Low-frequency combustion dynamics of a prefilming airblast injector were experimentally investigated in a laboratory-scale aero-engine gas turbine combustor operated with Jet A-1 fuel and air under idle/sub-idle conditions. Our measurements reveal that multiple modes – ranging from 45 to 292 Hz – can be excited in the non-premixed combustion system, including the Helmholtz, longitudinal, and hydrodynamic instabilities. The system’s propensity for mode selection depends strongly on combustor length and pilot equivalence ratio. A strong Helmholtz mode with a peak-to-peak amplitude of ~ 14 kPa occurs, provided that the combustor length is relatively short and the pilot equivalence ratio is high. A high-amplitude, intermittent burst at approximately 12 Hz always develops when the system is on the verge of transition to the Helmholtz instability. With a longer combustor length, on the other hand, the system transitions to the ¾-wave longitudinal mode via a discontinuous mode-hopping process. The system undergoes well-defined, limit cycle oscillations with an extremely large pressure amplitude of ~ 30 kPa, but the global $OH^*$ fluctuation amplitude is limited to merely 8.3%. Phase-resolved flame imaging measurements demonstrate that a ring-like partially-premixed flame emerges periodically in the dump region, when the non-premixed flame wrapping around the central recirculation zone is lifted off the fuel nozzle. The mutual interaction between the partially-premixed and non-premixed reaction zones is the dominant mechanism for intense sound generation. Under certain conditions, superposition of the Helmholtz and longitudinal modes occurs, leading to nonlinear interactions manifested by additional spectral peaks at the sum and difference of the two frequencies. In contrast to the other two cases, this superimposed tone is not characterized by limit cycle behavior, but by a noise-driven unsteadiness.
Nonlinear mode transition processes are presented using several analysis methods, including Fourier/Hilbert transforms, phase portraits, spectrograms, low-order analytic modeling, and high-speed flame visualization methods. Our results demonstrate that a non-premixed Jet A-1 spray flame yields an intermediate-amplitude, quasi-periodic L1 mode oscillation at 50 Hz for a low pilot equivalence ratio, and the frequency gradually increases with increasing fuel flowrates. Subsequent to a critical point (103 Hz), the system undergoes a discontinuous mode transition, giving rise to the formation of an extremely large pressure oscillation with an L2 mode structure at 244 Hz. Several key triggers were found to induce the mode shift: (i) generation of a high-intensity pulse in the flame’s heat release rate due to the spontaneous ignition of unburned reactant mixtures, (ii) emergence of a large-scale vortical structure and its interaction with a partially premixed flame front, and (iii) development of self-sustained limit cycle oscillations driven by periodic convection of a hot spot –a mechanism known as entropy wave propagation. Our study identifies the occurrence of a large-amplitude peak followed by a local minimum intensity, analogous to the activation energy concept, as an essential step for entry into a new state with large-amplitude limit cycle oscillations.
The influences of different pilot air swirler configurations were investigated since low-frequency growl instabilities have shown highly sensitive responses to variation of the pilot equivalence ratio. Two sets of pilot swirling geometries were considered in the experiments: the counter-rotating (CTR) and co-rotating (COR) pilot air swirl flows. Experimental results demonstrate that both cases can be excited to the strong instabilities with ~10% pressure oscillations and the instability frequencies were in a similar range from 40 to 250 Hz. However, it is important to note that the overall behavior on the Rayleigh criterion was significantly different between the two cases. The CTR case showed that a large number of data points with a peak-to-peak pressure amplitude of ~7 kPa violate the Rayleigh criterion. By contrast, most of the data points met the Rayleigh criterion in the COR case. The violation behavior in the CTR case was associated with the unsteady heat release rate with relatively high standard deviations during 10 cycles compared to the COR case. High-speed flame visualization and image post-processing results revealed that the violation behavior in the case of CTR was associated with the three major flame structures, i.e., a flame base, a spiral tail like diffusion flame, and convection of a hot spot in the downstream. From the Rayleigh index, in particular, high destructive interference was observed in the regions of a tail like diffusion flame and a hot spot in the downstream. Thus, it is apparent that flame dynamics in these regions induced the unsteady heat release oscillations responsible for the violation behavior. In the COR case, however, the unsteady flame structures in the downstream responsible for the irregular heat release rate waveforms were manifested by very weak luminous intensities in the case of COR. A periodic flame base movement characterized by high constructive interference was observed, leading to small standard deviations in the heat release rate oscillations. In both cases, constructive interference was observed in the upstream region of the flame base. This indicates that the dominant pressure oscillations are interrelated with the periodic movement of the upstream area of the flame base. In addition, when compared to the CTR case, the COR case showed more sensitive responses to the intermittent bursts during the transition to the Helmholtz instability.