Fatigue & Fracture of Engineering Materials & Structures

doi: 10.1111/j.1460-2695.2008.01308.x

The strength of annealed, heat-strengthened and fully tempered oat glass

F. A. VEER, P. C. LOUTER and F. P. BOS
Faculty of architecture, Delft University of Technology, Berlageweg 12628 CR Delft, the Netherlands Received in final form 21 November 2008

A B S T R A C T In the last two decades architectural glass has made an enormous leap from a secondary

material to a material that combines structural and cladding roles. The structural role is a new and problematic one. In contrast to most other engineering materials the strength of glass is not a material parameter but a parameter dependent on processing quality and damage to the glass surface. There is also no real agreement on how strong glass is. There is a concept Euronorm for structural glass that has values for the characteristic strength for annealed, heat-strengthened, fully tempered and chemically toughened glass. There is however no real agreement on the validity of these values for design of glass beams or columns. To provide an independent set of values a statistically significant series of four-point bending tests on glass were conducted using both lying and standing positions resulting in a set of values for the characteristic strength. Keywords strength; structural design; testing, glass.

INTRODUCTION

In the last two decades architectural glass has made an enormous leap from a secondary material to a material that combines structural and cladding roles. The structural role is a new and problematic one. In contrast to most other engineering materials the strength of glass is not a material parameter but a parameter dependent on processing quality and damage to the glass surface. There is also no real agreement on how strong glass is. The concept Euronorm for structural glass has some values for the characteristic strength for annealed, heatstrengthened, fully tempered and chemically toughened glass. These values have been adopted by national norms such as the Dutch NEN 2608.1 There is considerable discussion in the literature whether the approach used to obtain the Euronorm values is valid. The ring on ring method used here stresses the glass surface that is of much higher quality than the edges.2 Alternative methods for ring on ring testing are tests where the glass is contained in a vessel that is pressurized,3 or conventional bending tests.4,5 Bending tests have the advantage that they correspond closely to the tests conducted on other materials, are easily done, comparatively cheap and stress the edges of the

glass. There are however significantly different results if the specimens are tested standing or lying while the aspect ratio of the specimen also has an effect.6,7 One problem in the literature is that most data sets are too small or it is not certain that the data sets can be directly compared. To allow accurate comparisons glass specimens of 1000 100 10 mm were prepared from a single jumbo sheet of glass and cut and ground identically on a single processing line. Of these specimens one-third was fully tempered, one-third heat strengthened and onethird was in the original, annealed state. These specimens were tested in four-point bending. Of each group of specimen half were tested standing and half lying, resulting in six groups of data.
EXPERIMENTAL METHOD

Correspondence: F. A. Veer. E-mail: f.a.veer@tudelft.nl

Glass beams of size 1000 mm long and 100 mm wide were cut from a single glass plate with a thickness of 10 mm. These were cut on professional automated cutting machines and finished by automated grinding and polishing. One-third of the specimens were pre-stressed using full thermal tempering, and one-third of the specimens were pre-stressed using heat strengthening. In both cases the same automated tempering oven was used using the manufacturers software that controls the speed and duration with which the specimens enter the oven, remain there and the cooling rate after leaving the oven.

18

c 2009 The Authors. Journal compilation c 2009 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 32, 1825

STRENGTH OF ANNEALED, HEAT-STRENGTHENED AND FULLY TEMPERED FLOAT GLASS

19

Table 1 Four-point bending failure stress of all annealed specimens in MPa Test number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Annealed lying 25.8 28.3 30.7 31.2 32.1 36.1 36.6 37.7 38.4 38.8 39.9 40.9 41.4 41.7 42.6 Annealed standing 21.2 21.4 23.2 23.7 23.9 24.1 24.2 24.6 24.7 24.9 24.9 25.5 25.6 25.6 26.2 Test number 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Annealed lying 46.7 47.3 47.7 47.9 53.8 54.3 54.5 55.8 58.6 Annealed standing 26.7 27.4 27.8 31.6 32.7 33.5 38.2 39.1 22.6 23.0 37.8 21.9 28.4 42.9 27.3

Fig. 1 Test set-up with glass specimens lying.

