An Adaptive System Identification Framework for Early Detection of Themoacoustic Oscillation in Gas Turbine Combustor

Thermoacoustic oscillation is an unexpected and self-sustained high amplitude oscillation that can occur in landed gas turbines, causing severe vibration problems and potential structural failure of the combustor. Early detection of thermoacoustic oscillation is crucial to prevent such failures. In this study, experiments are conducted on a combustor working near the bifurcation point. The nonlinear dynamics of various working conditions are analyzed, and subcritical bifurcation is observed in the experiments. Then, we utilize an output-only system identification method to analyze the combustion system. This fluctuation is modeled as a stochastic Van-der-Pol (VDP) oscillator and its corresponding Stuart-landau equation. The deterministic properties of the combustion system are related to the statistical properties of the stochastic pressure fluctuation. The variational mode decomposition (VMD) algorithm is introduced to process the multi-frequency signal. The first two Kramers-Moyal coefficients of the Fokker–Planck equation are extracted to determine the transition of the probability density function. Finally, they are fitted by the VDP type equation, and the fitting result can be used to predict the occurrence of the Hopf bifurcation. The study provides new avenues to explore the development of early warning indicators to detect thermoacoustic instability in practical combustion systems.