1. Introduction
Acoustic resonance remains a classic problem in turbomachinery and aero-engines since Parker’s initial discovery in parallel-plate cascades (Parker Reference Parker1966) and annular cascades (Parker Reference Parker1968). It may cause severe noise (Parker Reference Parker1968) or non-synchronous blade vibrations (Fiquet et al. Reference Fiquet, Aubert, Brandstetter, Buffaz and Vercoutter2021), potentially leading to structural damage. As the acoustic resonance modes are inherently dissipative in a practical fan/compressor, they require external excitation to sustain large oscillations. Typically, resonance is excited by shed vortices, such as the unsteady vortex shedding from blunt trailing edges of the cascade. Natural frequencies of the acoustic resonance and the unsteady vortex shedding approach interact with each other, forming a lock-in process (Cumpsty & Whitehead Reference Cumpsty and Whitehead1971; Welsh, Stokes & Parker Reference Welsh, Stokes and Parker1984). Besides, unsteady tip leakage vortices provide another excitation source (Liu et al. Reference Liu, Willeke, Herbst, Yang and Seume2018).
More recently, however, it was discovered that the synchronous interactions between rotors and stators may also induce resonance phenomena. Kodama & Namba (Reference Kodama and Namba2016) observed high-level rotor–stator interaction noise at certain resonance frequencies. Aotsuka et al. (Reference Aotsuka, Oshio, Tateishi and Kusuda2023) reported high-amplitude unsteady blade loading at specific frequencies slightly below cut-off during numerical study on stator–rotor interaction. Although not mentioned in their paper, these high-loading frequencies perfectly match the characteristics of the single-blade-row Parker-type acoustic resonance modes (Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025).
Therefore, as a special form of stator–rotor interaction, the inlet-distortion–fan interaction should also be able to excite acoustic resonance, among which one common form is caused by the state-of-the-art boundary layer ingestion (BLI) aircraft concept (Ma, Lu & Li Reference Ma, Lu and Li2025). While it promises further increase of fuel efficiency, it also introduces significant inlet distortion (as shown in figure 1) across all operating conditions. This leads to strong interaction forcing on the inlet fan/compressor rotor of a BLI aero-engine. The pumpjet propulsion system in modern submarines also faces the same BLI problem (Xiong et al. Reference Xiong, Rui, Lu, Zhang and Huang2022), similar to figure 1(a). When such interaction forcing couples with the acoustic resonance modes, it may produce severe unsteady blade loading, as preliminarily shown by Mårtensson & Billson (Reference Mårtensson and Billson2024) in their numerical simulations. However, these acoustic resonances cannot be identified during the design stage using conventional Campbell diagrams, which consider only the structural natural frequencies and the synchronous blade forcing at different engine orders. The omission of acoustic natural frequencies may lead to unexpected forced vibrations of rotor blades. Therefore, it is of urgent necessity to develop reliable prediction methods for the acoustic resonances in rotor cascades.
Typical layouts of BLI fan-based propulsion systems (reproduced from figure 7 of Ma et al. Reference Ma, Lu and Li2025).

Previous studies reveal that acoustic resonance in turbomachinery and aero-engines can generally be divided into two categories. One is caused by axial wave reflections between adjacent rotor/stator cascades (Woodley & Peake Reference Woodley and Peake1999a , Reference Woodley and Peakeb ; Hellmich & Seume Reference Hellmich and Seume2008; Fiquet et al. Reference Fiquet, Aubert, Brandstetter, Buffaz and Vercoutter2021) or between the rotor and the inlet duct (Cooper & Peake Reference Cooper and Peake2000; Cooper, Parry & Peake Reference Cooper, Parry and Peake2004). The other category is caused by the interaction of acoustic fields with a single blade row (Duan & McIver Reference Duan and McIver2004; Koch Reference Koch2009). The latter was first discovered by Parker (Reference Parker1966, Reference Parker1967a , Reference Parkerb , Reference Parker1968), and is therefore hereafter named Parker-type acoustic resonance in this paper. These two kinds of acoustic resonance are of distinct mechanisms. The first kind must involve axial propagation of cut-on duct modes and acoustic reflections at separate components, whereas the second kind is dominated by axially cut-off acoustic modes and could occur in one isolated annular cascade. For the axially cut-on acoustic resonances, various two-dimensional (2-D) theoretical models (Woodley & Peake Reference Woodley and Peake1999a , Reference Woodley and Peakeb ; Cooper et al. Reference Cooper, Parry and Peake2004) and three-dimensional (3-D) numerical attempts (Liu et al. Reference Liu, Willeke, Herbst, Yang and Seume2018; Fiquet et al. Reference Fiquet, Aubert, Brandstetter, Buffaz and Vercoutter2021) have been made to predict their resonance frequencies. However, for the Parker-type acoustic resonances, all existing prediction models are limited to the stator case (e.g. Koch Reference Koch1983, Reference Koch2009; Duan & McIver Reference Duan and McIver2004). Although experimental evidence dates back to the 1960s (Parker Reference Parker1968), no theoretical predictions have been made on the rotor-blade-row Parker-type resonance modes. The closest attempt was made by Woodley & Peake (Reference Woodley and Peake1999b ), who developed a 2-D model for rotor–stator tandem cascades based on the Wiener–Hopf method. Nevertheless, they only captured Parker-type resonances of the stator rather than those of the rotor, possibly because that they mainly focused on the undamped resonances energised by vortical–acoustic couplings between cascades, and did not search for eigenfrequencies with dissipative imaginary parts.
For BLI applications, the prediction of the Parker-type acoustic resonances in rotor cascades is essential, as inlet distortions directly interact with the inlet rotor stage of an aero-engine. Therefore, we extend our previous work on Parker-type resonances in stators (Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025), which successfully considered both the non-zero mean flow and the critical 3-D effects, to a rotor blade row in this paper. A 3-D semi-analytic acoustic resonance model for a rotating annular cascade is first developed in § 2, then validated against Parker’s experiments in § 3. Theoretical results on the rotor-blade-row Parker-type acoustic resonance modes and their effect on the inlet-distortion–rotor interactions are discussed in § 4. Conclusions are summarised in § 5, emphasising the need for increased attention to inlet-distortion-induced acoustic resonances and the adoption of a modified Campbell diagram.
2. Theoretical modelling for acoustic resonance in rotor blade row
In this section, the mathematical formulations used in the theoretical model are derived. An analytic expression for the rotor-scattered pressured field is first obtained in § 2.1. The solution method for gust response of the rotor blade row is then introduced in § 2.2, which could be used to theoretically calculate the unsteady rotor response during inlet-distortion–rotor interactions. This follows the 3-D lifting surface developed by Namba (Reference Namba1972, Reference Namba1977), thus is outlined briefly in this paper. Based on these precedent derivations, an eigenvalue model for the prediction of complex acoustic resonance frequencies in a 3-D annular rotor cascade is developed in § 2.3. This eigenvalue model could solve for both the complex resonance frequencies for the rotor-blade-row Parker-type acoustic resonance eigenmodes and their spatial distributions in acoustic duct modes.
Schematic of the simplified rotor: (a)
$r{-}z$
meridian plane view, (b)
$\varphi{-}z$
azimuthal plane view.

