Flame-spray interaction modelling · 2017. 10. 9. · A. Snegirev, E. Kokovina, A. Tsoy. Coupled simulations of turbulent flame and pyrolysis of combustible material / Proc. European

Flame-spray interaction modelling

Alexander Snegirev [email protected] Fire Modeling and Flammability Group Department of Fluid Dynamics (Thermal Physics) Saint-Petersburg Polytechnic University

Objectives

To highlight ongoing research avenues

To present resent advances

To look for an overlap and synergy of research interests

This talk is therefore not strictly adhered to its title, albeit the flame-spray interaction remains the key issue

University of Brighton, UK, 10 August 2015 1

Research avenues

• Combustion physics

• Fire safety science

• Multiphase flows (under

development)

• Water spray fire suppression

• Halons, fuels

• Microscale combustion calorimetry

• Pyrolysis modeling

• Volatile oxidation

• Fire dynamics • Coupled

simulations • Fire

suppression

Fire modeling

Material flammability

Teaching Spray

modeling


What would happen during my talk if I consider all topics…


I did my best to select topics in this order…


An overview

Introduction Fire modeling Coupled

simulations Spray fire

suppression

Governing criteria

Spray model validation

Flame-spray interaction

Flame extinction

Fire suppression

modeling

Conclusions and future

work


Research methodology

Theory and modeling

Fire dynamics and smoke movement

Coupled simulations of

turbulent flame and pyrolysis

Water spray fire

suppression

Pyrolysis modeling

Experiment

Microscale combustion calorimetry

Sample cup


Publications

Snegirev A.Yu. Perfectly stirred reactor model to evaluate extinction of diffusion flame. Combustion and Flame 2015 (doi:10.1016/j.combustflame.2015.06.019)

Tsoy A.S., Snegirev A.Yu. Large eddy simulation of fine water sprays: comparative analysis of two models and codes. Thermophysics and Aeromechanics 22 (2015), In Press

A. Snegirev, E. Kokovina, A. Tsoy. Coupled simulations of turbulent flame and pyrolysis of combustible material / Proc. European Combustion Meeting – 2015, Paper P4-16, March 30 – April 2, 2015, Budapest, Hungary

Snegirev A.Yu., Tsoy A.S. Treatment of local extinction in CFD fire modeling. Proc. Combustion Institute 35 (2015) 2519-2526

Snegirev A.Yu. Fundamentals of Combustion Theory. St.-Petersburg, SPbPU Publ., 2014, 352 p. (In Russian)

Snegirev A.Yu. Generalized approach to model pyrolysis of flammable materials. Thermochimica Acta 590 (2014) 242-250

A.Yu. Snegirev, V.A. Talalov, V.V. Stepanov, A.S. Tsoy. Oxidation kinetics of pyrolysis volatiles and its implication to critical conditions of flame extinction. Proc. of the 10th International Symposium on Hazards, Prevention, and Mitigation of Industrial Explosions ISHPMIE-X (Bergen, Norway, June 9–14, 2014), 2014, pp. 71-84

A.Yu. Snegirev. Transient temperature gradient in a single-component vaporizing droplet. Int J Heat Mass Transfer 2013 (65) 80–94

A. Snegirev, V. Talalov, V. Stepanov, J. Harris. A new model to predict pyrolysis, ignition and burning of flammable materials in fire tests. Fire Safety Journal 2013 (59) 132-150

A.Yu. Snegirev, V.V. Talalov, V.V. Stepanov, J.N. Harris. Formal kinetics of polystyrene pyrolysis in non-oxidizing atmosphere. Thermochimica Acta 2012 (548) 17-26

Snegirev A.Yu., Frolov A.S. The Large Eddy Simulation of a Turbulent Diffusion Flame. High Temperature 2011 (49) 690–703 University of Brighton, UK, 10 August 2015 7

Building fires

Design fire scenario: a fire in the university classroom (2011)

Fire origin

Fire origin

Exit Time to escape


New Pulkovo terminal

45 MW design fire modeling

(SPbSPU, 2011)


Fires in the open atmosphere

City of Grozniy, 3/4/2013 Simulation (A. Tsoy, 2013)


Coupled simulations

Tight two-way coupling of turbulent diffusion flame and pyrolysis of combustible material governs fire growth rate

