Computational Modeling of Glass Curtain Wall Systems to Support Fragility Curve Development

With the increased push towards performance-based engineering (PBE) design, there is a need to understand and design more resilient building envelopes when subjected to natural hazards. Since architectural glass curtain walls (CW) have become a popular façade type, it is important to understand how these CW systems behave under extreme loading, including the relationship between damage states and loading conditions. This study subjects 3D computational models of glass CW systems to in- and out-of-plane loading simulations, which can represent the effects of earthquake or hurricane events. The analytical results obtained were used to support fragility curve development which could aid in multi-hazard PBE design of CWs.

A 3D finite element (FE) model of a single panel CW unit was generated including explicit modeling of the CW components and component interactions such as aluminum-to-rubber constraints, rubber-to-glass and glass-to-frame contact interactions, and semi-rigid transom-mullion connections. In lieu of modeling the screws, an equivalent clamping load was applied with magnitude based on small-scale experimental test results corresponding to the required screw torque. This FE modeling approach was validated against both an in-plane racking displacement test and out-of-plane wind pressure test from the literature to show the model could capture in-plane and out-of-plane behavior effectively.

Different configurations of a one story, multi-panel CW model were generated and subjected to in- and out-of-plane simulations to understand CW behavior at a scale that is hard to test experimentally. The structural damage states the FE model could analyze included: 1) initial glass-to-frame contact; 2) glass/frame breach; 3) initial glass cracking; 4) steel anchor yielding; and 5) aluminum mullion yielding. These were linked to other non-structural damage states related to the CW's moisture, air, and thermal performance. Analytical results were converted into demand parameters corresponding to damage states using an established derivation method within the FEMA P-58 seismic fragility guidelines. Fragility curves were then generated and compared to the single panel fragility curves derived experimentally within the FEMA P-58 study. The fragility curves within the seismic guidelines were determined to be more conservative since they are based on single panel CWs. These fragility curves do not consider: the effects of multiple glass panels with varying aspect ratios; the possible component interactions/responses that may affect the extent of damages; and the continuity of the CW framing members across multiple panels.

Finally, a fragility dispersion study was completed to observe the effects of implementing the Derivation method or the Actual Demand Data method prescribed by FEMA P-58, which differ on how they account for different levels of uncertainty and dispersion in the fragility curves based on analytical results. It was concluded that an alternative fragility parameter derivation method should be implemented for fragility curves based on analytical models, since this may affect how conservative the analytically based fragility curves become at a certain probability of failure level.