Tunnel fire

Application Challenge 4-04 © copyright ERCOFTAC 2004

Overview of CFD Simulations

A number of CFD simulations have been carried out of the Memorial Tunnel tests. That included here was carried out by Mott MacDonald on behalf of the French Centre d’Etudes des Tunnels and the Dutch Ministry of Transport and Public Works and considered only cases 615B and 621A. The test case considered in this document is 615B.

NAME	GNDPs		PDPs (problem definition parameters)		SPs (simulated parameters)
	Re	Fr	Fire size	Jet Fans	Detailed data	DOAPs
CFD 615B (jet fan fire test)	To $10^{6}+$		$\approx$ 100MV (follows fire curve)	Variable operation	T,V, Smoke	T,V

Table CFD-A: Summary description of test case

	SP1 Velocity $(ms^{-1})$	SP2 Temp $(^{0}C)$	SP3 Smoke Conc (%)	DOAPs, or other miscellaneous data
CFD 615B				Bulk Airflow $(m^{3}/s)$

Table CFD-B Summary description of simulated parameters

Simulation Case

Solution strategy

The CFD model was created using the commercial CFX code, Version 4.1, from AEA Technology Plc. This is a standard pressure-based, cell-centred, finite-volume code using a co-located variable arrangement (with Rhie-Chow pressure stabilization) and the SIMPLEC pressure-correction algorithm.

The problem was treated here as a time-varying, three-dimensional turbulent flow and equations were solved for pressure, velocity, static enthalpy (energy equation), turbulence kinetic energy and its dissipation rate and a relative smoke concentration. The fluid was defined as being weakly compressible (with density variations arising only from temperature variations).

The first-order Euler implicit scheme was used for the time discretization while for the convection terms the hybrid (central/first-order upwind) scheme was employed.

Computational Domain

The computational domain consisted of the whole length of the tunnel though with only half of the cross-section represented, exploiting the tunnel symmetry. A total of about 63,000 cells were used in a multiblock hexahedral mesh, with approximately 145 cross-sectional planes and 430 cells in each plane.

The longitudinal mesh spacing is shown below:

The diagram has been scaled by 15 in the vertical direction to enable easier viewing of the mesh. The white solid regions represent the location of the jet fans. The cell lengths are approximately 2m in the vicinity of the fire. The highest cell lengths are at the north portal, at 9.5 m. The cross-sectional mesh is shown below:

The circular and semi-circular regions indicate the location of the jet fans (plane of symmetry imposed on left-hand boundary).

Boundary Conditions

Fixed-pressure boundaries were applied at both tunnel portals with an ambient temperature of 280.2 K (7.2°.C).

A series of simulations were carried out using different numbers of jet fans in the absence of a fire, in order to match the results of the airflow tests performed in the tunnel prior to the fire tests. In order to obtain a close match, the roughness height of the tunnel wall was changed. The results were:

Number of Fans	Measured Flow	Smooth Wall	Rough Wall (Roughness Height 0.003)
1	275,000	305,335	289,553
5	600,000	634,128	595,318
10	835,000	900,717	849,882

The rough wall flows show less than 2% variation from the measured flows and were deemed accurate enough to enable the roughness height of 0.003m to be used in the fire simulation.

Prior to the fire test, a measured flow of 1.3 – 1.4 m/s was apparent in the tunnel, from the south to the north portal. In order to obtain this initial flow, the temperature of the tunnel walls was set to 9°.F (5°.C) above the ambient value. The predicted steady-state flow was 1.32 m/s. In the fire simulation therefore, this value of 5°.C above ambient was fixed on the walls.

The jet fans were represented as combinations of internal mass flow boundaries / inlets creating the appropriate volume flow rate. The volume flow rate for each fan was 43 m3/s (91 kcfm). The temperature of the air/smoke mixture and smoke concentration entering the fan were preserved at the fan exit. The specification of the mass flow and inlet boundaries at the jet fans required the fans to be modelled as at least two cells in the longitudinal direction.

The fire was represented as a source of heat and smoke concentration, varying with time. The measured fire curve for the test was used to obtain the heat generation at each time step in the simulation. A source and sink of mass was also used to model the effect of the generation of fire products and the oxygen consumption by the fire.

Time steps of 5 seconds were used throughout the simulation.

Application of Physical Models

The standard high-Reynolds number k-ε model with wall functions was used to represent the turbulence effects and no buoyancy corrections were employed. The flow was taken to be weakly compressible, with the density given by the equation of state.

Numerical Accuracy

In order to keep a check on the convergence history of the simulation, a monitoring point was specified 100m upstream of the fire source, at 1.3m from the floor and 0.75m from the symmetry plane at the left boundary of the domain. The number of iterations per time step was fixed at 200 at the start of the simulation and gradually reduced to 125 for the later stages.

Examination of the residuals indicates that an adequate level of convergence was achieved at each time step. Total mass residuals were reduced to the order of 10-1, with values of all the variables constant at the monitoring point to within 3 significant figures.

CFD Results

In order to enable comparison with the measurements obtained during the test, values of temperature, smoke concentration and velocity were obtained from the predicted flow field at each of the instrument locations used in the test. The data was extracted from the grid cell overlying or nearest to each of the instrument locations on each instrument tree.

For the temperatures, the data presented on the CD-ROM gives average temperatures at each measurement height for each instrument tree. Averages of the predicted temperatures at each height were therefore calculated in the simulation and tabulated in a similar fashion.

A similarly consistent practice was adopted for the velocity values (given at the centre-line of each measurement tree on the CD-ROM). Equivalent carbon monoxide concentrations have not been calculated, although the predicted smoke behaviour provides a qualitative comparison with the smoke diagrams given on the CD-ROM.

The numerical simulation data is given in the following files.

Q615CFD.dat File containing simulated bulk airflow data.

V615CFD.dat File containing simulated velocity data.

T615CFD.dat File containing simulated temperature data.

References

1. Castro, J D, Else, K, Rhodes, N: CFD Modelling of Memorial Fires, Mott MacDonald Report 50838/01/B, February 1998. (Work carried out for the Centre d’Etudes des Tunnels, France.)

2. Rhodes, N: CFD Modelling of Tunnel Fires, World Road Association (PIARC) World Road Congress, Kuala Lumpur, 1999.

3. CETU Report: Evaluation of Memorial Tunnel CFD Simulations, March 1999.

Contributors: Nicholas Waterson - Mott MacDonald Ltd

Site Design and Implementation: Atkins and UniS