A Study of Fire Durability for a Road Tunnel: Comparing CFD and Simple Analytical Models

Abstract

The durability of various typical tunnel sections in the event of a prescribed 100 MW fire has been assessed. Cast-iron sections, pre-cast concrete sections and in-situ concrete cut and cover sections are all considered to be part of a 1 km long road tunnel. An analysis of the tunnel constructions and surrounding geology (based on a real tunnel) has led to the estimation of failure temperatures for the structural elements, internal cladding systems, jet fans and their fixings. A commercial computational fluid dynamics (CFD) code was used to simulate various fire scenarios and calculate the times to failure of tunnel elements. Simulations were carried out for fires in different locations for the three section types. In parallel to the CFD study, an analytical model was devised to predict gas temperatures in the tunnel. Both models used the same input variables and general assumptions and great attention was given to establish the highest possible accuracy for all input variables and general assumptions. Comparing the predicted gas phase temperatures shows that there is less than a 20% difference between the complex CFD and the simple analytical model; this is well within the bounds of uncertainty inherent in either model and to the input parameters. Using both sets of gas phase temperatures, a detailed heat transfer study was carried out to calculate the temperature evolution of each of the tunnel elements. The differences in gas temperatures between the two modelling methods did not alter the conclusions regarding the time to failure of any tunnel elements. It is found that fire durability can be better analyzed by separating the fire environment into two zones, a near field close to the flames, where accuracy is defined by the assumptions, and a far field where the precision of the results is linked to the modelling method. This approach allows establishing that, for this particular case, failure of structural elements can only occur in the near field. This study shows that the detail of the calculations needs to be consistent with the accuracy of the input parameters and assumptions. Although CFD models can give highly detailed results, the implied accuracy of the results is defined by the assumptions inherent in the model setup, thus, there is the potential of a very costly and refined computation that leads to results of comparable accuracy to simple, less costly, models.