Study of premixed combustion instabilities using phase ...

14th Int Symp on Applications of Laser Techniques to Fluid Mechanics Lisbon, Portugal, 07-10 July, 2008

- 1 -

Study of premixed combustion instabilities using phase locked tomography PIV

Nathalie Guilbert, Arnaud Mura, Bastien Boust and Michel Champion

LCD, UPR9028 CNRS, ENSMA Poitiers, FRANCE, Author correspondence to: [email protected]

Abstract: The objective of this contribution is to understand the role of coherent structures in the development of combustion instabilities associated to lean premixed combustion. More precisely, the influence of the equivalence ratio (Φ=0.65, 0.70, 0.75 and 0.80) is studied for a fixed value of the Reynolds number (Re=25000) using the experimental device ORACLES composed of a sudden expansion of a 2D channel. This experiment is characterized by the existence of a natural instability whose frequency is about 50 Hz. The corresponding coherent structures are analyzed through a phased locked PIV method, allowing the application of the well known triple decomposition. Simultaneously, tomography is used to localize and analyze the flame front. It is observed that for leaner equivalence ratios i.e. Φ=0.65 and 0.70 no pulsation of the flame appears and the flame is located within the shear layer created by the step corresponding to the sudden expansion of the channel. Whereas, for higher values of the equivalence ratio (0.75 and 0.80), strong pulsations are present. In this latter case, a shift in the frequency is observed, namely 50 Hz for Φ=0.75, and at 53 Hz for Φ=0.80. The application of the triple decomposition to these cases shows that the coherent structure formation is the same for both equivalence ratios, with only a change in the period of the mechanism. The whole coherent structure is made of vortices formed at the step location and convected downstream, thus influencing the large scale topology of the flame. Vortices which rotate in the clockwise direction increase the fraction of fresh gases in the combustor. On the contrary anti-clockwise vortices, opposed to the mean direction of the flow, reduce it. The shift in frequency with equivalence ratio is due to the increase of local speed of sound with temperature, yielding to an acceleration of the vortex shedding process. The stochastic part of the velocity, of the same order of magnitude as the coherent part, wrinkles the flame front at small scales. A comparison of the flame topologies with and without pulsation is performed.

1. Introduction

In aeronautical combustion devices a common way to reduce pollutant emissions consists in

operating at low equivalence ratio. Unfortunately the corresponding conditions are very sensitive to

the development of combustion instabilities that can eventually lead to the destruction of the

combustion chamber. A key element in the development of such combustion instabilities is the

flame response to acoustic excitations. Heat release fluctuations can enter into resonance with the

acoustic perturbation so that the instability is self-sustained and may grow up to unacceptable

levels. The properties of such an instability and its development in turbulent reactive flows have to

be understood and the present study provides a new step towards this objective in specific ranges of

Reynolds number and equivalence ratio. Many experimental devices are available to perform such

investigations: some of them rely on a swirl flow, with conditions similar to those encountered in

aeronautical combustion chamber [1] but in such complex devices, it is sometimes quite difficult to

distinguish the contribution associated to the instability from turbulence. Another way is to

superimpose an acoustic forcing onto a very well defined flow configuration [2]. However,

considering the non natural nature of the excitation, conclusions are difficult to draw with respect to

practical configurations. In the present study, to reduce subsequent problems of analysis associated

either (i) to a special complexity of the flow or (ii) to the use of an artificial excitation, an academic

experimental device that displays a natural oscillation is used. This device, called ORACLES (One

Rig for Accurate Comparisons with Large Eddy Simulations), consists of two superimposed fully

developed turbulent channel flows fed with propane-air mixtures merging just upstream of a sudden

expansion. Former studies have highlighted that, for some inlet conditions, combustion gives rise to

the appearance of a natural periodic pulsation of the reaction zone whose amplitude is mainly

governed by the characteristics of the incoming reactants streams [3,4]. This allows to investigate

experimentally the influence of a large scale coherent motion superimposed on turbulent

combustion, thus leading eventually to a better understanding of the essential features and effects of


