Sei sulla pagina 1di 8

International Journal of Applied Glass Science 2 [1] 3946 (2011) DOI:10.1111/j.2041-1294.2011.00043.

Edge Strength Testing of Thin Glasses


Suresh T. Gulati* and John D. Helnstine
Corning Incorporated, SP-FR-06 Corning, New York 14831

Following a brief history of various tests for measuring edge strength, we present the theory and application of four-point vertical bend test (VBT). The latter is similar to a standard four-point bend test except that the specimen is turned on its edge so as to permit application of uniform tension to a majority of the edge aws. The article cautions against lateral buckling of the VBT specimen, resulting in failure initiation from the compression edge, if its dimensions are not optimized. We provide guidelines for designing the width and thickness of the VBT specimen relative to its length, which obviates the onset of lateral buckling. Three simple inspections to verify the validity of VBT include the shape of load versus deection curve, location of failure origin, and fracture pattern. Two examples, one involving edge quality of an aluminosilicate glass and the other involving factors that affect edge strength of automotive glass, help illustrate the application of VBT.

Introduction The study of the strength of glass has been a fascinating eld for generationsprobably because the rst piece was dropped and broken. Usually the study of glass strength involves a relatively large surface area and precludes edges. The ASTM standard C-1581 for exural strength (MOR) of glass is designed to minimize breaks that originate at the edge. Likewise biaxial strength measurements via ASTM F-394,2 or C-1499,3 to name a couple, are even more so designed to minimize edge origins. However, sometimes one wishes to study the strength of a single corner or edge of a piece of glass rather than its face. In a study of scoring of automobile windshield glass, Cleary and Gulati4 used the standard
*gulatist@corning.com r 2011 The American Ceramic Society and Wiley Periodicals, Inc.

four-point bend test with both the scored edges either on the tension side or on the compression side of the test specimen thereby focusing on a particular edge, or corner nish condition. Pantelides et al.5 also wished to test only one edge. As illustrated in their Fig. 5, they developed a exural test that put the rectangular cross-sectioned bar at an angle, along one of the diagonals, which put one corner at the maximum tension position of a exural shape. Then there are instances where the entire edge, for example the thin face, should be studied, not just one corner. For example, consider the thin face of a sheet of a 0.7-mm-thick LCD glass. That thin face could be considered as all edge that was processed in a simple grinding process. And, unfortunately, one cannot consistently determine which grinding process is less damaging by visual inspection, whether the damage is on the edge or near the edge on the face. That question can only be answered by a strength test that stresses the

40

International Journal of Applied Glass ScienceGulati and Helnstine

Vol. 2, No. 1, 2011

entire edge, and the near edge, more than the large face of the sheet. One method of measuring the strength of this whole edge is to cut the plate so that the thin face becomes the primary face subjected to uniform tension. Stahn6 studied the strength effects of different scoring and cutting methods of 10-mm-thick at glass by preparing a small 5 mm high bar from the edge of the original plate. This small scale MOR specimen could then be tested and analyzed by the standard process. Another method of testing the strength of such prepared edges is by stressing them thermally as was done by Inoue et al.7 and Gaume and Gy.8 This paper describes an edge strength test, which is essentially the standard four-point exure test with the specimen turned on its side. First, we will discuss the theoretical aspects of test and then focus on some of the practical issues in performing the test. It has been discovered that a similar test had also been used by Kerper et al.9 of NBS (now NIST) as early as 1962; see Fig. 2B of their paper for an illustration of the 15 in. 1 in. 1/2 in. 1/4 in. specimen. The paper will conclude with test results for several industrial applications. The Edge Test The edge test, as noted earlier, is a standard exure test with the specimen turned on its sideit could be called a vertical MOR cf. the standard horizontal MOR. The left side of Fig. 1 illustrates the normal fourpoint MOR conguration with largest faces of the specimen contacting the loading xture. The edge test conguration is illustrated on the right side of Fig. 1 where the specimen edge contacts the loading xture that is one of the narrowest specimen faces. At rst glance, the change in orientation of the specimen with respect to load and support xtures

should not change the strength calculation other than the height is now larger than the width. Dimensional nomenclature must be referenced to the orientation of the specimen, not to relative magnitudes of these dimensions. If the usual beam assumptions, that is small deection, plane sections remain plane, etc., still hold, the formula for calculating edge strength is1: S 3La bd 2

