See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/44087629

The Strength and testing of window glass

Article  in  Canadian Journal of Civil Engineering · February 2011

DOI: 10.1139/l90-088

CITATIONS READS

21 1,923

2 authors, including:

W. Alan Dalgliesh
National Research Council Canada
114 PUBLICATIONS   488 CITATIONS   

SEE PROFILE

All content following this page was uploaded by W. Alan Dalgliesh on 01 June 2014.

The user has requested enhancement of the downloaded file.

 Ser
TH1
. -

N21d
no. 1673
c.2
$1 National Research
ouncil Canada
Conseil national
de recherches Canada
BLDG. Institute for lnstitut de
--- -- Research in recherche en
Construction construction

The Strength and Testing of Window Glass

by W.A. Dalgliesh and D.A. Taylor

Reprinted from
Canadian Journal of Civil Engineering
Vol. 17, No. 5
October 1990
p. 752-762
(IRC Paper No. 1673)

NRCC 32341
 Abstract

This paper presents some of the background to the new Canadian General Standards Board (CGSB)
standard on the structural design of glass for buildings. It also describes strength testing of window
glass at the Institute for Research in Construction of the National Research Council of Canada using
three quite different testing machines. (i) A dynamic glass testing rig, with a large 2.35 m diameter
piston, applies dynamic air pressures to large panes of glass in the vertical position. Results of 107 tests
to failure show the considerable effect of loading rate on strength. (ii) A loading table applies linearly
increasing pressures to panes of glass in the horizontal position. Results from 47 fifteen-year-old
windows confirm assumptions about the deterioration of strength with in-service use of the new CGSB
standard on the structural design of glass for buildings. (iii) A ring-on-ring tester applies linearly
increasing loads to small 180 x 180 mrn samples; the equivalent of coupon tests for steel or cylinder tests
for concrete. The resulting uniform tensile stress fields avoid failures starting at the edges of the
samples. Some results are given. Testing provides the basis for the structural design of glass cladding
as it does for other structural materials. A great deal more testing of window glass is required.

Key words: glass strength, window strength, testing, strength deterioration, rate effect, dynamic
loading.

Ce document expose le contexte general d'elaboration de la nouvelle norme de l10fficedes normes

genbrales du Canada (Ongc) concernant la conception structurale du verre destine aux biitiments. I1
dkrit aussi le contrale de la resistance des vitres effectuk l'Institut de recherche en construction du
Conseil national de recherches du Canada B l'aide de trois appareils fort differents. (i) Un appareil de
contrdle dynamique du verre comportant un gros piston de 2,35 m de diametre applique des pressions
d'air dynamiques B de grandes vitres en position vertical. Les ksultats de 107 essais jusqu'a rupture
revelent l'effet considerable de la vitesse de mise en charge sur la rkistance. (ii) Une table de mise en
charge applique des pressions augmentant -
linbairement sur des vitres en position horizontale. Les
resultats des essais effectuks sv* " ' =theses formul6es dans
la nouvelle norme de l'Ongc c biitiments, concernant
la diminution de la rkistano u sur anneau applique
une charge augmentant line) : Nquivalant de l'essai
de compression simple pou champs uniformes de
contrainte de traction qui ards de 1'6chantillon.
Certains resultats sont e x r - conception structurale
des fapdes de verre, corn ~udraitfaire beaucoup
plus d'essais dans le d o m

Mots c l k : resistance du Aistance, effet de la

vitesse, mise en charge d y
 The strength and testing of window glass1
W. ALANDALGLIESH
AND DONALD
A. TAYLOR
Institute for Research in Construction, National Research Council of Canada, Ottawa, Ont., Canada KIA OR6
Received August 4, 1989
Revised manuscript accepted March 26, 1990

This paper presents some of the background to the new Canadian General Standards Board (CGSB) standard on the struc-
tural design of glass for buildings. It also describes strength testing of window glass at the Institute for Research in Construc-
tion of the National Research Council of Canada using three quite different testing machines. (i) A dynamic glass testing
rig, with a large 2.35 m diameter piston, applies dynamic air pressures to large panes of glass in the vertical position. Results
of 107 tests to failure show the considerable effect of loading rate on strength. (ii) A loading table applies linearly increasing
pressures to panes of glass in the horizontal position. Results from 47 fifteen-year-old windows confirm assumptions about
the deterioration of strength with in-service use in the new CGSB standard on the structural design of glass for buildings.
(iii) A ring-on-ring tester applies linearly increasing loads to small 180 x 180 mm samples; the equivalent of coupon tests
for steel or cylinder tests for concrete. The resulting uniform tensile stress fields avoid failures starting at the edges of the
samples. Some results are given. Testing provides the basis for the structural design of glass cladding as it does for other
structural materials. A great deal more testing of window glass is required.
Key words: glass strength, window strength, testing, strength deterioration, rate effect, dynamic loading.

Ce document expose le contexte gknkral d'klaboration de la nouvelle norme de I'Office des normes gknkrales du Canada
(Ongc) concernant la conception structurale du verre destink aux biitiments. I1 dkrit aussi le contrBle de la resistance des
vitres effectue h 1'Institut de recherche en construction du Conseil national de recherches du Canada h l'aide de trois appareils
fort diffkrents. (i) Un appareil de contrBle dynamique du verre comportant un gros piston de 2,35 m de diamttre applique
des pressions d'air dynamiques h de grandes vitres en position verticale. Les resultzits de 107 essais jusqu'h rupture rkvtlent
1'efPet considerable de la vitesse de mise en charge sur la rksistance. (ii) Une table de mise en charge applique des pressions
augmentant lintairement sur des vitres en position horizontale. Les rksultats des essais effectuks sur 47 fenetres vieilles de
15 ans confirment les hypothkses formulkes dans la nouvelle norme de l'Ongc visant la conception structurale du verre destint
aux bltiments, concernant la diminution de la rksistance avec les annees de service. (iii) Un essayeur anneau sur anneau
applique une charge augmentant linkairement sur des kchantillons de 180 x 180 mm; c'est l'huivalent de l'essai de compres-
sion simple pour le bkton ou de l'essai de traction pour l'acier. Les champs uniformes de contrainte de traction qui en rksultent
evitent une fissuration partant des bords de l'khantillon. vertains resultats sont exposks. L'expkrimentation constitue le
fondement de la conception structurale des fa~adesde verre, comme dans le cas de autres matkriaux de construction. II
faudrait faire beaucoup - plus
- d'essais dans le domaine des vitres.
Mots cl6s : resistance du verre, rksistance des fenetres, essais, diminution de la rtsistance, effet de la vitesse, mise en charge
dynamique.