Time to ignition, burning rate and flame spread are affected

Positive thermal feedback determines critical conditions of flame ignition and flame extinction

Mass flux of

volatiles

predicted by

Pyropolis

Pyrolyzing material

Mass

flow inlet

BC

Net heat flux

predicted by

Fluent


Coupled simulations with FDS (ignition)

PMAA ignition and burning 50 kW/m2

The effect of material

transparency and

volumetric

absorption

E. Kokovina, A. Snegirev ECM 2015


Coupled simulations with FDS (steady burning)

Steady burning rate subject to the external heat flux (10x10 cm plate)

Predicted burning rates are considerably different for two versions of FDS code because of strong sensitivity to the flame

The code may not be capable of predicting self-sustained burning and surface flame spread

,0

flame rrnet extfuel

g g g

extfuel

gSteady self sustained

burning rate

q qq qm

h h h

qm

h


Sprinkler fire protection in a factory

Two types of fire load (liquid and solid combustible)

Ethanol

Solid fuel

Sprinklers

Prescribed flame spread velocity

Sprinkler location and flow rate in accordance with national fire codes СП5.13130.2009

Fuel mass loss rate is reduced due to wetting


Sprinkler fire protection in a factory

No fire suppression: rapid blockage of evacuation pathways

When oxygen in fire room is exhausted, the flame ejects through the openings

Fire suppression: 5 sprinklers activated

Fire is extinguished in 2-3 min

No fire protection system Fire protection system activates


Fire suppression in the auxiliary power unit

FirEx project objectives funded by Airbus (2007-2010):

APU

Ultimate • To assist in design and certification of the fire

suppression system

Specific

• To develop efficient computational methodology and software tools to predict discharge, spraying and spread of the extinguishing agent throughout the APU compartment


Modeling methodology


Fire suppression in the auxiliary power unit

APU

Halon 1301 discharge (experiment: VNIIPO, 2009)

Halon discharge in protected compartment

(modeling: SPbSPU+ChemInform,

2009-2010)


Unresolved issues

Validation “backlog”

– Lack of well documented experimental cases

– Model complexity: close interaction of all the submodels. Errors introduced by each submodel obscure overall model performance. A comprehensive validation is required

– The effect of spray refinement needs to be quantified by means of governing criteria

Objectives

– To undertake validation studies

– To assess sensitivity to model parameters

– To identify systematic irregularities and model drawbacks

– To evaluate the effects of spray fineness

– To apply in practical cases


Validation matrix

High-pressure fine water spray

Turbulent flame

Bounded fire source

Low-pressure sprinkler

Gaseous flame suppression

Spreading fire

Reduction of pyrolysis rate due to wetting

Spray fineness


Initial droplet characterization

“Immediate” atomization model: initial droplet size distribution functions. Initial mean diameter and its variance must be pre-assigned

Initial droplet size PDF is assumed at a spherical surface centered at the nozzle outlet

0

0

2

l

PV

220

0 0 00 4w

DQ V K P D P

1/32

01/350 0

1/3 1/3

0 0 0

1We

lv

l

V DdC C

D P D

2/3 2

0 050 1/3 2/3

0

v

w

D Dd

P Q

Nozzle

diameter K-factor Liquid flow rate

Pressure drop Initial droplet

velocity

Initial median

droplet diameter

Depends on the

nozzle geometry


Modeling liquid injection

Combined log-normal and Rosin-Rammler

Hollow or full cone

Computational particles contain a number of real droplets

50

50

50

50

ln11 erf ,

2 2

1 exp ln 2 ,

v

v

v

v

d dd d

R dd

d dd

2

2 ln 2

dv50 = 0.102 mm


The effect of spray refinement: an example

Instantaneous droplet locations colored by droplet temperature

Instantaneous vapor mole fraction

LES predictions of vaporizing water spray (full-cone, cone angle 120°, initial droplet velocity 30 m/s, ambient air temperature 20°C, initial droplet temperature 20°C, flowrate 10 l/min)

50 1=50 , S 1vd 1

50 1=200 , S 10vd 2

50 1=500 , S 10vd

1

stopping distance

spray size

S dL

H


The effect of spray refinement: an example

Instantaneous droplet locations colored by droplet temperature

Scatter plots of the instantaneous droplet velocity (relative to the gas flow) versus droplet temperature colored by droplet residence time (increased from blue to red). Scatter size scales with droplet diameter

