Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-17T10:12:07.492Z Has data issue: false hasContentIssue false

Hierarchical equivalent source method with adaptive hydrodynamic filtering for robust aeroacoustic reconstruction

Published online by Cambridge University Press:  04 May 2026

Zhiteng Zhou
Affiliation:
State Key Laboratory of Nonlinear Mechanics, Institute of Mechanics, Chinese Academy of Sciences, Beijing 100190, PR ChinaSchool of Engineering Sciences, University of Chinese Academy of Sciences, Beijing 100049, PR China
Yi Liu
Affiliation:
State Key Laboratory of Nonlinear Mechanics, Institute of Mechanics, Chinese Academy of Sciences, Beijing 100190, PR ChinaSchool of Engineering Sciences, University of Chinese Academy of Sciences, Beijing 100049, PR China
Zhenyu Zang
Affiliation:
State Key Laboratory of Nonlinear Mechanics, Institute of Mechanics, Chinese Academy of Sciences, Beijing 100190, PR ChinaSchool of Engineering Sciences, University of Chinese Academy of Sciences, Beijing 100049, PR China
Shizhao Wang*
Affiliation:
State Key Laboratory of Nonlinear Mechanics, Institute of Mechanics, Chinese Academy of Sciences, Beijing 100190, PR ChinaSchool of Engineering Sciences, University of Chinese Academy of Sciences, Beijing 100049, PR China
*
Corresponding author: Shizhao Wang, wangsz@lnm.imech.ac.cn

Abstract

The equivalent source method (ESM) is one of the fundamental methods for reconstructing the far-field acoustic pressure and identifying the sources in aeroacoustics. However, it suffers from disturbances of near-field hydrodynamic pressure. We propose a hierarchical ESM (HESM) to suppress the disturbance by filtering out hydrodynamic pressure. The hydrodynamic pressure is filtered out using a frequency-domain convection operator. This operator replaces the prescribed convection velocity in Taylor’s frozen flow hypothesis with an adaptive complex convection velocity. The complex convection velocity is adapted in an iterative way to take into account the multiscale convection and spatial amplification of hydrodynamic pressure. The hydrodynamic pressure associated reconstruction error can thus be suppressed. The proposed HESM is compared with the Ffowcs Williams–Hawkings analogy method and the ESM that directly uses the pressure and the filtered pressure with the widely used uniform convection velocity. The use of complex convection velocity-based hydrodynamic pressure filtering mitigates the overprediction of acoustic pressure and enables precise reconstruction of both acoustic pressure directivity and spectra.

JFM classification

Information

Type
JFM Papers
Copyright
© The Author(s), 2026. Published by Cambridge University Press
Figure 0

Figure 1.Schematics of the ESM for prediction of sound generated from flows over a circular cylinder.

Figure 1

Figure 2.Flowchart illustrating the hierarchical reconstruction of acoustic and hydrodynamic pressure.

Figure 2

Figure 3.Schematics of the initial pressure field of the 2-D convecting vortex.

Figure 3

Figure 4.Comparison between the results of PESM and HESM for the non-dimensional frequencies at (a) $\omega = 0.21$ and (b) $\omega = 0.52$.

Figure 4

Figure 5.Contour of the (a) acoustic pressure and (b) hydrodynamic pressure at the vortex shedding frequency.

Figure 5

Figure 6.Contour of the (a) acoustic pressure and (b) hydrodynamic pressure at the second harmonic.

Figure 6

Figure 7.Comparison of the time derivative term ${\textrm {i}}\omega {p'_v}$ and the convective term ${{\boldsymbol {u}}_{\boldsymbol {v}}} \boldsymbol{\cdot }\boldsymbol{\nabla }{p'_v}$ at (a) the vortex shedding frequency and (b) the second harmonic.

Figure 7

Figure 8.Acoustic pressure of DNS results and predicted by using HESM, FESM, PESM and FW–H analogy method at (a) the vortex shedding frequency and (b) its second harmonic, respectively. The observers are located 14.5$D$ away from the circular cylinder centre. The free stream Mach number is 0.3.

Figure 8

Figure 9.Acoustic pressure of DNS results and predicted by using HESM, FESM and PESM at (a) the vortex shedding frequency and (b) its second harmonic, respectively. The observers are located 3.5$D$ away from the circular cylinder centre. The free stream Mach number is 0.3.

Figure 9

Figure 10.Comparison of acoustic pressure at observer location (0, 75$D$) using DNS results from the work of Inoue & Hatakeyama (2002) and the HESM, FESM and PESM approaches.

Figure 10

Figure 11.Acoustic pressure of DNS results and predicted by using HESM, FESM and PESM at (a) the vortex shedding frequency and (b) its second harmonic, respectively. The observers are located 7.5$D$ away from the circular cylinder centre. The free stream Mach number is 0.5.

Figure 11

Figure 12.Acoustic pressure of DNS results and predicted by using HESM, FESM and PESM at (a) the vortex shedding frequency and (b) its second harmonic, respectively. The observers are located 3$D$ away from the circular cylinder centre. The free stream Mach number is 0.5.

