Window ejecting plume is a common phenomenon associated with building fire hazards. The high temperature and the intensive radiant heat flux generated by the ejecting flame are the major factors inducing the fire to spread to the upper levels through the building façade. To mitigate or prevent such fire spreading, non-combustible vertical spandrel and horizontal projections are included in buildings with traditional façades. However, those fire-prevention features are seldom employed in modern buildings with glass façades. Considering that glass façades of both the single-skin and double-skin varieties are gaining popularity in commercial high-rise buildings, and that we currently lack the ability to deploy firefighting agents to such ultra-high structures, the need for a better understanding of ejecting plume behavior is critical as it would enable us to improve the firefighting systems in modern buildings.

The behavior of ejecting plumes in both single- and double-skin façade scenarios was studied in this work, with focus on ejecting plume trajectory, ejecting flame height, and ejecting plume temperature distribution along the plume trajectory.

The contents of the present work are divided into two parts: the first part focuses on the behavior of ejecting plumes in single-skin façade scenarios and the second part examines the behavior of ejecting plumes in the cavity of double-skin façade (DSF).

In the single-skin façade scenarios, the existing criteria for evaluating plume attachment were examined. Near wall square pool fire was used to simulate the upper part of a window plume and the burner was assumed to be installed at the position that the ejecting plume turned to the vertical direction. The existing criteria failed to predict plume attachment. As the criteria were all based on observations in compartment fires, it is believed that failure occurred because the criteria underestimated the induced pressure difference in large fire scenarios. Moreover, the heat release rate was found to have an influence on the ejecting plume trajectory.

To examine the effect of heat release rate, experiments using a reduced scale compartment-façade model were conducted. The heat release rate was found to be another factor with a strong impact on ejecting plume trajectory. Increasing heat release rates were associated with the plume approaching closer to the façade. Based on experimental observations, a temperature distribution model was proposed by direct application of continuous flame height as the characteristic length scale. However, in some cases, the vertical distance from the equilibrium point to the venting point was nearly equal to (or even larger than) the vertical distance from the equilibrium point to the attachment point, which cannot be explained by the effect of the induced pressure difference alone.

To determine the other factors accelerating the process of plume attachment after the equilibrium point, Fire Dynamic Simulator version 6 (FDS6) was employed to simulate the ejecting plume. Plume trajectories based on temperature field and velocity field were compared, and it was found that the two types of trajectories deviated from each other after the equilibrium point. In the near-wall region, the plume flowed upward nearly parallel to the wall, but the temperature trajectory kept approaching closer to the wall. The wall heat-blocking effect was identified as the main reason for the quick attachment of the plume trajectory after the equilibrium point.

To predict ejecting flame height and to examine the analogy between ejecting flame and wall fire, experiments were carried out using reduced-scale compartment models, both with and without a wall above the opening. A sharp increase of flame height was observed in cases both with and without wall, and thus the analogy between ejecting flame and wall fire is somewhat questionable. Increase of total combustion efficiency was found to be the major reason for the sharp increase observed in flame height, rather than blockage of the air entrainment rate by the wall (as would occur in a wall fire). Models were proposed to correlate the ejecting flame height and the heat release rate based on experimental observations.

The findings of part I suggest that the existing criteria for evaluating plume attachment, temperature models and flame height models all have limitations for applications to scenarios with large external fires. In such scenarios, influenced by the wall heat-blocking effect, ejecting plumes approach closer to the wall and attach to the wall at relative lower positions, which should be treated more rigorously in the design of firefighting system.

In the DSF scenarios, experiments were carried out using a reduced scale compartment-DSF model. Both cavity depth and heat release rate were varied. Increasing both the cavity depth and heat release rate led to the ejecting plume approaching closer to the interior wall; otherwise, the ejecting plume approached closer to the exterior wall. All plume trajectories with maximum horizontal position less than half the cavity depth attached to the interior wall, and vice versa. Temperature along the plume trajectory was found to depend on opening geometry and was almost independent of cavity depth. Models were proposed for prediction of plume trajectory temperature based on experimental observations.

To evaluate plume attachment, theoretical analysis and simulations were performed. An equation for estimating induced pressure difference was derived based on Bernoulli’s theory and mass conservation. Induced pressure difference changed sign when the plume centerline surpassed half of the cavity depth and drove the plume towards the exterior wall. Based on simulation results, criteria considering the effect of opening geometry, heat release rate, and cavity depth are proposed to estimate plume attachment.

The results of part II clearly indicate that the ejecting plume trajectories are determined by the coupled effect of HRR, opening geometry and cavity depth. For a given opening geometry and cavity depth, the ejecting plume trajectory can change dynamically with the variation of HRR and might attach to either façade skin. Thus, a wider protection area must be considered when the fire load inside a compartment can change substantially.