CERN European Organization for Nuclear Research SCGSPI Computer

CERN - European Organization for Nuclear Research SC-GS/PI Computer Fluid Dynamic Simulation of Fire and Evacuation Scenarios for Large Experiments of Physics Fabio Corsanego SC-GS/PI 5 th HEP Technical Safety Forum at SLAC 11 -15 Apr 2005 1

Topics • • SC-GS/PI Introduction CFD Codes Validation Applications • Evacuation Codes • Validation • Applications CERN - European Organization for Nuclear Research 2

How to assess fires . . let’s begin with the answer which we need; we want to know: … SC-GS/PI Gas Temperature slice plot Gas Velocity vector plot – …For structural integrity Walls and objects Surface Temperature plot • Long term wall temperatures • Hot gases temperatures – …For life protection • Early stage visibility • Smoke dilution & composition • Gas temperatures • (Radiant heat) Mixture fraction Iso-surface plot CERN - European Organization for Nuclear Research 3

How to assess fires (continued) …a glance at the CFD method… SC-GS/PI The idea is : – To run fully instrumented small scale real fire tests (record fire power, T, p, soot, chemical species and concentration) – to obtain results by interpolation of the real fire tests by means of semi-empirical equation Bigger precision and extrapolation of the small scale tests are possible, to a certain extent (*), with Computational Flow Dynamic (CFD) (*) CFD growing precision and validation means chance to downsize real test on materials and money saved! CERN - European Organization for Nuclear Research 4

Market Survey CFD Codes (most referenced) SC-GS/PI Here are reported the most referenced &used CFD codes for Smoke transportdiffusion modeling Dedicated Gen. Purpose FDS – (National Institute of Science and Technology –MD-US) (free) Phoenics (Concentr. Heat and Momentum Ldt, GB) (100 -2000 eu) Fluent (Fluent Inc. GB) (price requested) Flow 3 d (Flow Science Inc. NM-US) Smartfire (University of Greenwich) (2500 eu edu) Jasmine (Ies Limited GB) (12000 eu ) Sofie (University of Cranfield – GB) (150 eu) CFX CERN - European Organization for Nuclear Research 5

Validation effort at CERN and elsewhere on FDS 2 Heiss Dampf Nuclear Reactor at Karlsrhue – ~50, fully instrumented oil, propane, cables fire tests in the dismantled reactor vessel (20 m large, 50 m high) – FDS 2 was back-validated against these data at Maryland University Large scale test, well representative of large experiment halls SC-GS/PI Memorial Tunnel Fire Test in US – ~100 fully instrumented oil fire and ventilation tests – FDS 2 was back-validated against these data at CERN Long tunnel test, well representative of Accelerator Tunnels CERN - European Organization for Nuclear Research 6

FDS 2 Validation Heiss-Dampf Reaktor Karlsruhe SC-GS/PI Gasoil 2 -4 MW fire (Test T 52) Propane 1 MW fire (Test T 51) CERN - European Organization for Nuclear Research 7

HDR Propane TEST T 51 SC-GS/PI Data comparison ~5% ~11% FIRE room ceiling level layer Doorway ceiling level layer FIRE room floor level layer Doorway temperature profile ~15% ~100% CERN - European Organization for Nuclear Research 8

HDR TEST T 52 OIL 2 MW Data comparison Fire room temperature ~27% SC-GS/PI More on the geometry O 2 ~ 8% Upper Hatch temperature Doorway temperature ~45% ~12% Doorway velocity CERN - European Organization for Nuclear Research ~14% 9

FDS 2 MEMORIAL TUNNEL FIRE TEST VALIDATION SC-GS/PI 8 m high ng m lo 850 The Memorial tunnel test program has provided a huge amount of raw data on road tunnel fire Flight view inside tunnel Tests 8 m large CERN - European Organization for Nuclear Research 10

CASE T 501: 20 MW Natural Ventilation – (Temperature) 16’ from ignition TEMPERATURES OBSERVED: SC-GS/PI ~150 -300 F = 80 -150 C FDS 2 RESULT ~150 -300 F = 80 -150 C CERN - European Organization for Nuclear Research 11

Validation Overall Judgment SC-GS/PI The average results of the FDS 2 in different situations(*) situations is as follows : – Long term quasi steady Gas temperatures • errors of the order of 20 -30% of the DT – Gas Concentration • large errors, of ~ 1 order of magnitude – Velocity of transients, like smoke layers descent and movement • errors of the order of 40 -60 %. (*)different situations: situations • small room-small fire, • small room-big fire, • high bay-small fire, • high bay-big fire, tunnel fire, • pre-flashover, post-flashover, • poor ventilation, rich ventilation, • etc……!!!!!! CERN - European Organization for Nuclear Research 12