2.1. Pressure field scattered by the rotor
The physical model considered in this paper is shown in figure 2. The rotor blade row is assumed to consist of
$B$
blades and to be located in an infinitely long hard-walled annular duct, rotating at a constant angular velocity
$\varOmega$
. Axial reflections from duct ends or other components are neglected in this paper. The blades are simplified as zero-thickness, radially twisted sheets aligned with the local mean flow at zero incidence angle, neglecting mean blade loading and thickness noise. The axial chord length
$C_a$
is constant along the span. As shown in the experiments by Parker (Reference Parker1966), blade thickness has a negligible influence on the acoustic resonance frequency. A cylindrical coordinate system is adopted. The stationary and rotor-fixed coordinate systems are denoted as
$(r,\varphi ,z)$
and
$(r,\hat {\varphi },z)$
, respectively, where
and
$t$
is the observation time. Here and below, the hat symbol denotes quantities related to or measured in the rotor reference frame. The duct has constant hub radius
$R_h$
and tip radius
$R_d$
. The background flow is a basic axial flow of uniform velocity
$U$
and no swirl. The fluid is modelled as an inviscid perfect gas with uniform density
$\rho _0$
and speed of sound
$c_0$
.
The velocity triangle of the rotor is shown in figure 2(b). At radial position
$r$
, the stagger angle is given by
The positive direction of the unsteady blade loading
$\Delta P \exp (-\mathrm{i}\hat {\omega } t)$
is defined as pointing outwards from the pressure surface. Its azimuthal and axial components are then expressed as
$(\Delta P)_\varphi =\Delta P \cos \theta ,\ (\Delta P)_z=\Delta P \sin \theta$
. Due to geometry simplification, the normal vector of blade surface has no radial component. Based on the generalised Lighthill equation (Goldstein Reference Goldstein1976, pp. 189–192), the unsteady pressure field scattered by the rotor under high-Reynolds-number background mean flow can be expressed as (all unsteady perturbations are of frequency
$\hat {\omega }$
in the rotor frame)
The surface integral
$\int _{S(t')}({\cdot })\, \mathrm{d}S$
is at the blade positions of source coordinates
$(r',\varphi ',z',t')$
, where superscript
$'$
denotes coordinates of source points. The sound radiation caused by the viscous stress on blades is neglected under the high-Reynolds-number flows typical in turbomachinery. The quadrupole volume source term in the generalised Lighthill equation is also neglected, as the solid-boundary-related dipole sources dominate the quadrupole volume sources wherever the fluctuating velocities are small (Goldstein Reference Goldstein1976, pp. 220–222). This condition holds when the thin-blade assumption is applied such that the vortex shedding of blunt trailing edges is not considered.
If the inlet distortion has a circumferential periodicity of mode number
$m_0$
in the rotor reference frame, then the surface integral
$\int _{S(t')}({\cdot })\, \mathrm{d}S$
can be further simplified into the integral on one single blade. The circumferential coordinates on this reference blade are set as
$\hat {\varphi }'=-\varOmega z'/U$
without loss of generality. The Green’s function
$G$
of a hard-walled annular duct (Goldstein Reference Goldstein1976, pp. 62–66) can then be substituted into (2.3), after transforming stationary coordinates
$(r,\varphi ,z)$
into the rotor frame
$(r,\hat {\varphi },z)$
using (2.1). Following procedures similar to those in Shen et al. (Reference Shen, Wang, Sun, Zhang and Sun2022), the scattered pressure field is finally expressed as
where the pressure kernel function is
\begin{align} K_p(r,\varphi ,z,t\, \vert \, r',z') &= \sum _{s=-\infty }^{\infty }\sum _{n=0}^{\infty } \frac {B \exp (-\mathrm{i}\omega _s t)}{4\pi \kappa _{nm}}\,\phi _m(k_{\textit{mn}}r)\,\phi _m(k_{\textit{mn}}r') \exp (\mathrm{i}m\varphi ) \nonumber \\ &\quad \times \exp \left(\mathrm{i}m\frac {\varOmega }{U} z'\right) \left \{ \mathrm{H}(z-z')\, Q_{1}^p(r')\exp \left[\mathrm{i}\alpha _1^{\textit{mn}}(z-z')\right] \right.\nonumber \\ &\quad \left. +\, \mathrm{H}(z'-z)\, Q_{2}^p(r') \exp \,[\mathrm{i}\alpha _2^{\textit{mn}}(z-z')] \right \}. \end{align}
The coefficients are
Here,
$\phi _m(k_{\textit{mn}}r)$
is the normalised radial eigenfunction,
$k_{\textit{mn}}$
is the corresponding eigenvalue of circumferential mode number
$m$
and radial mode number
$n$
(Goldstein Reference Goldstein1976, pp. 62–66), and
$\mathrm{H}({\cdot })$
denotes the Heaviside function. Noted that the scattered circumferential mode number
$m$
satisfies
as was first derived by Tyler & Sofrin (Reference Tyler and Sofrin1962), and
$\omega _s$
denotes the frequency measured at the stationary reference frame as
The definition of the acoustic wave number
$k_0$
is
$k_0=\omega _s/c_0$
, where the stationary-frame frequency
$\omega _s$
is applied. The downstream (
$\alpha _{1}^{\textit{mn}}$
) and upstream (
$\alpha _{2}^{\textit{mn}}$
) axial wavenumbers for pressure waves are
where
$M_a=U/c_0$
is the axial Mach number of the mean flow, and
$\beta = \sqrt {1-M_a^2}$
. More details about the derivation process are given in Appendix A.
2.2. Linearised model for the gust response of a rotor
The 3-D lifting surface theory developed by Namba (Reference Namba1977) is applied to model the linearised unsteady response of the rotor blade row. The disturbance velocity on rotor blades is caused by acoustic waves when calculating the acoustic eigenfrequencies, while in the forced response problem of inlet–rotor interactions, the disturbance velocity is caused by inlet distortions. All unsteady pressure fluctuations, and the incident velocity disturbances when considering inlet distortion, are assumed to be small and linearisable.
In addition, although the model could be extended to supersonic relative tip speeds (Namba Reference Namba1972, Reference Namba1987), the present study is confined to the subsonic regime. This helps to simplify the resulting kernel function and the solution process of the integral equation, which are introduced below. Moreover, as will be demonstrated in § 4, the Parker-type acoustic resonance frequencies all go beyond cut-on (such that a genuine resonance mode no longer exists) before supersonic tip speed is reached. Therefore, our discussion on Parker-type acoustic resonances could be limited within subsonic rotor-tip-speed ranges in this paper.