- 2 -

the instability. The test rig also permits to build experimental databases useful for turbulent

combustion modelling purposes as it has been already used as a pertaining test case of validation for

either Large Eddy Simulation (LES) [5-7] or Reynolds Average Navier-Stokes (RANS) models

[8,9]. In this contribution, a two dimensional PIV (Particle Image Velocimetry) system is used to

study the mechanisms underlying the dynamics of a large scale coherent motion and stochastic

small scale turbulent fluctuations and their influences on the whole turbulent combustion process.

Conditional averaging procedures based on PIV [10] but also on LIF or chemiluminescence signals

are widely used to perform the analysis of such organised flows in reactive or non reactive turbulent

flows. In the present test rig, the observed combustion instability, well identified at 50Hz, modifies

the flow field not only inside the combustion chamber but also inside the incoming channel flows.

Thus, PIV phase-locked measurements can be performed by using the signal gathered from a

microphone positioned at the wall of one of the inlet channels. In ORACLES, the amplitude of the

instability is not very strong for all conditions. For a fixed value of the Reynolds number Re~25000,

the role of equivalence ratio in the development of the instability can be studied. To this respect our

main objectives are (i) to extract the energetic contribution and the spatial location of the coherent

structure for different phases of the pulsation, (ii) to study the influence of each contribution on the

global flame shape.

The present paper is organized as follows: the first section provides a detailed presentation of the

experimental test rig and associated diagnostics. The following two sections are devoted to a

general analysis of the flow field in both non reactive and reactive situations in pulsating or non

pulsating conditions and the presentation of the results ends with a more detailed investigation of

the coherent motion. Finally, conclusions are drawn from the analysis and perspectives for future

experimental investigations are presented.

2. Experimental setup and diagnostics

2.1 Experimental facility and operating conditions

The combustor called ORACLES, see Fig. 1, is a 10-meter long test facility that consists of two

superimposed turbulent streams of propane–air mixtures merging just upstream of a square section

combustion chamber. Each inlet reactants stream is generated as follows: 1) the air flow passes

through an electric heater that keeps its temperature constant (20°C) over time, 2) the propane flow

is added in a mixing chamber to create a homogeneous propane–air mixture, 3) a converging

section is followed by a 3-meter long rectangular section that generates a fully-developed turbulent

channel flow, 4) combustion occurs in a thermally insulated 2-meter long section fitted with quartz

windows that allow optical access at the side and bottom walls, 5) in the exhaust section, burned

gases are cooled by injection of pressurized water and extracted outwards by a fan.

Fig. 1 The experimental facility ORACLES

In this configuration, the combustion chamber is located downstream of backward facing steps

made by the sudden expansion of the rectangular cross section, see Fig. 2a, whose expansion ratio is

equal to 1.8. During the ignition phase, each stream is spark-ignited just downstream of the

1 2 3 4 5 3


- 3 -

expansion and flames can be stabilized thanks to the recirculation of burned gases, as illustrated by

Fig. 2b.

a) b)

Fig. 2 ORACLES combustion chamber: a) scales in mm, b) visualization of the combustion zone

ORACLES can be operated within a wide range of experimental conditions. In each stream, the

mass flow rate Q can be varied in the range 0.05–0.2 kg/s, and the equivalence ratio Ф can be

varied in the range 0.6–1.0. For the present study, experiments were carried out with the same mass

flow rate (0.0528 kg/s) and varying equivalence ratio (0.65, 0.70, 0.75 and 0.80) in both streams. In

order to keep these conditions constant over time in each stream, the mass flow rate of air is

regulated and the equivalence ratio of the propane–air mixture is controlled by using an infrared gas

analyser (Cristal 300, COSMA). The values of inlet parameters such as pressure, temperature, mass

flow rate and equivalence ratio are controlled and recorded at a frequency of 1 Hz by an acquisition

device (SA70, AOIP).

