Research Interests

B RE

C entre for

F ire Safety Engineering

“Fire modelling”

QMFSE lecture 8







Stephen Welch

Lecturer in Computational Modelling for Fire Safety Engineering

School of Engineering and Electronics

The University of Edinburgh

Content

• Fire Safety Engineering (FSE)

 Guidance documents

 Scope of FSE/models

• Fire models

 Simple calculation methods

 Zone models

 CFD models

 Comparisons

 Demo

Fire safety engineering - scope

Fire Risk









Fire Initiation

Passive Fire

and

Protection

Development



Fire

Suppression Fire Detection

Systems

Smoke

Movement





People and

Fire

Fire Safety Engineering (SFPE)

• Application of science and engineering

principles to protect people and their

environment from destructive fire

 analysis of fire hazards

 mitigation of fire damage by proper design, construction,

arrangement, & use of buildings, materials, structures,

industrial processes, & transportation systems

 the design, installation & maintenance of fire detection &

suppression + communication systems

 post/fire investigation & analysis

Provision for FSE – BS7974

1: Initiation and development of fire within the enclosure of

origin

2: Spread of smoke and toxic gases within and beyond the

enclosure of origin

3: Structural response and fire spread beyond the enclosure of

origin

4: Detection of fire & activation of fire protection systems

5: Fire service intervention

6: Evacuation

7: Probabilistic risk assessment

Eurocodes

Model types

• Deterministic

 one outcome for a given set of fixed initial

conditions and pre-determined inter-relationships

between events



• Probabilistic

 a weighted distribution of possible outcomes for a

given set of probabilities assigned to initial

conditions and later events

 Risk assessment e.g. METHOD, CRISP

 human egress e.g. SIMULEX, CRISP

Fire models - deterministic

• Simple calculation methods

ASKFRS, FPE tool, Design guides

• Zone models

Roofvent, CFAST, HAZARD

• CFD models

RANS:

 JASMINE, SOFIE, PHOENICS, CFX

LES:

 FDS, CFX

Fire models - probablistic

• Risk models

CRISP, METHOD

• Evacuation models

CRISP, SIMULEX, EXODUS, Gridflow, EXITT

Fire models - further info

• “Fire model survey”

http://www.firemodelsurvey.com/

Review papers

 SFPE Journal of Fire Protection Engineering, 13 (2),

2003

 SFPE Journal of Fire Protection Engineering, 4 (3),

1992

Web database

Fire modelling tools (BS7974)

• Level 1 - Deterministic: Calculations or

zone model (CFD model in some cases)

 Hazard analysis of specific aspect of design

• Level 2 - Deterministic: Zone or CFD

model

 Hazard analysis of overall design

• Level 3 - Probabilistic: Risk model

 Risk analysis of the overall design

The „Classic‟ Zone Model

• Single compartment with one fire and one opening







• Hot upper smoke layer

• Cool lower layer

• Fire plume

• Vent flow

Zone modelling concept

• Divide room/building into a small number of

components („zones‟)

 hot upper layer, cool lower layer, fire plume, vent outflow etc

 assume properties (e.g. temperature) uniform in each zone

• Describe interaction between zones using physics

and experimental observation

 mass, momentum and energy conservation

 supplement with experimental observation

• Solve the resultant set of equations

 predict temperature, flow rate, fire products etc in each zone

Basic zone model components

• Fire source • Vent flows

 heat release or pyrolysis rate  transfers mass & heat

 radiative fraction between rooms and to

 basic chemistry outside

 heat of combustion  doors & windows

 ceiling vents

• Cold lower layer

• Heat losses to walls

• Hot upper layer

 from hot layer into walls &

• Plume model ceiling

 transfers mass & heat

between lower & upper

layers

Advanced submodels

• Fire source • Heat losses to walls

 species yields for CO2,  composite materials

H2O, CO etc  solution of 1-D conduction

 multiple fire sources equation

• Ceiling jet • Multiple

• Radiation rooms/compartments

 from flaming region  vent flow from one room

feeds hot upper layer in the

 from hot layer

next room

 due to particulates, CO2 &

H2O • Flashover room

 transfer to objects and  single hot zone

boundaries

Current capabilities

• Today‟s state-of-the-art models may include some

(but not all) of the following:

 multiple compartments

 combustion chemistry

 radiation

 species yield & oxygen

 ceiling jet depletion

 heat conduction into walls  under-ventilated conditions

 post-flashover capability  detector response

 roof vents  ignition of remote objects

 multiple fire sources  HVAC

• Coupling to other hazard models

 e.g. human egress

Zone models in common use (1)

