Вы находитесь на странице: 1из 69

2011 ANSYS, Inc.

October 4, 2011 1
Combustion Modelling
Tools in CFD
Present and Future
Phil Stopford
ANSYS UK
Presentation to the DANSIS
Seminar on Combustion and
Reactive Flows, DONG Energy
Power, Denmark, 5
th
Oct. 2011
2011 ANSYS, Inc. October 4, 2011 2
Contents
Overview of combustion and reacting flow models
Applications illustrating recent trends:
Scale resolving turbulence models
Coupled models of reacting flows
Combustion models in fire simulation
Multiphase reacting flows
Recent developments
Conclusions
Questions
2011 ANSYS, Inc. October 4, 2011 3
Progress since 1980
Computing power increase
How big a calculation can I run over night?
- In 1980: Steady RANS case with 5k cells on mainframe supercomputer
- In 2011: Transient LES case with 10m cells on 500 parallel nodes
CFD software improvements
- Over-diffusive numerical methods (e.g. SIMPLE, RANS) superseded
- Better resolution of transient phenomena
- Coupled solvers giving better convergence
Physical/Chemical knowledge increase
- Better insights into turbulent flame dynamics, chemical mechanisms
- Laser-based non-intrusive diagnostics
2011 ANSYS, Inc. October 4, 2011 4
Fast Chemistry Combustion Models
Eddy dissipation
Uses reduced chemistry reaction mechanisms
Overall rate is minimum of turbulent mixing rate and chemical reaction rate
Applicable for all types of flame
Non-premixed
Use mixture fraction, f, and assumed PDF for fluctuations of f, instead of solving species
transport equations and reacting rates for equilibrium chemistry
Laminar flamelets for moderate non-equilibrium chemistry
Premixed
Reduce chemistry to reaction progress variable, c c equation model
Burning Velocity Model (BVM)
Also called Turbulent Flame speed Closure (TFC) model
Requires a turbulent flame speed correlation, e.g. Zimont, Peters, Glder
Enhanced Coherent Flame Model (ECFM)
Solves a transport equation for the flame surface per unit volume
Partially premixed
Combine non-premixed and premixed models
Assumptions in both models apply
2011 ANSYS, Inc. October 4, 2011 5
Finite Rate Chemistry Models
Laminar Finite Rate Model
Stiff chemistry solvers
Applicability of the finite-rate model
Flow regime: Laminar flow
Chemistry: Finite-rate chemistry
Flow configuration : Premixed, non-premixed, partially premixed
Eddy Dissipation Concept (EDC) model
General model of turbulence-chemistry interaction
Reactions take place in hot turbulent micro eddies
Compositional Transport PDF
PDF transport equations solved by Monte Carlo method
Mean reaction rate
dT dY dY P w w
N k k 1
0
1
0
1
0

}} }

=
2011 ANSYS, Inc. October 4, 2011 6
Scale Resolving
Turbulence Models
2011 ANSYS, Inc. October 4, 2011 7
Evolution of Turbulence Models
URANS
Gives unphysical single mode unsteady behavior
LES (Large Eddy Simulation)
Too expensive for most industrial flows due to high resolution
requirements in boundary layers
DES (Detached Eddy Simulation)
First industrial-strength model for high-Re with LES-content