All specimens were wrapped in PET foil for safety. For the heat-strengthened and fully tempered glass multiple layers PET foil were necessary. For annealed float glass a single layer of foil was sufficient. The beams were tested in four-point bending on a Zwick Z 100 universal testing machine with the specimen lying or standing. To avoid buckling of the standing specimens, these specimens were supported on the sides at five points along the length. One millimetre thick Teflon sheet was used as an intermediary between the metal supports and the glass to avoid inducing high contact stresses. The test rig is shown in Fig. 1. A displacement rate of 1 mm/minute was used for all tests. For a number of specimens in each series the pre-stress was measured using a scalp laser scanning device that measures the pre-stress level through the thickness. The prestress level was measured at the centre of the specimen where the mechanical test was conducted. All specimens for which the pre-stress was determined were tested standing. It is assumed that the average pre-stress levels were representative for all specimens.
RESULTS

Table 2 Four-point bending failure stress of all heat-strengthened specimens in MPa Test number Heat strengthened lying 58.8 65.0 71.6 74.5 81.4 81.6 84.2 85.9 90.7 95.7 99.6 99.9 103.1 104.1 106.3 Heat strengthened standing 54.9 56.1 57.1 57.3 57.7 59.3 59.6 64.3 66.2 68.9 69.5 70.1 72.9 74.0 74.6 Test number Heat strengthened lying 111.7 119.7 125.6 128.1 133.4 149.1 154.6 167.0 Heat strengthened standing 75.7 76.0 78.2 79.6 82.2 95.8 74.8 83.8 80.0 63.5 85.9 57.4 79.6 91.8

The results for the annealed specimens are given in Table 1. Table 2 gives the results for the heatstrengthened specimens. Table 3 gives the results for the fully tempered specimens. Annealed glass A Weibull plot of the results for annealed glass is given in Fig. 2. The values for the lying glass are given as diamonds and for the standing glass as plusses. The specimen failed by simple cracking as shown in Fig. 3. Higher failure stresses result in more cracks and more crack branching. Although the results of the lying tests suggest a reasonable Weibull fit, Fig. 2 suggests that both data sets are bilinear.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

16 17 18 19 20 21 22 23 24 25 26 27 28 29

The lines in Fig. 2 are Weibull fits to partial data sets, the exact slope depends on which point the authors have chosen to use. Heat-strengthened glass A Weibull plot of the results for heat-strengthened glass is given in Fig. 4. The values for the lying glass are given as diamonds and for the standing glass as plusses. The specimen failed by simple cracking from the edges, followed by secondary cracking resulting from the release of

c 2009 The Authors. Journal compilation c 2009 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 32, 1825

20

F . A . V E E R et al.

Table 3 Four-point bending failure stress of all fully tempered specimens in MPa Test number Fully tempered lying 96.1 107.7 120.2 130.9 133.6 135.3 139.2 145.5 146.8 147.1 147.2 147.3 147.7 153.0 156.2 166.4 Fully tempered standing 72.6 74.5 76.7 80.5 85.1 88.4 89.1 89.8 91.0 92.2 92.5 95.0 96.2 96.7 97.3 99.7 Test number Fully tempered lying 172.3 182.8 184.1 186.7 191.4 191.5 197.0 205.0 205.1 Fully tempered standing 100.2 102.8 103.4 103.5 106.6 107.4 109.4 122.1 106.3 111.7 127.8 93.7 125.1 110.5 88.6 99.6

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

Fig. 2 Weibull plot for annealed glass tests.

