When applied to Fire Safety Engineering, Computational Fluid Dynamics techniques enable designers to analyse proposed structures in far greater detail than would otherwise be possible. These methods are applicable to geometrically complex designs, whereas the more established zone model analysis technique is generally suited to multi-room rectangular type compartments only. Whilst CFD based models have provided engineers with the flexibility and power to perform detailed analysis on a given design, they have also added a level of complexity to the process. Far from being a mature technology, CFD models require considerable time and expertise to be used successfully. Construction and simulation of large models is often time-consuming, and can be compounded by design alterations (often based on predictions from previous simulations). Furthermore, given a fire hazard scenario, predicting the fire growth and transport of combustion products and heat, is a complicated problem requiring a consideration of turbulence, combustion and extinction, radiation and species transport.
Improvements to the modelling of these phenomena are belied by the vast quantity of possible fire scenarios and the sensitivity of the processes to input parameters, which themselves are the main impetus for the development of comprehensive models. Whilst reduced model simulation time is predominantly reliant on the development of faster computers and more efficient algorithms, model construction, redesign and simulation times may be reduced by limiting the level of geometric detail incorporated. Considering the variability of fire scenarios, the accuracy of the various 'sub-models' of the code, and the approximate results thus typically possible, it was proposed to investigate whether model construction time could be reduced by simplifying the modelled geometry. These simplifications were considered to be such that the uncertainties in all aspects of the model were similar so that the results remained practically significant.
Geometric simplifications were applied to a model based on a representation of the Queen Street Bus Terminal located in Brisbane, Australia. The work did not aim to assess the performance of this facility in the event of a fire and this work in no way reflects this. Studies investigating the influence of approximating curved walls with straight-wall segments; neglecting small tunnel elevation changes; approximating tunnel exit geometry; additional obstructions within the flow region; and truncation of the model extents were tested using Fire Dynamics Simulator. An additional simulation using Fluent was run to compare results between two different CFD codes.
The studies have shown that:
•Approximations to curved wall geometry are possible for initial investigative analyses. In these cases, the degree to which the correct flow 'streamlining' is represented is of most importance. Predicted flow properties (temperatures and velocities) may be expected to be similar (within the variability of the codes) for these cases.
•Changes in elevation variation should not be neglected as such changes form the basis of buoyancy driven compartment fire flows. The method of vertically truncating tunnel exits is a viable method of simplifying the geometry and reducing the size of the computational domain. These exits are preferable over ceiling holes as, in the latter case, an inverse plugholing effect can cause preference for the exiting gases (especially if the exit is not next to an end-wall).
•Internal flow obstructions which do not significantly block the flow path of gases in the smoke layer may, in the majority of cases, be neglected. Examples of such items are chairs, tables and shelves in room models, and larger obstructions such as vehicles or columns in tunnels. These approximations are valid provided the obstructions do not contribute a significant fraction of the total internal volume. Obstructions within the smoke layer itself should be included as stratification of this layer is of particular importance in the dispersal of combustion products throughout the domain.
•Truncating a particular model at a location which could not be considered an explicitly ‘open’ boundary is an efficient method of reducing model setup and simulation times. However, the influence of such truncations can be unexpected. Consideration should be given to the fact that, because the flow is approximately incompressible, the model geometry at the domain boundaries influences the entire model flow from almost the beginning of the simulation. Thus, it is not sufficient to model only the region encompassing the fire and ceiling jet after a specified duration. In particular, truncation of a simple straight tunnel or side tunnel which incorporates an unobstructed open exit is possible and may generally be considered to produce similar results to the full model. On the other hand, models should not be truncated if these side regions do not have open exits, have slightly restricted flow paths (for example, curves or any obstructions), have a varying elevation or are connected to other parts of the model. Although these geometric simplifications have been applied to a specific case study, it is considered that the results are of general significance (that is, applicable to other transit facilities). The variation in possible CFD model designs is limitless and thus this study is intended to form an example outlining the possible influences of geometric simplifications to compartment models. A selection of additional studies is given at the end of the report.
Although some level of simplification is made to all CFD models, the report states that there appears to be no significant documentation on the subject. Overall, it is considered that if the use of various simplifications is documented, then such time-saving measures may be used with some justification.