Sunteți pe pagina 1din 9

Twenty-Seventh Symposium (International) on Combustion/The Combustion Institute, 1998/pp.

1087–1095

EFFECTS OF TURBULENCE ON SPECIES MASS FRACTIONS IN


METHANE/AIR JET FLAMES

R. S. BARLOW and J. H. FRANK1


Combustion Research Facility
Sandia National Laboratories
Livermore, CA 94551-0969, USA

It is important that combustion models capture the effects of turbulent mixing on reaction zone structure
in non-premixed and partially premixed flames. A more complete understanding of the response of species
mass fractions to turbulent mixing is needed to improve predictive capabilities, particularly with regard to
combustion intermediates and minor species. Using the combination of Raman scattering, Rayleigh scat-
tering, and laser-induced fluorescence, simultaneous measurements of CO, OH, H2, and NO are obtained
along with the major species, temperature, and mixture fraction in a series of six piloted methane/air jet
flames. Flame conditions vary from laminar to turbulent with significant localized extinction. Two-photon
laser-induced fluorescence (TPLIF) is used to determine instantaneous CO concentrations, providing an
improvement over Raman scattering measurements of CO in methane flames. Conditional probability
density functions (cpdf ’s) of species mass fractions in the six flames are compared. Significant changes are
observed in the mass fraction cpdf ’s of several species. Results for H2O, CO2, H2, and OH are consistent
with the concept that turbulent transport becomes dominant over molecular diffusion within the range of
Reynolds numbers and axial locations considered in these experiments. The cpdf ’s of CO mass fraction
are broadened in the turbulent flames relative to the laminar flame. However, there is not an increase in
the maximum conditional mean value of the CO mass fraction as suggested by some previously reported
measurements in methane flames. The cpdf ’s of NO mass fraction at a given streamwise location in the
turbulent flames show NO levels decreasing significantly as jet velocity increases.

Introduction levels with increasing Reynolds numbers in a series


The effects of turbulent mixing on reaction zone of piloted methane/air jet flames.
structure in non-premixed and partially premixed Computational studies have been conducted to in-
flames have important implications for combustion vestigate mechanisms that might contribute to the
models, and a more complete understanding of the high measured levels of CO, H2, and OH. Mauß et
response of species mass fractions is needed, par- al. [7] considered the effects of unsteadiness on lam-
ticularly with regard to combustion intermediates inar flames and suggested that high CO concentra-
and minor species. Masri et al. [1] reviewed multi- tions result from an extinction–reignition process.
scalar experiments on piloted- and bluff-body sta- Chen and Dibble [8] observed that high CO levels
bilized flames of various fuels. Several studies of are also produced in perfectly stirred reactor calcu-
methane flames reported levels of intermediates that lations. Barlow and Chen [6] used laminar flame cal-
were higher than predicted by steady laminar flame culations to show that unsteadiness and differential
calculations or turbulent combustion models. CO diffusion can influence combustion intermediates.
levels measured by Raman scattering in piloted The present investigation of turbulent methane
flames of undiluted methane were significantly flame structure uses part of a comprehensive scalar
above those calculated in laminar flames [2]. Raman/ and velocity data set on piloted methane/air jet
Rayleigh measurements in bluff-body flames of flames that is described on the World Wide Web [9].
methane reported by Masri et al. [3] and Correa et Simultaneous measurements of CO, OH, H2, and
al. [4] have also shown CO levels higher than laminar NO were obtained along with the major species,
flame calculations and PDF model predictions. Står- temperature, and mixture fraction. Two-photon la-
ner et al. [5] studied piloted flames of dilute methane ser-induced fluorescence (TPLIF) was used to de-
and reported elevated levels of OH and H2. Barlow termine instantaneous CO concentrations, providing
and Chen [6] reported a progressive increase in OH an improvement over Raman scattering measure-
ments of CO. The influence of turbulence on con-
1Present address: Physical Sciences, Inc., 20 New En- ditional probability density functions (cpdf ’s) of spe-
gland Business Center, Andover, MA 01810, USA cies mass fractions is reported for six jet flames
1087
1088 NON-PREMIXED TURBULENT COMBUSTION