50 1=50 , S 1vd 1

50 1=200 , S 10vd 2

50 1=500 , S 10vd

LES predictions of vaporizing water spray (full-cone, cone angle 120°, initial droplet velocity 30 m/s, ambient air temperature 20°C, initial droplet temperature 20°C, flowrate 10 l/min)


Spray structure

1 2S 1, S 1

Vu

orin

en e

t a

l.,

FT

C 2

011 (

86

) 5

33

-561

1S 1

1S 1

Stopping distance, Ld

Cone

section

Neck

Turbulent

crown

2

18

lvelo

g

d

2 1velo i

g

gS

u

Spray structure depends on the stopping distance

Fine droplets are deflected by the gas flow if , where

Spray fineness criteria:


Integrated evaporation rate in a starting jet

Coarse spray => low vapor concentration => evaporation rate increases with droplet surface area

Fine spray => high vapor concentration => no increase of evaporation rate with spray refinement

Optimum droplet dispersion exists

About 0.2 mm for fire suppression by fine water sprays


Experiment by Ditch and Yu (2008)

Ditch B., Yu H.Z. Fire Safety Science 9, 2008, 541-552

– Nozzle diameter 0.7 mm

– Cone angle 76°

– Pressure 20.68 bar

– Measurements of liquid flow rates and mean droplet diameters (dv50, d32) at the distances of 0.4 and 0.62 m downstream the nozzle

42 cm

60 cm


The effect of numerical schemes

FDS5

FDS6

Gas velocity

Vapor

concentration


The effect of numerical schemes

z = 0.4 m z = 0.62 m

Axial distributions of total liquid/vapor flow rates

Radial distributions of liquid flow rates

FDS5

FDS6

Tsoy A., Snegirev A.

Thermophysics and

aeromechanics 2015

Vol. 22 (accepted)


Experiment by Santangelo (2010)

P.E. Santangelo, Exp Thermal Fluid Sci 34 (2010) 1353–1366

Nozzle diameter 0.5 mm

Pressure 60, 70, 80 bar, initial droplet diameter dv50 ~ 90 μ

Measured liquid flow rates and mean droplet diameters (dv50, d32) at the distance of 1 m downstream the nozzle

FDS5 FDS6

Vapor mole fraction


Simulation results

FDS5: Large-scale turbulence is under-resolved, too narrow spray is predicted, axial velocity is heavily over-estimated

FDS6: Better resolution of large-scale eddies, reasonable agreement with the experimental radial distributions

Tsoy A., Snegirev A. Thermophysics and

aeromechanics 2015 Vol. 22 (accepted)

60 bar 70 bar 80 bar


Unconfined buoyant turbulent diffusion flames

Small-scale and large-scale flames are considered

Experimental data

Q , kW D , m *

1/2 5/2

P

QQ

c T g D

* *2/5D Q D ,

m x , m

*D x

Schwille and Lueptow, 2006

15 0.18 1.0 0.18 0.018 9

Sandia, 2007 2500 2.00 0.4 1.38 0.075 18


15 kW burner fire

Schwille & Lueptow experiment (2006)

15 kW, 18 cm diameter burner, methane

dv50 = 0.6-0.8 mm

3–11.7 l/min

120º cone angle, 1.6 m height

Spray cone is much wider than the flame base

Investigated numerically by means of URANS (Snegirev, Lipjainen, 2008) and LES (Snegirev et al., 2010)


Experiment: flame height depends on water flow rate

15 kW burner fire

Undisturbed flame

(Schwille & Lueptow,

2006)

7.57 l/min

9.65 l/min 11.17 l/min

Nozzle

7.57 l/min

1.2 mm

(simulation)


15 kW burner fire: FDS simulation results

Water flow rate 7.57 l/min. The effect of spray fineness:

Time after nozzle activation 4 s 8 s 30 s

dv50 = 1.2 mm

dv50 = 0.9 mm

dv50 = 0.6 mm


Transient dynamics of flame-spray interaction

260 kW, 1 m diam., wood volatiles CH2O, mesh 88×88×128, water flow rate 10 l/min

Coarse spray, dv50 =

0.50 mm

Medium spray, dv50 =

0.20 mm

Fine spray, dv50 = 0.08 mm

Time step 0.05 s


Regimes of flame-spray interaction

Spray fineness number distinguish qualitatively different regimes of spray interaction with the flame

Coarse spray: S1 = O(1) => Spray drag number, SD

Fine spray: S1 << 1 => Spray momentum number, SM

Extinguishing a 260-kW flame above a 1-m diameter burner (0.2 s after nozzle activation): a) — coarse spray, dv50 = 0.5 mm; b) — medium spray dv50 = 0.2 mm; c) — fine spray 0.08 mm.

