Employing glass panels with rounded corners to mitigate seismic damage in architectural glass wall systems

Transcription

1 Employing glass panels with rounded corners to mitigate seismic damage in architectural glass wall systems A.M. Memari, P.A. Kremer Department of Architectural Engineering, The Pennsylvania State University, United States Abstract Racking motions during an earthquake can lead to serviceability failure (e.g., glazing gasket pullout. sealant damage. glass edge damage and glass cracking) or even ultimate failure (in the form of glass fallout that presents a threat to life safety) in conventionally glazed wall systems - even those that meet current building code provisions for nonstructural elements. A new approach to mitigation of seismic risk in conventionally glazed wall systems with architectural glass panels has been developed at the Building Envelope Research Laboratory (BERL) at The Pennsylvania State University. The essence of the approach is to modify the rectangular geometry of architectural glass panels at the corners through rounding. A pilot study at BERL has shown that rounding the corners of architectural glass panels can increase the drift capacity of the panels significantly and that maximum effectiveness is achieved by employing rounded corners with a 1 in. radius of curvature along with beveled and polished glass edges. Results of these inplane dynamic racking tests performed on full-scale mock-ups of curtain wall sections glazed with architectural glass panels of different glass types. employing various radii of curvature at the corners, various glass edge conditions. and varying glass-to-frame clearances are presented. 1 Introduction Recent earthquakes in the U.S. have focused attention on the vulnerability of architectural glass to seismic movements because of the widespread use of glass facades throughout seismic regions [l -51. Earthquake-induced damage to architectural glass components often: (1) necessitates expensive repairs. (2) exposes the contents of a building to weather. theft and vandalism. (3)

2 232 Eavihqunke Resistant Enplneer~ng Structures III causes a disruption in the activities within the building. and (4) presents a threat to life safety when glass falls from a damaged wall system. Although some cladding systems can offer an appropriate level of resistance to architectural glass damage during moderate to severe earthquakes. the majority of architectural glass cladding installations cannot. When potential liabilities associated with architectural glass damage are considered. it becomes clear that something should be done to protect these vulnerable buildings. and more importantly. the public from the specter of glass fallout. The Building Envelope Research Laboratory (BERL) at The Pennsylvania State University has developed andtor improved and tested several mitigation options for both new and retrofit wall system installations including: (1) seismically isolated wall systems for new buildings or recladding of existing buildings. (2) anchoring techniques for applied safety film for retrofit to existing windows. and (3) laminated glass for new architectural glass installations or replacement of existing glass panels. In fact, one of the systems under option 1 has been shown to provide complete resistance to earthquake-induced glass cracking and glass fallout [6]. Options 2 and 3 have been shown to provide excellent fallout resistance, but in most instances, depending on the glazing details, the glass panels still remain vulnerable to glass cracking at drift amplitudes commensurate with moderate earthquakes. Another option often used by curtain wall designers for newer cladding installations [7] is to adjust the glass-to-frame clearance and use appropriate spacer materials between the glass and frame. This option allows the glass panels to move independently within the racked frame during an earthquake without touching the frame. Despite the proven effectiveness of these options. they often exceed the cladding budget of the building owner. are not always practicable. and in some instances may limit the architectural aesthetics of the resulting design. Moreover, with no explicit code mandating an evaluation of the glass fallout potential of a given architectural glass installation (a problem that will be addressed with a new AAMA test method being developed [g]). building owners often select lower cost and vulnerable cladding systems for new installations or ignore the potential problems that exist with their existing cladding. A lower cost approach relative to the options discussed above has been conceived at BERL as yet another potential alternative for the mitigation of seismic risk of architectural glass panels in various wall systems. The essence of this approach is a slight modification of the rectangular geometry of architectural glass panels at the corners through rounding. Previous BERL studies [ l] have shown that earthquake-induced glass damage is initiated in the corners of architectural glass panels. In a similar manner as increased glass-to-frame clearances. corner rounding allows the glass to rotate more freely within the curtain wall frame when the frame is subjected to seismic movements. Thus. it is expected that corner rounding of architectural glass panels can increase both the serviceability (glass cracking) and ultimate (glass fallout) drift limit states of architectural glass panels.