A streamline coordinate system is first established, with streamwise coordinate
$z''$
and anti-lift-wise coordinate
$\hat {\varphi }''$
, as shown in figure 2(b). Based on (2.1), (2.3) and the
$\hat {\varphi }''$
-wise momentum equation, the dipole-induced upwash velocity on the reference blade (
$\hat {\varphi }=-\varOmega z/U$
) can be expressed as
where the velocity kernel function is expressed as
\begin{align} K_v(r,z,t\, \vert \, r',z') &= \sum _{s=-\infty }^{\infty }\sum _{n=0}^{\infty } \frac {-B \exp (-\mathrm{i}\hat {\omega }t)}{2\pi \rho _0 \sqrt {U^2+(\varOmega r)^2}}\,\phi _m(k_{\textit{mn}}r)\,\phi _m(k_{\textit{mn}}r') \nonumber \\ &\quad\times \exp \left[\mathrm{i}m\frac {\varOmega }{U} (z-z')\right]\left[ \mathrm{H}(z-z') (Q_{1}^v + Q_{3}^v)+ \mathrm{H}(z'-z) Q_{2}^v \right ]\!,\nonumber \\ m&=m_0-sB, \end{align}
\begin{align} Q_{1,2}^v&=\frac {M_a\beta ^2}{2\kappa _{nm}(\pm M_a\kappa _{nm}-k_0)} \left (\frac {m}{r'}+\frac {\alpha _{1,2}^{\textit{mn}} \varOmega r'}{U}\right )\left (\frac {m}{r}+\frac {\alpha _{1,2}^{\textit{mn}} \varOmega r}{U}\right )\nonumber \\ &\quad\times \exp \left[\mathrm{i}\alpha _{1,2}^{\textit{mn}}(z-z')\right]\!, \end{align}
The axial wavenumber for vortical waves is
$\alpha _3=(\hat {\omega }+m\varOmega )/U=\omega _s/U$
. Further derivation details are given in Appendix A. Also note that the vortical term
$Q_{3}^v$
contains
$\delta$
-function singularity at blade positions, i.e. at the lifting surfaces. This singularity corresponds to the jump of velocity potential across the lifting surfaces, and should be removed when calculating the velocity field in the fluid domain. Its removal technique by applying finite radial expansion has been given in Namba (Reference Namba1977).
At blade surfaces, the non-penetration boundary condition should be satisfied, which reads
Here,
$\hat {v}_{d}$
is the upwash velocity on the reference blade caused by the incident disturbances. After removing the singularities in the vortical term
$Q_{3}^v$
(Namba Reference Namba1977), (2.10) can be substituted into (2.14) to obtain an integral equation about the unknown blade loading distribution
$\Delta P(r',z')$
. This integral equation is numerically solved using the collocation method applied in Namba (Reference Namba1977) and Shen et al. (Reference Shen, Wang, Sun, Zhang and Sun2022). Details about the collocation process are given in Appendix B. Finally, the integral equation is transformed into a matrix equation for a series of undetermined loading-expansion coefficients
$B_{i'j'}$
:
where
$i,j$
are the axial and radial indexes for the collocation points, and the matrix
$ \unicode{x1D648}_{\textit{ij},i'j'}$
is calculated from the kernel function in (2.11). Here,
$\Delta P$
is determined by the expansion coefficients
$B_{i'j'}$
(see equation (B1) in Appendix B), and the scattered pressure field can then be reconstructed from coefficient vector
$\boldsymbol{B}_{i'j'}$
using (2.4).
2.3. Establishment of an eigenvalue model
Based on the above semi-analytic cascade model, an eigenvalue model can be established to predict the acoustic resonance frequencies of the rotor system. In the stationary reference frame, the acoustic waves propagating inwards and outwards from the rotor blade row can be expressed in duct modes, i.e.
\begin{equation} p_{\textit{in},\textit{out}}=\sum _{m=-\infty }^{+\infty }\sum _{n=0}^{+\infty } {D(A)}^{\textit{in},\textit{out}}_{\textit{mn}}\phi _m(k_{\textit{mn}}r) \exp \bigl[\mathrm{i} m\varphi +\mathrm{i}\alpha _{1(2)}^{\textit{mn}} (z-z_{0}) - \mathrm{i}\omega _s t\bigr]. \end{equation}
As shown in figure 2(a), the axial reference positions are taken at the leading edge (
$z_{0}=0$
for
$D_{\textit{mn}}^{in},A_{\textit{mn}}^{out}$
) and the trailing edge (
$z_{0}=C_a$
for
$A_{\textit{mn}}^{in},D_{\textit{mn}}^{out}$
) of the rotor. During the solution of the acoustic scattering process of a rotor, the rotor-frame frequency of the pressure waves
$\hat {\omega }$
is the independent variable, while the frequency in stationary frame
$\omega _s$
of the rotor-scattered waves follows (2.8) and varies with their circumferential mode number
$m$
. Accordingly, the rotor-frame frequency
$\hat {\omega }$
is used when searching for the acoustic resonance frequency of a rotor, because it is invariable in all scattered waves.
Denote the circumferential mode number of the main duct mode during acoustic resonances as
$m_0$
. We consider
$2M+1$
scattered circumferential modes in total, following (2.7) with
$s=-M,-M+1,\ldots ,M-1,M$
, such that all modes share the same inter-blade phase angle
$(2\pi m_0)/B$
. We consider
$N$
radial modes with
$n=0,1,2,\ldots ,N-1$
. To estimate the resonance frequencies of an isolated rotor, we neglect the interactions between the rotor and other upstream/downstream structures in aero-engines. Thus the disturbance velocity
$\hat {v}_d$
in (2.15) during acoustic scattering is caused by the particle velocity of the inward-propagating acoustic waves, i.e.
\begin{align} \hat {v}_d &=u_\varphi ''\nonumber\\&=-\frac {{D(A)}^{\textit{in},\textit{out}}_{\textit{mn}}}{\rho _0 \sqrt {U^2+(\varOmega r)^2}}\frac {\phi _m(k_{\textit{mn}}r)}{\left(\alpha _{1(2)}^{\textit{mn}}-\alpha _3\right)} \!\left ( \frac {m}{r} + \frac {\alpha _{1(2)}^{\textit{mn}} \varOmega r}{U} \right ) \exp \bigl[\mathrm{i} m \hat {\varphi }+\mathrm{i}\alpha _{1(2)}^{\textit{mn}} (z-z_{0}) - \mathrm{i}\omega t\bigr] , \end{align}
and the upwash velocity at the collocation points
$(r_{\!j},\xi _i)$
(see (B2)) on the reference blade can be written with matrix multiplication as
\begin{align} \unicode{x1D64F}^{D(A)}_{\textit{ij},{\textit{mn}}}&=\frac {\phi _m(k_{\textit{mn}}r_{\!j})}{\rho _0 \sqrt {U^2+(\varOmega r_{\!j})^2}}\left (\frac {m}{r_{\!j}}+\frac {\alpha _{1(2)}^{\textit{mn}}\varOmega r_{\!j}}{U}\!\right )\frac { \exp \bigl[-\mathrm{i}m\varOmega z_i/U + \mathrm{i} \alpha _{1(2)}^{\textit{mn}} (z_i -z_0)\bigr]}{\alpha _{1(2)}^{mn}-\alpha _3}. \end{align}
Note that on the reference blade, the circumferential coordinate
$\hat {\varphi }$
satisfies
$\hat {\varphi } = -\varOmega z_i /U$
.
The secondary pressure field scattered by the rotor is solved using (2.15) and (2.4), where the pressure integral in (2.4) can also be written in matrix form as
The integration matrix
$ \unicode{x1D64E}^{D(A)}_{\textit{ij},{\textit{mn}}}$
could easily be obtained by substituting (B1) into (2.4). Based on the superposition theorem, a matrix equation that relates the mode amplitude coefficients of inward- and outward-propagating acoustic waves can be derived as
\begin{equation} \left [ \unicode{x1D653}(\hat {\omega })\right ]_{2(2M+1)N\times 2(2M+1)N} \left [ \begin{array}{c} \boldsymbol{A}^{out}_{\textit{mn}} \\[3pt] \displaystyle \boldsymbol{D}^{out}_{\textit{mn}} \end{array} \right ]_{2(2M+1)N\times 1} = \left [ \begin{array}{c} \boldsymbol{D}^{in}_{\textit{mn}} \\[3pt] \displaystyle \boldsymbol{A}^{in}_{\textit{mn}} \end{array} \right ]_{2(2M+1)N\times 1} \triangleq \boldsymbol{0}, \end{equation}
following formulations similar to our previous work on stator resonances (Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025, equations (2.19)–(2.23). .The upwash matrix
$ \unicode{x1D64F}^{D(A)}_{\textit{ij},{\textit{mn}}}$
is now formed at the rotor-frame control points on the staggered blades and the integration matrix
$ \unicode{x1D64E}^{D(A)}_{\textit{ij},{\textit{mn}}}$
now calculates the pressure waves scattered by rotating dipole sources. As noted earlier, the rotor-frame fluctuation frequency
$\hat {\omega }$
is now taken as the independent variable because it does not change with
$m$
.