the pre-stress, as shown in Fig. 4. The heat-strengthened glass consistently fragmented into fragments of about 2 to 3 cm in length on average. This suggests a rather high pre-stress level for heat-strengthened glass. Specimens that fail at higher bending stresses do not fragment into smaller pieces. Figure 4 suggests that both data sets are essentially bilinear in their Weibull behaviour. The lines in Fig. 4 have been made in the same way as in Fig. 2. Fully tempered glass A Weibull plot of the results for fully tempered glass is given in Fig. 6. The values for the lying glass are given as diamonds and for the standing glass as plusses. The specimen failed by simple cracking from the edges, followed by massive cracking resulting from the release of the preTable 4 Statistical summary of results in Table 1, 2 and 3 Temper Orientation Average failure stress (MPa) Standard deviation (% of average) Average pre-stress (MPa) 4.5 4.5 64.3 64.3 100.6 100.6 Standard deviation of pre-stress (% of average) 37% 37% 3.8% 3.8% 12.9% 12.9% Fig. 3 Failure pattern of annealed glass tested lying, failing at 55.8 MPa.

stress, as shown in Fig. 7. The fully tempered glass consistently fragmented into fragments of less than 1 cm in length on average. The specimens that fail at lower bending stresses do not fragment into larger pieces. In contrast

Annealed Annealed Heat strengthened Heat strengthened Fully tempered Fully tempered

Lying Standing Lying Standing Lying Standing

42.0 27.5 104.0 71.3 157.4 98.0

21.8% 20.1% 27.7% 15.8% 18.9% 13.7%

c 2009 The Authors. Journal compilation c 2009 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 32, 1825

STRENGTH OF ANNEALED, HEAT-STRENGTHENED AND FULLY TEMPERED FLOAT GLASS

21

Fig. 4 Weibull plot for heat-strengthened glass tests.

Fig. 7 Failure pattern of fully tempered glass tested lying, failing at 107.7 MPa.

to the results for annealed and heat-strengthened glass the Weibull behaviour for the standing glass is reasonably linear, although there is a small hump in the Weibull plot of the lying glass. This is caused by a clustering of data at a certain stress level. The lines in Fig. 6 have been made in the same way as in Fig. 2 except that the full data set was used for the fully tempered lying glass. Failure stress in relation to pre-stress For a limited number of specimens the pre-stress through the thickness was measured using a scalp laser pre-stress measuring device. Measurements of the centre section were made. And the average pre-stress on the surfaces was calculated. These specimens were tested standing. The measured failure stresses and measured pre-stress levels are given in Table 5. A plot of failure stress against pre-stress is given in Fig. 8. The pre-stress measurements are quite reproducible. The general accuracy is about 5 MPa. It should be noted that the fragmentation pattern in these tests was fully consistent with that observed before and illustrated in Figs. 3, 5 and 7. The relatively high standard deviation in the measured pre-stresses annealed float glass should be looked at in terms of the experimental accuracy of the measurement, which is in the same range as the average measured value. Statistical description Although the Weibull approach is the most common statistical descriptor for failure strength in glass, the test results have shown that only for the fully tempered glass this gives a reasonable fit. Using the Matlab software package a large number of other statistical descriptors have been

Fig. 5 Failure pattern of heat-strengthened glass tested lying, failing at 111.7 MPa.

Fig. 6 Weibull plot for fully tempered glass tests.

c 2009 The Authors. Journal compilation c 2009 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 32, 1825

22

F . A . V E E R et al.