Increased complexity (grid sensitivity) due to explicit mix of two
modelling concepts
SAS (Scale-Adaptive Simulation)
Extends URANS to many technical flows
Provides LES-content in unsteady regions
Von Karman length scale occurs naturally in formulation
WMLES (Wall Modelled LES)
Analytic wall function for viscous sub-layer
Zonal LES
Only apply LES in the critical flow region RANS elsewhere
URANS
SAS-URANS
2 2
/
/
y U
y U
L
vK
c c
c c
=k
2011 ANSYS, Inc. October 4, 2011 8
Zonal LES: Model description
Model workflow
(U)RANS solution
User-defined LES zone
k
RANS
2011 ANSYS, Inc. October 4, 2011 9
Zonal LES: Model description
Model workflow
(U)RANS solution
User-defined LES zone
Source term in k-eqn. inside the LES zone:
C (k
LES
k
RANS
)
C big factor
Ensures LES eddy
viscosity in LES zone
k
RANS
k
LES
2011 ANSYS, Inc. October 4, 2011 10
Zonal LES: Model description
Model workflow
(U)RANS solution
User-defined LES zone
Source term in k-eqn. inside the LES zone:
C (k
LES
k
RANS
)
C big factor
Ensures LES eddy
viscosity in LES zone
Momentum source
at RANS-LES interface
to enforce fluctuations
Activated
simultaneously
k
RANS
2011 ANSYS, Inc. October 4, 2011 11
Zonal LES: Model description
Model workflow
(U)RANS solution
User-defined LES zone
Source term in k-eqn. inside the LES zone:
C (k
LES
k
RANS
)
C big factor
Ensures LES eddy
viscosity in LES zone
Momentum source
at RANS-LES interface
to enforce fluctuations
Low-turbulence part is
cut off automatically:

t
sensitive limiter
k
RANS
2011 ANSYS, Inc. October 4, 2011 12
Zonal LES: Model description
Near-wall modification: wall modelled LES ( WMLES )
Uses RANS in the inner part of the log layer
Enforces correct (high) turbulent K.E. near the wall
Mesh density at wall independent of Reynolds number
Practical LES method for industrial internal flows
2011 ANSYS, Inc. October 4, 2011 13
SAS: Gas-Fired Combustor Validation
Refs. Schildmacher, K.-U., R. Koch, R., ASME-Paper GT2003-38644, 2003, and
Schildmacher, K.-U., Koch, R., Wittig, S., Krebs, W., Hoffmann, S., ASME-Paper 2000-GT-0084, 2000.
Single burner configuration:
Swirl burner of an industrial gas turbine
Mounted in rectangular combustion chamber
Lean premixed main inlet (methane/air,
preheated)
Axial dilution air (preheated)
Atmospheric pressure
Experimental data:
Isothermal: x = 44, 138, 257 mm
Reacting: x = 42, 79, 103, 136, 259 mm
Temperature, velocity
Simulation: SAS-SST, EDM, 3.6M cells, 0.1s
2011 ANSYS, Inc. October 4, 2011 14
Simulation Results
2011 ANSYS, Inc. October 4, 2011 15
Axial Velocity
1.5 mm
2.5 mm
5 mm
7.5 mm
2011 ANSYS, Inc. October 4, 2011 16
Temperature
5 mm
10 mm
15 mm
20 mm
2011 ANSYS, Inc. October 4, 2011 17
Methane Mass Fraction
5 mm
10 mm
15 mm
20 mm
2011 ANSYS, Inc. October 4, 2011 18
Flame Front Dynamics
Iso-surface of premixed flame front coloured
by temperature
2011 ANSYS, Inc. October 4, 2011 19
Combustion Instabilities
In gas turbines applications, combustion instabilities can be described as
unsteady flow and pressure fields motion driven by the combustion process.