The eigenvalue problem is formulated by forcing the right-hand side of (2.21) to be zero, corresponding to zero inward-propagating acoustic waves and zero axial reflections of the outward-propagating waves caused by other upstream/downstream components. The commonly used singular value decomposition method (Stoer & Bulirsch Reference Stoer and Bulirsch2002) is applied to search for the complex eigenfrequency
$\hat {\omega }_k$
at which the matrix
$ \unicode{x1D653}(\hat {\omega }_k)$
is singular (Woodley & Peake Reference Woodley and Peake1999a
,
Reference Woodley and Peakeb
; Sun et al. Reference Sun, Liu, Hou and Sun2013). The real part
$\mathrm{Re}(\hat {\omega }_k)$
determines the oscillation time period of the acoustic resonance, while the imaginary part
$\mathrm{Im}(\hat {\omega }_k)$
represents the intrinsic temporal damping of the acoustic resonance eigenmodes. The results converge rapidly with the truncation numbers
$M$
and
$N$
, requiring only a limited number of cut-off modes.
In addition, the non-trivial solution for the outgoing wave mode amplitudes
$ [ \boldsymbol{A}^{out}_{\textit{mn}}\ \boldsymbol{D}^{out}_{\textit{mn}} ]^{\rm T}$
in (2.21) at eigenfrequency
$\hat {\omega }_k$
, which represents the eigenmodes of the Parker-type acoustic resonances, just corresponds to the eigenvector of matrix
$ \unicode{x1D653}(\hat {\omega }_k)$
with zero eigenvalue (the non-trivial basis vector of the null space of matrix
$ \unicode{x1D653}(\hat {\omega }_k)$
). This eigenvector could be solved numerically, such that a basic solution for
$ [ \boldsymbol{A}^{out}_{\textit{mn}}\ \boldsymbol{D}^{out}_{\textit{mn}} ]^{\rm T}$
, i.e. the spatial distribution of the Parker-type acoustic resonance eigenmode, can be obtained and further analysed.
3. Validation against Parker’s experiment
The relative flow speed
$W=\sqrt {U^2+(\varOmega R_m)^2}$
experienced by the rotor blade (Parker Reference Parker1968), at which the Parker-type acoustic resonances occur.

The experimental data from Parker (Reference Parker1968), which are obtained using a 22-inch-diameter compressor rotor with hub–tip ratio 0.9091, are used for validation of the eigenvalue model established in § 2.3. To facilitate direct comparison, we predict the resonance frequencies corresponding to figures 5 and 6 in Parker’s orginal paper, where the rotor-frame relative flow speed
$W=\sqrt {U^2+(\varOmega r)^2}$
, main circumferential mode number
$m_0$
and resonance frequency
$\hat {f}=\mathrm{Re}(\hat {\omega }_k)/2\pi$
are recorded in detail. Two sets of flat-plate-blade rotor are experimentally investigated, one with stagger angle
$\theta = 45^\circ$
, and the other with
$\theta = 64.5^\circ$
.
In the experiments (Parker Reference Parker1968), each acoustic resonance mode exists over a limited speed range, and the middle relative flow speed values
$W$
at which the acoustic resonances occur are redrawn and presented in figure 3. Based on the velocity triangle (as shown in figure 2
b) at mid-span
$r=R_m=(R_h+R_d)/2$
and the zero-incidence assumption, the axial flow speed
$U$
and shaft speed
$\varOmega$
at which the resonance modes are recorded can be determined from the experiment results following
$U=W\cos {\theta }$
and
$\varOmega =(U\tan {\theta })/R_m$
. Other geometry parameters of Parker’s test rig are
The background flow conditions
$\rho _0$
and
$c_0$
are not recorded in Parker’s original paper. They are taken as values of the sea-level standard atmosphere
$\rho _0=1.225\ \mathrm{kg\ m^{-3}},\ c_0=340.0\ \mathrm{m\ s^{-1}}$
in our calculations, because the intake of the stationary test rig was directly connected to the atmosphere, and the axial flow speed
$U$
was within the incompressible speed range (
$M_a\lt 0.15$
for all cases).
The comparative results are illustrated in figure 4. The close agreement with the experimental data not only validates the present model but also theoretically confirms the existence of the Parker-type rotating acoustic resonances in a rotor blade row. The predicted resonance frequencies (red lines) are just slightly lower than the experimental data (black lines), with relative errors below 2.5 % for the
$64.5^\circ$
staggered rotor and below 5 % for the
$45^\circ$
staggered rotor. These discrepancies may be attributed to the non-zero incidence angle (approximately
$5^\circ$
) in the experiments (Parker Reference Parker1968), whereas our model assumes zero incidence.
Comparison between rotor-frame theoretical resonance frequencies and Parker’s experimental results (figures 5 and 6 in Parker Reference Parker1968).

4. Results and discussion
The BLI fan used in Mårtensson & Billson (Reference Mårtensson and Billson2024) is investigated in our following studies. Its design point is at flight altitude 10 668 m (35000 ft) and at rotational speed
$N_r=4419$
rpm, corresponding to angular shaft speed
$\varOmega =\pi N_r/30=462.7566\ \mathrm{rad\ s^{-1}}$
. Other key parameters necessary for our calculation are
At design point, the rotor tip speed is slightly supersonic (
$316.54\ \mathrm{m\ s^{-1}}$
). As stated in § 2.2, when studying Parker-type acoustic resonances, our discussion could be limited to subsonic rotor tip speed. Thus in our following calculations, the shaft speed
$\varOmega$
is reduced while assuming constant flow coefficient to maintain zero incidence, such that the axial flow speed
$U$
linearly varies with
$\varOmega$
.
4.1. Parker-type acoustic resonances of a BLI fan
For the BLI problem, a circumferential Fourier expansion can be applied to the velocity field of any steady inlet distortion. Following (2.7) and (2.8), the pressure field excited by the
$m_i$
th harmonic of the distortion has circumferential mode number
$m=-m_i-sB$
and frequency
$\hat {\omega }=m_{i}\varOmega$
in the rotor frame. In the stationary reference frame, it reads
$m=|s|\,B \pm m_i$
(the plus sign is taken if
$s\gt 0$
) and
$\omega =|s|\,B\varOmega$
after adjusting all scattered frequencies to be positive. Consider that the
$s=0$
term corresponds to zero frequency in the stationary frame, the
$s=1$
scattered term with
$m_0=B-m_i$
should be responsible for the excitation of the acoustic resonances, because its frequency is closest to cut-on (while still being cut-off) and within the range of the Parker-type resonances (Koch Reference Koch2009; Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025).