3 Earthquake Reszstant Eng~neering Structures The results of a recent pilot study at BERL to evaluate the feasibility of architectural glass panels employing rounded corners to improve their resistance to seismic movements are presented in this paper. The objectives of this pilot study were to: (1) determine the optimum radii for rounding; (2) evaluate the effect of rounded corner surface condition on limit state performance; and (3) evaluate the effect of glass-to-frame clearance on limit state performance. 2 Experimental plan and dynamic crescendo test method In-plane dynamic racking crescendo tests as described below were performed on the architectural glass specimens in the test matrix in Table 1. Glass dimensions were chosen to allow comparisons with limit state data from previous studies conducted on similarly glazed specimens. General details of the curtain wall glazing components used to glaze the specimens are given in Figure 1. All glass panels were installed in a dry-glazed, Kawneer 1600TM wall system, which uses rubber gaskets between the glass and the aluminum curtain wall frame to secure each glass panel perimeter. Setting blocks and side spacers were modified for those specimens constructed with glass-toframe clearances less than the nominal 0.5 in. clearance typically used in the Kawneer 1600TM wall system. Because this study was exploratory in nature. the number of specimens of each glass type was kept to a minimum. The effect of three variables on the limit state performance of monolithic glass panels subjected to simulated seismic movements were evaluated - Configurations 1-6 were used to determine the effect of various radii of curvature at each corner of the AN monolithic glass panels (0 in. or rectangular. 'l2 in.. 1 in.. 2 in., 3 in., and 3 in./ 5 in.); Configurations 1. 3, 7, 9, 10 and l l were used to evaluate the effect of edge clearance (112 in. and 3/16 in. nominal glass-to-frame clearance); and Configurations 7 and 8 were used to evaluate the effect of glass edge condition (beveled and polished or "rough" seamed). Serviceability (glass cracking) and ultimate (glass fallout) limit state data were collected from the in-plane dynamic racking crescendo tests on each specimen as discussed later. In-plane dynamic racking crescendo tests were performed on the Dynamic Racking Test Facility described in greater detail by Behr and Belarbi [12]. Specimens were tested one at a time on the facility. Each specimen was centered between the sliding steel tubes of the test facility. and the curtain wall specimen vertical mullions were attached at all four corners to the facility's sliding steel tubes. These steel tubes slide on roller assemblies in opposite directions by means of a fulcrum and pivot arm mechanism. The bottom sliding steel tube was displaced by a computer controlled electrohydraulic servoactuator having a dynamic stroke capacity of + 3 in. (k 76 mm). The fulcrum and pivot arm mechanism attached to the top and

4 234 Earthquake Resistant Eng~neering Stmctwes III bottom sliding steel tubes doubled the effective servoactuator stroke capacity to k 6 in. (f 152 mm). The dynamic "crescendo test method" introduced by Behr and Belarbi [l21 and currently being developed as a standard test method by AAMA [8] to determine the A,,,,,, (glass fallout) limit state for storefront and curtain wall systems was used for the dynamic in-plane racking tests. The uninterrupted crescendo test drift time history is shown in Figure 2. The crescendo test consisted of a series of alternating "ramp up" and "constant amplitude" intervals, each comprised of four sinusoidal cycles at a nominal frequency of 0.8 Hz for drift amplitudes up to f 3 in. (+ 76 mm) and 0.4 Hz for drift amplitudes up to the facility limit of rt 6 in. (f 152 mm). The frequency adjustment during the crescendo tests was necessary because of hydraulic power supply and servoactuator volumetric flow limitations. Glass contact was established using foil sensors in conjunction with a custom-built contact sensor box. Interruptions of the crescendo test along with a synchronized video recording of each test were used to more precisely determine serviceability and ultimate drift limits for each specimen. Kawnscr 16W Shndnrd Top 14mm Bomm lomm I I Lsn 12mm R,gh, 12mm, Figure 1 : General glazing details for the specimens described in Table l.

5 Earthquake Rrsrsranr Enplrzeerlng Structures I Table 1: Test matrix for rounded corner dynamic racking tests Corner / Edge Conditions Nominal Glass-To- Frame Clearance No. of Specimens 2 radius rounded l Unseamed in, radius rounded l Unseamed in. radius rounded l Unseamed in. radius rounded l Unseamed in./ 5 in.' radius rounded l Unseamed in. FT 1 in. radius rounded l beveled and polished in. FT 1 in. radius "rough" rounded l seamed 1/2 in Rectangular l Unseamed 1 in. radius rounded l Unseamed 3116 in.' 3116 in. ' in. FT 1 in. radius rounded l beveled and polished in.' 1 ' All glass specimens were clear monolithic glass panels with 5 X 6 ft. outside dimensions. AN refers to annealed glass panels. and FT refers to fully tempered glass panels. AN glass panels had a surface compressive prestress of about 600 psi. and FT glass panels had a surface compressive prestress of about 15,000 psi. "Unseamed" refers to glass edges that have not been sanded. which is the typical edge condition of AN glass panels. "Seamed" refers to glass edges that have been sanded, which is the typical edge condition of tempered glass panels. ' A 3 in. radius of curvature was used along the horizontal (5 ft. side) portion of each corner and a 5 in. radius of curvature was used along the vertical (6 ft. side) portion of each corner. ' Glass edges had a 1116 in. bevel at each edge and were polished. Beveled and polished edges were used because doing so was thought to eliminate many flaws where cracking can initiate and to reduce the coefficient of friction between the glass and the frame. Configuration 9 clearance was actually in.. Configuration 10 clearance was actually 118 in.. and Configuration l l clearance was actually 5/32 in. due to variability in the fabrication of the specimen frames.