These instabilities are one of the most challenging problem in the development
of gas turbines (especially for land-based fully premixed systems)
The negative effects of combustion instabilities are varied:
Vibration (increased metal fatigue possible structure failure)
Enhanced heat transfer
Reduced operating performance
Deterioration of the pollutant emission performances
Pressure waves impact the:
Turbine
Compressor
Fuel feed
2011 ANSYS, Inc. October 4, 2011 20
7
th
Framework Programme of European Community
Marie Curie Initial Training Network - LIMOUSINE:
Limit cycles of thermo-acoustic oscillations in gas turbine combustors
Simulation of fluid-structure interaction
Project coordinator, Dr Jim Kok, University of Twente (NL)
Keele University (Staffordshire, UK)
Imperial College (London, UK)
CERFACS (Toulouse, France)
Brno University of Technology (CZ)
University of Zaragoza (Spain)
DLR (Stuttgart, Germany)
Ingenieurbro fr Thermo-Akustik (Munich, Germany)
Siemens Power Generation (Mlheim, Germany)
Electrabel/Laborelec (Brussels, Belgium)
ANSYS (UK)
Duration: 4 years from 1
st
Oct 2008
3 experienced researchers (post doc) and 17 early-stage researchers
2011 ANSYS, Inc. October 4, 2011 21
LIMOUSINE Test Burner
Rijke tube test burner at 4 different laboratories
2-D slice of the whole burner
Power ratings: 40 and 60 kW
Air factors, : 1.2, 1.4, 1.6 and 1.8
Structured hexahedral mesh of 37,500 elements
SAS-SST and LES scale resolving turbulence
models
Two-step EDM and BVM combustion models
Predictions:
short flame stabilized on the wedge
strong pressure oscillations for < 1.6
2011 ANSYS, Inc. October 4, 2011 22
Transient CFD Results
Two peaks at frequencies of
about 240 and 700 Hz
Lower intensity and quenching
for > 1.4
2011 ANSYS, Inc. October 4, 2011 23
CFD Results Profiles of resonant modes
Downstream region drives the instability
Weak coupling with upstream wavelength
mode reduces the frequency slightly
Higher resonant frequency consistent with
wavelength mode
2011 ANSYS, Inc. October 4, 2011 24
CFD Results v Experiment
power
[kW]
air factor
[-]
pressure oscillation frequencies
[Hz]
simulation simulation
end corrected
experiment 1 experiment 2 analytical
downstream
analytical
full
40
1.2 247 234 230 235 258 186
1.4 242 229 217 217 246 183
1.6 238 226 stable stable 241 181
1.8 216 205 stable stable 232 178
60
1.2 280 266 268 268 259 187
1.4 251 238 248 262 249 184
1.6 228 216 stable stable 243 182
1.8 225 213 stable stable 236 179
power
[kW]
air factor
[-]
pressure oscillation amplitude
[Pa]
simulation experiment 2
40
1.2 1539 743
1.4 1596 862
60
1.2 2940 3138
1.4 2312 2305
Lowest resonant frequency in
reasonable agreement between
CFD, experiment and wavelength
mode in downstream section (within
10%)
Amplitude of resonance also of
correct magnitude
2011 ANSYS, Inc. October 4, 2011 25
Modal
Static
Structural modal frequencies
[Hz]
full liner
bending
back-and-
forth
117 152
bending
lateral
255 331
torsion
531 606
2
nd
bending
638 -
plate
672 667
2
nd
plate
734 728
Conclusions:
Symmetric mode fundamental at 670 Hz
Window structure does not influence the plate modes
Structural modelling
2011 ANSYS, Inc. October 4, 2011 26 Work
package:2.5
Tufano/ Stopford
Two-way coupling:
1) Fluid to solid force transfer
2) Solid to fluid strain transfer
Solid deformation ~ microns does
not affect fluid flow in this case
Results:
Structure response is dominated by
wavelength forcing pressure at 670 Hz
as measured in experiment
Definition of damping is important
Fluid-Structure Interaction
2011 ANSYS, Inc. October 4, 2011 27
Double Skin Impingement
Cooled Combustor
Main Burner
Pilot Burner
PreChamber
Radial Swirler
DLE Burner and Combustor Technology
Courtesy of Siemens Industrial Turbomachinery Ltd.
