Investigation and Prediction of High Frequency Dynamics in Distributed Combustion Systems with LES

by Philip Bonnaire and Wolfgang Polifke

Motivation

Due to the energy transition, industrial turbines are required to have higher load flexibility as well as to further reduce NOx emissions. In addition, greater fuel flexibility is to be achieved, which enables the admixture of larger quantities of renewable hydrogen. Distributed Combustion Systems (DCS) which are characterized by a premixed air-fuel mixture that is induced to the combustion chamber in a secondary stage as reacting jet in cross flow (RJC) have already proven considerable potential to meet these requirements. However, axial fuel staging can result in high-frequency dynamics (HFD) that may lead to higher pollutant emissions and turbine damage. Since the HFD seem to differ in their mechanism of action from the low-frequency oscillations, which are already well understood, fundamental investigations must be carried out here.

Objectives and Strategy

The objective of this project is to obtain further knowledge about the mechanisms of HFD in order to develop remedial adjustments. Recent works have shown that large eddy simulation (LES) with high resolution is already sufficient to enable studies on high-frequency instabilities [1]. Therefore, simulations are carried out using the software STAR-CCM+ to investigate which operating conditions are prone to these dynamics and how exactly the second stage contributes to the occurrence of HFD. STAR-CCM+ offers a built-in CAD and meshing module as well as a large number of pre-implemented models for turbulent combustion in order to correctly reproduce the interaction between turbulence and chemistry, which will be very beneficial in the context of this work. Then a modal analysis is applied to the LES data to capture the dominant spatial and temporal structures. For this purpose, the common tools will be used such as (spectral) proper orthogonal decomposition (POD) and dynamic mode decomposition (DMD). POD uses a singular value decomposition (SVD) to create orthogonal eigenmodes that can be ranked by their eigenvalue. However, since this method only allows a spatial investigation of the problem, the use of DMD and SPOD is necessary to determine temporal characteristics. DMD offers the possibility of separating periodic behavior from a data set into frequency and grow rate [2]. A local analysis is also possible, which is very advantageous for the investigation of the RJC. Spectral POD combines POD and DMD, leading to reduced amount of space- and time dependent modes that represent the data snapshots [3]. These methods are also very well established for the analysis of low-frequency oscillations. However, the task is to investigate the validity of these decomposition techniques considering high-frequency dynamics.

Since LES is very computationally intensive, reduced order models (ROMs) are derived from the generated data. These models allow optimization studies to be carried out due to their reduced complexity and the associated savings in computational time. The main task is to identify a flame transfer function (FTF), which relates the heat release to acoustic fluctuations.  Although methods for system identification are already well tested in the field of thermoacoustics, the challenge is that these existing techniques may not achieve good results due to the complexity of the configuration examined within the scope of this project and the associated discretization of the LES. The extent to which these methods can be used for the DCS system must therefore also be determined.

The so-called linearized reactive flow (LRF) offers an alternative to the FTF in order to model the unsteady heat release. The thermoacoustic modes are calculated directly from linearized conservation equations. This method is not yet as popular as the FTF model but can also be considered if the previous approaches do not deliver the desired results.

References

[1] Mathieu Zellhuber, Joachim Schwing, Bruno Schuermans, Thomas Sattelmayer, and Wolfgang Polifke. Experimental and Numerical Investigation of Thermoacoustic Sources Related to High-Frequency Instabilities. International Journal of Spray and Combustion Dynamics, 6(1):1-34, March 2014.

[2] Peter J. Schmid. Dynamic mode decomposition of numerical and experimental data. Journal of Fluid Mechanics, 656:5-28, August 2010.

[3] Aaron Towne, Oliver T. Schmidt, and Tim Colonius. Spectral proper orthogonal decomposition and its relationship to dynamic mode decomposition and resolvent analysis. Journal of Fluid Mechanics, 847:821-867, July 2018.

Acknowledgement