UNITISED CURTAIN WALL WITH LOW THERMAL …oa.upm.es/38429/1/BELARMINO_CORDERO_DE_LA_FUENTE.pdf · muro cortina modular con marco de baja transmitancia tÉrmica integrado en el vidrio

Departamento de Construcción y Tecnología Arquitectónicas

Escuela Técnica Superior de Arquitectura

UNITISED CURTAIN WALL WITH LOW THERMAL

TRANSMITTANCE FRAME INTEGRATED WITHIN THE

INSULATING GLASS UNIT THROUGH STRUCTURAL ADHESIVES

MURO CORTINA MODULAR CON MARCO DE BAJA

TRANSMITANCIA TÉRMICA INTEGRADO EN EL VIDRIO

AISLANTE A TRAVÉS DE ADHESIVOS ESTRUCTURALES

TESIS DOCTORAL

Belarmino Cordero de la Fuente

Arquitecto por la Universidad Politécnica de Madrid

2015

Departamento de Construcción y Tecnología Arquitectónicas

Escuela Técnica Superior de Arquitectura

UNITISED CURTAIN WALL WITH LOW THERMAL

TRANSMITTANCE FRAME INTEGRATED WITHIN THE

INSULATING GLASS UNIT THROUGH STRUCTURAL ADHESIVES

MURO CORTINA MODULAR CON MARCO DE BAJA

TRANSMITANCIA TÉRMICA INTEGRADO EN EL VIDRIO

AISLANTE A TRAVÉS DE ADHESIVOS ESTRUCTURALES

Autor

Belarmino Cordero de la Fuente

Arquitecto por la Universidad Politécnica de Madrid

Director

Dr. Alfonso García Santos

Doctor Arquitecto por la Universidad Politécnica de Madrid

2015

Department of Construction and Technology in Architecture Unitised curtain wall with low thermal transmittance frame integrated within the insulating glass unit through structural adhesives

i

ACKNOWLEDGMENTS

From the Escuela Técnica Superior de Arquitectura de Madrid, I would like to thank

my supervisor Dr. Alfonso García Santos for giving me the opportunity to carry out

this work and for his encouragement and guidance. I am also grateful to the rest of staff

and my course friends. I would like to thank the Glass and Façade Technology research

group from the University of Cambridge, who have actively collaborated in the

chapters related to structural engineering. I am also thankful to the sponsors and

industrial partners that have helped with their financial support or have provided

materials for testing: Engineering and Physical Sciences Research Council (EPSRC);

Institution of Structural Engineers (IStructE); Fiberline; Excel Composites; Dow

Corning and Pilkington. Finally, I owe special thanks to my family and to my wife for

their continuing support.


ii


iii

CONTENTS

ACKNOWLEDGMENTS ............................................................................................. i

CONTENTS.................................................................................................................. iii

NOMENCLATURE .................................................................................................... vii

ABSTRACT (English) ................................................................................................. xi

ABSTRACT (Spanish)............................................................................................... xiii

INTRODUCTION .................................................................................................. 15

1.1 Background to insulated glass units (IGUs) ..................................................... 17

1.2 Background to curtain walling: stick and unitised systems .............................. 17

1.3 Curtain wall market trend ................................................................................. 20

1.4 Issues with conventional unitised curtain walls ............................................... 21

1.5 Description of proposed research ..................................................................... 22

1.6 Research hypothesis and objectives ................................................................. 22

1.7 Methodology and description of research tasks ............................................... 25

STATE OF THE ART ........................................................................................... 29

2.1 Pultruded GFRP ................................................................................................ 31

2.2 Adhesive connections ....................................................................................... 31

2.3 Composite structural action .............................................................................. 46

2.4 Thermal transmission ....................................................................................... 47

2.5 Analysis of similar existing products ............................................................... 48

2.6 Conclusion ........................................................................................................ 55

SCHEMATIC DESIGN DESCRIPTION AND COMPARISON WITH

CONVENTIONAL SYSTEM ............................................................................... 57

3.1 Manufacturing process and supply chain ......................................................... 59

3.2 Support condition and glass replacement strategy ........................................... 59

3.3 Sealing and drainage strategy ........................................................................... 62


iv

3.4 Fire performance and fire partition strategy..................................................... 63

3.5 Acoustic performance and strategies to limit flanking ..................................... 68

SCHEMATIC DESIGN STRUCTURAL ASSESSMENT BY

ANALYTICAL CALCULATION ........................................................................ 71

4.1 Method ............................................................................................................. 73

4.2 Results and discussion ...................................................................................... 83

4.3 Conclusion ........................................................................................................ 86

SCHEMATIC DESIGN THERMAL ASSESSMENT BY

NUMERICAL CALCULATION ......................................................................... 89

5.1 Method ............................................................................................................. 91

5.2 Results and discussion ...................................................................................... 98

5.3 Conclusion ...................................................................................................... 102

GFRP FRAME SELECTION BY 4-POINT BENDING TESTS .................... 103

6.1 Candidate materials ........................................................................................ 105

6.2 Method ........................................................................................................... 106

6.3 Results and Discussion ................................................................................... 110

6.4 Conclusion ...................................................................................................... 114

ADHESIVE SELECTION BY SINGLE-LAP SHEAR TESTS ..................... 117

7.1 Preliminary selection of candidate adhesives ................................................. 119

7.2 Method ........................................................................................................... 123

7.3 Results and discussion .................................................................................... 132

7.4 Conclusion ...................................................................................................... 143

DETAIL DESIGN DESCRIPTION ................................................................... 145

8.1 Design changes ............................................................................................... 147

DETAIL DESIGN STRUCTURAL ASSESSMENT BY NUMERICAL

CALCULATION ................................................................................................. 149

9.1 Method ........................................................................................................... 151


9.3 Conclusion ...................................................................................................... 160


v

DETAIL DESIGN THERMAL ASSESSMENT BY NUMERICAL

CALCULATION .................................................................................................. 161

10.1 Method ............................................................................................................ 163


10.3 Conclusion ...................................................................................................... 168

CONCLUSION AND FUTURE WORK ......................................................... 169

11.1 Conclusion ...................................................................................................... 171

11.2 Future work ..................................................................................................... 173

RELEVANT PUBLICATIONS / AWARDS .......................................................... 175

REFERENCES AND BIBLIOGRAPHY ............................................................... 177


vi


vii

NOMENCLATURE

Latin symbols

A Cross-sectional area of the structural element mm2

Ac Area of the segment above the cut line mm2

a Breadth of the section at the cut line being considered mm

Ac Centre-of-glazing area m2

Ae Edge-of-glazing area m2

Af Frame area m2

At Total area m2

E Modulus of elasticity GPa

E(t) Time dependent modulus of elasticity GPa

E0 Modulus of elasticity for time = 0 GPa

E∞ Modulus of elasticity for time = ∞ GPa

fb Limiting stress in bending MPa

fv Limiting stress in shear MPa

Gadhesive Shear modulus of the adhesive MPa

g-value Solar heat gain coefficient -

I Second moment of area mm4

L Length of curtain wall unit mm

l Span length mm

M Applied moment Nm

Mmax Maximum bending moment Nm

P Point load N

R Radius of curvature mm

R-squared Coefficient of determination -

ΔT Variation in temperature K

Ut Total U-value W/m2K

Uf Frame U-value W/m2K

Ue Edge-of-glazing U-value W/m2K


viii

Uc Centre-of-glazing U-value W/m2K

U-value Heat transfer coefficient W/m2K

V Applied shear N

Vmax Maximum shear force N

w Linear uniform load N/m

y Distance from most extreme fibre to neutral axis mm

y’ Distance from centre of area above the cut line to centroid of whole section

mm

Z Section modulus mm3

Greek symbols

αglass Coefficient of thermal expansion of glass K-1

ΑGFRP Coefficient of thermal expansion of GFRP K-1

γm Material safety factor -

ε Emissivity -

ƍmax Maximum deflection mm

λ Thermal conductivity W/mK

ρ Density kg/m3

σ Bending stress MPa

τaverage Average shear stress MPa

τbeam Beam shear stress MPa

τt Shear stress caused by differential thermal expansion MPa

Abbreviations

2D Bi-Dimensional

3D Tri-Dimensional

BMU Building Maintenance Unit


ix

CO2 Carbon Dioxide

FEA Finite Elements Analysis

IGU Insulating glazing unit

GFRP Glass Fibre Reinforced Polymer

PTFE Polytetrafluoroethylene

PVB Polyvinyl butyral

TSSA Transparent Structural Silicone Adhesive

UV Ultra Violet


x


xi

ABSTRACT (English)

Unitised curtain wall systems consist of pre manufactured cladding panels which can

be fitted to the building via pre fixed brackets along the edge of the floor slab. They are

universally used for high rise buildings because the factory controlled assembly of

units ensures high quality and allows fast installation without external access.

However, its frame is structurally over-dimensioned because it is designed to carry the

full structural load, failing to take advantage of potential composite contribution of

glass. Subsequently, it is unnecessarily deep, occupying valuable space, and protrudes

to the inside, causing visual disruption. Moreover, it is generally made of high thermal

conductivity metal alloys, contributing to substantial thermal transmission at joints.

This research aims to develop a novel frame-integrated unitised curtain wall system

that will reduce thermal transmission at joints, reduce structural depth significantly and

allow an inside flush finish. The idea is to adhesively bond a Fibre Reinforced Polymer

(FRP) frame to the edge of the Insulated Glass Unit (IGU), thereby achieving

composite structural behaviour and low thermal transmittance. The frame is to fit

within the glazing cavity depth.

Preliminary analytical structural and numerical thermal calculations are carried out to

assess the performance of an initial schematic design. 4-point bending tests on GFRP

and single-lap shear tests on bonded connections between GFRP and glass are

performed to inform the frame and adhesive material selection process and to

characterise these materials. Based on the preliminary calculations and experimental

tests, some changes are put into effect to improve the performance of the system and

mitigate potential issues. Structural and thermal numerical analysis carried out on the

final detail design confirm a reduction of the structural depth to almost one fifth and a

reduction of thermal transmission of 6% compared to a benchmark conventional

system. A flush glazed appearance both to the inside and the outside are provided

while keeping the full functionality of a unitised system.


xii


xiii

ABSTRACT (Spanish)

Los muros cortina modulares están constituidos por paneles prefabricados que se fijan

al edificio a través de anclajes a lo largo del borde del forjado. El proceso de

prefabricación garantiza buena calidad y control de los acabados y el proceso de

instalación es rápido y no requiere andamiaje. Por estas razones su uso está muy

extendido en torres. Sin embargo, el diseño de los marcos de aluminio podría ser más

eficiente si se aprovechara la rigidez de los vidrios para reducir la profundidad

estructural de los montantes. Asimismo, se podrían reducir los puentes térmicos en las

juntas si se sustituyeran los marcos por materiales de menor conductividad térmica que

el aluminio.

Esta investigación persigue desarrollar un muro cortina alternativo que reduzca la

profundidad estructural, reduzca la transmisión térmica en las juntas y permita un

acabado enrasado al interior, sin que sobresalgan los montantes. La idea consiste en

conectar un marco de material compuesto de fibra de vidrio a lo largo del borde del

vidrio aislante a través de adhesivos estructurales para así movilizar una acción

estructural compuesta entre los dos vidrios y lograr una baja transmitancia térmica. El

marco ha de estar integrado en la profundidad del vidrio aislante.

En una primera fase se han efectuado cálculos estructurales y térmicos preliminares

para evaluar las prestaciones a un nivel esquemático. Además, se han realizado

ensayos a flexión en materiales compuestos de fibra de vidrio y ensayos a cortante en

las conexiones adhesivas entre vidrio y material compuesto. Con la información

obtenida se ha seleccionado el material del marco y del adhesivo y se han efectuado

cambios sobre el diseño original. Los análisis numéricos finales demuestran una

reducción de la profundidad estructural de un 80% y una reducción de la transmisión

térmica de un 6% en comparación con un sistema convencional tomado como

referencia. El sistema propuesto permite obtener acabados enrasados.


xiv


15

INTRODUCTION

1. Introduction

2. State of the art

3. Schematic Design description

by comparison with conventional system

4. Schematic Design structural assessment

by analytical calculation

5. Schematic Design thermal assessment

by numerical calculation

6. GFRP frame selection

by 4-point bending tests

7. Adhesive selection

by single-lap shear tests

8. Detail Design description

9. Detail Design structural assessment


10. Detail Design thermal assessment


11. Conclusion and future works

SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


16


17

1.1 Background to insulated glass units (IGUs)

An IGU is an assembly consisting of at least two panes of glass, separated by one or more

spacers, hermetically sealed along the periphery, mechanically stable and durable (BS EN

1279-1, 2004). Compared to single glazing, the use of double glazing primarily reduces

energy transmission into and out of a building, but it can also reduce internal

condensation, improve thermal comfort and reduce noise transmission. The design and

manufacturing of the sealing along the edge of the unit determines its durability, the

extent of thermal bridging through the edge and the proportion of composite structural

action between the spacer and the glass panes. To ensure durability, the sealing has to

provide low moisture vapour transmission, guarantee material compatibility, have good

resistance to water, temperature changes and ultraviolet radiation and be sufficiently

flexible to accommodate differential thermal expansion between the glass panes and the

spacer and bowing caused by pressure variations (CWCT, 2010).

Figure 1: Typical insulating glass unit

1.2 Background to curtain walling: stick and unitised systems

Curtain walls are non-load bearing façade systems that hang from the structure of a

building. They comprise a supporting grid, generally made of metal profiles, and infill

panels, made of glass or other cladding materials. They have been widely used from the

Glass panes

Cavity

Edge spacer and seal


18

early 1970s due to their lightweight nature, simplification of temporary construction and

strong performance. They are classified into two main types: stick and unitised.

In stick systems, the components are assembled onsite, with individual mullions and rails

forming a supporting grid for curtain wall panels. The joints between adjacent units are

typically sealed during construction of the curtain wall by on-site application of wet

sealants to seal the gap between units. This requires external access to the curtain

wall/building during construction which reduces the speed of installation. Further, wet

sealants may not provide a consistently high-quality seal as their application relies upon

the standard of on-site work and so may vary.

Figure 2: Stick curtain wall (a) aluminium supporting grid fixed to the building slab (b) infill panels fixed to the supporting grid on site (c) schematic cross-section of glass panels fixed to aluminium

mullion

(a)

(b)

(c)


19

Unitised curtain walls consist of cladding units where panel and frame are pre-assembled

in factory and then easily transported and fitted to the building. The units normally span

from floor to floor hanging from pre-fixed brackets along the edge of the upper floor slab

and being horizontally restraint by the units below. The joints need to accommodate in-

plane differential movement between units while providing weather tightness. This is

resolved by introducing open grooves and overlapping gaskets along the perimeter of the

units that form pressure equalised and drained cavities between units once installed. On-

site application of wet sealants to seal the gap between units is thereby avoided. As a

result, external access is not required, higher quality control and speed of installation are

achieved and larger in-plane differential movement between units can be accommodated.

For these reasons, unitised curtain walls are the façade system of choice for high rise

buildings

Figure 3: Unitsed curtain wall (a) factory preassembly of glass panel and frame (b) preassembled units delivered on site (c) installation of preassembled unit (d) schematic cross-section of connection

between two preassembled units

(b)

(a)

(c)

(d)


20

1.3 Curtain wall market trend

Market demand has been shifting slowly from stick-built to unitized structures, which

rely less on skilled labour, provide more consistent quality, and are more capable of

offering advanced technological products. The following chart illustrates the market

segmentation of the global curtain wall industry by product type for the periods specified.

\

Figure 4: Market segmentation of the global curtain wall industry by product type for the periods specified. (Synovate Report)

The market share of unitized curtain wall structures in the total market increased from

approximately 48.7% in 2005 to 50.7% in 2009 and is expected to increase to

approximately 53.1% in 2012. On the other hand, the market share of stick-built curtain

wall structures decreased from approximately 30.9% in 2005 to approximately 28.7% in

2009 and is expected to further decrease to approximately 26.7% in 2012.

Historically, unitised systems have been selected for large commercial developments,

including high-rise, whereas stick systems are more synonymous with smaller and low-

rise schemes. There was a time when unitised systems were only selected for commercial

office developments. However the curtain walling market has evolved over the last ten

years and reached a level of maturity where unitised curtain walling is much more widely

available through an increasing number of sources.

The introduction of proprietary unitised curtain walling systems in the market creates

more opportunities for projects to benefit from the off-site prefabricated approach,

especially those projects where the façade area would not normally have been


21

commercially viable for a unitised approach or/and projects where the overall programme

for procurement would not have been sufficient to accommodate the lead-in period for

unitised curtain walling.

1.4 Issues with conventional unitised curtain walls

Edge of glazing and framing of conventional unitised curtain wall systems is inherently

inefficient, both thermally and structurally. It is based on the use of metal alloys with high

thermal conductivities, thereby leading to substantial thermal transmission at joints. The

thermal inefficiency has only recently come to the fore as the thermal performance of

glass units has steadily increased, so that the thermal performance of contemporary

curtain walls is governed by the edge-of-glazing and framing regions. Moreover, the

glazing spacers and curtain wall frames are structurally inefficient, and therefore over

dimensioned, as they fail to exploit the potential composite action with the glass panels.

This structural inefficiency also leads to space planning problems and aesthetic

weaknesses as the frames occupy valuable space, and protrude into buildings, causing

visual disruption.

Figure 5: (a) EN ISO 10077 Part 2 thermal transfer equation through curtain wall and

(b)thermographic image showing thermal bridging at joints (CWCT TN 49, 2007)

(a)

(b)


22

1.5 Description of proposed research This research addresses the thermal and structural inefficiency of edge of glazing and framing of conventional unitised curtain wall

systems through:

The utilisation of low thermal conductivity materials for frames to achieve low

thermal transmittance at joints;

The utilisation of structural adhesives between the glass and the frame to mobilise

composite structural action;

The integration of these technologies in the design of a novel high performance

unitised curtain wall.

The idea is to adhesively bond a Fibre Reinforced Polymer (FRP) frame to the edge of an

Insulated Glass Unit (IGU), thereby achieving composite structural behaviour and low

thermal transmittance. The frame is to fit within the glazing cavity depth. To illustrate the

differences with a conventional system, a conventional system and the proposed system

are represented one next to the other in figures 6 and 7. The internal views are compared

in figures 8a and 8b.

1.6 Research hypothesis and objectives

It is possible to demonstrate that the proposed system improves the structural and thermal

performance and the appearance of conventional unitised curtain walls. The following

measureable objectives have been set against conventional systems:

1. Reduction in structural depth through increased efficiency and, therefore,

increase of the available internal net floor area. Considering the high cost of space

in tall buildings, it would represent an important financial asset for developers;

2. Reduction of thermal transmission. Developers would be reassured that in the

future tall buildings will continue to meet increasingly stringent legislation

regarding energy performance and building owners would reduce their energy

bills.

3. Improved aesthetics. Achieving a seemingly frameless unitised curtain wall

would be possible by providing a flush glazed appearance both to the inside and

the outside while keeping the full functionality of a unitised system.


23

Figure 6: Conventional system schematic cross-section through mullion

Figure 7: Proposed system schematic cross-section thorugh mullion

Insulating glass unit

Spacer

Structural silicone secondary seal

Thermal break

Sealant

Structural silicone glass retention

Pressure equalised cavity

Gaskets

Aluminium frame

Gaskets

Structural glass

Spacer

Pressure-equalised cavity Structural adhesive

GFRP frame

Sealant


24

Figure 8: Visual appearance comparison between (a) conventional and (b) proposed systems

(b)

Frame protruding to the internal space

(a)

Flush glazed appearance to the internal space


25

1.7 Methodology and description of research tasks

In order to achieve these objectives the research has involved a series of cross-

disciplinary and multi-scale investigations. These investigations have comprised

theoretical, numerical and experimental techniques, ranging from micro-scale

investigations (e.g. to analyse the bonded interface surface), to numerical simulations at

panel level (e.g. to assess the reduction in energy transmission). This thesis is divided into

11 chapters; the introduction being chapter 1 and the conclusions and future work being

presented in chapter 11. The content of the main chapters (chapter 2 to chapter 10) are

briefly outlined below:

Chapter 2: State of the art

Before devising a novel curtain wall it seems pertinent to understand how existing

façade technologies have addressed thermal transmission, structural efficiency and

aesthetics. The main aspects that have been investigated are the utilisation of low

thermal conductivity materials and the utilisation of structural adhesives between

the glass and the frame. Similar products to the one proposed have been reviewed

and analysed.

Chapter 3: Schematic Design description and comparison with conventional

system

The proposed system is described and compared to a conventional system taken as

reference at a schematic level. Besides the structural and thermal performance,

which are assessed in depth in other chapters, the principal design issues and the

strategies to address them are reviewed for conventional unitised curtain wall

systems including: manufacturing process and supply chain; support condition,

installation process and glass replacement strategy; sealing and drainage

strategies; fire performance and fire partition strategy; acoustic performance and

strategies to limit flanking. The alternative strategies adopted by the proposed

system are then outlined.

Chapter 4: Schematic Design structural assessment by analytical calculation

Deflection, moment stress and shear stress induced in the proposed system by

wind load are predicted by means of simple bending theory (Euler-Bernouilli).


26

The results are compared with the performance of a conventional system taken as

reference. Sensitivity analysis is carried out to observe how varying the structural

depth of the system affects the stiffness and the bending stresses and how the

shear stresses at the adhesive-glass interface or the GFRP web influence the

design.

Chapter 5: Schematic Design thermal assessment by numerical calculation

Thermal transmittance and risk of condensation of the frame-integrated system are

assessed through comparative analytical and numerical thermal analysis with a

conventional system taken as reference.

Chapter 6: GFRP frame selection by 4-point bending tests

Four-point bending tests are performed on glass fibre reinforced polyester resin

and glass fibre reinforced phenolic resin specimens, some of the specimens being

previously heat soaked. The results provide information on the shear strength and

the time-dependent modulus of elasticity of the tested materials.

Chapter 7: Adhesive selection by single-lap shear tests

Single lap shear tests are performed on bonded connections between glass and

glass fibre reinforced polyester substrates using a range of candidate adhesives.

Two phases of testing take place with an intermediary analysis and adjustments in

the design of the connections. The results provide information on the failure mode

and a typical load versus shear displacement curve for each of the candidate

adhesives.

Chapter 8: Detail Design description

Based on the structural and thermal calculations on the schematic design and on

the results of the experimental tests, some changes are put into effect to improve

the performance of the system and mitigate potential issues.

Chapter 9: Detail Design structural assessment by numerical calculation

Numerical analysis has been carried out on the detail design to take into account

effects that were not considered in the initial analytical calculations such as shear


27

deformations, shear lag effect or time dependent properties of the GFRP and the

adhesive. A short term and a long tern load cases have been established to better

represent wind loading and to investigate if modelling the GFRP and the adhesive

with different Modulus of Elasticity affects the results. The maximum deflection

at edge of IGU, maximum tensile stress at glass and maximum shear stress at

adhesive and GFRP have been quantified.

Chapter 10: Detail Design thermal assessment by numerical calculation

Thermal transmittance and risk of condensation of the detail design are assessed

through comparative analytical and numerical thermal analysis with the

conventional system and the schematic design.