TABLE 1 The concentration of CO was measured by Raman


Relative standard deviations of scalars measured in flat scattering and by two-photon laser-induced fluores-
flames and estimated systematic uncertainties cence (TPLIF). Only the CO TPLIF results are in-
cluded here because they are less affected by inter-
Conditions Systematic ferences and have superior signal-to-noise
Scalar r (rms) (mass frac., T) Uncertainty characteristics. A calibration curve for the CO fluo-
rescence measurements was generated following an
T 1% 2140 Ka 3% approach similar to that for the Raman method. This
YN2 2% 0.73, 2140 Ka 3% curve was based on heater data for dilute CO in N2
YH2O 5% 0.12, 2140 Ka 4% at temperatures up to ;900 K and on rich, premixed
YCO2 6% 0.14, 2140 Ka 4% CH4/air flat flame measurements (;1900 K to
YH2 17% 0.003, 2020 Kb 6–12% ;2200 K). The shape of the polynomial through the
YOH 8% 0.0016, 2140 Ka 10% interpolated region was constructed to match ap-
YCO 6% 0.062, 2020 Kb 10–20% proximately the temperature dependence of a Boltz-
YNO 10% 8 ppm, 1760 Kc 10–15% mann fraction calculation (provided by M. Di Rosa,
Sandia). Further description of the CO TPLIF
aPremixed CH /air, f 4 0.96, uncooled (Hencken) method and comparisons of simultaneous CO mea-
4
burner. surements by Raman scattering and TPLIF are re-
bPremixed CH /air, f 4 1.27, uncooled (Hencken) ported elsewhere [13,14].
4
burner. We note here that comparison of measurements
cPremixed CH /O /N , f 4 0.72, cooled (McKenna) in lean and near-stoichiometric methane/air calibra-
4 2 2
burner. tion flames reveals a small but repeatable offset of
roughly 1 2 1015 mol/L in the TPLIF result above
the Raman result. This may be due to photodisso-
having fixed nozzle diameter and increasing jet ve- ciation of vibrationally excited CO2 at flame tem-
locities, corresponding to jet Reynolds numbers peratures [15]. It is not a significant effect for this
from 1,100 (laminar) to 44,800 (turbulent with sig- paper, because the cpdf ’s are for a mixture fractions
nificant localized extinction). Results suggest a interval where the Raman and TPLIF calibrations
strong influence of scalar transport on major and mi- have been matched. However, interference from
nor species. CO2 may be an important consideration for experi-
ments in lean premixed combustion, where lower
CO concentrations must be measured accurately
Experimental Methods
and direct calibration may be difficult.
Most aspects of the flow facility, diagnostic sys- Experimental uncertainties are summarized in Ta-
tems, and calibration procedures have been de- ble 1. Standard deviations of species mass fractions
scribed previously [10,11]. Spontaneous Raman in the calibration flames may be used to estimate
scattering of the beams from two Nd:YAG lasers contributions of photon shot noise and instrument
(532 nm) was used to measure concentrations of the noise to the widths of the cpdf ’s. Estimates of sys-
major species. The Rayleigh scattering signal was tematic uncertainties are based on analysis of the
converted to temperature using a species-weighted calibration methods, repeatability of calibrations,
scattering cross section, based on the Raman mea- considerations of calibration drift, and allowances for
surements. Linear laser-induced fluorescence (LIF) the interpolated calibration curves for CO and H2.
was used to measure OH and NO, and the fluores- With regard to comparison of cpdf ’s, the relative un-
cence signals were corrected on a shot-to-shot basis certainty (not including statistical noise or potential
for variations in Boltzmann fraction and collisional effects of spatial averaging) within each set of mea-
quenching rate. The spatial resolution was roughly surements in the five turbulent flames is estimated
0.75 mm. to be within 52% for the Raman measurements,
Temperature-dependent calibration functions for 55% for OH, 55% for CO, and 510% for NO.
each of the Raman channels were determined from The piloted burner [1] had a main jet diameter of
a series of measurements in flows above an electric d 4 7.2 mm and a pilot diameter of 18.2 mm. The
heater and a flat flame (Hencken) burner. OH mea- jet composition was 25% CH4 and 75% air, by vol-
surements were referenced to a premixed CH4/air ume, which produced a shorter, more robust flame
flat flame, in which the OH number density had than pure methane [1,5] with lower levels of inter-
been measured previously by laser absorption. The ference. The annular pilot burned a mixture of
NO calibration factor was determined by doping a C2H2, H2, air, CO2, and N2 having the same enthalpy
lean premixed CH4/O2/N2 laminar flat flame (Mc- and equilibrium composition as methane/air at 0.77
Kenna burner) with known concentrations of NO. equivalence ratio. A lean pilot was used to reduce
Flame calculations have shown that destruction of the flux of minor species from the pilot. One profile
the seeded NO in this flame is negligible [10,12]. (Flame D, x/d 4 15) was repeated with the pilot
TURBULENT METHANE/AIR JET FLAMES 1089

