Professional Documents
Culture Documents
IN
A SWIRLING FLOW}
In the final part of this dissertation, the theory of unsteady swirling
motion developed in the previous chapters is applied to a boundary-value
unsteady aerodynamic problem. To this end, we consider an imposed on a
swirling mean flow incident acoustic-vorticity wave (gust) evolving from
a
given upstream point of origin and interacting with an annular cascade of
blades. This interaction produces a near-field unsteady aerodynamic
response
coupled with far-field hydrodynamic and acoustic pressure fields.
For simplicity, we assume a stationary row of flat-plate non-loaded
blades
installed at zero angle of attack to the mean flow.
The blade chords are centered at $x=0$, and the incident gust plane
of origin (or measurement) coincides with the upstream computational
domain
boundary plane at $x=x_{in}.$
The boundary-value problem is governed by a set of equations
(\ref{G11})-(\ref{G12}). Without loss of generality, we consider a single
harmonic component of an upstream unsteady flow excitation, with a given
temporal
frequency $\omega .$ As usual, an arbitraty initial disturbance field at
$%
x=x_{in}$ is viewed as an inverse Fourier transform of the harmonic
solution
over the range of wavenumbers:
\begin{equation}
{\bf u}_{in}(x,r,\theta ;t)=\int_{-\infty }^\infty \sum_{n_g=-\infty
}^\infty {\bf a}(r)\exp \{i[k_g(x-U_0t)+n_g\theta ]\}dk_g \label{B1}
\end{equation}
In (\ref{B1}), $n_g$ is a circumferential wavenumber, and $k_g=\omega
/U_0$
is an axial wavenumber of an initial harmonic perturbation. The amplitude
vector ${\bf a}(r)$ is expanded in terms of Hankel functions of order one
as
indicated in formulae (\ref{I34})-(\ref{I36}) and (\ref{I75})(\ref{I77}):
\begin{equation}
a_{r,\theta }(r)=D_{r,\theta }\{H_1^{(1)}(\mu _nr)-\frac{H_1^{(1)}(\mu
_nr_h)%
}{H_1^{(2)}(\mu _nr_h)}H_1^{(2)}(\mu _nr)\}, \label{B2}
\end{equation}
where $D_{r,\theta }$ are given constants, and $\mu _n$ is the $n$-th
root
of (\ref{I35}).
Our present numerical analysis is restricted to the case of an imposed
upstream solenoidal perturbation. Thus, the axial amplitude component $%
\begin{equation}
\Delta p_s^{\prime }=\frac{D_0}{Dt}\Delta \phi _s=0 \hspace{0.5in} and
\hspace{0.5in} \Delta (\nabla \phi _s\cdot {\bf n})=0,\hspace{0.5in}{\bf
x\in
}{\cal W}, \label{B19}
\end{equation}
where the wake surface normal vector ${\bf n}$ is the same as in
(\ref{B15}%
). Also note that the wake conditions do not couple scattered and
incident
unsteady fields since the gust solution ${\bf Z}_g$ also satisfies the
same
conditions (\ref{B19}).
The first equation in (\ref{B19}) can be rewritten as
\begin{equation}
-ik_g\Delta \widehat{\phi }_s+\frac{\partial \Delta \widehat{\phi }_s}{%
\partial \zeta }=0,\hspace{0.5in}{\bf y\in }{\cal W} \label{B20}
\end{equation}
and integrated for $\zeta >1$ to give
\begin{equation}
\Delta \widehat{\phi }_s=\Delta \widehat{\phi }_{sTE}\exp \{ik_g(\zeta
-1)\},%
\hspace{0.5in}{\bf y\in }{\cal W}, \label{B21}
\end{equation}
where $\Delta \widehat{\phi }_{sTE}$ stands for the potential jump at the
blades trailing edges. Note that the Kutta condition is satisfied
automatically in this formulation.
