Towards 1D/3D coupling for fire and ventilation modelling in large underground infrastructures ?
Résumé
Risks in underground infrastructures are commonly associated with fire that is the most frequent phenomena that occurs in such places. A lot of studies were carried out to provide better understanding of fire behaviour in tunnels, mainly after the huge fire in the Mont Blanc and Tauern tunnel in 1999. Those studies include experimental tests, from small [1] to large scale [2], and numerical simulation, from analytical model or 1D model [3] to highly complex CFD approach [4, 5]. Based on all studies that were achieved, great improvements have been made in terms of fire knowledge and safety design capability. Mainly, computation precision was highly increased using both physical model improvement and informatics possibilities. However, even if current computer performances cannot be compared with the ones available ten years ago having the whole description of a complex underground infrastructure with a CFD model stays unrealistic. There was, during this last decade, several discussions relative to the ability of the CFD codes to predict real fire behaviour in tunnel [6] that have induce progress in such modelling. However, even with those improvements, two major issues still resist: the fire concerns the sub model used in the CFD code, the second concerns the time required in case of large infrastructure modelling. Of course, those two problems are closely linked considering that minimising the cell size, i.e. increasing the cell number, will induce a required time increase but also improve the quality of the results. The second approach that can be used for evaluating fire impact in tunnel and designing ventilation is 1D models [7,8]. Those codes are based on the hypothesis of a homogeneous distribution of physical quantities in the tunnel section. This means that some information is loosed, such as thermal stratification for example. Those codes are however able to give a precise description of the pressure losses distribution, that makes those tools very useful for ventilation system design. As described above, both approaches have benefits and drawbacks. It then appears as a great interest on coupling those two modelling level. Such a coupling will enable to have both a precise description of the physical phenomena that occur close to the fire and a global description of the whole infrastructure. If this can appear complex for simple infrastructure as road tunnel, such a coupling is crucial to model events in subway station. A coupling methodology was initiated using FDS, the well known fire code [9] and the INERIS 1D code VENDIS [8] for 1D modelling. The VENDIS code enable to compute pressure and flow rate equilibrium in highly complex network and can then be easily used for all underground infrastructures as tunnel, subway system but also mining network. As first evaluation test was made for a subway system with several stations as represented on Figure 1. This example is used to demonstrate the possibility of an FDS-VENDIS coupling enable to reach the above mentioned objectives, this means local information on the fire behaviour and global overview of the whole infrastructure.