CONTENTS 171 this case, the response of ameS3-as remains high. This ame also exhibits a large vertical oscillation of its leading edge and a peak value of the FTF gain at f = 96Hz. This analysis reveals that the mechanisms controlling the frequency response of swirled ames largely dier when they are stabilized by a blu body or when they are stabilized aerodynamically, the main dierences being related to the dynamics of the IRZ.
Conclusion
The dynamics of lean premixed swirling ames was investigated in this manuscript.
The main concern was to assess the eects of dierent injection conditions on the properties of ame stabilization, thermoacoustic instability and ame re-sponse to acoustic perturbations. In particular we considered eects of dierent (i) design of the swirler device, (ii) dimensions of the injector, (iii) bulk ow velocities and (iv) ame stabilization mechanisms.
Flame stabilization
A stabilization chart was obtained and three ame regimes were identied de-pending on the value of the swirl number and a normalized injector cross section area:
• at low level of swirl (S < 0.5), ames are elongated in the downstream direction with a shape similar to that of non-swirling V-ames
• in the intermediate swirl range (0.5< S <0.8) ames are well stabilized with a compact shape
• when the swirl number is too elevated (S > 0.8) ames ashback inside the injector.
These limits on the swirl number should only be taken as an indication and are also dependent upon the main dimensions of the injector. For a xed swirl number, an injector with an higher nozzle cross section has more propensity to ashback.
Thermoacoustic instability
One of the major drawback for the application of lean premixed combustion systems, is their sensitivity to thermoacoustic instabilities. In this work, we concentrated on the eects of dierent geometries of the injector on this phe-nomenon. In particular, we showed that it was possible to lower the intensity of the instability by properly modifying the distance between the swirler device and the combustion chamber backplane. On the other hand, when the dimen-sion of the injector are xed, a dierent design of the swirler did not lead to any important dierence on the properties of the thermoacoustic instability.
174 Conclusion
Flame response to acoustic perturbations
FTF measurements were conducted by submitting ames to well controlled acoustic perturbations. It was shown that the FTF gain curve of highly swirled ames is characterized by peaks and valleys while, when the swirl number is low (S = 0.20), the FTF behaves like a low-pass lter, with the gain increasing to reach a maximum higher than unity and then constantly decreasing towards zero for increasing forcing frequencies. Surprisingly, the same low-pass lter be-havior was observed for an high swirling ame fully aerodynamically stabilized, without the help of a blu-body. A low FTF gain value in the low frequency range could only be observed for highly swirled ames stabilized with the help of a blu-body. The FTF phase-lag is characterized by a linear behavior and by a jump with an inection point around the minimum gain frequency, if this minimum is present.
In terms of the eects of dierent injector geometries, it was observed that:
• as the swirl number is increased, either by an increase of the injector exit diameter, a decrease of the blu-body diameter or a dierent swirler device, the minimum gain value is decreased and the phase-lag jump increased
• by modifying the velocity inside the injector, either by a change of injector diameter or bulk ow velocity, it is possible to shift the FTF curves in the frequency range
• the same eect can be obtained by modifying the distance between the swirler and the combustion chamber backplane.
Mechanisms behind the acoustic response
Phase-locked (i) OH* chemiluminescence images, (ii) Particle Image Velocime-try (PIV) measurements and (iii) Large Eddy Simulations (LES) were exploited for an interpretation of the main mechanisms behind the acoustic response of lean premixed swirling ames. The focus was on the dierences between low and high value of the FTF gain, to assess those conditions possibly leading to a reduction of the sensitivity of LPM systems to acoustic waves.
Qualitatively, the motion of the investigated ames at these two peculiar fre-quencies is not dierent. The ame motion is the results of swirl number uctu-ations and interaction with vortical structures, leading to a modulation of ame length and angle. Swirl uctuations are due to the interaction of axial acous-tic and azimuthal convective perturbations, which are formed at the swirler exit when this is impinged by acoustic waves. Due to the convective nature of azimuthal perturbations, this interaction depends upon swirler-combustion chamber distance and mean ow velocity inside the injector. This explains the eects on the FTF of these parameters.
Conclusion 175 Quantitatively, uctuations are stronger at frequencies corresponding to a max-imum gain value. It was shown that this is due to a dierent strength of vortical structures at the two frequencies, which is not linked to ame/ow interaction since it was observed even in cold ow conditions. By post-processing of LES data, this was linked to a dierent phase-shift between the axial and the az-imuthal velocity perturbations, that lead to a dierent shape of the velocity proles at the injector rim. The low strength of vortical structures at the FTF gain minimum frequency is behind the low ame response and the drift from linearity of the FTF phase-lag.
