The prediction of high-frequency acoustic oscillations in gas turbine combustors is an important issue, related to engine performance, NOx emissions, component lifetime and engine operational flexibility. Different methods with increasing complexity and predictive ability have been discussed in a number of papers. Application of these methods requires large computational capacity and long computational times. Therefore, a limited number of variants of small combustor models or small sectors can be analyzed in a reasonable time. This paper presents an approximate approach, applicable under certain specific conditions. It is based on an understanding that the acoustic pressure oscillations are tied to the oscillation in heat release rate. The interaction is taking place in the heat release zone, independent of the type of the feedback mechanism. For a typical gas turbine combustion chamber, many acoustic modes exist in the frequency range of interest. However, only a few of these modes are excited by the combustion process and thus are relevant. The mode excitation depends both on combustion noise (due to flame excitation contribution independent of the acoustic field) and combustion instability (acoustic mode made unstable by the flame transfer function). With a flame surface obtained from steady state CFD simulation, and with acoustic mode shapes obtained from a Finite Element package, the forced acoustic response of the combustion system to the flame excitation was calculated. In a first validation step, this method has been tested on a single burner atmospheric test facility. In a second step, the method will be applied to an annular SEV combustion chamber of a GT26 ALSTOM gas turbine. The strength of this approach is that large models can be analyzed quickly to show the influence of changes in a flame position and effect of the combustor geometry. The weakness is that combustion instabilities can not be addressed by such a method. Furthermore, the phase relation of the excitation between different parts of the flame is frequency dependant and needs to be given as an input, which requires an experience and expert knowledge.

This content is only available via PDF.
You do not currently have access to this content.