You are on page 1of 14
Combustion and Flame 145 (2006) 481–494 www.elsevier.com/locate/combustflame The blowout mechanism of turbulent jet diffusion flames Chih-Yung
Combustion and Flame 145 (2006) 481–494 www.elsevier.com/locate/combustflame The blowout mechanism of turbulent jet diffusion flames Chih-Yung

Combustion and Flame 145 (2006) 481–494

Combustion and Flame 145 (2006) 481–494 www.elsevier.com/locate/combustflame The blowout mechanism of turbulent jet diffusion flames Chih-Yung

www.elsevier.com/locate/combustflame

The blowout mechanism of turbulent jet diffusion flames

Chih-Yung Wu a , Yei-Chin Chao a, , Tsarng-Sheng Cheng b , Yueh-Heng Li a , Kuo-Yuan Lee a , Tony Yuan a

a Institute of Aeronautics and Astronautics, National Cheng Kung University, Tainan, 701, Taiwan, ROC b Department of Mechanical Engineering, Chung Hua University, Hsinchu, 300, Taiwan, ROC Received 10 June 2005; received in revised form 25 November 2005; accepted 9 January 2006 Available online 10 March 2006

Abstract

The complicated flame stabilization mechanisms and flame/flow interactions in the blowout of turbulent non- premixed jet flames are experimentally studied using phenomenological observation, 2D Rayleigh scattering, 2D laser-induced predissociative fluorescence (LIPF) images of OH, and particle image velocimetry (PIV) techniques. The blowout process may be categorized into four characteristic regions: pulsating, onset of receding, receding, and extinction. Based on experimental findings, a blowout mechanism is proposed. The maximum “waistline” point of the stoichiometric contour, defined as the point where the radial distance between the elliptic stoichio- metric contour and the jet axis reaches a maximum value, can be regarded as the dividing point separating the unstable and stable regions for the lifted flame in the blowout process. If the flame base is pushed beyond the max- imum “waistline” point, the flame will step into the pulsating region and become unstable, triggering the blowout process. The triple flame structure is identified and found to play an important role in flame stabilization within the stable liftoff and pulsating regions. In the pulsating region, the stabilization point of the triple flame moves along the stoichiometric contour, stabilizing the flame where the flame base is bounded by the contours of lean and rich limits. If the flame is pushed beyond the tip of the stoichiometric contour, the stabilization point and triple flame structure vanish and the flame becomes lean. The flame then recedes downstream continuously and finally extinguishes. © 2006 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Blowout; Blowout process; Turbulent diffusion flames; Jet flames; Pulsation region; Receding; Triple flame

1. Introduction

For several decades [1], the blowout phenomenon was regarded as a special limiting point of the liftoff stability of laminar or turbulent jet flames. In the past, interest was mainly focused on predicting the

* Corresponding author. Fax: +886 6 2389940. E-mail address: ycchao@mail.ncku.edu.tw (Y.-C. Chao).

blowout limit and revealing the stabilization mecha- nism of the liftoff flames. Various models and phys- ical mechanisms have been proposed to delineate the liftoff behavior and blowout limits, including the early premixed combustion model [1], the flamelet extinction model [2], the large-scale mixing model [3,4], the combined premixed flame propagation and

flamelet extinction model [5], and the recent triple-

flame model [6–13]. Theories and models used to predict the blowout limits are similar to those de- scribing the stabilization mechanism of the liftoff

0010-2180/$ – see front matter © 2006 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

doi:10.1016/j.combustflame.2006.01.004

  • 482 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

Nomenclature

  • d jet diameter

d ε

effective diameter

f e

H L

acoustic excitation frequencies

theoretically predicted axial distance

from jet exit to lean flammability level

H s

contour

theoretically predicted axial distance

from jet exit to stoichiometric contour

H w theoretically predicted maximum

“waistline” distance, the axial distance

from jet exit to the location where the

radius of the elliptic stoichiometric con-

tour reaches a maximum value

h

s

measured distance from jet exit to stoi-

chiometric contour

h w measured maximum “waistline” dis-

tance

S u

U b

Y s

z

η

ex

η

ph

η

pl

θ ¯ c

θ ¯ e

θ ¯ s

ρ

e

ρ

laminar burning velocity

blowout velocity of pure fuel

stoichiometric mass fraction

axial distance from the jet exit

location where the lifted flame extin-

guishes completely

upper boundary of the flame pulsating

range

lower boundary of the flame pulsating

range

axial profile of mean fuel mass fraction

fuel mass fraction at the jet exit

fuel mass fraction at stoichiometric con-

tour

gas density at jet exit

ambient gas density

flames. Therefore, several existing blowout theories

and models stem from the stabilization mechanism of

the lifted flame. The concept of the premixed flame

model [1] was adopted by Kalghatgi [14] to delin-

eate the blowout velocity. Based on the experimen-

tal data, Kalghatgi [14] was able to empirically scale

the nondimensionalized blowout velocities with the

Reynolds number based on the mean stoichiometric

envelope and a universal formula was proposed. In the

meantime, Broadwell et al. [3] proposed that flame

stabilization occurs when hot gases, which have been

expelled to the edge of the jet by earlier large-scale

turbulent structures, are re-entrained and ignite non-

combusting eddies of the jet. In both theories, the

blowout velocity of mixture gases or diluted fuel also

can be estimated based on the initial conditions of

the fuel at the burner exit. Recently, the theories pro-

posed by Kalghatgi [14] and Broadwell et al. [3] were

examined using an extended database of methane,

propane, and hydrogen jet flames with various inert

dilutions [15]. The results showed that most of the

measured blowout velocities agree with the predic-

tions using a universal formula proposed by Kalghatgi

[14] and with those calculated using the large-scale

model [3] by including a Reynolds number effect.

