Challenging Glass, Conference on Architectural and Structural Applications of Glass

Bos, Louter, Veer (Organizing Committee)

www.bk.tudelft.nl/challengingglass

Improved computational methods for

determining wind pressures and glass
thickness in façades
Kenneth Zammit, Mauro Overend
University of Cambridge, UK. www.gft.eu.com
David Hargreaves
University of Nottingham, UK

Wind induced pressure is a major design consideration for determining the glass
thickness and glass selection in façades. However, the effects of wind loading
history on glass are largely neglected or grossly simplified. The use of
computational techniques, such as Computational Fluid Dynamics (CFD), to tackle
these issues is relatively untested in contrast to other fields of engineering where
CFD is used as a routine design tool. This paper firstly addresses the use of
boundary conditions which maintain the wind speed profile as it varies with height
above the ground, a problem afflicting several CFD studies in the atmospheric
boundary layer. It is then shown how CFD can be used together with wind tunnel
studies to tackle difficult design situations. Subsequently, the effects of fluctuating
wind loads on the structural strength of glass are assessed using transient,
geometrically non-linear analyses and improved glass failure prediction models.
Results are compared to those from current glass design standards where only peak
gust pressures are considered, where it transpires that such a detailed analysis can
give up to a 35% increase in efficiency.

Keywords: Glass strength, Wind, General crack growth model, CFD

1. Introduction

1.1. Background to the problem

New and exciting buildings are being constructed, pushing forward the boundaries of
engineering knowledge. The effects of wind flow over these tall and / or geometrically
complex structures are often outside the scope of wind loading codes of practice and
require detailed wind tunnel investigations in order to establish overall and local wind
pressures.

Modern wind loading codes of practice are based on wind tunnel studies with their
main focus dedicated to overall building loads. These are generally measured using a
high frequency base balance measuring overall shear and overturning moments at the
base of the building. In contrast, façade loads are measured using pressure taps
distributed over the model surface.

Façades are also becoming increasingly intricate, with diverse overall and detailed
geometry often stemming from requirements of energy efficiency. This, along with
pressures to achieve a high profile image desired by clients, has resulted in dramatic
increases in façade cost, sometimes exceeding £1500 per m2. In such façades, wind
tunnel testing can easily be justified for façade design given the significant cost benefit.

Although glass is the predominant material in modern façades, it is a relatively new

material when one considers that the float process was commercially developed in the
1960s. Consequently, glass design codes of practice are very often limited to specific
design situations with restricted design procedures. There is also little agreement
between codes of practice in different countries. As a result, there is a need in the
façade industry to bring wind engineering and glass design together to form a clear
basis for design. This is even more important when considering that the specialist
façade market is international with designers operating in different countries, often
leading to disputes when the design basis is not clearly defined.

1.2. Limitations of existing techniques

Wind loading codes of practice are limited to simple building geometries and offer
little to no guidance on complex façade geometries. Moreover, predicted façade
pressures tend to be considerably higher than those measured in wind tunnel tests as
shown in Table 1. Great care needs to be taken when simple codes are used in complex
situation since these can easily produce unsafe designs.

Table 1: Wind Loading Code wind speeds and Pressures [1]

Result BS 6399 Eurocode ESDU ASCE AS/NZS Measured CFD
[2] [3], [4] [5], [6] [7] [8] Pressures
Design gust speed
40.42 41.15 38.39 37.12 36.31 38.39 38.39
(ms-1)
Max. External
-1301.9 -1453.3 -1174.9 -1064.3 -1050.8 -858.55 -813.37
Pressure (Pa)

Glass codes of practice often use large factors of safety covering different aspects of
glass design making it difficult to adapt to new design situations. As a consequence of
this it is also difficult to address load combinations with different durations, although
this is currently being tackled within the development of prEN13474 [9]. All codes of
practice however, give very limited consideration to the wind-induced stress history
which gives rise to sub-critical crack growth in the glass. The effects of this stress
corrosion are not explicit in the code safety factors and are at best represented by a
generic equivalent load duration.

