FACULTY OF ENGINEERING NUMERICAL SIMULATION OF FLOW FIELDS IN CASE OF FIRE AND FORCED VENTILATION IN A CLOSED CAR PARK Xavier Deckers, Mehdi Jangi, Siri Haga and Bart Merci Department of Flow, Heat and Combustion Mechanics (Xavier.Deckers@Ugent.be) pag. 1
INTRODUCTION Design of smoke control systems in closed car parks: Past: prescriptive (simple rules of thumb) Now: performance based design Development of standards describing the required performance objectives in order to acquire a certain level of safety in terms of performance criteria. Possible Belgian standard: performance NBNobjectives: S-21-208-2 (2008) Protection Performance of requirement: the means of escape Protection access ofshall the structural be kept smoke-free elements by from limiting the public the maximum road to less temperature than 15 m from the fire Smoke Prescriptive clearance method: afterstandard a fire solution (Annex A) Assistance ventilation to firevelocities fighting: and flow rates prescribed Limit by-pass thephenomenon propagation of to smoke be avoided and heat Performance Permit safe based: access solution to firefighting validatedteams by CFDand calculation facilitate(annex their intervention B) Government Take intofunded account multi-disciplinary influence of beams research + useproject: of jet fans http://www.carparkfiresafety.be Performance criteria: smoke density < 0.01 g/m 3 at ceiling Improvement of existing ventilation design guidelines for closed car parks. European standard in preparation Fundamental design approaches for improvement of the fire safety in car parks
CASE STUDY Car park model: L x W x H = 30 m x 28.5 m x 2.4 m; HRR = 4 MW; Simulation settings in OpenFOAM buoyantfoam solver transient solver for buoyant, turbulent flow of compressible fluid for ventilation and heat transfer Structured mesh: cubic cells of 0.2 m Physical sub-models Turbulence model: standard k-ε RANS model No combustion model: fire included as volumetric heat source No radiation model: imposed 4 MW is convective HRR Boundary conditions Adiabatic walls: imposed heat entirely transferred to smoke Uniform inlet velocity: 1.3 m/s Outlet: fully advective Symmetry: half of car park Standard wall functions
CASE STUDY Car park model: L x W x H = 30 m x 28.5 m x 2.4 m; HRR = 4 MW; Simulation settings in OpenFOAM Investigation of parameters Physical parameters in the car park Ventilation flow rate Distance of the fire from the wall Presence of neighbouring cars Different sub-models in the simulations Turbulence model: RANS and LES (OpenFOAM) Combustion model (FDS) - VHS (OpenFOAM) Grid sensitivity for LES d
Influence of the imposed ventilation flow rate Vary the uniform velocity over the entire width Examine temperature distribution in car park: Capture the back-layering distance for different velocities (a) U in =1.3 m/s (b) U in =0.5 m/s lower ventilation velocity increase in back-layering length higher average temperature
Influence of distance of the fire source to the wall Design: worst case scenario for positioning of the car fire Entire car park as flow is not symmetrical Lf (a) L f = 6 m from wall (b) L f = 3 m from wall fire closer to the wall increase in back-layering length L f,crit below which the flow is significantly affected further research
Influence of neighboring cars Middle car on fire Scenario A Scenario B presence of cars will change flow pattern
Influence of neighboring cars blank space leads to a cavity flow between the car on fire and the upstream car temperature distribution at ceiling is clearly different (a) blank space between cars (b) no blank space Inclusion of neighbouring cars more realistic temperature distribution induces higher local velocities (obstructions)
Influence of neighboring cars The effectiveness of a smoke control system (back-layering) evaluated in an empty car park: conservative approach Proposed approach: dependent on objective of CFD analysis Determine the minimum required ventilation flow rate (smoke control design) empty car park Determination of temperature distributions (structural fire safety or fire spread) incorporate obstructions caused by cars
Influence of turbulence model RANS LES Unstructured grid (0.03-0.25 m) Volumetric heat source t LES = t RANS / 10 (in order to keep CFL < 1) Choice of grid in LES by estimating turbulent integral lentgh scale in k-ε results. LES: standard smagorinsky (C s =0.2) Unsteadiness Higher T with LES LES, t= 227 s K-ε, t= 225 s Use of symmetry BC with LES seems allowable (for determination of back-layering distance) remains to be confirmed with other CFD-packages LES, t= 227 s K-ε, t= 225 s
Influence of combustion model Mixture fraction based combustion model (FDS 5) VHS (OpenFoam) Uniform grid (0.2 m) LES (Smagorinsky C S =0.2) Symmetry plane General flow dynamics very similar For evaluation of the effectiveness of a smoke control system, VHS seems an adequate approach. (a) U in =1.3 m/s (b) U in =0.5 m/s
Effect of including the baroclinic generation of vorticity (FDS): Uniform mesh (0.2 m) Symmetry plane Regions where density gradient normal to pressure gradient. Inclusion of baroclinic generation of vorticity Creates more turbulent structures Leads to a decrease in back-layering length. Effect of baroclinic generation of vorticity: with (above) and without (below)
Grid sensitivity analysis LES (Standard smagorinsky, C s =0.2) Volumetric heat source Symmetry plane LES grid size: Unstructured grid (0.03-0.25 m) Coarse uniform mesh (0.2 m) Fine-mesh, Δ min = 0.03 m Coarse Uniform mesh, Δ= 0.20 m Qualitatively similar results More diffusion with coarser mesh Fine-mesh, Δ min = 0.03 m Coarse Uniform mesh, Δ= 0.20 m
CONCLUSION CFD simulation results (OpenFOAM) for fire in a closed car park distance of the fire source from a wall presence of cars objective= determine required ventilation flow rate (smoke and heat control) Empty car park is conservative objective= determination of temperature distributions (structural or fire spread) Incorporate possible obstructions Volumetric Heat Source inclusion of combustion model does not seem necessary (for determination of back-layering distances) determination of suitable minimum grid size for LES (using k-ε results) different results when using uniform coarser grid global pattern remain unchanged
Thank you for your attention! Questions? E-mail: Xavier.Deckers@Ugent.be