Can. J. Civ. Eng. 17, 752-762 (1990)

Introduction they compute the wind (and snow) loads, but the advice of
Structural testing is a necessary part of the development of others (glass companies) is commonly sought for the appropri-
methods for the design of window glass. Glass, in this regard, ate type and thickness of glass for the building.
is no different than other materials such as concrete, steel, The first comprehensive Canadian treatment of the struc-
timber, and masonry. Understanding their properties and tural design of window glass was released by the Canadian
behaviour required a great deal of testing and field observation General Standards Board (CGSB) in December 1989.
over many years, and the research continues. Similarly, test- Requested by the Associate Committee of the National Build-
ing of window glass defines its properties and strength, includ- ing Code, Structural Design of Glass for Buildings2 repre-
ing deterioration of strength with use. Field and laboratory sents 6 years of teamwork by researchers, designers, and
studies of the behaviour of windows and glass curtain walls are industry experts in both Canada and the United States. Its ori-
required for the satisfactory application of glass in buildings. gins actually go back 20 years, to a synthesis by W. G. Brown
In addition to providing a view, glass acts structurally as a at the National Research Council of Canada (NRC) of several
cladding material to separate the internal and external environ- failure theories as well as tests to failure of small and large
ments. As cladding, it is subject to building code require- windows (Brown 1969, 1974). He explained apparent dis-
ments. Clause 5.7.1.2 of the 1990 National Building Code of crepancies among the tests and proposed a design procedure.
Canada (NBC) requires glass to be designed to resist the loads Beason (1980) and Beason and Morgan (1984) developed
specified in the Code (Sect. 4.1) and installed according to Brown's approach into the failure prediction model used in the
good engineering practice. Many engineers and architects CGSB standard.
know little about the design of glass for buildings. Typically, CGSB 12.20-M89, "Structural Design of Glass for Build-
ings," gives tables of annealed glass resistance to lateral pres-
NOTE:Written discussion of this paper is welcomed and will be
received by the Editor until February 28, 1991 (address inside front 2Standard CGSB-12.20-M89 "Structural Design of Glass for
cover). Buildings." Denoted in the text as the Standard or the CGSB
'NRC No. 32341. standard.
Printed in Canada 1 Imprim6 au Canada
 DALGLIESH AND TAYLOR 753

sure as a function of area and shape for rectangular panes ratio is about 0.22, elastic modulus 70 000 MPa, and, from
simply supported on four edges. The recommended glass theoretical considerations, tensile strength should be around
strengths, computed by the failure prediction model, are sig- 14 000 MPa. In fact, the nominal tensile stress at failure under
nificantly lower on average (about 30%) than indicated by practical engineering conditions is well below 70 MPa. Sub-
tests of new glass because of allowances made for the loss of microscopic flaws, randomly distributed over the surface,
glass strength with in-service use. The CGSB committee, in grow slowly under tensile stress (stress corrosion) until a criti-
concert with a U.S. committee working on an ASTM standard cal level is reached at some particular flaw, at which point
for glass design, agreed that design should be based on the cracks propagate at speeds of up to 2000 mls. The number of
reduced strength of glass after some years of normal use. cracks formed increases with the amount of elastic energy
Test results available from around the world refer to about stored in the glass just before failure.
5000 panes broken in total (Orr 1957; Bowles and Sugarman ~ n ~ i n e e r i nstrength
g of glass depends on the interaction
1962; Hershey and Higgins 1973; Ishizaki et al. 1974; Jonsson between tensile stress and surface flaws normally invisible to
1977; Abiassi 1981; Johar 1981, 1982; Krall et al. 1981; the naked eye. Stress corrosion exacerbates the flaws under
Norville and Minor 1985; Kanabolo and Norville 1985). Of conditions of moisture and tensile stress. The experimentally
' these, only about 500 involved glass exposed to service condi- verified consequences of surface flaws and stress corrosion are
tions in buildings (Norville and Minor 1985). The lower variation in strength of 20 -25 %; strength reduction approxi-
strength found in tests of in-service glass can be related to sur- mately as the 117th power of surface area; strength decrease
face damage from normal aging and weathering and is approximately as the 1116th power of load duration under a
reflected in the parameters used to fit the failure prediction constant load; and strength loss as glass weathers and ages.
model to experimental data. Stress corrosion can be suppressed, and flaw severity con-
Another feature of the CGSB standard is that it explicitly trolled under laboratory conditions. For windows in normal
addresses the difference between fluctuating wind pressures use, polymer coatings have been proposed to prevent moisture
and the 1-min constant pressure commonly used to describe from entering the micro-flaws (Michalske and Bunker 1987),
glass strength. The dependence of glass strength on load dura- but the most common surface treatment to raise the threshold
tion had to be specified before the failure prediction model for occurrence of surface tensile stress is heat strengthening or
could be fitted to data from tests to failure at different loading even full tempering. Full tempering is not always favoured
rates, and before it could be used to predict resistance to wind because there is some loss of flatness, and some possibility of
effects. spontaneous breakage, i.e., failure internally generated by
This paper describes the tests aimed at relating loading rate impurities such as nickel sulfide inclusions within the glass.
and strength (Johar 1981, 1982) and discusses the fit of the Flaw growth due to stress corrosion is inhibited as long as ten-
failure prediction model to the data. It also presents additional sile stresses (induced thermally or by bending) are less than
new data on in-service glass strength. Small sample tests using the residual compressive stresses left by heat treatment.
concentric rings are explored as an economical way of adding For design, a theoretical framework is required. A theory
to the data base, and other research needed to improve the based on stress corrosion is a key to handling fluctuating pres-
Standard is recommended. sures, such as wind loads. Combined with Weibull's theory
The main purpose of the paper is, however, to explain some (1939) on the strength of brittle materials, the decrease in
of the background testing and theory used to develop the new strength with increasing load duration can be described by a
Standard (without attempting an authoritative discussion). damage criterion for failure.
Further, it aims to describe testing and testing machines used
in research at the Institute for Research in Construction (IRC) Dependence of glass strength on loading duration and rate
of the National Research Council of Canada in support of the Tests to failure of both small and large specimens of glass
Standard and in support of the construction industry's needs show strength decreasing with increasing duration of constant
for more information on window glass. load. Stress corrosion explains the effect of fluctuating pres-
sures as well as load duration and the varying loading rates
Mechanical properties of glass used in different tests. One way of describing the duration or
rate effect is to fit a straight line by least squares to a logarith-
Glass is an inorganic product of fusion that has been cooled mic plot of failure stresses, s f , vs. times to failure, tf:
to a rigid condition without crystallization. Most architectural
glazing is soda-lime glass: silica sand 72%, soda and potash
13% , lime 11% , magnesia 3 % , and alumina 1% . The compo-
where C1 is a constant.
nents are mixed with broken waste glass for better heat trans- Of particular interest is the second fitting parameter n, con-
fer and heated to 1500°C. Almost all flat glass is now "float sistently found to be 16 or 17 for soda-lime glass in small
glass" formed by floating it on molten tin, a process invented specimen tests, and in many tests of large panels as well. Some
30 years ago. For ease in cutting, after the temperature drops
large panel tests indicated larger values, and hence a weaker
to 550°C, further cooling is controlled to minimize residual
dependence on load duration (Ishizaki et al. 1974). As n plays
surface stresses, producing annealed float glass. After cutting a key role in the practical application of the failure prediction
to size, glass plates may be reheated and then their surfaces model on which the CGSB standard is based, further tests
"set" by rapid cooling for a short time with many air jets. The
were done to resolve this uncertainty about the rate effect for
central core cools more slowly, putting the hardened surfaces large panels. Before describing those tests, the role of n in
into compression and the core into tension. The result is heat adjusting for differing load histories is discussed briefly.
strengthened or prestressed glass, which is much stronger than
annealed glass. Brown 's damage criterion
Glass under load deforms elastically until sudden failure is Brown's design procedure (1969, 1974) includes a simple
I form of cumulative damage criterion to account for the growth
I initiated at a surface flaw under high tensile stress. Poisson's
754 CAN. J . CIV. ENG. VOL. 17, 1990