Dependence of the Parker-type acoustic resonance eigenfrequencies
$\hat {\omega }_k$
(a) observed in the rotor frame, (b) transformed into the stationary frame using
$\omega _k=\hat {\omega }_k+m_0\varOmega$
, on circumferential periodicity (at different mode numbers
$m_0$
).

Relative strength between (a) upstream duct mode amplitudes
$|A_{\textit{mn}}^{out}|$
, (b) downstream duct mode amplitudes
$|D_{\textit{mn}}^{out}|$
, of duct modes
$(m_1,0),\ (m_1,1)$
compared against the main duct mode
$(m_0,0)$
for the Parker-type acoustic resonance eigenmodes. Values equal to 0 represent that
$|A_{m_1 n}^{out}|=|A_{m_0 0}^{out}|$
, indicating that the
$(m_0,0)$
mode is only dominant at values below 0.

Accordingly, the complex resonance frequencies
$\hat {\omega }_k$
are calculated using positive
$m_0$
. As an example, the distributions of
$\mathrm{Re}(\hat {\omega }_k)$
and
$\mathrm{Im}(\hat {\omega }_k)$
with varying
$m_0$
at
$N_r=2400$
rpm (shaft speed
$\varOmega =2\pi \times 40\ \mathrm{rad\ s^{-1}}$
) are illustrated in figure 5. Similar distributions are observed at other shaft speeds. The cut-on frequencies for the main duct mode
$(m_0,0)$
and its scattered (aliasing) duct modes at
$m_1=m_0-B=-m_i$
and
$n=0,1$
are also illustrated. The results confirm that all Parker-type acoustic resonances occur at frequencies where the main resonant duct mode is cut-off. For resonances with
$0\lt m_0\le B/2$
,
$\mathrm{Re}(\hat {\omega }_k)$
approaches the cut-on frequencies of the main duct mode
$(m_0,0)$
until a genuine resonance mode ceases to exist due to axial acoustic propagation. This property resembles the Parker-type acoustic resonances in a stator (Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025).
However, for resonances with
$B/2\lt m_0\le B$
, the frequency distribution differs from that observed in a stator. In stator resonances, a frequency jump typically occurs between the
$m_0=B/2$
and
$m_0=B/2+1$
eigenmodes, whereas the lowest-order
$(m,0,0)$
resonances form a continuous branch in figure 5(a). (Here, we apply the same notation method as in a stator case (Koch Reference Koch2009; Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025), where the three numbers in
$(m,0,0)$
each denote the circumferential, radial and axial orders of the resonance eigenmode.) Moreover, an additional resonance branch also emerges at high
$m_0$
. In the stator case (Koch Reference Koch2009; Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025), its higher-frequency
$\mathrm{Re}(\hat {\omega }_k)$
usually suggests that it may correspond to resonances of higher axial order, i.e. the
$(m,0,1)$
resonances. However, as will be discussed in detail in the paragraphs below, in the rotor case, this resonance branch is formed by eigenmodes mainly distributed at duct modes of both mode number
$(m_0,0)$
and mode number
$(m_1,1)$
. These two duct modes differ in both circumferential and radial mode numbers, indicating that the Parker-type eigenmodes in a rotor are of more complex 3-D characteristics. Accordingly, the corresponding resonance branch in figure 5 is denoted as a ’mixed’ resonance branch. Another difference from the stator resonances (Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025) is that the eigenfrequencies
$\mathrm{Re}(\hat {\omega }_k)$
of these two Parker-type resonance branches in rotor cascades no longer increase monotonically with
$m_0$
. The same conclusion still holds after we transform the frequencies into the stationary reference frame using
$\omega _k=\hat {\omega }_k+m_0\varOmega$
, as shown in figure 5(b). Instead, they approach the cut-on frequencies of the scattered duct mode
$(m_1,0)$
or its high-radial-order mode
$(m_1,1)$
, as shown in figure 5(a), indicating significant influence from the
$m=m_1$
duct modes (as will be proved later when analysing the spatial distributions of the resonance eigenmodes). As for the imaginary parts
$\mathrm{Im}(\hat {\omega }_k)$
of the eigenfrequencies, they share similar shapes with
$\mathrm{Re}(\hat {\omega }_k)$
(approaching zero at both
$m_0 \to 0$
and
$m_0 \to B$
), but are of opposite sign and relatively smaller absolute magnitude compared to
$\mathrm{Re}(\hat {\omega }_k)$
.
The spatial distributions of the Parker-type acoustic resonance eigenmodes can then be analysed by solving the eigenvectors corresponding to zero eigenvalues of the singular matrices
$ \unicode{x1D653}(\hat {\omega }_k)$
, resulting in a basic solution for the duct-mode amplitude coefficients of the outward-propagating pressure waves
$\{ {A}^{out}_{\textit{mn}} \}$
and
$\{ {D}^{out}_{\textit{mn}} \}$
. The relative strength between the amplitudes of the scattered duct modes
$(m_1,0),\ (m_1,1)$
and the main
$(m_0,0)$
mode is evaluated and shown in figure 6. The amplitude of the
$(m_0,0)$
duct mode is taken as the reference, and the logarithms of the relative amplitudes, i.e.
$\log (|A_{\textit{mn}}^{out}|/|A_{m_0 0}^{out}|)$
and
$\log (|D_{\textit{mn}}^{out}|/|D_{m_0 0}^{out}|)$
, are calculated, such that when the curve exceeds zero value, the
$(m_0,0)$
duct mode no longer dominates. Results clearly demonstrate that as
$m_0$
increases, the dominant duct mode of the Parker-type acoustic resonances shifts from the
$(m_0,0)$
duct mode to duct modes of circumferential mode number
$m=m_1$
. This trend occurs in both the upstream and downstream pressure waves, and holds for both the
$(m,0,0)$
resonance branch and the mixed resonance branch. For the higher-frequency mixed branch, the dominant duct mode even switches to
$(m_1,1)$
, varied in both circumferential and radial mode number. These eigenmodes, especially at
$m_0=14$
or
$15$
, signify resonances where the energy is distributed across both circumferential and radial mode numbers, making them inherently more 3-D and complex than the classic Parker modes. The spatial distributions of the eigenmodes also further explain the reason why the resonance frequency
$\mathrm{Re}(\hat {\omega }_k)$
approaches the cut-on frequencies of the
$m=m_1$
duct modes when
$m_0$
approaches
$B$
(or equivalently, when the inter-blade phase angle approaches
$2\pi$
). The above properties are absent in stator resonances, where the dominant duct mode always follows the inter-blade phase angle
$\sigma$
as
$m_0=(\sigma B)/(2\pi )$
and does not shift to other scattered (aliasing) mode numbers.
It is also noteworthy that based on the results in figure 6, stationary resonance eigenmodes (i.e.
$|A_{m_1n}^{out}|=|A_{m_0 n}^{out}|$
and
$|D_{m_1n}^{out}|=|D_{m_0 n}^{out}|$
) do not occur in a rotor, even at
$m_0=B/2$
or
$m_0=B$
, while the stator resonances are of circumferential standing waves (Shen et al. Reference Shen, Wang, Li, Zhang and Sun2025) at these circumferential mode numbers. This observation challenges the conventional understanding of acoustic resonances, where the stationary
$\alpha ,\beta ,\delta, \gamma$
standing modes identified by Parker (Reference Parker1966) have long been regarded as basic types of acoustic resonance in turbomachinery cascades (Koch Reference Koch1983; Parker & Stoneman Reference Parker and Stoneman1989; König & Petry Reference König and Petry2012; Hong et al. Reference Hong, Wang, Jing and Sun2020). In a rotor blade row, each eigenmode is composed of counter-rotating spinning duct modes with non-equal mode amplitudes, such that the standing wave resonances disappear in the rotor case.
