Topic 20 Organizer Jose M GarciaOliver Subtopic 23 Coordinators Michele Bolla ETH Dan Haworth PSU Scott Skeen Sandia Subtopic 23 Contributors Experimental IFP Energy nouvelles ID: 697794
Download Presentation The PPT/PDF document "Subtopic 2.3: Soot Field" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Subtopic 2.3: Soot Field
Topic 2.0 Organizer
Jose M. Garcia-OliverSubtopic 2.3 Coordinators Michele Bolla, ETH Dan Haworth, PSU Scott Skeen, SandiaSubtopic 2.3 Contributors Experimental IFP Energy nouvelles Sandia Meiji University Modeling University of Wisconsin Politecnico di Milano ETH Zurich
POLIMI
Wisconsin
SandiaSlide2
Review of ECN 2 Soot Session
Dan Haworth provided discussed the physics of soot formation and CFD-based soot modeling, emphasizing the importance of radiation heat transfer (see Webex recording)Emre Cenker presented LII/LEM experiments for Spray A and a few parametric variantsPeak SVF of 2-4 ppm for Spray A (930 K, 21.8 kg/m
3)Peak SVF of 12 ppm at 1030 KSignal trapping considered to be negligibleTwo groups (ETH and U Wisconsin) submitted mean soot volume fraction data for Spray HModels reproduced measured soot levels and trends with variations in ambient O2 and densityNo definitive conclusions were drawn regarding the merits of the different modeling approachesRecommendations from ECN 2:Ambient temperature of ECN pre-combustion vessels should be well characterizedLII measurements exhibited significant statistical error due to jitter between the laser and camera. Future LII experiments must minimize jitter and account for it in the LII calibrationLong injection duration for measurements examining quasi-steady behaviorBegin looking at Spray A (n-dodecane)Modelers should perform systematic parametric studies to isolate and quantify the effects of individual physical processesTurbulence-Chemistry Interaction
Turbulence-Radiation InteractionNucleation, surface growth, agglomerationSlide3
Subtopic 2.3: Objectives
Soot Onset (Timing and Location)How to quantify for consistency between experiments and modelingParametric variation (850 K, 900 K, 1000 K) (13%, 15%, 21% O2)2-D Soot FieldTransient progression (1.5, 2.0, 2.5, 4.5 ms
ASOI)Compare IFPEN LII with extinction imaging from Sandia at available timingsEvaluation of signal-trappingStandardization of soot non-dimensional extinction coefficientSoot TemperatureComparison of 2-Color pyrometry (IFPEN) with Imaging Spectrometer (Sandia)Soot Particle SizeWhat is the primary particle size at the location of peak SVF?How does particle size change as a function of distance from the injector?“To improve the understanding of the physical/chemical processes of soot formation and oxidation under engine-relevant conditions and to distill this improved understanding into predictive CFD-based models.” -ECN3 GuidelinesSlide4
Sandia
Extinction Imaging Setup
Simultaneous ignition delay, quasi-steady lift-off length, and soot extinction measurementsTwo incident wavelengths has proven useful for understanding optical properties of soot
Soot Measurement Resolution
85 kHz 35 µs (2 wavelength) 23 µs (1 wavelength)
100 µm per pixel
Lower Detection Limit (Beam-steering)
< 0.5 ppm Slide5
Extinction Imaging
Spray A
Soot mass is proportional to measured optical thickness (
KL
)
High-speed extinction imaging measurements provide time-resolved
KL mapsTotal mass and axial resolved soot mass do not require tomography for comparison to modeled SVF results
Mass-based soot onset timing and location provide targets for modeling efforts
Inception of soot in spray head and its progression downstream provide a difficult modeling target
Mass