FIG. 1. Front view of dy the glass will be installed.

FIG.2. Side view of the dynamic glass testing rig showing the door open, the large piston, and the controller on the right.

of micro-flaws under tensile stress. If the stress at a particular reduces to the following expression of load duration effect:

*
flaw is given by a function of time, s(t), then failure is
assumed to be imminent at time when the integral of r(t)
raised to the nth power reaches a critical value:
[3] P60 = Pr[$]'

where Pf is the constant pressure causing failure after tf s and

[21 &it = 1: [s(t)ln dt P60 is the constant pressure causing failure after 60 s. P60 is
the "equivalent 60-second pressure" commonly specified for
Under constant pressure, stresses are constant, and [2] design.
 DALGLIESH AND TAYLOR

TABLE1. Test results for 107 panes of new glass 1524 x 2438 mm loaded to failure at
five loading rates

Equivalent
Loading Time Failure 60-second Fitted* Centre
rate to fail pressure pressure, pressure deflection
O<Pa/s) (s) (Wa) Pa (Wa) (Wa) (mm)
No. tested: 21
Mean
Std. dev.
Min
Max
cov (%)
No. tested: 21
Mean
Std. dev.
Min
Max
cov (%)
No. tested: 22
Mean
Std. dev.
Min
Max
cov (%)
No. tested: 24
Mean
Std. dev.
Min
Max
cov (%)
No. tested: 19
Mean
Std. dev.
Min
Max
cov (%)
*Fitted means: 5.90 kPa/(time to failure, s)0-067found by log-log least-square fit (Fig. 3). Expo-
nent for load duration effect on pressure is 110.067 = 14.9.

Structural testing to determine glass strength requires pre- Brown suggested that near failure, stress can be represented
ferably 20 - 30 specimens; the strongest in a group of 20 could approximately by a power law function of pressure. Stress in
easily be more than twice as strong as the weakest. Conse- [2] and [4] can then be replaced by pressure, but a different
I
quently, it is impractical to apply the same constant load to all value of n will apply.
and wait for each to break. For n = 16, if the weakest pane The damage criterion was used by Dalgliesh (1980) to con-
broke after 60 s, [3] indicates that the time to failure for one vert randomly varying wind pressures to equivalent 60-second
I twice as strong would be 45 days. loads. The object was to make a rational connection between
Pressures increasing linearly with time until failure are more wind load assessment and glass selection procedures. About
convenient for load tests, but unless stress is linearly propor- the same time, finite element and finite difference methods
tional to pressure, it is not permissible to substitute pressure were used to compute the relation between pressures and sur-
for stress as in [3]. A linear relation between pressure and face stresses as a function of plate geometry (Moore 1980;
stress is unlikely, since glass in windows deflects more than its Beason 1980). With these analytical methods as incentives,
thickness under lateral loads causing failure; large deflection financial support was found for IRC to conduct large-scale
theory applies and membranes stresses are mobilized. The tests to verify loading rate effects (and define n), but, first,
damage criterion does, however, also provide an expression equipment to do the testing was required. With some foresight
for load duration effect in terms of linearly increasing stress: it had already been procured.

Design of IRC's dynamic glass testing rig

Between October 1977 and April 1979, the Ontario Research
where Sm is the constant stress causing failure in 60 s. Foundation designed and built for the Institute for Research in
756 CAN. J. CIV. ENG. VOL. 17, 1990

'
1 ;7:,
.
1
1
*
I

t
I

-
\
f
m
I

a
I I I I

0
a
30
-
0

a
W
n
V . "

3.0-
-
I I I I I I

- - MEAN 2.41, cov 21 % (WEIBULL CURVE)

-
-
-
-
MEAN 2.43, cov 24 % (DATA)
I I I -
-

*\
.
CO
-
-

-
1
LEAU SQUARES LINE:
SLOPE - 0.067
I
3 +.
-. -
-
-
f - -
-
I
- -
I I I I I J I I
1. 6 0 20 40 60 80 100
0.1 1 10 100 1000
PROBABILITY O F FAILURE, %
TIME T O FAILURE, s
FIG.4. Ratio of failure load to 8 per 1000 load vs. probability of
FIG.3. Effect of loading rate. Strength of glass plates decreased as failure for 107 panes of new glass from one batch. The fitted Weibull I
loading rate decreased (i.e., as load duration increased). curve starts at a failure rate of 8 per 1000 using m = 7 and So = I
40 MPa.

Construction of NRC, first a small prototype, then a full-sized

WING NUT
glass testing machine (Stournaras 1979). Its initial use was to
study the effect of loading rate on the strength of window 4 mm GLASS
glass. The glass is loaded by evacuation of air and in order to
ensure fast and uniform loading, a big sealed piston 2.35 m in 9 mm REMOVABLE PLYWOOD
diameter and 300 mm thick forms most of the back face of the E
E
rig and generates the vacuum (Fig. 1). For safety, at rupture
z. 4