A more complete database about blowout limits of

a jet flame has been constructed with those inert-

diluted fuels. In addition, the analysis of diffusivity

effects on blowout limits shows that diffusive proper-

ties in terms of mass and thermal diffusivities are not

the dominant parameters in the blowout of turbulent

jet flames. In other words, the Schmidt number does

not play a major role in the turbulent blowout process.

Moreover, in a recent experiment [16], it is shown

that the blowout of a turbulent diffusion jet flame is

a transient process with a series of events occurring

consecutively, though it usually happens rapidly and

unpredictably. It is noticeable that the phenomena of

flame flickering and flame base pulsation can usually

be found before the flame blows out and it is generally

believed that flamefront instabilities play important

roles in the blowout process [17,18]. However, the de-

tailed mechanism of the blowout process is still not

clear.

To delineate the flow/flame interaction character-

istics responsible for blowout, proper tools and instru-

ments for observations and measurements are needed.

Flame instabilities in the blowout process are very

sensitive to external disturbance. Hence, nonintrusive

and simultaneously two-dimensional flame/flow mea-

surements, such as particle image velocimetry (PIV)

[19], are desirable to provide required information

for the blowout process. With the advancement of

high-power laser technology, it becomes feasible to

apply nonintrusive laser techniques in hostile com-

bustion environments. Laser-based diagnostic tech-

niques have been successfully developed to provide

nonintrusive spatially and temporally resolved mea-

surements of flow characteristics and chemical prop-

erties [20]. In the present study, Rayleigh scattering

imaging and laser-induced fluorescence (LIF) imag-

ing are applied to measure local mixture fraction

ahead of the flame base and to identify the flame

base location, respectively. Rayleigh scattering imag-

ing is one of the good methods for binary-mixture

flow visualization of the flow field under suitable as-

sumptions [21,22]. On the other hand, laser-induced

fluorescence (LIF) provides the ability to detect flame

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

483

radicals and pollutant species at ppm or even sub-ppm

level and has therefore received considerable attention

in flame measurements [23,24]. To resolve some im-

portant issues in combustion processes, simultaneous

Rayleigh imaging combined with LIF imaging [25]

can be utilized to delineate the spatial structure of re-

action zones and the existence of either thin flamelet

zones or broad distributed ones is of primary impor-

tance in modeling turbulent flames.

In a previous study [16], the blowout of a partially

premixed methane jet flame was delineated and clas-

sified. To simplify the initial conditions of the fuel

stream at the jet exit and to verify the effects due to

multiplex parameters when the jet flame blows out,

jet flames with inert-diluted fuel were used for a se-

ries of experiments. Blowout limits can be estimated

simply based on initial conditions of fuel at the jet

exit [15], although in the blowout process a reced-

ing liftoff flame is strongly affected by the upstream

flow. Local mixture fraction and thermal/flow charac-

teristics ahead of the flame base may vary with liftoff

height due to the turbulent cascade and mixing. To

look further into the intrinsic physical and chemical

properties of the flame base under blowout conditions,

and to probe into the relationship between local prop-

erties of flow ahead of the flame base and initial con-

ditions at the jet exit, the present study describes two-

dimensional laser diagnostics, including qualitative

tracing of the flame base, PIV, and simultaneous LIPF

and Rayleigh imaging of inert-diluted methane and

propane turbulent jet flames in the blowout process

and the results are quantitatively compared with the

theoretical prediction based on a model extended

from the universal formula of blowout velocity pro-

posed by Kalghatgi [14]. Furthermore, a new blowout

mechanism of a turbulent jet diffusion flame is pro-

posed based on the measurement and theoretical re-

sults.

2. Experimental setup

The experimental setup is shown schematically

in Fig. 1a. The jet flame burner consists of a well-

contoured circular nozzle 5 mm in diameter, from

which the fuel/diluent mixture emerges. The noz-

zle wall is contoured with a fifth-order polynomial

profile, and the area contraction ratio is 400. Fu-

els and diluents are metered by rotameters and elec-

tronic flowmeters. The accuracies of the rotameters

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 483 radicals and pollutant species

Fig. 1. Experimental setup: (a) essentials of the experimental arrangements; (b) arrangements of lasers and optics of PIV device; (c) schematic diagram of 2D Rayleigh and 2D LIPF.

  • 484 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

for fuel metering and electric flowmeter for air me-

tering are 0.5% of full scale and ±1.5% of full scale

from 10 to 100% of full scale respectively. We have

provided these data in the revision. Readings of ro-

tameters and electronic flowmeters are recorded to

calculate the experimental blowout velocity, defined

as the bulk fuel stream velocity when the flame blows

out. Compressed air from the tank and diluents and

fuel from the cylinders are filtered, metered, and pre-

mixed in the mixing chamber. In this study, 30%

nitrogen-diluted and argon-diluted methane and 50%

nitrogen-diluted and argon-diluted propane are inves-

tigated, and the flame conditions accompanied by var-

ious fundamental parameters tested in this study are

also listed in Table 1. Furthermore, noise reduction

and settling chambers are used to improve the flow

quality. The nozzle exit velocity shows a top-hat pro-

file, and the turbulence intensity at the jet centerline

is about 0.5%. The whole system is placed inside

an anechoic room. Characteristic frequency of the

lifted jet flame in the cold flow region is examined

by a probe microphone (B&K 4182). A non-catalytic-

coated R-type (Pt/Pt-13Rh) thermocouple of diameter

50 µm is used to measure the temperature and the

characteristic frequency at the flame base. Images of

flame structures and flame-base tracing are obtained

by a high-sensitivity three-chip color CCD camera

(Sony DXC-9000) with external triggering functions.

The frame rate can reach 60 frames per second, and

the serial images are digitized by the frame grabber

for further analysis.