28


29

STATE OF THE ART

The main aspects that have been investigated are pultruded GFRP; adhesive connections,

thermal transmission and composite structural action. Similar existing products have been

analysed.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


30


31

2.1 Pultruded GFRP

2.2 Adhesive connections

Adhesive bonding is the process of binding materials (called adherends or substrates)

together using any number of adhesive substances (called adhesives) which provide

sufficient strength and sufficient bonding to both substrates. In contrast to bolted

connections, adhesive connections can offer numerous advantages: In many cases, they

yield a quasi-uniform stress distribution in the substrates, often avoiding unfavourable

peak stresses in the glass, which after all remains a brittle material. Another significant

advantage is that boreholes are not needed. This is benefitial because they typically lead

to locally reduced glass strength and provoke thermal bridges. They can also connect very

thin elements avoiding the effect of bearing damage around bolts. Finally, from an

aesthetic point of view, bonded connections can be less apparent than bolted connections.

In most cases adhesives are synthetic materials which are characterised by a rather

complex material behaviour. Properties of adhesive bonds are influenced by a number of

circumstances which are hard to analyse, such as practical application conditions or

exposure to aggressive environments.

The detailed design of adhesive connections is not an easy task, and to date it is common

practice to work with significant safety margins. In addition, standard details are often not

yet available, which is why the application and design of adhesive bonds requires a

certain level of specialisation. Nonetheless, for certain applications adhesive connections

can provide extremely good solutions. However, an important condition for successful

application of this technology in the building industry is that sufficient attention should be

paid to the design and quality control of the bond.

2.2.1 Composition and classification

In terms of chemical properties, all commonly used adhesives are polymers. They mostly

consist of atoms of carbon, hydrogen and oxygen, and often nitrogen, chlorine and other

chemical elements form a part of the compound as well. One specific property of

polymers in comparison with other materials is the chain structure of their molecules.

This chain structure consists of the connection of the monomer units and is known as the

macromolecule. Based on its thermo-mechanical behaviour, polymers can be classified

into different groups:


32

Thermoplastics: They are solid at room temperature but become liquid when

heated. During the cooling process they solidify again. This procedure can be

repeated several times, but the properties of thermoplastics are affected by every

heat and cool process. Cyano-acrylates and PVB belong to this group.

Thermosets: They become plastic when they are processed for the first time, by

further heating they are cured by chemical reaction and this state is final. If this

procedure is repeated there are no more chemical changes. In case of over-heating

they can degrade and carbonize. Typical members of this group are epoxies and 2-

component polyurethanes.

Elastomers: Silicones are the most commonly used. They are very flexible in a

very high range of temperature. They are able to achieve large strains without any

change in macroscopic volume. The relation between stress and strain is

significantly non-linear due to progressive straightening of polymer chains in the

direction of the increasing strain. This phenomenon is reversible after unloading.

2.2.2 Adhesion and cohesion

Adhesion between an adhesive and a substrate is mainly based on a combination of three

principles, namely mechanical interlocking, diffusion and adsorption. Mechanical

interlocking occurs when the substrate is porous and the pores are filled with adhesive

material Diffusion is a bond on molecular level, a chemical reaction between molecules

of the adhesive and those of the substrate. Adsorption is a bonding mechanism due to

intermolecular forces working at the interface and causing a chemical bond.

Cohesion refers to intra- and intermolecular attraction between similar molecules (i.e. the

binding of the adhesive with itself).

2.2.3 Failure mechanisms

Failure of the connection can be by any of the following:

Substrate failure: Collapse of the glass member due to locally exceeding shear or

tensile strength of the substrate. Usually this case is considered to be favourable,

because the strength of the adhesive will not be the governing factor for the design

of the connection and better-known strength values can be used


33

Cohesive failure: Failure of the adhesive layer as a result of exceeding shear or

tensile strength of the adhesive itself.

Adhesive failure: Slippage or ripping of the adhesive layer from one of the

substrates due to insufficient adhesion to the substrate.

2.2.4 Mechanical actions on adhesives

The mechanical actions on adhesives are classified as below:

Compressive loads: They usually cause little or no problems for most adhesive

connections of common building materials. However, when dealing with stiff

substrates, it is extremely important that the adhesive is neither too compressible

nor too squeezable to avoid direct contact between both substrates, as this may

lead to large local stress concentrations and possibly even fracture of the glass.

Tensile loads: They seem theoretically very favourable for many adhesives if

applied centrically. However, in reality most loads will be eccentric, causing

additional bending moments and according peak stresses which may drastically

reduce the theoretical resistance of the connection. For this reason, normal tensile

actions on adhesive bonds are not ideal. However, in many cases normal tensile

forces cannot be avoided; still, acceptable levels of resistance can usually be

obtained if eccentricities are limited.

Peel stresses: They occur when tensile bending forces act perpendicularly to the

adhesive bonding surface, for instance due to highly eccentric tensile actions.

Consequently, very high peak stresses may appear near the edges. The major

influence factors determining the magnitude of these peak stresses are the

toughness and thickness of the adhesive, the overlap length, and the stiffness of

the construction. This type of loading should be avoided.

Thermal stresses: They will occur when two different materials are joined due

their different coefficients of thermal expansion. Generally, it will be more critical

if the stiffness of the adhesive is increased as the ability to deform will be limited.

A reduction of the thickness of the joint will have the same effect in most cases.


34

2.2.5 Stress distribution

The strain in the adhesive layer is larger near the edges especially on stiff connections, as

adhesives with a higher stiffness don‘t have a possibility to redistribute stresses within the

material itself. The stress distribution in the adhesive layer along the stiff joint is not

uniform and there is a tendency of stress peaks near the edges of glued elements.

Consequently, an increase overlap length will not necessarily result in a stronger

connection in the case of stiffer adhesives. Namely, from a certain overlap length

onwards, stresses near the mid-section of the connection will be virtually reduced to zero,

whereas almost full load transfer will take place close to the ends.

On the other hand, where a more flexible adhesive is used in a higher thickness, the stress

distribution can be more or less assumed as uniform.

2.2.6 Ageing

In general, ageing can lead to a significant decrease of stiffness and strength of adhesives.

Therefore it has to be investigated if the remaining strength of an aged adhesive is high

enough to carry the resulting shear stresses of the adhesive. Additionally it has to be

checked if the reduced stiffness of the adhesive due to ageing will lead to an increase in

local stresses in the glass, eventually leading to glass breakage. Regular monitoring is

required to identify signals of a possible damage in the adhesive (cracks, scratches, gaps,

bleeding, colour change, bubbles, traces of water, etc.) and reduce the risk of sudden

collapse by failure of adhesive connections. It is best to combine the periodical

inspections with the regular cleaning intervals. In external applications, the following

parameters may contribute to ageing:

Solar radiation including UV: UV-radiation is one of the main causes of damage

to organic materials. For structural glass connections it is important to choose UV-

resistant adhesives, because UV-radiation goes through the glass and can degrade

the adhesive. It can lead to damage of adhesive forces between glass and glue. It is

necessary to protect the adhesive by applying a coating on the glass if the UV-

resistance of the adhesive does not suffice.

Temperature variation: The external temperature range can range from -20°C to

+80°C. Thermal resistance of a polymer adhesive depends on the glass transition

temperature Tg. This temperature is a way to understand the molecular motion

that occurs in polymeric material. The degree of molecular motion affects the


35

adhesive and cohesive forces, the structure of the polymer chain, the degree of

cross-linking, the molecular weight, the brittleness and other polymer properties.

At temperatures less than Tg, a polymer behaves like a solid material in which the

molecular segments have moderate and independent motions. If the temperature

of polymer is increased, molecules become more flexible and mobile. Transition

of a polymer from glassy to rubbery state signifies that temperature is close to Tg.

If the temperature is raised above Tg, the distance between molecular segments is

increased and it is accompanied by increasing the specific volume of the

polymeric material. The glass transition temperature should be above the upper

service temperature for high bond strength values. Elastomers have usually a low

Tg (below 0°C). This leads to a low modulus of elasticity, low tensile and shear

strength and a high elongation at break. Rigid adhesives have usually a high Tg to

ensure high strength values during common temperatures, but they lose their

stiffness and strength if the temperature increases above their glass transition

temperature. Furthermore, adhesive forces between substrate and glue are reduced

at high temperatures, which can cause adhesive failure. However, when the

temperature decreases, this will cause increasing stiffness of the bonded

connection. At low temperature the joint will be prone to cohesive failure as a

result of the higher brittleness of the adhesive. During repeated temperature

changes adhesive layer has to be flexible enough to equalize different thermal

elongations of different joining materials. This can be achieved by using a flexible

and durable adhesive with an optimal layer thickness.

Humidity variation: Some adhesives absorb to a certain degree environmental

moisture or water and this causes swelling of the adhesive. On the other hand,

with decreasing relative humidity moisture can migrate out of the polymer, which

causes a volume decrease. The repetition of this process leads to a decrease in

adhesion. Absorbed moisture in the polymer can migrate to the interface between

the adhesive and the substrate and can accumulate in micro-cavities. This has also

a negative effect on the adhesion and can subsequently lead to adhesive failure.

Furthermore, water in combination with heat often leads to hydrolysis, a

phenomenon causing changes in the macromolecular structure of the polymer.

This leads to changes in the material properties of the polymer.


36

Repeated loading: When subjected to static actions, many adhesives will suffer

from creep or static fatigue, causing continued deformations without change in

load. Consequently, adhesive connections may fail even if the theoretical maximal

stresses are not reached. Also dynamic actions may cause fatigue. However, in

function of the specific products applied and compared to bolted connections,

adhesive connections will often be better resistant against dynamic fatigue.

Execution and setup imperfections: In addition, it is noteworthy that most

adhesives are relatively sensitive to execution and setup imperfections: poorly

executed adhesive bonds may have a significantly decreased lifetime or a lower

resistance.

2.2.7 Fire safety

An unprotected bonded connection will typically lose its strength and stiffness very early

in case of fire. In general, higher temperature enhances the molecular mobility and

therefore reduces the cohesive strength. Heating up to high temperatures first changes the

dimensional stability, then the chemical stability and it eventually leads to total

decomposition. The reduction of dimensional stability means that the adhesive joint

deforms largely without elastic spring back and starts to creep considerably under static

or dead load. The degradation of chemical stability is time-dependent (in most cases not

temperature-dependent) and goes along with a chemical reaction, e.g. oxidation or

cleavage. Here are some general regards on the behaviour of the adhesive in case of fire:

The majority of adhesives used for application in construction industry are mainly

organic; therefore most of them decompose at 120 to 150 °C. Thermoplastics are

earlier affected by higher temperatures or fire than adhesives with thermosetting

structure, which normally offer temperature resistances up to 200 °C. Silicones

sometimes offer maximum temperatures of 200 °C.

The duration of temperature or fire loads is essential. Sometimes a short loading

duration with high temperatures below the point of chemical decomposition and

flammability will not go along with a complete and sudden loss of carrying

capacity.

A very important factor is the degree of thermal conductivity of the substrates. A

high thermal conductivity of the substrate could lead to heat accumulation,


37

increasing the temperature in the substrate and the thermal stress of the adhesive

significantly.

Another important issue is the different thermal expansion of substrate and glass,

which can lead to great temperature stresses in the adhesive and can become

decisive at higher temperatures.

2.2.8 Types of adhesives

Scientific and industrial classifications are usually based on the curing method. As such,

two main categories can be distinguished: adhesives curing by means of a physical

process, or by means of a chemical reaction:

1. Adhesives curing by means of a physical process are in general not relevant for

structural applications. Curing takes place by evaporation of a solvent or by

solidifying; processes which usually are relatively time consuming. Other

disadvantages are relatively low strength, short lifetime and limited temperature

range. Examples are typical household glues such as cyano-acrylates or wood glue

and some polyurethanes.

2. Adhesives curing by means of a chemical reaction are better suited for structural

applications. These adhesives come typically in a liquid or viscous state and

solidify after a chemical reaction. This reaction can take place either between two

components that are part of the adhesive or between the adhesive and for instance

the humidity in the air. Examples are acrylates, epoxies, silicones, polyurethanes,

MS-polymers, etc.

Adhesive interlayers or laminates constitute a special case in such a classification. At

room temperature, these products are in solid estate and become liquid as a result of to an

increase of temperature or pressure (which is a physical process). A chemical reaction

follows enabling sufficient bonding to the substrates. Finally they solidify again while

cooling down. Examples are polyvinyl butyral (PVB) or SentryGlas® (SG).

2.2.8.1 Acrylates

Three different types of acrylate adhesives exist. Firstly, there is the family of

methacrylates, being two-component adhesives. Secondly and thirdly, there are


38

cyanoacrylates and UV-curing acrylates, respectively, both curing by a one-component

system.

As cyanoacrylates will usually not be used in structural applications, only methacyrlates

and UV-curing acrylates are described. Methacrylates are usually more durable than UV-

curing acrylates. The main advantages of the latter are transparency and swiftness of

execution. Both acrylate types are suitable to create adhesive connections characterised

by a high shear resistance; i.e. up to 20 to 30 MPa (without any safety coefficient).

Furthermore, it is known that UV resistance and creep behaviour is in most cases very

acceptable. They are adhesives that were initially designed to be applied in a very thin

layer. The recommended application thickness is typically below 0,5 mm. Consequently,

possibilities to cope with building tolerances or differential thermal expansions are

extremely limited. Therefore, a bigger thickness is usually needed. By varying the

thickness, shear stiffness and maximum possible elongation of the joint can be adjusted

but the ultimate load bearing capacity decreases. The behaviour of UV curing acrylates is

strongly dependent on temperature, which makes them poorly resistant against large,

quick or long-term temperature changes. In addition, resistance against moisture can be

problematic: the strength of most acrylates will be reduced after exposure to humidity.

Finally, the brittle nature of this type of adhesives is important for structural applications,

as they will give way all of a sudden and without significant prior deformation. This is

mainly problematic for UV-curing adhesives As peak stresses in the adhesive joint are

governing the design, the geometry should be designed to avoid peak stresses as much as

possible.

2.2.8.2 Epoxies

In general, epoxies are two component adhesives consisting of an epoxy resin and a

hardener. Based on the type of curing, they can be classified into UV-hardening and cold

and warm hardening epoxies. Curing of UV-hardening epoxies is analogous to UV-curing

acrylates, whereas the reaction of cold hardening epoxies is induced when resin and

hardener make mutual contact, and warm hardening epoxies are actually one component

adhesives curing under increased temperature. Hybrid epoxies have been developed over

the last years, yielding a more elastic material behaviour which is of interest for several

structural applications. They are also known as toughened epoxies, as they are tougher

and do not break in the same brittle way traditional epoxies do.


39

Similarly to acrylates, epoxies are very strong and stiff adhesives and are typically

applied in very thin layers. The strength of epoxies strongly depends on the specific

product and may reach characteristic values of up to 30 MPa. Two disadvantages are the

brittle nature of the adhesive joint and limited deformation capacity to deal with

differential thermal expansions. For this reason, efforts are done lately to modify epoxies,

e.g. with rubber particles, to increase their toughness and to make them a more elastic

material. In terms of dependency on temperature and humidity, epoxies will generally be

less sensitive compared to acrylates. In some cases extra attention should be paid to pre-

treatment. The curing time will normally be rather long, and warm hardening epoxies are

exposed to a risk of stresses due to shrinking. In general, the brittle nature of traditional

epoxies necessitates a connection design based on the avoidance of peak stresses.

2.2.8.3 Polyurethanes

Multiple types of polyurethanes exist. A differentiation may be made between physically

hardening polyurethanes and chemically hardening one- or two-component adhesives. For

structural applications, usually chemically hardening types are applied.

Properties of polyurethanes strongly depend on the exact ratio of the constituents. Curing

velocity, elasticity, adhesion, strength, etc. may vary by changing the proportion of the

different constituents. The strength of polyurethanes can vary dramatically, with

characteristic values typically ranging between 1 to 15 MPa. In general, strength and

stiffness are significantly lower compared to acrylates, but still relatively high compared

to typical silicones. Generally, polyurethanes have an average to low stiffness and are

applied with a relatively large thickness. Consequently, these adhesives possess good

gap-filling properties, and because of their flexibility they cope well with dynamic actions

and differential thermal expansions. However, the biggest problem for traditional

polyurethanes is their limited resistance to UV radiation. Regardless of some high-quality

exceptions, this is problematic for all polyurethanes. Although it is possible to protect the

adhesive by means of UV blocking primers or prints on the glass, in general this property

will lead to a limited lifetime of the connection.


40

2.2.8.4 Silicones

Two types of traditional silicones are available on the market: one-component and two-

component systems. One-component silicones cure when exposed to air by reacting to the

humidity in the air, whereas two-component silicones react more quickly due to the

addition of a chemical constituent. A transparent structural silicone adhesive (TSSA) film

has been recently developed. The latter cures in an oven at elevated temperatures and is

fully transparent.

Silicones are characterized by their relatively low characteristic strength, typically about

0,5 to 1,5 MPa in tension, and a relatively low stiffness. Generally, one component

silicone adhesives are slightly stronger and stiffer than two component silicones because

the first continue to cure over time. Because of their limited stiffness, silicones can

compensate very well building tolerances and they have good resistance to differential

thermal expansions. Silicones are normally applied with a joint thickness of at least six

mm. Moreover, they can be applied in a broader temperature range than most other

adhesives, and they have excellent resistance against UV radiation, ozone, humidity and

other external exposures. The most important drawbacks of silicones are the low curing

velocity of one component systems, compatibility issues with certain coatings and

laminated glass interlayers such as PVB, and silicones being a possible cause of corrosion

for the substrates. The relatively low strength may be a major disadvantage as well in a

structural context. Recently developed TSSA films reach strengths which are

considerably higher than those of traditional silicone products. TSSA was not developed

for linear applications, as were the standard silicone adhesives, but rather for use in point-

fixings.

2.2.8.5 Hybrid polymer adhesives

Hybrid polymers are adhesives in which modified silane molecules have been applied,

which is the reason that it is also known as modified silane (MS) polymer. Hybrid

polymer adhesives usually have a polyurethane and silicon basis, in an attempt to

combine the advantages of polyurethane and silicone adhesives in one product.

Many variants exist with specific properties in terms of curing, strength, stiffness, etc.

Again, one and two component products can be distinguished.


41

Hybrid polymer adhesives have a better resistance against UV-radiation and are

consequently more durable than typical polyurethanes. For the rest, properties of both

products are comparable: the stiffness is limited and average to low tensile strengths are

obtained. In addition, hybrid polymers are in general not very sensitive to surface

pretreatments, bonding well to a variety of different materials. Disadvantages of hybrid

polymer adhesives are the very slow curing of one component systems, strength levels

which sometimes are too low for structural applications, and a durability which is still

inferior to structural silicones. In addition, these adhesives have only relatively recently

been developed and there are several practical issues that require further research.

2.2.8.6 Adhesive interlayers

Initially they were developed for the bonding of glass to glass in laminated glass.

However, adhesive glass to metal bonds by means of interlayer foils have been applied

already in several projects and are subject of ongoing research. PVB is the genuine

interlayer foil originally developed for the automotive industry. Multiple types exist,

ranging from very flexible grades, for instance to increase the acoustic performance, to

relatively stiff ones, typically used for structural applications. In addition, PVB interlayers

bond relatively well to many materials. SG is an adhesive ionomer interlayer foil which

has a significantly higher stiffness compared to traditional PVB foil. SG is used for

architectural applications which require stiff laminates, for instance to enhance the post-

breakage behaviour.

Typical for adhesive interlayers is that curing requires a lamination process under

increased temperature and pressure in an autoclave. During such an autoclave cycle the

interlayer reaches a low viscous state and the final bond to the substrate surface is

realized by a chemical reaction. When the laminate is cooled down, the interlayer

solidifies again and the product reaches its final cohesive strength.

The strength of adhesive connections with traditional PVB is lower compared to SG, but

then again the latter product is more sensitive to differential thermal expansion as a

consequence of its higher stiffness. Disadvantages of both types of products are their

sensitivity to moisture and temperature.


42

2.2.9 Practical execution of adhesive bonds

2.2.9.1 Substrate cleaning and pre-treatment

Building materials are typically handled in an environment that is potentially

contaminated resulting in poor wetting and therefore a small contact surface, thus

yielding poor adhesion quality. Consequently, thorough cleaning and degreasing is a

standard procedure for virtually all adhesive processes. Typical degreasing products are

isopropyl alcohol, also known as isopropanol or IPA, and acetone. These products are

easily inflammable and may be harmful if no sufficient ventilation is provided.

Obviously, safety requirements should be provided in all cases. The typical application

procedure is firstly to apply the degreaser on a clean tissue that does not leave any fluff.

Subsequently, the surface is cleaned either starting from the centre and expanding in

circular way, or linearly in one way. Finally, the surface is dried using a clean and dry

tissue moving in the same way the degreasing was executed. This procedure ensures that

contaminations are optimally absorbed and removed by the tissue and the risk of applying

the adhesive on a wet surface is avoided.

For some products, manufacturers prescribe specific primers to improve adhesion to

certain substrates. Usually such primers combine additional degreasing with chemical

activation of the substrate surface to obtain a more profound reaction and an improved

bonding. For the application of primers it is also important to strictly follow the

manufacturer’s instructions, as specific procedures are usually required both in terms of

safety regulations and application of the primer. A point of attention is not to apply the

adhesive too quickly after the primer on the surface. The main reason is that some

solvents in the primer first need to evaporate to avoid bonding on a wet surface.

Finally, primers may also be applied for other reasons. A typical example for glass to

metal bonds is a UV blocking primer to protect adhesives behind the glass from UV

radiation. Obviously, this type of primer needs only to be applied in case the adhesive is

UV- sensitive.

Ozone treatments target an optimal cleaning of the surface which is to be bonded. Plasma

treatments are not only cleaning, but also chemically activating the substrate.

In practice, both techniques work out partially analogously: ozone or plasma particles are

shot towards the substrate and react with all contaminations present there. As a result, a


43

spotless surface is created which offers in most cases ideal bonding conditions. As a rule,

ozone treatments take about five minutes, whereas only about one minute is required for

plasma treatments. It is crucial that bonding takes place as soon as possible after the

cleaning process. The main reason is that extremely clean material will attract new

contaminations very quickly. Consequently, the cleaning effect may be counteracted

already after 30 minutes.

Both techniques require special equipment with the consequent main disadvantage that

such treatments are hard or even impossible to apply on a building site. Additional

techniques exist to enhance adhesive bonding on a surface; mechanical surface

roughening for example may create positive effects in this prospect. However, surface

roughening of the glass is not applied frequently because it is impractical, effects are

usually limited, and the glass strength will decrease. In contrast, what is done more

frequently is local removal of coatings or loose oxide layers. Coatings are preferably

removed to avoid incompatibility problems with adhesives, and risks of weak bonding on

the interface between coating and substrate. For safety’s sake, in such cases surface

coatings are to be removed chemically or physically, e.g. by sanding or sandblasting.

2.2.9.2 Adhesive preparation and application

When preparing two component adhesives, the mixture ratio between components should

be strictly respected. Usually this will be no problem because most glue guns are

designed to release and both components in the right ratio and blend them in a sufficiently

long nozzle. In addition, both components will usually have different colours, allowing an

easy visual check of colour uniformity and good mixture when a new cartridge is used.

Pot life is the useable life of an adhesive in a receptacle once it has been mixed. Some

adhesives will start to cure quicker than others, due to their different chemical makeup

and reactivity. Atmospheric and climatic conditions can also affect the pot life of a mixed

adhesive. If the useable pot life of an adhesive is exceeded, and the product is setting, it

should be discarded and a new batch should be mixed. Finally, for some adhesives also

the processing temperature should be respected strictly. For example, applying certain

adhesives at low temperatures may badly result their curing and therefore the adhesive’s

final characteristics.