Lw\-
2"x 6"JOIST @ 230 mm ON CENTRE
the glass is sucked in towards the rig. Initially a hydraulic
cylinder controlled by an electric servomechanism retracted
the big piston. In 1988, however, a large hydraulic pump with
'2 LAYERS OF 19 mm PLYWOOD
electronic controller was installed which can also drive the
large piston back and forth, up to 75 mm each way, applying \TO VACUUM CLEANER
alternating pressures if required. The pressure variations can
be sinusoidal, square wave, or triangular and can be applied FIG. 5. Section through loading table.
at many different rates. The large piston rides on an impervi-
ous inflatable seal which reduces friction and prevents small specimen tests of annealed soda-lime glass consistently
leakage. show an exponent of n = 16 to 17 (equation [2]) for tensile
The glass panels are mounted vertically on glazing tapes or stress at the site of the flaw causing failure, some large speci-
in their frames in the 3 X 3 m front face of the rig. The face men tests indicated that n = 22 (Ishizaki et al. 1974). In con-
is hinged at one side and can be opened like a door (Fig. 2) trast, an industry guideline (PPG 1979) for the relation of
to gain access to the interior. The "door" was built to take breaking stress to load duration indicated that n = 13.
1524 x 2440 mm windows but has been modified for 1524 x The first series of tests comprised tests to destruction of
1524 mm sizes also and can be changed to accommodate other 90 plates nominally 6 mm thick, using three different loading
sizes up to 3 x 3 m and perhaps beyond. It can also take other rates (0.15, 1.5, and 15 kPa/s) (Johar 1981). Batches of
types of cladding including roofing materials. 30 panes of glass were selected from each of three glass
The maximum negative pressure applied by the machine manufacturers (Canadian Pittsburgh Industries, Libbey -
during commissioning in 1979 was 18 kPa. Loading rates Owens -Ford, Pilkington Glass Industries). Ten of each batch 4
applied during the initial testing ranged from 0.0025 to were tested at each of the three loading rates. Nondestructive
55 kPa/s. These rates were sustained to failure of the glass - tests on a tempered, strain-gauged plate were also conducted.
0.25 s to 30 rnin. The next section describes the initial and The nondestructive tests established relations between uniform
other tests conducted using the dynamic glass testing rig. lateral pressure and surface tensile stresses at several locations
on the plate. The plates were 1524 x 2438 mm, tested verti-
Testing large glass plates using the dynamic glass cally against 12 mm wide neoprene gaskets (55 Durometer
testing rig hardness) with a constant clamping pressure of 1 N/mm.
Three glass manufacturers (Canadian Pittsburgh Industries, Lateral support was continuous along all four sides. The
Libbey -Owens - Ford, and Pilkington Brothers Canada) results of the first series showed significant correlation
joined with NRC's Division of Building Research (now IRC) between loading rate and breaking strength, but were unsatis-
in directing and funding two more contracts with Ontario factory as a means for narrowing down the estimate for n. One
Research Foundation from late 1979 to the summer of 1982 batch of 30 samples had a higher average strength than the rest
(Johar 1981, 1982). The objective of this joint research pro- (statistically significant at the 5 % level); although the average
gram was to reduce uncertainty about the application of the value of n for the other two batches of 30 samples was about
cumulative damage criterion to large glass plates. Although all 16, the stronger batch suggested a value as high as 21.
 DALGLIESH AND TAYLOR 757
A second series of tests (Johar 1982) on windows of the same r (long-to-short dimension). A is the area of the plate, and A.
size was more conclusive; 107 specimens from one batch, tested is a reference area taken as 1 m2.
as above but at five rates (nominally 0.0025, 0.025, 0.25, When pressure is applied to a glass plate, each flaw on the
2.5, and 25 P a l s ) , gave the following relation between the tension surface experiences a certain stress depending on its
natural logarithms of failure pressure, Pf, and time to failure, location and orientation. Severe flaws in high-stress regions of
tf (see [I]): the plate contribute most to the probability of failure, but all
combinations of flaw severity and tensile stress make their
contributions as well. The summation of all contributions over
Average thickness and dimensions between support lines, the whole plate is represented by the function Beason
used for calculation of the results given in this paper, are 5.82, (1980) and Beason and Morgan (1984) used a specially written
1492, and 2407 mm. Table 1 summarizes the results of the 107 finite difference program to compute a risk factor, R, from
tests, and the effect of loading rate is shown graphically in which Sm,p,rwas calculated, as a function of m, p, and r for
Fig. 3. Clearly, in accordance with Brown's criteria, as the ranges of aspect ratios, pressures, and m.
loading rate drops and load duration rises, the strength
decreases. The slope coefficient is 0.067 with a 95% confi-
dence interval of f0.16. The inverse of the slope gives a least- where E is the modulus of elasticity of glass and h is the plate
square estimate for n of 14.9 (see [5]). thickness. In their Table 2, Beason and Morgan (1984) tabu-
>
Using n = 14.9, equivalent 60-second pressures were calcu- lated values of R, the results of hundreds of runs, in nondimen-
lated (Table 1). Although the mean failure pressures for the sionalized, normalized form for simply supported rectangular
five rates were not identical, their coefficient of variation plates.
dropped from 24% to 8 % and the range from the highest to Failure data for in-service glass was collected previously
the lowest mean dropped from 2.84 to 0.71 kPa. Simiu and and fitted to [7]. It yielded the parameters m = 7 and So =
Lechner (1987) tested the validity of estimating 1-min loads 32.1 MPa for use in the CGSB standard. New glass can be
from the same data using n = 14, and concluded that the varia- expected to require a considerably higher value of So for best
tion among their five means was "shown statistically to be fit. Figure 4 demonstrates the generally satisfactory fit of the
relatively unimportant, though not necessarily negligible." Weibull distribution to the 107 tests of new glass at the Ontario
A value 14.9 for n applied to pressure can be reconciled Research Foundation, using m = 7 and So = 40 MPa. Since
with n = 16 or 17 for- stress by considering the relation the industry (and CGSB) standard allows 0.8% breakage rate
between pressure and stress in the final stages of loading to at the design load (8 failures per 1000 windows), the fitted
failure. Nondestructive tests were conducted on one strain- Weibull curve in Fig. 4 starts at a probability of failure of
gauged tempered pane (Johar 1981). The tests permitted inter- 0.8%.3
polation of stresses, a, corresponding to failure locations on The 107 tests have been used by others in the development
the 107 panes, for pressures, P, up to failure. A power law of design methods (Simiu and Lechner 1987; Simiu and
relation was fitted in each case: Hendrickson 1987).
Testing large panes using the loading table
where K is a constant. The exponent 0, which is a function of At the Institute for Research in Construction large panels of
aspect ratio and thickness, varied with failure location on the glass to be loaded to failure under ramp loads are tested in the
plate, but for most failure origins was in the range 0.85 < horizontal position on a loading table. Because of the difficulty
/3 < 0.95. To apply to pressure, the exponent for stress should of accommodating a variety of window sizes with one loading
be multiplied by /3. Thus, if n = 16 for stress, 13.6 < n < table, it is convenient and relatively inexpensive to build load-
15.2 for pressure; if n = 17 for stress, 14.4 < n < 16.2 for ing tables of the required sizes from wood.
pressure. Loading tables are shallow airtight boxes with the panes to
The failure prediction model uses the Weibull distribution, be tested forming their upper surfaces; two at IRC were
on theoretical grounds (Weibull 1939), to describe the sta- designed for glass failure loads ranging from 10 to 25 kPa.
tistical variation so evident within each loading rate group Approximately 24 cm deep (inside dimensions), they were
b (Table 1). The two-parameter Weibull distribution is defined made from kiln dried select structural pine for sides and
by the mode (the most frequently occurring value), So (MPa), purlins and one or two layers of 19 rnm fir plywood for the
and the nondimensional shape parameter, m. The two param- bottom as shown in Fig. 5. Sides and bottom were glued and
eters are interrelated; when the Weibull probability distribu- screwed together to make them airtight. A commercial glazing
tion is fitted to data, a change in m requires a change in So. gasket of neoprene rope surrounded by mastic was fixed to the
The cumulative distribution is given by top of the test frame so that the neoprene core was 6 mm in
from the edges of the glass when the pane was placed on top
[71 POF = 1 - exp(-B) of the frame. In short, the glass was simply supported. The
where POF is cumulative probability of failure, and edges of the glass were taped to the test frame to prevent leak-
age. A frame of wood with glazing gasket on its underside
-
3The 8 per 1000 load used for normalization in Fig. 4 was 1.57
kPa. Reference to Table 2 in Beason and Morgan (1984) to obtain this
The quantity Sm,p,rhas units of stress (MPa) and is a func- number will lead to inaccuracies due to interpolation. Beason recalcu-
tion of the Weibull shape parameter, m, the lateral pressure lated Table 2 for the CGSB committee to give smaller interpolation
applied to the plate, p ( P a ) , and the aspect ratio of the plate, intervals.
758 CAN. J. C N . ENG. VOL. 17, 1990