Arrangement of the PIV devices, including two

Nd:YAG lasers and optics, is shown in Fig. 1b. The

laser beams are aligned with optics through two po-

larizers and a wave plate. The resulting beam is then

expanded by three cylindrical lenses into a laser sheet

approximately 0.7 mm in thickness, which is actually

measured on the projection screen. The time interval

of the PIV system is controlled by a pulse signal/delay

generator. The fuel and air streams are seeded with

sieved ne Al 2 O 3 particles of sizes less than 10 µm.

A high-resolution, high-sensitivity, and low-dark-

current camera (SharpVision 1300DE) is used for

image recording. This CCD which is equipped with

a progressive scan interline CCD sensor is especially

suitable for PIV. The image array has 1300 × 1030

pixels, limited to 1280 × 1024 in practice, and the

pixel size is 6.7 × 6.7 µm. All images are captured,

digitized through a 16-bit digitizer, and recorded on

a hard disk for further analysis.

For identification of the reaction zone, the laser-

induced predissociative fluorescence (LIPF) of OH

molecules from v = 0 to v = 3 in the A 2 Σ X 2 Π

system is employed [26]. The laser beam spreads into

a thin sheet of height 34 mm and thickness 0.2 mm

by a single cylindrical lens (f = 1000 mm) and inter-

sects vertically through the flame axis. Only the 25-

mm central portion of the laser sheet, where the laser

intensity is high and uniform, is used for imaging. The

OH fluorescence signal is imaged onto an intensified

CCD camera (576 × 384 pixels) with a UV camera

lens (Nikkor, f = 105 mm, f/4.5). A 10-mm-thick

butyl acetate liquid filter is placed in front of the cam-

era to remove the Rayleigh scattering. The OH fluo-

rescence signal is collected at 297 nm, corresponding

to the fluorescence produced from the 3 2 tran-

sition. On the other hand, the 2D Rayleigh scatter-

ing system with KrF laser is employed to visualize

the upstream fuel–air mixing characteristics and to

obtain qualitative and quantitative concentration pro-

files. 2D Rayleigh imaging can be performed using an

experimental setup identical to the laser-induced pre-

dissociative fluorescence (LIPF) system, except that

a narrow-band filter of 248 nm is used instead of the

Table 1 Measured and theoretical estimated parameters of each condition for blowout process observation

 

30% N 2 -diluted CH 4

30% Ar-diluted CH 4

50% N 2 -diluted C 3 H 8

50% Ar-diluted C 3 H 8

U b (m/s) a

42.6

34.0

52.6

47.6

f e (kHz)

2.2

2.0

4.7

4.5

S u (cm/s)

31.6

32.3

36.9

37.7

H w (x/d )

24.0

22.0

32.0

30.0

h w (x/d )

20.0

19.0

26.0

25.0

η ph (x/d )

21.0

18.0

28.0

23.0

H s (x/d )

42.2

39.3

54.5

51.0

h s (x/d )

35.0

37.5

51.0

49.5

η pl (x/d )

32.0

38.0

50.0

49.0

H L (x/d )

63.8

59.3

82.4

76.9

η ex (x/d )

59.0

61.0

88.0

80.0

Note. d = 5 mm. a Proposed in previous work [15].

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

485

butyl acetate filter. In the present study, 2D LIPF and

2D Rayleigh imaging devices are applied and trig-

gered simultaneously to identify the mixture level of

fuel and oxidizer in the flame front. A schematic dia-

gram of 2D Rayleigh image and 2D LIPF systems is

shown in Fig. 1c.

  • 3. Results and discussion

3.1. Manipulation of blowout of a jet flame

Controlling the initial conditions of the jet flame

blowout process for repeatability in experiments is

possible with proper acoustic excitation [22]. Con-

ventionally, blowout velocity is measured by slowly

increasing the bulk jet exit velocity until the flame

blows out. However, it is very difficult to precisely

repeat the blowout process due to the inaccuracy of

manually controlled volumetric flow rates of fuel or

mixtures and the inertia associated with the pipeline

when triggering the blowout process. An effective

way to overcome this difficulty is to apply acoustic

excitation on the jet flame. The effects of acoustic

excitation on the jet have been reviewed by Ho and

Huerre [27]. They summarized and discussed effects

of major excitation parameters including forcing fre-

quency, forcing level, and phase angle on flow struc-

tures. Chao and his co-workers have conducted a se-

ries of jet flow and flame studies using acoustic ex-

citation [16,22,28,29]. They found that the blowout

limits can be extended and the irregular onset of the

blowout process can be controlled and repeated by

tuning the excitation frequency. This strategy of us-

ing acoustic excitation is to avoid the difficulty of

convective delays associated with slowly varying the

flow rates to reach blowout and to provide aligned

initial conditions for repeated measurements of the

dynamic blowout process. The function of turning off

the acoustic excitation is triggered by a slope detector,

a counter, and a relay switch. The residual acoustic ef-

fects after the acoustic excitation is turned off will not

affect the transition of the blowout process in prac-

tice because the sound propagation speed is far higher

than the flow velocity. In other words, it may be as-

sumed that the blowout process is triggered as soon

as the excitation is turned off.

To obtain the proper excitation frequency for this

study, the fundamental frequency of the fuel stream

at the blowout velocity must be determined first.

A microphone probe is applied to conduct the mea-

surements. However, to obtain a well-defined initial

condition, the best excitation frequency and ampli-

tude of each flame condition needs to be tested and

tuned manually. The chosen acoustic excitation fre-

quencies (f e ) and the blowout velocities (U b ) for each

flame condition are listed in Table 1. Observation and

measurement of the flame blowout process are con-

ducted with proper acoustic excitation to hold the

lifted jet flame at the blowout limits, which have been

measured in our previous study [16], and the blowout

transient is triggered when acoustic excitation is elec-

tronically turned off at a prescribed phase angle.

