Innovative design tool for the optimization of blast-enhanced facade systems

1Introduction

Threats to building occupants due to blast loads generally result from a combination of blast overpressure injuries due to a breach of the facade system and glass fragmentation injuries. As witnessed in the Oklahoma City bombing in 1995, approximately 40% of laceration and abrasion injuries were attributed to glass fragmentation (Norville et al., 1999). Current blast design related practices commonly refer to achieving a balanced design that is essentially focused on strength limit states whereby in a structural system subjected to blast, no element is appreciably weaker than the element that it supports, thus minimizing the risk of progressive collapse (Hinman, 2011). From the perspective of both structural integrity and human safety, a true balanced design requires a combination of both energy dissipation and glass fragment retention (Zobec et al., 2012). The Polyvinyl Butyral (PVB) interlayer used in laminated glass exhibits substantial fragment retention and post-elastic viscous deformation capacity. Given the combination of these two behavioural effects, laminated glass is the preferred glazing solution for blast-enhanced facades.

Figure 1 describes the various glass hazard classification levels according to ISO 16933:2007(2007). As the glazing is ruptured by the blast shock wave, numerous glazing fragments are projected into the occupied room. Limits for Low and High hazard regions have been defined according to the kinetic energy of the glass shards and most frequently occurring physiological damage thresholds. Eye injuries are presumed to occur at energy levels as low as 0.06 J/mm² or shard velocities of approximately 2 m/s (Low Hazard). Similarly, the potential for skin tissue penetration is presumed to occur at glass fragment kinetic energy levels of around 0.1 J/mm² or approximately at shard velocities of 9 m/s (Marchand et al., 2006), although this value has been found to vary significantly between annealed and toughened glass tests. In blast resistant facade design, the fragmentation hazards or spall is often the limiting factor, whilst the viscous-elastic energy dissipation can be maximized by the use of the PVB laminates. As a result, second generation blast-enhanced facades are usually referred to as being dissipative due to beneficial effects in terms of mitigating impulse transfer, but whilst it is obvious that hazard protection is still the dominant blast performance requirement, conformance to this criterion may in certain circumstances be detrimental for design optimization. None the less, performance criteria definitions are required in terms of impulse and energy reduction as building owners and occupants are increasingly requesting that the facade system is designed such that in the event of a blast the magnitude of peak reactions transferred to the main structure are minimised.

Currently, there is no consensus on the ultimate failure mode of cracked laminated glass due to large displacements. WINGARD 5.5 considers the primary failure mechanism for laminated glass to occur when the lateral in-plane edge displacement of the glass membrane exceeds twice the framing rebate depth based on arena tests of non-structurally glazed windows. Ellis (1991) and Ellis & Beak (1992) undertook water-bag testing of laminated glass panes to failure and reported the failure displacements and modes. The most common failure mechanism was cutting of the PVB membrane at the glass-frame edge interface due to excessive end rotations (Morison, 1999), generally dependent on the short span of the laminated glass and of the total thickness of the interlayer.

To take into account the sensitivity of the glass strength to the strain rate, Norvillle & Conrath (2001) have proposed a method to equate blast loads with an equivalent three second wind gust, enabling glazing make-up selection based on the glazing design standard ASTM E 1300–12 (2012). The inclusion of monolithic and laminated glass into a single method assumes that glass sizing is based on surviveability up to or just at the point of cracking (Morison, 2007). The method does not take into account large PVB deformations and energy dissipation. Thus, the design approach is overly conservative and uneconomical. The design of individual facade elements based on a ‘safe’ equivalent static load is straight forward and allows the analysis to be carried out rapidly. However, this approach is not capable of taking into account the dynamic interaction between the components of the facade system. Dynamic FEM analysis is the most accurate analysis tool but it often requires long computation time. The reliability of results is also limited to the accuracy of the input parameters. Under dynamic and impulsive loads, materials used in facades exhibit marked strain-rate dependent behaviour that must be considered and accurately modelled.

Equivalent SDOF methods have been used to analyse glass plates subjected to blast since its inception by Newmark (1956). The method requires two important parameters: the mass transformation factor and the resistance function. An accurate displaced shape is required to estimate the mass transformation factor, whilst the resistance function is dependent on stiffness parameters and the ultimate failure displacement. For a true balanced design that considers both resistance and building occupant safety, a rapid design tool capable of incorporating a dynamic analysis as well as accurately assessing the projection of glass fragments is required. The Permasteelisa Group has recently developed a glass fragmentation hazard tool used in conjunction with a FEM code, in order to improve the design optimization process as shown in Fig. 2.