TABLE2. In-service glass from Thompson Residence (exterior (weathered) surface tested in tension)

Failure
Span location
Failure Failure Loading Load
load time P, Thickness X, length Y, width X Y Deflection rate (actual1
No. (Wa) 6) (Wa) (mm) (mm) (mm) (mm) (mm) (mm) (Wals) calculated*)
1 3.46 47 2.84 4.10 1300 905 1300 0 10.8 0.074 1.54
2 5.38 77 4.57 4.00 1300 905 220 180 15.9 0.070 2.54
3 2.79 39 2.26 4.10 1300 905 260 635 10.7 0.072 1.23
4 6.17 85 5.27 4.00 1300 905 340 275 16.0 0.073 2.93
5 5.29 73 4.47 4.05 1300 905 1275 15 15.4 0.072 2.46
6 4.87 68 4.10 4.00 1300 905 14.6 0.072 2.28
7 6.55 91 5.62 4.07 1298 897 180 240 17.6 0.072 3.07
8 4.89 68 4.12 3.90 1300 930 1050 855 17.8 0.072 2.37
9 6.17 90 5.29 4.00 1300 929 145 710 20.4 0.069 2.97
10 5.88 81 5.01 4.00 1300 930 470 182 21.1 0.073 2.81
11 5.29 73 4.47 3.93 1300 928 1050 185 21.0 0.072 2.55
12 6.69 93 5.75 3.95 1300 925 1225 250 20.0 0.072 3.26
13 4.39 60 3.66 3.84 1300 925 335 195 16.0 0.073 2.14 I
14 6.30 86 5.39 3.90 1300 924 385 800 19.4 0.073 3.09
15 5.63 79 4.79 3.95 1300 925 620 765 0.071 2.71
16 6.64 91 5.70 3.88 1300 930 1220 785 18.8 0.073 3.29
17 6.68 91 5.73 4.04 1300 925 480 815 19.9 0.073 3.18
18 7.06 95 6.08 3.93 1300 900 60 0 17.5 0.074 3.44
19 5.40 75 4.58 3.90 1300 900 1310 0 15.2 0.072 2.61 ,

20 5.63 76 4.77 4.09 1300 900 210 725 0.074 2.60

21 6.04 86 5.16 4.01 1300 895 65 930 15.3 0.070 2.86
22 3.53 51 2.92 4.00 1300 900 0.069 1.62
23 4.05 59 3.38 3.97 1300 899 1140 180 14.0 0.069 1.89
24 6.00 88 5.14 3.86 1300 930 1230 200 18.5 0.068 2.99
25 6.44 94 5.54 4.00 1300 900 1180 800 17.2 0.069 3.08
26 7.14 103 6.18 3.96 1300 975 lo00 190 20.1 0.069 3.58
27 5.89 86 5.04 4.00 1300 975 1230 150 19.9 0.068 2.89
28 5.89 84 5.03 3.96 1300 975 1080 880 18.4 0.070 2.92
29 4.80 69 4.04 3.96 1300 975 1170 170 16.3 0.070 2.34
30 5.87 85 5.02 3.91 1300 975 310 810 18.2 0.069 2.95
31 2.69 40 2.19 3.83 1340 916 320 190 13.7 0.067 1.32
32 3.00 45 2.46 3.77 1340 916 380 265 16.6 0.067 1.50
33 3.10 45 2.54 3.69 1340 916 185 780 14.6 0.069 1.59
34 4.27 61 3.57 3.75 1342 916 220 800 17.1 0.070 2.20
35 3.32 49 2.73 4.00 1357 1300 1090 230 20.1 0.068 1.97
36 3.75 54 3.11 4.05 1356 1300 1120 1180 0.069 2.20
37 3.75 55 3.11 3.93 1356 1300 1150 190 21.4 0.068 2.29
38 5.45 81 4.64 4.01 1358 1300 160 1270 21.8 0.067 3.34
39 5.67 82 4.83 4.81 1358 1300 210 1150 21.6 0.069 2.82
40 3.23 49 2.66 3.80 1374 1342 1130 360 21.0 0.066 2.12
41 2.31 35 1.86 3.83 1374 1342 140 145 17.4 0.066 1.47
42 3.64 53 3.01 4.05 1300 1062 175 220 16.0 0.069 1.78
43 3.93 57 3.27 3.94 1300 1062 1335 60 17.0 0.069 1.99
44 5.04 76 4.27 4.04 1300 1065 1245 230 20.1 0.066 2.54
45 5.31 78 4.51 3.86 1300 1065 0 0 21.0 0.068 2.81 I
46 4.66 86 3.98 3.87 1300 1065 1200 155 19.9 0.054 2.48
47 3.71 54 3.08 3.74 1300 1065 1150 940 17.7 0.069 1.99
*Calculated load from CGSB standard.