44

Several guidelines should be respected when applying adhesive joints. For example, for

one component adhesives, which typically cure when exposed to air, it is essential that the

joints are not closed off from air. Furthermore, a triangular cross-section of the adhesive

joint is preferred to a round one. The main reason behind that is that a triangular joint will

burst open when the second substrate is placed into position, whereas a round joint will

usually only be flattened, which results in a smaller bonded surface. The application

process must take place without bubbles or entrapped air. It is recommended to include a

small overflow of the adhesive to ensure the joint is filled completely over the whole

length. Another major factor for the application is the viscosity of the adhesive. The

viscosity defines the type of application (e.g. casting or application with static mixing

tubes) and can be increased with higher temperature, although not in great extent.

Some adhesives are very sensitive to deviations of their optimal application thickness.

Good workmanship needs a method to maintain and accurately guarantee a constant joint

thickness. To achieve this, special accessories exist, such as small glass spheres with a

certain diameter that are mixed through the adhesive. Optimal thickness of adhesive layer

has to be chosen not only with respect to required stiffness and load carrying capacity but

also to provide sufficient elongation (or shear strain), if that is required. In some cases it

must also compensate possible geometrical imperfections and balance tolerances of the

connected surfaces. In addition, the open time of an adhesive has to be respected. Open

time is the time interval after application at which the second substrate can be embedded

in the applied adhesive and the bond still meets tensile adhesion strength requirements. If

the open time expires, the spread adhesive should be scraped off and discarded.

When substrates and adhesive materials have been positioned, the intended joint should

be sufficiently filled. In practice, a sufficient amount of adhesive is put into place and all

materials are carefully positioned to avoid air inclusions. An advantage of using glass as a

substrate is that in many cases visual inspection of the adhesive is possible through the

glass. Substrates may be shifted carefully back and forth to additionally remove possible

air inclusions as long as the adhesive did not start to cure yet. This process requires some

experience and skill. Finally, excess adhesive should be removed by rounding off the

edges to become a visually smooth and clean appearance, and to avoid peak stresses that

are more easily obtained in case of sharp edges. Rounded edges can be made by using a

spatula before the adhesive has cured.


45

2.2.9.3 Curing and compatibility

Different curing methods such as by air humidity, UV, mixing of two components or

booster, will significantly influence the open time. Consequently, some fast-curing two

component adhesives will not be suitable to use in a large area or long bonds as they may

have already partially cured by the time the full area is filled with adhesive.

Another important point is the potential incompatibility of the different materials

involved. For example, during the glass lamination process, silicones or derived products

should never be used in the same room as they are incompatible with most adhesive

interlayers. Furthermore, incompatibility may also be an issue when combining adhesives

and certain coatings.

2.2.9.4 Quality control

Certain types of adhesives are very sensitive to the execution parameters described above,

meaning that their mechanical resistance may yield significant dispersion. In practice it is

extremely difficult to check whether or not all subsequent steps in the production process

have been carried out with sufficient precision, as non-destructive testing methods are not

readily available. This is one of the reasons why, compared to traditional building

materials, very large safety factors are common practice in structural adhesive design

today. It is strongly recommended to consult the adhesive producer and to pay attention to

the data sheets and work instructions. In addition the best before date must be respected

and the storage of the adhesives must be clearly defined (e.g. for moisture curing or UV-

hardening adhesives or polyurethanes susceptible to humidity).

2.2.10 Adhesive selection

The choice of adhesive is of vital significance for the design of load bearing glass

structures. Besides the different mechanical values of the adhesive, the substrate

characteristics, ageing and temperature resistance as well as the application, flow and

curing properties are crucial for the manufacturing process and the final load-bearing

capacity of the cured joint. Besides that the allowable tolerances of individual

components and hence the resulting thickness of the adhesive layer are important for the

mechanical behaviour.

All adhesives should also be chosen regarding to their open time and pot-life, which are

important in respect to fabrication criteria. Some of adhesives can be applied by gap-


46

filling, but other more viscous ones have to be compressed by the components that have

to be connected. Generally shape as well as size and geometry of glued joints also affect

selection of adhesives type with respect to their liquidity.

2.3 Composite structural action

If a panel is freely placed on top of a beam and friction between the plate and the beam is

assumed to be negligible, the plate and the beam will act separately to resist flexural

action – the layered limit (Figure 9c). Their separate actions give rise to a longitudinal

slip between the plate and the beam and also results in large deflections. If, however, the

plate and the beam are somehow interconnected, the longitudinal slip can be reduced

consequently resulting in reduced vertical deflections. Thus by interconnecting two

elements, their combined bending stiffness can be increased considerably. This

phenomenon of two components working together as opposed to separately is known as

composite action. The stiffness of the connection determines the degree of composite

action achieved. A connection that is as stiff as the constituent components results in full

composite action – the monolithic limit (Figure 9a).

Figure 9: The concept of composite action (Lukaszewska, 2009)

Composite structures have found wide applications especially in the aerospace industry

where the first mass production of sandwich units made of thin veneer faces with a balsa


47

core happened during World War Two. The pioneering research into these applications at

the time was the work of Gough et al. (1940) and Williams et al. (1941). In the

construction industry, the technology has mainly been applied through the use of timber-

concrete or steel-concrete composite floors. Steel-concrete slab composite systems are

well established and preferred over other types of floor systems due to their advantages

which include being simpler, faster, lighter and economical constructions (Andrade,

2004).

Figure 10: Composite slab (SMD Stockyards, 2015

2.4 Thermal transmission

The performance of ten different spacer bars in Insulated Glass Units (IGUs) mounted on

frames of four different materials was assessed by Elmahdy (2003), from the National

Research Council of Canada, to determine the factors that affect the thermal transmission

at joints.


48

Figure 11: Thermal resistivity of the combination of 4 different frame materials and 10 different edge

of glazing designs (Elmahdy, 2003)

This study concluded that the overall U-value of a window assembly is dependent on the

type of spacer bar, frame material (and design), and glazing, and is particularly affected

by the thermal properties of the frame material. For example, for the same conventional

spacer, the U-value of the assembly varied by 20% depending on the frame and ,for the

same wooden frame, the U-value of the assembly varied by 15% depending on the spacer.

Muñoz and Bobadilla (2012) undertook the task of developing a range of thermally

efficient façade systems through a process that involved U-value and condensation risk

assessment calculations. The performance of the initial design was assessed and then

modifications were proposed in an iterative process. This study demonstrated the

importance of addressing thermal bridges as well as overall U-values to properly

characterize a façade system.

2.5 Analysis of similar existing products

The concept of bonding the glass panels to the framing members is not in itself new. In

fact one form of unitized curtain wall system, known as structural silicone glazing, uses

low stiffness silicone adhesives with a bond line thickness ≥ 6mm. This produces a

relatively flexible joint that accommodates the differential thermal expansion between the

glass panels and the metal framing members. The disadvantage of this flexible joint is

that it is too compliant and it therefore mobilizes an insignificant amount of composite


49

structural action between the panes and the framing members. The novelty of the

composite unitized system discussed in this paper is that the façade framing members

consist of GFRP pultrusions that have a coefficient of thermal expansion similar to that of

glass, this makes it possible to use stiffer adhesives with thinner bond lines thereby

generating significant composite structural action between glass panels and the frames.

There are a number of products that have similarities with the proposed system. A search

has been carried out to identify curtain wall and IGU designs which are similar to the one

that is being proposed. A cross examination has be carried out against the following

points which are considered essential features in the proposed design:

Low thermal conductivity frame

Integration of the frame within the glazing depth and projection in elevation

without the need of additional framing

Structural contribution of glass by composite action with frame through the use of

structural adhesives

No requirement for external access in the installation

Double line of defence, pressure equalisation and provision for drainage

Glass units designed to be replaceable individually without dismounting other

units

No requirement for application of wet sealants to seal the gap between units

Integrated brackets for mounting to a building or structure and/or for acting as an

anchor point for forcing the IGU out of its natural plane - Cold bending would be

an attractive feature that could make the invention more distinct.

Similar products are listed below together with notes stating the differential aspects. This

notes are then summarised in table 1.

2.5.1 Glass panel for external enclosures (Rico Jaraba, 2006)

No composite action and requires additional framing on the inside.


50

2.5.2 Wall panel, method for manufacturing same and use of the panel in a curtain

wall (Van Herwijnen, 2003)


2.5.3 Edge seal gasket assembly for a multiple glazing unit (Thomas E. Kennedy,

1993)

It claims that it does not require frame but does not mention adhesive bonding nor

composite action. It features the gasket fixed to the unit but does not provide

double line of defence, pressure equalisation nor provision for drainage.

Frame

Frame


51

2.5.4 Isolierglass-Doppelscheibe, insbesondere für Treibhäuser ( Gerresheimer Glas

AG, 1973)

Relies on front sealing. Does not provide double line of defence, pressure

equalisation nor provision for drainage.

2.5.5 Curtain wall system wherein a special connection system is used for plate

materials such as glass, aluminium sheet, etc. (Gokdemir and Yilmaz, 2012)


Joint with single gasket overlap, without pressure equalissation nor provision for drainage

Joint relying on wet sealants and external access, without pressure equalissation nor provision for drainage


52

2.5.6 Attaching panes of glass on to frames made of aluminium, PVC or the like

(Pierre, 1992)

Incorporates aluminium frame which is not integrated within the glazing unit.

Frame

Frame


53

2.5.7 Window unit (Shea et al, 1982)

Relies on back sealing. Does not provide double line of defence, pressure

equalisation nor provision for drainage. Incorporates additional framing at top and

bottom.

Frame at top and bottom

Joint relying on wet sealants, without pressure equalissation nor provision for drainage


54

Table 1: State of the art analysis chart

PROPOSED DESIGN

Features typical of high performance Insulating Glass Units

Features typical of unitised curtain walls

Inte

grat

ed b

rack

ets f

or m

ount

ing

to a

bu

ildin

g or

stru

ctur

e

Low

con

duct

ivity

FR

P fr

ame

Inte

grat

ion

of f

ram

e w

ithin

gl

azin

g w

ithou

t add

ition

al

fram

ing

Com

posi

te a

ctio

n th

roug

h st

ruct

ural

adh

esiv

es

No

exte

rnal

acc

ess i

n in

stal

latio

n

Dou

ble

line

of d

efen

ce,

pres

sure

equ

alisa

tion

and

prov

isio

n fo

r dr

aina

ge

Gla

ss u

nits

rep

lace

able

in

divi

dual

ly w

ithou

t di

smou

ntin

g ot

her

units

No

wet

seal

ants

to se

al g

ap

betw

een

units

Rico Jaraba (2006)

Not addressed

Not addressed

Van Herwijnen (2003)

Not addressed

Not addressed

Thomas E. Kennedy, (1993)

Not addressed

Not addressed

Not addressed

Not addressed

Not addressed

Gerresheimer Glas AG (1973)

Not addressed

Not addressed

Not addressed

Not addressed

Not addressed

Gokdemir and Yilmaz (2012)

Not addressed

Not addressed

Pierre (1992)

Not addressed

Not addressed

Not addressed

Shea et al (1982)

Not addressed


55

Moreover, all of these products and investigations have been limited to proof of concept

prototypes and there appear to be no systematic investigations and/or test data of these

proposals. There are also no reported validated analytical or numerical models of the

mechanical response of such GFRP-glass composite units. This research redresses this

shortcoming by characterizing the framing materials and adhesives through mechanical

testing and subsequently used this material-level test data in a numerical model of a

typical GFRP-glass composite unitized panel subjected to realistic loads. This approach is

based on the investigations by Overend et al (2011) on five candidate adhesives for load

bearing steel–glass connections where mechanical testing and numerical modelling were

used to predict the performance of adhesive connections. Nhamoinesu and Overend

(2012) applied a similar methodology to assess the mechanical performance of adhesives

for a steel-glass composite façade system.

The initial selection of the candidate materials is made on the basis of technical data

provided by the manufacturers is used to select (Fiberline, 2003; Huntsman, 2007; 3M

Scotch-Weld, 1996; Dow Corning, 2013). The selection of candidate adhesives is further

aided by the findings of: (a) Belis et al (2011), who screened a broad range of glass-metal

bonds featuring silicones, polyurethanes, MS-polymers, acrylates, and epoxies, and; (b)

Peters (2006) who carried out investigations on the bonding of fiberglass and glass.

2.6 Conclusion

The proposed system is an innovative combination of existing technologies applied to

facades: spacers made of low thermal conductivity materials such as FRP, composite

structural action between glass and FRP and unitised curtain wall systems. In essence, the

proposal consists takes high performance IGUs and modifies them to build a unitised

curtain wall. There are several technical issues that had to be resolved to make this

possible:

Unlike conventional unitised, in the proposed design the IGUs cannot be detached

from the frame. Therefore, it would not be possible to replace them following the

traditional structural scheme where the units are top hung from the slab and

interlocked with the units below. They had to be fixed to the slabs at the top and at

the bottom, removing any structural interlock between units;

Unlike conventional unitised, the proposed design does not have protruding

frames to bolt the brackets. Moreover, FRP is not anisotropic as aluminium and


56

does not work well with bolted connections. Therefore, the brackets would have to

be directly fixed to the IGU;

Unlike conventional unitised, the proposed design does not have protruding

frames to fit the gaskets. They had to be integrated directly within the IGU to

provide a double line of defence, pressure equalisation and provision for drainage.

Finally, unlike previous work, this research validates the design both thermally and

structurally through testing and modelling.


57

SCHEMATIC DESIGN DESCRIPTION AND

COMPARISON WITH CONVENTIONAL SYSTEM

The proposed system is described and compared to a conventional system taken as

reference at a schematic level. Besides the structural and thermal performance, which are

assessed in depth in other chapters, the principal design issues and the strategies to

address them are reviewed for conventional unitised curtain wall systems. The alternative

strategies adopted by the proposed system are then outlined.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


58


59

3.1 Manufacturing process and supply chain

3.2 Support condition and glass replacement strategy

Brackets or fixings form the link between curtain wall and building structure. Each panel

in an unitized system is prefabricated as a statically determinate structural element.

The intention is for the brackets or fixings to be designed so that no potentially damaging

internal forces are generated in the panels after they have been attached to the structure.

Transfer of loads

Normally a conventional curtain wall is supported in front of the structural frame with the

provision of a buffer zone in between to accommodate tolerances.

Two types of connection are generally used. Vertical load supports resist gravity loads

and provide primary connections to the structural frame. These can be at the top or

bottom of the primary curtain wall framing member. Secondary connections at the top

and bottom of the curtain walling framing member provide resistance to horizontal loads.

The vertical load supports frequently provide horizontal restraint as well.

Vertical loads include the self-weight of the cladding and any internal and external

attachments thereto. Horizontal loads include wind pressure and suction, external impact

loads from cleaning and maintenance personnel and cradles and internal impact from

building users.

Accommodation of movement

Movement is change in dimension arising from material properties. It can be permanent

or reversible. It applies to the curtain wall and the structural frame behind. The

movements in the curtain wall may work with or against the structural frame and/or vice

versa. They can be summarised as movements resulting from: applied load (including

dead and live loads), settlement, creep, temperature change, moisture change, shrinkage,

etc.

The design of the brackets or fixings connecting the curtain wall to the structural frame

needs to take account of these movements to avoid imposing loads on the curtain wall for

which it has not been designed; resulting in deformation and breakage, and in extreme

cases pieces falling off the building and/or imposing larger than anticipated movements


60

on the curtain wall; resulting in breakdown and failure of seals and disengagement of

pieces of curtain walling.

Accommodation of tolerance

Deviations are differences between specified nominal dimensions and actual measured

dimensions. Induced deviations are permanent deviations arising from variations and

errors deviations, which the bracket or fixing design must accommodate.

The common problems can be summarised as follows:

Deviations larger than tolerances.

Devices attached to the structural frame to receive curtain wall brackets

wrongly positioned.

Strong points in the structural frame to receive curtain wall brackets in the

wrong place.

The design of the curtain wall brackets or fixings connecting it to the structural frame

need to accommodate the agreed tolerances and deviations.

Design of brackets or fixings to allow for movement for conventional system

Generally, panels have two vertical load carrying supports. Occasionally, one is

completely fixed to the slab edge and the other allows horizontal sliding movement in the

plane of the wall letting the panel expand and contract due to temperature change. More

usually the panels are hooked on to brackets or fixings attached to the structural frame.

One of the hooks can slide horizontally in the plane of the wall, the other is fixed. In/out

restraint is provided by interlocking sections or splice plates linking the panels and

making them move together, or brackets or fixings back to the structural frame resulting

in the panels moving independently of each other.

Where restraint is provided by interlocking sections, slab edge deflection causes one load

bearing bracket to drop relative to its neighbour resulting in the whole panel rotating in

the plane of the wall. This produces changes in the sizes of the vertical and horizontal

joints between the panels, some open and some close depending on where the panel sits

on the span of the slab edge.

Where restraint is provided by splice plates, when the slab edge deflects, the panels stay

vertical and some wracking occurs and some panel interlocking occurs resulting in panels

lifting off their load support brackets. This lifting effect can result in the entire panel load


61

being carried on one load support bracket rather than two. Building sway and differential

settlement cause the panels to rack.

Figure 12: Movement diagram of conventional unitized curtain wall caused by building frame

deflection due to vertical load and wind sway. (Source: CWCT TN 54, 2007)

Where restraint is provided by brackets or fixings back to the slab, slab edge deflection

results in the same effects as for interlocking joints. However, the restraint brackets and

fixings have to be provided with slots and oversized holes to allow for the upper slab

being loaded differently from the one below and therefore moving differently. Normally

the lower restraints are designed to allow movement in any direction in the plane of the

wall. Similar effects occur when the structural frame settles and/or sways.

Detailing of brackets or fixings to allow for movement for the proposed system


62

The transfer of loads and accommodation of movement would follow similar pattern as

explained above. It is envisaged that panels would have two vertical load carrying

supports similar to the conventional systems. The in/out restraint would be provided by

additional brackets or fixings back to the structural frame instead of interlocking or

splices, resulting in the panels moving independently, used in conventional systems.

Glass replacement strategy

The characteristics of all conventional curtain walling, unitized or ‘stick’ and mechanical

restraint or structural glazed is that the frame remains in place when the glass is being

replaced. In all of them the system is designed to allow the glass to become detached

from the frame. In the majority of the cases, the replacement is done from the outside

requiring external access for the operators and for transporting the glass.

On the proposed system the frame is integrated in the glazing, this made the traditional

way of replacing the glass and maintaining the frame not suitable.

It is envisaged that the glass replacement on the proposed system would also be carried

out from outside. The glass unit along with the frame would become detached from the

adjacent units leaving only the brackets in place. To achieve this the system has been

designed with a fixation that can be unbolted independently without the need of removal

of perimeter elements. In the same way, the replaced glass unit with the frame integrated

would be able to be located and bolted back in place without disturbing adjacent glazing

units. The saddle gaskets that provide the weathering protection of the system are a

flexible material that can be re-inserted into the grooves provided in the framing by the

operators working from outside

3.3 Sealing and drainage strategy


63

3.4 Fire performance and fire partition strategy

Curtain wall systems can be fire rated, but generally they are used at building envelope

areas where fire resistance is not a requirement. However curtain wall systems need an

appropriate detailing to limit the spread of fire and provide compartmentation. The main

requirements for curtain wall systems are:

the provision of fire stopping between the external wall and compartment floors

and walls; and,

the limitation of combustibility of materials used in the wall.

In some situations there may also be a requirement to provide fire protection to brackets

supporting the wall.

3.4.1 Compartmentation

Many buildings are divided into compartments to restrict fire spread. Where an external

wall abuts a compartment wall or floor, it is necessary to provide fire stopping between

the external wall and the compartment wall or floor to restrict fire spread through the

junction.

Overview of the building regulation code

In some countries, such as Spain, it is mandatory to provide a 1m high band with a fire

resistance of 60 minutes for insulation and integrity at the interfaces with the floor slab as

shown on the figure below.

Figure 13: Spanish building code requirement for floor compartmentation (Source: CTE-DB-SI

(2010). Seguridad en caso de incendio. Ministerio de Fomento: Madrid, Spain page 2-2)

However, there are other countries, such as United Kingdom, where it is assumed that the

curtain wall would be collapsed in a fire event. Therefore the gap between floor and


64

external wall only requires to stop the heat and smoke spread between compartment

floors during the period that the external wall stands. As stated in the UK Buildings

Regulations Approved Document part B, Clause B8.25, where a compartment floor meets

an external wall, the junction should maintain the fire resistance of the compartmentation.

Fire-stopping and sealing systems shall be proprietary products which have been shown

by test to maintain the fire resistance of the wall.

Figure 14: UK building regulations for floor compartmentation (Source: UK Building Regulations

Approved Document part B, Clause B3 Diagram 33. NBS RIBA Enterprises, Newcastle Upon Tyne, United Kingdom )


65

Detailing of conventional curtain wall system at floor compartment

Although the curtain wall is not required to be fire resisting, the effectiveness of the fire

stop will depend on the performance of the curtain wall. It is worth noting that relatively

early in a fire, the temperature of hot smoke can be as high as 500ºC. Solutions include

removing a strip of the insulation to allow the fire stop to continue to the back of the

glazing or metal spandrel panel, a fire resisting lining on the back of the insulation against

which the fire stop can interface or a fire resisting insulation for the whole of the spandrel

panel. If the interface of the fire stop with the curtain wall is aligned with the transom

location, the transoms may require protection by fire resisting boards to extend the fire

resisting construction to the glazing. Fire stopping products are required to prevent

transfer of heat and smoke. Proprietary materials are available which are generally based

on rock fibre to control the passage of heat and aluminium foil or a liquid applied

membrane to control the passage of smoke. Fire stops should be tested to demonstrate

performance. Fire stops are often tested to BS 476-20 or EN 1366-4 with the fire stop

positioned between fire resisting constructions. In addition, to form a good seal, fire stops

generally need to be compressed. The amount of compression required depends on the

nature of the fire stop materials and should be as required by the fire stop manufacturer.

Figure 15: Typical detail for conventional unitized curtain walling system for floor compartmentation


66

Detailing of proposed system at floor compartment

The detailing of the proposed system at floor level will vary depending on the country on

which the building would be located:

In the case of most of the countries in continental Europe, such as Spain, a 1m

band insulation would need to be added at the back of the inner glass pane to

provide de 60 minutes fire integrity and insulation. A fire resistance board can

also use to form the band. This might require protruding frames to be fixed to. The

glass can be back painted or fritted on this area to avoid the fire stopping material

being seen from the outside.

In the UK and countries with similar requirement, a detail as used in the

conventional systems would work. The gap between glass and floor slab would be

filled with insulation providing the fire resistance and a metal sheet or a liquid

applied membrane to control the passage of smoke.

Figure 16: Proposed system detailing for floor compartmentation


67

3.4.2 Combustibility

To avoid the spread of fire along the envelope surface it is key to limit the combustibility

of the materials used in the wall.

Overview of the building regulation code

Building regulation codes vary from country to country. Below are the extracts and

comparison of the Spanish and UK regulations:

Spanish Building Code (Código Técnico de la Edificación CTE-DB-SI (2010).

Seguridad en caso de incendio. Ministerio de Fomento: Madrid, Spain page 2-2)

states that the fire reaction of the materials that cover more than 10% of the

envelope area shall be B-s3-,d2 in accordance with BS EN 13501-1 up to a height

of 3.5m or on the entire envelope area when the building is higher than 18m above

ground level.

In accordance with UK Building Regulations Approved Document part B, Clause

B4 12.7 Insulation Materials/products: In a building with a storey 18m or more

above ground level any insulation product, filler material (not including gaskets,

sealants and similar) etc. used in the external wall construction should be of

limited combustibility. It requires to be any material/product classified as Class

A2-s3,d2 or better in accordance with BS EN 13501-1.