4.2. Inlet–rotor interactions to excite Parker-type acoustic resonance eigenmodes
Based on (2.15) alone and without the eigenvalue model, theoretical analysis on inlet-distortion–rotor interactions of the BLI fan is first performed. For simplicity, a non-rotating inlet distortion inducing only axial velocity disturbances is considered. To vary
$\hat {\omega }$
at a fixed rotor speed
$\varOmega$
, the distortion is assumed to fluctuate at frequency
$\omega _0$
in the stationary frame, analogous to the cross-wind inlet disturbances studied by Chennuru et al. (Reference Chennuru, Corral, Rodríguez-Blanco, Vahdati and Zhao2025). The resulting rotor-frame perturbation frequency is
$\hat {\omega }_d=\omega _0+m_{i}\varOmega$
, and the upwash disturbance velocity on the rotor blades is
where the circumferential coordinate satisfies
$\hat {\varphi }=-\varOmega z/U$
. To stimulate the BLI effect, a radially increasing amplitude (
$\max 0.1U$
) is adopted for the upwash disturbance velocity, although the shape of this radial amplitude distribution does not influence the qualitative trends in the following results. The unsteady blade loading is quantified by the radially averaged lift coefficient, defined as
where the non-dimensional lift coefficient is
Figure 7 shows the variation of
$\overline {|C_L|}$
with the rotor-frame disturbance frequency
$\hat {f}_d=\hat {\omega }_d/2\pi$
for
$m_0=1,3,6,9,12,15$
(
$m_i=B-m_0$
). The rotational speed is fixed at 2400 rpm as in § 4.1. Pronounced loading peaks occur near the acoustic resonance frequency
$\mathrm{Re}(\hat {\omega }_k)/2\pi$
(marked by dashed lines), except for
$m_0=1$
where a genuine resonance mode is absent due to the cut-on effect. Two
$\overline {|C_L|}$
peaks appear for the
$m_0=15$
case, corresponding to the two resonance frequencies below the cut-on frequency of the
$(15,0)$
duct mode in figure 5. These peaks correspond to forced oscillations of the acoustic resonances, whose driving force comes from the inlet–rotor interaction. Although an abrupt change in
$\overline {|C_L|}$
is also observed near the duct-mode cut-on frequency (marked by dotted lines), these
$\overline {|C_L|}$
peaks caused by cut-on frequencies are markedly smaller than those related to acoustic resonances. This again verifies that the duct-mode cut-on frequency is not a reliable indicator for resonance conditions in inlet–rotor interactions (Parker Reference Parker1983; Woodley & Peake Reference Woodley and Peake1999a
), underscoring the need for a dedicated model to predict acoustic resonance.
Magnitude of unsteady blade loading response at different rotor-frame disturbance frequencies
$\hat {f}_d=\hat {\omega }_d/2\pi$
during inlet–rotor interactions.

To further demonstrate the relationship between these
$\overline {|C_L|}$
peaks and the Parker-type acoustic resonance eigenmodes of the rotor blade row, the spatial distributions of the unsteady pressure fluctuations during inlet-disturbance–rotor interactions at
$\overline {|C_L|}$
peak frequencies are recovered using (2.4) and shown in figures 8 and 9. Figure 8 presents the
$\varphi{-}z$
pressure distributions, where we can clearly observe resonant patterns in the inter-blade regions. At the
$\overline {|C_L|}$
peak frequency corresponding to the
$m_0=9$
resonance, the rotor-scattered pressure field in figure 8(a) clearly shows a two-channel periodicity in the circumferential direction, corresponding to inter-blade phase angle
$(2\pi m_0)/B$
. Whereas in figure 8(b), the pressure distribution near
$m_0=15$
acoustic resonance shows a six-channel periodicity, indicating that the dominant circumferential mode number satisfies
$|m|=3$
. This result justifies our previous observation that at high
$m_0$
, the resonance eigenmode is dominated by duct modes with circumferential mode number
$m=m_1=m_0-B$
.
Figure 9 presents two
$r{-}z$
pressure fields both at resonance frequencies with
$m_0=15$
, but one (figure 9
a) is in the
$(m,0,0)$
resonance branch, while the other one (figure 9
b) is in the mixed resonance branch. They show distinct distribution patterns, where a radial pressure node is visible in figure 9(b), demonstrating that the second radial mode (
$n=1$
) dominates the pressure field. This also follows the previously obtained results on resonance eigenmodes of the mixed branch in § 4.1. In addition, an axial node also exists in the pressure pattern in figure 9(b), verifying that the eigenmodes of the mixed branch are also of non-zero axial order.
Calculated
$\varphi{-}z$
pressure distribution at mid-span
$r=R_m=(R_h+R_d)/2$
during inlet–rotor interactions: (a)
$m_0=9$
(
$m_i=9$
),
$\hat {f}_d = 405\ \mathrm{Hz}$
; (b)
$m_0=15$
(
$m_i=3$
),
$\hat {f}_d = 341\ \mathrm{Hz}$
. The real part of the non-dimensionalised complex pressure value
$\tilde {p}$
is presented, where unsteady pressure is
$p=\tilde {p}\rho _0 U^2 \exp (-\mathrm{i}\hat {\omega } t)$
.

Calculated
$r{-}z$
pressure distribution at inter-blade streamline plane
$\varphi =(\pi /B) - (\varOmega z/U)$
(at the middle between two blades) during inlet–rotor interactions with
$m_0=15$
(
$m_i=3$
): (a)
$\hat {f}_d = 341\ \mathrm{Hz}$
, (b)
$\hat {f}_d = 622\ \mathrm{Hz}$
. The real part of the non-dimensionalised complex pressure value
$\tilde {p}$
is presented, where unsteady pressure is
$p=\tilde {p}\rho _0 U^2 \exp (-\mathrm{i}\hat {\omega } t)$
. The green dashed lines show the axial positions of the leading edge and the trailing edge of the rotor blades.

4.3. Modified Campbell diagram
Example of a modified Campbell diagram for the BLI fan, which was originally studied in Mårtensson & Billson (Reference Mårtensson and Billson2024). The Parker-type acoustic resonance frequencies
$\mathrm{Re}(\hat {\omega }_k)$
of the rotor system are added.

Accordingly, a modified Campbell diagram is proposed in figure 10. The rotor-frame Parker-type acoustic resonance frequencies
$\mathrm{Re}(\hat {\omega }_k)$
at different rotational speeds
$N_r$
(different
$\varOmega$
) are predicted and added to the conventional Campbell diagram, introducing a third kind of natural frequency for the rotor system. These acoustic resonance frequencies at different
$m_0=B-m_i$
are illustrated by coloured curved lines, which transform to thin black dash-dotted lines after the resonance mode disappears (becomes cut-on, as discussed in § 4.1). The coloured lines are mostly of negative slope, but exceptions occur for high-
$m_0$
resonances, which are affected by the decreasing cut-on frequencies of their
$m_1=m_0-B$
scattered duct modes as
$\varOmega \to 0$
. Meanwhile, synchronous excitation frequencies
$\hat {\omega }_d=m_i \varOmega$
of steady BLI inlet distortion are represented by the coloured dotted lines of constant positive slope.