Table 5 Measured pre-stress and failing stress for standing specimens Number Annealed failure stress (MPa) 22.6 23.0 37.8 21.9 28.4 42.9 27.3 29.1 28.1% Pre-stress (MPa) 6 5.4 4.0 2.5 2.5 4.3 6.8 4.5 37.0% HS failure stress (MPa) 74.8 83.8 80.0 63.5 85.9 57.4 79.6 91.8 77.1 15.0% Pre-stress (MPa) 59.7 62.9 65.9 65.1 67.6 65.4 62.9 65.2 64.3 3.8% FT failure stress (MPa) 106.3 111.7 127.8 93.7 125.1 110.5 88.6 99.6 107.9 12.9% Pre-stress (MPa) 94.6 109.9 101.4 97.8 104.9 97.6 102.4 96.1 100.6 5.1%

1 2 3 4 5 6 7 8 Average Std/average

Table 6 Summary of results Glass condition Average failure strength 42.0 27.5 104.0 71.3 157.4 98.0 Minimum test value (MPa) 25.8 21.2 58.8 54.9 96.1 72.6 Maximum test test value (MPa) 58.6 42.9 167.0 95.8 205.1 122.1 Calculated Weibull strength (MPa) 24 20 54 52 88 77

Annealed lying Annealed standing Heat strengthened lying Heat strengthened standing Fully Tempered lying Fully tempered standing

results imply that only a complex multiparameter model can describe all these data in a single model. There is, however, no theoretical framework for such a model.
DISCUSSION

Fig. 8 Relation between pre-stress and failure stress.

tried using the same data. Figure 9 shows a number of probability plots for annealed glass and Fig. 10 for fully tempered glass, both tested standing. In the case of annealed glass no single-parameter statistical descriptor was found that describes the data. For fully tempered glass, the Weibull, normal distribution and lognormal distribution are all reasonable to good statistical descriptors. These

From an engineering point of view the ideal result would be a simple formula to calculate the strength of glass for any condition. The results make it clear that no simple solution exists. The results show clearly that the strength of glass is not only dependent on the orientation of the glass versus the load, other criteria such as aspect ratio also play a role. For the specimens used, the strength standing is about 67% of the strength lying. Even a large number of identical specimens tested identically do not give a unique result. In Fig. 11 all data are Weibull plotted together and the different behaviour of the six data sets is clearly visible. The results show considerable spread. This spread cannot be described easily by normal statistics. The commonly used Weibull approach is valid for fully tempered glass, but other statistical descriptors such as the lognormal distribution fit as well or better. The Weibull function has the advantage of a failure-based physical model behind it. As the Weibull function cannot describe the behaviour of annealed or heat-strengthened glass

c 2009 The Authors. Journal compilation c 2009 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 32, 1825

STRENGTH OF ANNEALED, HEAT-STRENGTHENED AND FULLY TEMPERED FLOAT GLASS

23

Fig. 9 Different probability distributions for annealed glass tested standing.

adequately the validity of the underlying model for glass has to be questioned. Assuming that the lower strength values can be used to provide a safe Weibull modulus, a 1 in 1000 failure probability can be used to calculate the Weibull strength values as given in Table 6. These should not be used as design values, as standing specimen with higher aspect ratios than those used will result in a lower Weibull strength. The given values are only for allowing comparison between the six sets. More interesting is the underlying problem of the material mechanics involved. The results indicate that a lot of our understanding about the failure of glass is incorrect. The consistent bilinearity of the Weibull plots implies that there is no single source of failure. A single source of failure, such as voids expanding into cracks as found in certain ceramics,8 should result in a single Weibull line. Bi-linearity implies that there are two data sets mixed up together. This conclusion is supported by the work of Whittle et al.,9 which shows different Weibull moduli for different flaw types in glass. More important is the non-correlation between failure strength and fragment size of heat-strengthened and fully tempered glass. Half of the heat-strengthened glass spec-

imens are stronger than the weakest fully tempered glass specimen. This half, however, consistently gives larger fragments of consistent size. The low strength fully tempered specimens fragment consistently to small fragment of consistent size. Fragment size is thus a good indicator of the pre-stress that has been introduced, but no indicator of strength. The pre-stress level itself is thus also no indicator or guarantee of failure strength. As the pre-stress level is consistent within a series the large variation in strength implies that failure cannot be initiated by the simple concept of (defect size) (applied stress + pre-stress) gives a stress intensity that is greater than the critical one. The commonly accepted principle that failure stress = applied stress pre-stress is also shown to be false. As a general statistical result this might approximately hold for glass tested lying, as the averaged results in Table 4 show. This is clearly not true for the results of the standing experiments. In several individual cases the failure stress is smaller than the measured pre-stress. The lying annealed glass has a failure stress ranging from 25 to 60 MPa. The lying heat-strengthened glass has a failure stress ranging from 60 to 170 MPa. The variations in failure stress are about three to four times

c 2009 The Authors. Journal compilation c 2009 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 32, 1825

24

F . A . V E E R et al.