This paper initially outlines the distinctive background of the analysis method used with current approaches. Secondly, the results of original research outcomes derived from testing of laminated glasses of various thicknesses, treatments and aspect ratios and critical parameters affecting performance are established. The findings include equivalent interlayer post elastic properties, transformation factors and allowable post cracked laminated glass deflection limits. During dynamic analysis and subsequent hazard level assessment, the glass fragment size is identified as the key parameter affecting the results of the tests. This parameter also has a significant influence during the various stages of the design process. Finally, an original fragmentation model, implemented in the analysis tool is presented.

2SDOF approach and MDOF models

3Experimental investigations into the post-cracking behaviour of laminated glass

Permasteelisa, in collaboration with the University of Technology of Sydney and The University of Melbourne, undertook an extensive water-bag testing program (Fig. 5). A statistically broad number of samples of various glass thicknesses and heat treatments have been considered in order to derive relative failure displacement values. Furthermore, the deflected shape of the glass was progressively mapped at various displacement intervals in order to derive the non-linear shape as well as the stiffening effects of glass fragmentation. Non-linearity effects as a result of non-uniform water pressure distribution due to the varying water head; g, over the glass surface were identified and corrected using a numerical FEM calibration procedure (Lumantarna et al., 2012). Current methods to derive the transformation factors in the post-cracked phase assume the PVB to behave as a pure membrane with no effects of glass fragmentation on shape.

4The dissipative role of the facade

Knowledge of the post cracking behaviour of laminated glass is important in order to design a facade that performs as required in terms of maximizing hazard protection and blast energy dissipation. Whilst maximizing hazard protection (which will be discussed in the subsequent chapter of this paper) is well understood and protection strategies and requirements have been defined by various statutory organizations over the past twenty years, blast energy dissipation is not yet clearly defined.

Although second generation blast-enhanced facades are currently being optimized emphasizing blast energy dissipation, clear guidelines on how to best evaluate dissipative effects do not yet exist. By contrast, building owners are currently defining in project specifications specific performance requirements in terms of reaction reduction, strictly related to actual specific limitations of the main structure in its ability to transfer facade loads. As a result, there is a real need for clarification and design criteria should be specified more effectively in order to classify the dissipative performance of blast-enhanced facades.

The initial question that needs to be addressed, is whether or not characterizing the facade in general terms, as being ‘dissipative’ is an appropriate term? How much quantifiable energy does the facade need to dissipate?.

It is true that a significant proportion of blast energy is dissipated through plastic and viscous-elastic deformations of the facade components as well as damping effects, but the most important role performed by the facade is the ‘postponement’ of energy transfer or rather effectively distributing the blast impulse over a longer time duration. This effect is particularly beneficial with regard to load transfer to the main structure. As a result each mechanical component of the facade is subject to less severe loading when the same impulse can be absorbed over a longer time (as can be evidenced by the iso-damage curves behaviour). Inertial effects play an important role in reducing the reaction transfer at the initial facade response phase, whilst the plastic/visco-elastic deformations are the primary factors that result in specific force reductions during the secondary phase, when the inertial effects commence to be detrimental.

In general the facade can be seen as a system capable of transforming the blast input duration and peak pressure over a different combination of the two parameters, but for the majority of cases without significant changes to their product (impulse). For this purpose Fig. 9 describes the effect of transfer from the Blast Load (BL) to an Equivalent Reaction Load (ERL) of a lower iso-damage level.

For instance, Fig. 10 and Fig. 11 describe the reaction time histories transferred to the structure for two different glazing configurations as well as differences in the complexity of the facade analytical model subjected to blast load condition, GSA D. The difference between the two glazing configurations remains in the use of annealed or tempered glass. The first three analytical models are (Multiple-Degree-of Freedom) MDOF whose component deformability is (1) only glass is deformable, (2) both glass and supporting mullions are deformable and (3) glass, mullions and bracket (dissipative) are deformable. Model 3 involves a bespoke two-step modelling procedure, combining the MDOF glazing reaction history analysis and subsequently using this load file to model the facade framing members behaviour using FEA.

Tables 1 and 2 summarize the results in terms of two indices:

The use of annealed glass is interesting to note as the impulse ratio remains around a unitary value, whilst the beneficial effect increases proportionally as other deformable (retardant) components are also simulated. By contrast, the use of tempered or fully toughened (FT) glass reduces the beneficial effect of the facade in terms of RMS reaction reduction in addition to a certain degree of impulse amplification. Although different outlooks are present in the literature, the authors believe that for this reason tempered or FT glass is detrimental with respect to the mitigation strategies, at least in terms of blast reaction transfer. Obviously completely different evaluations of the two types of glass behaviour can be given according to protection performance requirements.

