Data-driven analysis of the flame dynamics in a stochastically forced model annular gas-turbine combustor

Despite significant research, thermoacoustic instabilities in combustors remain a crucial issue in the development and operation of continuous combustion systems. In this research, we study a model annular gas-turbine combustor for analyzing intrinsic thermoacoustic instabilities of a turbulent combustion system. Specifically, we design and verify a combustor with swirl-stabilized methane-air lean premixed flame under the atmospheric pressure condition. We first observe the dynamics within the combustor while varying the equivalence ratio and observe several thermoacoustic modes associated with the intrinsic acoustic modes of the combustor. We then transversely force the combustor with the Gaussian white noise, aiming to observe the noise-induced dynamical behaviors of the combustor. Recognizing the coherently resonating flame under the stochastic excitation, we conduct the system identification for computing linear and nonlinear parameters that govern the pressure oscillation. Van der Pol-type oscillator equation is postulated for this purpose, phenomenologically reproducing the self-sustained pressure oscillation of a thermoacoustic system. We also take high-speed snapshot images of the flame in an annular gas-turbine combustor and analyze the reacting flow field with the spatio-temporal analytical method, namely the dynamic mode decomposition. We find that a wide variety of thermoacoustic dynamics can successfully be analyzed via the data-driven methods used in this study.