• ASET-B • ASET

 widely used  ASET-B, plus elementary

 written in 1980s for PCs hazard & detection analysis

 very simple

 user specifies:

 room area & height

 fire heat release rate

 calculates temperature and

hot layer height in a single

room with closed doors &

windows

 available from NIST (free)

Zone models in common use (2)

• The HARVARD family of codes

 assumes hot and cold layers in fire compartment

 includes treatment for radiation heat transfer

 relatively sophisticated chemistry

 allows prediction of heating and possible ignition of targets

 HARVARD 5 and FIRST are single-room zone models

 FIRST is available from NIST (free)

 HAZARD 1 is available from NFPA ($250)

 includes a human egress model

 HARVARD 6 is a multi-room version

Zone models in common use (3)

• The CFAST family of codes

 together with ASET-B probably the most widely used

 multiple rooms/compartments and fire sources

 combustion species predictions

 radiation from hot gas layer

 FAST/CFAST is available from NIST (free)

 uses latest CFAST for the zone modelling

 includes some life threat calculations due to heat and toxic gases

 includes graphical user interface

 The Firewalk Project

 coupling of CFAST with a virtual environment/reality suite

Some other zone models

• LAVENT • FISBA

 models activation of  combustion in upper layer

sprinklers and ceiling vents • FireWind

 includes a ceiling jet  developed from FireCalc

component/zone

 egress model (WayOut)

 available from NIST (free)

• Other zone models:

• CCFM

 ARGOS

 multiple rooms & vents

 MRFC

 natural or forced ventilation

 FLAMME_S

 wind & stack effects

 FIGARO

 available from NIST (free)

 CSTBZI

CFD models

• Most general sense: • Fluid flow

 all aspects of computer- conservation equations

based simulation of fluid  mass

flow phenomena  momentum

• More specifically:  energy

 computer simulations  principally heat in fire

solving fluid flow problems

conservation equations  other properties

 using an established  e.g. chemical species

numerical methods for • Discretized space

second-order partial  Cells or control volumes

differential equations

Navier-Stokes equations

• A generic equation holds for all the main

conserved properties associated with fluid flow

    u j 

  

  S

(  ) 

t x j 

   x j 



source terms (e.g.

convection

heat from fire)

time rate of

diffusion

change





 satisfied for each conserved property at each control volume

 generates a very detailed (and potentially accurate) solution

Basic CFD fire model elements

• Heat sources • Smoke exhaust

 fires, HVAC, radiators  naturally or mechanically

• Solid boundaries ventilated

 arbitrary geometric shapes • External wind

for building elements and • Radiation heat transfer

internal obstacles  surface to surface

 heat losses to walls etc  from „smoke‟ layer

• Ventilation openings

 doors, windows etc

 any number or configuration

Zone v CFD models

• Zone 1 • CFD

2 2



 room(s) divided into a few,  geometry divided into lots of

uniform regions (zones) small regions (cells)

 one value of temperature,  one value of temperature etc

smoke concentration etc in at each cell

each zone  valid for complex shapes

 valid for simple shapes  detailed solution generated

 mass, heat, combustion  capture variation within large

products transferred regions used in zone models

between zones according to  conservation equations

 scientific laws  less reliance on empiricism

 experimental observation

Zone models - strengths

• Capture main features of a room fire

• Run quickly on a PC

 many alternative simulations

1

 Monte Carlo studies

2 2

 sensitivity analysis straightforward

• Relatively easy to lean to use

 models now available with graphical user interfaces

 but require knowledge of fire safety science/engineering

Zone models - weaknesses

• Usually assume a well-ventilated enclosure fire

 not appropriate for complex buildings or very large spaces

 some models handle under-ventilation/flashover and larger spaces

• Each zone is relatively large

 assumed spatial uniformity may bypass important features

• Significant reliance on empirical coefficients

 e.g. plume entrainment formulae

• Physical phenomena 1

 loosely coupled 2 2

 less general

• Tells us nothing about the flowfield!