3.2. Blowout process of a jet flame

Observation of the blowout of a jet flame can be

achieved by tracing the flame base images. The con-

tinuous blowout images are recorded by a CCD cam-

era with a frame rate of 30 frames per second, and the

images are digitized by the frame grabber for analysis.

The flame base is then identified with digital image

processing, and the flame-base trace can be expressed

as a function of time. According to the dynamic char-

acteristics of jet flames, the transient blowout process

can be divided into four regions: pulsating, onset

of receding, receding, and extinction regions [16].

In this study, to further understand the characteris-

tic blowout processes of different fuel/oxidizer mix-

tures, flame-base traces of inert diluted methane and

propane flames are made, and typical traces and flame

images in the four regions of the 30% nitrogen-diluted

methane jet flame are shown in Fig. 2. Similarly to

the previous study [16], the initial condition of the

blowout process is controlled by proper acoustic ex-

citation. As soon as the acoustic excitation is turned

off at a prescribed phase, the process engages and

the flame remains, and then the flame base starts to

pulsate in the axial direction at roughly 1 Hz [16].

The whole flame is blue in color and looks simi-

lar to a lifted flame. This is the pulsating region, as

seen in photograph α in Fig. 2. The flame tip is obvi-

ously affected by the buoyancy-induced vortex. When

the flame base reaches the height of x/d = 34, the

flame simply moves downstream continuously, being

unable to come back. Red-hot stripes can be found

on the blue flame. This is the onset of the reced-

ing region as shown in photograph 2β . When the

flame moves further downstream, the flame tip be-

comes red, yet the flame base remains blue. This is

the receding region as illustrated in photograph 2γ .

Finally, when the height reaches at x/d = 59, the

flame becomes shorter and the base is blue, as noted

in photograph 2δ . Flame extinction occurs in this re-

gion. In the process red stripes and spots are found in

the flames, indicating local low temperature and un-

burned regions.

The pulsating frequency of the flame is mea-

sured using an ion probe placed at a fixed position

close to the flame base before blowout. Fig. 3 shows

the pulsating frequency for the 30% nitrogen-diluted

methane flame. Low pulsating frequencies, around

  • 486 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

486 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 Fig. 2. Typical flame-base

Fig. 2. Typical flame-base trace in blowout process of the 30% nitrogen-diluted methane flame with typical flame image in each region: pulsating, onset of receding, receding, and extinction.

1–5 Hz, similar to those previously reported [16] are

observed in the flame base. Similar phenomena are

also found in the 30% argon-diluted methane flame

and the 50% nitrogen- and argon-diluted propane

flames, despite the apparent differences in Lewis

number [15]. The measured pulsating frequency

is also similar to the pulsating frequency of non-

premixed jet flames near extinction proposed by Füri

et al. [17], but less than the characteristic frequency

(10–18 Hz) of buoyancy-induced large toroidal vor-

tices, which is a type of Kelvin–Helmholtz instabil-

ity [30]. For the current turbulent blowout flames, the

thermal-diffusive instabilities may not be as dominant

in the blowout process [15] as they are in the lami-

nar flames [31] and may be suppressed by increasing

convective velocity of flow [17]. The mechanism for

instability as the flame moves downstream of the

maximum waist of the stoichiometric contour is still

not clear and cannot be given based on current results.

Other instability modes may exist and play a role in

the blowout process. It is unclear now and needs to be

clarified by further experiments.

  • 3.3. Fuel/air mixing and statistics of flame behaviors

It is widely accepted that the universal formula

proposed by Kalghatgi [14] based on the premixed

flame model [1] can provide an accurate estimate of

the blowout velocities for a variety of fuels with dif-

ferent degrees of dilution for various inert species [15].

In the formula, the position of the flame base closely

corresponds to the locations of the stoichiometric con-

tour. Kalghatgi’s model can be extended to predict

major characteristics of the blowout process. First of

all, to identify the locations of the stoichiometric con-

tour, the theoretical model proposed by Birch et al.

[32] for the concentration distribution issued from

a free jet is adopted. A normalized axial profile,

θ ¯ c /

θ ¯ e ,

in a free jet can be shown as

θ ¯ c /

θ ¯ e = k 1 d ε /(z + a 1 ), d ε = d(ρ e ) 0.5 ,

(1)

where θ ¯ , ρ , z , and d ε

are the mean fuel mass frac-

tion, density, and axial distance from the jet exit and

effective diameter, respectively, and subscripts e and

indicate properties at the jet exit and ambient con-

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

487

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 487 Fig. 3. Pulsating frequency

Fig. 3. Pulsating frequency of the 30% nitrogen-diluted methane flame.

ditions. Reported values of the constants k 1 and a 1

are 4.0 and 5.8d , respectively. Hence, the distance

along the jet axis from the jet exit to the location

where the mean fuel concentration falls to the stoi-

chiometric level, H s , can be derived from Eq. (1) as

H s = 4

¯

θ e

¯

θ s

1/2 + 5.8 d,

ρ

e

ρ

(2)

where subscripts s indicate properties at stoichiomet-

ric level. Equation (2) can also be used to calculate the

distance of the lean flammability limit, H L , by replac-

ing

¯

θ s by the lean flammability-limit mass fraction.

Distances estimated with Eq. (2) for the stoichiomet-

ric level and the lean flammability limits are listed in

Table 1. Furthermore, for the radial profiles of the jet,

the normalized profiles of mean mass fraction concen-

tration are accurately described by the Gaussian-type

function [32,33], and can be shown as

¯

¯

θ/ θ

c

= e {−D(r/z) 2 }

,

(3)

where D is found to be 73.6. Substituting Eq. (1) into

¯

¯

Eq. (3) and replacing θ by θ s , the stoichiometric con-

tour can be expressed as a function of r and z and is

shown as

r

2 +

z 2

¯

ln

θ s (z 5.8d)

¯

73.6

4

θ 0 d ε

= 0.