2.2 Physical diagnostics

The goal of the present work is to study the combustion instabilities occurring in ORACLES by

investigating both (i) the velocity field and (ii) the associated flame front topology, in relation with

the acoustic behaviour of combustion. For this purpose, a coupled diagnostics is used. From the

point of view of acoustics, pressure variations are recorded using two microphones as well as a

piezoelectric transducer (Kistler 601A). The sensors are flush-mounted upstream of the combustion

section (see Fig. 3). The corresponding signals are recorded using a 2-channel oscilloscope

(TDS220, Tektronix). Simultaneously, the recording of both flame front and velocity field is

performed by using PIV in the vertical symmetry plane of the combustion zone (see Fig. 3). Each

stream is seeded with droplets of grape-pip oil, which allows to delineate the flame front location

visualized by the droplets vaporization surface. As a result, the flame front is detected by

tomography while the velocity field is simultaneously obtained in the fresh gases through a PIV

cross-correlation technique. To this purpose, we use an integrated PIV system (Flowmaster,

Lavision) including a 12-bit CCD camera (Flowmaster 3S) of 1280×1024 pixels, and a dual-cavity

Nd:YAG laser (Gemini, New Wave Research, 120 mJ) that generates a planar laser sheet at a

wavelength 532 nm with a thickness of approximately 500 µm. In order to reach an optimum

signal-to-noise ratio, an interferential filter selecting the desired wavelength (532 ± 5 nm) is

mounted on the lens (Nikon AF Nikkor, 50 mm, f/1.8). In addition, a liquid crystal shutter is used to

avoid an overexposure of the second frame resulting from integrating the flame luminosity. The

system is operated by using the software Davis 6, and can be a priori run at a frequency up to 8 Hz.

However, in the situation that we investigate, the acquisition timing is synchronized with the

acoustic signal. As already stressed in the introduction, the ORACLES setup displays a regime of

pulsating combustion within a specific range of conditions. In order to capture the desired data at a

given phase with respect to the position of the flame, the pseudo-periodic signal delivered by the

microphones is amplified, filtered and analyzed by an electronic device. When this signal exceeds a

given threshold value, then the PIV system is triggered (see Fig. 4). A digital delay generator (DG

535, Stanford Research Systems) is used to avoid an overexposure of the camera chip due to

x

y

A

A

A – A


- 4 -

successive triggers. The delay between the acoustic signal and the trigger of PIV acquisition,

denoted as ∆t, is tuned to vary the phase at which the flame is captured. The time between the first

and second PIV frames is adjusted so that the maximum displacement of particles does not exceed

25% of the cross-correlation window.

Fig. 3 Diagnostics performed on ORACLES

Pressure

Microphone

Microphone

+ Filter

Delay

generator

Trigger for

PIV system

∆t ∆t

Camera exposure

Fig. 4 Synchronization of the PIV system with the acoustic signal

This so-called “phase-locked” PIV technique allows the recording of successive velocity fields and

flame fronts having the same phase with respect to the flame pulsation. The resulting sequence of

PIV data can be used to carry out a statistical study of both the velocity field and the flame front, for

a given value of the phase i.e. similar shapes of the flame. For each operating point of ORACLES,

several sequences of PIV fields are recorded, in order to study the flame at different phases of its

pulsation. For each sequence, more than 600 PIV fields are recorded. The velocity vector

calculation is performed thanks to the Davis 6.2 software. To avoid vector calculation in burned

gases where no seeding droplets are present, an image mask is created for each couple of images.

The mask is obtained through an automatic contour extraction procedure based on a grey-level

histogram threshold using an index of fuzziness. Such a method is often used to obtain the flame

contour in order to characterize the local flame structure and associated flame surface density [11].