TABLE 2
Flow parameters

Ujet—cold Upilot—burnt Ucoflow


Flame Rejet (m/s) (m/s) (m/s)

A ;1100 2.44 none 0.9


B ;8200 18.2 6.8 0.9
C ;13400 29.7 6.8 0.9
D ;22400 49.6 11.4 0.9
E ;33600 74.4 17.1 0.9
F ;44800 99.2 22.8 0.9

Fig. 2. Scatter plots of measured temperature at x/d 4


5 in the laminar Flame A and at x/d 4 30 in the turbulent
Flames B–F. All 3000–5000 samples from each flame are
included. The line shows results of a strained, opposed-
flow laminar flame calculation with a strain parameter of a
4 50 s11 and is included as a visual guide.

affect flame structure, and the flame behaves as a


diffusion flame, with a single reaction zone near the
stoichiometric mixture fraction of fs 4 0.351 [5,16].
Profiles were measured at x/d 4 5 and x/d 4 10 in
the laminar flame and at x/d4 15 and x/d 4 30 in
the five turbulent flames. To minimize errors due to
Fig. 1. Scatter plots of measured temperature at x/d 4 calibration drift, the flames were measured out of
5 in the laminar Flame A and at x/d 4 15 in the turbulent order, wind-tunnel air measurements were acquired
Flames B–F. All 3000–5000 samples from each flame are before and after each profile, and the first profile was
included, and in Flame A the clusters of points from sepa- repeated at the end of the series. A total of 3000–
rate locations are apparent. The line shows results of a 5000 samples was collected at four to six radial lo-
strained, opposed-flow laminar flame calculation with a cations (more in the laminar case) within each flame
strain parameter of a 4 50 s11 and is included as a visual at each streamwise station.
guide. Scatter plots of temperature versus mixture frac-
tion are shown in Figs. 1 and 2. The calculated tem-
perature curve for an opposed-flow laminar flame
flow rate reduced by 40% to assess the influence of
with a strain parameter of 50 s11 is included to fa-
the pilot on minor species concentrations, and there
cilitate visual comparisons. These scatter plots pro-
was no significant change in the cpdf ’s.
vide a qualitative indication of the probability of local
extinction, which is characterized by samples with
Results and Discussions strongly depressed temperatures. At x/d 4 15 (Fig.
1) there is no evidence of extinction in Flame B.
Flame Characteristics Flames C and D have very small probability of local
Measurements were obtained in six flames, as extinction, and extinction becomes more important
listed in Table 2. At these flow conditions the fuel- in Flames E and F. At x/d 4 30 (Fig. 2), the scatter
rich premixed chemistry is too slow to significantly plots for Flames C, D, and E all show low probability
1090 NON-PREMIXED TURBULENT COMBUSTION

There is not sufficient information on scalar gra-


dients in the present flames to quantify the effects
of spatial averaging on the cpdf ’s. However, we note
that the cpdf ’s are determined for mixture fraction
intervals corresponding to peak conditional-mean
mass fractions. Thus, spatial averaging across react-
ing samples would tend to reduce the apparent mass
fractions. We find that the cpdf ’s of several species
shifted toward higher mass fractions with increasing
jet velocity, suggesting that spatial averaging does
not have a strong influence on the present results.
The mixture fraction was calculated following the
method of Bilger et al. [18], but with oxygen ex-
cluded from the expression:

2(YC 1 YC,2)/wC ` (YH 1 YH,2)/2wH


f4
2(YC,1 1 YC,2)/wC ` (YH,1 1 YH,2)/2wH
Here, Y’s are elemental mass fractions, w’s are
atomic weights, and subscripts 1 and 2 refer to the
main jet and coflowing air stream, respectively. The
partial premixing of fuel and air leads to boundary
Fig. 3. Measured conditional means of temperature and conditions for YO that are relatively close, causing
the mass fractions of CO2, H2O, CO, OH, NO, and H2 in excessive noise in the mixture fraction if oxygen is
Flame B at x/d 4 15 (symbols). Lines show results of a included in the calculation.
strained, opposed-flow laminar flame calculation with a Figure 4 includes cpdf ’s from x/d 4 5 and x/d 4
strain parameter of a 4 100 s11. This calculation includes 10 in Flame A and from x/d 4 15 in Flames B
full transport. The mixture fraction intervals used in form- through F. The latter location corresponds to the
ing the cpdf’s are indicated for each species. “neck” region of the turbulent flames, where local
extinction is most probable in the highly strained
cases. The cpdf ’s of YH2O in the turbulent flames
of extinguished samples, indicating a healthy reig- align with those in the laminar jet flame, and the
nition process. In contrast, Flame F still has high progression with increasing Reynolds number in-
probability of local extinction at x/d 4 30, and the volves an increase in the population of extinguished
characteristic bimodality of the temperature distri- samples without a significant change in YH2O for re-
bution in lean and near-stoichiometric samples [1] is acted samples. In contrast, the cpdf of YCO2 shows a
evident. progressive decrease in the most probable value of
The rms fluctuation in temperature near the stoi- YCO2. The cpdf of YOH shows a trend similar to that
chiometric mixture fraction in the laminar flame is reported in Ref. [6]. As the Reynolds number in-
40–45 K, which is greater than the 20–25 K mea- creases, the curves broaden and shift toward higher
sured in the calibration flames. This is attributed to mass fractions. This trend is clearly present before
small variations in the location of the reaction zone, local extinction becomes important, and it cannot be
which do not significantly add to the relative fluc- attributed to reignition.
tuations in species mass fractions. For CO, the main effect of increased turbulence
is a broadening of cpdf ’s relative to Flame A. This
Conditional Probability Density Functions broadening is mainly in the direction of lower mass
fractions, but the trend is less obvious than for OH.
In this section, cpdf ’s of the mass fractions of H2O, At x/d 4 15 in the turbulent flames, the cpdf ’s of
CO2, CO, OH, H2, and NO in the six flames are YH2 display a nonmonotonic response to increasing
compared. For each species, the cpdf is formed from Reynolds number. The curve for Flame B is similar
samples within a mixture fraction interval corre- to that from Flame A. YH2 tends to be higher in
sponding to the peak mass fractions of that species. Flame C but decreases significantly in Flames E and
These intervals are shown in Fig. 3, which includes F. The cpdf ’s of YNO show mass fractions decreasing
conditional means of measured scalars at x/d 4 15 as the Reynolds number increases, and this may be
in Flame B. The location of the peaks can change primarily an effect of residence time.
slightly in these flames. However, small changes in Figure 5 includes the cpdf ’s from x/d 4 30, which
position or width of the mixture fraction intervals do is roughly two-thirds the Favre-average stoichiomet-
not significantly alter the cpdf ’s or conclusions re- ric flame length (Ls 4 47d) for the turbulent cases.
ported here. Here, cpdf ’s of YH2O in Flames B through E are
TURBULENT METHANE/AIR JET FLAMES 1091

Fig. 4. Conditional pdf’s of the measured mass fractions of H2O, CO2, OH, CO, H2, and NO from x/d 4 5 (A5) and
x/d 4 10 (A10) in the laminar jet flame and from x/d 4 15 in the turbulent flames B–F. Mixture fraction intervals used
in forming the cpdf’s are given in each graph.

nearly identical, with the most probable values of explain, at least in part, the distinct shift to higher
YH2O being about 10% higher than in Flame A. The mass fractions that is seen in the cpdf ’s of YH2 at x/
cpdf for Flame F is bimodal, but the portion cor- d 4 30. A comparison of laminar flame calculations
responding to reacted samples aligns with the other at the same strain rate (a 4 100 s11) shows the peak
turbulent flames. The cpdf ’s of YCO2 in Flames B–E mass fraction increasing by ;10% for H2O, decreas-
show a most probable mass fraction about 10% lower ing by ;9% for CO2, and increasing by ;60% for
than in Flame A. These differences in major species H2 when equal diffusivities are specified instead of
are believed to reflect the increasing importance of the normal CHEMKIN treatment of transport prop-
turbulent transport relative to molecular diffusion as erties.
Reynolds number and streamwise distance are in- The cpdf ’s of YOH at x/d 4 30 in Flames B–F
creased [19]. A similar effect on YH2O and YCO2 is continue to show higher mass fractions than in
observed in strained laminar flame calculations [17] Flame A. This may also be related to transport ef-
when all diffusivities are set equal to the thermal fects, because the specification of equal diffusivities
diffusivity. This transport effect is also believed to in the a 4 100 s11 laminar flame calculation causes
1092 NON-PREMIXED TURBULENT COMBUSTION

