Volume 42, Issue 10 - October 2007
Guide to Glass
Tracking Down Glass Baluster Failure
By William J. Nugent and Mark K. Schmidt
The use of glass balusters to support handrails and fulfill other code mandated guard requirements in building structures has been common for more than 20 years. For indoor applications, glass balusters typically are tempered glass panels that support a continuous wood or metal handrail at their top edge and, in turn, are continuously supported along their bottom edge in a metal channel. Usually, the glass is fixed firmly in the channel with a Portland cement-, epoxy-, or gypsum-based grout.
We recently investigated extensive and costly failures of glass balusters in two outdoor stadiums. Based on these investigations, it is clear that the type of glass product, the structural design approach and the support detailing that have been used successfully for years in indoor glass baluster systems may require modifications for adaptation in outdoor stadiums where maintenance (e.g., high-pressure washing) and moisture and temperature cycling create a different and more demanding environment.
In both cases we investigated, the balusters were nominal 14-mm thick laminated glass panels, each comprised of two sheets of 6-mm thick heat-strengthened glass laminated to a 1.5-mm thick polyvinyl butyral (PVB) interlayer. The typical panels measured about 750 mm high by 1,500 mm long and were grouted continuously at their base into aluminum channels. Within 12 months of installation, multiple glass panels had either cracked, become delaminated or both in both stadiums.
Site observations and documentation, field and laboratory testing and analyses were performed to determine the cause of the cracking and delamination and to understand the structural behavior of the balusters.
Delaminations also were occurring in the baluster panels, with one or both glass lites separating partially or completely from the interlayer. The delaminations resulted in clouded or hazed areas and typically were located at the bottom corners of glass panels at the grout line.
Field examination also revealed a number of locations where the silicone sealant at the top of the grouted panel base was in poor condition. Instances of missing sealant, sealant exhibiting adhesion failure at glass or aluminum surfaces and sealant that was bubbled or blistered were noted.At some locations, the grout was found to be soft. At other locations, the grout was relatively firm. Cracked and delaminated baluster panels were located almost exclusively where the grout was firm. In many of these cases, the sealant was also missing or in poor condition. Where the grout was soft, no visible cracks or delaminations were noted.
Based on field observations, several sections from each of the two glass baluster systems were removed for detailed laboratory examination and testing. These sample sections included glass panels that were cracked with no delaminations, glass panels that were delaminated with no cracks, glass panels that were both delaminated and cracked and glass panels that contained no visible cracks or delaminations.
A number of additional grout samples also were removed for laboratory analysis. Samples were selected from areas where the sealant above the grout was either in good or poor condition. Additionally, samples were taken both where the grout was soft and where it was firm.
Grout Pocket Examination
Examination of the grout pocket in samples that contained the relatively firm grout revealed the presence of a white deposit between the aluminum and the grout. Grout pockets that contained relatively soft grout showed little evidence of the white deposit.
Chemical analysis revealed that the white deposit was aluminum hydroxide—a by-product of aluminum corrosion. The thickness of the deposits ranged from 0.75 mm to 1.5 mm. The white deposits occurred primarily where the grout was in contact with the aluminum. At locations where the channel had been filled with grout incompletely and where there was no grout in contact with the aluminum, very little evidence of corrosion was noted.
In the confined space of the base channel, expansive pressures were created within the channel by the growth of corrosion by-product. The origins of the glass panel fractures examined in the laboratory coincided with locations where intermittent contact of the grout and corrosion by-product created non-uniform pressures along the bottom edge of the glass. In panels that were not cracked, the corrosion by-product was either not present, or the grout and corrosion by-product provided relative uniform contact/support along the glass edge. Thus, pressures exerted on the bottom edge of the glass were more uniform.
Samples of the grout were analyzed using x-ray diffraction to identify their crystalline components. The analysis indicated that at least two different types of grout were used. Portland cement-based grout had a firm consistency. Gypsum-based grout was found to be soft.
The alkalinity of each grout type was evaluated by measuring its pH. The pH of the Portland cement-based grouts was typically above 12.5 (highly alkaline). The pH of the gypsum-based grout samples ranged from 9 to 12 (more neutral).All the glass panel samples with delaminations at the bottom corners were determined to have Portland cement-based grout. Upon removal of the base channel in the laboratory, all panels set in Portland cement-based grout were found to contain further delaminations typically extending 10 to 25 mm inward from the edges that were in contact with the grout. Panels set in the gypsum-based grout did not exhibit visible delaminations in the field and, when the bottom edges were exposed in the laboratory, generally exhibited very limited delamination along the edges that were in contact with the grout.
A preliminary analysis for the stadium installations in question showed that the governing condition was created by the concentrated load acting horizontally (perpendicular to the plane of the baluster panels) at a top free corner. The analysis also showed that, depending on assumptions, the bending stresses created at the base of the glass by the governing loads could be greater than the allowable stress.In light of this analysis, load tests were performed to gain an understanding of the structural behavior of the laminated glass balusters and determine their fitness for purpose.
Baluster panels of the same type, size and thickness as those in the stadiums were set up in the laboratory with the same aluminum base channel, grout and other details used in the stadiums. A loading apparatus capable of applying either a uniform horizontal line load or a concentrated horizontal load to the top edge of the glass was employed.
The loading apparatus also had the capability of applying the load as an impulse (on and off in approximately 3 seconds) or for a sustained period (20 minutes or more). Prior to testing, instruments were applied so that bending strains at the base and horizontal deflections at the top could be measured. A series of proof load tests performed in accordance with IBC criteria showed that single glass balusters could support more than two times the code-specified uniform and centrally located concentrated loads without failure or permanent deformation. For concentrated loads applied at panel corners, similar test results were achieved by simulating the in-situ handrails, which allowed adjacent glass panels to contribute in resisting the applied load. The limited testing also provided data about the degree of composite behavior in the panels and the manner in which stresses created by concentrated loads applied at the top edge are distributed across the width of the panel base.
Table 1 is a summary of the data on composite behavior. For the impulse-type loadings (on and off in approximately 3 seconds), the laminated panels generally behaved as if they were monolithic. For longer duration loads (20 minutes or more), the degree of composite behavior averaged approximately 83 percent.
Based on the distribution of bending stresses across the width of the panel base under application of the concentrated load at the top center and at the top corner, respectively, the maximum glass tensile stress measured during testing can be calculated using an effective base width determined by a load distribution angle of approximately 15 degrees. The corresponding section modulus of the laminated glass panel should be calculated using the sum of the minimum glass thicknesses and the interlayer thickness.We concluded that both the cracking and the delamination of the laminated glass baluster panels was caused by the combination of moisture and a highly alkaline Portland cement-based grout in the glass panel base support channels. These conditions fostered corrosion of the aluminum, creating nonuniform expansive pressures that led to the glass fractures. The highly alkaline grout also attacked the PVB interlay resulting in visible delaminations.
Table 1. Summary of Composite Behavior Data
This data summarizes composite behavior. For impulse-type loadings (on and off in approximately 3 seconds), laminated panels behaved as if they were monolithic. For longer duration loads (20 minutes or more), the degree of composite behavior averaged approximately 83 percent.
William J. Nugent is president and principal, and Mark K. Schmidt is unit manager and senior consultant of Wiss, Janney, Eistner Associates Inc., in Northbrook, Ill., which focuses on solving structural, architectural, and construction materials problems. This article is adapted from a presentation made at Glass Processing Days 2007 in Tampere, Finland.
For an expanded version of this article, click here.