To calculate vectors, an adaptative cross-correlation algorithm with a decreasing size of the

Pressure

transducer

Micro-

phones

Laser

Laser

sheet

Mirror

PIV +

Tomography


- 5 -

interrogation window from 64x64 to 16x16 and a 50% overlap is used. More than 98% of the

vectors are validated for each step and the obtained spatial resolution is about 470 µm [12].

Moreover, the field of view was optimized to eliminate peak locking effects. Because of the

reflexions on the step and the wall chamber, the overview is limited to -0.55<x<10.5 cm and -

5.75<y<2.64 cm. Notice that the frame of reference (x,y) is defined such as the splitter plate is

located at (0,0) and the wall at -6.53 cm.

3. Global observations

3.1 Non reactive flow field

Before analyzing the combustion process, the non reactive flow field is analyzed briefly. In Fig. 5

the mean velocity field downstream of the lower part of the sudden expansion is displayed.

Downstream of the splitter plate, a mixing layer develops but does not modify the overall structure

of the flow field created by the backward facing step. At the external boundary of the wake

produced by the step, a shear layer is also developing. Moreover, a recirculation zone can be

observed at the bottom of the step, even if the optical access becomes restricted in the vicinity of the

wall. Previous investigations performed with Laser Doppler Anemometry [3] have shown that

another recirculation zone appears in the upper part of the test section but the dimensions of the two

recirculating zones are quite different, giving rise to an asymmetrical flow field. The influence of

the area expansion ratio on this type of asymmetry has been largely studied in such a kind of

backward facing step configuration [13]. However, it must be noticed that in the presence of

combustion, a symmetric flow field is recovered.

Fig. 5 Non reactive flow: average velocity field

In the next subsection, the reactive flow field is described for different values of the equivalence

ratio of incoming premixed reactants.

3.1 Reactive flow field: characteristic frequency of pulsating modes

The reactive flow is studied for four distinct equivalence ratios Φ (0.65, 0.7, 0.75, 0.8), for a fixed

value of the Reynolds number Re~25000 with a mass flow rate of 52.8 g/s. A preliminary direct

observation realized with a classical CCD camera and a short exposure time (1/5000s) shows that a

strong oscillation appears only for Φ=0.75 and 0.8. This observation is confirmed by pressure

measurements carried out with a precision of 0.4 Hz. In Fig. 6 the corresponding spectral density

energy are reported for the four values of the equivalence ratio. We can notice that oscillations are

well defined for Φ=0.75 and Φ=0.8 with a very high spectral energy density level whereas this level

becomes negligible for Φ=0.65 and Φ=0.7. Moreover, a shift in the characteristic frequency is

observed when the equivalence ratio is varied, going from about 50 Hz for Φ=0.75 to 53 Hz for

Φ=0.8. This shift can be related to the increase of the speed of sound with temperature. Indeed, the

energy released by the flame modifies the characteristic pulsation: the temperature of burned gases


- 6 -

increases with equivalence ratio and, as a consequence, the acoustic oscillation frequency increases

as a result of the variation of the local speed of sound. Similar conclusions have been drawn from

the recent experimental work of DiTaranto et al. [14] for a lower Reynolds number flow (Re=3500).

In the experimental setup studied here, frequencies are close to those associated to the acoustic

modes of the combustion chamber (f0=47.5 Hz). From these preliminary investigations, we realize

five snapshots along the cycle of the pulsation for Φ=0.75 and Φ=0.80 using the phase-locked PIV

procedure described above. In former studies on ORACLES configuration, [3,4] it was observed

that, whatever the conditions, the reaction zone consists of two symmetrical flame-brushes anchored

in the shear layers that develop at the step corners. To reinforce this conclusion, we notice that even

in the case of pulsation, perfectly symmetrical lower and upper flame regimes are observed.

Accordingly we analyze only the lower flame as the most convenient to investigate experimentally.

