Wavelet based reduced-order-modelling of intermittent combustor flows

We study the reduced-order-modelling of intermittent flow events encountered in real engineering applications. Characterizing such intermittent events using Fourier-based methods, such as, for instance, spectral-proper orthogonal decomposition (SPOD), would require a large set of modes. Representing such flow features using a smaller subset of modes would require a basis that can optimally capture events localised in time, such as a wavelet basis. We therefore use wavelet-based decompositions to model the flow. Two separate methods are presented: (1) POD on wavelet-filtered data and (2) POD on the coefficients obtained from a wavelet-transform. We apply these decompositions on a turbulent swirl nozzle combustor (Re~20,000) flow field which shows intermittent precessing vortex core (PVC) oscillations. The time-resolved data for the analysis is obtained from large eddy simulation (LES). The low-rank models obtained from this analysis can potentially explain the mechanisms that cause such intermittent behaviour in the flow. This could in turn impact future efforts towards building control strategies for such flows.