soot
=
pixel area
Slide6
Time Sequence of LII vs. Time-Resolved Extinction
*
Tamb: 930 K *ρamb: 21.8 kg/m3
Can compare progression of total soot mass as an indicator of soot onset
Appears to be a mismatch in reacting vapor penetrationSlide7
Soot Onset: Timing and Location
Mass-based soot onset timing and location provide targets for modeling efforts
Based on a soot mass threshold of
0.5 µg
for total mass
Based on a soot mass threshold of
10 ng
for axial resolved mass
Rate of total soot mass increase is very similar for IFPEN LII data and Sandia Extinction Imaging Data
200 µs difference in soot onset potentially explained by uncertainty in IFPEN vapor penetrationSlide8
Soot Onset: Timing and Location
15%
850 K15%1000 K
T
amb
[K]
850
900
1000
Mean
Soot Mass [µg]
(quasi-steady)
2
14
42Slide9
Soot Onset: Timing and Location
15%
850 K15%
1000 K
T
amb [K]
850
900
1000
Mean
Soot Mass [µg]
(quasi-steady)
2
14
42
Full soot field was not captured, so numbers are considered low relative to realitySlide10
Soot Onset: Timing and Location
13%
900 K21%900 K
O
2,amb
[%]
13
15
21
Mean
Soot Mass [µg]
(quasi-steady)
10
14
11Slide11
Soot Onset: Timing and Location
13%
900 K
21%900 K
O
2,amb [%]
13
15
21
Mean
Soot Mass [µg]
(quasi-steady)
10
14
11
Full soot field was not captured, so numbers are considered low relative to realitySlide12
Soot Timing and Location Relative to Ignition
Parametric variation around Spray A in temperature and O
2
concentration show a predictable trend in the time between high-temperature ignition and soot onset and the location of high-temperature ignition and soot onset.Slide13
Time-Resolved Total Soot Mass
Higher ambient temperature and O
2
lead to better performance of UW model
UW model scales similarly later during quasi-steady period for AR and O3 cases
Between 1 and 2
ms
ASOI, POLIMI model scales similarly for all but the 21% O
2
case
WisconsinSlide14
Ensemble Averaged SVF (IFPEN/Sandia)Slide15
Ensemble Averaged SVF
sdf
LII n-heptane: 15% O2, 1000 K, 1500 bar, 30 kg/m3, 100 µm orifice
With sufficient statistics, ensemble average of single-shot LII yields axisymmetric images similar to time- and ensemble-averaged extinction imaging dataSlide16
Radial Profiles of
fv
Signal trapping may cause plateau in LII dataCorrection must be applied to raw LII signal before integration and calculation of fv
IFPEN used a 425 nm +/- 15 nm
bandpass
filter for collection of LII signal
Extinction measurements at Sandia using 406 nm incident light showed a mean
KL
of ~0.9 between 55 and 60 mm (
KL
= 0.45 for half the path length)
Signal trapping could result in 36% of the signal blocked along the centerline
Must also consider the effect of
k
e
Sandia
KL
using
406 nm incident lightSlide17
Non-dimensional Extinction
Coeff., ke
Primary Particle Diameter, dpke (N=5)ke
(N=75)ke (N=150)
[nm]
unitlessunitlessunitless
107.03
7.08
7.12
16
7.04
7.21
7.28
20
7.06
7.33
7.47
30
7.12
7.77
8.04
40
7.258.37
8.77507.459.089.62607.729.9010.6
Standard ke was updated from 4.9 to 8.7 for 632.8 nm extinction measurements
ke computed from Rayleigh-Debye-Gans theory for fractal aggregates is differentRefractive index 1.75-1.03i from Williams et al. Int. J. Heat and Mass Transfer (2007) Np primary particles per aggregate, dp primary particle diameterIncident wavelength of 632.8 nmGreater effect of Np for larger primary particle sizeSmall particles sizes in Spray A measured by TEM means uncertainty in assumption of constant Np is reducedGreatest uncertainty remains in the refractive index of sootSlide18
Non-dimensional Extinction
Coeff., ke
Primary Particle Diameter, dpke (N=5)ke
(N=75)ke (N=150)
[nm]