Zone models - typical use

• Compartment smoke filling

 time-dependent change in upper („smoke‟) layer temperature,

height, species

 smoke exhaust capacity calculation 1

 available time for egress 2 2

• Smoke spread between rooms

 generally through connecting doorways in a multi-room model

• Compartment temperature

 average temperature of compartment walls and ceiling

 incident fluxes

 relevant particularly for post-flashover analysis

 single-zone modelling

CFD models - strengths

• Details of fluid flow represented

as accurately as required

 resolving details within large regions

• Treat arbitrary geometries (in principle!)

 any size or shape

 any number and location of compartments & openings

• Physical phenomena

 comprehensive treatment

 coupled combustion, soot, radiation, turbulence!

 empiricism reduced

• Powerful post-processing and analysis

 detailed flow patterns, temperature distributions etc

CFD models - weaknesses

• High computational requirements

 typically days or weeks

 but running codes on PCs is now realistic

• Requires a greater knowledge of fluid dynamics

and numerical procedures

 but knowledge of fire safety science/engineering is the most

important pre-requisite for ANY computational fire modelling

• Simulations require care and „nursing‟

 fire simulations prone to numerical divergence and error

 skill and experience is critical

 general purpose

 specialist codes

CFD fire model limitations

• Fire source • Turbulence

prescription  an ever present issue with

 fire burning rate is generally all CFD applications

prescribed by the user  approximations needed

• Chemistry  mesh resolution

 simplified combustion

only CO2 & H2O

 single fuel

• Heat conduction into

solids

 mesh resolution!

CFD models - typical use

• Smoke movement problems

 Can handle large complex spaces

 Airport terminals

 Atria

 Typically 100k-1M cells adopted

 Quasi steady simulations

 Smoke control system design

 size of extract fans

CFD models - combustion

• Combustion models

 distributed heat release

 c.f. heat source models

 combustion products predictions

 raditaion model input

 soot formation and oxidation

 allows accurate modelling of radiation from flaming region

 turbulence effects accommodated

 RANS codes

 LES codes

Turbulent flow - fire









From: Cox G & Chitty R, “Some Stochastic Properties of Fire Plumes”, Fire & Materials, 6, 127-134, 1982

CFD - RANS & LES

• RANS

 Reynolds averaged Navier Stokes

 “Ensemble averaging” of turbulence motions

 Quasi steady solutions

• LES

 Large eddy simulation

 Explicit representation of turbulent motions at grid scale

 Subgrid scale model

Why do we model turbulence?

• In principle flow can be solved directly

 Direct Numerical Simulation (DNS)

 Requires mesh sufficient to resolve smallest scale

 Estimate the number of nodes required (3D):

ReT Nodes Iterations

3

 lo  12,300 6.7x106 32,000

N nodes     Re lo

9/ 4

l  30,800 4.0x107 47,000

 k 61,600 1.5x108 63,000

230,000 2.1x109 114,000



 Hence most DNS focuses on low-Re applications!

