Modeling Thermal and Mechanical Behaviors of Glass Facades Under Fire


Student thesis: Doctoral Thesis

View graph of relations


Awarding Institution
Award date15 Aug 2022


Glass facades are utilized extensively in buildings due to their advantageous criteria of transparency and illumination. Glass facades are also characterized by practicality, aesthetics, and economic high performance. However, glass is a brittle material and can break easily when exposed to elevated temperatures during fire. Accordingly, glass facades suffer from low fire resistance, which can lead to huge losses of life and money, as recorded in many fire accidents. For example, a fire in London’s Grenfell Tower in 2017 caused 72 deaths and several million dollars of damage. Double-skin facade (DSF), an important development in glass facades, can increase risk and damage when they fail during fire, since the gap between glass layers may provide a channel to spread the fire and smoke to other levels and buildings. Although water film is applied from the building’s sprinklers to improve the fire performance of glass facades and prevent glass failure, the down-flowing water film interaction with fire and glass is a challenging topic. Ultimately, understanding the fire performance and mechanical behavior of glass during fire is crucial to increasing fire safety. This thesis aims to investigate (1) the thermal behavior of double-skin glass facades during fire, while considering the effects of venetian blinds, (2) the glass cooling mechanism when subjected to a down-flowing water film, and (3) failure and crack evolution in glass panes under fire loads.

First, a numerical framework to simulate the thermal performance of DSF under fire conditions was proposed. The framework was based on the smoothed particle hydrodynamics (SPH) technique, and it can be used to compute numerical solutions and thus simulate thermal degradation of DSF under fire conditions. The numerical model was validated by comparing the predicted response parameters in fire-exposed DSF systems with those measured in fire experiments. The validated numerical model was then employed to derive empirical equations linking temperature with both time and location along the interior and exterior glass panes of the DSF. Additionally, numerical simulations were conducted for the same DSF configuration, but this time equipped with venetian blinds to examine their influence on fire performance in glass DSF. Results generated from these numerical simulations clearly infer that “blind tilt angle” has a significant influence on fire growth and fire spread characteristics in DSF and thus need to be accounted for in the design of DSF systems in high-rise buildings.

Second, we investigated the cooling mechanism of glass panes with down-flowing water film during fire outbreak by simulating the heat energy conservation equation using the SPH method. The meshfree nature of the SPH method allows us to predict the temperature distribution efficiently in continuous flow problems, in contrast to mesh-based methods. The proposed SPH model was validated by comparing the results from our simulation under specific conditions with experimental measurements and results from commercial software packages. The new SPH model was also utilized to simulate the effects of heat flux variation, down-flowing velocity, and thickness of water film on temperature distribution of glass during fire. The developed SPH model is sufficiently able to describe glass cooling under different conditions.

Subsequently, a new meshfree algorithmic model was presented to simulate the thermal behavior of laminated glass (3D) during fire under the effects of down-flowing water film. The proposed model was based on the finite difference method (FDM) in a meshfree scheme to solve the heat transfer equation in laminated glass while considering the effects of water film and air convection. A newly efficient algorithm was developed within the model to boost processing speed. Our FDM model outperforms the commercial software packages, specifically Autodesk CFD, as follows: (1) our model is 21 and 71 times faster than Autodesk CFD in heating and cooling scenarios, respectively; (2) the results produced by our FDM model show greater accuracy than those obtained by the commercial software packages when compared with experimental results; and (3) the number of nodes (particles) used in our model is more than the number used in Autodesk CFD by about 2 times, meaning we can simulate heat transfer in larger-scale problems in a faster and more efficient manner. The FDM model was well validated by experiments, and a detailed comparison study between our FDM model and Autodesk CFD was conducted. Next, a new parametric study was established to investigate the effects of PVB thickness and water film release time (WFRT) on the cooling behavior of laminated glass during fire.

In turn, a phase-field thermomechanical modeling framework was constructed for predicting the complicated behaviors of thermal cracking in glass panes under fire. We incorporated the proposed mathematical model, which calculates the exact deformation of the mesh elements, into the variational phase-field model to simulate thermal fracture behavior in glass panes in an effective manner. The developed model improves upon previous attempts to predict thermal cracking in the following ways: (1) in a major departure from the classical phase-field simulation of thermomechanical fracture, crack evolution can be predicted using only temperature distributions—the phase-field formulations are kept fixed to overcome mesh dependency and convergency; (2) the new modeling framework directly transforms temperature variations into thermal strains (rate of loading) using fewer mesh elements and a larger time step, thus substantially reducing the computational effort; and (3) the proposed model can simultaneously predict multiple cracks distributed in any arbitrary space in the glass panes more realistically than the previous numerical models, regardless of glass pane type and size, fixation method, or thermal loading variation. The proposed coupling model was validated through comparisons against experimental observations and ANSYS software simulations. Moreover, the validated model was used for the first time to examine the effect of influential real engineering conditions, namely the heating rate, glass pane size ratio under non-uniform thermal loading, and glass pane fixation with a frame on three sides, on thermal cracking behavior.

Finally, a new algorithm based on peridynamic modeling was proposed to simulate crack evolution in homogenous and heterogeneous materials in the most efficient manner. The main purpose of constructing the super-fast peridynamic (SFPD) algorithm in this study was to simulate the thermal cracks in glass more efficiently, which can then be extended in the future to study the cracking behavior in laminated glass. The proposed SFPD algorithm surpasses the previous models and algorithms as follows: (1) the SFPD is hundreds of times faster than the regular peridynamic approach for the same computational task, where every particle interacts with up to 28 neighboring particles; (2) peridynamic convergency is dramatically enhanced (despite using fewer particles) by adopting a new strategy of time step calculation; (3) the SFPD algorithm can simulate the cracks in a wide range of scales, including the subatomic scale; and (4) the prediction of crack velocity using the SFPD algorithm is far superior to other models’ predictions when compared with experimental measurements. The philosophy behind our SFPD algorithm is to decrease the computational time needed at every simulation step rather than increasing the time step, since the peridynamic stability and convergence depend primarily on a small time step. Various problems of crack initiation and propagation were investigated using our algorithm in both homogenous and heterogeneous materials under different boundary conditions. The newly proposed algorithm was validated, and a computational cost (CC) comparison was conducted between our SFPD algorithm and other numerical studies to prove its high computational efficiency. The SFPD algorithm was then employed to simulate the cracking in glass during fire.