Wind tunnel testing is based on scaled models (about 1:200) of the entire building,
giving a physically limited number of pressure taps over the whole façade area. This
results in substantial extrapolations of pressure distribution which cannot give an
accurate representation of pressure integration over façade areas. In order to address
the effect of non-simultaneous action of peak pressures, these can be averaged in time
instead of being averaged over an area. However, there currently is lack of agreement
among different countries on such temporal averaging, which can give substantially
different figures. The scales used in wind tunnels also give rise to modelling
limitations, where complex façade geometries cannot be represented, once again
leaving loading values open to interpretation. Recent advances in computational power
enable the capture of transient and simultaneous load data, however these are not
normally used in façade design.

1.3. Paper outline

This paper firstly outlines the developments in Computational Fluid Dynamics (CFD),
addressing simulation issues which often afflict computational wind engineering. This
section also shows how information extracted from CFD analyses can be used to tackle
some of the limitations of wind tunnel testing. The second part of the paper describes
the transient analysis carried out using existing time varying field measurements, to
assess the effects of fluctuating wind loads on the structural strength of glass within a
single skin façade. This is undertaken using transient, geometrically non-linear
analyses and improved glass failure prediction models. Finally, results are compared to
those from current glass design standards where only peak gust pressures are
considered.

2. CFD Simulation

2.1. Existing and future modelling

Although simulations in wind tunnels can give accurate measurements, such
information is limited to a number of particular pressure taps on the building surface or
anemometers placed around the buildings. Instrumentation is expensive and requires
skill for use and calibration. As an alternative, simplified analytical methods can be
used to describe very simple flow problems. However, for more complicated flow
problems, the finite differences method is used to solve the Navier-Stokes equations for
fluid flow which are discretised and solved iteratively.

CFD offers little advantage over experimental tests when overall wind forces for
building stability calculations are required, as the effects of localised differences of
wind pressure on the building surface tend to cancel each other out. However, when
the detailed and localised flow structure is required, such as when pressures are
required for façade design, the set up used in experimental techniques becomes
complicated and the design loading data is very sensitive to errors in localised
pressures. In such cases numerical methods can be useful for determining the detailed
pressures over a façade. Mean pressures are generally much better predicted than peak
pressures in CFD simulations.

Indeed, much of this is due to the prevalent use of steady Reynolds-Averaged Navier-
Stokes (RANS) turbulence models, which were predominantly designed for
streamlined flow in the aerospace, automotive and chemical industries. When surface-
mounted bluff bodies such as buildings are present, RANS models perform particularly
poorly by failing to predict, in particular, the correct separation and reattachment on the
roofs of low-rise buildings [10]. Recently, Hanjalic [11] wrote a critical review of the
future of RANS models for all engineering applications. He indicated that while RANS
models had found successful niches in certain applications, this was not true for all –
indeed this can be said for wind engineering. The future of CFD modelling in the
prediction of structural loads in wind engineering lies firmly in the use of unsteady
simulations, making use of unsteady RANS, Detached Eddy Simulations (DES) and
Large Eddy Simulations (LES). This will put a strain on computer resource but the
switch to unsteady simulations also opens up a number of other possibilities to the
modeller. Rather than focussing on the mean pressures on a building in a well-
characterised Atmospheric Boundary Layer (ABL), unsteady CFD modelling could be
used to assess the validity of the quasi-steady theory. Modelling single, high speed
gusts could become possible, initially only for a few seconds of real time, but with
longer periods and more gusts being added as simulation capabilities increase with time.

2.2. Consistent inlet boundary conditions

Both wind tunnel testing and CFD methods are very sensitive to the boundary
conditions which define the simulated ABL. This means that in either approach, the
mean and gust wind speed alone are not sufficient to perform a simulation. The
‘growing’ of an artificial boundary layer within a wind tunnel is quite well understood
and validated from full scale measurements.

In CFD simulations, this is replaced by formulae defining the inlet conditions for the
flow [12], [13]. However this practice is not widely and accurately adopted,
particularly for transient turbulence models such as LES. Indeed, LES simulations
require the specification of inlet conditions that vary both spatially and temporarily.
The generation of such conditions is the focus of much current research (e.g. Xie and
Castro, [14]) and represents a real sea change in the approach to the specification of
inlet conditions.