As can be seen from the above extracts, both the Spanish and UK Regulations have

similar requirements with regards to the prevention of the spread of fire. The use of

materials with limit combustibility is mandatory. In both cases, it is assumed that there

might be some materials that need to be used in the system that might be combustible. In

this case the Spanish code limits the percentage of area of these materials while the UK

code explicitly mentions which materials are excluded of this requirements.

Combustibility on conventional curtain wall system

Typical materials used in conventional curtain wall systems are glass and aluminium in

the largest quantity. Both of them meet the requirements of limit of combustibility as

stated in the building regulations. However conventional systems also use rubbers, EPDM

products or sealant to provide the weathering requirements. These materials do not

usually meet the level of combustibility required but as their extent of use in the façade is


68

limited, the codes allow their use because the spread of fire along them would also be

limited.

Combustibility of proposed system

The key material used in proposed system is glass, which is considered as non-

combustible. The gasket used for weathering protection are similar to conventional

system and are generally excluded of the non-combustibility requirement due to the

limited extend on their use. The fire behaviour of the FRP frames would depend on

several factors including their reaction due to its position within the glazed panes.

3.5 Acoustic performance and strategies to limit flanking

The performance of a wall has to be considered in terms of the external, internal and the

adjacent spaces. The aim is to provide a building envelope that gives the required sound

pressure levels within a room or other internal space.

The noise level within a room will depend on:

the amount of sound energy transmitted through the wall

the inter-reflection of sound inside the room.

he second parameter is related to the internal properties of the space, such as finishes

and furniture and it does not depends on the external envelope system type. Therefore it is

not analysed in detail in this comparison with conventional system. This assessment is

focused on the first item, the amount of sound transmitted through the wall, which is

mainly divided into two components:

Airborne or direct sound transmission;

Flanking transmission.

3.5.1 Airborne or direct sound transmission

As described in the CWCT technical note 39, the sound transmission through a whole

wall is established by calculating an apparent sound reduction index (SRI) for the wall.

This is used to determine the difference in sound between the outside and inside.


69

Direct sound transmission on conventional curtain wall system

The direct sound transmission through the curtain walling is transmitted through the

glazing and frames. As the glazing usually covers the majority of the area, it is the key

element to reduce the sound coming inside a building.

Direct sound transmission on proposed system

Proposed system can accommodate ‘acoustic’ glazing as the conventional systems.

‘Acoustic’ glazing is generally double or triple glazing with at least one pane of laminated

glass.

Figure 17: Diagram of the direct sound transmission through glazing.

3.5.2 Flanking transmission.

As described in the CWCT technical note 39, flanking transmission is the transmission of

sound through the wall by adjacent elements, such as partitions or floors. For sound

travelling through an external façade, flanking would involve vibration of the façade

being transmitted to internal walls and floors, which would then radiate sound into rooms.


70

Flanking transmission on conventional curtain wall system

In the interface of a conventional curtain wall system with the internal partition, flanking

sound can be transmitted through the glazing and through the mullion, being the mullion

the weak part. Depending on the requirement for the project, it might need to be filled up

or clad to reduce the amount of sound transmitted.

Figure 18: Diagram of flanking transmission through conventional system.

Flanking transmission on proposed system

The benefit of proposed system in terms of flanking transmission is that the frame which

is the weak part is reduced significantly when compared to a conventional system. In

addition, the remaining frame is integrated within the glass panes adding more mass

resistance to the transmission.

Figure 19: Diagram of flanking transmission through Proposed system system


71

SCHEMATIC DESIGN STRUCTURAL

ASSESSMENT BY ANALYTICAL CALCULATION

Deflection, moment stress and shear stress induced in the proposed system by wind load

are predicted by means of simple bending theory (Euler-Bernouilli). The results are

compared with the performance of a conventional system taken as reference. Sensitivity

analysis is carried out to observe how varying the structural depth of the system affects

the stiffness and the bending stresses and how the shear stresses at the adhesive-glass

interface or the GFRP web influence the design.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


72


73

4.1 Method

The calculations are based on simple bending theory (Euler-Bernouilli), assuming that no

shear deformations take place and that the cross sections of the beam remain planar and

normal to the deformed axis of the beam. It is assumed that the whole width of the glass

is playing an equal role, excluding shear lag effects. A linear approach has been taken

ignoring time and temperature dependant properties of materials. The assumptions made

on the properties of the materials, the geometry of the structural sections and the loading

and support conditions are first described. A summary of the calculations carried out

follows.

4.1.1 Material properties

The table below summarises the main mechanical properties of the materials.

Table 2: Mechanical properties of materials

E Modulus of Elasticity [GPa]

fb Limiting stress for bending [MPa]

fv Limiting stress in shear [MPa]

Aluminium alloy Type 6063 T6 (extrusion) British Standards (BS 8118: Part 1: 1991)

70 160 95

γm = 1.2 Material safety factor to be applied in bending and shear stress calculations, not deflection

Glass Fibre Reinforce Polyester Fiberline Design Manual for GFRP, which is in accordance with EUROCOMP Design Code. The calculation methods and safety philosophy of which are in accordance with Eurocode 1, section 1, Bases for projecting and stress on supporting structures.

23 240 25

γm = 1.6 Material safety factor (Table 2 - short-term load at 80 °C). This partial coefficient is dependent on the production method, degree of postcuring, certainty of dimensional stability and operating temperature. Applied in bending and shear stress calculations, not deflection

Float glass American Standards ( ASTM E1300 - 12ae1) for float glass

70 Allowable surface stress of glass [MPa] (dependent on load duration)

Annealed Heat strengthened Toughened

23.3 46.6 93.1

Huntsman Araldite 2047 Epoxy (Nhamoinesu and Overend 2010)

0.6035


74

4.1.2 Geometry of the structural sections

In order to show the effect of varying the depth of the section on the stiffness and the

moment stresses, a range of varied depths was defined for each system. The geometry of

the sections and their respective properties are described in the following figures and

tables.

Figure 20: Structural cross-section of the conventional system

In general practice the frames are sized to bear all the load, disregarding the structural

contribution of glass mobilised through the structural silicone bonds. This is a

conservative approach taken partly to avoid having to rely on the execution of the bond or

its durability and partly due to the fact that structural silicones are deemed too flexible to


75

mobilise significant composite action between the aluminium frame and the insulating

glass unit. In fact, the Shear Modulus of DC 993, a typical structural silicone, is two

orders of magnitude lower than the adhesives that are considered in the proposed design

(Nhamoinesu and Overend, 2012). Therefore, for the purpose of this study, the

contribution of the glass to the stiffness of conventional system unit is ignored and the

second moment of area is calculated based exclusively on the geometry of the aluminium

frame. The progressive system depths are achieved by stretching the middlle portion of

the aluminium extrusions. This way, the front and back of the system are the same for all

the different depths as shown in figure 20.

Table 3: Structural section properties for a range of depths of the conventional system

System depth I Second moment of area [mm4]

y Distance from most extreme fiber to neutral axis [mm]

Z = I/y Section modulus [mm3]

A Cross-sectional area of the structural element [mm2]

200 mm 4. 38 x 106 78.70 55.63 x 103 1 796.00

210 mm 5. 22 x 106 84.02 62.10 x 103 1 876.00

220 mm 6.15 x 106 89.32 68.86 x 103 1 956.00

230 mm 7.18 x 106 94.59 75.91 x 103 2 036.00

240 mm 8.31 x 106 99.84 83.25 x 103 2 116.00

250 mm 9.55 x 106 105.08 90.88 x 103 2 196.00

260 mm 10.90 x 106 110.29 98.79 x 103 2 276.00

270 mm 12.36 x 106 115.500 106.99 x 103 2 356.00

280 mm 13.93 x 106 120.68 115.46 x 103 2 436.00

290 mm 15.63 x 106 125.86 124.22 x 103 2 516.00

300 mm 17.46 x 106 131.03 133.25 x 103 2 596.00

Monolithic behaviour of the assembly of glass panes, GFRP frames and adhesive has

been assumed. A linear approach has been taken ignoring time and temperature

dependant properties of adhesives and GFRP and the shear lag along the width of the

glass flanges have also been ignored. Preliminary calculations indicated that the

contribution of the GFRP frames and the adhesive to the effective second moment of area


76

of the whole section was less than 1%. For this reason, the second moment of area of the

proposed sytem was calculated based exclusively on the geometry of the glass flanges.

Figure 21: Structural cross-section of the proposed system

The cut line at the adhesive-glass interface as shown in figure 21 is considered valid for

shear stress calculations both at the adhesive-glass interface and at the GFRP web. This is

due to the area of GFRP being considered negligible compared to the area of glass.

Table 4: Structural section properties for a range of depths of the proposed system

System depth I Second moment of area [mm4]

y Distance from most extreme fibre to neutral axis [mm]

Z = I/y Section modulus [mm3]

𝒚′ Distance from centre of area above the cut line to centroid of whole section [mm]

51 mm 12.73 x 106 25.50 499.18 x 103 20.50

53 mm 13.98 x 106 26.50 527.41 x 103 21.50

55 mm 15.28 x 106 27.50 555.75 x 103 22.50

57 mm 16.65 x 106 28.50 584.12 x 103 23.50

59 mm 18.07 x 106 29.50 612.71 x 103 24.50

61 mm 19.56 x 106 30.50 641.31 x 103 25.50


77

4.1.3 Applied load

Shear stress induced on the adhesive by differential longitudinal thermal expansion of the

glass and the GFRP has been investigated. As can be seen in equation 1, it is directly

related to the difference between their coefficients of thermal expansion:

τt = Gadhesive(L/2)ΔT(αglass−αGFRP)

x [1]

Where:

τt is the shear stress caused by differential thermal expansion

Gadhesive is the shear modulus of the adhesive

L is the length of the curtain wall unit

ΔT is the variation in temperature

αglass is the coefficient of thermal expansion of glass

αGFRP is the coefficient of thermal expansion of GFRP

x is the thickness of the adhesive bond

The coefficient of thermal expansion of glass may vary slightly between 8 x 10-6 m/mK

and 9 x 10-6 m/mK depending on the exact composition. Conversely, the coefficient of

thermal expansion of pultruded glass fibre reinforced polyester may vary significantly.

According to Fiberline (2014), it may vary between 8 x 10-6 m/mK and 14 x 10-6 m/mK

depending on the specific profile. Experimental research by Sengupta and Spurgeon

(1992) confirms that the expansion of the composites is s directly related to the volume

filling fraction and modulus of elasticity of the resin and inversely related to the volume

filling fraction and modulus of elasticity of the fibres. Moreover, the thermal expansion of

polyester resin in non-linear. By adjusting the arrangement of the fibres and their

proportion respect of the resin in the GFRP, it is possible to approximate its coefficient of

thermal expansion to that of float glass, thereby, keeping differential thermal expansion to

a minimum. It is worth noting that the shear stress induced by differential thermal

expansion could be reduced if necessary by increasing the thickness of the bond.

In a conventional system, the weight of each insulating glazing unit usually rests on the

transom through two setting blocks. It is transmitted through the transoms to the mullions

and then goes up the mullions and is transferred to the slab through the brackets. The

moment and shear at the transoms is sometimes the critical case and may be an important


78

design driver. However, in the proposed system the weight of the glass panes does not

rest on the transoms. It is transferred to the mullions directly through the glass panes

themselves. Therefore, dead load does not influence the design of the the frames except

for the brackets. It does induce shear stress on the bonded connection but the value of this

stress is negligible due to the large surface of the linear bond.

Apart from the aforementioned loads, the main loads that would normally be considered

in a comprehensive structural assessment would be wind load, barrier loads and impact

loads. Where applicable, sand or snow superimposed loads, blast or seismic loads should

also be considered. For the purpose of this comparative study, wind load has been

considered the critical case and has been the only load to be investigated in depth.

A façade pressure of 3 KPa has been applied on a vertical façade as shown in figure 22

As load safety factors may vary for different load cases, the load is unfactored. A

conservative assumption has been made that all the wind load is taken directly by the

mullions, ignoring the structural contribution of the transoms.

Figure 22: Elevational area of wind load assigned to each mullion


79

4.1.4 Support

In conventional systems the units normally span from floor to floor hanging from pre-

fixed brackets along the edge of the upper floor slab and being horizontally restrained by

the units below, as shown in figure 23(a). In the proposed system, the horizontal restraint

is provided directly by the edge of the lower slab, as shown in figure 23 (b). Figure 23(c

and d) provides the shear and moment diagrams of a simply supported beam under

uniformly distributed load assuming static loading conditions.

Figure 23: Structural diagrams (a) cross-section showing out-of plane load distribution and support

condition for conventional system; (b) cross-section showing out-of plane load distribution and support condition for proposed system; (c) shear distribution and (d) bending moment distribution

for simply supported beam

(a)

(b)

(c)

(d)

Bracket connection to the upper slab

Upper unit horizontally restrained by unit below

Bracket connection to the lower slab

Vmax = wl

2

Mmax =

w𝑙2

8


80

In the conventional system, if the stack joint was located at the inflection point, where the

bending moment is 0, a continuous beam support condition could be assumed as shown in

figure 24. However, to do so, the stack joint would have to be located roughly at 21% of

the span. This would be the most efficient scenario but would imply raising the transom

roughly 735 mm above the slab obstructing the view. In order to respect the architectural

intent of full height glazing, a simply supported beam condition has been assumed for

both the conventional and the proposed system.

Figure 24: Structural diagrams (a) cross-section showing out-of plane load distribution and support condition for conventional system with a raised stack joint; (b) shear distribution and (d) bending

moment distribution for continuous beam

(a)

(b)

(c)

Bracket connection to the upper slab

Raised stack joint

Vmax = wl

2

Mmax =

w𝑙2

12


81

4.1.5 Deflection calculations

The maximum deflection at framing members is calculated for a range of depths for both

systems using equation 2.

ƍmax = 5w𝑙4

384EI [2]

Where:

ƍmax is the maximum deflection

w is the linear uniform load

𝑙 is the span length

E is the modulus of elasticity

I is the second moment of area

The data obtained is used to plot a typical depth vs. deflection curve for each system. The

curves are then related to the the maximum permissible deflection for frame members due

to wind load established in section 3.5.2.5 of the CWCT Standard for systemised building

envelopes (CWCT, 2005). The recommendation for four-edge supported double glazed

units is 1/175 of the length unit along the unit edge, or 15 mm, or more restrictive limits if

set by the manufacturer, whichever is the lesser. The value of 15 mm is taken as

guidance.

4.1.6 Bending moment calculations

The maximum bending moment is calculated using equation 3:

Mmax = w𝑙2

8 [3]

Where:

Mmax is the maximum bending moment



The bending stresses are then calculated for a range of depths for both systems using

equation 4:


82

σ = M

Z [4]

Where:

σ is the bending stress

M is the bending moment

𝑍 is the section modulus

The data obtained is used to plot a typical depth vs. bending moment curve for each

system. The curves are then related to the limiting stresses of the materials used.

4.1.7 Shear calculations

The maximum shear force is calculated using equation 5:

Vmax = wl

2 [5]

Where:

Vmax is the maximum shear force



The average shear stress is then calculated for a range of depths for the conventional

system using equation 6.

τaverage = V

A [6]

Where:

τaverage is the average shear stress

V is shear force

A is the cross-sectional area of the structural element


83

The local shear stress may exceed the average shear stress at certain locations depending

on the geometry of the structural section. This was not important in the conventional

system because the shear strength of aluminium is well above the calculated average

shear stress. However, the structural section of the proposed system is composed of

different materials, some of them with relatively low shear strength. For this reason beam

shear stresses have been calculated at critical locations for both the 55 mm deep and a 61

mm deep options using equation 7

τbeam = VACy′

Ia [7]

Where:

τbeam is the beam shear stress at a given location

V is the shear force

AC is the area of the segment above the cut line

y′ is the distance from centre of area above the cut line to centroid of whole section

I is the second moment of area of whole section

a is the breadth of the section at the cut line being considered

As the area of GFRP is negligible compared to the area of glass, the cut line at the

adhesive-glass interface shown in figure 21 has been considered valid to calculate both

the shear stresses at the adhesive-glass interface and the GFRP web. The stresses have

been calculated for a range of different thicknesses of the GFRP web. . The data obtained

has been used to plot a breadth of section vs. shear stress typical curve for each of the 55

mm deep and the 61 mm deep options. The calculated stress at the adhesive-glass

interface is used to determine the minimum shear strength required from candidate

adhesives. The calculated stress at the GFRP web is related to the the limiting stress of

GFRP.

4.2 Results and discussion

The relationship between the depth of the system and the deflection of the frame under a

fixed wind load of 3 KN/m2 is illustrated in figure 25 for both the conventional and the

proposed systems.


84

Figure 25: System depth vs. deflection of the frame curves

In order to meet the recommend deflection limit of 15 mm, the conventional system needs

to be 241 mm deep while the proposed system only needs to be 43 mm deep. Similarly,

the stiffness of the 55 mm deep proposed system is matched by a 288 mm deep

conventional system. In both cases, the proposed system requires five times less structural

depth than the conventional system to achieve an equivalent stiffness.

The relationship between the depth of the system and the bending stress under a fixed

wind load of 3 KN/m2 is illustrated in figure 26 for both the conventional and the

proposed systems. In the conventional system, the limiting stress of aluminium is reached

below system depths of 193 mm. In the proposed system, the limiting stress of heat

strengthened glass is only reached below system depths of 22 mm. Bending stresses are,

therefore, less critical than deflection of the frame in both systems and do not drive the

design.

15.00, 42.79

15.00, 240.80

8.22, 55.00

8.22, 287.37

y = 533.85x-0.294

y = 132.37x-0.417

0

25

50

75

100

125

150

175

200

225

250

275

300

325

0 5 10 15 20 25 30

Sy

ste

m d

ep

th [

mm

]

Deflection [mm]

Deflection limit

Conventional system

Proposed system


85

Figure 26: System depth vs. bending stress curves

The relationship between the breadth of the section at the cutline and the shear stress

under a fixed wind load of 3 KN/m2 is illustrated in figure 27 for the proposed system.

The shear stresses at the section through the adhesive bond and the GFRP web are

investigated. The breadth of the section considered at the adhesive bond is 52 mm which

corresponds to a 26 mm bond for each of the two frames. The breadth of the section

considered at the GFRP web is 4 mm which corresponds to a 2 mm web thickness for

each of the two frames.

The two curves corresponding to the 55 mm deep and the 61 mm deep options are very

similar, the 61 mm deep option yielding slightly lower shear stresses than the 51 mm deep

option. Therfore, increasing the depth of the system reduces the shear stress but only

marginally. For the 55 mm deep system, the shear stress is 3.31 MPa at the adhesive and

43.04 MPa at the GFRP web. This value is used in the following chapters as target shear

strength for candidate adhesives. The value at the GFRP web almost threefolds the

permissible shear stress of GFRP of 15.63 MPa. The most effective way of reducing the

shear stress at this location would be to increase the thickness of the GFRP webs. To

reach a permissible shear stress value, the thickness of each GFRP web would need to be

increased to 5.5 mm. Increasing the depth of the system could also help.

133.33, 192.53

46.60, 21.33 y = 1873.2x-0.465

y = 332.64x-0.715

0

25

50

75

100

125

150

175

200

225

250

275

300

325

0 20 40 60 80 100 120 140

Sy

ste

m d

ep

th [

mm

]

Bending moment stress [MPa]

Limiting stress aluminium

Limiting stress heat strengthened glass

Conventional system

Proposed system


86

Figure 27: Breadth of the section at the cutline vs. shear stress

4.3 Conclusion

Subjected to wind load, the proposed system requires five times less structural depth than

the benchmark conventional system to achieve an equivalent stiffness. Bending stresses

are not the critical case and do not drive the design. The shear stress at the adhesive-glass

interface is 3.31 MPa. The shear stress at the GFRP web almost threefolds the permissible

shear stress of GFRP. The most effective way of reducing the shear stress at this location

is increasing the thickness of the GFRP webs. Increasing the depth of the system also

helps to a lesser extent.

The results obtained are very useful to achieve an understanding of how variables such as

the system depth, the width of the adhesive bond or the thickness of the GFRP web

influence the design and they provide a clear direction as to how to develop the design

further. However, they need to be interpreted with caution since they are based on

numerous assumption as listed below:

• Basic beam theory (Euler-Bernouilli) – No shear deformation, cross sections of

beam remain planar and normal to main axis of beam;

• No shear lag accounted for;

• No time and temperature dependent properties accounted for;

3.31, 52.00

15.63, 11.01

43.04, 4.00y = 172.09x-1

y = 152.46x-1

0

5

10

15

20

25

30

35

40

45

50

55

0 5 10 15 20 25 30 35 40 45

Bre

ad

th o

f th

e s

ec

tio

n a

t c

ut

lin

e [

mm

]

Shear stress [MPa]

Breadth at adhesive bond

Breadth at GFRP web

Limiting stress GFRP

55 mm deep system

61 mm deep system


87

• Only wind load considered: dead load, thermal load, impact load, etc. are ignored.

This is the considered acceptable because it is the critical case;

• Local buckling not considered;

• Contribution of gplazing to stiffness in conventional systems considered to be

negligible.

These results, therefore, need to be validated through experimental testing and numerical

analysis in the following chapters.


88


89

SCHEMATIC DESIGN THERMAL ASSESSMENT

BY NUMERICAL CALCULATION

Thermal transmittance and risk of condensation of the frame-integrated system are

assessed through comparative analytical and numerical thermal analysis with a

conventional system taken as reference.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


90


91

5.1 Method

Curtain walls often contain different kinds of materials, joined in different ways, and can

exhibit numerous variations of geometrical shape. With such a complex structure, the

likelihood of producing thermal bridges across the curtain wall envelope is quite high. For

this reason, standard procedures have been established to calculate the thermal

transmittance of curtain wall structures. BS EN 12631 (2012) and ANSI/NFRC 100

(2014) are the reference standards in Europe and America respectively. They both

describe overall system U-value calculation methods based on area weighting the U-

values of the different components. Using validated computer software is industry

common practice to obtain specific U-values for bespoke systems.

Figure 28 illustrates the complete process map. On the left hand side, the input

parameters that had to be defined: environmental conditions, materials and geometry. On

the right hand side, the output data: U-values and risk of condensation. The details have

been drawn in Autocad (http://www.autodesk.co.uk/products/autocad/overview). The U-

value calculations have been carried out following ANSI/NFRC 100 (2014). State-of-the-

art computer software developed at Lawrence Berkeley National Laboratory has been

used: WINDOW 6.3.9.0 (http://windows.lbl.gov/software/window/window.html) to

model the glazing build-up and calculate the U-value at the centre-of-glazing. The glazing

has been then imported into THERM 6.3.19.0

(http://windows.lbl.gov/software/therm/therm.html) to be modelled with the rest of

components. This software has been selected because it has been especially tailored to

model products for NFRC certification. It provides flexibility to model bespoke details by

allowing to import CAD drawings. It is a robust software developed by a prestigious

university and funded by the United States Government. It is available for free on line and

widely used in industry.


92

Figure 28: Process map

The U-value of the glazing edge and the frames has been obtained for each of the main

cross-sections of the conventional and the proposed systems. The overall façade U-value

has been calculated by area weighting these U-values in accordance with the equation

shown below taken from THERM 6.3 / WINDOW 6.3 National Fenestration Rating

Council Simulation Manual (Lawrence Berkeley National Laboratory, 2011).