Fig. 5. Conditional pdf’s of the measured mass fractions of H2O, CO2, OH, CO, H2, and NO from x/d 4 5 (A5) and
x/d 4 10 (A10) in the laminar jet flame and from x/d 4 30 in the turbulent flames B–F. Mixture fraction intervals used
in forming the cpdf’s are given in each graph.

a 25% increase in peak YOH. However, interpretation transport effects, and various branches of the NO
is complicated by the fact that the most probable production and reburn chemistry. The prediction of
values of YOH in the cpdf ’s for Flame A are lower these NO results will be an interesting challenge for
than predicted for full-transport laminar flames. It is combustion models.
possible that radiative loss and the convective history Returning to the results at x/d 4 15 (Fig. 4), the
of radical recombination contribute to the observed response of the cpdf ’s of all species to increasing
reduction in OH levels relative to the adiabatic op- Reynolds numbers appears to be consistent with the
posed-flow calculation. The cpdf ’s of CO in Flames combined (and sometimes offsetting) effects of in-
B–E are broader than in Flame A but are centered creases in scalar dissipation and increases in the rela-
at roughly the same mass fraction. The results for tive importance of turbulent transport. Laminar
YNO in the turbulent flames show decreasing mass flame calculations show that the peak mass fractions
fractions with increasing Reynolds numbers. Resi- of H2O, CO2, H2, and NO decrease with increasing
dence time may be the main effect here. However, scalar dissipation or strain rate, whereas the peak
it is difficult to discern the relative importance of values of CO and OH are relatively insensitive to
convective residence time, local scalar dissipation, strain rate. When diffusivities are set equal in the
TURBULENT METHANE/AIR JET FLAMES 1093

