• Means of escape is not affected by smoke and/or heat from the fire;
• The structure survives the burnout of a realistic, worst case fire and significance collapse is prevented;
• The fire is contained within the compartment of its origin; and
• Fire fighting intervention can be carried out safely.

1. The prediction of the maximum fire temperatures and duration using the concept of natural fire, which considers the behaviour of real fire. It takes into account the fire load density, ventilation, compartment geometry and properties of enclosure. When considering the fire scenarios, an realistic office fire load density of office was adopted as the base value. Even though an automatic sprinkler system would be installed, its operation was not account for when determining the maximum fire size. It was found that ventilation is the dominating factor when assessing the natural fire curves. An initial study on the behaviour of the glazed façade under fire condition was conducted. It was demonstrated that the glazing would have failed much earlier than the structural steel elements and therefore additional ventilation can be introduced as a result. Further consideration was also given to possible inner rooms within the office which prevent substantial ventilation being introduced. The fire temperatures and duration can be calculated for the entire fire period for each ventilation case. A typical time temperature history plot can be found in Fig. 5. A sensitivity study was also conducted varying parameters such as the fire load density, wall material and fire compartment geometry. The adoption of natural fire temperatures is a departure from conventional design but are considered more realistic in the performance based design compared to a BS476 Standard Fire Curve. More importantly, it demonstrates that the building is being able to survive a full burnout of its contents without collapse. The real fire temperature curves form the basis of the time line plot.
   
2. The 3-dimensional structural fire performance was studied using a finite element program Vulcan developed at the University of Sheffield. It is a non-linear finite element program which has been validated against results from large scale fire tests. Initial studies were conducted on small subframe utilising symmetry and Fig. 2 shows the deflected shape of the subframe at various heating stages. For the purpose of this study, the subframe analysis was heated to a BS476 Standard Fire Curve to enable appropriate comparison to be made in the sensitivity study conducted at a later stage.

Figure 2: Vulcan Model: Small Subframe Analysis after 30 Minutes (Left) and 120 Minutes (Right) ©FEDRA

The results from one of the analyses are plotted in Fig. 3 showing the vertical deflections and internal forces in accordance to BS476 standard fire curve heating. The analysis assumes all the structural steel elements are fully protected except the intermediate beams (beams not connected to columns). Such a protection profile is not new and the Steel Construction Institute and BRE have published relevant documents on the concept 1. This is enabled by the enhanced composite slab performance in fire primarily due to membrane action.

Figure 3: Results from Small Subframe Analysis click to see an enlarged figure ©FEDRA

The results of the subframe analysis show that some membrane action formed in the early stage when the slab is heated and a compression ring formed around the protected primary beams. Due to the relatively large aspect ratio of the assumed grid, the slab enhancement delivered the membrane action is limited and the unprotected intermediate beams are in tension as early as the 32nd minute. The primary beams remain in compression until the 90th minute where the majority of the beams are in tension and supported by catenary action. The catenary forces in the protected primary beams peak at approximately 120th minute.

The main objective of the analysis was to investigate the optimum fire protection profile to achieve the performance criteria set out. Sensitivity studies were conducted using the small subframes, varying the protection thickness, heating profile and sizes of structural elements (reinforcement mesh, concrete grade, beam sizes etc.). Once the preferred solutions were identified, a full-frame model was created comprising a quarter of the entire floor slab. The deflected shape of the model produced is shown in Fig. 4. The displacements and forces were studied carefully to ensure the performance criteria were achieved. For example, the vertical displacements of the beam and slab were checked against their limit particularly where vertical compartmentation exists. Using this approach ensures the overall structural stability including the interactions between beams, columns and their connections. For instance the additional catenary action experienced by the primary beams in fire was checked against the column strength the connection design so that there is sufficient strength and ductility for energy absorption thereby reducing the risk of progressive collapse.

Figuer 4: Vulcan Model – (top) widespread fire across the entire floor, (bottom) small compartment fire ©FEDRA

As a result of the studies, a best value solution was reached by increasing the performance of the slab in exchange for the omission of fire protection for the intermediate beams. Where passive fire protection is required, the material and thicknesses were carefully selected to give the required performance in fire, with sufficient robustness over time in service.
The solution increases the inherent fire resistance of the superstructure, relies less on the passive fire protection, and therefore is less susceptible to damage or failure (due to less maintenance requirement). In other words, it is a more robust solution. The additional benefit is the saving generated from the omission of passive fire protection and the associated time savings in construction stage.

Results

Fig. 5 shows one of the integrated plots that considers the elements discussed above. Three key events are identified that represent three different scenarios and their consequences.

Figuer 5: Performance based approach using time line ©FEDRA

	The first scenarios assumes the automatic sprinklers are fully operational and the fire size is successfully controlled. As a result, flashover does not take place and structural deformation is not incurred.
	This scenario assumes the automatic sprinkler system fails and a fully developed fire takes place. The fire temperature curve shown in Fig. 5 was calculated based on a typical ventilation condition. The maximum fire temperature achieved is approximately 1300°C and the heating phase lasts for 50 minutes. The fire will eventually burnout and the vertical displacement for the beams at the time would be approximately 350mm. These deflections are not shown in the plot.
	Further analyses were also conducted using the BS476 Standard Fire Curve. This was to provide additional comfort in the design and so that comparisons could be drawn with a solution that would be derived using a more traditional prescriptive method. The predicted beam displacements shown in Fig. 5 demonstrate a fire resistance period of 130 minutes. This can also be seen as a factor of safety against the unlikely event where the heating phase continues