Fig. 6 Energy density spectra for Q=52.8 g/s and four values of Φ

3.2 Reactive flow field: flame characteristics in non pulsating modes

The spectral analysis reported in Fig. 6 shows that, for leaner cases, the flapping of the flame

becomes negligible. No flame oscillation is observed and PIV measurements can be performed

without any phase-locked procedure. In Fig. 7, the mean velocity fields for Φ=0.65 and 0.70 are

displayed: the flame stabilizes, thanks to the hot gases recirculation zone, in the external shear layer.

The magnitude of the flame flapping is found to be too small to be observed, with no significant

effect on the flame shape. Velocity levels are of the same order of magnitude as in the non reactive

case except further downstream where the flow is significantly accelerated by thermal expansion.

Moreover, no significant differences are observed between the two cases associated to Φ=0.65 and

Φ=0.70 respectively.

(a) (b)

Fig. 7 Average velocity field Φ=0.65 (a) and Φ=0.70 (b)


- 7 -

3.3 Reactive flow field: flame characteristics in pulsating modes

In presence of oscillations, a conditional averaging must be performed to ensure a realistic

statistical analysis. Using the experimental diagnostic setup described above, we verified that no

jitter affects the flame location. With tomographic PIV measurements, the probability of presence

of unburned gases can be obtained, see Fig. 8 for four different phases of the flapping flame period

at Φ=0.75. This quantity is nothing else but the conditional average progress variable, or

equivalently, it provides the shape of the conditional mean flame brush for a given value of the

phase. The limited thickness of the resulting flame brush confirms the quality of the phase-locked

measurement: for a given value of the phase, the flame stays within the same region from one

snapshot to another. The first observation that can be made from Fig. 8 concerns the drastic change

of the flame location over the pulsating cycle. A large scale motion influences the flame location in

a longitudinal direction near the recirculation zone. The flame is first located near the backward

facing step (0°), then pushed downstream by unburned reactants (180°). A large pocket of fresh

reactants is created upstream and then convected further downstream, thus modifying deeply the

flame front location. This observation clearly confirms the presence of a large scale coherent

motion that strongly affects the global shape of the flame. The same conclusions hold for Φ=0.80.

This change in the value of the equivalence ratio only accelerates the phenomenon (f=53 Hz) but

does not change the phenomenology. A similar evolution has been observed numerically through

the Large Eddy Simulations of the ORACLES experiment performed by Duwig and Fureby [7].

(a) (b)

(c) (d) Fig. 8 Probability of presence of unburned reactants

for several phases Φ=0.75, ϕ=0° (a), 90° (b) ,180° (c), 270° (d)

The influence of this large scale coherent structure onto the local flame topology can be

subsequently analyzed by considering the flame surface that separates fully burned products from

fresh reactants. The flame contours, as obtained from the method described in section 2, can be used

to discriminate this influence. Such representative instantaneous flames contours are reported in

Fig. 9a for Φ=0.65 and Φ=0.70 and for different values of the phase (with a fixed equivalence ratio

Φ=0.75) in Fig. 9b.


- 8 -

(a) (b)

Fig. 9: Instantaneous flame contour at Φ=0.65 and Φ=0.70 (a), and Φ=0.75 for several phases (b)

The typical length scales associated to the wrinkled flame front remain of the same order of

magnitude whatever the considered experimental conditions i.e. with (b) or without (a) pulsation.

The flame locations for cases without pulsation are approximately the same (Fig. 9a) whereas, as

previously observed, the flame location is strongly modified by the superimposed coherent motion

for pulsating conditions (Fig. 9b). The presence of this large scale motion seems to only affect the

global flame shape at the largest scale and not the flame wrinkling levels which seem to remain

controlled by smaller scale stochastic motion associated to turbulence. Further analyses are now

under way to confirm these preliminary conclusions through a systematic and quantitative analysis

of the joint curvature and strain rate statistics.