laminar calculations (the analog of turbulent trans- REFERENCES


port being dominant), the peak mass fractions of
OH, H2, and NO increase significantly, whereas 1. Masri, A. R., Dibble, R. W., and Barlow, R. S., Prog.
those of H2O and CO increase modestly, and that of Energy Combust. Sci. 22:307–362 (1996).
CO2 decreases. For CO2, the effects of increased 2. Chen, J.-Y., Kollmann, W., and Dibble, R. W., Com-
turbulent transport and increased scalar dissipation bust. Sci. Technol. 64:315–346 (1989).
are additive, contributing to the progression ob- 3. Masri, A. R., Dibble, R. W., and Barlow, R. S., in
served in Fig. 4. At x/d 4 15 these effects tend to Twenty-Fourth Symposium (International) on Com-
offset each other for H2O, H2, and CO. Unsteadi- bustion, The Combustion Institute, Pittsburgh, 1992,
ness may also contribute to the measured trends, but pp. 317–324.
computational studies [6,7] have required unusual 4. Correa, S. M., Gulati, A., and Pope, S. B., in Twenty-
scenarios of unsteady strain to produce significant Fifth Symposium (International) on Combustion, The
changes in species mass fractions. Combustion Institute, Pittsburgh, 1994, pp. 1167–
1173.
5. Stårner, S. H., Bilger, R. W., Dibble, R. W., and Bar-
low, R. S., Combust. Sci. Technol. 70:111–133 (1990);
Conclusions Combust. Sci. Technol. 72:255–269 (1990).
6. Barlow, R. S. and Chen, J.-Y., in Twenty-Fourth Sym-
These experiments provide a detailed character- posium (International) on Combustion, The Combus-
ization of the effects of turbulence on species mass tion Institute, Pittsburgh, 1992, pp. 231–237.
fractions in piloted methane jet flames. Results show 7. Mauß, F., Keller, D., and Peters, N., in Twenty-Third
significant changes in the cpdf ’s of some species as Symposium (International) on Combustion, The Com-
the jet Reynolds number is increased from the lam- bustion Institute, Pittsburgh, 1990, pp. 693–698.
inar regime through a series of turbulent flames. The 8. Chen, J.-Y. and Dibble, R. W., Combust. Sci. Technol.
following conclusions are drawn: 84:45–50 (1991).
9. www.ca.sandia.gov/tdf/Workshop.html.
1. Conditional pdf’s of OH show a distinct broad- 10. Nguyen, Q. V., Dibble, R. W., Carter, C. D., Fiechtner,
ening and shift toward higher mass fractions as G. J., and Barlow, R. S., Combust. Flame 105:499–510
the Reynolds number is increased. This trend is (1996).
consistent with previous experiments in similar 11. Barlow, R S., Fiechtner, G. J., and Chen, J.-Y., in
flames. Twenty-Sixth Symposium (International) on Combus-
2. Results for H2O, CO2, H2, and OH all appear to tion, The Combustion Institute, Pittsburgh, 1996, pp.
reflect a shift in importance from molecular dif- 2199–2205.
fusion to turbulent transport within the range of 12. Reisel, J. R., Carter, C. D., and Laurendeau, N. M.,
flow conditions and streamwise locations consid- Combust. Sci. Technol. 91:271–295 (1993).
ered. 13. Frank, J. H. and Barlow, R. S., in Twenty-Seventh
3. Mass fractions of CO in the turbulent flames are Symposium (International) on Combustion, The Com-
not significantly higher than those measured in a bustion Institute, Pittsburgh, 1998, pp. 759–766.
laminar jet flame. Rather, YCO in the highly 14. Barlow, R. S., Frank, J. H., and Fiechtner, G. J., “Com-
strained neck region of the piloted turbulent parison of CO Measurements by Raman Scattering
flames (x/d 4 15) tends to be lower than in the and Two-Photon LIF in Laminar and Turbulent Meth-
laminar jet flame. ane Flames,” paper WSS/CI 98S-19 (1998).
4. NO mass fractions tend to decrease with increas- 15. Nefedov, A. P., Sinel’shchikov, V. A., Usachev, A. D.,
ing Reynolds numbers. However, it is difficult to and Zobnin, A. V., Appl. Opt., in press.
discern the relative importance of local scalar dis- 16. Tanoff, M. A., Smooke, M. D., Osbourne, R. J.,
sipation, convective residence time, transport ef- Brown, T. M., and Pitz, R. W., in Twenty-Sixth Sym-
fects, and the various branches of the NO pro- posium (International) on Combustion, The Combus-
duction and reburn chemistry in establishing this tion Institute, Pittsburgh, PA, 1996, pp. 1121–1128.
measured trend. 17. Chen, J.-Y., personal communication of results of op-
posed-flow laminar flame calculations using the Sandia
code and the Miller-Bowman (1989) mechanism, with
Acknowledgments
boundary conditions matching the present jet and cof-
low streams.
This research was supported by the United States De- 18. Bilger, R. W., Stårner, S. H., and Kee, R. J., Combust.
partment of Energy, Office of Basic Energy Sciences. The Flame 80:135–149 (1990).
contributions of G. Fiechtner and C. Carter to the devel- 19. Smith, L. L., Dibble, R. W., Talbot, L., Barlow, R. S.,
opment of the CO LIF system are gratefully acknowl- and Carter, C. D., Combust. Flame 100:153–160
edged. (1995).
1094 NON-PREMIXED TURBULENT COMBUSTION