Figure 12

Figure 13.Relative error of filtered pressure predicted by using HESM at (a) the vortex shedding frequency (b) its second harmonic, respectively.

Figure 13

Figure 14.Comparison of convection velocity’s amplitude at (a) the vortex shedding frequency (b) its second harmonic, respectively. The black solid line denotes the mean velocity over 10 vortex shedding periods. Different blue lines mark the fifth to thirtieth epochs. The green dash line represents the convection velocity amplitude of pressure source. The grey long-dash line represents the strength of the vortical sources.

Figure 14

Figure 15.Schematic of the instantaneous turbulent flow field around a circular cylinder. The number and positions of the equivalent sources and collocation points are shown schematically as adjustable parameters; their influence on the results is analysed in Appendix B.

Figure 15

Figure 16.Schematics of the set-up of the equivalent sources and collocation points. Note that the aspect ratio of the sketch is shown schematically.

Figure 16

Figure 17.(a) Surface pressure coefficient and (b) mean streamwise velocity at the wake centreline. Experimental and B-spline simulation results for surface pressure coefficient are sourced from Kravchenko & Moin (2000), while DNS results are from Li et al. (2021). Experimental mean streamwise velocity results come from Ong & Wallace (1996), with B-spline and DNS results sourced from Kravchenko & Moin (2000) and Li et al. (2021), respectively.

Figure 17

Figure 18.Acoustic pressure spectra of LES results and predicted by using the HESM, FESM and PESM at the observers: (a) $(8.46D,8.51D,1.57D)$; (b) $(-8.55D,8.42D,1.57D)$.

Figure 18

Figure 19.Directivity of the SPL during $0.1\lt St\lt 1.0$.

Figure 19

Figure 20.Comparison of the relative error of different position of collocation points between the predicted acoustic pressure by using the HESM, FESM and PESM with the numerical results from CFD. Panels (a) and (b) represent the results of extrapolation at the non-dimensional frequency St and 2St, where St is the non-dimensional vortex shedding frequency. Panels (c) and (d) represent the results of interpolation at the non-dimensional frequency St and 2St.

Figure 20

Figure 21.Comparison of the relative error of different position of equivalent sources between the predicted acoustic pressure by using the HESM, FESM and PESM with the numerical results from CFD. Panels (a) and (b) represent the results of extrapolation at the non-dimensional frequency St and 2St. Panels (c) and (d) represent the results of interpolation at the non-dimensional frequency St and 2St.

Figure 21

Figure 22.Comparison of the relative error of different numbers of collocation points between the predicted acoustic pressure by using the HESM, FESM and PESM with the numerical results from CFD. Panels (a) and (b) represent the results of extrapolation at the non-dimensional frequency St and 2St. Panels (c) and (d) represent the results of interpolation at the non-dimensional frequency St and 2St.

Figure 22

Figure 23.Comparison of the relative error of different number of equivalent sources between the predicted acoustic pressure by using the HESM, FESM and PESM with the numerical results from CFD. Panels (a) and (b) represent the results of extrapolation at the non-dimensional frequency St and 2St. Panels (c) and (d) represent the results of interpolation at the non-dimensional frequency St and 2St.

Figure 23

Figure 24.Comparison of the relative error in far-field sound extrapolation at the vortex shedding frequency for flow over a 2-D circular cylinder at free stream Mach number 0.5. The error is plotted as a function of (a) the number of collocation points, (b) the number of equivalent sources, (c) the position of the collocation points and (d) the position of the equivalent sources.

Figure 24

Figure 25.Comparison of the relative error in far-field sound extrapolation at the vortex shedding frequency for flow over a 3-D circular cylinder at Reynolds number 3900. The error is plotted as a function of (a) the number of collocation points, (b) the number of equivalent sources, (c) the position of the collocation points and (d) the position of the equivalent sources.

Figure 25

Figure 26.Comparison of reconstructed pressures at the vortex shedding frequency. Panels (a) and (b) are the reconstructed hydrodynamic pressure for $N_s=10$ and $N_s=30$, respectively. Panel (c) is the difference between the hydrodynamic pressure magnitudes, defined as $p'_{1st, error}=|p'_{1st, N_s=10}|-|p'_{1st, N_s=30}|$.

Figure 26

Figure 27.Schematics of the instantaneous flow field of the round jet and set-up of the equivalent sources and collocation points.

Figure 27

Figure 28.(a) Centreline axial velocity and (b) its streamwise decay. The solid symbols denote data from the DNS of Freund (2001). The dashed line in (b) represents the theoretical decay formula for the jet centreline velocity. The parameter $y_0$, denoting the virtual origin of the jet, is set to $2.075D$ in accordance with Freund (2001).

Figure 28

Figure 29.Acoustic pressure spectra of CFD results and predicted by using the HESM and PESM at the observers: (a) $(17D,10D,0)$; (b) $(37D,10D,0)$.