30

Inlet

25 15m downstream of inlet

Outlet
20
Height (m)

15

10

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4

U / Uref

Figure 1: Decaying ABL from CFD simulation showing varying wind speed with height

The ABL described by inlet conditions is often not maintained in CFD simulations as
represented in Figure 1. This results in errors easily in the region of 36% on overall
pressure distribution where turbulent flow separation occurs. However pressures on
the windward and leeward faces are predicted with much greater accuracy as shown in
earlier work by the authors [1]. By introducing a shear stress at the top of the boundary
layer as suggested by Richards and Hoxey [15] and modifying the standard definition
of the near wall treatment of the rough surfaces, Hargreaves and Wright [12] manage to
maintain the ABL over a fetch of 4km.

These improvements to the CFD boundary conditions and the move to unsteady
simulations are currently being implemented to assess the effects on surface pressure
distribution accuracy.

2.3. The use of CFD simulation to augment wind tunnel data

As noted in Section 1.2, wind tunnel testing, despite its accuracy has physical
limitations, particularly when addressing façade pressures. When assessing overall
overturning loading, forces due to the action of wind are measured at the base of the
model. This automatically summates the action of pressures over all surfaces without
the need for accurate pressure measurements at particular locations. In this case, CFD
can offer little additional information as long as the building geometry has a single base
and is not aero-elastic both of which complicate wind tunnel testing procedures.

When measuring local pressures for façade design, we are interested not only in
discreet pressure values at particular points over the surface, but also in the variation of
pressure over the surfaces. Only where the building geometry is very simple, with
known distributions of flow extracted from detailed testing, can the wind tunnel be
used to give accurate local pressure distributions for use in structural analysis.

(a) (b)

Figure 2: CFD Pressure distribution diagrams (a) Wind normal to cube (b) Wind 45˚ to cube

In contrast, CFD can give a very detailed description of surface pressures and wind
flows around buildings at any point in the computational model. Therefore, if wind
tunnel testing is used to calibrate a CFD model, it would take relatively little effort to
predict accurate pressure distributions, once standard modelling procedures are defined.
Such information could be used to predict forces on larger cladding elements, façade
secondary structures and also local main structural elements.
Moreover, an accurate CFD model is not limited by scale. Once the pressure
distribution is calibrated, CFD be used to predict pressures on intricate external façade
elements such as shading devices and externally ventilated double skin façades. This
would address a much needed aspect of façade engineering where codes of practice
give very little guidance.

3. General Crack Growth Model transient simulation

3.1. Existing research

A number of studies have been carried out by other authors in attempt to characterise
the effects of fluctuating wind load on glass structural design. Holmes [16] showed
that although upwind velocity fluctuations can be described by a normal (Gaussian)
distribution, pressure histories for side walls of buildings where flow separation occurs
are characterised by different probability distributions. This phenomenon also holds
true for windward walls of low rise buildings where higher turbulence intensities occur.
Holmes [17] then also postulated that this pressure distribution had a significantly
different effect on glass strength, although geometrically non linear effects due to large
plate deflections were ignored.

Das et al. [18] carried out dynamic analysis for vibrating glass plates in simulated
fluctuating wind. They proceeded to show that wind acceleration contributes very little
to fluctuating wind pressure. Maximum principal stresses only were considered in this
study and surface pressure fluctuations were assumed to follow the parent wind speed
fluctuations, which is not the case particularly for side faces of buildings. Dynamic
amplification of stresses is also low on glass plates used in practice, since these have
high natural frequencies.

Reed [19] integrated the pressure fluctuations modelled using a statistical simulation to
estimate the damage on glass panels. Geometrically non linear effects due to large plate
deflections were considered using analytical solutions. In agreement with Holmes,
Reed showed that a non-Gaussian model of windward pressure causes more damage to
glass. However, the statistical superposition of peak stress and critical flaw was only
tackled qualitatively.