(directly above the gasket under the glass) was placed on top failure in tension, the application of tape or other films on the
of the glass at the edges and bolted down to prevent the corners compression side of the loaded pane does not affect its strength
of the pane from lifting during loading. This frame was similar during testing.
to that specified in ASTM-E997-84 (ASTM 1984). A heavy-duty vacuum cleaner is used with a voltage con-
A clear plastic tape is applied to the top surface of all panes. troller to evacuate the air from the test box such that the pres-
This provides safety when the glass implodes into the testing sure on the glass increases at a predetermined constant rate
box. It also keeps the pieces together so the failure origin can until failure. The rate is chosen to precipitate failure in about
be determined. Panes are placed on the test frame with this 60 s on average. Pressure in the test box and center deflection
film on the compressive (top) side of the glass during testing. of each pane are measured continuously by a pressure trans-
As glass is an elastic material which suffers brittle (sudden) ducer and LVDT (electronic deflection gauge) and recorded on
 A N 0 TAYLOR 759

I I I I I I I I I
0 20 40 60 80 100 8 %OF LENGTH OF CLEAR SPAN (LONG SIDE)
I PROBABILITY OF FAILURE, %
FIG.7. Failure locations for 45 panes of 15-year-old glass from
Ro. 6 . Probab~litydlstribuuon for 47 tests to failure of 15-year-old Thompson Residence - two break origins were not identified.
annealed glass. The failure load is divlded by the design load from
"6 the new CGSB standard (1989) (So = 32.1 MPa, m = 7).
break. Use of the design value according to the Standard for
computer and tape recorder. Pressure is also recorded on an normalization assumes that [71 and [8] properly account for
XY plotter. Failure origins are noted and used to compute variations in aspect ratio and area from window to window.
approximate maximum principal tensile stresses in the glass at The results are listed in Table 2 and are compared in
failure. If the glass is heat-strengthened, samples of broken Fig. 6 with the theoretical Weibull probability distribution for
glass are kept from each pane. They are used to determine glass designed according to the CGSB standard. It appears that
thickness and the locked-in prestress due to the heat treatment. the reduction in strength of the glass from the Thompson Resi-
dence is close to the reduction obtained from the data used to
Previous tests by others develop the Standard. Table 2 also includes the X and Y coor-
As noted in the introduction, researchers over the years have dinates (long and short side, respectively) of the initial break-
acquired the results of over 5000 tests to failure conducted on age locations. Break origins were concentrated in the highly
loading rigs of various designs. The tests included samples of stressed comers of the plates where theoretical bending
all commercially available thicknesses and ranged in area from stresses are highest; Fig. 7 shows them all plotted in one
0.5 to 10 m2, with 10-40 or more replications for each size. corner.
Most of that glass was new, however. The comparable data Fifteen percent of the failures started at the cut edges of the
base for weathered or aged glass is only one tenth as large, and panes indicating that edge quality is important, even for plates
sample sizes (replications) are usually smaller too. Because supported on all sides.
used glass has resale value, it is difficult to interest renovators
in donating windows for testing after taking the trouble to Testing small samples: ring-on-ring tests
remove them with the necessary care.
The characteristic strength used in the design does reflect The testing described so far using the dynamic glass testing
rig and the loading table is time consuming and expensive. A
the small data base on in-service glass. To take account of the
variability in test results, the Weibull probability distribution method of obtaining glass strength by testing small samples,
rather than the more familiar normal, or Gaussian, distribution cheaply, would clearly be welcomed. As the most severe
is fitted to the data. As noted before, the characteristic strength imperfections on a glass plate are at the cut edges, edge
" parameter, So, was set at 32.1 MPa based on the earlier tests strength could be assessed, to help determine thermal break-
of old glass, whereas for new glass, So is 40 MPa or more. age stresses (Pilette and Taylor 1988), by bending narrow
strips of glass (Walker and Muir 1984). Failures would cer-
h-service glass - Thompson Residence tainly begin at the edges. To assess the strength of a window
0 In 1986, IRC acquired 47 windows removed from the under lateral loading, however, requires that the failure start
University of Ottawa's Thompson Residence after 15 years of on the surface of the glass, not at an edge. Nevertheless, under
service. Nominally 4 mm thick, they ranged in area from 1.2 flexural loads, a certain percentage of breaks do start at the
to 1.9 m2 and in aspect ratio from 1 to about 1.5. They were edges in the corner regions as discussed in the previous section
loaded to failure, with the weathered (exterior) surfaces in ten- (Pig. 7).
sion, on a specially built loading table at a rate of 0.07 P a l s , The ring-on-ring test of small glass plates (180 mm square)
so that failure occurred in roughly 40- 100 s depending on the appears to provide an economical alternative to tests of full-
strength of the specimen. sizedpanes (Norville and Minor 1984, Simiu eral. 1984). The
Although these results were not available in time to be con- main benefits, however, appear to be for comparative testing.
sidered in setting So for the CGSB standard, they now provide For example, in the Ottawa area, aged glass is available from
an opportunity to check its calibration. Each failure pressure double glazed windows that have failed. Either the seals have
(after conversion to the equivalent 60-second pressure) was failed and the windows are cloudy or the windows have
divided by the expected design value according to the Stan- cracked owing to thermal stresses. If the latter is the case, the
dard. That design pressurc conforms to the traditionnl risk of cracked windows can only be evaluated by cutting them up
breakage of 8 per thousand. That is, of every 1000 specimens into small samples, as for the ring-on-ring tests, and loading
loaded to their design pressure, on average 8 are expected to these to failure. To assess the deterioration in strength of the
 CAN. I. C N . ENG. VOL. 17. 1990

I ALTERNATIVE RUBBER
LOAD,CELL LOADING RlNG

I
STEEL LOADING RlNG

O-RING

STEEL REACTION RlNG ' I

FIG.8. Rubber and steel loading rings (51 mm diameter) and 127 mm reaction ring. The loading rings are bolted to high-precision load cells.

The test is performed by placing a small glass plate on a cir-

cular reaction ring of 127 mm diameter and applying, on its
opposite surface, a load transmitted through a smaller, 51 mm
diameter loading ring until failure occurs (Fig. 8). As glass
fails in tension, the purpose is to produce with this test a
uniform tensile stress field on one face of the plate within
the smaller ring. Stresses outside the loading ring decrease
to almost zero at the cut edges so failures should not start
there (Fig. 9). The major disadvantage of ring-on-ring tests is
that only a relatively small area (within the smaller ring) is
tested. Larger rings can be used, however, for greater glass
-0.50 1 1 thickness.
0.0 20.0 40.0 60.0 80.0 ,. 100.0 Over 400 ring-on-ring tests have been conducted at IRC.
RADIUS, mm
Loads were increased linearly at 0.11 kN/s (25 lbls). Series of
FIG.9. Typical radial and tangential tensile stress ratios. Radius is tests were conducted comparing loading rings made of steel as
measured from centre of loading ring. Edge of loading ring is at shown in Fig. 8 with those where a 5.5 mm rubber O-ring
25.5 mm and reaction ring is at 63.5 mm. 51 mm in diameter applied the load to the glass. The aim was
to check whether the steel loading ring wascausing stress con- '
TABU 3. Ring-on-ring tests of in-service glass from a double-glazed centrations that reduced the failure stress and conseouentlv the
window area of uniform tensile stress field. A finite element analysis
of the sample under steel and rubber loading rings confirmed t
Outside pane (side 1) Inside pane (side 4) what strain gauging of a sample had shown. In 6 mm thick
glass, stress concentrations on the tension surface of the
No. tested 23 20 sample (6 mm) below both steel and mbber loading rings were
Successes 20 19 >
Success ratio (%) 87 only 1%-2% larger than tensile stresses within the 51 mm
95
Mean failure load (kg) 607.8 532.6 loading ring (Fig. 9). Thus data from samples where breaks
COV failure load (%) 19.5 28.8 started at any location under and within the 51 m m loading
Thickness (mm) 5.886 5.859 ring could be used and considered to be successes. The tests
Avg. time to fail (s) 53.8 47.1 also showed that the variability of results was less using the
COV time (%) 19.8 28.2 steel ring and the proportion of successes was higher. Thus the
rubber loading ring was abandoned.
Table 3 shows results of in-service glass of unknown age
aged glass, results from as many ring-on-ring tests as possible removed from NRC buildings. Two panes of glass from a
can be compared either with the equivalent tests of new glass double-glazed window are shown. The surfaces of the glass
or with the strength of the (unweathered) inside surfaces of are labelled I,?, 3 , 4 , from outside to inside the building. Sur-
double- and triple-glazed units. Eventually, in this way an esti- faces 2 and 3 face each other inside the window and were not
mate of the influence of in-service use will emerge. tested in this case. Table 3 illustrates that side 4 of the inside
 OALGLlESH AND TAYLOR

