You are on page 1of 3

The Strength and Failure of Glass in Bending

Fred Veer*, J. Zuidema**, F.P.Bos*


* Building Technology group, faculty of Architecture, Delft University of
Technology, www.zappi.bk.tudelft.nl
** laboratory of materials science, Delft University of Technology

Keywords

1=Strength 2=bending 3=tempered glass 4=failure patterns

Abstract Figure1
Experimental setup
The strength of glass as used in
architectural engineering is one of those
things about which too little is known.
To determine what would be a safe
design stress for glass in a completely
glass structure, tests have been done on
panels of annealed oat glass and fully
tempered glass using different sizes and
both lying and standing positions. The
results demonstrate that the strength
of both annealed and tempered glass
is dependent on the orientation of
loading, the commonly held strength
values for tempered glass are a gross
over estimate for beams under bending
loading, the strength data cannot be

Building Projects
described by Weibull statistics.

Case Studies
Introduction Specimen size Tempered position Failure stress* s.d. (%)
Glass is a material that is used in all (MPa)
branches of engineering because it 60050*10 mm No Standing 52.4 16.1
is the only rigid transparent material 60050*10 mm No Lying 71.4 19.6
available. The last decades have seen 100012510 No Standing 43.7 11.8
increasing use of glass in automobiles mm
100012510 Yes Standing 140.1 14.8
and buildings. The tendency in modern
mm
buildings to increase the amount of
100012510 No Lying 71.0 20.1
glass used has resulted in the desire to mm
use the glass in a load bearing manner. 100012510 Yes Lying 154.5 20.6
Glass however is a brittle material that mm
fails unpredictably in tension. Although 100025010 No Standing 39.9 24.3
the strength of glass can be described mm
in certain cases by Weibull statistics 100025010 Yes Standing 87.6 18.3
and probabilistic strengths calculated, mm
as shown earlier by Veer and Zuidema 100025010 No Lying 51.1 35.3
[1,2], considerable uncertainty exists mm
in the literature about the exact 100025010 Yes Lying 101.2 22.3
parameters and the allowable design mm
strength. For this reason 4 point * average of 16 tests
bending tests have been conducted on Table 1
standing glass panels of 1000250 mm test results for all specimens
and 1000125 mm using annealed oat
glass and fully tempered oat glass.

Experimental method
Glass beams of size 1000 mm long and layer of foil was sufcient, for tempered an intermediary between the metal
125 or 250 mm wide were cut from a oat glass three layers of foil were supports and the glass to avoid inducing
single glass plate with a thickness of 10 needed. high contact pressures. The test rig is
mm. Specimens of 600 mm long and 50 The beams were tested in 4 point shown in gure 1.
mm were cut later to provide additional bending on a Zwick Z 100 universal
data. These were professionally cut testing machine with the specimen Results
on professional cutting machines and standing or lying. The cut and processed
The results of the tests are summarized
nished by grinding and polishing. Half edges were thus directly stressed in the
in table 1.
of the specimens were pre-stressed standing specimens. To avoid buckling
using full thermal tempering. All the specimen was supported on the
specimens were wrapped in pet foil for sides at 5 points along the length. 2
safety. For annealed oat glass a single mm thick sheet nylon was used as

GLASS PROCESSING DAYS 2005 - www.gpd.fi 1


Discussion Figure 2
Weibull plot for annealed
The data for the 100012510 mm oat glass 1000125 mm
specimens which were tested in a
standing position is plotted in Weibull
diagrams and normal probability plots in
gures 2 and 3.
Although earlier research by Veer
and Zuidema suggested that the failure
behaviour of glass could be described by
Weibull statistics these results suggest
that this conclusion is not valid in all
cases. In the annealed oat glass there
is a clear deviation at the lower and
higher strength levels while the data for
the fully tempered glass does not even
approach a weibull or normal probability
distribution. Of the specimens tested
only the 250 mm high annealed oat
glass specimens approach a reasonable
Weibull distribution. In practice most Figure 3
of the data sets cannot be described by
Weibull plot for fully
a single statistical function. In addition tempered oat glass
there is a clear effect of size and 1000125 mm
specimen orientation. The higher the
specimens the lower the strength as is
shown in gure 4. Standing specimens
are also signicantly weaker than lying
specimens as is seen in gure 5.
Considering the spread in results this
implies design strengths of 25 MPa and
80 MPa for the 125 mm high annealed
respectively fully tempered oat glass.
For the 250 mm high specimens this
would be 16 and 50. These values are
Building Projects

