: Point-wise bonding with silicones). Stahlbau Vol. 77, Issue 11,Ernst & Sohn Verlag, Berlin, Germany (2008b)

Hagl, A.: Silicone bonded point supports – behaviour undercyclic loading, Engineered Transparency—InternationalConference at glasstec, Düsseldorf, Germany (2010)

Held, B.: Beurteilung von mit Silikonen verklebter Bauteile aufder Basis neuer Erkenntnisse aus der Werkstoffforschung(Translation: Evaluation of bonded building elements basedon new findings of materials research). Diplomarbeit, FHMünchen, (2007)

Hess, Dr R. (Glasconsult): 8142 Uitikon, Swizzerland. “Testingof a U-type-bonding as safety devices of the glass panelsof various glass Soundbarrier Walls along freeways in thevicinity of Zürich” (2006)

Mullins, L.: Effect of stretching on the properties of rubber. Rub-ber Chem. Technol. 21(1948), 281–300 (1948)

Pröbster, M.: Elastisch Kleben (Translation: Elastic bondings),Springer Vieweg Wiesbaden, Germany (2013)

Schadow, T.: Beanspruchungsgerechtes Konstruieren von Kleb-verbindungen in Glastragwerken (Translation: Appropriateconstruction of adhesive bondings by glass structures), Dis-sertation, Technische Universität Dresden, (2006)

Treloar, L.R.G.: The Physics of Rubber Elasticity, 3rd edn.Clarendon Press, Oxford (2005)

Wolf, A.T., Descamps, P.: Determination of Poisson’s ratio ofsilicone sealants from ultrasonic and tensile measurements.In: Johnson, P.G. (ed.) Performance of Exterior BuildingWalls, ASTM STP 1422. American Society for Testing andMaterials, West Conshohocken (2002)

Wolf, A.T., Cleland-Host, H.L.: Material properties for use inFEA modeling: sealant behavior with Ambient LaboratoryClimate Aging. In: Wolf, A.T. (ed.) Durability of Build-ing and Construction Sealants and Adhesives, ASTM STP1453. ASTM International,West Conshohocken, PA (2004)

Wolf, A.T.: Preliminary Evaluation of the mechanical proper-ties and durability of transparent structural silicone adhe-sive (TSSA) for point-fixing in Glazing in preparation forthe derivation of static and dynamic design strength val-ues. In: Fourth Symposium on durability and constructionsealant and adhesives, June 16–17, Ananheim (CA), (2011)

123