Identifying wavemakers of self-excited flow oscillations using information theory and complex network analysis

Flow regions sustaining large scale self-excited coherent oscillations are called wavemakers. Identifying these regions can aid informed engineering design decisions aimed at suppressing these oscillations when they are not beneficial. Structural sensitivity maps determined from linear stability analysis (LSA) is a physics-based approach for identifying the wavemaker. Geometrically complex flows in engineering applications whose time-averaged flows may not have symmetries, can yield large and computationally challenging LSA eigenvalue problems. We present an alternative data-driven approach using complex network analysis (CNA) to determine the wavemaker of flows with large scale oscillations, using only time-series data determined from large eddy simulation (LES) or experimental measurements. A network representation of the dataset is constructed using measurement points in space as nodes. Two ways of determining node connectivity: using correlation and mutual information between velocity fluctuations at the nodes are evaluated. Nodes with high weighted closeness centrality in the network are mapped back into physical space to identify the wavemaker. Using LES data from a turbulent swirl combustor (Re~20,000) with a coherent precessing vortex core (PVC) oscillation, the CNA predicts the wavemaker region in good agreement with structural sensitivity analysis. Disrupting the flow field in the wavemaker region suppresses the PVC oscillations, thereby, validating the CNA approach.