6 236 Eartlzquahe Reszstant Engineering Strz~ctzwes I11 m P a U Q Time (sec) 2.4 g g -2.4 $: Figure 2: Drift time history for dynamic racking tests. (1 mm = in.) 3 Test results Serviceability data recorded during the crescendo tests included: (1) glass-toframe contact, (2) the "serviceability drift limit." defined by Behr [9] as the drift required to cause observable glass cracking (a condition that would necessitate glass replacement, but one that would not pose an immediate life safety hazard): and (3) the "ultimate drift limit." defined by Behr [g] as the drift required to cause fallout of glass fragments with a surface area greater than one square inch (a condition that could pose a life safety hazard to building occupants and pedestrians). These serviceability data are presented in Figure 3. Glass panels with rounded corners did not, in general. change the glass damage mechanism observed in previous studies for rectangular glass panels glazed in conventional curtain wall frames [5. 6, 9-11]. Glass-to-frame contact was observed to occur at a drift amplitude of about 1 in. for those configurations with corner rounding of 1 in. or less. This is not surprising because the tip alone of rectangular glass corners does not typically impinge upon the frame, but rather. contact occurs over a length of glass in the corner region. Interestingly, the glass cracking and glass fallout limit states for annealed glass panels with rounded corners are nearly the same: however, in annealed rectangular glass panels, the limit states often differ by a large amount such as the 51"" difference in Configuration 1. In rectangular annealed glass panels. cracking in the corner regions is typically initiated as fragments in the corners are sheared from the main panel. This localized fracture behavior at the onset of glass cracking in AN glass panels is due to the low residual surface stress in the glass. Of course. in HS and FT glass panels. which are observed to fracture more extensively because of their increased residual surface stress. glass cracking coincides with glass fallout most of the time in HS glass and always in the case of FT glass. This shearing of small fragments at the corners of rectangular glass panels acts in a similar manner to corner rounding in that the panel is then able to rotate somewhat more freely; hence, fallout is typically delayed until the panel is

7 bound into a position that results in a load transfer to the panel of sufficient magnitude to cause additional fracturing and subsequent fallout in the glass panel. Configurations 1-6 were used to establish an approximate optimum degree of corner rounding for AN glass panels. As shown in Figure 3. maximum glass cracking and glass fallout limit state performance in the specimens tested occurred within those specimens with 1 in. radius corners (Configuration 3). Although not enough specimens were tested at each radius to allow rigorous statistical comparisons. there were enough specimens tested up to and including the 1 in. radius specimens (Configurations 1-3) to substantiate the observed trend through these specimens in Figure 3. It is fortuitous that the maximum benefit does not appear to occur beyond 1 in. corner radii because specimens with corners rounded beyond a 1 in. radius (Configurations 4-6) would in most wall systems require some modification to the wall system in the corner regions to close the air gap that the rounded corners would leave. In fact, as the radius was increased beyond 1 in. the glass panels were observed to move perhaps too freely within the racking curtain wall frame. In the case of the specimen with 3 in. rounded corners (Configuration 6), the excessive movement led to earlier glass cracking and glass fallout because the panel moved itself more quickly into a bound position for load transfer to the glass panel to occur. The Configuration 6 specimen (combination of a 3 in. radius along the horizontal portion of each corner and a 5 in. radius along the vertical portion of each corner) also exhibited excessive movement within the racking frame. but the increased rounding along the vertical edge at each corner delayed load transfer from the frame (and thus, glass cracking and fallout) as compared to the 3 in. radius specimen. It is also quite clear from Configurations 1-6 in Figure 3 that rounded corners can have a significant effect on the glass cracking limit state and somewhat less of an effect on the glass fallout limit state in AN monolithic glass panels. Most notably, a 95% increase in the glass cracking limit state and a 22% increase in the glass fallout limit state were observed for the 1 in. rounded corner specimens (Configuration 2) as compared to the performance of those specimens with no corner rounding (Configuration 1). This performance for 1 in. rounded corner specimens is on par with the in. glass cracking and glass fallout limit state performance determined through previous testing by Behr [9] for FT monolithic rectangular glass panels. No attempt was made to dress the edges of the AN monolithic glass panels through edge seaming, but doing so would be expected to incrementally improve the limit state performance of the AN glass panels further as first noted by Behr et al. [10]. Configurations 7 and 8 were used to evaluate the effect of edge and corner condition on the drift capacity of FT monolithic glass panels employing the optimum (of those radii considered in this study) 1 in. radii corners. Two edge conditions were considered: (1) beveled and polished edges