2011 ANSYS, Inc. October 4, 2011 28
Transient CFD analysis
Combusting cases:
Case 1 without casing
Case 2 with casing
Numerical models:
2-step Eddy Dissipation Model
SAS turbulence model
Compressible solver
without casing
Case 1:
9M tet/prism elements
with casing
Case 2:
16M tet/prism elements
2011 ANSYS, Inc. October 4, 2011 29
Transient Combusting Simulation
Pressure
Temperature
Case 2: Aerodynamic and combustion instability in can
induces pressure fluctuations in casing
2011 ANSYS, Inc. October 4, 2011 30
Effect of Casing
Casing influence
Without casing
With casing
2011 ANSYS, Inc. October 4, 2011 31
Precessing Vortex Core (PVC)
A PVC of a double helical structure has been
identified in non-reacting case
The frequency based on Strouhal number is 411
Hz compared to the CFD prediction of 432 Hz
PVC is NOT the most serious source of
instability as it is almost destroyed by
combustion
CFD predictions, non-reacting case
Experimental results, non-reacting case
Data from SGT-100 tested at German Airspace Centre DLR
2011 ANSYS, Inc. October 4, 2011 32
Comparison with engine test data
Good agreement with the
experimental data.
Major shift in frequency
assumed to be due to
simplifications
Wide frequency peak in
CFD is due to limited
number of running cycles
Ref. G Bulat et al, Paper GT2009-59721, Proc. ASME TURBO Expo 2009: Power for Land,
Sea and Air, June 8-12, 2009, Orlando, FL.
2011 ANSYS, Inc. October 4, 2011 33
Coupled Models
2011 ANSYS, Inc. October 4, 2011 34
Coupling CFD to a boiler model
Modern 800 MW
e
coal-fired boilers in South Korea
must manually optimise burner and furnace
settings for each coal blend
CFD model can show how heat flux changes with
burner/furnace settings
1-D boiler model will give plant efficiency as function
of heat fluxes
Therefore models
must be coupled to
simulate the
complete plant
Courtesy of KEPRI and KOSEP
2011 ANSYS, Inc. October 4, 2011 35
Coal Furnace/Boiler Interface
Link made between CFX and a steam-side boiler model, PROATES Off-Line from
E.ON
Wall Heat
Flux
Wall
Temperatures
CFX
PROATES
Courtesy of E.ON Engineering
2011 ANSYS, Inc. October 4, 2011 36
CFX Model
Conventional CFX pulverised coal model
Lagrangian particle transport model
One-step devolatilization
Inhomogeneous char oxidation
Discrete transfer radiation
NOx post-processing
Coal sulphur model
Computational time: 24 hours on 2 dual processors (Xeon 5500s)
2011 ANSYS, Inc. October 4, 2011 37
Validation Results
Plant data PROATES CFX
Final S/H steamout pr. (bar)
Final S/H steamout temp. ()
Final R/H steamout pr. (bar)
246.2
566
40.46
246.3
566
40.54
Final R/H steamout temp. () 567 579
Evaporator steamout temp. () 411 405
Final R/H spray water flowrate (kg/s) 4.1 7.8
After Secondary Economiser:
Water out temp. () 336 336
Gas exit temp. () 380 378 387
Gas exit O
2
(molar fraction) 0.0341 0.0356 0.0364
Gas exit CO (molar fraction) 8 10
-6
3 10
-6
Gas exit SO
2
(molar fraction) 1.79 10
-4
1.82 10
-4
Gas exit NO (molar fraction) 2.19 10
-4
2.19 10
-4
2011 ANSYS, Inc. October 4, 2011 38
Measured and Predicted Heat Absorption
Heat absorption (MW) Plant data PROATES
Final S/H tube bank 137 137
Platen S/H tube bank 139 149
Division S/H tube bank 294 287
Secondary Economiser tube bank 124 125
Final R/H tube bank 158 156
Lowtemp. R/H tube bank 198 199
Water wall 674 678
Total 1,724 1,731
2011 ANSYS, Inc. October 4, 2011 39
CFD Results & Summary
Combined model used fully-coupled, in one-way coupling mode or in standalone mode
as appropriate
Software currently used for:
What-if questions
Troubleshooting
Engineer training
Ref. H-Y Park et al., Fuel, Vol. 89, Issue 8, pp 2001-2010, Aug 2010
2011 ANSYS, Inc. October 4, 2011 40
Fire and Structural Integrity
U.S. Federal Building and Fire Safety Investigation of the World Trade Center Disaster, NIST J. Hill
http://www.nist.gov/public_affairs/ncstmin_dec2-3.htm

1
MN
MX
CTRUSS-13f (Gravity + temperature w/ const seq : static + dynamic analysis)
-42.11
-37.357
-32.603
-27.849
-23.095
-18.342
-13.588
-8.834
-4.081
.673211
DEC 2 2004
12:45:04
NODAL SOLUTION
STEP=28
SUB =10
TIME=1800
UZ (AVG)
RSYS=0
DMX =42.222
SMN =-42.11
SMX =.673211

Vertical displacement contour at 700 C
Comprehensive investigation of the collapse of the
World trade center have revealed the Fire-Induced
Thermal Stress and Structural Failure Analysis.