(4)

The stoichiometric contour, Eq. (4), can be used to

find the maximum “waistline” point of the stoichio-

metric contour by setting dr/dz to zero, and one can

obtain

 
 

¯

¯

= −

 

2 ln

θ s

5.8d

θ s

z

.

(5)

¯

z

¯

 
 
  • 4 θ 0 d ε

  • 4 θ 0 d ε

z 5.8d

Equation (5) is solved numerically for z to find the

maximum “waistline” distance, H w , along the axis

from the jet exit to the location where the radius of the

elliptic stoichiometric contour reaches a maximum

value. These two parameters, H s and H w , character-

ize the stoichiometric contour and can be shown to

play important roles in the proposed mechanism of

the blowout process.

The theoretical prediction above can be verified

by experimental measurements. Two-dimensional

Rayleigh scattering imaging is applied to define the

stoichiometric contour for the inert-diluted methane

and propane cold jets. Fig. 4a shows the averaged

Rayleigh imaging of the 30% inert-diluted methane

jet. In this figure, each Rayleigh image is an aver-

age of 30 single-pulse measurements and the figure is

composed of 15 images measured at different heights.

Due to the limitations of the translation stage, the

measured flame height can only reach up to x/d = 60.

Being proportional to the Rayleigh scattering cross

section of mixtures, the color level in stoichiomet-

ric conditions, as well as the lean and rich flammable

limits, can be estimated from the binary mixing of the

Rayleigh scattering signals and is also shown in the

figure. Fuel concentration decreases axially and radi-

ally from the jet exit due to entrainment and turbulent

mixing with ambient air. Furthermore, the two major

  • 488 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

488 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 Fig. 4. (a) Averaged

Fig. 4. (a) Averaged mixing level distribution of the 30% nitrogen-diluted methane flame; (b) probability distribution of transient flame base location during blowout process; (c) flame-base propagation velocity of the 30% nitrogen-diluted methane flame based on sequential images.

characteristic parameters of the elliptic stoichiometric

contour, h w and h s , in which lower case is used here

to distinguish the measurement data, can be identi-

fied via the quantitative Rayleigh measurements. The

measured h w and h s are listed in Table 1. To fur-

ther identify the flame pulsating range, the probability

distribution of the transient flame base location dur-

ing the blowout process using 10 samples of flame

base tracing is shown in Fig. 4b and is marked as the

pulsating region in which the probability is greater

than 2%. The locations of the upper and lower bound-

aries of the flame pulsating range, η ph , and η pl , are

also tabulated in Table 1 for further comparison with

other parameters. Moreover, η ex , defined as the lo-

cations where the lifted flame is extinguished com-

pletely, are also determined based on statistical results

with flame base tracing of the blowout process and are

listed in Table 1. To further delineate the flame/flow

interaction during the blowout process, typical flame

base propagation velocities of 30% nitrogen-diluted

methane flames are plotted in Fig. 4c against the flame

base locations calculated from sequential images. The

results show that the flame base propagation veloci-

ties scatter randomly between 1 and 1 m/s when the

flames are in the pulsating region and x/d is less than

the onset position of receding flames. The flame base

oscillation may be affected by pulsating instabilities

and flow interactions so that the propagation veloc-

ities jitter irregularly. On the other hand, the flame

base propagation velocity increases almost linearly

with x/d from the onset of receding to extinction.

A similar trend was found in the 30% argon-diluted

methane and the 50% nitrogen- and argon-diluted

propane flames.

The measured and predicted results of the charac-

teristic parameters listed in Table 1 are also compared

to verify the model described above. By careful exam-

ination of the data in Table 1, one can find that there

are very good correspondences between the measure-

ment maximum “waistline” point h w and the lower

boundary of the pulsating region η pl , as well as the tip

of the stoichiometric contour h s and the upper bound-

ary of the pulsating region η ph for all the flame cases

tested. Even the predicted lean flammability limit

data H L have good agreements with the measurement

extinction point data η ex for all the cases. Compar-

isons of the measured and predicted results in Ta-

ble 1 indicate that the measured maximum “waistline”

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

489

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 489 Fig. 5. Simultaneous two-dimensional

Fig. 5. Simultaneous two-dimensional OH image superposed with Rayleigh scattering image of the 30% nitrogen-diluted methane flame in the pulsating region.

point h w and the onset of receding h s takes place at

the axial position slightly upstream of the theoretical

predictions (H w and H s ) for all inert-diluted methane

and propane flames. Though the model slightly over-

predicted these parameters, fair agreement between

the predicted and measurement data is achieved.

The propagation and formation of a lifted non-

premixed jet flame was studied by Lyons and Wat-

son [34]. It is believed that the propagation of a lifted

flame and flame shape is related to the stoichiomet-

ric mass fraction contour Y s . For the blowout of

a jet flame, based on the premixed model, Vanquick-

enborne and van Tigglen [1] have shown that the

blowout process can be triggered when the flame base

is pushed downstream where the stoichiometric mass

fraction Y s contour reaches its maximum radial width.

In view of the good correspondence between h w and

η pl , it becomes reasonable to consider the location

where the Y s contour reaches its maximum radial

width as a dividing point separating the stable and

unstable regions. Instability in terms of pulsation usu-

ally takes place in a flame during the blowout process.

