Installation of exposed fire sprinklers may cause inconvenience in areas where architectural and interior presentation is significant. In order to overcome this inconvenience, recessed and concealed sprinklers were created and are applied widely. Response Time Index (RTI) and C-factor are the thermal sensitivity (intrinsic parameters) used to characterise a sprinkler. They are also used as input parameters in computer fire models to simulate sprinkler response time. However, the RTI and C-factor are not published by the manufactures. Therefore the RTI and C-factor of the recessed and concealed sprinklers have been analysed and determined in this research. In order to obtain the RTI of the recessed and concealed sprinklers, four of the most commonly used sprinkler models (two recessed and two concealed) in New Zealand have been investigated in plunge test experiment by using a wind tunnel in this research. The UC3 wind tunnel used to conduct the plunge test has been fabricated in this research. This work has demonstrated that the UC3 wind tunnel could provide a very stable and uniform temperature profile in the test section. However, the velocity uniformity of the tunnel needs to be improved in the future. The "apparent" RTI for different recessed and concealed sprinkler models (two recessed and two concealed) have been determined in the plunge test experiment. It should be noted that the "final calculated RTI" for each tested recessed and concealed sprinklers has been denoted as "apparent RTI" in this study. BRANZFIRE computer model has been used to model the fire scenarios in the full scale fire tests conducted by Bill and Heskestad (1995). The best input fire object location, the best input sprinkler distance below the ceiling and the input "apparent C-factor" in BRANZFIRE for the flush, recessed, concealed and the recessed sidewall sprinklers have been determined in this research. This work has generally improved the guidance available to fire safety engineers for the RTI and C-factor of the recessed and concealed sprinklers.

Prescriptive fire safety codes regulate the use of combustible room linings to reduce fire risk. These regulations are based on classification systems which designate materials according to their relative hazard when exposed to a standard fire scenario. However, no quantitative data sets on the fire risk of wooden lining materials exist which take into account relevant uncertainties, such as movable fuel loads in compartments. This work is a comparative risk analysis on the influence of wooden linings on the time to flashover in a compartment, considering uncertainties in the fuel load configuration. A risk model is set up for this purpose using B-RISK, a probabilistic fire design and research tool currently under development at BRANZ (Building Research Association of New Zealand) and the University of Canterbury. The risk model calculates fire spread in a compartment between fuel load items and from fuel load items to combustible linings. Multiple iterations are performed considering varying fuel load arrangements and input values sampled from distributions (Monte-Carlo simulation). The functionality and applicability of the risk model is demonstrated, comparing the model with experiments from the literature. The model assumptions are described in detail. Some of the model inputs are defined as distributions in order to account for uncertainty. Parametric studies are conducted in order to analyse the sensitivity of the results to input parameters which cannot be described as distributions. Probabilistic times to flashover are presented and discussed for an ISO 9705 compartment considering varying movable fuel loads and different lining configurations. The fuel load is typical for a hotel room occupancy. Effects of suppression measures are not considered. It is shown that flashover occurs approximately 60 seconds earlier if walls and ceiling are lined with wooden materials than if all linings are non-combustible. This value refers to the 5th percentiles of the time to flashover, i.e. in 5% of the cases flashover has occurred and in 95% of the cases flashover has not (yet) occurred. Referring to 50th percentiles (median values), the difference is approximately 180 seconds. Furthermore it is shown that with wooden wall and ceiling linings in approximately 95% of the iterations flashover occurs, whereas with non-combustible linings 86% of the iterations lead to flashover. After 900 seconds, in 90% of the iterations flashover occurs if walls and ceiling are lined with wooden materials, and in 77% of the iterations if the linings are non-combustible. Using different wooden lining materials (non-fire retardant plywood, fire retardant plywood, and MDF) has no significant effect on the probabilistic times to flashover. Varying the fuel load energy density has an influence only when all linings are non-combustible and when the fuel load energy density is relatively low (100–200 MJ/m2). This work contains recommendations regarding the further development of B-RISK, the research into the fire risk connected with wooden room linings, and suggestions regarding the further development of prescriptive fire safety codes.

This research report examines the use of zone modelling compared with CFD modelling to determine when zone model approximation is valid and when a CFD model might be required. A series of computer simulations with enclosures and fires of various sizes was performed to compare the capabilities and limitations of the two computer methods. The relationship between the size of the enclosure space and the size of the fire has been demonstrated in a dimensionless form. The zone model BRANZFIRE and the CFD model FDS were used for simulating smoke development. The simulations included various full-scale experimental data on both small and large spaces found in the literature. Further simulations of large exemplar spaces with a range of fire sizes were performed to investigate different variables, which have not been examined in full-scale experiments. The simulation results have been compared based on the smoke layer height and the average layer temperature. Zukoski’s smoke filling equation was also used to compare the layer height predictions against BRANZFIRE and FDS. It was found that different data reduction techniques gave different approximations to the layer height. A perfect match between the experimental data and the model output was very difficult to achieve. FDS showed a large uncertainty of the smoke layer height and temperature in the early stages of fire across the enclosure space. In the later stages, this uncertainly became minimised where the smoke layer height and temperature were fairly uniformly developed across the space. For fire enclosures with instantaneous steady-state fires, the predictions between BRANZFIRE and FDS agreed well with each other if the fire size and the enclosure size were within a reasonable range. From the modelling of the full-scale experiments, FDS showed favourable layer-height comparisons against the full-scale experimental tests. However, the output results from BRANZFIRE are less comparable with those of FDS for the experiments with fire growth. An appropriate smoke transport time lag should be included for Zukoski’s smoke filling equation and BRANZFIRE; otherwise, they gave conservative estimates of the layer height to smaller fires with a growth phase. In general, the data reduction methods and zone models should not be used if the fire is too small relative to the enclosure size. A very low temperature rise within the enclosure space would give invalid predictions of the layer height and average layer temperature. This is because there is no clear indication of a separation between the upper and lower smoke layers or temperatures. Single point data of smoke concentrations and temperatures from CFD models should be considered through the entire space or at the specified location of interest. This also applies to an extremely large fire relative to the enclosure size where temperature distribution across the space might not be very homogenous. CFD models could also be used to investigate the details of the smoke properties in the early stages of growing fires, in which the smoke transport lag and the plume effects cannot be seen in BRANZFIRE. This research is intended to provide guidance for fire engineers by determining which of the computer methods can be used confidently and appropriately as a design tool.