Filtering acoustic from hydrodynamic velocity using modal decomposition methods on an acoustic liner under grazing turbulent flow

• Comparison of POD, CCD, and SPOD for acoustic–turbulent field separation. • Triple decomposition is ineffective for turbulent, low-AHFR conditions. • POD works for dominant acoustics but fails with noise or energetic peaks. • SPOD provides frequency-resolved modes but requires long time-resolved data. • CCD isolates coherent modes if a clean reference signal is available. The separation of acoustic-induced velocity from the turbulent velocity fluctuations is tackled on a numerical database representing a segment of an acoustic liner subjected to a grazing acoustic wave and turbulent flow. This scenario is meaningful due to the challenge of distinguishing sound-induced fluctuations from aerodynamic ones, and it has practical implications on the estimation of the liner’s impedance and sound absorption properties. The separation is performed employing modal decomposition methods: proper orthogonal decomposition (POD), spectral proper orthogonal decomposition (SPOD), and canonical correlation decomposition (CCD). The acoustic-induced velocity is reconstructed by selecting a limited number of modes representative of the acoustic-induced flow. All the decomposition methods are influenced by the acoustic-to-hydrodynamic fluctuation ratio, i.e., the relative amplitude of the acoustic waves and turbulent fluctuations. The CCD and SPOD outperform POD when the acoustic amplitude is low compared to the flow turbulence intensity. The acoustic forcing frequency must be known a priori or easily identifiable in the spectrum for SPOD. CCD better captures non-linear effects, e.g., due to the vortex shedding at high sound pressure levels, which are associated with high-order modes.