8 238 Earthyzmke Resistant Engineering Str.uctwes 111 (Configuration 7) and (2) seamed edges and roughly rounded glass corners. Roughly rounded corners were characterized by glass protrusions along the rounded corners rather than the smooth and continuous curvature which was characteristic of the other configurations with rounded corners. The limit state data collected by Behr [9] along with that from subsequent studies for identically glazed and sized FT monolithic panels with rectangular corners provides a good baseline for Configurations 7 and 8 comparisons. These previous studies (not discussed here) indicated that the glass cracking and glass fallout limit states of FT monolithic glass with rectangular corners are about in. Thus. an increase in the glass cracking and glass fallout limit states of 54% was observed for Configuration 7 as compared to FT rectangular glass panels, and a decrease of 37% in the cracking and fallout limit states was observed for Configuration 8 as compared to FT rectangular glass panels. These data suggest that edge condition substantially affects the limit state performance of rounded corner glass panels and that glazing specifiers would need to check that rounded corners are constructed properly to ensure the desired performance of panels employing rounded corners. The drift index (drift index was determined by dividing the limit state in in. by the glass panel height of 81 in. (2057 mm)) associated with glass cracking and glass fallout for Configuration 7 was 5.5%. which suggests that FT glass panels with 1 in. radius beveled and polished rounded corners could offer serviceable performance during severe earthquakes. Another benefit of the beveled and polished edges is reduced frame damage when glass panel failure occurs. Glass panel failure often leaves behind significant scraping and gouge marks within the glazing pocket of aluminum frames such as those used in this study, but the glass panels with polished edges caused only minor cosmetic frame damage due to permanent deformations at the points where the glass engaged the frame during panel failure. Although no annealed specimens were prepared with beveled and polished edges. it is expected that they would also derive additional limit state performance increases and frame damage reductions than rounding alone can provide. The effect of reduced glass-to-frame clearance on limit state performance of both rectangular and rounded corner specimens was explored in Configurations and 11. These specimens were to be constructed with a nominal in. glass-to-frame clearance, but as noted in Table 1. the actual edge clearances varied somewhat from in. In particular, Configuration 10 had only a 118 in. edge clearance. Nonetheless. these limited data points are invaluable because very little information is available regarding the effect of edge clearance on limit state performance. Comparison of the limit state data for Configurations1 and 9 (AN glass panels with rectangular corners) suggests that the 59% reduction in edge clearance yields a 17Y0 decrease in the cracking limit state and a 40% decrease in the fallout limit state. Comparison of the limit state data for Configurations 3 and 10 (AN glass panels with 1 in. rounded corners) suggests that a 75%) decrease in edge

9 Earthquake Res~stant E11plneerq Structures III 239 clearance yields a 57% decrease in the cracking limit state and a 51% decrease in the fallout limit state. Finally, comparison of the limit state data for Configurations 7 and 11 (FT glass panels with 1 in. rounded corners and polished and beveled edges) suggests that a 69% decrease in edge clearance yields a 38% decrease in both the cracking and fallout limit state. Despite the decrease in limit state performance with decreased edge clearance for Configuration 11 as compared to Configuration 7, its limit state performance was still quite good. In fact. the limit state performance of Configuration 11 was only 4% less than that for lt monolithic glass panels with rectangular corners and a 112 in. edge clearance. The observed drift index of 3.4'% at glass cracking and fallout for Configuration 11 suggests that even with reduced edge clearances. glass panels with rounded corners and polished and beveled edges could still remain serviceable in severe earthquakes. Figure 3: Pilot test results showing the effects of corner rounding on 114 in. annealed and fully tempered monolithic glass panels glazed in Kawneer 1600TM curtain wall frames with (a) 112 in. nominal glassto-frame clearances. and (b) 3/16 in. nominal glass-to-frame clearances. 4 Conclusions This pilot laboratory study of the behavior of architectural glass panels ~ ith rounded corners when subjected to dynamic racking crescendo tests has indicated that rounded corners demonstrably increase both the serviceability (glass cracking) and the ultimate (glass fallout) limit states of monolithic