Wide-angle nozzle (120° cone), nozzle height 3.0 m above the floor, water flow rate 10 l/min. LES, mesh 88×88×128. Light colour surface — vapour mole fraction 0.01.

Coarse spray: Drops penetrate inside the flame

and evaporate there =>

Fine spray: Vaporized jet suppresses the flame provided jet momentum is high enough

S1 = 0.6 S1 = 0.15 S1 = 0.035

a) b) c)

DS spray droplet drag

plume momentumMS

spray momentum

plume momentum


Fire Laboratory for Accreditation of Models and Experiments (FLAME) – Sandia Labs, NM, USA (2007)

2.5 MW pool fire, 2 m diameter (JP8 – C11H21)

Nozzle: 30º cone angle, 5 m height

Spray directed to the pool center (90° and 45°)

Pressure drop affects water flow rate, initial droplet velocity and initial droplet diameter (nozzle diam. D0 = 8.74 mm)

Critical pressure (critical flow rate) separates regimes with and without flame extinguishment

Large-scale flame suppression

Nozzle

Fuel pool


FDS simulation results (Tsoy et al., Proc. ISFEH7, 2013)

Large-scale flame suppression

Low pressure

spray

High pressure

spray


Need to improve local extinction model

Time to extinction decreases as the pressure increases

Agreement with the experiment improves with grid refinement and increased number of computational droplets

Flame extinction is not resembled

Determined by reduction

of flame height Local extinction

model

Nozzle activates

Flame is down


Modeling of extinction in diffusion flames

Activation energy asymptotics for the strained laminar diffusion flamelet (Lecoustre et al, 2010; Narayanan et al, 2011; Vilfayeau et al, 2015)

Adiabatic PSR extinction model: Magnussen et al, 80s; Hewson et al, 2003; Snegirev & Tsoy, Proc. Combust. Inst. 35 (2015)

– Extinction due to excessive strain rate (high-strain limit only)

New non-adiabatic PSR extinction model (Snegirev, 2015):

– Extinction due to excessive strain rate (high-strain limit = blow-off)

– Extinction due to radiative losses from flame (low-strain limit = quenching)

Snegirev, Tsoy, PCI 35

(2015) 2519-2526

Snegirev, Combustion

and flame, 2015


The PSR extinction model

Only cell-averaged quantities are available

Reaction zone is not resolved; it occupies a small part of the cell volume

The reaction zone is treated as a PSR fed by the fuel and oxidizer streams

Temperature and composition in the feeding streams is expressed via the cell-averages

Residence time in the PSR is governed by the local (subgrid) strain

Single-step global oxidation reaction, finite rate kinetics

Critical subgrid strain exists causing extinction in the PSR

Such a critical strain depends on cooling and dilution, thereby incorporating all the extinction mechanisms


PSR theory to model local extinction

Fuel and energy balance at steady state

Reaction rate

Max flame temperature

Mixture specific heat

Flame temperature at steady state

,0fuel fuel fuel resY Y r 0 loss resh h q

exp

fuel oxn n

fuel fuel oxfuel

fuel ox

M Y Y Er A

M M T

R

,0 ,0

max 0

min ,

ˆ

fuel ox ox lossC res

P

Y Y hT T

q

c

, ,ˆ ˆ ˆ1P P no dil dil P dil dilc c Y c Y

0

ˆ

exp

fuel ox

fuel ox

n n

fuel P res

res n nfuel ox C fuel

l

o

s

x

osM c T T

AM M h Y T Y T E T

q R

The residence time is

coupled with the local

strain


PSR theory to model local extinction

Flame temperature at steady state

Critical condition

Extinction occurs, if the subgrid mixing time is less than the critical residence time

Flame temperature at extinction is not constant, being dependent on the amount of a diluent (EGR in engines and furnaces, fire suppressant)

1

0

*

max exp

n

res n

T T TA

T T E T

R

2

max 0

1 1

ext ext ext ext

E n n

T T T T T T

R Extinction temperature

is not constant!