Fig. 1. The classic, standard MOR specimen conguration is shown on the left hand side with specimen face in contact with the loading xture. In the edge conguration shown on right, the loading xture contacts the edge.

where S is the edge strength, MPa (psi), L is the failure load, N (lbf), a is the moment arm between one support contact and the nearest load contact, mm (in.), b is the width of specimen, mm (in.), d is the thickness (height) of specimen, mm (in.). For a detailed discussion of the standard assumptions see Baratta et al.10 The standard test method for glass, ASTM C-158, recommends that the ratio of the width to thickness of standard MOR specimens be between 2:1 and 10:1. Because the above nonstandard vertical MOR specimens will generally fall outside that range, what are the consequences and are there any other limits to be considered? First we will discuss why we want the height to be as large as possible and then discuss potential problems for such a specimen conguration. Consider, rst, Fig. 2 which shows the fracture surface of two edge strength specimens where most of the specimen cross section, including the neutral axis, is below the photo, out of sight. The highest tensile stressed portion of these specimens is the surface, the thin edge, at the top of the photo. Note that the break origin of the left specimen is on the side rather than on the top edge as in the specimen on the right. This suggests that this particular edge preparation process may have caused some face damage near the edge that is worse than any damage on the edge. Let us recall that the stresses in this exure test decrease linearly to zero at the neutral axis as sketched in Fig. 3. Thus the fracture of the right-hand side specimen in Fig. 2 started at the maximum stress value while for the left-hand specimen it started at a stress lower than the maximum. If we are interested in potential aws, for example due to undesired grinding damage, slightly away from the edge, we need a stress gradient to stress the area in question as uniformly as possible. The taller the specimen, the lower the stress gradientcompare the two sketches in Fig. 3and hence more likely to nd these near-edge aws.

www.ceramics.org/IJAGS

Edge Strength Testing of Thin Glasses

41

Fig. 2. The break source of the right specimen is farthest from the neutral axis while that of the left specimen is on the face, closer to the neutral axis than the edge of specimen, suggesting the edge processing generated damage near the edge. The arrows near the bottom help the eye nd the width, b, of each specimen.

A question arises as to whether one should simply use the maximum strength reached at the edge as the strength to be used in further analysis or should one calculate the actual strength of that aw. The maximum strength, when the origin is on the face, could be considered to be a censored, minimum edge strength. And, if the prime need for the test data is to rank, or improve the edge processing, then the authors feel using the basic maximum calculated strength is appropriate and has been used in the rest of this paper. By having a large height, the stress gradient from edge to neutral axis is relatively low. However, one cannot just make the specimen very high as there are limits to the goodness of simple beam theory. For example, if the beam is too high relative to its span, both the analysis and experimental work by Conway et al.,11 Kaar,12 and Saad and Hendry13 show that the linear stress dis-

tribution in a simple beam becomes convoluted such that Mc/I is not a good predictor. In these authors studies of three-point exure they concluded that the classic stress distribution started to deviate at a span to height ratio of about 2. For very deep beams their calculations predict a lower stress value at the edge and possibly a higher stress value away from the edge. Even though the applied loads in four-point exure test are more uniformly distributed than in three-point exure test, maintaining the span to height ratios 42 would provide accurate enough edge strength data via the classic bending beam formula. A fairly obvious problem with a tall and thin exure specimen, not normally a concern in typical testing, is that it could buckle sideways under high load; see Fig. 4. Lateral instability of this nature could induce higher tensile stresses on the top edge of the buckled face than on the bottom due to the smaller local bending radius. Unfortunately, this mode of failure has been observed. Because the edge strength is not known a priori, guidance formulas on appropriate thickness, height, and spans ratios were developed to be conservative enough without the expense of more exact nite element analysis taking xturing and the loading process into account. We start with published solutions to boundary conditions that are rough approximations of these tests. The formulas used are from Roarks Formulas for Stress and Strain14 and the software used for calculating stresses for the plots is TK Solver 4.0 from Universal Technical Systems Inc. (Rockford, IL). Consider a straight uniform beam of width b, height d, and length l. Two standard cases appear appropriate as estimators. The rst case has the beam subjected to a pure bending moment M with its ends held vertically but not xed. The second case assumes that the beam is loaded with the force P on the top face in

