Identification of Combustion Noise and Flame Dynamics of Confined Turbulent Flames

by Malte Merk and Wolfgang Polifke

Motivation

In various industrial applications like aeronautical engines or gas turbines, confined turbulent combustion processes yield a high level of combustion noise. Thereby the combustion noise is an undesirable by-product of combustion processes, as it decreases the combustion efficiency or even leads to combustion instabilities. The noise is mainly caused by unsteady heat release rates resulting in unsteady volumetric expansions which in turn generate sound waves in the reactive regions.

Compared to a laminar combustion regime, an accurate estimation of the combustion noise level in an turbulent application is a challenging task, since it is the result of two different contributions: Firstly, the core noise which is directly associated to the turbulent combustion, acts like a source of noise. Secondly, the flame is disturbed by the generated sound waves itslef that are reflected by the surrounding confinement ( combustor walls, liners, compressor and turbine stages ) and impinging again upon the flame. To enable a reliable prediction of the combustion noise level these two effects must be distinguished in order to model the core noise and the flame-acoustic interaction seperatly. However, the distinction between the effects is noticably impeded by turbulence.

Objectives and Strategy

Setups with an intense flame-acousting coupling, like it is the case for most confined turbulent combustion processes, cannot be handled properly yet by existing models. It is then arduous to determine accuratly the source terms without knowledge of the flame dynamics and vice versa. Therefore the overall objective is to develop methods that allow the concurrent determination of the combustion noise source and the flame dynamics.

This will be realized by performing numerical experiments by means of Large Eddy Simulations ( LES ) of the combustion process while applying a broadband excitation signal. The generated input and output time series can be combined with System Identification techniques ( SI ) to infer models of the noise source and the flame dynamics, namely the Flame Transfer Function ( FTF ). So as to cross-validate the acquired results, experimental data will be used that is provided by the Laboratoire EM2C, École Centrale, Paris.

The combination of SI and LES allows to predict from a single LES computation the FTF not only for one, but for a wide range of frequencies. Repeated computational expensive LES can thus be avoided.

Once the combustion noise sources and FTF are correctly identified, they can be joined with a suitable low order network model of the combustor acoustics in order to predict the combustion noise level in confined systems, see figure 1.

Vid. 1: Turbulent flame undergoing self-sustained themoacoustic oscillations

Acknowledgment