*res A

0extT T

d dT

Classical S-curve


Model calibration

Three model parameters: n, A, E

Counter-flow flame measurements of extinction strain and temperature can be used to derive n and A for the given E

Methane-air: Sext = 430 s-1, Text = 1760 K

E = 121 kJ/mol (Coffee et al., 1983,

laminar flame speed modeling)

PSR model: n = 1.69

Heptane-air: Sext = 460 s-1, Text = 1620 K

E = 110 kJ/mol (Seiser et al., 1998,

counter-flow diffusion flame)

PSR model: n = 2.01


Implementing local extinction in CFD

Critical flame temperature (CFT) model

Perfectly stirred reactor (PSR) model

2 2O ,0 ,O

max 0ˆ

C

P

Y hT T

c

max 0, , ,extT T T E n

*

1ext extT

A

ext SGSC 1

SGS extS C

Estimating maximum

flame temperature

Solving equation for

the extinction temperature

Evaluating residence

time at extinction

Extinction?

2 2O ,0 ,O

max 0ˆ

C

P

Y hT T

c

CFTT max CFTT T

Estimating maximum

flame temperature

Assume constant

critical flame temperature

Residence time is not considered

Extinction?


Flammability diagram: comparison to the CFT model

Conventional model

New model

2

2

0,

O , 0,

,O

P CFT ext

ext ext

C

c T TY T

h

Polystyrene volatiles

(oxidation kinetics is

derived from MCC data)

2

2

max, 0,

O , 0,

,O

,P ext ext

ext ext

C

c T S TY T S

h

Local strain


Implementing in CFD

Schwille & Lueptow experiment (2006)

FDS 6

7.57 l/min, dv50 = 1.2 mm

(simulation)

Nozzle

15 kW, 0.18 m

Methane burner


The non-adiabatic PSR extinction model

,0fuel fuel fuel resY Y r

0 loss resh h q

04 4

0

ˆ14

P

loss

loss

c T Tq T T

2 2 2 20 CO CO CO CO H O H O soot vP X T X T X T C f T

exp

fuel oxn n

fuel fuel oxfuel

fuel ox

M Y Y Er A

M M T

R

2 2,0 O ,0 O

max 0

min ,

ˆ

loss refuel

P

scY Y s h qT T

c

,0 ,0 maxDa , ,fuel fuel ox resr Y Y T


Fundamental flammability limit (CH4 and C7H16)


Fundamental flammability limit (pyrolysis volatiles)

Two-distinct flammability limits (high- and low-strain) merge at the fundamental limit (MEC point)

The fundamental limit is obtained from the first principles (competition of the finite rate kinetics and radiative losses from the reaction zone)

Existing techniques typically measure either high- or low-strain limit. Lack of the experimental data for the fundamental limit, particularly for the pyrolysis volatiles

PMMA


Fire suppression modeling

Thermal activation of a sprinkler:

Reduction of pyrolysis rate due to surface wetting

2

RTI RTI RTI

gg mg

Heat exchange with Heat losses Cooling byfire plume to the mount water droplets

uT T T TdTu С С

dt

actT T

,0

,

expfuel fuel w w

Mass ofBurning ratewaterno wetting

m t m t e m t dt


Suppression of radially spreading fire

Sprinkler activation

– Conventional (thermal)

– Gefest Ltd., St-Petersburg: group enforced activation (thermal activation of the first sprinkler followed by group assignment and enforced activation of the group after 10 s delay)

Modeling

– FDS 5.5

– Large compartment, ceiling clearance 5 and 10 m

– Fire load: 300 kW/m2, max surface area 6x6 m2

– Flame spread velocity 0.01 m/s (α = 0.094 kW/m2, “fast” NFPA204M)

– Flow rate per sprinkler 29 l/min

– Reduction of pyrolysis rate by water ew = 0.3 (m2/kg)/s

– RTI = 140 m1/2s1/2, Tact = 68°C

Heating element

Ignition in

the centre


Conventional sprinkler activation

Conventional (thermal) sprinkler activation cannot prevent fire spread, the heat release rate increases to a high value. 10 m ceiling clearance