Ut =Σ(Uf∗Af)+Σ(Ue∗Ae)+Σ(Uc∗Ac)

At [8]

Where:

Ut = Total U-value [W/m2K]

At = Total area [m2]

Uf = Frame U-value [W/m2K]

Af = Frame area [m2]

Ue = Edge-of-glazing U-value [W/m2K]

Ae = Edge-of-glazing area [m2]

Uc = Centre-of-glazing U-value [W/m2K]

Ac = Centre-of-glazing area [m2]


93

Assumptions regarding elevation dimensions, material properties and environmental

conditions are identical for both systems compared.

Figure 29 represents a schematic elevation of a curtain wall unit with the dimensions that

have been considered for the calculation of the frame, the edge-of-glazing and the centre-

of-glazing areas according to the proposed design and NFRC standards.

Figure 29: Projected areas in elevation

The IGU has been modelled as triple glazing with 16 mm Argon-filled cavities and high

performance coatings. Table 5 describes the precise build up based on products which are

available in industry. The centre-of-glazing U-value result is 0.67 W/m 2K.

Mullion edge of glazing width Mullion width

Centre of glazing area

Transom edge of glazing width

Transom width


94

Table 5: WINDOW 6.3.9.0 modelled IGU description and centre-of-glazing results

Table 6 : Material properties as modelled in THERM 6.3.19.01

Figures 30 and 31 represent the main cross sections of the conventional system and the

proposed system respectively as modelled indicating the materials. It should be noted

that, being a 2D analysis, there are elements that are not modelled such as weep holes,

setting blocks, corner keys, etc. These elements are usually disregarded in U-value

calculations as their influence is negligible. Each material has been modelled with a

Material Component Thermal Conductivity λ [W/mK]

Emissivity ε [-]

Aluminium alloy (anodised) Frame 160 0.90

Butyl rubber Spacer primary seal 0.17 0.90

Cavities modelled as Frame cavity NFRC 100 or Frame cavity Slightly Ventilated NFRC 100

Ethylene propylene diene monomer Gasket 0.25 0.90

Fibreglass Frame 0.30 0.90

Glass (soda lime) Stepped glass 1.0 0.84

IGU imported from WINDOW 6.3.9.0. with 0.67 W/m 2K centre-of-glazing U-value

Polyamide (nylon) Thermal break 0.25 0.90

Silica gel (loose fill) Spacer desiccant 0.13 0.90

Silicone Adhesive, sealant 0.35 0.90

Stainless steel Vapour barrier foil 15 0.20

Structural adhesive Structural adhesive 0.4 0.90

Styrol acryl nitrile copolymer Spacer main body 0.16 0.90


95

determined thermal conductivity λ [W/mK] and emissivity ε as shown in table 6. A

thermal conductivity of 0.4 W/mK has been assigned to a standard structural adhesive.

The design of the spacers is based on the information provided by a recognised warm-

edge spacer manufacturer (Swissspacer, 2008) with a main body of composite plastic and

a stainless steel foil that functions as vapour barrier.

Figure 30: Conventional system as modelled in THERM 6.3.19.0 (a) mullion and (b) transom

Ethylene propylene diene monomer gasket

Polyamide (nylon) thermal break

Structural silicone

Aluminium alloy (anodised) frame

Stainless steel vapour barrier (0.01 mm)

Styrol acril nitril copolymer spacer (1 mm)

Rubber butyl primary seal

Stepped glass

Cavity

Insulating glazing unit

(a)

(b)

Silica gel dessicant


96

Figure 31: Proposed system as modelled in THERM 6.3.19.0 (a) mullion and (b) transom

(b)

(a)


Cavity slightly ventilated

Glass fibre reinforced polyester frame


Styrol acril nitril copolymer spacer (1 mm)


Stepped glass

Cavity non ventilated



Silicone sealant Structural adhesive


97

The conditions assumed are NFRC standardized environmental conditions for U-factor

calculations for product ratings listed in table 7.

Table 7: Environmental Conditions for NFRC Simulations for U-factor calculations

Variable Assumed values

Outside Temperature -18 ºC

Inside Temperature 21 ºC

Wind Speed 5.5 m/s

Wind Direction Windward

Direct Solar 0 W/m2

Sky Temperature -18 ºC

Sky Emissivity 1.00

For condensation analysis the same procedure has been followed except for the

environmental conditions, for which typical conditions would normally be assumed based

on the climate at the building location and the use of the building. In this case, -5 ºC have

been assumed as external temperature while typical office conditions of 21 ºC and 40%

have been assumed as internal temperature and internal relative humidity.Condensation

occurs on a surface if the surface temperature is below the dew-point temperature. A

temperature profile has been generated through each cross-section analysed to obtain the

inside surface temperature. Based on the assumed internal temperature of 21 ºC and

relative humidity of 40%, a dew-point Temperature of 279.9 K (6.9 ºC) has been derived.


98


Heat flow and U-values for the conventional and proposed systems are described in tables

8 and 9. Against the conventional system, the proposed system achieves a reduction of the

heat flow through the frame area of 45% while the heat flow through the edge of glazing

area is increased by 33%. The total area-weighted U-value of the system is reduced by

10%.

Heat flow profiles of representative transom and mullion sections for conventional and

proposed systems are illustrated in figure 32. The heat flow across the conventional

system is unevenly distributed with the aluminium frame concentrating large peak values

reaching over 2000 W/m2. Where the aluminium frame is thermally broken, the heat flow

is bypassed through the edge of the IGU, mainly through the steel vapour barrier and the

structural silicone. The heat flow across the proposed system is more evenly distributed

with highest values just over 300 W/m2 located at the stainless steel vapour barrier, the

GFRP frame and the structural adhesive. While the conventional system concentrates the

heat flow at the frame area, the proposed system distributes the flow between the frame

and the edge of glazing areas.

Table 8: Heat flow comparison between proposed and conventional systems

1.54

0.85

0.42

0.63

2.84

2.84

0.00

1.00

2.00

3.00

4.00

5.00

6.00

Conventional system Proposed system

He

at

flo

w [

W/°

K]

Centre of glazing

Edge of glazing

Frame


99

Table 9: U-value comparison between proposed and conventional systems

Temperature profiles of representative transom and mullion sections for conventional and

proposed systems are illustrated in figure 33. Eliminating the frame protrusions to the

inside in the proposed system implies that less surface is exposed to the inside. Moreover,

the distance between the external and internal surfaces is reduced provoking a steeper

temperature gradient. These facts result in lower inside surface temperatures for the

proposed system than for the conventional system. For the assumed environmental

conditions, the lowest surface temperature is 10.2 ºC and is located at the transom gasket.

This temperature is still above the calculated dew-point temperature of 6.9 ºC so there is

no risk of condensation.

3.62

0.70 0.67

0.91

1.99

1.05

0.670.82

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

Frame Edge of

glazing

Centre of

glazing

Total

U-v

alu

e [

W/m

2°K

]

Conventional system

Proposed system


100

Figure 32: Heat flow profiles for (a) conventional system mullion (b) proposed system mullion (c)

conventional system transom (d) proposed system transom

(a)

(b)

(c)

(d)

frame

frame

glass edge

glass edge

glass edge

glass edge

Structural silicone

Aluminium frame

Stainless steel vapour barrier

GFRP frame and structural adhesive


frame

glass edge

glass edge

frame

glass edge

glass edge


101

Figure 33: Temperature profiles and location of areas with minimum inside surface temperature for

(a) conventional system mullion (b) proposed system mullion (c) conventional system transom (d) proposed system transom

16.8 ºC

10.2 ºC

17.4 ºC

11.8 ºC

(a)

(b)

(c)

(d)


102

5.3 Conclusion

Against the conventional system, the proposed system achieves a total area-weighted U-

value reduction of 10%. This is achieved due to substantial improvement at the frame

area and despite substantial worsening at the edge of glazing area. The improvement

provided by the proposed system against the conventional system would be more

pronounced if the proportion of frame area against glazing area was higher, which is

generally the case with most curtain wall units comprising intermediate frames.

Moreover, as the thermal performance of glazing improves in the future, the relevance of

the performance of the frame will increase further.

The heat flow across the conventional system is unevenly distributed with the aluminium

frame concentrating large peak values. Where the aluminium frame is thermally broken,

the heat flow is bypassed through the edge of the IGU, mainly through the steel vapour

barrier and the structural silicone. The heat flow across the proposed system is more

evenly distributed with moderate peak values located at the stainless steel vapour barrier,

the GFRP frame and the structural adhesive. While the conventional system concentrates

the heat flow at the frame area, the proposed system distributes the flow between the

frame and the edge of glazing areas.

The proposed system presents lower temperatures in the inside surfaces than the

conventional system. This is due to the fact that the proposed system is narrower,

provoking a steeper temperature gradient. Moreover, by eliminating the frame protrusions

to the inside, the heat transfer surface of the frame is reduced. Nevertheless, the

temperature is still above the calculated dew-point temperature so there is no risk of

condensation.

While the performance of the conventional system is close to its full potential, the design

of the proposed system is at a schematic stage and there is still scope for optimization.

The design could be improved by reducing the thickness of the GFRP frame web.

However, doing this would increase the shear stress in the GFRP frame web as seen in the

previous chapter. Integrating the frame with the spacers could also reduce the thermal

transmission.


103

GFRP FRAME SELECTION BY 4-POINT

BENDING TESTS

Four-point bending tests are performed on glass fibre reinforced polyester resin and glass

fibre reinforced phenolic resin specimens, some of the specimens being previously heat

soaked. The results provide information on the shear strength and the time-dependent

modulus of elasticity of the tested materials.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


104


105

6.1 Candidate materials

Forty pultruded GFRP bars of dimensions 150 mm x 20 mm × 5 mm with the glass fibres

in the longitudinal direction as shown in figure 34 were tested.

Figure 34: 4-point bending test specimen

Two types of GFRP were chosen as the candidate materials for the frame: glass fibre

reinforced polyester resin and glass fibre reinforced phenolic resin. These two types of

composite differ in the resin used as matrix. Their principal advantages are listed on table

10 below.

Table 10 : Advantages of glass fibre reinforced polyester resin and glass fibre reinforced phenolic

resin (Hartley, 2002)

Some of the specimens were stored and tested at ambient conditions while others were

heat soaked in an oven at 130ºC for 30 minutes and then cooled down to ambient

conditions prior to testing. This was done to identify variations in performance after being

subjected to high temperatures. It is worth noting though, that the performance has been

measured after and not during the time the specimens were subjected to high

temperatures. This decision was made due to the difficulty to manipulate hot specimens

and achieve consistent temperatures while testing. The temperatures applied in the test

could typically be caused by solar radiation or adhesive curing processes while the

Glass fibre reinforced polyester resin Glass fibre reinforced phenolic resin

Very versatile Can be fine-tuned to particular specifications Low cost Good physical / mechanical properties Good electrical properties Excellent pigmentability Good chemical resistance

Excellent fire performance Excellent temperature performance Very Low smoke

Direction of the glass fibres


106

temperatures caused by fires could reach over 1000°C (ASTM E 119). Therefore, the test

is valid to confidently predict serviceability performance after temperatures tipically

caused by solar radiation or to validate temperatures of up to 130°C maintained during 30

minutes in the manufacturing process. The test is not valid to assess integrity or measure

performance in the event of a fire. Table 11 shows the full scope of testing specifying the

amount of specimens tested for each matrix resin type and whether they were heat soaked

before testing or not.

Table 11 : Scope of testing for 4-point bending test

Matrix resin type Polyester resin Phenolic resin

Ambient conditions prior

to testing Non heat soaked

Heat soaked Non heat soaked Heat soaked

Number of specimens 10 10 10 10

6.2 Method

Four-point bending tests were performed instead of three-point bending tests to prevent

stress concentrations in the middle region of the GFRP bars, what could cause cracks. The

tests were performed on an Instron 5567 testing machine with a 30 KN load cell at a

loading rate of 2 mm/minute. The GFRP bar was placed on two round supports of equal

height and spaced 135 mm apart. The two round supports were placed such that their

centre line aligned with the centre line between the two crossheads connected to the

Instron. The two crossheads were spaced 75 mm apart. A stiff steel strip was clamped in

the centre of the GFRP using a toolmaker clamp so that the displacement in the centre of

the GFRP could be measured by placing a displacement gauge on the steel plate. Another

displacement gauge was placed on the crosshead to double check the displacement

reading from the Instron.

The test setup is shown in figure 35. Figure 36 describes the load distribution used in the

test and the resulting shear and moment diagrams. Figure 37 describes the geometry of

the specimens’s cross-section and its properties indicating the location of the cut line

considered for shear calculations.


107

Figure 35: Specimen being tested on Instron 5567 machine

Figure 36: Structural diagrams (a) cross-section section showing load distribution and dimensions between supports and crossheads; (b) shear distribution and (c) moment distribution for simply

supported beam

Instron applied load distributed to two

crossheads

First displacement gauge at steel strip clamped to the

centre of the GFRP bar

Second displacement gauge at crosshead

(c)

Vmax = P

Mmax = 30mm x P

(b)

(a)


108

Figure 37: Geometric properties of the specimen’s cross-section and cutline

6.2.1 Shear strength calculation

The shear strength is calculated through basic beam theory principles as per equation 9:

τbeam = VACy′

Ia [9]

Where:

τbeam is the beam shear stress at a given location

V is the shear force which corresponds to half of the load applied by the Instron at the

moment of failure (figure 35b)

AC is the area of the segment above the cut line (figure 37)

y′ is the distance from centre of area above the cut line to centroid of whole section

(figure 37)

I is the second moment of area of whole section (figure 37)

a is the breadth of the section at the cut line being considered (figure 37)

6.2.2 Young’s Modulus calculation

The Modulus of Elasticity is calculated through basic besam theory principles as per

equation 10:

E =MR

I [10]

Second moment of area I= 208.33 mm4 Distance from centre of area above the cut line to centroid of whole section y’ = 1.25 mm


109

Where:

E is the Modulus of Elasticity

M is the applied moment (figure 35c)

I is the second moment of area (figure 37)

R is the radius of curvature, calculation described in figure 38 where:

x is half the distance between crossheads

y is the displacement read by the gauges

A = tan−1(𝑥/𝑦)

B = 180° − 2A

R =x

SinB

Figure 38: Radius of curvature diagram

Displacement gauges recorded readings every 0.25 seconds and for each instant the

displacement reading was taken, the modulus of elasticity is calculated. For the initial

displacement readings, the calculated modulus of elasticity fluctuates considerably in the

range from 100 GPa to infinity. This is due to the large percentage error in displacement

readings when the displacement is small. Therefore, all the readings that give a modulus

of elasticity higher than 100 GPa are removed. In addition, when the load begins to

decrease, it is a sign that the GFRP starts to behave plastically. All the readings after the

load starts dropping are also excluded.


110

6.3 Results and Discussion

All specimens failed to horizontal shear stress provoking delamination in the specimens

as shown in figure 39.

Figure 39: Horizontal shear stress failure in (a) polyester resin and (b) phenolic resin specimens

The average shear strength test results are summarised in figure 40. The average shear

strength was very close for all four specimen types tested, ranging between 17 MPa and

19 MPa. Without prior heat soaking, it was 9% higher for the phenolic specimens than for

the polyester specimens. By heat soaking prior to testing, the shear strength of the

polyester specimens increased slightly while that of the phenolic specimens decreased

resulting in an average shear strength 4% higher for the polyester specimens than for the

henolic specimens. The modulus of elasticity test results are plotted in the following

pages.

Figure 40: Shear strength summary results

Delamination provoked by shear stress

(a) (b)

17.47 17.76

19.04

17.08

1011121314151617181920

Polyester resinNon heat-soaked

Polyester resinHeat soaked

Phenolic resinNon heat-soaked

Phenolic resinHeat-soaked

Shea

r Stre

ngth

[MPa

]


111

Table 12: Heat soaked phenolic resin modulus of elasticity vs. time curves

Table 13: Non heat soaked phenolic resin modulus of elasticity vs. time curves

Exponential fit E = Eve−βt + E∞ (95% confidence bounds) Ev = 48.38 (47.66, 49.1) β = 0.01681 (0.01635, 0.01726) E∞ = 29.11 (28.88, 29.34) R-squared: 0.7808 / Adjusted R-squared: 0.7807

Power fit E = atb + c (95% confidence bounds) a = 149.8 (147.3, 152.4) b = -0.2118 (-0.2332, -0.1904) c = -17.32 (-23.5, -11.14) R-squared: 0.7637 / Adjusted R-squared: 0.7636




112

Table 14: Heat soaked polyester resin modulus of elasticity vs. time curves

Table 15: Non heat soaked polyester resin modulus of elasticity vs. time curves


Power fit E = atb + c (95% confidence bounds) a = 274.9 (262.7, 287) b = -0.1696 (-0.2041, -0.135) c = -79.37 (-104.5, -54.22) R-squared: 0.6482 / Adjusted R-squared: 0.6481

Exponential fit E = Eve−βt + E∞ (95% confidence bounds Ev = 63.73 (62.63, 64.83) β = 0.01369 (0.01331, 0.01407) E∞ = 23.35 (23.04, 23.66) R-squared: 0.6584 / Adjusted R-squared: 0.6583



113

Best fit lines are calculated for each of the four specimen types and plotted with two

different equations:

Exponential fit 𝐸 = 𝐸𝑣𝑒−𝛽𝑡 + 𝐸∞ where 𝐸𝑣 = 𝐸0 − 𝐸∞. Previous empirical

evidence from material viscoelasticity suggests that this equation represents the

correct behaviour. However, it is clearly shown in the following plots that this

equation does not represent the behabiour of the GFRP material well when the

displacement readings are small. It tends to understimate the modulus of elasticity

GFRP.

Power law 𝐸 = 𝑎𝑡𝑏 + 𝑐. Power law is introduced since it represents better the

behaviour for small displacement readings.

By comparing adjusted R-square, it is clear that the exponential equation fits better than

the power equation, but not by much. Generally, the exponential equation fits better when

t > 100s while the power equation fits better when t < 100s. Therefore, for the purpose of

calculating bending induced by wind loads, the modulus of elasticity should be calculated

using the power fit equation.

The modulus of elasticity summary of results can be seen in figure 41. The modulus of

elasticity applicable to short term loads is notably higher than that applicable to long term

loads. The short term load was established in 5 seconds since the readings were not

reliable below this load duration due to the large percentage error when measuring small

displacements. This difference is especially remarkable in polyester specimens where the

stiffness is reduced by over 70% for long term loads. It reduces from 104 MPa to 23 MPa

in non heat soaked specimens and from 130 MPa to 25 MPa in heat soaked specimens. In

phenolic specimens the stiffness is reduced by over 50%, from 69 MPa to 30 MPa in non

heat soaked specimens and from 81 MPa to 29 MPa in heat soaked specimens. The

polyester specimens are stiffer than the phenolic specimens to short term loads but less

stiff to long term loads.

Heat soaking prior to testing did not affect the stiffness to long term loads of neither the

polyester nor the phenolic specimens. It did, however, increase the stiffness to short term

loads of the polyester specimens by 25% and of the phenolic specimens by 17%.


114

Figure 41: Modulus of elasticity summary results

6.4 Conclusion

On the ground that there were no significant divergences between the mechanical

properties of the polyester and the phenolic specimens even after heat soaking and

considering its lower cost and better appearance, polyester is selected as the matrix for the

GFRP frame.

All specimens failed to horizontal shear stress provoking delamination in the specimens.

Moreover, the average shear strength was, for all four specimen types tested, ranged

between 17 MPa and 19 MPa. This is below the 25 MPa indicated by the manufacturer

(Fiberline, 2003). These results, together with the results of the analytical predictions,

point out shear strength of GFRP as a potential weakness that needs to be addressed in the

design. The thickness of the GFRP web will need to be increased despite the fact that

doing this will increase the thermal transmittance.

The modulus of elasticity for permanent loads ranged between 23 MPa and 30 MPa

depending on the specimen. These results are in line with the information provided by the

manufacturer (Fiberline, 2003). However, for instant loads, the values ranged between 69

MPa and 130 MPa, duplicating and is some cases triplicating the permanent load values.

Polyester resin

Non heat-

soaked

Polyester resin

Heat soaked

Phenolic resin

Non heat-

soaked

Phenolic resin

Heat-soaked

E(t=5) 104.29 129.86 69.31 89.21

E(t=infinity) 23.35 24.62 29.72 29.11

0

10

20

30

40

50

60

70

80

90

100

110

120

130

140

E (

GP

a)


115

Therefore, it is important to establish the duration of the load that should be contemplated

for wind load as this has a large effect on the modulus of elasticity of GFRP.

It is worth noting the following considerations:

The Modulus of Elasticity results for the short duration loads are not as reliable as

it would be desirable because of the large displacement error in the measured

displacements

The performance has been measured after and not during the time the specimens

were subjected to high temperatures. The test is valid to confidently predict

serviceability performance after temperatures typically caused by solar radiation

but not valid to assess integrity or measure performance in the event of a fire


116


117

ADHESIVE SELECTION BY SINGLE-LAP

SHEAR TESTS

Single lap shear tests are performed on bonded connections between glass and glass fibre

reinforced polyester substrates using a range of candidate adhesives. Two phases of

testing take place with an intermediary analysis and adjustments in the design of the

connections. The results provide information on the failure mode and a typical load

versus shear displacement curve for each of the candidate adhesives.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


118


119

7.1 Preliminary selection of candidate adhesives

The choice of adhesive is of vital significance for the design of load bearing glass

structures. Besides the mechanical properties of the adhesive and substrates, there are

many other considerations that need to be taken into account such as thickness of the

bond, exposure to UV radiation, temperature and moisture changes, creep under long

term loads or application process. These considerations are summarised in table 16 and

explained in the following paragraphs in more detail.

Table 16: Considerations and target performance required from candidate adhesives

Shear resistance 5-10 MPa Thickness 2-3 mm UV resistance Reasonable durability (can be protected by applying frit to the glass)

Temperature variation Range from -20°C to 80°C Retain 75% of shear strength at 80°C while not too brittle at -20°C Good durability

Exposure to moisture Dimensional stability Good durability

Long term load Adhesive with limited creep and visual signs before failing Provide dead load fail-safe mechanism in case of failure of the bond

Execution Easy and economical process that allows quality control

The shear stress in the adhesive caused by windload is 3.31 MPa, as per the analytical

calcutions in previous chapters. The candidate adhesives should provide a higher shear

strength value to provide some flexibility to accommodate more onerous wind loads,

thermal loads, unit geometry or safety factors. The target values considered desirable

range between 5 to 10 MPa.

The thickness of the adhesive is a fundamental design decision. By increasing it, the

deformation capacity of the bond is increased. Conversely, the shear stiffness and

brittleness are reduced. In this application the minimum permissible thickness is dictated

by the out-of-plane manufacturing tolerances of glass and GFRP. A bond thickness of 2

to 3 mm is considered.


120

This is an external application exposed to UV radiation and humidity variations. The

adhesive can be protected from UV radiation by applying a ceramic frit on the glass

surface and can be protected from exposure to moisture by using sealants. However, the

design should not rely solely on these measures and reasonable durability against each of

these factors is required from the adhesive. Moreover, the adhesive should provide

dimensional stability against moisture changes.

The temperature can range from -20°C to +80°C. At high temperatures adhesive tend to

lose stiffness and strength while at low temperatures they tend to increase stiffness and

brittleness. It is a requirement for the candidate adhesives that they retain at least 75% of

their shear strength at +80°C.

When subjected to static actions, many adhesives will suffer from creep or static fatigue,

causing continued deformations without change in load. Consequently, adhesive

connections may fail even if the theoretical maximal stresses are not reached. There is not

much information about performance under long term permanent loads at present. There

are experimental investigations under way at the University of Cambridge but the results

are not yet available. Until then, the system is to be designed with a fail safe mechanism

that would take the glass dead load in case of failure of the bond. In addition, it would be

useful to select an adhesive that gives out visual signs such as changes in coulour before

failing.