Calderone [20] used a slight modification to Browns integral [21] for load duration on
glass. This was done to convert stress history to load history in order to avoid non-
linear transient analyses. Unfortunately, since stress and pressure are not linearly
related, this would require calibration for different load distributions, pane geometry
and glass thickness. A log-normal distribution is also proposed by Calderone and
Jacob [22] to achieve a better fit to weathered glass tests. They propose the use flaw
density and distribution of flaw depths as design parameters, particularly for limit state
design. In these studies, pressures for integration were based on free wind speed
records as opposed to surface pressure data. This may sometimes be unsafe since
pressure distributions are distorted by local turbulence and vary over the surface of a
building as shown by Ko et al. [23]. The latter authors used wind tunnel pressures
from the side face of a building and showed that Gaussian simulation of pressure
fluctuations again give unsafe results for glass strength proposing a statistical
simulation pressure simulation that gives a better fit to data.
Despite this, Calderone and Jacob [22] postulates that pressure history from previous
storms do not contribute glass strength reduction with age. However, it appears that
full meteorological data was not used in their simulations of 20 to 50 year glass
strengths.

3.2. Description of the analysis

In the work carried out by the authors of this paper, façade pressures taken at 5Hz for a
period of 1 hour from existing field measurements on the Silsoe cube [24] were used.
This was done in order to eliminate any modelling errors involved in scale models and
to ensure that a sufficient number of pressure taps were present on the surfaces of
interest. The mean wind speed at the time of measurement was of 9.17ms-1. This was
converted to a 1 in 50 year storm wind speed of 26.74ms-1 expected at the same site of
the cube. The converted pressure history on the windward face of the cube is
represented in Figure 3.

1000
Transient Pressure
Mean Pressure
800

600
Pressure (Pa)

400

200

0
0 10 20 30 40 50 60

-200

Time (min)

Figure 3: Wind Pressure History

Pressures were measured at multiple locations over the façades as shown in Figure 4.
In wind tunnel tests, there are insufficient pressure taps to average readings over the
glazing area. In the UK, temporal averaging is used in to overcome this. BS6399 [2]
requires the averaging of pressures over 1s (in full scale) for cladding areas of up to 5m
in diagonal length. In order to avoid any errors associated with temporal averaging,
pressures measured simultaneously over a height of 3m, were applied to the glass
plates analysed. This also avoided temporal scaling effects due to the required shift of
wind speed from an hourly average of 9.17ms-1 to 26.74ms-1 used in the analysis.
 Figure 4: Silsoe cube tapping points

This gave 18000 pressure readings used for each time step in the Finite Element (FE)
transient analysis, carried out using LUSAS version 14 [25]. The model, shown in
Figure 5, consisted of a 3m x 2m plate, 8mm thick and simply supported on four edges.
Based upon comparison with analytical solutions of non-linear plate analysis, a mesh of
8 x 12 elements was adopted wherein each element was an 8 noded, quadrilateral, thin
shell element with quadratic inter-nodal interpolation.

Membrane stresses induced by the large deflections of the glass plate were taken into
account using the total lagrangian non-linear solver of LUSAS. Ten load steps were
sufficient to give accurate static results, however these were increased to take into
account the variation of pressure within the 18000 data points.

The FE analysis results were post processed using a Visual Basic scripting algorithm
developed by the authors, based upon the work of Overend [26] to calculate the
equivalent uniform stress σp acting on the stressed glass surface (Equation 1). This is
based upon fracture mechanics principles and the random distribution of surface flaws
on the glass. A detailed description of the theory upon which this is based can be
found in Overend et al. [27].

1
1 m
σ p =  ∫ (cbσ max )m dA
 A area  (1)
 Entity: Top Stress
Component: S1

1.5
3.0
4.5
6.0
7.5
15.1
9.0
10.5
12.0
13.5 Z

15.0
X Y
Max 15.1 at Node 120

Figure 5: FE Analysis output

Once the equivalent uniform stress for each load step was calculated, the effects of
crack growth during each 1/5 of a second were integrated using Equation 2, based on
the well known Brown’s integral [21] for load duration effects on glass. This
effectively completes the transformation of the real-world surface stress distribution
caused by the transient wind pressures into an equivalent uniform stress σpe, acting
constantly for a given load duration Teq. This stress would cause the same amount of
crack growth in the glass surface as that in the real world condition.