signicantly lower than those assumed


Case Studies

in the relevant design standards (3). The


values in these design standards are
based on experiments with the glass
lying. It has been noted by Hess, (5) that
Figure 4
glass tested standing has a strength
about 40% less than that tested lying. effect of standing speci-
men height on average
These result suggest that the fall in failure stress
strength can be signicantly greater
than 40% and is also dependent on
specimen size.
No clear direct reason for the failure
of the Weibull distribution is evident.
The most logical and simple explanation
is that there are several types of failure
mechanism, possibly associated with
different types of defects causing the
initial failure. The specimens that fail at
lower strength values having some rare
type of defect. Fractographic analysis so
far has not provided any evidence for
this theory. This theory would however
explain some of the anomalous data
found earlier by Veer and Zuidema (1,2)
Figure 5
and which is also shown in the design
standards (4). The small size of critical effect of orientation on
failure strength for an-
defects in glass and the transparent nealed oatglass
nature of glass make it difcult to study
the edges and fracture surfaces.
The cracks generally show up in a
fan shaped form emanating from the
point in the tensile zone where fracture
started. This fan shaped pattern is
visible in both annealed and tempered
oat glass, as can be seen in gures 6
and 7. The presence of the fan shaped
pattern in tempered oat glass is
actually contrary to what is commonly
stated in the literature, (3). Tempered
glass is supposed to disintegrate
immediately on overloading. In the

2 GLASS PROCESSING DAYS 2005 - www.gpd.fi


lying condition which is normally used Figure 6
for testing no special crack pattern crack pattern in weakest
can be distinguished. In the standing and strongest 125 mm
high annealed oat glass
position used in these tests the fan specimens
shaped crack pattern implies that failure
in tempered glass is caused by the
overloading of edge defects similar to
those in annealed glass and that the Figure 7
cracks causing this failure run at such crack pattern in weakest
a high speed that the specimen fails and strongest 125 mm
before it disintegrates due to the release high tempered oat glass
specimens
of the pre-stress energy. The unstable
cracks from the overloading must thus
be signicantly faster than the cracks
caused by the release of the pre-stress.
In the literature, (3,6), a relationship Figure 8: cracks
branching from point
is usually assumed between the number of failure for 125 mm
of cracks in glass and the stress at high glass
failure. In gure 7 the number of cracks
in the fan shaped area of the annealed
and tempered oat glass specimens has
been plotted against the failure strength
for the 125 mm high specimens. The
data for the 125 mm high specimens
approaches a single line. The data for
the 250 mm high specimens is less clear.

Conclusions
From the results the following is
concluded :

The failure strength of most of the


test results does not t the Weibull
distribution function.

Building Projects
The deviation from the Weibull There is some correlation between
distribution is especially signicant [5] Hess, R.

Case Studies
the number of cracks that develop Glastrager, Bericht 20 des instituts for hochbau
at the bottom of the failure strength from the point of failure and the techniek, ETH Zurich , 2000
distribution. failure strength. This correlation [6] Janssen M., Zuidema J., Wanhill R.
The results suggest that this deviation works best in specimens with a low Fracture mechanics, 2nd edition, 2002
can only be explained by the proportion of height to length.
presence of multiple failure Acknowledgements
mechanisms with different probability References The material assistance of the van
distributions. Noordenne glass groep and SIKO bv in
The design strengths of glass beam [1] Veer F.A., Zuidema J, van Kranenburg C
The strength of glass providing us with glass and adhesives
specimens tested standing is Proceedings 2002 European Conference on to conduct our research is gratefully
signicantly lower than that reported Fracture (ECF14), Krakau acknowledged.
in the literature. [2] Veer F.A. , Zuidema, J
The experimental work for this
The failure of tempered glass The strength of glass, effect of edge quality
Glass processing days / educational glass paper by S. Broersma, E. Brogt. J.
specimens tested standing is by crack conference. (pp. 106-109). Tampere, Finland: Kaars Sijpesteijn, S. de Richemont and
growth from an initial defect. Glass processing days / tamglass ltd.Oy. D.Vosmaer is acknowledged.
Only after the specimen has failed by [3] Gulati S.T., course notes, design principles for
overloading does cracking due to the glass products , 5th glass processing days ,
Tampere 1997
release of the pre-stressing energy [4] Draft European standard for mechanical design
take place. of glass, EN 1288-2

GLASS PROCESSING DAYS 2005 - www.gpd.fi 3

You might also like