The high-amplitude forced oscillations of the acoustic resonance modes induced by BLI–rotor interaction can therefore be predicted by the intersections between the two sets of coloured lines. As discussed in § 4.1, a distortion of circumferential order
$m_i$
interacts only with the acoustic resonance mode
$m_0=B-m_i$
. Each corresponding pair of frequency lines is marked by the same colour in figure 10, with their intersection points marked by red circles. These intersections indicate situations where a steady inlet distortion may excite acoustic resonance, potentially resulting in high-amplitude unsteady blade loading and forced blade vibration. They are distributed across the entire rotational speed range, corroborating the numerical studies in Mårtensson & Billson (Reference Mårtensson and Billson2024). Note that only the lowest-order
$(m,0,0)$
resonances are shown in figure 10, because higher-order resonances have higher natural frequencies and do not intersect with the BLI–rotor interaction synchronous frequencies.
A positive finding is that for low
$m_i$
(high
$m_0$
) cases, where the inlet distortion is relatively stronger, the intersections occur at high shaft speeds where the ideal acoustic resonance mode is cut-on and disappears. This will suppress resonance and reduce the unsteady blade loading response during inlet–rotor interactions. However, for higher
$m_i$
cases, genuine intersections occur between the Parker-type acoustic resonance frequencies and the synchronous BLI excitation frequencies. Under such cases, the Parker-type acoustic resonance eigenmodes will be excited by the inlet–rotor interactions, leading to a high-amplitude unsteady blade loading response. As shown in figure 10, some of the intersections even lie close to the second structural frequency for this particular BLI fan (marked by the black dashed line), making them potentially more dangerous from an aeroelastic perspective.
5. Conclusions
This paper establishes a 3-D semi-analytic model for predicting acoustic resonances in rotating annular cascades, providing comprehensive estimation for the Parker-type resonance eigenmodes in a rotor. Results reveal that rotor acoustic resonances differ fundamentally from the stator cases due to rotation effects, particularly for high-
$m_0$
resonance modes. The well-known stationary Parker modes are also absent in rotor configurations. A particularly critical scenario, where acoustic resonances are excited by inlet distortions in a BLI aero-engine, is investigated. Analysis of inlet-distortion–rotor interaction confirms that unsteady blade loading peaks at the acoustic resonance frequency rather than at the duct-mode cut-on frequency, necessitating dedicated prediction models for acoustic resonances in a fan/compressor. Moreover, a modified Campbell diagram incorporating these acoustic natural frequencies is proposed. It identifies widespread intersections between synchronous excitations and acoustic resonance modes across the operational speed range, where high-amplitude unsteady blade loading and consequent forced vibrations of rotor blades may be induced.
This work highlights the critical need to consider acoustic resonance during rotor design, especially for BLI configurations where intense inlet distortions exist. The proposed diagram and 3-D prediction model provide a practical framework for avoiding acoustic-resonance-induced vibrations. Future work will further investigate the more critical scenario in which all three natural frequencies coincide in the modified Campbell diagram. Besides, limited by the availability of analytical Green’s functions, the effects of complex 3-D swirl flows caused by non-zero incidence or mean loading cannot be rigorously considered in our semi-analytic model. They may be further studied using a numerical method, and the semi-analytic model established in this paper should serve as a good benchmark reference for numerical solutions of acoustic resonances.
Funding
This work is supported by the National Natural Science Foundation of China (grant no. 52476024).
Declaration of interests
The authors report no conflict of interest.
Appendix A. Detailed derivation of a rotor-scattered field
In the stationary cylindrical coordinate system illustrated in figure 2, the Green’s function for a hard-walled annular duct with axial flow speed
$U=M_a c_0$
can be expressed as (Goldstein Reference Goldstein1976, pp. 62–66)
\begin{align} G(r,\varphi ,z,t\,\vert\, r',\varphi ',z',t') ={}&\frac {1}{4\pi ^2}\sum _{m=-\infty }^{+\infty }\sum _{n=1}^{+\infty } \frac {\phi _m(k_{\textit{mn}}r)\, \phi _m(k_{\textit{mn}}r')}{2\pi } \exp (\mathrm{i}m\varphi ) \exp (-\mathrm{i}m\varphi ') \nonumber \\ &\mbox{}\times \int _{-\infty }^{+\infty } \int _{-\infty }^{+\infty } \frac {\exp \,[\mathrm{i}\alpha (z-z')]}{\beta ^2\alpha ^2 + 2M_ak_0\alpha -k_0^2+k_{\textit{mn}}^2} \nonumber \\ &\mbox{}\times \exp \,[-\mathrm{i}\omega (t-t')]\, \mathrm{d}\alpha\, \mathrm{d}\omega , \end{align}
where
$\phi _m(k_{\textit{mn}}r)$
is the normalised radial eigenfunction of the hard-walled annular duct satisfying the orthogonality as
In this equation,
$\delta _{jl}$
represents the Kronecker delta. After transforming the source coordinates in (A1) into a rotor-fixed reference frame using
$\hat {\varphi }'=\varphi ' - \varOmega t'$
, the Green’s function (A1) could be substituted into (2.3) to obtain
\begin{align} p(r,\varphi ,z,t) &= -\frac {\mathrm{i}B} {4\pi ^2} \sum _{s=-\infty }^{+\infty }\sum _{n=1}^{+\infty } \int _{S_1(t')} \phi _m(k_{\textit{mn}}r)\, \phi _m(k_{\textit{mn}}r') \exp (\mathrm{i}m\varphi ) \exp (-\mathrm{i}m\hat {\varphi }')\, \Delta P_\varphi \nonumber \\ &\quad\times \exp \,[-\mathrm{i}(\hat {\omega }+m\varOmega )t] \int _{-\infty }^{+\infty } \frac {\exp \,[\mathrm{i}\alpha (z-z')]}{\beta ^2\alpha ^2 + 2M_ak_0\alpha -k_0^2+k_{\textit{mn}}^2} \nonumber \\ &\quad\times \left (\frac {m}{r'} + \frac {\alpha \varOmega r'}{U}\right ) \mathrm{d}\alpha\, \mathrm{d}S_1,\quad m=m_0-sB, \end{align}
where the surface integral is simplified into the surface of one reference blade
$S_1$
using circumferential periodicity of the unsteady blade loading, and the infinite integral about frequency
$\omega$
is performed, following similar procedures in Shen et al. (Reference Shen, Wang, Sun, Zhang and Sun2022). For the flat-plate blades shown in figure 2, the surface integral
$\int _{S_1(t')}$
on the blade surface could be further simplified as
$\int _{S_1(t')}(\boldsymbol{\cdot })\mathrm{d}S_1 = \int _{R_h}^{R_d}\int _{0}^{C_a} (\boldsymbol{\cdot })\mathrm{d}r'\mathrm{d}z'/\cos {\theta }$
, such that (A3) is simplified as
\begin{align} p(r,\hat {\varphi },z,t) & = -\frac {\mathrm{i}B \exp (-\mathrm{i}\hat {\omega }t)} {4\pi ^2} \sum _{s=-\infty }^{+\infty }\sum _{n=1}^{+\infty } \int _{R_h}^{R_d}\int _{0}^{C_a} \phi _m(k_{\textit{mn}}r) \phi _m(k_{\textit{mn}}r') \exp (\mathrm{i}m\hat {\varphi }) \nonumber \\ &\quad\times \exp (-\mathrm{i}m\hat {\varphi }') \Delta P(r',z')\int _{-\infty }^{+\infty } \frac {\exp \,[\mathrm{i}\alpha (z-z')]}{\beta ^2\alpha ^2 + 2M_ak_0\alpha -k_0^2+k_{\textit{mn}}^2} \nonumber \\ &\quad\times \left (\frac {m}{r'} + \frac {\alpha \varOmega r'}{U}\right )\mathrm{d}\alpha \mathrm{d}r'\mathrm{d}z',\quad m=m_0-sB, \end{align}
after transforming its circumferential coordinate
$\varphi$
into the rotor reference frame using
$\varphi =\hat {\varphi } + \varOmega t$
.