unitlessunitlessunitless
107.03
7.08
7.12
16
7.04
7.21
7.28
20
7.06
7.33
7.47
30
7.12
7.77
8.04
40
7.258.37
8.77507.459.089.62607.729.9010.6
Standard ke was updated from 4.9 to 8.7 for 632.8 nm extinction measurements
ke computed from Rayleigh-Debye-Gans theory for fractal aggregates is differentRefractive index 1.75-1.03i from Williams et al. Int. J. Heat and Mass Transfer (2007) Np primary particles per aggregate, dp primary particle diameterIncident wavelength of 632.8 nmGreater effect of Np for larger primary particle sizeSmall particles sizes in Spray A measured by TEM means uncertainty in assumption of constant Np is reducedGreatest uncertainty remains in the refractive index of soot
O
3 (21% O
2
)Slide19
Signal Trapping
Correction based on Sandia extinction data improves plateau somewhat
Correction actually decreases mass along chosen cross section by 4%
Use uncorrected
f
v
as
I
LII
(
x
,
y
)
, make correction based on Gaussian
KL
from Sandia data, re-integrate new
KL
LII
Correction increases mass by a factor of 1.8Slide20
Total Soot MassIFPEN calibrated with 632.8
HeNe laser extinctionke = 8.7 was standard at the time of publicationSandia extinction imaging with 406 nm LEDke = 7.76 based on RDG theory with d
p = 16 nm and Np = 150Slide21
Total Soot MassIFPEN calibrated with 632.8
HeNe laser extinctionke = 8.7 was standard at the time of publicationke = 7.28 from RDG theory with dp = 16 nm,
Np=150 as in Imaging Extinction work (20% increase in fv and soot mass)Slide22
Summary
Extinction imaging measurements have provided useful targets for modeling efforts including:Soot onset timeSoot onset locationSoot mass and/or soot volume fractionTransient progression of the 2D soot field with high temporal resolution (35 µs)Need to increase field of view and further reduce effects of beam steering
Comparison of LII/LEM measurements from IFPEN and Sandia’s Extinction Imaging measurementsSimilar rate of soot mass increase for Spray ADifferences in reacting penetration may explain difference in soot onset timeDifferences in SVF lessened by accounting for signal trapping (~400 nm)Differences in SVF lessened further by considering ke derived from Rayleigh-Debye-Gans theoryPrimary particle size as measured by IFPEN/Meiji ranges from 10-20 nmSmall primary particle sizes reduce the error associated with our assumption of constant Np throughout the soot field.Slide23
Dirty Laundry-Nozzle Aging (injector 370)
Similar lift-off lengths and total soot mass, slightly short ignition delay time for later data, significantly shorter soot onset timeMass measurements and pressure traces indicate change in discharge coefficient (more mass in later experiments)
Tamb = 905 KLift-off: 16.09τig =
404 µs (chemi)τig =
400 µs (press)
Tamb = 902.5 KLift-off: 16.23
τig = 344 µs (
chemi
) faster camera
τ
ig
=
370
µs
(press)Slide24
Outline: Soot modeling
Presentation
soot
models used (3 contributors)UW, POLIMI
and ETHAnalysis C2H2 as
soot
„
initial
condition
“
C2H2 total mass in time (UW, ETH, POLIMI
and
UNSW)
Spatial
distribution
at 1.5 ms and 4 ms (UW, ETH, POLIMI, UNSW and ANL)Analysis soot results for reference
caseTotal soot mass in timeSoot spatial extent
at 1.5/2.0/2.5 ms compared to KL (qualitative)SVF comparison at 4 ms (quantitative)Mean particle size at
4msAnalysis Soot onsetEvolution of soot
mass and locationSensitivity analysis soot modelSurface growth rateConclusionsOutlook Slide25
Overview ECN Soot modeling
ECN 1: No soot results presented
ECN 2: Only Spray H (n-heptane) consideredTwo contributors: UW and ETHBoth used two-equation soot modelUW: G. Vishwanathan et al.,
Comb. Sci. and Tech.