Fig. 3. As illustrated in these two diagrams, the specimen with the larger height has a lower stress gradient so that a larger surface area is stressed close to the maximum. The red dash-dot line indicates the neutral axis where the exural stresses in the beam are zero.

Fig. 4. An unfortunate side effect of testing a bar on its narrow edge, is the tendency to bow outward leading to different bending stresses and alternate failure modes often known as lateral instability.

42

International Journal of Applied Glass ScienceGulati and Helnstine

Vol. 2, No. 1, 2011

the middle of the span l (i.e., a three-point bend test). The ends in both cases are simply supported but constrained from twisting. The bending moment M is half the applied load multiplied by the moment arm which, as explained earlier, is half the difference in support and load spans. The length l represents support span, which is obviously less than the undened specimen length. If we substitute into these formulae the basic beam bending stress, s 5 Mc/I, we can estimate the edge stress developed at the initiation of the lateral buckling.  2  s    b d b sC EG 1 0:63 d l d where E and G are the elastic and shear moduli of beam material respectively and C is equal to p for the pure bending case and equal to 4.23 for the three-point exure loaded case. The more conservative value of C for our specimen and xture design size is p. The importance of the ratio b/d now becomes clearer as we design a specimen that will fail in bending rather than in buckling, that is the stress given by the above equation, should be higher than the expected bending strength. To help understand this formula further, Fig. 5 shows a plot of log(l/d) vs log(d/b) for a generic glass. Thus, given the width b of a glass specimen, its height d can be easily determined. Similarly,

the y-axis can be used to determine the length l from the height d. Unfortunately, sometimes one must iterate these specimen dimensions until proper failure modes are observed. The authors recommend that all specimens should not be made before verifying proper failure modes are observed.

Experimental Setup A basic four-point bend xture that meets the requirements of, say, ASTM C-1581 or ASTM C-116115 is a good starting point. However, several modications need to be made to control both the specimen and loading process. Figure 6, a photograph of one of the xtures is used as an illustration. First, side supports for test specimen need to be added to control vertical alignment along with lateral stability. These supports, positioned at the support span rollers, need to be close tting, yet have very little friction. A low friction polymer like Teont (DuPont, Wilmington, DE) or a steel pin with space to put a Teont lm between the pin and test specimen is appropriate. Because not all the specimens have identical thickness, the supports need to be adjustable or replaceable with conforming diameters. Alternatively, multiple

Fig. 5. To reduce the possibility of lateral instability during an edge test, the specimen and xture dimensions should be on left of the glass estimated strength line and above the length/height ratio of 2.

www.ceramics.org/IJAGS

Edge Strength Testing of Thin Glasses

43

Fig. 6. Vertical MOR test xture for measuring edge strength with side supports at support rollers, and lead sheet cushions between load rollers and the specimen.

xtures can be made to accommodate different specimen widths. In the standard MOR test xtures both the load span and support span knife edges/rollers are designed to rock parallel and normal to the long axis of specimen. This allows adjustment to make good contact in imperfect specimens that could be tapered either along the length or across the height. For the edge test, however, any sideways loading may induce lateral buckling at lower loads than those with pure vertical loading. Fortunately, any taper across a thin specimen is not likely to be large. Thus, to accommodate specimens with tapered height, a soft insert is used between the load roller and specimen. Appropriate materials for such an insert include lead, polymers, etc. To reduce any side loading, the load-span contacts should be rigid and parallel to the support span contacts. Because lateral buckling is a possibility during the edge test, the failure mode of each specimen needs to be veried whether it is a valid (near) edge failure or a result of side buckling with a top edge failure. Inspection of break origins, crack/fracture pattern, and the load deection curve are the primary means of ensuring valid failures. The loaddeection curve is the easiest one to inspect because it is readily observable during the test. A proper test will exhibit a straight line to failure as illustrated by the straight line in the Fig. 7 schematic. If the specimen undergoes lateral buckling, the loaddeection curve will bend over as illustrated by the dotted curve in the Fig. 7 schematic. However, the bent over loaddeection curve may not be as dramatic as illustrated. Sometimes the specimen will fail well before the