The appearance of the pulsating instability may trig-

ger the blowout process. In addition, the correspon-

dence between h s and η ph may imply that the on-

set of receding in the blowout process has a strong

connection with the diminution of the stoichiometric

concentration and the fuel-lean condition plays an im-

portant role in the flame recession and extinction of

the blowout process, as the agreement between H L

and η ex may imply.

To verify the level of fuel/air mixing and to iden-

tify the reaction zone near the flame base, simulta-

neous measurements of LIPF-OH superposed with

Rayleigh images of the flame base are performed.

Single-shot images of Rayleigh and OH for the 30%

nitrogen-diluted methane flame in the pulsating re-

gion are shown in Fig. 5. With regard to the noise in

Rayleigh images, the Mie scattering from surround-

ing dust cannot be avoided completely. In a Rayleigh

image, the distinct noise dots can be identified with

simple image-processing principles and are removed

in the calculation. In other words, the calculation is

based mainly on regions where noise dots do not exist.

Despite these limitations of the Rayleigh technique,

it is still the best tool to image the distribution of

fuel/air mixing levels in nonreacting areas. In addi-

tion, since the flame thickness in the jet flame base

is thin and the intensity is weak, the LIPF-OH im-

age becomes a little bit noisy. Here, we have no in-

tention of using the Rayleigh scattering signals for

temperature measurements [35], as the variation of

the Rayleigh cross section between fuel and air is

very large in the current flames. Simultaneous mea-

surements of Rayleigh and OH images provide useful

information about fuel/air mixing and its associated

reaction zones. Fig. 5 indicates that fuel and air are

well mixed to stoichiometric levels or even lean con-

ditions before approaching the flame base. Similar

results of inert-diluted propane flames are also found,

but the images are less noisy because the Rayleigh

cross-section of methane is much smaller than that

of propane. Large cross-sectional differences between

fuel and air offer better signal-to-noise ratio and re-

duce measurement errors. For simultaneous measure-

ments of Rayleigh and OH images, the heat release

from the flame base [36,37] may affect the proper-

ties of upstream unburned reactants near the flame

base. It is well known that the Rayleigh scattered sig-

nal depends on the composition and temperature. If

the composition is fixed, the signal is inversely pro-

portional to the temperature. Namely, the heat release

may cause underprediction of equivalence ratio. How-

ever, as compared with the major length scale in the

jet flame blowout process, which is described in terms

of the stoichiometric contour, the length scale of the

heated region is much smaller and the induced error

  • 490 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

due to streamline divergence of the heated region can

be neglected for the present purpose.

3.4. Simultaneous fuel/air characteristics near

flame gbase

To further understand the relationship between

flow velocity and flame base propagation, typical

plots of simultaneous PIV velocity distributions ahead

of the flame zone (marked by dashed lines) for

the 30% nitrogen-diluted methane flame in the typ-

ical pulsating, onset of receding, and receding re-

gions are shown in Figs. 6a–6c, respectively. Simi-

larly to the method used by Schefer and Goix [10]

and Muñiz and Mungal [7], the decrease in particle

density and scattering cross section in flame can be

490 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 due to streamline divergence
490 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 due to streamline divergence

Fig. 6. Velocity distribution ahead of flame base of the 30% nitrogen-diluted methane flame: (a) 14 < x/d < 26; (b) 27 < x/d < 39; (c) 41 < x/d < 53.

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

491

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 491 Fig. 6. ( continued

Fig. 6. (continued )

used as a marker for the high-temperature zone. In

Fig. 6a the velocity of the unburned gases of the cen-

terline at this height is about 8 m/s and decreases

with x/d . The velocity just ahead of the flame base

decreases approximately to the laminar burning ve-

locity of a few tens of centimeters per second. In

Fig. 6b, the flame base is located at the onset posi-

tion of recession, and the velocity just ahead of the

flame base is also approximately equal to the laminar

burning velocity of the 30% nitrogen-diluted methane

flame. However, in Fig. 6c, the velocity ahead of the

flame base is about 2–3 m/s higher than the lam-

inar burning velocity. Similar results for the 30%

argon-diluted methane flame and the 50% nitrogen-

and argon-diluted propane flames are also obtained.

Based on the PIV measurements, the velocity distri-

butions at typical heights of the stabilization point in

the three characteristic blowout regions can be cal-

culated. Typical results for the 30% nitrogen-diluted

methane flames are shown in Fig. 7. Briefly, when

the flame base is located below the onset position of

recession, the flame base always stabilizes at a po-

sition where local velocities are around the laminar

burning velocity (0.8–3 S u ) and the flame may pul-

sate at a frequency of roughly 1 Hz. However, when

the flame base is pushed downstream beyond the on-

set position of recession, the stabilization point of

the flame moves toward the center of the jet, where

the flame suffers from much higher flow velocity

and becomes thinner and weaker. In addition, in the

recession region the local velocity at the flame stabi-

lization point increases with the height until the flame

vanishes. These phenomena are also found in other

inert-diluted flames.

The triple flame structure of turbulent lifted flames

is often distorted by vortices [38] so that the three

branches may not always be found. For the purpose

of the present study, we do not intend to involve in

detailed diagnostics of concentration and velocity of

triple flame (edge flame) structure; instead, results

from others, such as Lyons and Watson [34] for con-

centration and Muñiz and Mungal [7] for velocity, are

borrowed to compare with the present results. Due to

the high distortion of the flame base, especially in the

blowout process, and the difficulty of observation via

planar laser diagnostics, the major evidence of exis-

tence of a triple flame structure in turbulent flame base

may be found from propagation velocity. We have

found that the velocity just ahead of the flame base

is approximately equal to the laminar burning veloc-

ity when the flame base is located within the tip of the

stoichiometric contour, which agrees with results of

Muñiz and Mungal [7] and the triple flame model of

Kioni et al. [9] and Ruetsch et al. [36]

3.5. The blowout mechanism

Similar to the results reported in several previous

studies using PIV [7–10], important evidence that the

fluid velocity just ahead of the lifted flame base is

low, close to the laminar flame speed, and the fact

that the velocity decreases continuously to the laminar

flame velocity along the stream up to the triple point

of the triple flame have been described in this study.