Fire below steel I-beam and concrete roof as a
two-way fluid-structure interaction
J.Penrose & Y. Sinai, Interflam2010
2011 ANSYS, Inc. October 4, 2011 41
Coupling of ANSYS to Process Simulation
Software
APECS: Advanced Process Engineering
Co-Simulator
Funded by US DOE with partners
including NETL, Aspen Technology
and ALSTOM
Two-way data exchange through
standard CAPE-OPEN interface
Compliant with Aspen Plus, Aspen
HYSYS, ChemCad, COFE, gPROMS,
ProSim Plus, UniSim Design,
INDISS, PRO/II
Reduced order models based of neural
networks and principal
component analysis
Examples of CFD models
linked to Aspen Plus
2011 ANSYS, Inc. October 4, 2011 42
Combustion models on
fire simulation
2011 ANSYS, Inc. October 4, 2011 43
Droplet trajectories coloured by residence time
Courtesy of Nanomist Systems LLC, US
Large water droplet (150m) mass fraction
Efficiency of fire suppression device is critical
to minimize fire impact as soon as it is detected
Droplet size modeling in possibly complex
geometry are simulated by our users in order to
optimize the location and the effectiveness of
fire suppression equipment
Temperature iso-surfaces and droplet trajectories before
extinction, in a ships machinery space
British Crown Copyright 2007/MOD.
Published with the permission of the Controller of Her Britannic Majesty's Stationery Office
Fire Suppression
2011 ANSYS, Inc. October 4, 2011 44
European Commission Firenet project
Combination of modified EDM and detached
eddy simulation (DES) turbulence model
Good agreement with experiment
Horvat et al, Comb. Sci. Tech. 2008
Animation
Temperature, fireball
Simulation of Backdraft
2011 ANSYS, Inc. October 4, 2011 45
Multiphase Reacting
Flows
2011 ANSYS, Inc. October 4, 2011 46
Two stage, up flow, prototype entrained flow gasifier
Models used in the simulation
Discrete Phase: Lagrangian particle transport
Turbulence: Standard k-c model
Radiation: Discrete Ordinates model
Reactions: Finite Rate/Eddy Dissipation model
Stochastic tracking: Discrete Random Walk model
Operating pressure 2.84 MPa
Boiling point of water 502K
Coal, water and oxygen inlets
Oxygen + Nitrogen 2 x 11.44 kg/s, 440K
Oxygen mass fraction 0.944
Fuel (Combustible Discrete Phase) 2 x 10.93 kg/s, 450K
Water (Evaporating Discrete Phase) 2 x 4.53 kg/s, 450K
Coal, water inlet
Fuel (Combustible Discrete Phase) 6.17 kg/s, 450K
Water (Evaporating Discrete Phase) 2.56 kg/s, 450K
Outlet Pressure outlet
Post processing
surface
Entrained Flow Gasifier
2011 ANSYS, Inc. October 4, 2011 47
Heterogeneous (particle surface) reactions
Char oxidation
C<S> + O
2
CO
2
CO
2
gasification
C<S> + CO
2
2 CO
H
2
O gasification
