59718 - Flow Response of an Industrial Gas Turbine Combustor to Acoustic Forcing Extracted From Unforced Data 

Linearized methods take as input the temporal mean of a flow and yield insight in the underlying flow dynamics. We apply these methods to the flow in a state-of the art industrial swirl burner. A common problem when using experimental mean flow data as input to these methods is that not all the flow is accessible to the measurement techniques, which truncates the domain for the linearized calculations. The goal of this study is the assessment of the accuracy of different linearized methods to reveal the nonsteady flow effects, in regard to pre-processing, the method applied as well as the type of data. To gain further insights on the robustness, the linearized methods are applied to numerical data from large eddy simulations for comparison. Experimental data for different operating conditions was measured using particle image velocimetry and is used to compare the results against.

We start the examinations with the Linear Stability Analysis (LSA), which examines the global stability of the flow. The goal of this part of the study is to examine the Precessing Vortex Core, a helical flow instability, which occurs in the injector flow. For the numerical mean flow the linear stability analysis produces correct results. For the experimental mean flow however, the stability spectrum is erroneous and does not show the PVC as an unstable mode.

The second and third linear methods are the Resolvent Analysis (RA) and what we call the Directly Forced Linear Analysis (DFLA). Both methods can be used to investigate the Kelvin-Helmholtz vortex shedding in the outer shear layer of the swirling jet when it is acoustically forced from upstream direction. While the LSA only showed correct results for the numerically obtained mean flow, the RA and the DFLA show satisfactory results for both the experimental and numerical mean flow. Finally, this can be explained by the resolvent analysis: It shows that the Kelvin-Helmholtz mechanism is much more sensitive to forcing than all other dynamic mechanisms in the flow for both the experimental and the numerical mean flow. Consequently, it is the only mode, which is to a significant extent excited by the acoustic forcing, independently of the domain.

This study highlights the applicability of linearized methods, to gain deeper understanding of flow dynamics in industrial fuel injection systems, allowing for a time and resource-saving exploration of large parameter spaces.