TABLE
4. Ring-on-ring tests of in-service glass from a triple-glazed window
Outside pane Middle pane Inside pane
Side 1 Side 2 Side 3 Side 4 Side 5 Side 6
No. tested 11 11 12 13 24 23
Successes 6 8 11 8 13 12
Success ratio (%) 54.5 72.7 92 61.5 54.2 52
Mean failure load (kg) 697.5 754.8 1051.3 777.9 751.5 650.9
COV failure load (%) 11.7 30.7 9.5 21.3 19.2 37.4
Thickness (mm) 6.17 6.173 6.290 6.290 6.257 6.260
Time to fail (s) 61.8 66.4 93.5 68.8 66.7 57.5
COV time (%) 12.0 30.5 9.3 22.0 19.3 37.6

>
pane of glass is weaker than the outside (side 1); abrasion due Hard body impacts due to windborne debris may result in
to cleaning side 4 may have weakened it. The variability of broken windows in tall buildings and so may earthquakes.
* ) samples from the inside pane is higher, indicating a more Palling glass breaks other windows below and causes
uneven distribution of microcracks. Natural weathering of hazardous conditions at street level. Research is required on
side 1 appears to he more uniform. As loading rates for all ways to keep broken windows from falling out. This may
tests were 0.11 kN/s (25 lbls), nu corrections to failure loads entail comparative testing of the behaviour of laminated glass,
were required. heat-strengthened glass, and glass with plastic surface fiims,
Similar tests were conducted on a triple-glazed window for which the dynamic glass testing rig is ideally suited.
removed from an NRC building. Samples of each side of each The strength of thin heat-strengthened laminated windows
pane were tested (Table 4). Again the weakest side, 6, was under lateral loads need study. Tests on the loading table will
inside the room, and side 1, outside. Samples from the pro- indicate how these windows compare in strength and stiffness
tected sides 2 to 5 were tested and were found to be signifi- with nonlaminated heat-strengthened windows of the same
cantly stronger. These results from two windows are given as total thickness.
examples. Clearly, results from many such windows are
required before conclusions can be drawn. Conelusions
Ring-on-ring tests must be done carefully, with the rings
centred precisely, or results will be quite variable. Their 1. Structural testing is a necessary part of dealing with
appeal, as noted previously, is in salvaging useful data on the glass. The engineering properties of glass and the behaviour
strength of in-service glass from broken or damaged windows. of windows, glass curtain walls, and spandrel glass have to he
However, when an adequate number of undamaged windows obtained by tests in the laboratory or in the field.
come available they should he tested, full-sized, on a loading 2 . The most important question to he answered by testing of
window glass is how much allowance should be made in
table.
- -
" for the deterioration in elass streneth with service. A
desien
---"------------
Research nperlr companion question is, should the glass in a building be
replaced at some stage?
The deterioration of window strength throughout its service 3. structural testing should be conducted in prescribed
life is not well understood and requires further research. The ways. Because glass is so variable, many replications of tests
in-service strength reflected by the CGSB standard is based On should be done (at least 20); the loading history must be care-
a limited number of tests of glass 8-25 Years old, from the fullv controlled and recorded: windows must be tested in a
0
mid-western United States. The effects of different exposures reaionablv clean environmentto avoid jab-induced scratches,
and the rate of deterioration with time are largely unknown. 4. The bynamic glass testing rig, the loading table, and ring-
At least 20 panes are normally tested to establish a mean on-ring test facility are all important research tools for evaluat-
breaking strength. What is more, of the total of about 500 tests ing the structural strength of window glass.
mentioned earlier some were tests of interior surfaces of 5. Glass is another material that can be understood and
sealed insulating glass units not exposed to abrasion or wet- "engineered." To assist the designer, a new CGSB standard
ness. The 279 panes on which the CGSB estimate is based has recently been published. With it the structural engineer is
yielded six separate evaluations, from only four different in a good position to expand his expertise from wood, steel,
building sites. concrete, and masonry, to include glass.
More studies of the edge quality of annealed window glass
are required. To get improved resistance to thermal breakage,
heat-strengthened glass is often selected. This is expensive. Acknowledgments
Development of ways of improving edge quality of annealed The authors would like to thank Preston and Lieff Glass of
windows would enhance their performance against thermal Ottawa who donated 47 windows removed from the Thomp-
breakage. son Residence at the University of Ottawa during a renovation
Further research on impact is required. Snow slides off program. They would also like to express appreciation for the
roofs onto skylights below and falling icicles ricochet off assistance of Pat Daly who conducted most of the testing and
obstacles and hit windows. Experimental research should be Claude Pilette, at the time a co-op student from the University
conducted on the impact resistance of various sizes of win- of Ottawa, who did the finite element analysis of the ring-on-
dows with a selection of gaskets and frames. ring tests.
 ABIASSI, J. J. 1981. The strength of weathered window glass using ORR,L. 1957. Engineering properties of glass. In Windows and glass
surfacecharacteristics. Institute for Disaster Research, Texas Tech in the exterior of buildings. Publication 478, Building Research
University, Lubbock, TX (NTIS Accession No. PB82-172735). Institute, National Academy of Sciences, National Research Coun-
ASTM. 1984. Structural performance of glass in exterior windows, cil, Washington, DC, pp. 51-62.
curtain walls, and doors under the influence of uniform static loads PILBTTE, C. F., andTnywn, D. A. 1988. Thermal stresses indouble-
by destructive method. ASTM E 997-84, Vol. 4.07, 1988 Annual glazed windows. Canadian Journal of Civil Engineering, lS(5):