10 240 Enrtlzquake Resistant Engineering Strwtures III architectural glass panels. Of the various radii tested, 1 in. radius corners offered the most significant gains in serviceability, yet even 112 in. rounded corners offered a substantial increase in serviceability as compared to conventional, rectangular corners. Thus, it is expected that no modifications to most wall system glazing details would be necessary to accommodate architectural glass panels employing rounded corners. This study has also underscored the importance of edge condition on the drift capacity of monolithic architectural glass panels, particularly for fully tempered glass panels. In fact, it was shown that the glass cracking and glass fallout limit states of FT monolithic glass panels employing roughly rounded corners are reduced by 40% as compared to rectangular cornered panels. However, those specimens with corners that were smoothly rounded and glass edges that were beveled and polished. offer superior performance than those that are just smoothly rounded. Moreover, the beveled and polished rounded corner specimens also performed well with minimal edge clearances. Although these tests were limited to monolithic glass panels. insulating. laminated and filmed glass units are also expected to exhibit improved serviceability and ultimate limit state performance when lites with rounded corners are used in their construction. The authors are currently planning follow up studies to: (1) investigate the application of rounded glass corners in monolithic. insulating, laminated and filmed glass units within a variety of wall systems (e.g., curtain wall, storefront and structural silicone glazed) as well as potential performance changes due to the misalignment of lites in laminated and IGU units. (2) determine more precisely the optimum rounding radii; (3) determine if rounded corners with a larger radii for the vertical than the horizontal component of the corner can improve the limit state performance further; (4) assess the performance of rounded corners in bent glass panels and (5) assess the effect of weathering and other in service conditions on the drift capacity of monolithic glass panels with rounded corners. It is clear from this limited study that architectural glass panels with rounded corners, if manufactured properly, may offer the glazing specifier an attractive option that can resist glass damage due to earthquake-induced movements in both retrofit and new cladding installations. The follow up studies described above will provide the additional guidance necessary for the glazing specifier to use architectural glass panels with rounded corners in practice. 5 Acknowledgement Major funding for this study was provided by the National Science Foundation under Grant No. CMS as part of the first author's NSF Career Award. The support of NSF is gratefully acknowledged.

11 6 References [l] Lingnell, A.W. Initial survey and audit of glass and glazing system performance during the earthquake in the Los Angeles area on January Final report submitted to Primary Glass Manufacturers Council. Lignell Consulting Services, [2] EERI. Northridge earthquake January 17, 1994: preliminary reconnaissance report, ed. John F. Hall. Earthquake Engineering Research Institute. Oakland, CA , [3] Evans, D. and Ramirez. F.J.L. Glass damage in the September 1985 Mexico City earthquake. Lessons Learned from the 1985 Mexico City Earthquake, Vitelmo Bertero (ed.), Earthquake Engineering Research Institute. Oakland. CA, p [4] EERI. Northridge earthquake reconnaissance report. Vol. 1. Earthquake Spectra, Supplement C to Vol. l l [5] Wang, M. L., Sakamoto, I. and Bassler. B. L. Design of cladding for earthquakes. Chapter 4, Cladding, Council on Tall Buildings and Urban Habitat. McGraw Hill. Inc.. pp , [6] Brueggeman. J.L.. Behr. R.A., Wulfert, H., Memari. A.M.. and Kremer, P.A. Dynamic racking performance of an earthquake-isolated curtain wall system. Earthquake Spectra, 16 (4). pp [7] Beason. W.L., and Lingnell. W.A. Chapter 8, Emerging uses for window glass. Emerging Materials for Civil Infrastructure State of the Art, eds. R. A. Lopez- Anido and T.R. Naik, American Society for Civil Engineers, Reston. pp [8] American Architectural Manufacturers Association. AAMA 501.X-200x - recommended dynamic test method for determining the seismic drift causing glass fallout from a wall system. Document under ballot in April 2001 by the AAMA Methods of Test task group. Possible publication byaama later in [9] Behr. R. A. Seismic performance of architectural glass in mid-rise curtain wall. Journal of Architectural Engineering. 4 (3). pp [l01 Behr. R. A.. Belarbi. A. and Brown, A. T. Seismic performance of architectural glass in a storefront wall system. Earthquake Spectra. 11 (3). pp [Ill Behr. R. A., Belarbi. A. and Culp. J. H. Dynamic racking tests of curtain wall glass elements with in-plane and out-of-plane motions. Journal of Earthquake Engineering and Structural Dynamics. 24 (l), pp [l21 Behr. R. A. and Belarbi. A. Seismic test methods for architectural glazing systems. Earthquake Spectra. 12 (1). pp , 1996.