182 (2010)
ETH: M. Bolla
et al.,
Comb. Sci. and Tech.
185
(2013)
Comparison of quasi-steady soot only
ECN 3: Spray A (n-
dodecane
) considered
Three contributors: UW, ETH and POLIMI
All used two-equation soot model
UW and ETH used the same soot model as ECN 2
Soot modeling for Spray A at early stage (to-date no publication)
Comparison of soot
temporal
and spatial evolution
Focus on soot onset evolutionSlide26
Two-equation soot model
ACETYLENE / PAH
PRODUCTS
Inception (1)
Coagulation (5)
Surface
Growth
(2)
Surface oxidation (3-4)
FUEL
Chemical
mechanism
(0)
Solve transport equation for soot mass fraction and number
density
Accounts for inception, surface growth, coagulation and surface oxidation
Calibrated reaction rates (semi-empirical)
Mono-disperse spherical soot particles assumed
Agglomeration neglectedSlide27
Two-equation soot model
ACETYLENE / PAH
PRODUCTS
Inception (1)
Coagulation (5)
Surface
Growth
(2)
Surface oxidation (3-4)
FUEL
Chemical
mechanism
(0)
(1) Particle Inception
(5) Particle Coagulation
(2) Particle Surface Growth
(3) Particle Oxidation by O
2
(4) Particle Oxidation by OH
ETH and POLIMI:
UW:Slide28
Modeling Approach
Temp [K]
800850900
1000
1100
1200
O
2
[vol%]
15
13/15/17/21
13/
15
/17/21
13/15/17/21
13/15/17/21
13/15/17/21
Density [kg/m
3
]
22.8
7.6/15.2
/
22.8/30.4
7.6/15.2
/
22.8
/30.4
7.6/15.2
/
22.8/30.4
7.6/15.2
/
22.8/30.4
7.6/15.2
/
22.8/30.4
P
inj
[
MPa
]
150
50/100/150
50/100/
150
50/100/150
50/100/150
50/100/150
Computational grid
Related sub-models
Lift-off length
Onset of the averaged OH concentration
Ignition delay
Maxmium
d
T
/
d
t
Maxmium
d
OH
/
d
t
Phenomenon
Model
Spray breakup
KH-RT instability
Evaporation
Discrete multicomponent (DMC)
Turbulence
Generalized
RNG k−
ε
model
Combustion
SpeedChem
Droplet collision
ROI model
Near
nozzle flow
Gas-jet
model
Soot
formation
Multi-step phenomenologicalSlide29
Physical
process
Expression
Inception
:
A4
soot
C
2
H
2
surface growth
Coagulation
O
2
oxidation
OH oxidation
PAH condensation
Transport
equations
G.
Vishwanathan
et al., Combustion Science and Technology, 2010, 182(8):1050-1082.