[ ]
1
 t f σ (t ) n dt  n
σ pe = ∫  (2)
 tr Teq 
However, a lower stress limit exists below which sub-critical crack growth does not
occur. This was taken as 34.6% of the glass strength for instantaneous loads based on
the work of Fischer-Cripps and Collins [28]. A probability of failure, Pf of 1/1000 was
used, giving a minimum equivalent uniform stress σf of 5.4Nmm-2 which is the sub-
critical crack growth threshold stress on a glass plate of Area, A = 6m2 derived from
equations 3 and 4.

K ISCC σ f
k mod = = ≥ 0.346
K IC σs (3)

Pf = 1 − exp( − kA σ sm )
(4)
3.3. Results
Table 2 gives the equivalent constant stress from Equation 2 for different static load
durations, which would give the same amount of crack growth on annealed glass, as
that caused by the transient pressure of a 1 hour storm. Therefore the stress values in
the table and the stresses induced by the transient wind pressures have an equal
probability of failure. Data reported here is generally based on pressures measured on
the windward face of the Silsoe cube with wind incoming perpendicular to this face.
The maximum load acting on the glass plate was identified in the load history, and the
corresponding maximum equivalent uniform stress was of 11.78Nmm-2. If this
maximum stress is used, as is commonly done in design situations, it can be seen from
Table 2, that its equivalent static load duration is of just under 5 seconds.

Table 2: Static Stresses equivalent to 1 hour of wind loading on the windward face
Load duration, Teq Equivalent Static
Stress, σpe (Nmm-2)
10 minutes 8.64
60 seconds 9.98
5 seconds 11.65
1 seconds 12.89

If one were to apply a 10 minute load duration to the glass strength parameters in
design, as recommended in prEN 13474 [9], the appropriate equivalent static stress to
use would be that of 8.64 Nmm-2. It would therefore be very conservative to take the
maximum measured wind pressure, and assume a 10 minute static load duration for the
same load.

It should also be noted that crack growth was only detected for 30% of the duration of
the 1 hour storm. In the remaining 70% of the storm duration, the wind induced stresses
where below the sub-critical crack growth threshold. This is however specific to the
intensity of storm combined with the glass configuration chosen. In the case of a storm
with the most likely annual maximum wind speed (with an annual probability of 0.632
[2]), there would be no effect on crack growth as the maximum stress would be lower
than 5.4Nmm-2. The use of safety factors on the applied wind pressure also
dramatically reduce the effects of weaker storms since the design stresses are higher
than those resulting from the expected storm intensities. Table 3 gives results of the
analysis performed using data from different surfaces and wind direction. All analyses
were performed using the same wind speed, therefore the maximum stress reflects the
varying external surface pressure coefficients (Cpe).

Earlier comparisons by the authors [29], using different international codes of practice
for wind loading and glass design, showed that glass thickness required for a 3m x 2m
annealed glass pane at the site of the Silsoe cube, could vary between 6mm to 12mm
thick, depending on which codes of practice are used. In most cases, prEN13474 gave
the higher thicknesses.
 Table 3: Results for different data sets
Cube Surface Stress due to max. Equivalent duration % of time where
pressure (Nmm-2) of max. stress (s) crack growth occurs
Windward - wind @ 90º 11.78 4.3 30%
Leeward - wind @ 90º 7.75 1.3 0.6%
Roof – wind @ 90º 14.55 1.9 65%
Windward – wind @ 45 º 9.79 1.8 10%
Leeward – wind @ 45 º 8.11 5.6 5%
Roof – wind @ 45 º 10.19 4.4 22%

4. Conclusions and Future Work

Effects of the load history experienced by glass panes in wind storms are largely
simplified or ignored by current glass codes of practice. This has been investigated by
assessing the damage occurring on the surface of a glass pane in terms or crack growth.
Using existing full scale field data and transient FE analysis, equivalent uniform glass
stresses have been calculated, giving an equivalent static load duration of just under 5
seconds for wind pressures.