To apply the jump condition numerically, we recall the cascade
quasiperiodicity relation (\ref{B8}) which allows to calculate the
potential
jump across the wake as
\begin{equation}
\Delta \widehat{\phi }_s(\zeta ,\xi ,\vartheta )=\widehat{\phi }_s(\zeta
,\xi ,\vartheta )-\widehat{\phi }_s(\zeta ,\xi ,\vartheta +2\pi
/N_b)e^{-i\sigma } \label{B22}
\end{equation}
In the numerical analysis, equation (\ref{B21}) is applied along the
surface
$\vartheta =2\pi /N_b$ whereas the second relation in (\ref{B19}) is
implemented for the grid points at $\vartheta =0.$ Since the
quasiperiodicity condition couples the normal velocity components on both
sides of the blade passage domain,
\begin{equation}
\frac{\partial \widehat{\phi }_s(\zeta ,\xi ,\vartheta )}{\partial
n}=\frac{%
\partial \widehat{\phi }_s(\zeta ,\xi ,\vartheta +2\pi /N_b)}{\partial n}
%
e^{-i\sigma }, \label{B23}
\end{equation}
the continuity of{\it \ }$\partial \widehat{\phi }_s/\partial n$ across
the
cluster far enough from the critical layer. In some cases (see examples
below) the nearly-convected modes may carry away the total unsteady
pressure
response pertaining to a certain modal number $m$. Therefore, such modes
should not be neglected in the formulation of the outflow conditions in
terms of the far field expansion (\ref{B29}) but they should also be
filtered to satisfy the outgoing wave criteria (\ref{B30}) and
(\ref{B31}).
A numerical implementation of the inflow/outflow conditions involves the
application of the pseudospectral technique for an eigenvalue analysis of
equations (\ref{G11})-(\ref{G12}) (see Chapter 7 for details). In
contrast
with a typically sparse filling of the matrix elements pertaining to the
governing equation and boundary conditions, the far field conditions may
fill up the corresponding matrix rows densely depending on the number of
terms taken in the truncated series (\ref{B29}). More non-zero entries
have
to be included at the downstream boundary where the far field condition
is
formulated for a scattered pressure function
$\widehat{p}_s=(ik_g-\partial
/\partial \zeta )\hat \phi _s,$ to be consistent with the wake
conditions.
\subsection{\bf \ Validation and Comparison with 2D Linear Cascade
Theory.}
The results of numerical solution for the boundary-value problem are
presented below in comparison with a flat plate linear cascade code based
on
the integral method (Ventres, \cite{Vent}). A brief discussion of the
conventional gust definition applied to model an upstream disturbance
field
in two-dimensional unsteady aerodynamics will be appropriate here (for
details, a reader can refer, e.g., to Goldstein \cite{AEROAC}, p.222).
Following this, an important note on the comparability of the present and
linear cascade analyses will be made.
\begin{figure}
\vspace{4in}
%\centerline{\psfig{figure=../Gvan/FORT/BLADES/AIAA97/lincas.eps,height=4
in,width=5in}}
\caption{Unsteady inflow in a linear cascade model (adapted from
\cite{AEROAC}.}
\label{bvp2}
\end{figure}
An infinite linear cascade can be obtained by ''unrolling'' the blade row
of
a fan at its mean radius $r_m.$ If we assume that the fan is rotating
with
angular velocity $\widetilde{\Omega }$ than the cascade will move with
velocity $U_c=\widetilde{\Omega }r_m\,$between two infinite lines
connecting
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/cp_uvent06.eps,height=3i
n,width=5in}}
\caption{Aerodynamic response to a gust with $|a_\theta (r)|=\cos\pi s$,
at blade strips: --,$s=0.25$;
-,$s=0.5$; -.-,$s=0.75$; -,$s=1,$ compared to 2-D cascade analysis
(dots). $k_{g}=6.5$ (a), 7 (b), 7.5 (c), 8 (d); $n_{g}=0$. $M_0=0.6,
\chi =0.$ }
\label{bvp8}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/eig_u06.eps,height=3in,w
idth=5in}}
\caption{Eigenvalues for upstream/downstream unsteady response. $k_{g}=1$
(a), 3 (b), 6 (c), 9 (d); $n_{g}=0$. $M_0=0.6, \chi =0.$ }
\label{bvp9}
\end{figure}
The gust (\ref{B2}) used henceforth produces an unsteady lift which
matches
well with the linear cascade result up to frequency $k_g\simeq 6$ but
then
does not exhibit any resonant peaks; it appears that the comparison is
improving for higher frequencies as the lift is decreasing.