Fig. 7. Load versus time (or cross-head) curves for a valid test, solid line, and probable buckled invalid test specimen (dotted line).

loaddeection curve is developed due to more severe aws at the top edge; however, this can be alleviated by improving the quality of the top edge. To understand this buckling phenomenon further, a 0.7-mm-thick edge strength specimen was strain gaged with two gages near the top edge of the 19.0 mm (3/4 in.) high specimen, near the normal compression edge of a bent beam. As the load increased, the specimen exhibited lateral instability whereby one gage went nonlinearly in tension and the gage on the other side went into nonlinear compression, see Fig. 8. This side buckling mode is not as abrupt as the classic buckling behavior, but nevertheless is obvious enough to be used as a validity check. Another specimen, only 12.7 mm (1/2 in.) high, strain gaged similarly did not exhibit this load/strain bifurcation; suggesting in this case that a reduction in the height could provide control over the buckling failure mode. The next inspection step is to check the fracture pattern for cracking direction, that is whether fracture initiated at the desired edge or not. Figure 9 shows two examples of valid test patterns. Invalid tests have origins on the top (nominal compressive) edge of the specimen. Taping the specimen before testing with a compliant transparent tape outside the high tensile stress region is recommended. If the crack pattern is not sufcient to determine the direction of crack growth, then the nal test is to look for fracture origin using a magnifying lens or microscope.

44

International Journal of Applied Glass ScienceGulati and Helnstine

Vol. 2, No. 1, 2011

Fig. 8. Strain gage output for buckled vertical MOR specimen 19 mm high 0.7 mm thick. Note that compressive strain is considered positive.

Applications of Edge Strength Test Several applications have been reported elsewhere in more detail about the use of this exural edge test. Gulati et al.16 conducted a round robin study of edge strength measurement for annealed soda-lime-silica glass which included different test methods. The importance of edge nish to thermal tempering of the same glass was also studied by Gulati et al.17 Because the aw severity on ground edges may be different than that on specimen faces, the fatigue behavior of edge aws was reported by Akcakaya and Gulati18 for soda-lime-silica glass. A couple of other example applications, including

acid etching of 1.1-mm-thick glass, and the manufactured edge strength of automobile windshield glass, are discussed below as illustrative of the utility of this vertical exure test.
Improving Edge Quality of Aluminosilicate Glasses

A glass substrate with scored edges was breaking from the edge more often than acceptable in production. One proposed solution was to strengthen the edges by acid etching. Such a treatment could reduce aw severity by reducing aw depth and possibly increasing its tip radius. An experiment was conducted using

Fig. 9. Valid fracture patterns in annealed (A) and heat strengthened (B) specimens. Invalid fracture would start at the opposite edge, the nominal compression side, due to buckling.

www.ceramics.org/IJAGS

Edge Strength Testing of Thin Glasses

45

(i) standard wash, (ii) 10 min etch with 2% HF acid, and (iii) 10 min etch with 5% HF acid. The edge test specimens were 1.1 mm wide, 12 mm high, and about 75 mm long. The support span was 50.8 mm and the load span was 25.4 mm. These dimensions yield height to width ratio of 10.9 and length to height ratio of 4.2, which are well within our guideline. Figure 10 shows Weibulls distribution of edge strength for the three different treatments. The mean strengths were 92 MPa (standard wash), 104 MPa (2% HF etch), and 151 MPa (5% HF etch). While the mean strength following 5% HF treatment is signicantly higher (at 95% condence level) than that with either standard wash or 2% HF etch, the Weibull distribution plot suggests that mean strength does not tell the whole story. Let us note that high strength data are separated whereas the low-strength data tend to congregate together, thus suggesting that the largest aws were not reduced in their severity by either of the acid etch treatments. Perhaps identifying the source of original aws and minimizing them might be more benecial than any postaw-generation treatment.
Edge Strength of Automotive Glass