Book of ASTM Standards, pp. 807-814. ..
-.
Xn7-1114
BEASON, W. L. 1980. A failure prediction model for window glass. PPG. 1979. PPG glass thickness recommendations to meet architects'
Institute for Disaster Research, Texas Tech University, Lubbock, specified l-minute wind load. Technical Services, Flat Glass Divi-
TX (NTIS Accession No. PB81-148421). sion, PPG Industries, Pinsburgh, PA.
BEASON, W. L., and MORGAN, J. R. 1984. Glass failure prediction SMIU,E., and HENDRICKSON, E. M. 1987. Design criteria for glass
model. ASCE Journal of Structural Engineering, 110(2): 197- cladding subjected to wind loads. ASCE Journal of Structural
212. Engineering, 113(3): 501-518.
BOWLES,R., and SUGARMAN, B. 1962. The strength and deflection SIMIU,E., and LECHNBR, J. A. 1987. Experimental glass failure data
characteristics of large rectangular glass panels under uniform and equivalent one-minute loads. ASCE Journal of Structural
pressure. Glass Technology, 3(5): 156-170. Engineering, 113(12): 2503-2508.
BROWN,W. G. 1969. A load duration theory for glass design. SIMIU,E., Reen, D. A,, YANCY,C. W., MARTIN,J. W., HENDRICK- ,i
Proceedings, Annual Meeting of the International Commission on SON,E. M., GONZALEZ, A. C., KOIKB,M., LECHNER, J. A,, and
Glass, Toronto, Ont., pp. 75-78; NRCC 12354, Division of BATTS,M. E. 1984. Ring-on-ring tests and load capacity of clad- ~ 1.
Building Research, National Research Council of Canada, Ottawa, ding glass. NBS Building Series 162, National Bureau of Stan-
Ont. dards, Gaithersburg, MD. f
-1974. A practical formulation for the strength of glass and S~URNARAS, C. J. 1979. Development and evaluation of a rig for the
its special application to large plates. NRCC 14372, Division of study of dynamic fatigue of flat glass. Final Report (67005), !I
Building Research, National Research Council of Canada, Ottawa, Ontario Research Foundation, Mississauga, Ont.
Ont. WALKER, G. R., and MUIR,L. H. 1984. An investigation of the bend-
CGSB. 1989. Structural design of glass for buildings. CANICGSB- ing strength of glass louvre blades. Proceedings, Ninth Australian
12.20-M89, Canadian General Standards Board, Ottawa, Ont. Conference on the Mechanics of Structures and Materials, Univer-
DAWLIESH,W. A. 1980. Assessment of wind loads for glazing sity of Sydney, Sydney, Australia, pp. 122-131.
design. Proceedings, IAHRIIUTAM Symposium on Practical %BULL, W. 1939. A statistical theory of the strength of materials.
Experiences with Flow-Induced Vibrations (September 3-6, Royal Swedish Institute for Engineering Research, No. 151, Stock-
1979). Edited by E. Naudascher and D. Rockwell. Springer- holm, Sweden.
Verlag, Berlin, Heidelberg. pp. 696-708.
HERSHEY,R. L., and HIGGINS,T. H. 1973. Statistical prediction List of symbols
model for glass breakage from nominal sonic boom loads. Report
No. FAA-RD-73-79, Booz-Men Applied Research, Inc., A area of glass plate
Bethesda, MD (NTIS Accession No. AD-763-594). A, reference area of glass plate (= 1.0 m2)
ISHIZAKI, H., MIYOSHI,S., MIURA,T., IDA,M., KASAOI, S., and B function which reflects the risk of failure of a glass
~RAYAMA, T. 1974. Fatigue effect on wind resistance of glass plate
panels. Tenth International Congress on Glass, July 1974, The C1 a constant
Ceramic Society of Japan, Kyoto, Japan, pp. 10-50 to 10-55. D,,,, measure of critical resistance to failure of a surface
J~HAR S., 1981. Dynamic fatigue of flat glass. Phase 11, Final Report flaw exposed to tensile stress and water vapour
(670391, Ontario Research Foundation, Mississauga, Ont. E modulus of elasticity of glass
--- 1982. Dynamic fatigue of flat glass. Phase El,Final Report h thickness of glass plate
(67049), Ontario Research Foundation, Mississauga, Ont. .
JONSSON, B. 1977. Glas i byggnader. Glassets tjocklek. Rapport R67: K a constant
1977 Statens rad for hyggnadsforskning, Stockholm, Sweden (in m shape parameter of two-parameter Weibull distribution
Swedish).
KANABOLO, D. C., and NORVILLE, H. S. 1985. The strength of new
n
p, P latkral pressure on glass plate
-
exoonent. describes sloue of loading rate eauation

window glass plates using surface characteristics. Glass Research Pf constant pressure causing failure after tf seconds A'

and Testing Laboratory, Texas Tech University, Lubbock, TX. pm constant pressure causing failure after 60 s , also called
KRALL, W. R., SISKOS, W. R., STEWART, R. A,, and SPINDLER, R. G. "equivalent 60-second pressure"
1981. The behavior of float glass under uniform wind loading. POF cumulative probability of failure I
Preprints of 4th U.S. National Conference on Wind Engineering r aspect ratio of glass plate (long-to-short dimension)
Research, July 27-30, Seattle, WA, pp. 155-159.
MICHALSKE, T. A., and BUNKER, B. C. 1987. The fracturing of glass. R risk factor in Beason and Morgan (1984), measures
Scientific American, December, Dalton, MA, pp. 122- 129. contribution of flaws over the whole surface of a plate
Moone, D. M. 1980. Proposed method for determining the glass under pressure to the failure load
thickness of rectangular glass solar collector panels subjected to s m s , r risk function, measures contribution of flaws over the
uniform normal pressure loads. JPL Publication 80-34, Jet Propul- whole surface of a plate under pressure to the failure
sion Laboratoly, Pasadena, CA. load
NBC. 1990. National building code of Canada 1990. NRCC 30619, S(f) nominal tensile stress at surface flaw at time t
Associate Committee on the National Building Code, National constant stress causing failure in 6 0 s
sm
Research Council of Canada, Ottawa, Ont. mode of two-parameter Weibull distribution
$0
NORVILLB, H. S., and MINOR,J. E. 1984. Glass strength evaluation nominal tensile stress at failure
using ring-on-ring tests. Proceedings, American Society of Civil Sf
Engineers Specialty Conference on Engineering Mechanics in tf time to failure
Civil Engineering, University of Wyoming, Laramie, WY, pp. 0 surface stresses at failure locations under pressures P
- ...
1437-144(1.
- .. P exponent which relates stress and pressures at breakage
-1985. Strength of weathered window glass. American Cer-
amic Society Bulletin, 64(11): 1467-1470.

View publication stats