Fig. 10 Different probability distributions for fully tempered glass tested standing.

haviour/flaw type/flaw size distribution than by the prestress level.

CONCLUSIONS

From the results it is concluded:

Fig. 11 Weibull plot of all test results. From left to right at 0.96 probability, annealed standing, annealed lying, heat strengthened standing, heat strengthened lying, fully tempered standing and fully tempered lying.

the variations in pre-stress levels for heat-strengthened and fully tempered glass. This implies that the failure behaviour of heat-strengthened and fully tempered glass is still controlled more by the combination of material be-

- The Weibull descriptor is only valid in certain cases as descriptor for the failure strength of glass. Other statistical descriptors can be equally valid in these cases. - It is doubtful if a certain guaranteed minimum strength can be defined. The strength is however dependent on edge quality, orientation of glass relative to load, aspect ratio and pre-stress level. - Fragment size after failure is a good indicator of the prestress level. However it is no indication of the strength of the glass. - The physical process of the failure of pre-stressed glass is not controlled by the combination of defect size and the sum of the applied stress and (negative) pre-stress. - The failure strength is not the sum of average intrinsic strength plus pre-stress level. In certain cases the glass fails at an applied load that is smaller than the pre-stress level.

c 2009 The Authors. Journal compilation c 2009 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 32, 1825

STRENGTH OF ANNEALED, HEAT-STRENGTHENED AND FULLY TEMPERED FLOAT GLASS

25

- The measured pre-stress levels are not very consistent implying that the production process is less controlled than commonly assumed. REFERENCES
1 Nederlands Normalisatie Instituut, NEN 2608:1997/C1:2007 nl, Vlakglas voor gebouwen Weerstand tegen windbelasting Eisen en bepalingsmethode. Pellice, S., Gilabert, U., Solier, C., Castro, Y. and Duran. (2004). Mechanical behaviour of glass reinforced with SiO 2 hybrid sol-gel coatings. J. Non-Cryst. Solids 348, 172. Kondratieva, N. and Zubkov, V. (2006) Experimental and theoretical study of flat glass strength at cross bending. In: Proceedings Glass Processing Days. China, Peking 2006 as published on www.glassfiles.com.

7 8 9

Froli, M. and Lani, L., Probabilistic assessment of the bending strength of chemically and thermally tempered glass specimens. Available from www.glassfiles.com. Rodichev, Y., Maslov, V., Netychuk, A., Bodunov, V. and Yevplov, Y. (2007) Bending strength and fracture of glass materials under the different loading conditions. In: Proceeding Glass Performance Days Tampere 2007. Available from www. glassfiles.com. Veer, F. A., Bos, F. P., Zuidema, J. and Romein, T. (2005) Strength and fracture behaviour of annealed and tempered float glass. In: Proceedings 11th International Conference on Fracture, Turin, Italy). Veer, F. A. Heron. (2007) The strength of glass, a non transparent value. Heron 52(1), 87104. Kendall, K. et al. (1986) Influence of toughness on Weibull modulus of ceramic bending strength. J. Mat. Res. 1, 120. Whittle, B. R. et al. (2002) A water based coating for the strength of glass. Phys. Chem. Glasses 43c, 207212.

c 2009 The Authors. Journal compilation c 2009 Blackwell Publishing Ltd. Fatigue Fract Engng Mater Struct 32, 1825