Evacuation Simulation M. Survey SC-GS/PI Evacuation simulation codes: • Simulex (I. E. S. International) – License 2500€/yr • Exodus 3 D Greenwich University (Fire Safety Group) – License 5000 €/yr CERN - European Organization for Nuclear Research 13

What we have to assess on fires: (continued) ‘…The construction works must be designed and built in such a way that in the event of an outbreak of fire: . . SC-GS/PI – …(d) occupants can leave the construction works or be rescued – …(e) the safety of rescue teams is taken into consideration Will occupants leave the structure before the smoke/fire produces untenable conditions? Ship muster point video CERN - European Organization for Nuclear Research 14

Exodus Validation References SC-GS/PI • The Validation of Evacuation Models. Authors: E Galea. CMS Press, Paper No. 97/IM/22, ISBN 1 899991 22 0, 1997 • The EXODUS Evacuation Model Applied to Building Evacuation Scenarios. Authors: M Owen, E Galea, P Lawrence. Journal of Fire Protection Engineering 1996, Vol. 8(2), pp 65 -86 • The Collection and Analysis of Pre-Evacuation Times from Evacuation Trials and their Application to Evacuation Modelling”. Gwynne S, Galea E. R. , Parke J, Hickson J. Fire Technology, Kluwer Associates, US, pp 173 -195, vol 39, number 2, 2003. CERN - European Organization for Nuclear Research 15

SC-GS/PI APPLICATIONS AT CERN - European Organization for Nuclear Research 16

The fire SC-GS/PI • Parabolic ramping fire power 0 -> 5 MW then steady • Two different growth rates ( 5 MW in 700 s , 1500 s) • Located in 3 different position (bottom, top, side) • 6 different runs per each experimental cavern CERN - European Organization for Nuclear Research 17

ATLAS CERN - European Organization for Nuclear Research SC-GS/PI 18

Atlas Fires Locations In front of one exit Under Detector SC-GS/PI Top of Detector CERN - European Organization for Nuclear Research 19

ATLAS Videos SC-GS/PI ATLAS THE CONSTRUCTION OF THE MODEL 2. avi ATLASLARGE. avi CERN - European Organization for Nuclear Research 20

ATLAS Slices (Fire under det. 5 Mw 300 s) SC-GS/PI t=150 s Smoke fills upper level t=100 s t=250 -300 s Smoke at level of upper exits t=200 s Smoke starts to descend on upper gangways CERN - European Organization for Nuclear Research 21

CMS CERN - European Organization for Nuclear Research SC-GS/PI 22

Smoke development and transport within the cavern SC-GS/PI CMS model showcase CMS Fire Under Detector 5 MW CERN - European Organization for Nuclear Research 23

Smoke layer position on the xz plane of the cavern 3 rd level gangway flooded by smoke SC-GS/PI t=3’ t=5’ 1 st level gangway flooded by smoke t=7’ 2 rd level gangway flooded by smoke t=6’ CERN - European Organization for Nuclear Research 24

Evacuation modelling SC-GS/PI 100 people in CMS 130 people in ATLAS Snapshots of the CMS Model Snapshot of the Atlas Evacuation Simulation CERN - European Organization for Nuclear Research 25

Results Overview ATLAS SC-GS/PI ATLAS: Smoke propagation to exit level gangways: 300 s (medium propagation) to 480 s (moderate propagation) Evacuation Time: [120 s for discovery] + 90 to 160 s= 210 s to 280 s Evacuation Validated CERN - European Organization for Nuclear Research 26

Results Overview CMS SC-GS/PI ATLAS: Smoke propagation to exit level gangways: 250 s (medium propagation) to 360 s (moderate propagation) Evacuation Time: [120 s for discovery] +85 to 110 s= 205 to 230 s Evacuation Validated CERN - European Organization for Nuclear Research 27

Conclusions SC-GS/PI CFD and evacuation simulation tools allowed simulation and reasoned validation of complex and non standard geometries The findings show that the gap between evacuation and smoke propagation is not high Several measures have to be in place in order to minimize delays and difficulties in case of emergency …(but this is by itself a different and huge subject) CERN - European Organization for Nuclear Research 28