0.95

0.9

0.85
kmod

0.8

0.75

0.7

0.65

0.6
0 100 200 300 400 500

Storm duration (hours)

Figure 6: Variation of kmod with extrapolated storm duration

Following the recommendations of prEN13474, a 5 second pressure has a load duration

factor kmod of 1. Even if the 1 hour storm history used is repeated 3 times, kmod remains
above 0.94. The recommended effective load duration of 10 minutes in prEN13474
(giving kmod = 0.74) is comparable to a storm lasting 128 hours with no reduction in
intensity over time, as seen in Figure 6. The recommended values therefore seem
largely conservative leading to higher glass thicknesses. A load duration closer to the 1
to 3 second design gusts as given by wind loading codes seems more appropriate,
although further research is required to quantify this precisely.

When designing glass, difficulties may arise in obtaining wind load histories for new
buildings, however conservative assumptions can be taken to cover a range of
situations. Wind tunnel tests for façades are now common practice for large projects
due to direct cost benefits. Data from these tests can be used directly to extract the
wind loading history to design glass more efficiently. In addition, advances in CFD
make it possible to give further information on pressure distribution over façades,
particularly over intricate geometries such as brises soleil and double skin façades.

4.1. Future work

The research reported in this paper represents on-going work, as such the reader should
be aware of the following limitations:
• The effects of multiple storms and crack healing were not considered
• Limited wind data sets were used

Research in the is field is being undertaken jointly by the Glass & Façade Technology
Research Group at the University of Cambridge and the Environmental Fluid
Mechanics Research Group at the University of Nottingham. These studies will address
some of the above-mentioned limitations by:
• Incorporating multiple storm statistics and critical surface flaw design into the
current model
• Performing analysis on further full scale and wind tunnel data sets
• Assessing the adequacy of wind tunnel test temporal averaging for glass
design
• Performing weathered glass testing to refine the failure prediction models
• Performing IGU and laminated glass testing to extend the results of these
findings to different products