It is noteworthy that most adhesives are relatively sensitive to execution and setup

imperfections: poorly executed adhesive bonds may have a significantly decreased

lifetime or a lower resistance. Therefore, an adhesive that is easy to apply and whose

application is easy to monitor and control is required.

Many properties of the adhesive are unknown or are not accurately described by the

manufacturer. Therefore, the selection process is based on a combination of the

manufacturers’ technical datasheets and previous research on adhesive-metal and

adhesive-GFRP bonds. One acrylate, two epoxies and one silicone have been selected to

provide a varied range of candidate adhesives.


121

7.1.1 Acrylate: Huntsman Araldite 2047

Acrylates are suitable to produce adhesive connections with high shear resistance of up to

20 MPa to 30 MPa when applied in their optimum application thickness, which is

typically below 0.5 mm. At the same time, the brittle nature of this type of adhesives may

provoke sudden failure without significant prior deformation, what is not desirable. For

this application, a minimum bond thickness of 2 mm is considered necessary to cope with

the manufacturing tolerances of glass and GFRP. By increasing the thickness, the shear

strength is likely to decrease while the joint becomes more flexible. Application is

normally easy and the curing time required is relatively short. Acrylates may be sensitive

to temperature and moisture changes.

Nhamoinesu and Overend (2012) performed a series of single lap shear tests on various

steel-adhesive-toughened glass joints. Among the six adhesive tested, Araldite 2047

showed the best results. All specimens failed cohesively after substantial plastic

deformation. The joints were relatively flexible yet they carried significantly high loads.

According to the manufactuer’s technical datasheet (Huntsman Advanced Materials,

2007), Araldite 2047 – GFRP bonds can achieve a lap shear strength of 6 MPa.

7.1.2 Epoxies: 3M Scotch-Weld DP 490 and 2216 B/A

Similarly to acrylates, epoxies are very strong and stiff adhesives and are typically

applied in very thin layers. The strength of epoxies strongly depends on the specific

product and may reach characteristic values of up to 30 MPa. Two disadvantages are the

brittle nature of the adhesive joint and limited deformation capacity to deal with

differential thermal expansions. Again, by having a 2 mm thich bond, the shear strength is

likely to decrease while the joint becomes more flexible. Moreover, efforts have been

done lately to modify epoxies, e.g. with rubber particles, to increase their toughness and

to make them a more elastic material. These hybrid epoxies are also known as toughened

epoxies, as they are tougher and do not break in the same brittle way traditional epoxies

do. In terms of dependency on temperature and humidity, epoxies will generally be less

sensitive compared to acrylates. In some cases extra attention should be paid to pre-

treatment. The curing time will normally be rather long, and warm hardening epoxies are

exposed to a risk of stresses due to shrinking. In general, the brittle nature of traditional

epoxies necessitates a connection design based on the avoidance of peak stresses.

According to the manufacturer technical datasheet (3M, 1996), 3M Scotch-Weld DP 490

shows good adhesion to many plastic surfaces even by simply solvent wiping. Its bond


122

with glass fibre reinforced phenolic resin fails at 30.3 MPa by cohesive failure. In

addition, Stefan Peters (2006) performed a series of tests on GFRP-glass joints where 3M

Scotch-Weld DP 490 ranked first in fracture tensile strength, fracture shear strength,

fracture tensile strength at 80°C and fracture shear strength at 80°C, and second in

fracture tensile strength after natural ageing, fracture tensile strength at -20°C and fracture

shear strength at -20°C. Although Nhamoinesu and Overend (2012) deemed DP 490

unsuitable for steel-glass bond applications due to its lack of flexibility and lack of plastic

deformation, GFRP is much more flexible than steel, what could substantially increase

the flexibility of the joints.

3M Scotch-Weld 2216 B/A was tested in a second phase. It has similar properties to 3M

Scotch-Weld DP 490. According to Nhamoinesu and Overend (2012), 3M Scotch-Weld

2216 B/A specimens showed good flexibility but the load carrying capacity was relatively

low with maximum loads of only 7.3 kN due to adhesion failure at the steel-adhesive

interface. Despite premature joint failure, it deformed significantly before failure, what is

ideal for accommodating building tolerances and differential thermal expansion between

substrates.

7.1.3 Dow Corning TSSA Silicone

Silicones are characterized by their relatively low characteristic strength, typically about

0.5 MPa to 1.5 MPa in tension, and a relatively low stiffness. Because of their limited

stiffness, silicones can compensate very well building tolerances and they have good

resistance to differential thermal expansions. Silicones are normally applied with a joint

thickness of at least 6 mm. Moreover, they can be applied in a broader temperature range

than most other adhesives, and they have excellent resistance against UV radiation,

ozone, humidity and other external exposures. The most important drawbacks of silicones

are the low curing velocity of one component systems, compatibility issues with certain

coatings and laminated glass interlayers such as polyvinyl butyral (PVB), and silicones

being a possible cause of corrosion for the substrates. The relatively low strength may be

a major disadvantage as well.

A transparent structural silicone adhesive (TSSA) film has been recently developed. Its

strength is considerably higher than that of traditional silicone products. According to the

manufacturer, its shear strength ranges from 4 MPa to 5 MPa. It is applied in a sheet

format which is 1mm thick. For the purpose of this application, TSSA needs to be applied

in 3 mm thick. Hence, three layers need to be applied and its shear strength will most


123

likely be compromised. Finally, it needs to be cured at 130°C and 0.7 MPa for 30

minutes.

7.2 Method

7.2.1 Pre-dimensioning of the bond area

A preliminary check was carried out on the area of the adhesive bond. Assuming an even

distribution of stresses, a bonded connection of 40 mm x 30 mm would be sufficiently

small to ensure that cohesive failure of the bond would occur before failure of the glass or

the GFRP substrates.

Table 17: Preliminary check of the dimensions of the bond

Max. load glass pane = 167 kN 𝐭𝐞𝐧𝐬𝐢𝐥𝐞 𝐬𝐭𝐫𝐞𝐧𝐠𝐭𝐡

𝐦𝐚𝐭𝐞𝐫𝐢𝐚𝐥 𝐟𝐚𝐜𝐭𝐨𝐫 𝒙 𝒂𝒓𝒆𝒂

Max. load GFRP bar = 30 kN 𝒕𝒆𝒏𝒔𝒊𝒍𝒆 𝒔𝒕𝒓𝒆𝐧𝒈𝒕𝒉

𝒎𝒂𝒕𝒆𝒓𝒊𝒂𝒍 𝒇𝒂𝒄𝒕𝒐𝒓 𝒙 𝒂𝒓𝒆𝒂

Permissible load bond = 24 kN 𝒔𝒉𝒆𝒂𝒓 𝒔𝒕𝒓𝒆𝒏𝒈𝒕𝒉 𝒙 𝒂𝒓𝒆𝒂

7.2.2 Specimen Preparation

The single lap shear test specimens consisted of two pultruded glass fibre reinforced

polyester bars bonded to opposite sides of a toughened glass pane as shown in figure 34.

Figure 42: Single-lap shear test specimen

Assuming: Tensile strength = 120 MPa Material factor = 1.8 Area = 250 mm x 10 mm

Assuming: Tensile strength = 240 MPa Material factor = 1.6 Area = 40 mm x 5 mm

Assuming: Shear strength = 20 MPa Area = 40 mm x 30 mm

Toughened glass Pultruded glass fibre reinforced polyester bar Adhesive

bond


124

7.2.3 Huntsman Araldite 2047, 3M Scotch-Weld DP 490 and 2216 B/A

The preparation of the specimens with the adhesives Huntsman Araldite 2047 Acrylate,

3M Scotch-Weld DP 490 and 2216 B/A was very similar. Two threaded holes 3.5 mm

deep were drilled near the short edge of the bars to be able to attach the displacement

transducers during testing. Another smaller hole was drilled through the centre of the

adhesive bond region to allow extra adhesive to flow out when this was applied.

Figure 43: Drilling of two threaded holes in the GFRP bar to attach the displacement transducers and a smaller hole to allow extra adhesive to flow out when applied

It was necessary to fabricate bespoke jigs and blocks to apply the adhesive in a consistent

manner across all the specimens. The jigs and blocks were designed to fit the glass

precisely leaving an exposed the area where the adhesive would be applied. The jigs

ensured alignment in plan and avoided the adhesive from spilling around the edges. The

blocks were used to control the depth of the adhesive bond. Several blocks with different

heights were fabricated to be able to produce bonds with different thicknesses. The jigs

and blocks were manufactured using modeling board and were coated with

polytetrafluoroethylene (PTFE) to prevent the adhesive from sticking to them. Figure 44

illustrates their dimensions and how they were adjusted to the glass panes.

Transducer clamped to steel plate screwed to pultruded bar

Excess adhesive flowing out


125

Figure 44: Adjustment of jigs and blocks to the glass pane (a) photograph (b) schematic section and (c) 3D sketch

Block with varied heights depending on the thickness of the adhesive bond Jig

Toughened glass pane

Area left to apply the adhesive

PTFE wrapped block

PTFE wrapped jig

(a)

(b)

(c)


126

The steps listed below were followed to apply the adhesive and are illustrated in figure

45:

1. The glass pane and the GFRP bars were thoroughly cleaned with acetone;

2. The jigs and the blocks were fitted onto the glass pane;

3. A gun of particular mixing ratios was used to apply the adhesive in excess;

4. The GFRP bars were placed gently on the adhesive;

5. Weights were placed on top of the GFRP bars so that the GFRP bars rested fully

on the underlying block, thus achieving the proposed thickness. In the process, the

adhesive filled all the gaps and the excess flowed out through the drilled hole;

6. The adhesive was left to cure as indicated by the manufacturers’ technical

datasheet to reach handling strength;

7. After the specimen reached handling strength, the jigs and blocks were removed

and any adhesive overflow was cut off;

8. The adhesive was left to cure further to reach full strength as indicated by the

manufacturers’ technical datasheet.

Figure 45: Adhesive application process

1(a)

1(b)

2

3

7

5(b)

5(a)

4


127

After curing, two metal L plates were bonded to the glass to provide a perpendicular

surface upon which the displacement gauges could be placed.

Figure 46: Bonding of two metal L plates to place displacement gauges

7.2.4 Dow Corning TSSA

The adhesive Dow Corning TSSA requires pressure and temperature for curing. This led

to having to make some modifications to the preparation of the specimens described in

the previous section.

The use of an autoclave is recommended by the manufacturer to achieve the conditions of

0.7 MPa and 130°C. However, since an autoclave was not available at the time of testing,

the manufacturer suggested an alternative procedure. This consisted in pre-pressing the

bond to ensure sufficient substrate contact and introducing the specimen in the oven at

130°C for 30 minutes. Pressure was applied through the use of clamps but could not be

measured. The block had to be adjusted to be able to compress the adhesive. Two rubber

plates were glued at the short edges of the top surface of the block protruding 3 mm. The

function of this rubber was to allow the application of pressure on the adhesive while still

preventing the adhesive from flowing out. This modified arrangement is shown in figure

47.

Metal L plates bonded to glass with UV curing adhesive

UV light lamp


128

Figure 47: Adjustment of jigs and blocks to the glass pane for Dow Corning TSSA specimens (a and

b) photographs (c) schematic section

(a) (b)

(c)

Clamp to apply pressure on the adhesive

Clamp to keep the jig tightly fixed to the glass pane


129

The steps listed below were followed to apply Dow Corning TSSA:

1. 1. The glass pane and the GFRP bars were thoroughly cleaned with acetone and

allowed to dry;

2. The jig was fitted onto the glass pane and clamped tightly to it using a toolmaker

clamp;

3. The TSSA was taken from the refrigerator, laid flat and allowed to equilibrate

with ambient temperature for 30 minutes;

4. 92-023 primer was applied to the glass and the GFRP surfaces to be bonded with a

fine, particulate free paper cloth (saturated with the primer) and allowed to dry for

5 minutes;

5. The TSSA sheet was cut into 40 mm x 30 mm rectangles;

6. The bottom liner of the TSSA was removed and the TSSA was placed carefully on

the glass surface. Any air bubbles were removed by pressing gently. Then the top

liner was removed. This step was repeated three time to achieve a triple layer;

7. The GFRP bar was gently place on top of the adhesive;

8. Steel plates were positioned above and below the assembly;

9. Pressure was applied through clamps;

10. After repeating the same procedure on the other side of the glass, the whole

assembly was introduced in a preheated oven at 130°C for 30 minutes;

11. The assembly was removed from the oven and left to cool down for 30 minutes

before removing the jigs and auxiliary components. Any adhesive overflow was

cut off.

7.2.5 Testing Procedure

The single lap shear tests were performed on an Instron 5500R testing machine with a

150 kN load cell. Specimens were clamped to the testing machine. A displacement gauge

was attached to the end of each GFRP bar with the gauge probe resting on a steel L plate

glued 80 mm from the edge of the glass as shown in figure 48. The displacement gauge

was attached to the end of the GFRP bars to exclude the effect of GFRP elongation in the

measurements. The displacement gauges measured the vertical displacement in each

adhesive joint separately. All tests were displacement controlled with a displacement rate

of 0.2 mm/minute. Photographs were taken before, during and after each test.


130

Figure 48: Single-lap shear test setup (a) schematic elevation (b) specimen being tested on Instron 5500R machine

All specimens were tested to failure and the reason for failure was classified into one of

the following categories:

Substrate failure: Collapse of the glass member due to locally exceeding shear or

tensile strength of the substrate. Usually this case is considered to be favourable,

because the strength of the adhesive will not be the governing factor for the design

of the connection and better-known strength values can be used.

Cohesion failure: Failure of the adhesive layer as a result of exceeding shear or

tensile strength of the adhesive itself.

Adhesion failure: Slippage or ripping of the adhesive layer from one of the

substrates due to insufficient adhesion to the substrate.

The mean load, extension and shear stress at failure were also registered. Testing was

carried out in two phases with intermediary analysis and adjustments in the specimens as

described in table 18.

(a)

(b)

Steel L plate

Displacement transducer

GFRP bar Clamp

Toughened glass pane


131

Table 18 : Scope of testing for single-lap shear test

PHASE 1 Adhesive type Acrylate Epoxy Silicone

Manufacturer Huntsman 3M Scotch-Weld Dow Corning

Name Araldite 2047 DP 490 TSSA

Bond thickness 3 mm 3 mm 3 mm

Number of specimens 10 10 10

INTERMEDIARY ANALYSIS Poor adhesion at GFRP-adhesive

interface provokes slippage

High stiffness of the adhesive provokes

high stress concentrations that

break the glass

Invalid results due to inadequate

specimen preparation

ADJUSTMENTS The GFRP surface was abraded to

increase adhesion at the GFRP-

adhesive interface

The thickness of the bond was increased to 5 mm to reduce the stiffness of the

connection

No further testing without autoclave.

Adhesive temporarily

disregarded and replaced by another

adhesive

PHASE 2 Adhesive type Acrylate Epoxy Epoxy

Manufacturer Huntsman 3M Scotch-Weld 3M Scotch-Weld

Name Araldite 2047 DP 490 2216 B/A

Bond thickness 3 mm 5 mm 3 mm

Number of specimens 3 3 3


132


7.3.1 Phase 1

Table 19: Summary of results Phase 1

a) Failure mode b) Mean load at failure c) Mean extension at failure d) Mean shear strength

Huntsman Araldite 2047 / 3mm thick a) Adhesion GFRP- adhesive (10/10) b) 1.32 kN c) 0.81 mm d) 1.10 MPa

3M Scotch-Weld DP 490 / 3mm thick a) Whole glass breakage (9/10)

Glass peeling off at the adhesive joint (1/10) b) 5.38 kN c) 0.11 mm d) 4.49 MPa

Dow Corning TSSA / 3mm thick a) Cohesive preceded by substantial adhesive

plastic strain (9/10) Adhesion GFRP- adhesive (1/10)

b) 0.26 kN c) 24.09 mm d) 0.26 MPa


133

7.3.1.1 Huntsman Araldite 2047 (3 mm)

All ten specimens presented adhesion failure at the GFRP–adhesive interface. As shown

in figure 49, the bonded connection surface in the detached GFRP bar was clear from any

traces of adhesive, which remained compact and adhered to the glass plate.

Figure 49: Adhesive failure at GFRP-adhesive interface in Huntsman Araldite 2047 3mm thick bond

The bonds registered a mean shear stress of 1.32 MPa at failure. The average extension at

failure was 0.81 mm. However, only about 0.10 mm could be attributed to the elastic

shear strain of the adhesive. After the adhesion failure, the load dropped gradually while

the GFRP slowly slipped from the adhesive showing some residual load bearing capacity

before detaching completely.

The GFRP surface was believed to be too smooth, provoking adhesion failure and

preventing the adhsive from reaching its potential. According to the manufacturer

(Huntsman), it could potentially achieve a shear strength of 6.3 MPa in bonded glass-

GFRP connections. In order to improve adhesion of the GFRP-adhesive interface,

abrading the surface of the GFRP to increase adhesion thorugh mechanical interlock was

proposed for Phase 2.


134

7.3.1.2 3M Scotch-Weld DP 490 (3 mm)

All ten specimens experienced glass substrate failure. Nine specimens suffered whole

glass failure as shown in figure 50(a) while in one specimen the glass peeled off at the

edge as shown in figure 50(b).

Figure 50: Glass substrate failure in 3M DP490 Epoxy 3 mm thick bonds (a) whole glass breakage

and (b) glass peeling off at the edge

The bonds registered a mean shear stress of 4.49 MPa at failure, which was the highest

among all the candidate adhesives and not far from the requirement of 5.7 MPa. The

average extension at failure was 0.11 mm. No plastic deformation was observed before

failure, which was sudden and immediately followed by loss of loading capacity. Glass

failure in this test was believed to be caused by peak shear stress in the adhesive that was

then transferred to the glass resulting in a local stress concentration in the glass that

caused breakage.

Shear stress peaks are caused by a phenomenon named differential shear (Adams et al,

1997). Assuming a single lap joint with rigid substrates and the adhesive only deforming

in shear, then the shear stress is uniformly distributed across the adhesive as shown by the

(a)

(b)

Origin of failure Glass chip peeled off from pane


135

parallelograms in figure 51. Assuming a similar joint but with elastic substrates, the

elastic substrates present differential tensile strain that generates a differential shear stress

distribution across the adhesive as shown by the distorted shapes in figure 52. For the

upper substrate, the tensile stress is a maximum at A and falls to zero at B. Thus, the

tensile strain at A is larger than that at B and this strain must progressively reduce over

the length l. The converse is true for the lower substrate. The shear strain in the adhesive

must progressively decrease to zero in the middle and increase again along the length l.

Shear stress is proportional to shear strain. Hence, the maximum shear stress occurs at the

ends.

Figure 51: Uniform shear stress distribution (Adams et al, 1997)

Figure 52: Differential shear stress distribution (Adams et al, 1997)

In addition to differential shear, the edges are also the weakest area in toughened glass as

they are not as toughened as the centre. This contributes to trigger failure near the edges.

The exact location of the origin of the glass failure is determined by local flaws in the

glass. Figure 50(a) shows the exact origin of failure in the circled region. This is adjacent

to the edge, but not right on the edge.

The reason why the glass peeled off at the edge in one of the specimens as shown in

figure 50(b) instead of suffering whole glass failure was not clear. An explanation for this


136

could be that the eccentric load path through the bonded connection induced bending

stress at the edge.

The strain in the adhesive layer is larger near the edges especially on stiff connections, as

adhesives with a higher stiffness don‘t have the possibility to redistribute stresses within

the material itself. Since 3M DP490 Epoxy is a very stiff adhesive, increasing the

thickness of the adhesive from 3 mm to 5 mm was proposed for Phase 2 in an attempt to

achieve a more flexible joint.

7.3.1.3 Dow Corning TSSA (3mm)

Nine specimens presented cohesion failure as shown in figure 53. One specimen

presented adhesion failure at the GFRP–adhesive interface.

Figure 53: Cohesive failure in Dow Corning TSSA Silicone 3 mm thick bond

The bonds registered a mean shear stress of 0.26 MPa at failure. The average extension at

failure was 24.09 mm. Although the cohesive failure mode is desirable and the adhesive

showed remarkable flexibility, the low shear strength is too low.

There is an important discrepancy between the tested shear strength of 0.26 MPa and the

shear strength of 4-5 MPa provided by the manufacturer. This might be due to two

Remains of adhesive on both glass and GFRP substrates


137

reasons. The first reason is that three layers of film were used against the manufacturer’s

recommendation of using only one layer. The second reason is that the adhesive storage

conditions and specimen preparation process were not in line with the manufacturer’s

recommendations. The recommended storage temperature was exceeded. Moreover, the

curing pressure and temperature conditions were achieved by means of clamps and an

oven instead of an autoclave. This non-ideal specimen preparation process probably

undermined the strength of the adhesive bonds. Due to the lack of access to an autoclave,

TSSA was disregarded for Phase 2.


138

7.3.2 Phase 2

Table 20: Summary of results Phase 2

(

a) Failure mode b) Mean load at failure c) Mean extension at failure d) Mean shear strength

Huntsman Araldite 2047 / 3mm thick / Abraded GFRP

a) Glass peeling off combined with some plastic strain in the adhesive (2/3)

Adhesion GFRP- Adhesive (1/3) b) 4.28 kN c) 0.28 mm d) 3.57 MPa

3M Scotch-Weld DP 490 / 5mm thick a) Glass peeling off combined with cohesive failure

(2/3) Whole glass breakage (1/3)

b) 5.64 kN c) 0.18 mm d) 4.70 MPa

3M Scotch-Weld 2216 B/A / 3mm thick a) Adhesion Glass-Adhesive (2/3)

Glass peeling off (1/3) a) 2.26 kN b) 0.46 mm c) 1.88 MPa

Changes from Phase 1


139

7.3.2.1 Huntsman Araldite 2047 (3 mm) with abraded GFRP

Two specimens presented glass substrate failure with glass peeling off from the edge as

shown in figure 54(a). One specimen presented adhesion failure at the GFRP-adhesive

interface as shown in figure 54(b).

Figure 54: Failure mechanisms in Huntsman Araldite 2047 Acrylate 3 mm thick bonds with abraded

GFRP (a) glass peeling off at the edge with signs of adhesive plastic deformation and (b) adhesive failure at GFRP-adhesive interface

The specimen that suffered adhesion failure was excluded from average calculations since

it was deemed to have been poorly executed. The bonds in the other two specimens

registered a mean shear stress of 3.57 MPa at failure. The average extension at failure was

0.28 mm. After the glass failure, the load dropped gradually showing some residual

loading bearing capacity before detaching completely. Glass failure was preceded by

some plastic deformation in the adhesive as shown in figure 54(a). The white region in

the circled area shows the plastic deformation zone

(a)

(b)

Whitening of the adhesive denotes plastic deformation prior to failure

Glass chip peeled off from pane


140

Major improvements are observed when comparing to Phase 1 results for the same

adhesive. The plastic deformation of the adhesive before glass breakage and the structural

residual loading capacity after breakage are important qualities for façade applications.

Visible plastic deformation would help to prevent glass breakage identifying the problem

before it happens. In case of glass failure, residual load bearing capacity would provide

time to replace the unit before complete collapse.

7.3.2.2 3M Scotch-Weld DP 490 (5mm)

All three specimens presented substrate failure. Two specimens had the glass plucking at

the edge combined with cohesive failure in part of the adhesive as shown in figure 55(a).

One specimen presented whole glass breakage as shown in figure 55(b).

