An Experimental Investigation of Self-Excited Combustion Dynamics in a Single Element Lean Direct Injection (LDI) Combustor
The management of combustion dynamics in gas turbine combustors has become more challenging as strict NOx/CO emission standards have led to engine operation in a narrow, lean regime. While premixed or partially premixed combustor configurations such as the Lean Premixed Pre-vaporized (LPP), Rich Quench Lean burn (RQL), and Lean Direct Injection (LDI) have shown a potential for reduced NOx emissions, they promote a coupling between acoustics, hydrodynamics and combustion that can lead to combustion instabilities. These couplings can be quite complex, and their detailed understanding is a pre-requisite to any engine development program and for the development of predictive capability for combustion instabilities through high-fidelity models.^ The overarching goal of this project is to assess the capability of high-fidelity simulation to predict combustion dynamics in low-emissions gas turbine combustors. A prototypical lean-direct-inject combustor was designed in a modular configuration so that a suitable geometry could be found by test. The combustor comprised a variable length air plenum and combustion chamber, air swirler, and fuel nozzle located inside a subsonic venturi. The venturi cross section and the fuel nozzle were consistent with previous studies.^ Test pressure was 1 MPa and variables included geometry and acoustic resonance, inlet temperatures, equivalence ratio, and type of liquid fuel. High-frequency pressure measurements in a well-instrumented metal chamber yielded frequencies and mode shapes as a function of inlet air temperature, equivalence ratio, fuel nozzle placement, and combustor acoustic resonances. The parametric survey was a significant effort, with over 105 tests on eight geometric configurations. A good dataset was obtained that could be used for both operating-point-dependent quantitative comparisons, and testing the ability of the simulation to predict more global trends.^ Results showed a very strong dependence of instability amplitude on the geometric configuration of the combustor, i.e., its acoustic resonance characteristics, with measured pressure fluctuation amplitudes ranged from 5 kPa (0.5% of mean pressure) to 200 kPa (~20% of mean pressure) depending on combustor geometry. The stability behavior also showed a consistent and pronounced dependence on equivalence ratio and inlet air temperature. Instability amplitude increased with higher equivalence ratio and with lower inlet air temperature. A pronounced effect of fuel nozzle location on the combustion dynamics was also observed. Combustion instabilities with the fuel nozzle at the throat of the venturi throat were stronger than in the configuration with fuel nozzle 2.6 mm upstream of the nozzle.^ A second set of dynamics data was based on high-response-rate laser-based combustion diagnostics using an optically accessible combustor section. High-frequency measurements of OH*-chemiluminescence and OH-PLIF and velocity fields using PIV were obtained at a relatively stable, low equivalence ratio case and a less stable case at higher equivalence ratio. PIV measurements were performed at 5 kHz for non-reacting flow but glare from the cylindrical quartz chamber limited the field of view to a small region in the combustor.^ Quantitative and qualitative comparisons were made for five different combinations of geometry and operating condition that yielded discriminating stability behavior in the experiment with simulations that were carried out concurrently. Comparisons were made on the basis of trends and pressure mode data as well as with OH-PLIF measurements for the baseline geometry at equivalence ratios of 0.44 and 0.6. Overall, the ability of the simulation to match experimental data and trends was encouraging. Dynamic Mode Decomposition (DMD) analysis was performed on two sets of computations - a global 2-step chemistry mechanism and an 18-step chemistry mechanism - and the OH-PLIF images to allow comparison of dynamic patterns of heat release and OH distribution in the combustion zone. The DMD analysis was able to identify similar dominant unstable modes in the combustor.^ Recommendations for future work are based on the continued requirement for quantitative and spatio-temporally resolved data for direct comparison with computational efforts to develop predictive capabilities for combustion instabilities at relevant operating conditions. Discriminating instability behavior for the prototypical combustor demonstrated in this study is critical for any robust validation effort Unit physics based scaling of the current effort to multi-element combustors along with improvement in diagnostic techniques and analysis efforts are recommended for advancement in understanding of the complex physics in the multi-phase, three dimensional and turbulent combustion processes in the LDI combustor.^
William E. Anderson, Purdue University.
Aerospace engineering|Mechanical engineering