Comparison of Racking Test versus FE Results

Because the silicone sealant is not directly visible in the video images (Fig. 19), another approach must be taken to calculate the elongation of the sealant. The

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.

Lower Left Center Lite – Sealant Bead Elongation

(L4 Camera)

Da, Mockup Lateral Displacement (in.)

Da, Mockup Lateral Displacement (mm)

FIG. 21—Enlarged sealant elongation versus mockup displacement.

elements that are visible and measurable in the videos are the edge of glass and the intersection of the horizontal and vertical mullions behind the glass. Prior to beginning the mock-up testing, adhesive rulers were attached to each corner of the center glass lite and the surrounding glass lite corners (Fig. 19). Video cameras were located at each of the four corners of the center lite of glass. After the testing was completed, still images were extracted from the videos at known testing displacement values. These images were imported into a CAD program. Because the rulers adhered to the glass provide a known dimension, the still images were then scaled so that direct measurements could be taken from the images. By measuring the location of a corner of the glass lite (relative to the intersection of the mullions) from the initial condition and then at a known dis­placement, the relative movement in the horizontal and vertical directions of the corner of the glass lite can be calculated. Because the sealant is adhered to both the glass and the aluminum mullions, it must move the same amount. With this information we can calculate the elongation of the sealant at known displacements of the mock-up.

For this paper, the lower left corner of the center lite of glass (Fig. 3) was focused on. The two graphs (Figs. 20 and 21) show the elongation in the sealant bead at the corner of the glass as a function of the overall mock-up displace­ment. Calculated elongations for the same corner of the glass for all three tested boundary conditions are shown. In addition, the sealant elongation for the same corner of the glass lite for the “racked" boundary condition as predicted by the FE model is also shown.

There are a number of findings from these graphs to be discussed. To begin with, for a given displacement, the elongation in the sealant for boundary condi­tion 1 (sway) is lower than for boundary condition 3 (rack). At a displacement

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.

of 2 in, (51 mm), the sealant elongation in boundary condition 1 is 2.5 %, while the sealant elongation in boundary condition 3 is 10% (Fig. 20). The sealant elongation in the hybrid condition, boundary condition 2, at the same displace­ment is 15 %. Although one would expect to see elongation for boundary condi­tion 2 to be less than that of boundary condition 3, there were other factors that influenced such an outcome. These factors include failure of the index clips con­necting the two parts of the split mullion resulting in rocking action of the pan­els, and lack of corner restraints during test on boundary condition 1.

The comparison between the sway and rack conditions, which were the two main boundary conditions for this study, yields useful sealant performance in­formation, summarized in Table 5. Specifically, in the sway condition, very little strain is placed on the sealant, resulting in minimal stress being transferred through the sealant. In reviewing Fig. 12, it can be seen that 2.5 % strain would produce a sealant stress of approximately 5 psi (34 kPa). In the fully racked con­dition, at 10 % movement in the sealant, approximately 35 psi (241 kPa) stress in the sealant is generated, when taken in a worst case scenario assuming tensile behavior of the sealant. In comparing the sway condition to the rack condition at this 2 in, (51 mm) displacement, the stress on the sealant can be expected to be approximately seven times greater in the rack condition, up to 35 psi (241 kPa) in tension under racking.

In actuality, the stress generated falls somewhere between the shear and tensile conditions, and the shear stress at 2 in, (51 mm) of racking is only 7.5 psi (52 kPa). The sealant, as evaluated in either tension or shear, is well within its capability anywhere from the 7.5 psi (52 kPa) to 35 psi (241 kPa) bounded condi­tions given an ultimate strength of 133-149 psi (917-1027 kPa) in shear and ten­sion, for the two-part sealant, respectively. In looking at the rack condition at the design displacement level of 3 in, (76 mm), the strain in the sealant is approximately 15% (Fig. 20), resulting in a sealant stress of 15 psi (103 kPa) in shear and 45psi (310kPa) in tension (Fig. 12), again well within the sealant capability for this most stringent mode of testing. This analysis, summarized in Table 5, compares actual strains generated in the sealant per racking testing and calculates stresses in the sealant based on coupon testing (Figs. 11-14), which was carried out according to industry standard test procedures.

TABLE 5—Summary of mock-up displacements, sealant strains, and associated stresses.

Boundary

Condition

Displacement

(mm)

Actual Strain (per video analysis)

Correlated Sealant Stress Range Shear to Tensile Stress Range, psi (kPa)

Sway

1 (25)

1.8%

1-3 (7-20)

Sway

2(51)

2.5%

2-5 (14-34)

Sway

3 (Afallout) (76)

4.0%

5-15 (34-103)

Rack

1 (25)

5.0%

7-17(48-117)

Rack

2(51)

10.0%

7.5-35 (52-241)

Rack

3 (Afallout) (76)

15.0%

15-45 (103-310)

FE model

1 (25)

15.0%

15-45 (103-310)

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.