The mechanical integrity of an automotive windshield depends not only on surface but also edge aws. The latter arise from scoring and nishing processes of single glass plies before sagging and lamination with the polyvinyl butyral (PVB) layer. The edges experience

signicant vibrational and impact stresses during the windshield manufacturing process, which might lead to fracture unless the edges are nished properly. To quantify aw distribution, 30 specimens, each measuring 2.4 mm thick, 25.4 mm high, and 152 mm long, were prepared from around the perimeter of single ply of soda-lime-silica glass (without disturbing the scored and nished edge) and tested on a support span of 102 mm and load span of 51 mm.19 Inspection of the failure strength distribution, the failed glass, and the cutting process determined that a poor-quality diamond nishing wheel combined with improperly supported edges of the large and thin glass ply lead to a poor strength distribution. These ndings helped windshield manufacturer modify both the nishing process and support xture, which improved the edge nish and cut down production losses. The edge nish can also play an important role in thermal tempering of 3-mm-thick sidelites and backlites for passenger cars, SUVs, and light-duty trucks. If the temporary tension during initial quenching of sodalime-silica glass exceeds the edge strength, production yields are signicantly affected. One way to minimize such failures is to heat treat edges in the transformation temperature range thereby healing the edge aws and improving the edge strength sufciently. The effect of different heat treatments on edge quality was examined by measuring edge strength of 3 mm thick, 20 mm high, and 110 mm long specimens using a support span of 100 mm and load span of 50 mm. A sample size of 10 or 15 specimens was used for each of the heat treatments.17 These data are also plotted as Weibulls distributions in Fig. 11.

Fig. 10. Weibulls distribution of edge strength of aluminosilicate glass following three different edge treatments.

Fig. 11. Weibulls distribution of strength data for specimens before and after different heat treatments.

46

International Journal of Applied Glass ScienceGulati and Helnstine

Vol. 2, No. 1, 2011

It is clear from Fig. 11 that the best strength improvement was obtained after the heat treatment at 6601C where aw healing was signicant due to lower viscosity. The minimum strength, which controls premature failure during tempering, increased by 15%. The fast heat treatment, despite higher temperature, did not affect minimum strength due, possibly, to a temperature lag of the relatively massive glass specimens which did not heal aws as well as did the slow heat treatment at 6301C. Summary The edge strength test described in this paper, the four-point vertical bend test (VBT), offers the following advantages: (i) it encompasses almost all of the edge and nearedge aws associated with edge nishing; (ii) it is a relatively simple test and similar to the conventional MOR exure test; (iii) it can be carried out in various environments much like the MOR test; (iv) it is a convenient test for measuring dynamic fatigue of edge aws; and (v) it can be used for any material that exhibits elastic behavior up to failure. The paper also provides guidelines for designing test xtures and specimen dimensions to minimize lateral instability, notably for thin specimens, which might lead to lateral buckling type failure before exural failure. The quality of the compression edge may also need sufcient care to minimize buckling failures. Several applications of edge strength test are provided for commercial glasses that help quantify the edge nishing process for annealed, heat-strengthened and tempered glasses. Such data can help glass manufacturers optimize edge nish thereby minimizing production losses while improving mechanical reliability simultaneously. The edge strength test is gaining popularity, notably for thin and ultrathin glasses used in consumer electronics as display or cover glass. Acknowledgments The authors are grateful to Tim Roe and Jim Brown for designing test xtures and to Tim Roe, Mike Hobczuk, Marlene Hillman, and Mark Neuman

for measuring the edge strength of specimens of different commercial glasses. The authors also express their gratitude to Cornings S & T Division for approving the publication of this work which was presented on October 16, 2002 at The American Ceramic Society Glass & Optical Materials Division Fall Meeting in Pittsburgh.