Soot Modeling
ApproachSlide30
Non-reacting mixing
Soot
modeling results
Reacting conditionsSlide31
Total C2H2 mass
Large
differences in peak C2H2 mass (factor 4)All simulation predict a plateau after approx. 3 msDelays in start of C2H2
production coincides with differences
in ID
Different ID:UW 0.82 msETH 0.48 ms
POLIMI 0.62 msUNSW 0.70 ms
EXPERIMENT 0.41
ms
IDSlide32
C2H2 comparison at 1.5 and 4 ms
1.5
ms4 msr=0mmr=0mm
LOLSlide33
Total soot mass
Comparison
total soot massOnset of soot formation
UW
and
ETH show
a comparable
magnitude
and
shape
Experimental
first
soot
bump
not
captured
by
the models
Delays in start
of soot formation coincides with differences in ID
IDSlide34
Temporal evolution soot
region
: 1.5/2.0/2.5 ms1.5 ms2 ms2.5 ms
Qualitative
Soot
region
in qualitative
agreement
Differences
in
soot
spread
and
tip
penetration
Simulation
has
shorter penetration
at 2/2.5 ms
Experiment: KL signalSimulation: normalized SVFSlide35
Soot volume fraction at 4 ms
r=0mm
z=60mmQuantitative
Soot
region
in qualitative agreement
Different axial
offsets
LOL-
soot
UW
and
ETH
show
comparable
results
UW
tighter
in
radius
->less
soot
volumeLOL[ppmv]Slide36
Computed mean particle size at 4 ms
[nm]
UW and ETH models predict largest
particles
of 17-18
nm
Largest
mean
particle
size
at
peak
sootSlide37
Soot onset: Evolution axial soot mass
UWETH
EXP
ID=0.82 ms
ID=0.48 ms
ID=0.41 ms
For
soot
onset
analysis
„
reset
processes
“
->
Consider
time after ID
ETH
shows
a good shape,
soot 2 times
lowerUW is 2 times lower than ETH->
Comparable
SVF but
lower
spread
of
the
soot
region
UW
overpredicts
location
of
soot
onset
-> due
to
larger ID (0.82 vs. 0.41
ms
)Slide38
Soot onset: Evolution SVF simulation
UW
ETHID=0.82 msID=0.48 ms
Evolution
of
SVF is
comparable
UW
reaches
half SVF
max
after ID+0.7ms
and
ETH
takes
0.8
ms
(quasi-
steady
SVF
max
is
6 ppmv)Slide39
Soot onset: Mean particle size evolution
UW
ETHID=0.82 msID=0.48 ms
UW shows
a strong particle
size
peak
at
ID+0.1
ms
ETH
shows
a
more
smooth
increase
at
the
beginning
(ID+0.1-0.2 ms)Fast stabilization
of
particle size upstreamSpray A TEM60 mmIFPEN/MeijiSlide40
Sensitivity analysis: Surface growth -33%
Soot
mass is most sensitive w.r.t. surface
growth
(cf. e.g. Bolla
et al., CST 2013)->
most
illustrative
sensitivity
study
A 33%
reduction
in
surface
growth
decreases
total
soot
mass but not
the
shapeBoth UW
and ETH react
analogously: reduction of soot mass by 40-50%Radial SVF profiles
are
nearly
down-
scaled
->
Soot
region
remains
the
sameSlide41
Summary and conclusions
Detailed
analysis of soot formation performed
for
reference
case
Large
differences
in C2H2
and
soot
onset
-> DIFFERENT ID
Soot
onset
:
first
soot
peak not reproducedProbably
mixing related
(Tip vortex dynamics) -> LES needed?Quasi-steady soot
fairly
well
captured
(same
as
ECN 2)
Sensitivity
analysis
on
surface
growth
assessed
Consistent
results
with
and
without
TCI
Soot
spatial
extent
remains
unchanged
->
Mostly
mixture
fraction
determines
where
soot
is
Before
looking
at
TCI
and
more
complex
soot
models
one
should
:
Assure
accurate
tip
penetration
and
mixture
fraction
distribution
Improve
IDSlide42
Outlook - Topic 2.3 Soot field
Experimental
Soot:Extinction Imaging in constant flow vessel (build up statistics for time-resolved tomographic reconstruction)Gas sampling (can we measure acetylene axial profile?)Combined laser-induced incandescence with extinction imaging
Spectrally resolved laser-induced fluorescence (progression of PAH growth)
Quantify soot in Spray A with other injectors
Multiple injections
Spray B
Soot
modeling
:
Keyword
for
future
: TRANSIENT
Short
injection
, multiple
injection
Understanding
the
first
soot bump
Need for more accurate chemical mechanisms – ID must be
improved
Alternatively
:
re-visit
n-
heptane
sprays
in
more
detail
?Slide43Slide44
LIF 355: consideration CH2O and PAH (first impression)
First
impression of simulation compared to LIF 355
CH2O
is
more
upstream
and
PAH(A4)
is
more
downstream
than
exp
.