The large scale coherent structure which has been highlighted for pulsating combustion modes

clearly drive the global flame characteristics and in particular its location. The analysis of this large

scale contribution is refined below by considering the triple decomposition formalism [15]. This

will allow to identify more precisely the role of the coherent structure versus the one of turbulence.

4. Coherent contribution to flame oscillations

4.1 Basic principles of the triple decomposition

The conditional averaging method is one current approach to study flows containing deterministic

or periodic structures [10,15]. In turbulent flows, conditional averaging allows the measured signal

to be split into distinct contributions: the mean, the coherent and the stochastic components. To

obtain the corresponding contributions, an approach based on the so-called triple decomposition is

used and the measured turbulent signal )(tV is split into its time-averaged value V and its

coherent and stochastic components ( )ϕtVC and ( )tV ' respectively [15]:

( ) ( ) ( )tVtVVtV C

'++= ϕ

The subscript ϕ denotes the coherent component phase corresponding to time t. Measuring the

conditionally averaged value of the quantity ( )ϕtVC actually implies averaging over an ensemble

of deterministic flame location phase. The mean component V is measured thanks to a random

average velocity of flame oscillation and the coherent component of pulsations is then obtained as

follows:

( ) ( )∑=

∞→−=

K

k

k

CK

C VtVK

tV1

1lim ϕ

Here, ( )k

C tV ϕ is the kth

measurement for a given value of the phaseϕ .


- 9 -

To perform conditional averaging over an ensemble of coherent structures, the phase of the

coherent structure must be unambiguously determined at the moment of measurement. This

splitting can be performed as long as the coherent and stochastic pulsations are assumed to be

independent.

(a) (b)

(c) (d)

Fig. 10 Example of triple decomposition for Φ=0.75 at ϕ=90°, (a) instantaneous, (b) mean, (c) coherent and

(d) fluctuating velocity fields

4.2 Triple decomposition: analysis of the coherent contribution

An example of application of the triple decomposition formalism is reported in Fig. 10 for Φ=0.75

and ϕ=90°. The instantaneous (a), the time average (b), the coherent (c) and stochastic velocity

fields (d) are displayed. The most important contribution to the instantaneous velocity field is

associated to the mean velocity component with a level in between 5 and 11 m/s. The stochastic and

coherent contributions are less important but they are approximately of the same order of magnitude

between 1 and 3 m/s. Concerning their spatial organisation, the three components exhibit significant

differences. As expected the stochastic velocity is found to be randomly distributed in fresh gases

and, as expected, the associated fluctuation length scales does not appear to be fully resolved in the

present experimental conditions. The mean velocity contribution displays a privileged direction and

it is very similar to the mean velocity field obtained in the absence of any oscillation i.e. for Φ=0.65

and 0.7, as shown by Fig. 7. Concerning the coherent part, it seems to be rather well organised with

two identified structures localised between the mixing and shear layers. The first coherent structure

comes from the inlet channel with a large velocity (more than 3 m/s) and the second one is

convected downstream with a decreased velocity (between 1 and 2 m/s). These are counter-rotating

structures. This finding confirms our previous analysis on the influence of large structures on the

flame shape.

The coherent structure clearly plays an important role on the global large scale shape of the flame.

The first structure, observed at the left of the figure 10c, tends to push the fresh mixture away in the

vicinity of the bottom wall to create a pocket of fresh reactants that the flame cannot burn. However

the characteristic length scale of these structures (larger than 3 cm) is too large to contribute

significantly to flame wrinkling and it is more realistic to conclude that the stochastic velocity field


- 10 -

is responsible for the observed flame wrinkling. This confirms our conclusions concerning the

flame wrinkling levels in the presence of a pulsating mode (or not). Coherent and stochastic

components seem to be rather independent as required when applying the principles of the triple

decomposition. The same conclusion holds when Φ=0.8. Therefore, even if the equivalence ratio

plays a role in the acceleration of the phenomenon, the essential mechanisms of the formation of the

coherent structure and its contribution to the instantaneous velocity field do not seem to be

significantly modified when Φ is changed.

