Discussion of the AAMA Test Protocol and Test Results

The AAMA 501.6 testing protocol requires that the mockup consist of what is determined to be the critical lites of glass. More specifically, the protocol requires test mockups to include those lites of glass in the curtain-wall system with the largest glass area, the thinnest glass, the most vulnerable glass type and glazing system type, the smallest glass-to-frame clearances, the smallest height-to-width ratio, and the largest drift index. It was determined that the patient room vision lite (5 ft, 2 in. x 5 ft, 3 in. (1575 mm x 1600 mm)) and the shadow box lite above (2 ft, 2 in. x 5 ft, 3 in. (660 mm x 1600 mm)) were the most critical lites to be tested. Figure 7 shows the mockup that was constructed for preliminary testing. The mockup is of stick-built type for this initial testing to cause the mockup to be racked and to determine the code required minimum delta fallout displacement (Afallout). According to ASCE 7-05 [2], Afaiiout shall be larger than the product of 1.25 Ip Dp, where Ip is the important factor and Dp is the design relative story displacement (drift). For this project, and more spe­cifically this mockup, the product of 1.25 Ip Dp was determined to be 3.75 in. (95.2 mm). This was the displacement that the mockup needed to reach under racking load without glass fallout to pass the test.

Because the curtain-wall system for this project is largely unitized and therefore de-coupled from movement in adjacent floors, testing to a racking dis­placement is conservative as the actual sway displacement will be less. There were four primary reasons why this decision was made. First, this was a prelim­inary test to assure OSHPD that a four-SSG system was safe enough to consider for an essential service building like a hospital. As such, it was necessary to es­tablish that the system could perform beyond any drift that it was reasonably expected to experience under the design criterion. Secondly, OSHPD was con­cerned that while the primary mode of behavior for each unit was to sway, that there might be some friction or binding along the sill of the unit, which would force some level of racking to occur. Third, on this project, we have some punched windows that are completely framed into the metal stud system, which also supports the ACM panels and is designed to rack. While not required to fully rack because of the perimeter caulk joint and sliding head channel, these smaller windows will undergo more racking than the unitized units. Finally, the decision was made to construct the corner unitized units in an L-shaped

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014

Downloaded/printed by

Rochester Institute Of Technology pursuant to License Agreement. No further reproductions authorized.

configuration to wrap the corners of the building. As such, these units would be required to undergo more racking than the typical unitized units. With all of these issues in mind, it was determined that the course of action that would most likely establish the performance capability of the system would be to rack the mockup.

With the mockup configuration testing method selected (rack, not sway), it was imperative that the glass, SSG bead size and type, and mullion paint (Kynar) be determined to exactly match what will be installed on the building. Sealant type, bead geometry, surfaces to be adhered to, and glass edge clearance all had to match what would be used on the project (Fig. 4) for the results of the testing to be considered valid for determining the performance of the final design. The resulting details for a horizontal and vertical section through the mullions for the preliminary mockup are shown in Fig. 10.

The AAMA 501.6 test can be used to validate a number of key design elements of the curtain-wall system including: adequacy/performance of the SSG struc­tural sealant bead, adequacy of the glass edge clearance, drift associated with the first evidence of glass cracking (Acracking), which is a serviceability limit state and drift associated with glass fallout (Afallout), which is an ultimate limit state. Of course, the primary objective of the AAMA 501.6 test is to determine Dfallout.

According to AAMA 501.6 protocol, three replicates of a given mockup shall be tested on a dynamic racking test apparatus to determine the drift authorond – ing to glass fallout. For this project, the test apparatus was located at the ATI test facility in York, PA and is shown in Fig. 11 with a typical mockup mounted. Based on the AAMA 501.6 loading protocol, crescendo racking test consisting of a concatenated series of “ramp up" intervals and “constant magnitude" intervals each consisting of four sinusoidal cycles shall be applied to the specimen. The in-plane racking displacement steps between constant amplitude intervals shall be.25 in. (6.4 mm). The test shall be carried out at a frequency of 0.8 Hz for dis­placement amplitudes of 3 in. (76.2 mm) or less and at a frequency of 0.4 Hz for larger amplitudes. This means that for a displacement of 3 in. (76.2 mm) at 0.8 Hz, the displacement rate for top of the mockup would be 9.6 in. (244 mm) per second (576 in. (14630 mm) per minute). For the sealant, however, the strain rate is much lower as mentioned subsequently. Glass fallout drift (Afallout) is defined as the drift corresponding to a piece of glass at least 1 in. [2] (645 mm2) in area breaking away and falling out of the mockup. For this project, the test was stopped after each step to inspect the specimen and, therefore, the con­catenated displacement-time history used is shown in Fig. 12.

An important objective in these tests was to also determine the drift capacity of the structural silicone at glass fallout limit state. For this reason, as has been mentioned, the mockups were designed and attached to the test facil­ity as stick-built systems. The mockups tested had dimensions of 5 ft, 7.5 in. (1715 mm) wide by 13 ft, 11 in. (4242 mm) high. Figure 13 shows one mockup on the test facility. The glass panels used in the mockups were 1 in. (25.4 mm) thick IGU for both vision and spandrel lite with ceramic frit. The glass type used was.25 in. (6.4 mm) thick heat strengthened.