The potential
$\psi$
of the rotor-scattered field is introduced to simultaneously solve for both pressure and velocity fluctuations. In the rotor-frame streamline coordinate system shown in figure 2(b),
$\psi$
satisfies
where
${\mathrm{D}}/{\mathrm{d}t} = {\partial }/{\partial t} + W\, {\partial }/{\partial z''}$
is the material derivative, and
$W=\sqrt {U^2+(\varOmega r)^2}$
is the streamwise background velocity. By the causality condition, the unsteady perturbations at infinite upstream can be regarded as zero (Namba Reference Namba1977), thus
$\psi$
could be solved by integrating
$p$
along streamlines to infinitely far upstream,
$z''=-\infty$
:
\begin{align} \psi &=-\frac {\exp \left (\mathrm{i}\hat {\omega }z''/W \right )}{\rho _0W} \int _{-\infty }^{z''} p \exp \left (-\mathrm{i}\hat {\omega }z''/W\right ) \mathrm{d}z'' \nonumber\\&= -\frac {\exp \left (\mathrm{i}\hat {\omega }z/U \right )}{\rho _0U} \int _{-\infty }^{z} p \exp \left (-\mathrm{i}\hat {\omega }z/U\right ) \mathrm{d}z. \end{align}
Note that
$\mathrm{d}z''=\mathrm{d}z / \cos {\theta } = \mathrm{d}z\, U / W$
. Substituting (A4) into (A6), then applying the residue theorem to solve for the infinite integral of wavenumber
$\alpha$
, the final expression for the rotor-scattered potential can be derived as
\begin{align} \psi (r,\hat {\varphi },z,t) &= -\frac {\mathrm{i}B \exp (-\mathrm{i}\hat {\omega }t)} {2\pi } \sum _{s=-\infty }^{+\infty }\sum _{n=1}^{+\infty } \int _{R_h}^{R_d}\int _{0}^{C_a} \phi _m(k_{\textit{mn}}r)\, \phi _m(k_{\textit{mn}}r') \exp (\mathrm{i}m\hat {\varphi }) \nonumber \\ &\quad\times \exp (-\mathrm{i}m\hat {\varphi }') \Delta P(r',z') \left [ \mathrm{H}(z-z') Q_{1}^\psi + \mathrm{H}(z'-z) Q_{2}^\psi + \mathrm{H}(z-z') Q_{3}^\psi \right ]\nonumber \\ &\quad\times \mathrm{d}r'\,\mathrm{d}z',\quad m=m_0-sB, \end{align}
with coefficients
\begin{align} Q_{1,2}^\psi &=\frac {M_a\beta ^2}{2\kappa _{nm}(\pm M_a\kappa _{nm}-k_0)} \left (\frac {m}{r}+\frac {\alpha _{1,2}^{\textit{mn}} \varOmega r'}{U}\right ) \exp \,[\mathrm{i}\alpha _{1,2}^{\textit{mn}}(z-z')], \end{align}
Following (A5) and
$\boldsymbol{v}=\boldsymbol{\nabla }\psi$
(where
$\boldsymbol{v}$
is the vector form of the rotor-induced velocity), (2.4) and (2.10) can both be derived from (A7).
Appendix B. Solution of the integral equation
As stated in § 2.2, (2.10) can be substituted into (2.14) to obtain an integral equation about the unknown blade loading distribution
$\Delta P(r',z')$
. This integral equation is numerically solved using the collocation method. Following Namba’s procedures (Namba Reference Namba1977, Reference Namba1987), the unsteady pressure loading can expanded as
\begin{equation} \Delta P(r',z')=\sum _{j'=1}^{J} \phi _{j'}^{(\infty )}(r') \left [ B_{1j'} \cot \left (\frac {\xi '}{2}\right ) + \sum _{i'=2}^{I}B_{i'j'}\sin ((i'-1)\xi ') \right ] , \end{equation}
truncated with
$I$
axial terms and
$J$
radial terms. Function
$\phi _{n}^{(\infty )}(r')$
is the
$m\to \infty$
asymptotic approximation of the radial eigenfunction
$\phi _m(k_{\textit{mn}}r')$
(Namba Reference Namba1987; Shen et al. Reference Shen, Wang, Sun, Zhang and Sun2022). Glauert’s transformation has been used on the axial coordinate as
$z'={C_a(1-\cos \xi ')}/{2},\ z'\in [0,C_a],\ \xi '\in [0,\pi ]$
. The unsteady Kutta condition is inherently applied here (Rienstra Reference Rienstra1992), and the convergence of the numerical solution is checked a posteriori.
The
$I \times J$
evenly spaced collocation points are then applied at the surface of the reference blade, which are positioned at
$(r_{\!j},\xi _i),\ i=1,\ldots ,I,\ j=1,\ldots ,J$
, where
\begin{align} z_{i}= C_a\frac {(1-\cos {\xi _{i}})}{2}, \ \xi _i=\frac {(2i-1)\pi }{2I}, \ i&=1,\ldots ,I, \\ r_{\!j}= R_h + \frac {(2J-1)(R_d-R_h)}{2J}, \ j&=1,\ldots ,J. \end{align}
At these observation points, the integral equations are forced to be precisely satisfied. After substituting (B1) into (2.10), the integration in the radial direction could be analytically performed using the orthogonal property of
$\phi _{n}^{(\infty )}(r')$
(Namba Reference Namba1987; Shen et al. Reference Shen, Wang, Sun, Zhang and Sun2022), therefore only the axial integral
$\int _{0}^{C_a}({\cdot })\,\mathrm{d}z'$
in (2.10) needs to be numerically calculated using the trapezoidal rule. To precisely calculate the Cauchy principal value when integrating the axially singular kernel function
$K_v$
, the axial coordinates of the source points are selected as
such that the observation position
$\xi _i$
is axially placed at the middle between two source points. Here,
$I' = (2\eta +1) I$
, and
$\eta$
is an adjustable integer coefficient, which is taken as
$\eta =4$
in our calculations.
At the collocation points
$(r_{\!j},\xi _i)$
described above, the integral equation (2.14) finally leads to a matrix equation for the undetermined loading-expansion coefficients
$B_{i'j'}$
, as shown in (2.15). Vector
$\hat {\boldsymbol{v}}_{d,\textit{ij}}$
is constructed from the complex values of the upwash velocity at
$(r_{\!j},\xi _i)$
caused by inlet disturbances. The matrix
$ \unicode{x1D648}_{\textit{ij},i'j'}$
corresponds to the integral of the kernel function in (2.10) following (B1) and the procedures described above.






