C<S> + H
2
O CO + H
2
H
2
gasification
C<S> + 2 H
2
CH
4
Reactions
Homogeneous reactions
CO combustion
CO + 0.5 O
2
CO
2
Water-gas shift
CO + H
2
O CO
2
+ H
2
CO
2
+ H
2
CO + H
2
O
H
2
combustion
H
2
+ 0.5 O
2
H
2
O
CH
4
combustion
CH
4
+ 1.5 O
2
CO
2
+ 2 H
2
O
CH
4
reforming
CH
4
+ H
2
O CO + 3 H
2
CO + 3 H
2
CH
4
+ H
2
O
Tar combustion
TAR + CO n CO
2
Volatile break-up (Volumetric reaction)
Volatile a CO + b H
2
S + c CH
4
+ d H
2
O + e H
2
+ f N
2
+ g TAR
2011 ANSYS, Inc. October 4, 2011 48
Test Results
Carbon Monoxide (CO)
Carbon Dioxide (CO
2
) Water Vapor (H
2
O)
Static Temperature
Species Mass Fraction
CO 0.614
H
2
O 0.285
N
2
0.028
CH
4
0.039
H
2
0.005
CO
2
0.014
H
2
S 0.015
Syngas Composition at outlet
2011 ANSYS, Inc. October 4, 2011 49
DOE-NETL: Gasification Modeling using Euler-Granular
Model
Syamlal and Bissett (1992) have reported gasification modeling with
step-by-step processes
Included drying, devolatilization, tar cracking, steam gasification, CO
2
gasification, methanation, water-gas shift, char combustion reactions
Gasification UDF for Euler-Granular model
Developed with funding from NETL
Based on the work reported by Syamlal and Bissett (1992) and Wen et. al.
(1982)
H
2
and CO combustion reactions also included
Used heterogeneous stiff chemistry solver of Fluent12 to take care of the
stiffness of these reactions
2011 ANSYS, Inc. October 4, 2011 50
Heterogeneous reactions
- Char combustion
2C<S> + O
2
2CO
- CO
2
gasification (reversible)
C<S> + CO
2
2 CO
- H
2
O gasification (reversible)
C<S> + H
2
O CO + H
2
- H
2
gasification (reversible)
C<S> + 2 H
2
CH
4
Homogeneous reactions
- Water-gas shift (reversible)
CO + H
2
O CO
2
+ H
2
- CO combustion
CO + 0.5 O
2
CO
2
- CH
4
combustion
CH
4
+ 1.5 O
2
CO
2
+ 2 H
2
O
- H
2
combustion
H
2
+ 0.5 O
2
H
2
O
- Tar Combustion
Tar + O
2
CO
2
+ H
2
O
Reactions
Initial stage reactions
- Moisture release
- Devolatilization
- Tar cracking
2 3 2 2
2 4 2
Cl NH S H O H
H CH CO CO TAR Volatile
2 3 2 2
2 4 2
Cl
d
NH
d
S H
d
O H
d
H
d
CH
d
CO
d
CO
d d
| | | |
| | | | o
+ + + +
+ + + +
3 2 2
2 4 2
NH S H O H
H CH CO CO C Tar
3 2 2
2 4 2
NH
c
S H
c
O H
c
H
c
CH
c
CO
c
CO
c c
| | |
| | | | o
+ + +
+ + + +
O(v) H Moisture
2

2011 ANSYS, Inc. October 4, 2011 51


Published Results: Shaoping et. al. (2007)
Transport Gasifier at Power System Development Facility (PSDF),
(a) Flow path lines (b) Solid velocity vectors (c) Carbon combustion rate
(a)
(b) (c)
The gas composition at the outlet of PSDF
The mean gas temperature along the PSDF
2011 ANSYS, Inc. October 4, 2011 52
Biomass Gasification
Prediction of reacting flow in a static bed with air injection at base
