Computer simulation of fire-sprinkler interaction
The research work presented herein addresses the problem of the mathematical modelling of fire and fire-sprinkler scenarios. This involved the numerical simulation of two-phase, three-dimensional, buoyant, turbulent, recirculating flows. The simulations were carried out in two successive and distinct stages. The first stage dealt with the modelling of buoyant hot turbulent gas flows generated by a fire within room-sized compartments. These single phase studies were based on the field modelling approach to fire simulation. The second part of the study involved the introduction of the cold water droplets through a single sprinkler head, thus, extending the scenario into the more complex two-phase regime. This led to expanding the single-phase fire model to take account of two concurrently present phases, ie. gas/liquid. The strategy used to model the twophase process was the Eulerian-Eulerian technique or volume-fraction method. In order to take into account the physics involved in this process, interphase friction or drag was considered. Furthermore, due to the large difference in temperatures between the hot gases and the cold water droplets it was necessary to introduce interphase heat transfer. Due to the subsequent evaporation of the water droplets interphase mass transfer was also accounted for. Models for both steady-state and time-dependent situations were developed, whereby experimental results of transient fire-sprinkler tests were used for validation. The simulations performed indicated the creation of extremely complex flow fields within the compartments both prior and during sprinkler activation. Phenomena such as the significant cooling of the hot combustion gases caused by the active sprinkler and the evaporation of water droplets have also been predicted. This has been verified by the experimental data. Thus, it can be concluded that the models outlined herein are capable of simulating the complex two-phase fire-sprinkler scenarios. The need for subsequent investigative studies into such areas as the effect of using different auxiliary relationships eg. heat transfer, sprinkler characteristics and gridspacing has been highlighted. In order to complete the validation process, further experimental data needs to be made available. This two-phase technique has proven to be very computationally intensive with simulations requiring days of CPU time. 'Mis is clearly unacceptable. However, it is suggested that parallel computing technology may provide a means for reducing the CPU time involved to hours. It can be concluded that though the model developed requires further investigation and refinements, it provides a basis for a practical and useful fire engineering tool.