The presented work focuses on the experimental spray structure and the combustion of a linearly-arranged 5-swirler array. The aerodynamics and spray characteristics of a non-reacting single swirler are reported first as a baseline, followed by those of a 5-swirler array to investigate the effect of swirling flow interactions on aerodynamics and combustion. For the baseline single swirler, the smaller droplets follow the air flow more closely and further dispersed away at the exit of swirler. Thus, the mean diameter of droplet increases with the flow developing further downstream. However, in the central portion of a 5-swirler array, the droplet size remains similar. It is attributed to that swirling flow interactions might provide better air/fuel mixing and the additional shear stress to break up droplet continuously and is evident by the higher turbulent intensity in the aerodynamic measurement. Due to the influence of gas phase, the distribution of liquid phase in center toroidal recirculation zone (CTRZ) is non-uniform in a 5-swirler array. The center swirler of a 5-swirler array features a larger CTRZ which is accompanied by two smaller CTRZs from its neighbors. The flame anchored by the center swirler of a 5-swirler array is richer than the other two neighboring flames when the inter-swirler spacing is 2D, where D is the diameter of swirler exit diameter. However, when the inter-swirler spacing is increased to 2.5D, all swirlers feature a similar flame, which is different from what is expected from non-reacting flow studies reported previously. The unexpected result should be attributed to the difference in swirling strength between non-reacting and reacting flows. Moreover, the high speed imaging is employed to investigate the flame spreading during ignition process for a 5-swirler array. The high-speed movies show that the directional mechanism of flame spreading along lateral direction remains basically the same and is independent of the investigated test parameters including: two inter-swirler spacings, five fuel flow rates, five air pressure drops across swirlers, and five upstream air temperatures. An empirical correlation incorporating normalized inter-swirler spacing, air/fuel ratio, Reynolds number, and normalized air temperature is proposed and validated through a normalization procedure within around ± 10% error. The increase of Reynolds number and normalized air temperature has favorable impact on the flame spreading, which is stated by the empirical correlation.

An experimental study was conducted to perform an analysis of the effect of the geometric modifications of the venturi on the non-reactive and reactive flow behavior using a counter-rotating radial-radial swirler. In the non-reactive flow tests, measurements were taken in a central vertical plane and horizontal (cross-sectional) plane at the exit of the swirler, using a High-Speed, Two Dimensional, Particle Image Velocimetry (2D PIV) system. The size of the swirler used in the non-reactive flow tests is a 4.76X scaled size of the swirler used in combustion. The 4.76X swirler models were tested in air flow seeded with olive oil at Re = 51,500, corresponding to the pressure drop across the 1X swirler models of 4% of atmospheric pressure at ambient conditions. Compared with the 1X swirler models, the 4.76X swirler models provide high spatial and temporal resolutions from the enhanced visibility of the flow characteristics and lower velocities at the same Re. Four swirler configurations of high swirl number (SN ≈ 1.0) were used, with no modification for the baseline configuration (configuration 1), and with the chevrons on the venturi for the straight chevrons configuration (configuration 2). The design of the inclined venturi was used for the converging venturi configuration (configuration 3), and chevrons were added on the converging venturi for the converging chevrons configuration (configuration 4). In the combustion tests, the 1X swirler models were tested using 478K preheated air at 4% pressure drop across the swirler, and different chamber lengths. Measurements were conducted using a regular camera to capture the flame image, and dynamic pressure transducers to obtain the acoustic pressure oscillations. Four configurations were studied and compared in the non-reactive and reactive flows with the objective of understanding the mechanisms responsible in reducing the extent of the combustion instabilities. Results of this study show that the converging venturi in configuration 3 appears to be the best design in eliminating the combustion instabilities in the fuel-lean region as compared to the other configurations. This indicates that the prevention of the frequencies coupling between the heat release rate and acoustic oscillations has been achieved by using the design of the converging venturi.

No abstract

Self-excited combustion instability in an annular combustor with low-swirl flames is studied with the combination of LES and acoustic solvers. Acoustic analysis with a Helmholtz solver provides an estimate of frequencies and modal structure in the annular combustor. LES gives detailed modal dynamics at specific instability modes. Combustion instabilities in the annular combustor including longitudinal, spinning and standing modes are successfully captured in single LES. Numerical results show the instability modes are not constant, they switch among these modes randomly and rapidly. The flow oscillates back and forth in phase with the largest pressure amplitude locating near the outlet of the injectors for the longitudinal mode. The azimuthal instability oscillates at the 1A2L mode of the annular system. In the spinning mode, the pressure anti-nodes move forwards while the modal structure keeps constant. For the standing mode, the locations of pressure anti-nodes are fixed in the annular combustor and the fluctuations at the pressure anti-nodes keep out of phase. The near-zero value of mean spin ratio indicates the dominant azimuthal mode is the standing mode. The azimuthal modes captured by LES are in good agreements with that predicted by Helmholtz solver in terms of frequency and modal structure. The maximum deviation of the predicted frequency is less than 5%. This adds values before putting the low-swirl injector into the actual annular combustor.