  • 492 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

492 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 Fig. 7. Velocity distribution

Fig. 7. Velocity distribution just ahead of the 30% nitrogen-diluted methane flame at different heights.

Moreover, in our previous study [15] of an identical

flame configuration, we clearly identified the triple

flame structure in the OH images. Hence, in the pul-

sating region, a triple flame structure is identified with

a propagation velocity very close to the laminar flame

speed near the stabilization point of the triple flame at

the flame base. The flame base in the pulsating region

is found mostly in the axial locations corresponding to

the range from the maximum “waistline” point to the

tip of the stoichiometric “elliptic” contour. However,

in the receding region, the flame base is pushed down-

stream and follows local flow velocity. Furthermore,

the results of simultaneous fuel/air mixing level mea-

surements show that the flame base propagates along

the stoichiometric layer in the pulsating region. These

instantaneous and simultaneous 2D mixing level and

velocity results provide specific and sound evidence

to verify the existence of the triple flame structure and

the stabilization point of the triple flame [7] serves to

provide a suitable flame base stabilization mechanism

in the pulsating region during the blowout of turbulent

jet flames.

The above findings of the flame base behavior

and evolution characteristics in the blowout process

can be used to construct the blowout mechanism of

a turbulent jet diffusion flame. To illustrate the mech-

anism of stabilization, propagation, and breakdown of

the triple flame structure in a lifted jet flame in the

blowout process, the bold dashed and dotted curves

denoting contours of the stoichiometric level and the

lean flammable limit are compared in Fig. 8, and the

corresponding heights, H s and H L , as well as the rich

flammable limit contour are also shown in Fig. 8.

In the proposed mechanism, if the lifted flame is

initially stabilized in the range upstream of the max-

imum “waistline” point of the stoichiometric contour

while the jet exit velocity (U o ) is between the liftoff

velocity (U l ) and the blowout velocity (U b ) (Fig. 8b),

the lifted flame is stable and remains lifted (Vanquick-

enborne and van Tigglen [1]). In this region, very

good agreement between experimental and theoreti-

cal estimates of the liftoff heights has been reported

in previous work [39–42]. In general, the liftoff height

is proportional to jet exit velocity (U o ) and inversely

proportional to the square of the maximum laminar

flame speed. As shown in Fig. 8c, if the flame base

of the lifted flame is pushed beyond the maximum

“waistline” point while the jet exit velocity is equal

to the blowout velocity, the flame will step into the

pulsating region of the blowout process and become

unstable. In either the stable or unstable region, as

shown in Figs. 8b and 8c, the jet flame is hollow

in structure. In the pulsating region, the triple flame

structure is found in the flame base and the fast sto-

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

493

C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494 493 Fig. 8. Schematic of

Fig. 8. Schematic of the proposed blowout process mechanism.

ichiometric branch of the triple flame in the flame

base, which moves along the stoichiometric contour,

serves as the stabilization point of the triple flame to

provide the essential element to stabilize the flame.

The flame base is bounded by the boundaries of lean

and rich limits. The pulsating region corresponds to

the axial range from the maximum “waistline” point

to the tip of the stoichiometric contour. The unsta-

ble counterbalance of the local flow velocity and the

flame propagation speed at the stabilization point of

the triple flame result in whole flame pulsation in the

pulsating region. As the flame base is pushed down-

stream, the flame base moves toward the center along

the contour and suffers from higher flow velocity. As

soon as the flame is pushed by the flow beyond the tip

of the stoichiometric contour (Fig. 8d), the stabiliza-

tion point and the stoichiometric branch of the triple

flame structure vanish and the flame becomes lean.

In this region, the hollow-cone structure disappears

and the jet flame base becomes disklike. The flame

recedes downstream continuously and finally extin-

guishes (Fig. 8e).

Triple flame stabilization plays an important role

in flames in stable liftoff in the pulsating regions

of the blowout process. The stabilization and pul-

sation of the stabilization point of the triple flame

along the stoichiometric contour constitute the ma-

jor dynamic behavior in the pulsating region of the

blowout process. The stoichiometric contour can be

determined by the initial gas properties and the initial

velocity at the jet exit. Therefore, for a turbulent flame

not only the blowout limit [15] but also the blowout

process and the accompanied dynamic behaviors can

be estimated and characterized based on the initial ve-

locity/Reynolds number and gas properties at the jet

exit.

4. Conclusion

Through phenomenological observation and de-

tailed measurements of the mixing and velocity distri-

butions using 2D Rayleigh scattering, 2D LIPF-OH,

and PIV techniques, a blowout mechanism is pro-

posed to delineate the dynamic flame-base behavior

and evolution characteristics in each characteristic re-

gion, i.e., the pulsating, onset of receding, receding,

and extinction regions, of the blowout process of a tur-

bulent jet diffusion flame. The mechanism is primary

based on the findings that triple flame structures are

found in the flame base in the pulsating and onset of

receding regions and the correspondence of the flame

base locations in each region with the stoichiometric

and lean limit contours of the premixed model. The

stabilization and pulsation of the stabilization point

  • 494 C.-Y. Wu et al. / Combustion and Flame 145 (2006) 481–494

of the triple flame along the stoichiometric contour

constitute the major dynamic behavior in the pulsat-

ing region and the diminution of the stoichiometric

branch of the triple flame and the fuel-lean condition

lead to the recession and extinction of the flame in the

blowout process. The proposed blowout mechanism

based on triple flame and stoichiometric contour also

provides an explanation for the fact that the blowout

process of a turbulent jet diffusion flame can be es-

timated and characterized based on the initial veloc-

ity/Reynolds number and gas properties at the jet exit

without knowing the local flame/flow conditions of

the liftoff flame near blowout.