Figure 55: Failure mechanisms in 3M DP490 Epoxy 5 mm thick bond (a) glass peeling off at the edge

combined with cohesion failure and (b) whole glass breakage

The bonds reached an average shear stress of 4.70 MPa at failure. This is the highest

registered in all the tests. The average extension at failure was 0.18 mm.

(a)

(b)

Origin of failure

Adhesive cohesion failure



141

A much higher proportion of the specimens failed by glass plucking at the edge compared

to Phase 1 results of the same adhesive. This is believed to be caused by the induced

bending moment at the edge which was incremented by the increase in thickness of the

adhesive.

As shown in Figure 56, a bending moment is created by the non-alignment of the loads

applied on the two substrates. This bending moment tends to peel one substrate off the

other from the edge. Although the two pulling forces applied at GFRP plates in this test

were aligned, the force was transmitted through the body of the glass, which was not

aligned and a bending moment was induced. This bending moment is directly

proportional to the thickness of the adhesive. The similarities between the deformed joint

illustrated in figure 56(b) and the glass peeling off combined with cohesion failure in the

opposite corner of the adhesive illustrated in figure 55(a) confers consistency to this

explanation.

Figure 56: Induction of bending moment in single shear lap joint

This induced bending moment is an unwanted side effect of single-lap shear tests with

thick bonded connections. This effect should not be critical in the real façade application.

Even with the induced moment, the shear strength achieved is above the requirements.

Therefore, it is reasonable to affirm that 3M Scotch-Weld DP 490 Epoxy would meet the

requirements of this façade application. In the future, it would be good to carry out further

tests eliminating or reducing the induced moment. One option would be reducing the

thickness of the adhesive to 2 mm. Another option would be to carry out double-lap shear

tests.


142

7.3.2.3 3M Scotch Weld 2216 B/A (3mm)

Two specimens presented adhesion failure at the glass-adhesive interface as shown in

figure 57(a).One specimen had the glass plucking at the edge as shown in figure 57(b).

Figure 57: Failure mechanisms in 3M Scotch Weld 2216 B/A Epoxy 3mm thick bonds (a) adhesive

failure at GFRP-adhesive interface and (b) glass peeling off at the edge

The bonds reached an average shear stress of 1.88 MPa at failure. The average extension

at failure was 0.46 mm. All the extension was due to elastic shear strain.

The reason for the glass plucking is the induced bending moment explained in the

previous section. The adhesion failure at the glass-adhesive interface is in contradiction

with previous research (Nhamoinesu and Overend, 2012) that yielded higher shear

strength results without adhesion failure at the glass-adhesive interface. This unexpected

failure could be due to inadequate surface preparation or to large bond thickness. In any

case, the results were far off from the required shear strength and this adhesive was

considered inadequate for this façade application.

(a)

(b)



143

7.3.3 Comparative stiffness

Figure 58 illustrates the load vs. extension typical curves for each of the bonded

connections tested. The highest load was achieved by the 3M Scotch-Weld DP 490 bond

with 5 mm thickness. It is visually noticeable how the bond became more flexible by

increasing the thickness of the bond from 3 mm to 5 mm. The mean shear strength was

4.70 MPa and the mean extension at failure was 0.18 mm. When the surface of the GFRP

was abraded, the Huntsman Araldite 2047 bond achieved reasonable load bearing

capacity with some plastic deformation before failure. The mean shear strength was 3.57

MPa and the mean extension at failure was 0.28 mm. Dow Corning TSSA and 3M

Scotch-Weld 2216 B/A had very low load bearing capacity.

Figure 58: Load vs. extension curves of bonds (Dow Corning TSSA is excluded for clarity)

7.4 Conclusion

3M Scotch-Weld 2216 B/A did not display enough load bearing capacity. The results for

Dow Corning TSSA are not valid because the storage and specimen preparation

conditions were not ideal.

The 3M Scotch-Weld DP 490 bonds achieved the highest shear strength values. The

failure mode was glass breakage. The high stiffness of this adhesive contributes to peak

0

1

2

3

4

5

6

7

8

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Loa

d [

kN

]

Displacement [mm]

Araldite 2047 Acrylic

3M Scotch Weld DP 490 Epoxy

Araldite 2047 Acrylic (Abraded GFRP)

3M Scotch Weld DP 490 Epoxy (5mm)

3M Scotch Weld 2216 B/A Epoxy


144

shear stresses being transmitted to the glass causing local stress concentrations and

breakage. By increasing the thickness from 3 mm to 5 mm, the flexibility and shear

strength of the bond was increased. The increased thickness allows a more uniform

distribution of the stresses and reduces the peak loads transmitted to the glass. The

application of the adhesive was easy and it took around one week to be cured.

The Huntsman Araldite 2047 bonds achieved reasonable load bearing capacity. The

failure mode was the safest for this application. Glass failure was preceded by some

plastic deformation in the adhesive which was visible because the adhesive turned white.

After the glass failed, the load dropped gradually showing some residual loading bearing

capacity before detaching completely. The application of the adhesive was easy and it

took around one day to be cured.

The results of the single-lap shear tests validate both 3M Scotch-Weld DP 490 and

Huntsman Araldite 2047 as suitable adhesives for this application. The safer failure mode

of the Araldite 2047 makes it the best option to develop the design further. The mean

shear strength of the 3mm thick bond was 3.57 MPa.

The following considerations should be noted:

Bending moment was induced at bonded connections due to eccentric loading.

This may have produced premature failure in samples. Double-lap shear tests

would be preferable in future investigations

Loading pattern is deflection driven so is not representative of real wind loading.

This will be resolved in future work by building and testing full scale prototype


145

DETAIL DESIGN DESCRIPTION

Based on the structural and thermal calculations on the schematic design and on the

results of the experimental tests, some changes are put into effect to improve the

performance of the system and mitigate potential issues.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


146


147

8.1 Design changes

Based on the structural and thermal calculations on the schematic design and on the

results of the experimental tests, the following changes are put into effect to improve the

performance of the system and mitigate potential issues:

GFRP frame is 5 mm thick instead of 2 mm thick (5 mm thick intermediary glass

pane instead of 3 mm too)

Adhesive bond is 2mm thick instead of 3mm thick

GFRP frame and spacer integrated into single element. Stainless steel vapour

barrier formerly attached to spacer is now attached to GFRP.

Gaskets at mullion are overlapping gaskets, not kissing gaskets, and leave glass

edges exposed

These changes can be observed in the figures over the next page which represent the

schematic design against the detail design.


148

Figure 59: Scheme design

Figure 60: Detail design

Gaskets

Structural glass

Silica gel


GFRP frame

Sealant

Gaskets

Structural glass

Spacer


GFRP frame

Sealant


149

DETAIL DESIGN STRUCTURAL ASSESSMENT

BY NUMERICAL CALCULATION

Numerical analysis has been carried out on the detail design to take into account effects

that were not considered in the initial analytical calculations such as shear deformations,

shear lag effect or time dependent properties of the GFRP and the adhesive. A short term

and a long tern load cases have been established to better represent wind loading and to

investigate if modelling the GFRP and the adhesive with different Modulus of Elasticity

affects the results.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


150


151

9.1 Method

The assumptions made on the modelling, the applied load and the properties of the

materials are described.

9.1.1 Software and model

A three-dimensional FEA model for the Glass-GFRP composite unit was constructed

using LUSAS v14.5; a non-linear elastic model with the geometry shown in figure 61a

below. A four node tetrahedral element type (TH4) which is a 3-dimensional

isoparametric finite element with linear interpolation order was used. The element output

is obtained at both the element nodes and Gauss points and consists of a stress output

(direct and shear stresses) and strain output (direct and shear strains). The stresses are

obtained by integrating the constitutive relationship at the element Gauss points. Nodal

stresses are then obtained by extrapolation from the Gauss points. This is achieved by (i)

defining a fictitious element with nodes at the element Gauss points and then (ii)

extrapolating the stress or strain to the nodal points of the real element using the shape

function of the fictitious element (equations 11 and 12).

N

I

IiiIiii N1

,, [11]

N

I

IiiIiii N1

,, [12]

Where:

N is the number of Gauss points

i denotes nodal point values

I denotes Gauss point values

The averaged nodal stresses are then obtained by evaluating the mean of the extrapolated

nodal values. Since the Glass-GFRP composite unit is symmetrical about the mid-length

(line CD in figure 61a) and mid-width (line AD in Figure 61a), only a quarter of the

composite units were modelled. For the boundary conditions, a z-direction restraint was

applied on the end of the GFRP E-section at point B in Figure 61a to represent the

connection of the GFRP to the main supporting structure. Symmetrical boundary


152

conditions were assigned on the yz-plane along CD and on the xz-plane along AD (figure

61a).

Figure 61: Glass-GFRP composite unit (a) full model and (b) magnified view showing meshing detail

(a)

(b)

C

B

A

D

1750m

750mm

y

x

750mz

𝑞 (N/mm2)

57mm

10mm

37mm

10mm

33mm

26mm

Glass pane (10mm)

Adhesive (2mm)

GFRP profile (5mm)

Adhesive (2mm)

Glass pane (10mm)


153

A mesh density of four element divisions across a 10 mm thickness was adopted for the

glass panels. A mesh density of one element division across a 5 mm and 2 mm thickness

was adopted for the GFRP E-section and for the adhesive respectively (figure 61b). The

reason for the smaller divisions in the adhesive was to account for the relatively larger

displacements therefore proportionally higher stress gradients. Reducing mesh density in

the glass also reduced computational time.

A uniformly distributed load q was applied on the top glass panel (figure 61a). A

geometric and material non-linear analysis was run with an updated Lagrangian approach.

The analysis had a total of 5 increments, with 4 iterations per increment. The increase was

automatic with a starting load factor of 0.1 and a maximum total load factor of 1.

9.1.2 Applied load

As in the schematic design chapter, wind load has been considered to be the critical case

and has been the only load to be investigated in depth. However, for the detail design two

cases with different load durations have been analysed: a short duration higher wind load

and a long duration lower wind load. This is to try to capture the effect of the variation of

the modulus of elasticity of the GFRP and the adhesive with different load durations.

Wind loads on façades are often calculated using building codes. These calculations are

based on wind velocity to which a number of factors are applied such as gust effects,

internal pressures, building height, etc. The wind velocity considered may vary depending

on the code. BS EN 1991-1-4 (2005) bases its calculations on a 10 minute mean wind

velocity with an annual risk of being exceeded of 0.02. These calculations are a

simplification of the dynamic action of wind into a static action. In reality, the velocity of

the wind changes constantly. Wind gusts provoke instantaneous fluctuations from the

mean wind velocity as shown in figure 62.


154

Figure 62: Typical fluctuation of wind velocity in time

A one second load duration pressure calculation would use the peak velocity pressure

while the 10 minute pressure calculation would use the basic velocity pressure. Equations

13 and 14 taken from NA.2.17 (NA to BS EN 1991-1-4, 2011) represent the peak and

main basic velocity pressures in a town terrain:

For one second gust: qp = Ce(z) Ce,T qb [13]

For 10 minutes wind: qp = qb [14]

Where:

qp: peak velocity pressure

qb: reference mean basic velocity pressure

Ce(z): value of exposure factor NA.7 (NA to BS EN 1991-1-4, 2011)

Ce,T: value of exposure correction NA.8 (NA to BS EN 1991-1-4, 2011)

Ce(z) and Ce,T depend on the distance upwind from shoreline and the height. Since the

basic velocity pressure is taken, by definition, at 100m above ground, it is logical to use

the same height to calculate an equivalent peak velocity pressure. Assuming a distance of

10 km upwind from the shoreline, the proportion between basic velocity pressure and

peak velocity pressure would be 4. Based on this proportion and tying with the 3KPa

pressure that was considered standard in schematic design, the following wind pressures

have been assumed:

Load case 1 - one second gust: qp = 3 KPa


155

Load case 2 -ten minutes wind: qp = 0.75 KPa

9.1.3 Material properties

The GFRP-Glass composite unit has been run as a non-linear-elastic model: the glass has

been run as a linear elastic material but the Araldite A2047 adhesive and the E-profile

GFRP have been modelled as elastic-perfectly plastic materials. Geometric non-linearity

is also addressed by default by the software. The table below summarises the main

mechanical properties of the materials as modelled on each of the two load cases:

Table 21: Mechanical properties of materials

Load case 1 Load case 2

Wind load duration (s) 600 (10 minutes) 1

Wind pressure (N/m2) 750 3000

Glass E (GPa) 70

Glass v (-) 0.23

GFRP E (GPa) 23.37 Derived from 4-point bending test

100 Not representative to consider the peak wind pressure directly from t=1 because the initial loading starts from a mean velocity façade pressure, not from 0. This will be resolved in future work by building and testing full scale prototype and applying real wind pressure patterns

GFRP v (-) 0.3

GFRP yield stress (MPa) 75

Araldite A2047 E (GPa) 142.35 Approximated by % extrapolation

634.90 (Nhamoinesu and Overend, 2012)

Araldite A2047 v (-) 0.43 (Nhamoinesu and Overend, 2012)

Araldite yield stress 3.61 (Nhamoinesu and Overend, 2012)


156


9.2.1 Maximum deflection at the edge of the IGU

The maximum displacement at the edge of the panel occurs at mid-span and is 2.83 mm

for Load Case 1 and 7.65 mm for Load Case 2. The ratio between Load Case 2 and Load

Case 1 is 2.7. The allowable deflection that is being considered is 15mm following

guidance from section 3.5.2.5 of the CWCT Standard for systemised building envelopes

(CWCT, 2005). Therefore, the calculated deflection is lower than the allowable

deflection.

Figure 63: Glazing displacement contour plot for Load Case 1 (600s; 750N/m2)

Figure 64: Glazing displacement contour plot for Load Case 2 (1s; 3000N/m2)

Mid-span tensile stress =

2.24MPa

Maximum deflection at edge at midspan = 2.83mm

Maximum deflection at edge at midspan = 7.65mm


157

9.2.2 Maximum tensile stress at the glass

The tensile stress at the glass occurs at mid-span on the long edge of the glass (where the

highest tensile stresses in the glass are expected if stress concentrations are discounted)

and is 4.5 MPa for Load Case 1 and 29.11 MPa for Load Case 2. The ratio between Load

Case 2 and Load Case 1 is 6.5. The allowable surface stresses are 46.6 MPa on heat

strengthened glass and 93.1 MPa on toughened glass (ASTM E1300 - 12ae1). Therefore,

the calculated tensile stresses are lower than the allowable tensile stresses.

Figure 65: Principal Stress Contour Plot for Load Case 1 (600s; 750N/m2)

Figure 66: Principal Stress Contour Plot for Load Case 2 (1s; 3000N/m2)

Tensile stress at midspan = 4.5MPa

Tensile stress at midspan = 29.11MPa


158

9.2.3 Maximum shear stress at the adhesive

The maximum shear stress at the adhesive occurs near the end and is 0.57 MPa for Load

Case 1 and 1.97 MPa for Load Case 2. The ratio between Load Case 2 and Load Case 1 is

3.5. The calculated shear stresses are below the average shear stress at failure of the tested

connection with Araldite 2047.

Figure 67: Adhesive Shear Stress Contour Plot for Load Case 1 (600s; 750N/m2)

Figure 68: Adhesive Shear Stress Contour Plot for Load Case 2 (1s; 3000N/m2)


2.24MPa

Maximum shear stress = 0.57MPa



159

9.2.4 Maximum shear stress at the GFRP

The maximum shear stress at the GFRP occurs at the end of the profile at the web area is

12.35 MPa for Load Case 1 and 43.25 MPa for Load Case 2. The ratio between Load

Case 2 and Load Case 1 is 3.5. The shear strength of GFRP is 17 MPa according to the

experimental tests. Therefore, the GFRP profile would fail to shear in the short duration

load case before reaching the values in the figures below.

Figure 69: GFRP Shear Stress Contour Plot for Load Case 1 (600s; 750N/m2)

Figure 70: GFRP Shear Stress Contour Plot for Load Case 2 (1s; 3000N/m2)


2.24MPa




160

9.3 Conclusion

The short duration load is the critical case for all the calculated parameters. The

fundamental reason for this is the higher façade pressures that have been considered

which are four times higher in the short duration load case than in the long duration load

case.

The design is compliant with the criteria set for maximum deflection at the edge of the

IGU, maximum tensile stress at the glass and maximum shear stress at the adhesive.

Moreover, the eduction of structural depth to almost one fifth compared to benchmark

conventional system as calculated in initial analytical calculations is confirmed.

On the other hand, the shear stress at the GFRP exceeds the shear strength of the GFRP in

the short duration load case. This occurs only in a very small area adjacent to the support

condition while the majority of the profile remains within the permissible values so

should not be a major issue. It could be resolved by reinforcing the profile at such

locations with either steel plates or by incorporating glass fibres in multiple directions at

the ends of the profiles. It should be noted that the longitudinal shear strength of the

pultruded GFRP is relatively low because the fibres are mostly unidirectional.

Incorporating glass fibres in multiple directions has the potential to increase the shear

strength.

The low reliability of the 4-point bending results for short duration loads makes it

difficult to determine the Modulus of Elasticity to be modelled. Moreover, it would not be

representative to consider the peak wind pressure directly from t=0 because the initial

loading starts from a mean velocity façade pressure, not from 0. This will be resolved in

future work by building and testing full scale prototype and applying real wind pressure

patterns.


161

DETAIL DESIGN THERMAL ASSESSMENT BY

NUMERICAL CALCULATION

Thermal transmittance and risk of condensation of the detail design are assessed through

comparative analytical and numerical thermal analysis with the conventional system and

the schematic design.

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


162


163

10.1 Method

The thermal transmittance and condensation numercial analysis has been calculated for

the detail design following the same procedure as for the schematic design. The

assumptions are also the same except for the design changes listed below and illustrated

in figure 71:

The IGU has a 5 mm thick intermediary glass pane instead of 3 mm thick (table

22);

The GFRP frame is 5 mm thick instead of 3 mm thick. The GFRP at the transom

is concealed behind the glass pane leaving exposed glass edges.

The GFRP frame and the spacer are integrated into a single element. The stainless

steel vapour barrier formerly attached to the spacer is now attached to the GFRP

frame.

The gaskets at the mullion are overlapping gaskets, not kissing gaskets, and leave

the glass edges exposed.

Table 22: WINDOW 6.3.9.0 modelled IGU description and centre-of-glazing results


164

Figure 71: Proposed system as modelled in THERM 6.3.19.0 (a) mullion and (b) transom


Heat flow and U-values for the conventional system, the schematic design and the detail

design are described in tables 23 and 24. Against the conventional system, the detail

design achieves a reduction of the U-value of the frame area of 53% while the U-value of

the edge of glazing area is increased by 66%. The total area-weighted U-value of the

system is reduced by 6%.

(b)

(a)



Glass fibre reinforced polyester frame



Stepped glass

Cavity non ventilated


Silicone sealant

Structural adhesive


165

Table 23: Heat flow comparison between between conventional system, schematic design and detail design

Table 24: U-value comparison between conventional system, schematic design and detail design

Heat flow profiles of representative transom and mullion sections for the schematic

design and the detail design are illustrated in figure 72. Both profiles are similar with

maximum heat flow values just over 300 W/m2 located at the stainless steel vapour

barrier, followed by the GFRP frame and some portions of the gasket. Both systems

distribute the flow between the frame and the edge of glazing areas.

1.54

0.85 1.01

0.42

0.630.69

2.84

2.842.83

0.00

1.00

2.00

3.00

4.00

5.00

6.00

Conventional

system

Initial design

proposal

Design

development

He

at

flo

w [

W/°

K]

Centre of glazing

Edge of glazing

Frame

3.62

0.70 0.67

0.91

1.99

1.05

0.670.82

2.37

1.16

0.670.86

0.00

0.50

1.00

1.50

2.00

2.50

3.00

3.50

4.00

Frame Edge of

glazing

Centre of

glazing

Total

U-v

alu

e [

W/m

2°K

]

Conventional system

Initial design proposal

Design development


166

Figure 72: Heat flow profiles for (a) schematic design mullion; (b) detail design mullion; (c) schematic

design transom and (d) detail design transom

(b)

frame glass edge

glass edge

GFRP frame Stainless steel

vapour barrier EPDM gasket

(d)

frame

glass edge

glass edge

(a)

frame glass edge

glass edge

GFRP frame and structural adhesive


(c)

frame

glass edge

glass edge


167

Temperature profiles of representative transom and mullion sections for the detail design

are illustrated in figure . The lowest surface temperature is 11.3 ºC and is located at the

transom gasket. This temperature is lower than the equivalent temperature calculated for

the conventional system (16.8ºC) but higher than that calculated for the schematic design

(10.2 ºC). In all cases the calculated temperatures are above the dew-point temperature of

6.9 ºC so there is no risk of condensation.

Figure 73: Temperature profiles and location of areas with minimum inside surface temperature for detail design (a) mullion and (b) transom

(a)

(b)

12.2 ºC

11.3 ºC


168

10.3 Conclusion

Against the conventional system, the detail design achieves a total area-weighted U-value

reduction of 6%. This is 4% less reduction than that achieved by the initially proposed

design. The benefit of integrating frame and spacers is counterbalanced by the structural

need to increase the thickness of the GFRP web.

The heat flow profiles of the schematic and the detail designs are similar with maximum

heat flow values just over 300 W/m2 located at the stainless steel vapour barrier, followed

by the GFRP frame and some portions of the gasket. Both designs distribute the flow

evenly between the frame and the edge of glazing areas.

The lowest inside surface temperature for the detail design is 11.3 ºC and is located at the

transom gasket. This temperature is lower than the equivalent temperature calculated for

the conventional system (16.8ºC) but higher than that calculated for the schematic design

(10.2 ºC). Removing the portions of GFRP frame and gaskets that were covering the glass

edges in the initial design mitigates thermal bridging and contributes to a slight rise of

peak low superficial inside temperatures. In all cases the calculated temperatures are

above the dew-point temperature of 6.9 ºC so there is no risk of condensation.


169

CONCLUSION AND FUTURE WORK

1. Introduction

2. State of the art

















SCHEMATIC

DESIGN

DETAIL

DESIGN

EXPERIMENTAL

INVESTIGATIONS


170


171

11.1 Conclusion

The objectives of the investigation to develop a unitised curtain wall that would reduce

the thermal transmission, reduce the structural depth and improve the aesthetics of

conventional systems have been demonstrated:

11.1.1 Reduction in structural depth

The analytical calculations carried out on the schematic design reflected that, subjected to

wind load, the proposed system required almost five times less structural depth than the

conventional system to achieve an equivalent stiffness. Moreover, they pointed out shear

stress at the GFRP web and the adhesive as the critical cases in the structural design.

Consequently, the thickness of the GFRP web was increased in the detail design and the

properties of a range of GFRP and adhesives were studied in depth by carrying out 4-

point bending tests on GFRP and Single-Lap Shear tests on adhesive connections between

glass and GFRP.

4-point bending tests were used to select polyester resin based GFRP as the preferred

framing material and to characterise its viscoelastic behaviour by calculating its variable

Modulus of Elasticity both for short duration and for long duration loads. The results also

confirmed the low shear capacity of the selected pultruded GFRP in the longitudinal

direction of the fibres. Single-Lap-Shear tests were used to select Araldite 2047 as the

preferred adhesive for the application based on its reasonable load bearing capacity, and

its safe failure mode, which was preceded by visual signs of plastic deformation and

showed residual loading bearing capacity before detaching completely. The shear strength

of the connection was also quantified.