It is understood that the strain rate during an actual seismic event is faster than the industry standard coupon test method pull rate of 0.5 in, (13 mm)/min. In fact, the equivalent strain rates experienced by the sealant beads on the mock-ups ranged from 2 in, (51mm)/min up to 28 in, (711mm)/min. These strain rates are a function of the test frequency (Fig. 2) and the boundary condi­tions (Fig. 8). As stress gauges cannot be installed in the sealant joint to take measurements during the racking testing, these stresses can only be estimated using FE modeling or stress strain graphs from the sealant coupon testing. The estimated stresses can then be validated by observing actual sealant perform­ance and behavior on the mock-up after racking movement is induced.

The hysteresis curves, generated from the load sensor on the test apparatus, documented in Figs. 22 and 23, show very tight elastic behavior in the mock-up all the way up to a displacement of 3.00 in, (76 mm). While stresses in the seal­ant may be higher than predicted by the coupon testing, the overall perform­ance of the sealant is comparable. Because, in general, faster loading rates result in materials showing higher strengths, it is expected that if the sealant coupons were pulled in shear at a higher strain rate, the resulting stress-strain curve for “fast" shear behavior would be steeper and probably closer to the ten­sile test curves presented here. Coupon shear testing at a higher strain rate would be a more direct correlation to sealant behavior in a racking testing and can be considered for future studies.

Finally, the FE modeling results show a more rapid increase in sealant elon­gation as the mock-up displacement is increased. Because this FE model is linear-elastic, it cannot accurately model the “softening" of the sealant with

FIG. 22—Hysteresis curve racking boundary condition 3 (BC3).

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

increased displacement. The “softening" of the sealant (reduction in modulus) with increase strain/stress is clearly illustrated in the hysteresis curve in Fig. 22. As the mock-up is cycled through larger displacements, the force required to reach a given displacement decreases. This “softening" behavior is also somewhat evident in the coupon test results (Figs. 13 and 14). Up to about 1 in, (25 mm) of mock-up displacement, the FE results follow the tested results for boundary con­dition 3. As displacements get larger, the FE results overestimate the actual seal­ant elongation as seen in the mock-up testing. This further documents that the FE modeling is a conservative approximation of the elongations in the sealant.

Summary and Conclusions

This study has provided new full-scale experimental data for the performance of a unitized four-sided SSG curtain wall system including the effects of a corner condition. The study has developed racking test data on continuous silicone sealant behavior and provided comparison with coupon test results and FE modeling results. When curtain walls are attached in such a manner that they are allowed to sway, as do fully unitized systems, the stress the sealant experien­ces at the allowable drift ratios per ASCE 7 are very minimal (well below 20 psi (138 kPa)).When curtain wall systems are required to rack, emulating stick-built conditions, sealant stress will be higher and sealant damage may occur at rack­ing displacements beyond the seismic design drift level. However, at design dis­placement drift levels in a fully racking system, there should be no sealant

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.

damage in a properly designed curtain wall system based on the results of this study. Also, the corner condition was shown not to be a limiting factor under racking movement. This is a key finding of this research because typical AAMA 501.6 testing does not require the inclusion of out-of-plane corner conditions.

This paper discusses the history of sealant design level stresses and sup­ports the conclusions made in 1996, that 50psi (345 kPa) is a reasonable design stress level for seismic events. The racking testing showed that the sealants can cycle up and down (in tension) to 50 psi (345 kPa) without adversely affecting the ultimate properties of the sealant. Based on coupon test results, this study also showed that the ultimate strength level of the sealants was only minimally affected by environmental conditions, including extreme temperatures and 5000 hr of QUV exposure. The silicone sealants are modulus stable across these environmental conditions. Also, by comparing tensile and shear stress values, this study shows that predicting curtain wall performance using tensile sealant design values is conservative when considering seismic-induced movement.

Silicone offers a flexible anchoring method when used on all four sides and, through curtain wall system racking testing, it has been shown to perform satisfac­torily at high racking displacement, with acceptable damage levels that leaves the system still operable, as required by the building code. Coupon sealant test results on individual sealant samples show the long-term durability of silicones, as well as their suitability for use in seismic regions based on how they cyclically move and return without damage (sealant remains in an elastic deformation zone).

Finally, this study documents that linear elastic FE modeling is a valid way to conservatively predict the sealant stresses and system behavior, up to allow­able drift levels, prior to implementing full physical racking tests (AAMA 501.4 or 501.6). By analyzing a FE model together with static racking testing (AAMA 501.4), system suitability for seismic areas can be acceptably predicted. Dynamic racking tests (AAMA 501.6) are very useful to new system technologies or new material verification and projects involving essential service structures.

Shear testing of sealant coupon samples at varying strain rates would pro­duce stress-strain curves that would be meaningful and more readily comparable to actual seismic testing results. This should be considered for future work. Sam­ples could be traditional “H piece" configurations to represent a sealant joint, or could be lap shear joints, or both to compare the effects of sealant joint configura­tion. Cyclic testing could be repeated at a higher number of cycles using the cho­sen joint design and a higher strain rate to more closely mimic actual seismic testing instead of correlating results through tensile testing. Furthermore, for follow-up studies, it is recommended that nonlinear FE modeling packages be used to more accurately predict the sealant behavior at higher strain levels.