The complex behavior of $C_L(k_g)$ plots in Figure \ref{bvp7} is
connected
with acoustic cut-on of modes in the far field of the blade row. The
unsteady lift is known to be very sensitive and exhibit large variations
when the frequency approaches the cut-on conditions (see, e.g.,
\cite{FA}).
The near and far field unsteady responses thus strongly affect each
other,
which emphasizes the importance of the correct formulation for the
inflow/outflow conditions. For $k_g=1,4.5,5.5,8.5,$ Figure \ref{bvp9}
shows
a set of eigenvalues defining the modes able to propagate in both
directions. For $k_g=1,$ the acoustic response includes two modes
propagating upstream and downstream, and a set of evanescent waves in the
upstream direction. A new acoustic cut-on occurs at $k_g\simeq 4.5,$ when
a
new {\it pair} of modes starts propagating upstream. This near-cut-on
situation is discussed in Appendix E in connection with ill-posed
upstream
boundary conditons. The pair of modes finally splits in the upstream and
downstream directions at $k_g\simeq 5.5.$ Another acoustic resonance
occurs
at $k_g\simeq 8.5$ when a new-born modal pair appears upstream in
Figure \ref{bvp9}d .
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/cp_u06.eps,height=3in,wi
dth=5in}}
\caption{Aerodynamic response at blade strips: --,$s=0.25$;
-,$s=0.5$; -.-,$s=0.75$. $k_{g}=1$ (a), 3 (b), 6 (c), 9 (d); $n_{g}=0$.
$M_0=0.6, \chi =0.$ }
\label{bvp10}
\end{figure}
The variation of pressure coefficient $C_p^s$ along the mid-span blade
chord
is shown in Figure \ref{bvp10} for the gust (\ref{B2}), for reduced
frequencies $k_g=1,3,6,9.$ These results will be compared below for the
effect of swirl on the cascade unsteady pressure response.
\subsubsection{\bf Swirling Flow}
The effect of a $45^0$-swirl on the unsteady aerodynamic response of the
annular cascade is investigated for two cases corresponding to potential
and
vortical swirling flows. At its initial position $x=-T_x$, an unsteady
vortical disturbance is imposed on a mean flow in the form of a
sinusoidal
gust,
\begin{equation}
{\bf u}_{-\infty }=[a_x(r),a_r(r),a_\theta (r)]\exp [i\{k_gx+n_g(\theta %
\frac{xM_s}{rM_0})-k_gM_0t\}] \label{B47}
\end{equation}
where the amplitudes are defined by (\ref{B2}) and (\ref{B3}). For
simplicity, we take $D_r=0$ and thus $a_r(r)=0.$
In terms of the linear cascade analogy in Figure \ref{bvp2}, one may
think
of a wake-induced convected velocity defect (\ref{B41}) imposed on a
uniform
inflow with $\nu =0$ and interacting with a rotor which induces a swirl
angle $\mu =\pi /4$. Unloaded rotor blades are thus installed at a
stagger
angle $\chi =\pi /4.$ Another analogy suggests a secondary flow - induced
gust (\ref{B43}) imposed on the same mean flow geometry; however, the
continuity condition (\ref{B33}) requires a radial gust component to
enter
the problem. Nevertheless, although {\it initially} $a_r(r)=0$ is
assumed,
the radial component will {\it develop} in a swirling flow as a results
of
the gust evolution downstream (discussed in Chapter 8), which makes the
secondary flow analogy reasonable.