Group enforced sprinkler activation

Group enforced activation suppresses material gasification and stops burning, fire is localized and quenched. 10 m ceiling clearance


Radially spreading fire suppression: the effect of spray refinement

dv50 = 500 μ dv50 = 100 μ

Thermal activation, 5 m ceiling clearance

When the fire growth rate is too high, finer spray cannot suppress fire, being less efficient than the coarse spray

Spray momentum is

high enough to deflect

the spray


Radially spreading fire suppression: the effect of cross-wind

Due to water spraying beyond the fire origin, suppression efficiency is reduced

Advantage offered by the group activation is also reduced


Group activation Thermal activation

No cross-wind

Cross-wind 1.2 m/s

Radially spreading fire suppression: the effect of cross-wind

Скорость бокового ветра: слева – 0 м/с, справа – 1.2 м/с


Group activation Thermal activation

Heat release

rate

Number of

activated

sprinklers

Sprays in selected engineering applications

Fire suppression IC Engine Pressurized liquid

discharge

Liquid Water Liquid fuels LPG, LNG,

Halocarbons etc.

Initial mean droplet size

~0.05 – 1.0 mm ~0.004-0.02 mm 0.01-0.04 mm

Nozzle diameter ~1-10 mm ~0.2 mm Unknown in

accidental releases, ~20 mm in FSS

Atomization mechanism

Mechanical Mechanical Flashing

Thermal state of the liquid

Sub-boiling Sub-boiling Superheat, possibly

supercritical

Ambient atmosphere

Atmospheric air, hot combustion

products

Hot combustion products

Atmospheric air, hot combustion

products

Jet flow Sub-sonic,

~10-100 m/s Trans-sonic,

~100-500 m/s Super-sonic, choked,

~50 m/s


The diesel spray

Injector make Bosch (1 hole)

Hole diameter 0,2 mm

Fuel volume 30 mm3

Fuel density 808 kg/m3

Gas temperature 540 K

Injection pressure 160 MPa

Wall temperature 360 K

Chamber diameter 50 mm

Chamber length 80 mm

0.3 0.4 0.5 0.6 0.7 ms

2 MPa 6 MPa

2 MPa, RR-LN, dv50 = 8μ, γ = 2.25

Crua et al., Univ of

Brighton, UK, 2008


Conclusions

Use of either commercial or open-source software to simulate turbulent evaporating sprays often entails considerable errors in gas/droplet flow rates and velocity distributions

In LES, it is crucially important to ensure capability of the numerical scheme to replicate the large scale turbulent fluctuations

Failure to do so causes underestimated radial jet spread, overestimated jet velocity at the axis. In case of co-axial spray and flame-induced flows, capability of the spray to suppress fire is overestimated

Beyond the droplet movement and evaporation, spray-flame interaction modeling encounters the need to consider:

– Flame extinction at a subgrid level – the new model is proposed

– Reduction of solid fuel gasification due to wetting – spray-material coupling


Conclusions

Despite of dramatic difference of quantitative characteristics of sprays in different engineering applications, a similarity exists in their structure and dynamics

Such a similarity can be interpreted in terms of two dimensionless criteria which can also be considered as the spray fineness criteria

– S1 is the ratio of a stopping length for the characteristic droplet to the external length scale (e.g. the distance to the obstacle)

– S2 is the ratio of the droplet sedimentation velocity to that of the external gas flow

Two distinct criteria govern the spray-flame interaction

– In the course spray regime (S1 is of order of 1) the governing criteria is the ratio of the spray droplet drag to the plume momentum

– In the fine spray regime (S1 << 1) the governing criteria is the ratio of the spray momentum to the plume momentum

Optimum initial droplet diameter corresponds to the value below which further spray refinement does not result in a considerable increase of the evaporation rate


Future work

Non-combustible sprays

Flame-spray coupling

Flame-material coupling

Spray-material coupling

Combustible sprays

Ignition and extinction

Turbulent flame

modeling


Documents

Flame-spray interaction modelling · 2017. 10. 9. · A. Snegirev, E. Kokovina, A. Tsoy. Coupled simulations of turbulent flame and pyrolysis of combustible material / Proc. European