References
1. ASTM. Standard Test Methods for Strength of Glass by Flexure (Determination of Modulus of Rupture), ASTM C-158-02. ASTM, West Conshohocken, PA, 2007. 2. ASTM, Standard Test Method for Biaxial Flexure Strength (Modulus of Rupture) of Ceramic Substrates, ASTM F-394-78 (withdrawn 2001), ASTM, West Conshohocken, PA, 1996. 3. ASTM. Standard Test Method for Monotonic Equibiaxial Flexural Strength of Advanced Ceramics at Ambient Temperature, ASTM C-149909. ASTM, West Conshohocken, PA, 2010. 4. T. Cleary and S. Gulati, Inuence of Glass Score and Seam Orientation of Edge Strength of Multi-Layered Glass Articles, Fractography of Glasses and Ceramics IV, Ceramic Transactions, Vol. 122, eds., J. R. Varner and G. D. Quinn. American Ceramic Society, Westerville, OH, 327341, 2001. 5. C. P. Pantelides, G. P. Sallee, and J. E. Minor, Edge Strength of Window Glass by Mechanical Test, J. Eng. Mech., ASCE, 120 [5] 10761090 (1994). 6. D. Stahn, Increased Strength of Flat Glass Articles by a New Cutting Method, Strength of Inorganic Glass, ed., C. R. Kurkjian. American Ceramic Society, Westerville, OH, 443452, 1985. 7. M. Inoue, T. Ono, and I. Sato, Evaluation of Strength and Edge Flaw Severity of Barium Borosilicate Glass Substrate by Thermal Downshock Testing, J. Cer. Soc. Jpn., 101 [2] 149153 (1993). 8. O. Gaume and R. Gy, Measurement of Edge Strength using a Special Thermal Test, Vol. XX. ICG Mtg., Kyoto, Japan, 2004. 9. M. J. Kerper, T. G. Scuderi, and E. H. Eimer, Strength of Glass as Related to Edge Finish, Progress Report No. 1, Covering the Period April 1962 to June 1962, NBS Report No. 7595. 10. F. I. Barrata, W. T. Matthews, and G. D. Quinn, Errors associated with Flexure Testing of brittle Materials, U.S. Army Materials Technology Laboratory Technical Report MTL TR 8735, July 1987. 11. H. D. Conway, L. Chow, and G. W. Morgan, Analysis of Deep Beams, J. Appl. Mech,, 18, 163172 (1951) see also discussion on 421423. 12. P. H. Kaar, Stresses in Centrally Loaded Deep Beams, Proc. S. Exp. Stress Analyses, XV [1] 7784 (1957). 13. S. Saad and A. W. Hendry, Stresses in a Deep Beam with a Central Concentrated Load, Experim. Mech., 1 [6] 192198 (1961). 14. W. C. Young, and R. G. Budynas ed. Roarks Formulas for Stress and Strain, 7th edition, McGraw-Hill, New York, 2002. Table 15.1, Cases 11 & 13, p. 728; or 6th ed. Table 34, p. 680. 15. ASTM. Standard Test Method for Flexural Strength of Advanced Ceramics at Ambient Temperature, ASTM C-1161. ASTM, West Conshohocken, PA. 16. S. T. Gulati, et al. Round Robin Test Results for Edge Strength of Float Glass, Vol. XX. ICG Mtg., Kyoto, Japan, 2004. 17. S. T. Gulati, T. Roe, and J. Vitkala. , Importance of Edge Finish to Thermal Tempering, GPD Conference, Tampere, Finland, 2001 18. R. Akcakaya and S. T. Gulati, Effect of Edge Finish of Float Glass Products on their Strength and Fatigue Behavior. ICG Mtg, Amsterdam, the Netherlands, 2000. 19. S. T. Gulati and M. A. Khaleel, Design Considerations for Lightweight Windshields, Symposium on International Automotive Technology (SIAT 2001SAE Conference), Pune, India, SAE Paper 2001-01-0029, January 1013, 2001.

Potrebbero piacerti anche