Analysis of time-signal of an industrial gas turbine combustor under an unstable condition

This paper analyzes the time-signals of pressure sensors mounted in an industrial gas turbine combustor under an unstable condition. The present investigation is aimed at the discussion of the sudden increase in amplitude due to the limit-cycle oscillations and of its temporal evolution. To this purpose, different post-processing tools are described and adopted: i.e. wavelet transform, cross-correlations, time-space Fourier transform and proper orthogonal decomposition (POD). The properties of the wavelet transform are used in order to identify the time of occurrence and the frequency of the limit-cycle oscillations. They occur at the second harmonic of the natural frequency of the annular combustion chamber. The amplitude of the pressure fluctuations at this characteristic frequency increases to a critical value with very large amplitude in about 0.15s that corresponds to about 26 periods of the phenomenon. Within this period, the pressure signals from two neighboring burners have a quite large and increasing degree of correlation as it is observed from the cross-correlation of the signals. The time-space Fourier transform suggests that the instability couples with a natural mode of the combustion chamber. The azimuthal wave length of such mode is half of the combustion chamber circumference (this corresponds to an azimuthal mode 2). According to this findings, the POD is used to provide an identifier for the occurrence of the limit-cycle oscillations. In fact, POD is known to isolate the deterministic fluctuations based on an energy rank. Hence, the first POD mode isolates the effect of specific frequency forcing and its energy content is retained in the first POD eigenvalue which is used as identifier.