1 Department of Civil Engineering, Technical University of Denmark2 Section for Building Design, Department of Civil Engineering, Technical University of Denmark3 Technical University of Denmark4 Exponent Inc.5 Imperial College London6 Imperial College London
The HVAC component of FDS 6 was used to divide a 1.2km tunnel into a 3D near fire area and a 1D area further away from the fire in order to investigate the feasibility of multiscale modeling of tunnel fires with this new feature in FDS. The two sub-models were coupled directly. The results were compared with reference works on multiscale modeling and the outcome is considered positive, with a deviation of less than 5% in magnitude of relevant parameters, yet with a significant reduction of the simulation runtime. As such, the multiscale method is deemed feasible for simulating tunnel fires in FDS6. However, the simplifications that are made in this work require further investigation in order to take full advantage of the potential of this computational method. INTRODUCTION Multiscale modeling for tunnel flows and fires has previously been studied using RANS general purpose CFD software and it has yielded satisfactory results in comparison to full scale CFD simulations [1-3]. It combines a 3D domain for the near fire zones, which are characterized by large temperature and pressure gradients, with a 1D network approach for the far field, where the flow is treated as a mono-dimensional quantity. The present study aimed to analyze whether or not the multiscale modeling approach for tunnel fires could be successfully applied in Fire Dynamics Simulator 6 (FDS6), an open source, fire-specific CFD software  that is easily accessible to modeling specialists. METHOD The implementation of multiscale modeling in FDS used a mono-directional road tunnel with a rectangular cross-section of 8m width and 6.5m height. Its total length of 1200m was split into 400m of 3D domain with two 1D ducts of 400m on either side. The 3D domain was divided into 17 meshes in order to reduce the runtime of the simulations. The tunnel walls were considered adiabatic and the model used ambient conditions at the portals. The 1D model was defined using the novel HVAC component of FDS6 as two ducts connected to the FDS domain and to the ambient by two nodes each. The ducts had a cross sectional area equal to the one of the tunnel in the 3D domain. A fixed flow was specified in the tunnel to induce the flow given by the jet fans. The ventilation system consisted of 5 jet fan pairs in each duct, but only the jet fans from one side were activated simultaneously. The main fire scenario involved a 30MW fire, as this exemplifies the peak release rate of a burning bus. DISCUSSION The results obtained using FDS showed good agreement with the ones obtained in the reference work by Colella et al . While the multiple meshes introduced an error of approx. 10ºC and velocity errors of less than 0.5m/s, the results showed deviation of less than 5% from the results obtained using a full CFD solution in the reference work. Figure 2 presents veraged temperature results in the 3D model from the simulation involving the main scenario. The durations of the simulations were significantly reduced using the multiscale model with a cell size of 0.4m, with one simulation requiring around 6h to complete on an 18-core computer using multiple meshes. CONCLUSION Using the results of the reference work as validation, it was concluded that it is feasible to use multiscale modeling of tunnel fires in FDS6. This method provides a significant reduction in run time and computational resources, while maintaining an accuracy similar to the one given by using a full CFD solution.
Book of Abstracts: Fire Safety Day 2014, 2014, p. 32-33