The hysteresis, or load-displacement curves, for the mockup is shown in Fig. 14. There are four loading cycles shown in this figure, which authoronds to:

Copyright by ASTM Int’l (all rights reserved); Tue May 6 Г2:07Г08 EDT 2014

Downloaded/printed by

Rochester Institute Of Technology pursuant to License Agreement. No further reproductions authorized.

FIG. 11—ATI test facility for preliminary testing.

0—2 in. (0—50.8 mm), 0—3.75 in. (0—95.3 mm), 0—4.25 in. (0—108 mm), and 0—4.5 in. (0—114.3 mm) displacement levels. 2 in. (50.8 mm) displacement rep­resents the design level seismic event for this building, i. e., 1.25% drift, whereas 3.75 in. (95.3 mm) represents the displacement that building code requires to be obtained without glass fallout. Both of these displacement cycles show very tightly spaced data. This indicates that there is very little strength loss in the sili­cone sealant as the mockup is cycled back and forth. The general slope of the 0—3.75 in. (0—95.3 mm) displacement cycle is less than that for the 0—2 in. (0—50.8 mm) displacement cycle. This indicates a softening of the silicone

FIG. 12—Displacement-time history used for preliminary racking testing.

Copyright by ASTM Int’l (all rights reserved); Tue May 6 12:07:08 EDT 2014 Downloaded/printed by

sealant resistance as it is cycled through larger displacement levels. This is con­sistent with the sealant test data (Figs. 5 and 6), which shows a reduction in the Young’s Modulus as the elongation is increased. The sealant strain rate is a function of the actual sealant strain experienced at a given drift, the sealant bead thickness and the frequency of the racking. Based on a recent study [18], at a displacement of about 3 in. (76 mm) applied at a frequency of 0.4 Hz, the strain rate for a 9/16-in. (14-mm) – thick bead can vary between 2 in. (51 mm) per

minute to 28 in. (711 mm) per minute. This strain rate is higher than the coupon strain rate of.5 in. (12.7 mm) per minute, but is closer to what would be expected to be experienced in an earthquake as the racking frequencies (0.8 Hz and 0.4 Hz) are representative of the natural (fundamental) frequencies of midrise buildings. It should be noted that, in general, materials show higher strengths at larger load­ing rates. This means that in an actual earthquake, it is expected that the sealants would show higher strength than what is shown by the coupon tests. Therefore, using coupon test results at lower strain rates is actually on the conservative side. The 0—4.5 in. (0—114.3 mm) displacement cycle represents the point at which the largest glass lite fell from the mockup. This data is represented by the color green and is more widely spaced than the previous displacement cycles. This wider spac­ing represents the strength loss as the sealant fails. The load-displacement cycle for 0—4.25 in. (108 mm) was included to show the performance of the sealant and mockup just prior to the cycle in which the glass fell out.

Overlaid on this data are two lines that represent the load-displacement data generated by the finite-element models. One of the lines represent the sili­cone sealant with a Young’s modulus, E = 400 psi (2758 kPa), whereas the sec­ond one represents an E = 100 psi (690 kPa). A Young’s modulus of 400 psi (2758 kPa) more closely represents the tension properties of the sealant (Table 4), whereas a Young’s modulus of 100 psi more closely represents the shear properties of the sealant (Table 3).

According to ATI test report [19], the results show that no glass fallout occurred in any of the three mockups at the target drift of 3.75 in. (95.2 mm). The overall Afallout for the mockup was reported to be 4.25 in. (108.0 mm) drift based on bottom vision lite fallout, which is 13% larger than the design drift of 3.75 in. (95.2 mm). As for sealant performance, minor sealant tear is reported on the exterior side at 3 in. (76.2 mm) drift and on the interior side at 3.75 in. (95.2 mm) drift. Figure 15 shows typical sealant tears at such drift levels. There­fore, as the test results indicate, at drifts close to the design drift, some sealant

tearing occurred but not sufficient for any glass to become disengaged. The seal­ant tearing progressed at drifts beyond the design drift.


The conclusion from the test program is that even under the highly unlikely condition that the unitized system’s stack joints do not function as designed and cause the curtain wall to rack as a stick-built system, this curtain-wall system satisfies ASCE 7-05 seismic provision of Dfallout > 1.25 Ip Dp. Of course, because the final design will be of unitized construction, the glass fallout is certainly not expected to occur under the design drift. Based on the test results, one can con­clude that the sealants are expected to experience some tear at drifts close to the target drift of 3.75 in. (95.2 mm) for a stick-built construction. However, for a unitized system wherein the stack joints allow the adjacent panels to slide past one another in a sway mode, structural sealant damage is not expected to occur at this target drift. The overall conclusion from the study is that four-sided SSG curtain-wall systems can be designed to satisfy the seismic provisions of the building code even in a stick-built construction system. However, because four­sided SSG systems are generally shop-glazed and mostly unitized system is employed, the seismic code provisions with respect to glass fallout are expected to be satisfied more readily and sealant damage (if any) is expected to be much less compared to stick-built systems.