4.3 Temporal evolution of the coherent structure

In Fig. 10 we have shown the presence of two well identified structures contributing to the coherent

part of the velocity field. Then it is of interest to take a closer look at their evolution during the

cycle. Accordingly to Fig. 11 we present the coherent structure velocity fields ( )ϕtVC for Φ=0.75

at different values of the phase along the pulsating cycle, namely ϕ=0°, 90°, 180° and 270°. It may

be stated that the coherent structure is formed by large scale vortices which are created in the inlet

channel and convected further downstream during the flapping period. For each value of the phase

two neighbouring structures can be distinguished. The first is counter-clockwise rotating and it is

opposed to the mean flow direction whereas the second one is clockwise rotating in the same

direction as the main flow. In all cases, vortices in the clockwise direction increase the fraction of

fresh gases present in the combustor volume whereas vortices in the counter-clockwise direction

reduce it. Simultaneously both types of vortices are convected. These large pockets of fresh gases,

where stand the two vortices, either slow down or accelerate the combustion rate thus modifying the

flame response to the perturbation. As a consequence, a modification of the location of the reaction

zone under the influence of the coherent structure appears. This mechanism of formation and

transport of the coherent structure is associated with the large variations of the flame shape with

time over the pulsation period.

(a) (b)

(c) (d)

Fig. 11 Temporal evolution of the coherent structure, Φ=0.75, ϕ=0°, 90°, 180°, 270°

In the case Φ=0.80, not reported here, the organisation of large scale vortices associated to the


- 11 -

coherent motion is similar. The corresponding flow pattern is obtained in the same manner but is

characterized by a larger frequency.

4.4 Energy distribution analysis

As previously mentioned the main advantage of the conditional averaging method is the possibility

that it provides to discriminate the fluctuations, between a coherent contribution and its stochastic

counterpart. As an example Fig. 12 shows the distribution of the kinetic energy, for each velocity

components, namely the mean component as well as coherent and stochastic fluctuating

components. This splitting allows also the evaluation of the specific contribution due to coherent

structures and other turbulent scales to the momentum transfer processes as well as their influence

on the mixing processes in the flow. Figure 12 shows that kinetic energies of coherent and

stochastic motions are of the same order of magnitude. On the other hand the energy of the mean

component is an order of magnitude higher than the coherent and stochastic ones. The coherent

energy is produced between the shear and mixing layer and hence influences the global flame

shape. The highest level of energy associated to the stochastic motion is occurring near the flame

front, which confirms its influence on the flame wrinkling process. It is worth noticing here that, to

evaluate the corresponding energy levels, we used only the two measured components of velocity

and we do not attempt to estimate the third component, thus avoiding the use of any arbitrary

assumption. This result confirms the relevance of the distinction between the coherent and the

stochastic contributions to describe the turbulent flame development.

(a)

(b) (c)

Fig. 12 Energetic contribution of mean (a), coherent (b) and stochastic field (c) for Φ=0.75 and at ϕ=180°

5. Conclusions and perspectives

In the present contribution an experimental study of a turbulent premixed flame stabilized

downstream of a double sudden expansion is reported for a fixed Reynolds number Re~25000 and

several equivalence ratios Φ (0.65, 0.70, 0.75 and 0.80). A low frequency instability has been

evidenced for Φ=0.75 and Φ=0.80 with a characteristic frequency of 50 Hz at Φ=0.75 and 53 Hz at

Φ=0.80, whereas for leaner mixtures no significant instability has been observed. The

corresponding oscillation is associated to the existence of a well identified coherent structure that

deeply modifies the flame shape. To highlight its influence, PIV phase-locked measurements were