The numerical results are compared below with the integral method
\cite{Vent}
for a linear cascade. Although the gust definitions in both cases are
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/eig_ugo003.eps,height=3i
n,width=5in}}
\caption{Eigenvalues for upstream/downstream unsteady response. $k_{g}=1$
(a), 3 (b), 6 (c), 9 (d); $n_{g}=0$. $M_0=0.03, \bar{M}_{\Gamma}=0.01,
\bar{M}_{\Omega}=0.02, \bar{\chi}=45^{o}.$ }
\label{bvp15}
\end{figure}
The variation of $C_p^s(x)$ along the span of a blade in Figures
\ref{bvp12}
and \ref{bvp13} for potential and vortical swirling flows, is more
significant than in the similar cases of a uniform flow (Figure
\ref{bvp6})
but still moderate compared to large variations for high-speed flows.
Note
that the integral method in fact compares well with the present results
as
far as general trends are concerned; the amplitudes are different due to
different gust normalization in (\ref{B46}). However, the linear cascade
results do not show oscillations along the chord for higher $k_g.$ Such
an
oscillatory behavior, less pronounced for vortical swirl, may at least
partially be attributed to the appearance of downstream evanescent modes
in
the eigenmode analysis of unsteady swirling flows. The wavenumbers
corresponding to propagating or decaying waves in the cases of potential
and
vortical mean swirl components, are shown in Figures \ref{bvp14} and \ref
{bvp15}, respectively. As for the case of no swirl, there are one
upstream
and one downstream propagating acoustic modes. In addition, the Doppler
shift causes infinite sets of evanescent ''pressure-dominated''
eigenmodes
to propagate in both directions, with all radial modes spinning in the
direction of mean swirl, decaying downstream, and those spinning in the
opposite direction, decaying upstream.
The difference between two cases appears in the nearly-convected region
of
eigenvalues. For the potential swirl, the ''vorticity-dominated''
eigenmodes
have {\it convected} double eigenvalues and do not contribute to the
unsteady pressure field far downstream. On the contrary, certain {\it %
nearly-convected} modes in the vortical swirl may produce an unsteady
pressure field which should be accounted in the downstream outflow
conditions. Such modes typically have eigenvalues which cluster far
enough
from the critical layer. A special numerical procedure selects such modes
and includes them in the series expansion (\ref{B29}) based on the
criterion
of pressure norm comparison for evanescent ''pressure-dominated'' and
\caption{Radial
variations of gust vortical velocity and pressure amplitudes: -, real
parts, --, imaginary parts, '.', absolute values. Top: initial station;
bottom: mid-chord station. $k_{g}=1, n_{g}=0$. $M_0=0.03,
\bar{M}_{\Gamma}=0.01,
\bar{M}_{\Omega}=0.02.$ }
\label{bvp18}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gr_ugo003_9.eps,height=3
in,width=5in}}
\caption{Radial
variations of gust vortical velocity and pressure amplitudes: -, real
parts, --, imaginary parts, '.', absolute values. Top: initial station;
bottom: mid-chord station. $k_{g}=9, n_{g}=0$. $M_0=0.03,
\bar{M}_{\Gamma}=0.01,
\bar{M}_{\Omega}=0.02.$ }
\label{bvp19}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gx_ug06_1.eps,height=3in
,width=5in}}
\caption{Downstream evolution
of real parts of gust vortical velocity and pressure amplitudes, at
radial strips: --, $s=0.25$, -, $s=0.5$, '.', $s=0.75$. $k_{g}=1,
n_{g}=0$. $M_0=0.03, \bar{M}_{\Gamma}=0.03.$ }
\label{bvp20}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gx_ug06_9.eps,height=3in
,width=5in}}
\caption{Downstream evolution
of real parts of gust vortical velocity and pressure amplitudes, at
radial strips: --, $s=0.25$, -, $s=0.5$, '.', $s=0.75$. $k_{g}=9,