COMMENTS
George Kosàly, University of Washington, USA. It ap- all measurements (roughly 0.75 mm). Therefore, the ef-
pears to me that the effects you are showing can be some- fective spatial resolution is about the same for all measure-
what better understood via the Damköhler number than ments in flame B. The laminar flame calculation at a 4
via the Reynolds number. 100 s11 in Fig. 3 was selected because it yielded reasonably
good agreement with the measured temperature and major
Author’s Reply. It is correct that several trends in the species. These calculated curves are plotted mainly as vi-
data can be best explained in terms of an appropriately sual aids to help show the widths of the peaks relative to
defined Damköhler number. The increasing probability of the mixture fraction intervals used for the conditional
localized extinction in this series of flames, as shown in PDFs for each species. The very close agreement on CO
Figs. 1 and 2, is an obvious example of the turbulent mixing is gratifying but should not be assigned significance beyond
rates becoming too fast for the chemical reaction rates (de- the fact that agreement is within the estimated experimen-
creasing Damköhler number). The trends in NO mass frac- tal uncertainty. The close agreement near the NO peak is
tion are believed to be mainly influenced by the ratio of fortuitous. OH and NO have the narrowest peaks in the
the thermal NO formation time scale to the convective res- mixture-fraction plots and are the most likely species to
idence time in each flame, which is also a Damköhler num- suffer from spatial averaging effects. However, such effects
ber effect. The decrease in OH mass fraction between x/d are difficult to estimate in flames B–F without information
4 5 and x/d 4 10 in the laminar flame A may be correlated on instantaneous scalar gradients. As noted in the paper,
with a Damköhler number formed from the time scale of the CPDFs for YH2O, YH2, and YOH show increasing mass
the three-body radical recombination reactions and the fractions with increasing jet velocity, which is opposite to
convective time between these measurement locations. A the trend one would expect if spatial averaging was signifi-
difficulty in describing these effects is that there is not a cant. There is some spatial averaging in these measure-
unique definition of the Damköhler number that can be ments, but the qualitative conclusions of this paper do not
associated with all the different trends in the experiment. appear to be sensitive to this potential source of error.
The Reynolds number is used in this paper partly as a con-
venience because it may be defined simply, and the extra ●
words required to define multiple flow and chemical time
Mohy S. Mansour, Cairo University, Egypt. The scatter
scales are avoided. However, the main conclusion of this
plots in flames D, E, and F show different trends at the
paper is that the evolution in the relative importance of
lean and rich side of the reaction zone. How do you explain
turbulent transport vs. molecular diffusion is of central im-
these differences? Does mixing play an important role at
portance in understanding the measured trends in the con-
the lean side while extinction is more dominant at the rich
ditional PDFs of species mass fractions. This evolution is
side?
most appropriately described in terms of Reynolds number
rather than Damköhler number. The cold jet exit Reynolds
Author’s Reply. Two experimental details should be con-
number may be less appropriate than a local turbulence
sider when interpreting scatter plots in Figs. 1 and 2. First,
Reynolds number or some other definition, but it is used
the data are from 4 or 5 radial locations in each flame, and
here for simplicity.
the PDF of mixture fraction may be different for each
flame. This can have some influence on visual impressions.
● Second, the measurement uncertainty in mixture fraction
is roughly 4 B15% of the local value (standard deviation),
Normand M. Laurendeau, Purdue University, USA. Be- and this causes the apparent spreading or blooming of the
cause the effective probe volume would be smaller for CO, scatter plots on the rich side even when there is no extinc-
might this explain the good agreement between your CO tion. Beyond these experimental artifacts, there is a clear
measurements and laminar flame predictions near the peak trend that temperature samples on the rich side drop pro-
of your CO vs. mixture fraction plot as compared to other gressively below the calculated laminar flame curve as the
such plots that you report? In particular, for the OH and jet velocity is increased for flames D, E, and F. This trend
NO mixture-fraction plots, how much of the observed de- is apparent at both streamwise locations, and it is believed
pression at the peak might be related to probe effects? to result from the extinction/mixing/reignition process that
becomes progressively more important in these flames.
Author’s Reply. The experimental data in Fig. 3 that you When there is localized extinction, some mixing occurs
refer to are from flame B, which has a relatively low Reyn- with little or no reaction, and fluid elements move from
olds number and might best be described as having an un- being close to the fully reacted envelope toward the inte-
steady laminar reaction zone. Consequently, the reaction rior of the temperature vs. mixture fraction domain.
zone is typically not far from being normal to the laser Information on the original fuel-side boundary condition
beams. While the beam diameter for CO LIF is smaller becomes lost in this process. When reaction is re-estab-
than for the other measurements, the imaged length of the lished near the stoichiometric mixture fraction, these
probe volume along the axis of the beams is the same for mixed interior elements become the boundary conditions
TURBULENT METHANE/AIR JET FLAMES 1095

for the reacting layer. This phenomenon must also affect the trajectory marked by the laminar flame calculation. Ex-
the lean side, and there is some indication of this in the tinction and reignition are relevant near the stoichiometric
scatter plots for flame F. However, on the lean side, the condition, while mixing is active on both rich and lean
coflowing air provides an ever-present source of fluid at the sides. It is the constraints on the mixing process that lead
original air-side conditions, so one still finds samples near to the foregoing trend in the scatter plots.

S-ar putea să vă placă și