and top feed of solid biomass pellets
Uses Eulerian-Eulerian multiphase model with heterogeneous
reactions coded in CFX expression language
Multiphase reactions:
char + CO2 -> 2 CO (char oxidation by CO2)
char + H2O -> CO + H2 (char oxidation by steam)
char + 2 H2 -> CH4 (pyrolysis)
char + 0.5 O2 -> CO (char oxidation)
Gas phase reactions:
H2 + 0.5 O2 -> H2O (hydrogen combustion)
CH4 + 1.5 O2 -> CO + 2 H2O (methane combustion)
CO + 0.5 O2 -> CO2 (CO oxidation)
CH4 + H2O -> CO + 3 H2 (methane reforming)
CO2 + H2 <-> CO + H2O (gas water shift reaction)
Figure shows cooling effect of adding increasing amounts of water
vapour to injected air stream
2011 ANSYS, Inc. October 4, 2011 53
Recent Developments
2011 ANSYS, Inc. October 4, 2011 54
Improved models that use information from scale resolving
turbulence models: LES, SAS, DES, ...
1) G-Scalar model Norbert Peters
Level set method
Flame-front reconstruction
Re-initialisation of G to obtain distance from flame-front
2) Thickened Flame model Thierry Poinsot
Artificially broaden flame-front so that is can be resolved
Based on DNS modelling at lower Reynolds number
Premixed Turbulent Combustion
2011 ANSYS, Inc. October 4, 2011 55
G-equation model
Progress variable: c-equation in BVM
Mean (RANS) or filtered (LES) reaction progress
Replaced by level set method: G-equation
Mean distance to the flame front
Interface tracking method
Spatially filtered thin flame remains thin
Suitable for use with scale resolving turbulence models
c
~
s )
x
c
~
D (
x
) c
~
u
~
(
x
) c
~
(
t
t u
j
t
j
i
j
V +
c
c
c
c
=
c
c
+
c
c

( ) G
~
D s ) G
~
u
~
(
x
) G
~
(
t
t t u i
j
V =
c
c
+
c
c
k
2011 ANSYS, Inc. October 4, 2011 56
G-equation Model
Advantages of G-equation over c-equation
Can include the effects of (mean) curvature on the flame speed
(Karlovitz number)
c-equation predicts continually increasing flame brush thickness
Especially bad for LES
Disadvantages of G-equation over c-equation
Must be run in unsteady mode, so computationally expensive
G-equation good model for unsteady flames (LES)
Reference
Norbert Peters, Turbulent Combustion, Cambridge UP,2000
2011 ANSYS, Inc. October 4, 2011 57
Hamamoto Test Case Setup
Ref. Hamamoto et al. (1988)
Diameter 125 mm
Height 35 mm
| 1.0
p
ini
243 kPa
T
ini
325 K
CFD Setup
3D Hex.
Structured
3 sec.
Level Nodes Cells
1 1,536 705
2 6,144 2,945
3 24,576 12,033
Turbulence SST Model
s
T
Ewalds Closure (2006)
Ignition r
T
= 3 mm,V
ini
= 0.2 mm
3
2011 ANSYS, Inc. October 4, 2011 58
Hamamoto Test Case Results
2011 ANSYS, Inc. October 4, 2011 59
Thickened Flame Model
Proposed by Thierry Poinsot et al.