By considering the time scales of these two dominant mechanisms (interferences between the acoustic pulsation and the azimuthal ow disturbances generated at the swirler outlet and interaction of the ame with vortical structures gener-ated at the burner rim), a nondimensionalization of the FTF results obtained with dierent injection conditions could be obtained.
The ame motion of a fully aerodynamically stabilized ame, analyzed at last, is somehow dierent. In this case, a vertical oscillation of the ame leading edge is observed as a consequence of the vertical oscillation of the internal recirculation region (IRZ) while, when the blu-body is present, this vertical oscillation is hindered and the IRZ mainly oscillates laterally.
Perspectives
• All the results presented in this manuscript refer to fully premixed system at a xed value of the equivalence ratio (Φ = 0.82). It would be interesting to see if the mechanisms observed can be conrmed for dierent values of Φ. Even more interesting would be to conduct a same analysis for technically premixed systems.
• FTF measurements are a powerful tool to analyze the response of the ame to incoming perturbations, but are time consuming. A step forward towards the identication of the main parameters controlling this response was taken in this manuscript, but work still need to be done. In particular, the dierent behavior of ames stabilized with or without the help of a blu-body, deserves further attention.
• The mechanisms leading to a dierent strength of vortical structures at the FTF gain minimum and maximum frequencies were identied through PIV and LES in cold ow conditions. It would be interesting to check whether the same conclusions could be drawn when performing the same analysis in reactive conditions.
• FTF measurements were performed only in the low and intermediate frequency range in this work. It would be interesting to extend these
176 Conclusion
data to the high frequency range and see if new minimum or maximum FTF gain values are present. In this case, it would be worth to conduct the same analysis at these other peculiar frequencies and check if the same mechanisms are in action.
• It would be interesting to compare LES with experimental data for the conguration without the blu-body, in cold but also reactive condi-tions. This analysis would show the dierent ow properties inside the injector, leading to the dierent responses of swirling ames stabilized with/without blu-body.
Appendix A
Comparison of naturally
unstable and externally excited ame motion
Flame transfer function are used in low-order models, in conjunction with de-scriptions of the burner acoustic eld, to predict stability maps of combustion systems. However, they do not give access to qualitative informations about the ame motion, in response to the acoustic excitation. To analyze the mecha-nisms behind the ame response, phase-locked images of the ame motion were exploited in this thesis.
The experimental setups, used to characterize combustion instabilities and to mimic them by external excitation, are sketched in Fig.A.1. In the rst case, the bottom of the burner was closed by a plate and the exhaust length could be modied by adding tubes of length T = 220 mm. Up to three tubes, 3T = 660 mm, were mounted during this study. Doing so, the thermoacoustic state of the burner was modied and instabilities could be triggered.
The uctuating velocity was measured with an hot wire probe and pictures of the ame motion were taken with an ICCD camera, synchronized by the HW signal.
In the second case, the exhaust tubes were removed and a loudspeaker was mounted at the bottom of the burner, to introduce the same acoustic excita-tion measured during self-triggered instability, controlled by the HW signal.
Pictures of the ame motion were then taken with the ICCD camera, synchro-nized by the HW signal.
Figure A.2 shows the ame motion obtained with these two dierent setups.
The phase angles indicated in the gure, refers to the hot wire signal. The ge-ometrical conguration analyzed was presented in Ch.5. The ame movement shown at the left of the symmetry axis in Fig.A.2, obtained by introducing the acoustic excitation with the loudspeaker, correctly reproduces the ame motion observed at the right of the symmetry axis and measured during self-triggered
178
instabilities. This was checked for two conguration with dierent exhaust tube length.
This check was important to conrm that the analysis of ame motion con-ducted by externally exciting the ow with a loudspeaker, correctly reproduces the same mechanisms observed during self-triggered instabilities.
ICCD + OH* filter HW
Loudspeaker
Closed plate Exhaust tube
(2T or 3T) HW
ICCD + OH* filter
Figure A.1: Setup used for the imaging of naturally unstable (top) or externally excited (bottom) ames.
Appendix A - Comparison of naturally unstable and externally
excited flame motion 179
FigureA.2:PhaseaveragedOH*chemiluminescenceoftheameduringonecycleofoscillationfortheforcedame(leftofthesymmetry axis)andfortheself-triggeredame(rightofthesymmetryaxis).Top:2Texhaustlength,fI=181Hz.Bottom:3Texhaustlength, fI=164Hz.Geometricalconguration:SW3,Do=20mm,C=10mm,δ=50mm.FulldetailsweregiveninCh.5.