LIF 355
coincides
approx
. with UW simulated C2H2
Simulation UW at
4 msExperiment IFPENLIF 355 at 4.7 msSlide45
Sandia constant-volumeSteady
soot
Comparable soot volume fractionDI tight, CMC broad
distribution
Experiment is in
between
DI
CMC
Exp
.
42 bar
85 bar
Source:
Bolla
et al.,
Comb
.
Theory
Modelling (2014)Slide46
Sandia constant-volumeQuasi-
steady soot
Soot formation rate is comparableDI predicts 500 times larger soot oxidation rateCaused by limited mixture fraction co-existance range
Formation
Oxidation
PDF
soot
O
2
C
2
H
2
soot
PDF
DI
DI
CMC
CMC
Source:
Bolla
et al.,
Comb
.
Theory
Modelling (2014)Slide47
Sandia constant-volumeTransient soot
DI
overpredicts soot oxidation after end of injection
1
2
3
4
1
2
3
4
12% O
2
, 14.8 kg/m
3
, 1000 K
DOI=1.8
ms
Source Exp.:
Idicheria
and Pickett, IJER (2011)Slide48
Pyrometry
IFPEN 2-Color SetupCollected 425 +/- 15 nm and 676 +/- 14.5 nmCalibrated with Santoro burner inside vessel at 1 atmEliminates uncertainties associated with soot emissivity15 images at 3.5 ms ASOI, ensemble averaged
Spray A,
T
sootSlide49
Pyrometry
Sandia Imaging Spectrometer SetupSystem images only the central 1.4 mm along spray axisCollects emission from entire spray eventExposure derived from high-speed imagingSpectra quantified using a calibrated integrating sphereSlide50
Pyrometry
Two very different pyrometry approachesIFPEN: 2-color, 2 camera pyrometrySandia: Imaging Spectrometer, long exposure, center 1.4 mm along spray axisSlide51
Soot Subtopic 2.3 Contributors
ExperimentalSandiaextinction imaging: Time-resolved KL maps, soot mass, and fv maps during quasi-steady periodSoot pyrometry (Imaging Spectrometer):
Spatially resolved soot particle temperature and KL along central axis of spray flame + total radiation from broadband soot emissionIFPENLaser-induced Incandescence & Laser Extinction: Time sequence of fv along central plane of spray flame, ensemble averaged fv during quasi-steady periodTwo-camera, Two-color pyrometry: 2-D map of soot particle temperatureIFPEN/MeijiSoot sampling/TEM analysis: Soot particle sizingSlide52
Subtopic 2.3: Overall Objectives
What is the soot distribution for Spray A?How is it modified with different parametric variables?How do different measurement techniques compare?How accurate do different modeling approaches predict the soot field?
“To improve the understanding of the physical/chemical processes of soot formation and oxidation under engine-relevant conditions and to distill this improved understanding into predictive CFD-based models.” -ECN3 GuidelinesHigh-speed Extinction Imaging, Spray A, n-dodecaneSlide53
Soot Onset: Timing and Location
Soot mass is proportional to measured optical thickness (
KL
)
High-speed extinction imaging measurements provide time-resolved
KL
maps
Total mass and axial resolved soot mass do not require tomography for comparison to model results
Mass-based soot onset timing and location provide targets for modeling efforts
Based on a soot mass threshold of
0.5 µg
for total mass
Based on a soot mass threshold of
10 ng
for axial resolved mass
Mass
soot
=
pixel area
T1 (800 K)
Extinction due to beam steering helps define threshold. Soot extinction not detected for 800 K case. Soot mass attributed to beam steering equivalent to approx. 0.25 µg