Combustion model for scale-resolving turbulence model, e.g. LES, DES or SAS
Laminar flame is too thin to resolve in 3-D
Laminar flame thickness, o ~ 1mm
Need about 10 points to resolve internal flame structure
CONCEPT:
Artificially thicken the flame but keep the laminar flame speed constant
Increase flame thickness by the thickening factor, F
Multiply laminar mass/heat diffusivities by F
Divide laminar reaction rate by F
2011 ANSYS, Inc. October 4, 2011 60
Thickened Flame Model
Thickening factor calculated as:
A is the cell length (V
1/3
)
N is the user specified number of grid points in flame (typically 10-15)
o is the user specified laminar flame thickness (usually 0.1 to 0.3 mm)
Dynamic TF: Limit diffusivity increase to near flame
Prevent enhanced mixing up or down stream
Dynamic thickening factor, O, is calculated as:
so
= spatially filtered absolute reaction rate
| is a model constant that controls the transition between thickened and
unthickened regions (typically 10)
O + = ) 1 ( 1
max
F F
o
A N
F
max
=
2011 ANSYS, Inc. October 4, 2011 61
Thickened Flame Model
Dynamic effective diffusivity calculated as:
Far from flame (O = 0 ) , D
eff
= D
lam
+ D
turb
Near flame (O = 1 ), D
eff
= D
lam
E F
E is the efficiency factor
Small scale flame wrinkles are suppressed by thickening
Increase the laminar flame speed by E
Multiply both diffusivities and reaction rate by E
Calculated from the turbulent
flame speed at the length scales l
t
= A and l
t
= FA using e.g. Zimont
correlation
2011 ANSYS, Inc. October 4, 2011 62
TFM Example: Chen (1996) Premixed
Flame F3
Instantaneous (filtered) temperature
TFM Laminar
x/d=2.5
x/d=8.5
LES with Re = 24,000
2011 ANSYS, Inc. October 4, 2011 63
TFM Example: Chen Flame F3
TFM F, Oand E
F
O
E
2011 ANSYS, Inc. October 4, 2011 64
TFM example: Chen flame F3
TFM vs LAM vs EXP mean axial velocity
x/d=2.5
x/d=4.5
x/d=6.5 x/d=8.5
2011 ANSYS, Inc. October 4, 2011 65
TFM example: Chen flame F3
TFM v LAM v EXP mean temperature
x/d=2.5 x/d=4.5
x/d=6.5 x/d=8.5
2011 ANSYS, Inc. October 4, 2011 66
TFM Observations
Advantage:
Generally applicable to all types of turbulent combustion using tabulated or
reduced chemistry schemes
Issues:
1) Thickening destroys influence of small scale turbulence need E factor
2) All species must be transported so more computationally expensive than
a progress variable + mixture fraction method
References:
Legier, Poinsot and Veynante, Centre for Turbulence Research, Proc Summer
Program 2000, p157
Kuenne, Ketelheun and Janicka, Combustion and Flame 158 (2011) 1750
2011 ANSYS, Inc. October 4, 2011 67
Time-Scale Separation (TSS) Model
Innovative pollutant model to predict CO
Problem: How to represent fast formation of CO at flame front and relatively
slow post-flame oxidation?
Time-Scale separation used to represent the two processes:
Data extracted for and from PDF chemistry tables
CO T
Front
CO
CO
cS c s Y
Dt
DY
+ V =
CO formation at flame front
CO oxidation to CO
2
Front
CO
Y
CO
S
2011 ANSYS, Inc. October 4, 2011 68
2011 ANSYS, Inc. October 4, 2011 69
Future Trends
Scale-resolving turbulence models used to predict combustion dynamics, e.g.
noise and vibration
Trend towards coupled models of CFD with thermal and stress analysis, and also
1-D process models
Increased use of combustion models, rather than inert models, to simulate fires
Growing use of multi-phase reacting flow models, both Eulerian and Lagrangian,
for simulation of coal and biomass gasifiers
Rapid development of unsteady premixed combustion models for use with
scale-resolving turbulence models, e.g. TFM and G-Equation models, but
some work remains to be done

Вам также может понравиться