In parallel, CFD analysis and wind tunnel testing will be used together to enable the
accurate design of double skin glass façades. This will be done through calibration of
CFD models using wind tunnel test data of large scale façade models. The goal of the
research is to use the results of this calibration to give a universal glass façade design
method combining physical wind tunnel testing, CFD simulation and glass failure
prediction.
5. References
[1] Overend, M. and Zammit, K., Wind loading on cladding and glazed façades. Proceedings of the 2nd
International Symposium on the Application of Architectural Glass, Munich, 2006.
[2] BS 6399-2:1997. Loading for buildings – Part 2: Code of practice for wind loads. BSI.
[3] EN 1991-1-4:2005. Eurocode 1: Actions on structures – Part 1-4: General actions - wind actions. CEN.
[4] NA to BS EN 1991-1-4, Draft 4.10a, 2005. U.K. National Annex to Eurocode 1: Actions on structures
– Part 1-4: General actions - wind actions. BSI.
[5] ESDU Data Item 82026, Revision E, 2002. Strong winds in the atmospheric boundary layer. Part 1:
hourly-mean wind speeds. London: ESDU International.
[6] ESDU Data Item 83045, Revision C, 2002. Strong winds in the atmospheric boundary layer. Part 2:
discrete gust speeds. London: ESDU International.
[7] ASCE 7-05. Minimum Design Loads for Buildings and Other Structures. American society of civil
engineers.
[8] AS/NZS 1170.2:2002. Structural design actions. Part 2: Wind actions. Standards Australia, Standards
New Zealand.
[9] prEN 13474-3:2007. Unpublished Draft. Glass in building - Determination of the strength of glass
panes - Part 3: General method of calculation and determination of strength of glass by testing. CEN.
[10] Richards, P.J., Quinn, A.D. and Parker, S. A 6m cube in an atmospheric boundary layer flow. Part 2:
Computational studies. Wind and Structures, 5(2-4), 2002, 177-192.
[11] Hanjalic, K. Will RANS survive LES? A view of perspectives. ASME J.Fluids Eng., 127, 2006, 831-839.
[12] Hargreaves, D.M. and Wright, N.G. On the use of the k-ε model in commercial CFD software to model
the neutral atmospheric boundary layer. J. Wind Eng. Ind. Aero., 95, 2007, pp.355-369.
[13] Blocken, B., Stathopoulos, T. and Carmeliet, J. CFD simulation of the atmospheric boundary layer:
wall function approach. Atmospheric Environment, 41(2), 2007, pp. 238-252.
[14] Xie, Z.T and Castro, I.P. Efficient generation of inflow conditions for large-eddy simulation of street-
scale flows. Flow, Turbulence and Combustion, 2008, under review.
[15] Richards, P.J. and Hoxey, R. Appropriate boundary conditions for computational wind engineering
models using the k–ε turbulence model. J. Wind Eng. Ind. Aero. 46, 47, 1993, pp. 145–153.
[16] Holmes, J.D. Non-Gaussian characteristics of wind pressure fluctuations. J. Wind Eng. Ind. Aero. 7,
1981, pp. 103-108.
[17] Holmes, J.D. Wind action on glass and Brown’s integral. Eng. Struct., 7, 1985, pp. 226-230.
[18] Das, Y.C., Vallabhan, C.V.G. and Zhang, Y. Nonlinear dynamic response of window glass plates
subjected to fluctutation wind pressure . J. Wind Eng. Ind. Aero. 41-44, 1992, pp. 2437-2447.
[19] Reed, D.A. Influence of non-gaussian local pressures on cladding glass. J. Wind Eng. Ind. Aero. 48,
1993, pp. 51-61.
[20] Calderone, I. New developments in predicting the strength of glass after 20 or more years in use in
buildings. Proceedings of Glass Processing Days, Tampere, Finland 2003.
[21] Brown, W.G. Apracticable formulation for the strength of glass and its special application to large
plates. National Research Council of Canada, Pub. No. NRC 14372, Ottawa, November 1974.
[22] Calderone, I. and Jacob, L. The dangers of using a probabilistic approach for glass design.
Proceedings of Glass Processing Days, Tampere, Finland 2005.
[23] Ko N., You K. and Kim Y. The effect of non-gaussian local wind pressures on a side face of a square
building, J. Wind Eng. Ind. Aero. 93, 2005, pp. 383-397.
[24] Richards P.J., Hoxey R.P. and Short L.J., Wind pressures on a 6m cube. J. Wind Eng. Ind. Aero. 89,
2001, pp. 1553-1564.
[25] FEA, LUSAS modeller for Windows 98/ME & NE/2000/XP Version 14.0-5. FEA Ltd., 2007.
[26] Overend, M. The appraisal of structural glass assemblies. PhD Thesis. University of Surrey, UK. 2002.
[27] Overend, M., Parke, G.A.R. and Buhagiar, D. Predicting failure in glass – a general crack growth
model. J. Struct. Engrg., 133, 8, 2007, pp. 1146-1155.
[28] Fischer-Cripps, A.C. and Collins, R.E. Architectural glazing: Design standards and failure models.
Building and Environment, 30,1,1995, pp.29-40.
[29] Overend, M., Zammit, K. and Hargreaves, D.M. Applications of computational wind engineering in the
design of glass façades. Proceedings of Glass Processing Days, Tampere, Finland 2007.
6. Notation
A surface area tf load duration
cb biaxial stress correction factor tr time at which the applied
k surface strength parameter tensile stress exceeds the
KIC critical stress intensity factor surface pre-compression
(plane strain fracture toughness) σ(t) stress at time t
KISCC plane strain fracture toughness σpe constant equivalent failure
for sub-critical crack growth stress acting for a time Teq
kmod stress corrosion ratio σf surface tensile strength of
m surface strength parameter, glass
Fourier series numerical factor σmax maximum major tensile stress
n static fatigue constant σp equivalent uniform stress
Pf probability of failure σs instantaneous failure stress
Teq equivalent static load duration