1. Fire scenarios of CMS underground installations and model building Michael Plagge Physics Department Compact Muon Solenoid Experiment Group European Organization for Nuclear Research (CERN) Institute of Process Equipment and Environmental Engineering Department of Process Design and Safety Otto-von-Guericke-University Magdeburg February 16 th 2012

2. Ventilationfirecap modelopensmallonoffbigonoffclosedsmallonoffbigonoff Further possibilities: Intermediate plug (yes/no)? UP55, UPX56, concrete block level -3 (open/closed)? R56, R54, …. (yes/no) …… What kind of fire scenarios should we consider? As long as each entry has two options: 2 n = number of simulations (combining every option)

3. What are our expectations from the results? p,T, U distribution Smoke production, distribution and tracking Smoke concentration levels Contamination level …?  Improvement of air management (application)  Combustion model for contaminated solids and their products (theory)  …?

4. How to model combustion? Combustion model Mixture-fraction Gas burner, specified HRR or MLR Computationally “fast” Specified pyrolysis properties, MLR Only one gas species used for combustion. Reason: cost Finite-rate All species and reactions are defined and computed Computationally very expensive NIST FDS User Guide Version 5 – Chapter 11

5. FDS Mixture-fraction-model Depending on preset values: Standard reaction scheme following: Combustion reaction must be known in detail. FDS uses propane as surrogate.  Fractions varying from 0.. 1, experimental data needed (or just use default).  Default soot fraction:  Depending only on the molar mass of the fuel, the molar mass of the soot (default is carbon) and a preset soot yield (experimental estimated). NIST FDS User Guide Version 5 – Chapter 11

6. From soot to smoke – how to visualize? 1 st option: Using (massless) particles following the velocity field 2 nd option: Amount of light obscured by smoke depends on the amount of smoke between the observer and the background NIST Smokeview User Guide Version 5 – Chapter 4

7. 1 st option: tracer particles Streak lines: current location of a particle plus all locations where it has been. NIST Smokeview User Guide Version 5 – Chapter 4

8. 2nd option: volumetric method Beer’s law = Beer-Lambert law NIST Smokeview User Guide Version 5 – Chapter 4

9. SOOT_YIELD: Note that this parameter does not apply to the processes of soot growth and oxidation, but rather to the net production of the smoke particulate from the fire Colored scalar or ??? How to model smoke? NIST FDS User Guide Version 5 – Chapter 11  With respect to theoretical aspect (contamination model) the above missing phenomena could be quite important (?) FLUENT brochure “To simulate the smoke? I put a passive scalar as smoke with velocity and temperature are equal : 1m/s and 700 K.” OpenCFD Forum The fire was modeled using a volumetric heat source represented by the experimental heat release data as a function of time. The soot concentration and optical density were calculated using a given soot yield value implemented in the cells of the volumetric heat source.” Hagglund et al 1998 in IAFSS Asia-Oceana “FLUENT, Propane and air, Two combustion models namely, the Magnussen and Hjertager [21] eddy-dissipa- tion model and the mixture fraction/PDF. It was shown that the PDF approach used less CPU time and provides a better prediction of the #ame shapes. ” Wu et al 2000 in Fire Safety Journal

10. Substitute species?