Acknowledgment

Financial support by the National Science Coun-

cil, Republic of China, through Projects NSC90-

2212-E-006-120, NSC91-2212-E-006-039, and NSC-

92-2212-E-006-058 is gratefully acknowledged.

References

[1] L. Vanquickenborne, A. van Tigglen, Combust. Flame 10 (1966) 59–69.

[2] N. Peters, F.A. Williams, AIAA J. 21 (1983) 423–429. [3] J.E. Broadwell, W.J.A. Dahm, M.G. Mungal, Proc. Combust. Inst. 20 (1984) 303–310. [4] W.J.A. Dahm, R.W. Dibble, AIAA Paper 88-0538,

1988.

[5] R.C. Miake-Lye, J.A. Hammer, Proc. Combust. Inst. 22 (1988) 817–824. [6] C.M. Müller, H. Breitbach, N. Peters, Proc. Combust. Inst. 25 (1995) 1099–1106. [7] L. Muñiz, M.G. Mungal, Combust. Flame 111 (1997)

16–31.

[8] A. Liñán, Acta Astronaut. 1 (1974) 1007–1039.

[9] P.N. Kioni, B. Rogg, K.N. Bray, A. Liñán, Combust. Flame 95 (1993) 276–290. [10] R.W. Schefer, P.J. Goix, Combust. Flame 112 (1998)

559–574.

[11] K.A. Watson, K.M. Lyons, J.M. Donbar, C.D. Carter, Combust. Flame 117 (1999) 257–271. [12] K.A. Watson, K.M. Lyons, J.M. Donbar, C.D. Carter, Combust. Flame 119 (1999) 199–202. [13] E.F. Hasselbrink, M.G. Mungal, Proc. Combust. Inst. 27 (1998) 213–221. [14] G.T. Kalghatgi, Combust. Sci. Technol. 26 (1981) 233–

239.

[15] Y.-C. Chao, C.-Y. Wu, K.-Y. Lee, Y.-H. Li, R.H. Chen,

T.S. Cheng, Combust. Sci. Technol. 176 (2004) 1735–

1753.

[16] Y.-C. Chao, Y.-L. Chang, C.-Y. Wu, T.S. Cheng, Proc. Combust. Inst. 28 (2000) 335–342. [17] M. Füri, P. Papas, P.A. Monkewitz, Proc. Combust. Inst. 28 (2000) 831–838. [18] E.W. Christiansen, C.K. Law, C.J. Sung, Proc. Com- bust. Inst. 29 (2000) 807–814. [19] R.J. Adrian, Appl. Opt. 25 (1986) 3855–3858. [20] A.C. Eckbreth, Laser Diagnostics for Combustion Temperature and Species, second ed., Gordon and Breach, 1996. [21] M.C. Escoda, M.B. Long, AIAA J. 21 (1983) 81–

84.

[22] Y.-C. Chao, C.-Y. Wu, T. Yuan, T.S. Cheng, Combust. Sci. Technol. 174 (2002) 87–110.

[23] D.R. Crosely, G.P. Smith, Opt. Eng. 22 (1983) 545–

553.

[24] K. Kohse-Höinghaus, J. Prog. Energy Combust. Sci. 22 (1994) 203–279. [25] K.A. Watson, K.M. Lyons, J.M. Donbar, C.D. Carter, Combust. Flame 123 (2000) 252–265. [26] P. Andresen, A. Bath, W. Gröger, H.W. Lülf, G. Meijer, J.J. ter Meulen, Appl. Opt. 27 (1988) 365–378.

[27] C.M. Ho, P. Huerre, Annu. Rev. Fluid Mech. 16 (1984)

365–424.

[28] J.M. Han, Ph.D. thesis, National Cheng Kung Univer- sity, Taiwan (1992). [29] Y.-C. Chao, M.S. Jeng, Proc. Combust. Inst. 24 (1992)

333–340.

[30] J. Buckmaster, Y. Zhang, Combust. Theory Modeling 3 (1999) 547–565. [31] J. Lee, S.H. Chung, Combust. Flame 127 (2001) 2194–

2204.

[32] A.D. Birch, D.R. Brown, M.G. Dodson, J.R. Thomas, J. Fluid Mech. 88 (1978) 431–449. [33] H.A. Becker, H.C. Hottel, G.C. Williams, J. Fluid Mech. 30 (1967) 285–303. [34] K.M. Lyons, K.A. Watson, J. Energy Res. Technol. 123

(2001) 221–227. [35] R.W. Dibble, R.E. Hollenbach, Proc. Combust. Inst. 18 (1981) 1489–1499. [36] G.R. Ruetsch, L. Vervisch, A. Liñán, Phys. Fluids 7 (1995) 1447–1454. [37] A. Upatnieks, J.F. Driscoll, C.C. Rasmussen, S.L. Cec- cio, Combust. Flame 138 (2004) 259–272. [38] D. Veynante, L. Verisch, T. Poinsot, A. Liñán, G. Ruetsch, in: Proceedings of the Summer Program, Stanford/Ames Center of Turbulence Research, 1994. [39] G.T. Kalghatgi, Combust. Sci. Technol. 41 (1984) 17–

29.

[40] C.J. Chen, W. Rodi, Vertical Turbulent Buoyant Jet—

A Review of Experimental Data, Pergamon, Oxford,

1980.

[41] W.M. Pitts, Proc. ASME 1 (1987) 123.

[42] W.M. Pitts, Combust. Flame 76 (1989) 197–212.