- 12 -

performed. The subsequent use of the triple decomposition on the obtained measurements confirms

that the pulsation is created by a large coherent motion generated upstream and convected

downstream. This coherent structure influences only the global shape of the flame while the flame

wrinkling remains essentially driven by the stochastic motion. The levels associated to the energetic

contribution of the coherent structure are found to be of the same order of magnitude as those

associated with the stochastic motion. Moreover, for both values of the equivalence ratios Φ=0.75

and Φ=0.80, the mechanisms of emergence of the coherent structure is the same. The only

difference lies in the value of the frequency of generation of these vortices since temperature plays a

role in accelerating the phenomenon but it has no effect on the mechanism of formation of the

vortices itself. To improve the understanding of this phenomenon will require further measurements

of the velocity field in the burned gases coupled with LIF measurements to extract the flame front

location.

References

[1] Paschereit C.O., Gutmark E., Weisentein W. (2000), Excitation of thermoacoustic instabilities by

interaction of acoustics and unstable swirling flows, AIAA Journal 38(6):1025-1034

[2] Demare D., Baillot F. (2004), Acoustic enhancement of combustion in lifted non premixed jet flames,

Combust. Flame 139(4):312-328

[3] Besson M., Bruel P., Champion J.L., Deshaies B., (2000), Experimental analysis of combusting flows

developing over a plane-symmetric expansion, J. Thermophys. Heat Transf. 14(1):59-67

[4] Nguyen D., Bruel P. (2003), Turbulent reacting flow in a dump combustor, AIAA Paper2003-0358

[5] Duchamp de Lageneste L. (2000), PhD Thesis, Ecole Centrale de Lyon

[6] Domingo P., Vervisch L., Payet S., Hauguel R. (2005), DNS of a premixed turbulent V-flame and LES of

a ducted-flame using a FSD-PDF subgrid scale closure with FPI tabulated chemistry, Combust. Flame

143(4):566-586.

[7] Duwig C., Fureby.C (2007), Large Eddy Simulations of unsteady lean stratified premixed combustion,

Combust. Flame 151:85-103

[8] Robin V., Mura. A, Champion M, Plion. P. (2006), A multi Dirac presumed PDF model for turbulent

reactive flows with variable equivalence ratio, Combust. Sci. Technol. 118(10-11):1843-1870

[9] Kurenkov A., Oberlack M. (2005), Modelling turbulent premixed combustion using the level set

approach for Reynolds averaged models, Flow Turb. Comb. 74:387-407

[10] Alekseenko S.V., Bilsky A.V., Markovich D.M. (2004), Application of the method of Particle Omage

Velocimetry for analyzing turbulent flows with a periodic component, Inst. Exp. Tech. 47(5):703-710

[11] Renou B., Mura A., Samson E., Boukhalfa M. (2002), Characterization of the local flame structure and

the flame surface density for freely-propagating premixed turbulent flames at various Lewis numbers,

Combust. Sci. Technol. 174(4):143-179

[12] Foucault J.M., Carlier J., Stanislas M. (2004), PIV optimisation for the study of turbulent flow using

spectral analysis, Measurement Sci. Tech. 15:1046-1058

[13] Abbot D.E., Kline S.J. (1962), Experimental investigation of subsonic turbulent flow over single or

double backward facing steps, J. Basic Eng. 84D:317-325

[14] DiTaranto M., Jorgen H. (2006), Combustion instabilities in sudden expansion oxy-fuel flames,

Combust. Flame 146:493-512.

[15] Hussain A.K.F.M, Reynolds W.C (1972), The mechanics of an organized wave in turbulent shear flow,

Part 2. Experimental results, J. Fluid. Mech. 54:241-261

Acknowledgments

The authors thank the Région Poitou-Charentes for its financial support within the PPRIMME

project. They are also indebted to D. Falaise, C. Losier and V. Bertin for their technical assistance.

Study of premixed combustion instabilities using phase ...

Documents