Reynolds averaging

• Time averaging T

t0 

 Reynolds (1895): 1 ~ 2

U (t 0 ) 

T  U (t )dt

T

t0 

2



4

~ Instantaneous

U

3 Average

Velocity, U









2

1

t=t0

0 T T

 Time, t

0

2 5 10 15

2

Reynolds averaged equations

• Decompose velocity: ~

ui  U i  ui

 Ui is a mean component

 ui is a fluctuating component

• Substitute into differential equation:

~~

ui u j 

 U i  ui U j  u j   0

xi xi

 Average: 

U i  ui U j  u j   0 Ui  Ui ,Uiui  0

xi



xi

 

U iU j  2ui u j  0

Reynolds averaged equations

• Momentum – ensemble averaged form:

U k U i  P   U i 

     u k ui 

xk xi xk  xk 

• Similar to instantaneous form but with additional

(unknown) terms, the Reynolds stresses:



 Normal stresses:  ui 2







 Shear stresses:  ui u j

• Referred to as “RANS”

 Reynolds Averaged Navier Stokes

Boussinesq Approximation (1877)



• Defined a modelled “local eddy viscosity” or

“turbulent kinematic viscosity”, relating Reynolds

stress to mean strain rate

• By analogy with viscous stress in laminar flow:

 U i U j 2 U k 

 ij  l     ij 

 x xi 3 xk 

 j 

 Here the viscous stress related to a mean strain rate

• Hence, Reynolds stresses expressed:

 U i U j 2 U k  2

 ui u j  T     ij   k ij

.  x xi 3 xk  3

 j 

Viscosity models

• Linear Eddy viscosity models

 Algebraic (zero equation) models

 Single transport equation models

 Two transport equation models



• Industry standard k- model

 Buoyancy modifications

 Some known weaknesses

 Axisymmetric plumes

CFD - RANS & LES

CFD - RANS & LES

CFD models - advanced models

• Sprinkler sprays

 principally cooling of hot gases by water droplets

• Structures in fire

 thermal response

 mechanical response

• Flame spread

 real materials

 mixed fuels

CFD - fire specific features

• Buoyancy driven flow

 Froude number (V2/gD ) ~ 1

• Multiple and complex fuels

 Surrogates

• Soot formation

 Fuel specific

 Affects radiative heat loss

• Range of lengthscales

 Kolmogorov (1mm)

 Building-scale (100m)

CFD - code choice

• General purpose • Specialist

 sophisticated pre- & post-  developed by fire scientists

processing  limited range of other modelling

 wide range of modelling options options



 well documented  focussed on fire issues

 user support  generally less advanced pre- &

post-processing

 expensive

 but maybe simpler to use

 parallel processing

 validated against fire

 Fluent, ANSYS-CFX, Star-CD, experiments

Phoenics

 JASMINE, SOFIE,

SMARTFIRE, KOBRA-3D,

Kameleon Fire, FDS

Modelling process

Input building  Retain only sufficient detail

geometry





Generate mesh

 Mixture of manual and automatic

techniques



Apply boundary

conditions and  Ambient conditions, fire sources,

fluid/solid ventilation sources, wall properties

properties etc



 CFD solver - need to supervise

Generate solution numerical convergence



 Solution OK? - if so extract

Analyse results

required results

Thermal/mechanical analysis

• Advanced/general methods

 “based on acknowledged principles and assumptions of the

theory of heat transfer”

 consider temperature-dependent material properties

 moisture influence “may conservatively be neglected”

 validation

 on basis of relevant test results

 sensitivity analysis on critical parameters

STELA solver

Modelling hierarchy

• Simple tools

 Plume entrainments

• Zone models

 Smoke filling times

• CFD

 Smoke movement in complex spaces

 Flame spread?!

• Field-zone network models

Model exploitation

• Advanced models (CFD)

 Efficiencies possible, eliminating redundancies

 Optimise protection via whole frame analyses

 Caution is needed!

 Some “non-linear” features

 Short-hot versus long-cool

 Improve standards of FSE, through training & education

 Maintain equivalent or improved levels of safety!



• Conservative assumptions/worst cases?

 Can defeat purpose of using model!

 Progressively reduce uncertainties

Conclusions

• Fire safety engineering is maturing

 Becomes more useful as accuracy improves

 Better information on basic properties

 Overcome uncertainties in inputs

 More extensive sensitivity studies

 Improved computational procedures

 Leverage of simple model results

 Faster hardware

 Parallel processors