Numerical analysis was carried out on the detail design to take into account effects that

were not considered in the initial analytical calculations such as shear deformations, shear

lag effect or time dependent properties of the GFRP and the adhesive. The viscoelasticity

of the GFRP and the adhesive was accounted for by establishing one short and one long

duration load cases based on the wind velocity profile in time, with peak wind velocity

pressures for one second wind gusts and mean wind velocity pressures for ten minutes

loads. The respective Modulus of Elasticity for each material and load case were

extracted from the experimental tests and previous research. The final results indicate that

the short duration load is the critical case for all the calculated parameters due to the


172

higher façade pressures associated to the peak wind velocity. The design criteria set for

maximum deflection at the edge of the IGU, maximum tensile stress at the glass and

maximum shear stress at the adhesive are met. Moreover, the reduction of the structural

depth to almost one fifth compared to the conventional system as calculated in the initial

analytical calculations is confirmed. This proportion may vary for different conditions but

a significant reduction in structural depth is evident. However, the shear stress at the

GFRP exceeds the shear strength of the GFRP in the short duration load case. This occurs

in a small area adjacent to the support while the majority of the profile remains within the

permissible values so should not be a major issue. It could be resolved by reinforcing the

profile at such locations with either steel plates or by incorporating glass fibres in

multiple directions at the ends in the manufacturing of the profiles. It should be noted that

the longitudinal shear strength of the pultruded GFRP is relatively low because the fibres

are most unidirectional. Incorporating glass fibres in multiple directions has the potential

to increase the shear strength significantly.

11.1.2 Reduction of thermal transmission

The schematic design achieved a total area-weighted U-value reduction of 10%. In detail

design, the performance was optimized by integrating into one single element the frame

and the spacer what would have reduced the thermal transmission. Unfortunately, the

thickness of the web had to be increased for structural reasons, so the final reduction

value is 6%. It should be noted that the benchmark conventional system considered is the

best that could be found in the market, with the best performing triple glazing and an

overall U-value of 0.91 W/m2K while regulations typically require values in the area of

2.0 W/m2K (Building Regulations Part L1A, 2013). The lowest inside surface

temperature for the proposed design is above the calculated dew-point temperature so

there is no risk of condensation.

11.1.3 Improved aesthetics

A seemingly frameless unitised curtain wall has been designed providing a flush glazed

appearance both to the inside and the outside while keeping the full functionality of a

unitised system.


173

11.2 Future work

In the research process, several areas have been identified that would require further

investigation in the product development of the frame-integrated unitised curtain wall:

Investigate reinforcing GFRP profile at ends with either steel plates or by

incorporating glass fibres in multiple directions to increase its shear strength

Perform destructive testing of wider range of adhesive bonds in shear and

bending, including long term performance and accelerated weathering

Develop fire and acoustic compartmentation details

Assemble prototypes for testing:

o Air and water tightness and deflection under wind load / Response to

dynamic loads / Performance in fire / Durability of panels (particularly to

vapour infiltration within the cavity)

o Prototypes composite beam and then full size prototype


174


175

RELEVANT PUBLICATIONS / AWARDS

Cordero, B. et al (2015). Thermal Performance of novel-frame integrated unitised

curtain wall. Revista de la Construcción vol.14 no.1. Santiago. Chile

Paper in process for Proceedings of the ICE – Construction Materials

International Patent Application No. PCT/GB2014/053567

Finalist in the Council on Tall Buildings and Urban Habitat (CTBUH) Awards

2012

Pump-prime funding from the Engineering and Physical Sciences Research

Council (EPSRC)

Grant from the Institution of Structural Engineers (IStructE)

Bright IDEAS Awards (EPSRC)


176


177

REFERENCES AND BIBLIOGRAPHY

3M, 1996. Scotch-Weld DP 490 Technical Data Sheet [online]. Available from:

http://polymerteknik.com/pdf/datablad/DP_490_Tek.pdf [Accessed 08 March 2014]

Adams, R.D. et al., 1997. Structural Adhesive Joints in Engineering. 2nd edition.

Chapman & Hall, Chapter 2.2.1, pp.17 – 18, London, UK.

Andrade, V (2004). Standardized composite slab systems for building constructions.

Journal of Constructional Steel Research, 60, 493–524

ANSI/NFRC 100 (2014). Procedure for Determining Fenestration Product U-factors.

American National Standards Institute / National Fenestration Rating Council: Greenbelt,

United States

ASTM E 119 -12a. Standard Test Methods for Fire Tests of Building Construction and

Materials.

ASTM E1300 - 12ae1. Standard Practice for Determining Load Resistance of Glass in

Buildings

BS 476-20 (1987) Fire tests on building materials and structures. Method for

determination of the fire resistance of elements of construction (general principles).

British Standards Institution: London, United Kingdom

BS 8118-1 (1991) Structural use of aluminium. Code of practice for design. British

Standards Institution: London, United Kingdom

BS EN 1279-1 (2004). Glass in building - Insulating glass units. British Standards

Institution: London, United Kingdom

BS EN 13501-1 (2002). Fire classification of construction products and building

elements, Part 1 – Classification using data from reaction to fire tests. British Standards

Institution: London, United Kingdom.

BS EN 1366-4 (2006) Fire resistance tests for service installations. Linear joint seals.

British Standards Institution: London, United Kingdom

http://polymerteknik.com/pdf/datablad/DP_490_Tek.pdf


178

BS EN 13947 (2006). Thermal performance of curtain walling - Calculation of thermal

transmittance. British Standards Institution: London, United Kingdom.

BS EN 1991-1-4 (2005). Eurocode 1: Actions on structures – Part 1-4: General actions –

Wind actions. British Standards Institution. London, United Kingdom

CTE-DB-SI (2010). Seguridad en caso de incendio. Ministerio de Fomento: Madrid,

Spain.

CWCT (2003). Technical note 39 - Sound transmission through building envelopes.

University of Bath: Bath, United Kingdom

CWCT (2005). Standard for systemised building envelopes. Part 3: Air, water and wind

resistance. University of Bath: Bath, United Kingdom.

CWCT (2007). Technical Note 49 - U-value of curtain walls .Centre of Window and

Cladding Technology, University of Bath. Bath, United Kingdom.

CWCT (2007). Technical note 56 – Accommodation of structural movement. University

of Bath: Bath, United Kingdom

CWCT (2010). Technical note 62 - Specification of insulating glass units. University of

Bath: Bath, United Kingdom

CWCT (2011). Technical note 73 – Fire performance of curtain walls and rainscreens.

University of Bath: Bath, United Kingdom

Elmahdy, A. H. (2003). Effects of improved spacer bar design on window performance.

National Research Council of Canada. IRC Construction Technology Update No. 58:

Ottawa, Canada.

EN ISO 10077-2 (2012) Thermal performance of windows, doors and shutters --

Calculation of thermal transmittance

Fiberline, 2003. Design Manual. Second edition. Kolding, Denmark.

Fiberline, 2014. Typical thermal properties for Fiberline GFP profiles [online]. Available

from: http://www.fiberline.com/vinduer-doere-facader/profiler-til-vinduer-og-doere-

/tekniske-egenskaber/termiske-egenskaber/typical-thermal-properties-fiberlin [Accessed

09 March 2014]

http://www.fiberline.com/vinduer-doere-facader/profiler-til-vinduer-og-doere-/tekniske-egenskaber/termiske-egenskaber/typical-thermal-properties-fiberlin

http://www.fiberline.com/vinduer-doere-facader/profiler-til-vinduer-og-doere-/tekniske-egenskaber/termiske-egenskaber/typical-thermal-properties-fiberlin


179

Gerresheimer Glas AG. (1973). Isolierglass-Doppelscheibe, insbesondere für

Treibhäuser. Deutsches Patentamt DE 25 32 633 B1. Germany

Glover, M., Field, S. (2002). Fenestration sealed frame, insulating glazing panels. US

Patent 6 401 428 B1. United States

Gokdemir, H., Yilmaz, M. (2012). Curtain wall system wherein a special connection

system is used for plate materials such as glass, aluminium sheet, etc. European Patent 2

444 579 A1. European Patent Office.

Gough, G. S., Elam, C. F., de Bruyne. N. D (1940) The stabilisation of a thin sheet by a

continuous supporting medium. J. Roy. Aero. Soc., Vol 44, Jan. 1940, pp 12-43.

Hartley, J.R., 2002. Composite Profile. Fibreforce, Issue 1, Rev 2. Runcorn, UK.

Huntsman Advanced Materials, 2007. Araldite 2047 Technical Data Sheet [online].

Available from: http://www.intertronics.co.uk/data/ara2047.pdf [Accessed 07 March

2014]

Jin, Q. et al (2009). The performance and optimization of a novel facade panel for energy

efficiency building. In: CISBAT, Lausanne, Switzerland

Kassnel-Henneberg, B. (2011). Isolierglasscheibe. DE 10 2009 057 156 A1. Germany

Kragh, M., 2009. Novel Integrated Curtain Wall Concept using GFRP Composite

Materials. In: GPD, Finland.

Lawrence Berkeley National Laboratory (2011) THERM 6.3 / WINDOW 6.3 National

Fenestration Rating Council Simulation Manual: Berkeley, United States

Lukaszewska, E. (2009) Development of Prefabricated Timber-Concrete Composite

Floor. PhD thesis, Luleå University of Technology, Sweden.

LUSAS (2012) Finite Element System, Lusas theory manual, FEA Ltd, UK, (2012).

Muñoz Viveros, C., Bobadilla Moreno, A. (2012). Simulación y evaluación de puentes

térmicos. Revista de la Construcción Volumen 12 No 22, 92-111

NA to BS EN 1991-1-4 (2011). UK National Annex to Eurocode 1 – Actions on

structures – Part 1-4: General actions –Wind actions. British Standards Institution.

London, United Kingdom

http://www.intertronics.co.uk/data/ara2047.pdf


180

NFRC 100 (1997). Procedures for Determining Fenestration Product U-factors. Equation

4. National Fenestration Rating Council: Greenbelt, United States

Nhamoinesu, S., Overend, M., 2012. The mechanical performance of adhesives for a

steel-glass composite façade system. Challenging Glass 3. TU Delft.

Overend, M. et al., 2011. The selection and performance of adhesives for a steel–glass

connection. International Journal of Adhesion & Adhesives, Volume 31, Issue 7, 587-

597.

Peters, S., 2006. Bonding of fiberglass and glass for constructional applications. Thesis

(PhD). University of Stuttgart Press, Stuttgart, Germany, Chapter 4.4 – 4.8, pp.101-117.

Petersen, L. (2013). Building façade structure with joined pultruded elements. US Patent

8,402,705 B2: United States

Pierre, K. (1992). Attaching panels of glass on to frames made of aluminium, PVC or the

like. FR 2662739 A. France.

Rico Jaraba, C. (2006). Glass panel for external enclosures. European Patent 1 726 766

A8. European Patent Office.

Seele, G. et al. (2010). Insulating glass unit. US Patent 2010/0011703 A1. United States

Sengupta, L.C. and Spurgeon, W.A., 1992. Thermal expansion measurements of polymer

matrix composites and syntactics. US Army Materials Technology Laboratory.

Massachusetts.

Shea, Jr. et al. (1982). Window unit. US Patent 4 464 874. United States.

SMD Stockyards. (2015). Available from: http://www.smdstockyards.co.uk/product/r51-

floor-decking/ [Accessed 22 March 2014]

Standard Test Methods for Fire Tests of Building Construction and Materials (ASTM E

119). American Society for Testing and Materials, Philadelphia.BS EN 13947:2006:

Thermal performance of curtain walling - Calculation of thermal transmittance

Swisspacer, 2008. Datasheet Swisspacer V [online]. Available from:

http://www.swisspacer.com/en/sites/default/files/products/documents/Swisspacer%20V%

20datasheet.pdf [Accessed 22 March 2014]

http://www.smdstockyards.co.uk/product/r51-floor-decking/

http://www.smdstockyards.co.uk/product/r51-floor-decking/

http://www.swisspacer.com/en/sites/default/files/products/documents/Swisspacer%20V%20datasheet.pdf

http://www.swisspacer.com/en/sites/default/files/products/documents/Swisspacer%20V%20datasheet.pdf


181

THERM 6.3 / WINDOW 6.3 NFRC, 2011. Simulation Manual. Lawrence Berkeley

National Laboratory.

Thomas E. Kennedy, M. (1993). Edge seal gasket assembly for a multiple glazing unit.

US Patent 5 584 143. United States.

UK Building Regulations Approved Document Part B Volume 2. Fire safety. Buildings

other than dwelling houses (2010). NBS RIBA Enterprises, Newcastle Upon Tyne,

United Kingdom.

Van Herwijnen, J. (2003). Wall panel, method for manufacturing same and use of the

panel in a curtain wall. European Patent 1 335 079 A1. European Patent Office.

Wurm, J. (2007). Glass structures: design and construction of self-supporting skins,

Basel: Birkhäuser, ISBN: 978-3-7643-7608-6 / 978-3-7643-8317-6


182


183

List of Tables

Table 1: State of the art analysis chart ................................................................................................... 54

Table 2: Mechanical properties of materials .......................................................................................... 73

Table 3: Structural section properties for a range of depths of the conventional system ....................... 75

Table 4: Structural section properties for a range of depths of the proposed system ............................. 76

Table 5: WINDOW 6.3.9.0 modelled IGU description and centre-of-glazing results ........................... 94

Table 6 : Material properties as modelled in THERM 6.3.19.01 ........................................................... 94

Table 7: Environmental Conditions for NFRC Simulations for U-factor calculations .......................... 97

Table 8: Heat flow comparison between proposed and conventional systems ...................................... 98

Table 9: U-value comparison between proposed and conventional systems ......................................... 99

Table 10 : Advantages of glass fibre reinforced polyester resin and glass fibre reinforced

phenolic resin (Hartley, 2002) ............................................................................................. 105

Table 11 : Scope of testing for 4-point bending test ............................................................................ 106

Table 12: Heat soaked phenolic resin modulus of elasticity vs. time curves ....................................... 111

Table 13: Non heat soaked phenolic resin modulus of elasticity vs. time curves ................................ 111

Table 14: Heat soaked polyester resin modulus of elasticity vs. time curves ...................................... 112

Table 15: Non heat soaked polyester resin modulus of elasticity vs. time curves ............................... 112

Table 16: Considerations and target performance required from candidate adhesives ........................ 119

Table 17: Preliminary check of the dimensions of the bond ................................................................ 123

Table 18 : Scope of testing for single-lap shear test ............................................................................ 131

Table 19: Summary of results Phase 1 ................................................................................................. 132

Table 20: Summary of results Phase 2 ................................................................................................. 138

Table 21: Mechanical properties of materials ...................................................................................... 155

Table 22: WINDOW 6.3.9.0 modelled IGU description and centre-of-glazing results ....................... 163

Table 23: Heat flow comparison between between conventional system, schematic design and

detail design ......................................................................................................................... 165

Table 24: U-value comparison between conventional system, schematic design and detail

design ................................................................................................................................... 165


184

List of Figures

Figure 1: Typical insulating glass unit ................................................................................................... 17

Figure 2: Stick curtain wall (a) aluminium supporting grid fixed to the building slab (b) infill

panels fixed to the supporting grid on site (c) schematic cross-section of glass panels

fixed to aluminium mullion ................................................................................................... 18

Figure 3: Unitsed curtain wall (a) factory preassembly of glass panel and frame (b)

preassembled units delivered on site (c) installation of preassembled unit (d)

schematic cross-section of connection between two preassembled units .............................. 19

Figure 4: Market segmentation of the global curtain wall industry by product type for the

periods specified. (Synovate Report) ..................................................................................... 20

Figure 5: (a) EN ISO 10077 Part 2 thermal transfer equation through curtain wall and

(b)thermographic image showing thermal bridging at joints (CWCT TN 49, 2007) ............ 21

Figure 6: Conventional system schematic cross-section through mullion ............................................. 23

Figure 7: Proposed system schematic cross-section thorugh mullion.................................................... 23

Figure 8: Visual appearance comparison between (a) conventional and (b) proposed systems ............ 24

Figure 9: The concept of composite action (Lukaszewska, 2009) ......................................................... 46

Figure 10: Composite slab (SMD Stockyards, 2015 .............................................................................. 47

Figure 11: Thermal resistivity of the combination of 4 different frame materials and 10 different

edge of glazing designs (Elmahdy, 2003) .............................................................................. 48

Figure 12: Movement diagram of conventional unitized curtain wall caused by building frame

deflection due to vertical load and wind sway. (Source: CWCT TN 54, 2007) .................... 61

Figure 13: Spanish building code requirement for floor compartmentation (Source: CTE-DB-SI

(2010). Seguridad en caso de incendio. Ministerio de Fomento: Madrid, Spain page 2-

2) ............................................................................................................................................ 63

Figure 14: UK building regulations for floor compartmentation (Source: UK Building

Regulations Approved Document part B, Clause B3 Diagram 33. NBS RIBA

Enterprises, Newcastle Upon Tyne, United Kingdom )......................................................... 64

Figure 15: Typical detail for conventional unitized curtain walling system for floor

compartmentation .................................................................................................................. 65

Figure 16: Proposed system detailing for floor compartmentation ........................................................ 66

Figure 17: Diagram of the direct sound transmission through glazing. ................................................. 69

Figure 18: Diagram of flanking transmission through conventional system. ........................................ 70

Figure 19: Diagram of flanking transmission through Proposed system system ................................... 70

Figure 20: Structural cross-section of the conventional system ............................................................. 74


185

Figure 21: Structural cross-section of the proposed system ................................................................... 76

Figure 22: Elevational area of wind load assigned to each mullion ....................................................... 78

Figure 23: Structural diagrams (a) cross-section showing out-of plane load distribution and

support condition for conventional system; (b) cross-section showing out-of plane

load distribution and support condition for proposed system; (c) shear distribution and

(d) bending moment distribution for simply supported beam ................................................ 79

Figure 24: Structural diagrams (a) cross-section showing out-of plane load distribution and

support condition for conventional system with a raised stack joint; (b) shear

distribution and (d) bending moment distribution for continuous beam ................................ 80

Figure 25: System depth vs. deflection of the frame curves .................................................................. 84

Figure 26: System depth vs. bending stress curves ................................................................................ 85

Figure 27: Breadth of the section at the cutline vs. shear stress ............................................................. 86

Figure 28: Process map .......................................................................................................................... 92

Figure 29: Projected areas in elevation .................................................................................................. 93

Figure 30: Conventional system as modelled in THERM 6.3.19.0 (a) mullion and (b) transom .......... 95

Figure 31: Proposed system as modelled in THERM 6.3.19.0 (a) mullion and (b) transom ................. 96

Figure 32: Heat flow profiles for (a) conventional system mullion (b) proposed system mullion

(c) conventional system transom (d) proposed system transom .......................................... 100

Figure 33: Temperature profiles and location of areas with minimum inside surface temperature

for (a) conventional system mullion (b) proposed system mullion (c) conventional

system transom (d) proposed system transom ..................................................................... 101

Figure 34: 4-point bending test specimen ............................................................................................ 105

Figure 35: Specimen being tested on Instron 5567 machine ............................................................... 107

Figure 36: Structural diagrams (a) cross-section section showing load distribution and

dimensions between supports and crossheads; (b) shear distribution and (c) moment

distribution for simply supported beam ............................................................................... 107

Figure 37: Geometric properties of the specimen’s cross-section and cutline ..................................... 108

Figure 38: Radius of curvature diagram ............................................................................................... 109

Figure 39: Horizontal shear stress failure in (a) polyester resin and (b) phenolic resin specimens ..... 110

Figure 40: Shear strength summary results .......................................................................................... 110

Figure 41: Modulus of elasticity summary results ............................................................................... 114

Figure 42: Single-lap shear test specimen ............................................................................................ 123

Figure 43: Drilling of two threaded holes in the GFRP bar to attach the displacement

transducers and a smaller hole to allow extra adhesive to flow out when applied ............... 124


186

Figure 44: Adjustment of jigs and blocks to the glass pane (a) photograph (b) schematic section

and (c) 3D sketch ................................................................................................................. 125

Figure 45: Adhesive application process ............................................................................................. 126

Figure 46: Bonding of two metal L plates to place displacement gauges ............................................ 127

Figure 47: Adjustment of jigs and blocks to the glass pane for Dow Corning TSSA specimens (a

and b) photographs (c) schematic section ............................................................................ 128

Figure 48: Single-lap shear test setup (a) schematic elevation (b) specimen being tested on

Instron 5500R machine ........................................................................................................ 130

Figure 49: Adhesive failure at GFRP-adhesive interface in Huntsman Araldite 2047 3mm thick

bond ..................................................................................................................................... 133

Figure 50: Glass substrate failure in 3M DP490 Epoxy 3 mm thick bonds (a) whole glass

breakage and (b) glass peeling off at the edge ..................................................................... 134

Figure 51: Uniform shear stress distribution (Adams et al, 1997) ....................................................... 135

Figure 52: Differential shear stress distribution (Adams et al, 1997) .................................................. 135

Figure 53: Cohesive failure in Dow Corning TSSA Silicone 3 mm thick bond .................................. 136

Figure 54: Failure mechanisms in Huntsman Araldite 2047 Acrylate 3 mm thick bonds with

abraded GFRP (a) glass peeling off at the edge with signs of adhesive plastic

deformation and (b) adhesive failure at GFRP-adhesive interface ...................................... 139

Figure 55: Failure mechanisms in 3M DP490 Epoxy 5 mm thick bond (a) glass peeling off at

the edge combined with cohesion failure and (b) whole glass breakage ............................. 140

Figure 56: Induction of bending moment in single shear lap joint ...................................................... 141

Figure 57: Failure mechanisms in 3M Scotch Weld 2216 B/A Epoxy 3mm thick bonds (a)

adhesive failure at GFRP-adhesive interface and (b) glass peeling off at the edge ............. 142

Figure 58: Load vs. extension curves of bonds (Dow Corning TSSA is excluded for clarity) ............ 143

Figure 59: Scheme design .................................................................................................................... 148

Figure 60: Detail design ....................................................................................................................... 148

Figure 61: Glass-GFRP composite unit (a) full model and (b) magnified view showing meshing

detail .................................................................................................................................... 152

Figure 62: Typical fluctuation of wind velocity in time ...................................................................... 154

Figure 63: Glazing displacement contour plot for Load Case 1 (600s; 750N/m2) ............................... 156

Figure 64: Glazing displacement contour plot for Load Case 2 (1s; 3000N/m2) ................................. 156

Figure 65: Principal Stress Contour Plot for Load Case 1 (600s; 750N/m2) ....................................... 157

Figure 66: Principal Stress Contour Plot for Load Case 2 (1s; 3000N/m2) ......................................... 157

Figure 67: Adhesive Shear Stress Contour Plot for Load Case 1 (600s; 750N/m2) ............................. 158


187

Figure 68: Adhesive Shear Stress Contour Plot for Load Case 2 (1s; 3000N/m2) ............................... 158

Figure 69: GFRP Shear Stress Contour Plot for Load Case 1 (600s; 750N/m2) .................................. 159

Figure 70: GFRP Shear Stress Contour Plot for Load Case 2 (1s; 3000N/m2) .................................... 159

Figure 71: Proposed system as modelled in THERM 6.3.19.0 (a) mullion and (b) transom ............... 164

Figure 72: Heat flow profiles for (a) schematic design mullion; (b) detail design mullion; (c)

schematic design transom and (d) detail design transom ..................................................... 166

Figure 73: Temperature profiles and location of areas with minimum inside surface temperature

for detail design (a) mullion and (b) transom ...................................................................... 167

Documents

UNITISED CURTAIN WALL WITH LOW THERMAL …oa.upm.es/38429/1/BELARMINO_CORDERO_DE_LA_FUENTE.pdf · muro cortina modular con marco de baja transmitancia tÉrmica integrado en el vidrio