Finally it should be noted that the RRR parameter properly expresses the beneficial effect of the dissipative role of the facade when quasi-static criteria are used for the design of the support structure (main building structure), whilst additional considerations in terms of frequency behaviour must be addressed when dynamic responses are not negligible and critical resonances could be further excited.

5An original fragmentation model

Where applied to laminated glass, current flight models for glass hazard assessment have generally proven to be overly conservative (Marchand et al., 2006). This results in an uneconomical design with regards to both physical performance and architectural aesthetics.

In response to such cost impacts as well as hazard level assessment reliability, a fragmentation tool has been developed based on the stochastic behaviour of glass, fracture mechanics principles for crack tip opening and strain energy density in order to derive the major characteristics of the fracture pattern.

The targets of the investigation are to understand the:

Two factors are required in order to define the most important features for a new hazard level assessment model. They are:

6Beyond current standards

Current standards used for hazard level assessment in commercial buildings, have origins based on common experimental military applications. Direct application to continuous facades should consider typical civilian occupancy and usage in terms of the number of persons, relative positions and distances from the facade. There should be a distinction in hazard level classification for podium facades typically characterized by large vertically spanning glass units with crowded though transient movement of persons in contrast to typical office conditions where occupants may be stationary for extended time periods.

Figure 15 represents the Isodamage curves B/S (breaks safely), interlayer rupture and high hazard according to the new fragmentation tool for a 1220 × 1520 × 6.6.4AN-glass. The figure shows that the impulsive region of the P-I curves background is the region where the ‘loss of design threat area’ is most significant. The reason is clear. In the impulsive region (constant impulse), acceleration conditions are more favourable for shard detachment from the interlayer with different levels of shard velocities possible, with the consequence of every possible outcome within different hazard levels. For quasi-static behaviour (constant pressure), the high hazard curve intersects the interlayer failure curve. This occurs because ‘no-spall’ high hazard is reached without significant acceleration conditions. In this case, shard detachments are unlikely.

The chart highlights how the ‘threat area lost’ region corresponds to the most typical Satchel bomb conditions according to ISO16933. However in this specific case a more typical vehicle bomb threat such as ISO-EXV19 lies within the transition area. It can be easily understood that applying overly restrictive limitations on fragmentation within the proximity of the facade (as per GSA, assuming spall effects) results in expensive solutions. These solutions generally exhibit stiffer behaviour with less blast energy dissipation and subsequently higher risk of overall facade structural collapse due to excessive bracket reactions to the main structure. Similarly, structural collapse risk is greater due to non-optimisation of facade blast energy dissipation capabilities.

As shown in Fig. 16, under the blast load ISO-EXV19, the single glass 6.6.4AN would be inadequate as it results in high hazard levels. In order to maintain the glass performance within the hazard region with a high degree of confidence, a 10.10.4AN is preferable. In terms of reactions to the main structure through the facade connection bracket, this would result in an approximate 50% increase in reactions assuming rigid quasi-static behaviour of the bracket system.

Other benefits of the fragmentation model would be to create a new perception of probable hazard levels in order to improve design cost effectiveness without loss of safety. Estimation of fragment size and velocity is also an important tool for injury assessment once the specific occupant working (or living) conditions within the vicinity of the facade are defined. Given that blast effects on facades are not uniform and high pressure-impulse loads are generally localized, further significant risks of structural collapse can also be found by balancing enhanced levels of facade energy dissipation with acceptable injury probability.

7Conclusions

In this paper, the features of the design optimization process assessed by a numerical tool developed by the Permasteelisa Group have been presented. Experimental testing to determine the post-cracking behaviour, failure modes and limits of laminated glass shows that the PVB membrane provides significant levels of protection against breach of the blast wave and hence overpressure injuries. The limiting factor though for the design of blast-enhanced glass is the risk of injury due to glass fragmentation. Current models are overly conservative, resulting in not only oversized glass but also higher reactions to the main structure with less possibility of effectively utilizing the energy dissipation potential of the facade. A new fragmentation modelling approach assuming Griffith energy balance criteria shows excellent correlation with blast arena tests. A design balance between PVB membrane dissipation le-vels and acceptable glass fragmentation hazards is required for both sufficiently safe yet economical design of blast-enhanced glazing.