n_{g}=0$. $M_0=0.03, \bar{M}_{\Gamma}=0.03.$ }
\label{bvp21}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gx_ugo003_1.eps,height=3
in,width=5in}}
\caption{Downstream evolution
of real parts of gust vortical velocity and pressure amplitudes, at
radial strips: --, $s=0.25$, -, $s=0.5$, '.', $s=0.75$. Left: total
values; right: amplifying components. $k_{g}=1, n_{g}=0$. $M_0=0.03,
\bar{M}_{\Gamma}=0.01, \bar{M}_{\Omega}=0.02.$ }
\label{bvp22}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gx_ugo003_9.eps,height=3
in,width=5in}}
\caption{Downstream evolution
of real parts of gust vortical velocity and pressure amplitudes, at
radial strips: --, $s=0.25$, -, $s=0.5$, '.', $s=0.75$. Left: total
values; right: amplifying components. $k_{g}=9, n_{g}=0$. $M_0=0.03,
\bar{M}_{\Gamma}=0.01, \bar{M}_{\Omega}=0.02.$ }
\label{bvp23}
\end{figure}
The major difference between the cases of potential and vortical flows is
noticed for the gust unsteady pressure evolution. At the mid-chord loci,
the
maximum pressure response for $k_g=1$ in the potential swirl is about two
times smaller than in the vortical swirl, and rapidly decreases for
higher
reduced frequencies. For the vortical swirling flow, such a response
stays
at about the same level for the range of $k_g=1...10,$ with a significant
contribution coming from amplifying nearly-convected modes, for lower
reduced frequencies.
For an incident 2--D\ non-axisymmetric acoustic-vorticity wave with $%
n_g=r_mk_g,$ a similar set of results is presented in Figures
\ref{bvp24}-%
\ref{bvp36}. The unsteady lift coefficients $|C_L^{0.5}|$ show a
significant
divergence for potential and vortical mean swirling flows for small
reduced
frequencies but compare well for $k_g\geq 5.$ Both the phase dependence
and
amplitudes of the curves are different from the linear cascade results,
although at least the amplitude deviation appears to decrease for higher
reduced frequencies. Among the possible reasons for these differences, we
mention the effects of a non-axisymmetric gust evolution in swirling
flows.
Recall that the circumeferential wavenumbers in the 2-D and 3-D cascade
models are not completely compatible, so that different cascade
quasiperiodicity conditions, along with a variable cascade spacing, may
induce a phase deviation for the lift coefficient curves. These results
are
further illustrated in Figures \ref{bvp25}, \ref{bvp26} which compare the
pressure jump coefficients for the potential and vortical mean flows. In
both cases, a large variation of $C_p^s(x)$ along the blade span is noted
at
the leading edge, which occurs only for small reduced frequencies.
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/cls003ng.eps,height=3in,
width=5in}}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gr_ugo003ng_1.eps,height
=3in,width=5in}}
\caption{Radial
variations of gust vortical velocity and pressure amplitudes: -, real
parts, --, imaginary parts, '.', absolute values. Top: initial station;
bottom: mid-chord station. $k_{g}=1, n_{g}=5$. $M_0=0.03,
\bar{M}_{\Gamma}=0.01,
\bar{M}_{\Omega}=0.02.$ }
\label{bvp31}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gr_ugo003ng_9.eps,height
=3in,width=5in}}
\caption{Radial
variations of gust vortical velocity and pressure amplitudes: -, real
parts, --, imaginary parts, '.', absolute values. Top: initial station;
bottom: mid-chord station. $k_{g}=9, n_{g}=45$. $M_0=0.03,
\bar{M}_{\Gamma}=0.01,
\bar{M}_{\Omega}=0.02.$ }
\label{bvp32}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gx_ug003ng_1.eps,height=
3in,width=5in}}
\caption{Downstream evolution
of real parts of gust vortical velocity and pressure amplitudes, at
radial strips: --, $s=0.25$, -, $s=0.5$, '.', $s=0.75$. $k_{g}=1,
n_{g}=5$. $M_0=0.03, \bar{M}_{\Gamma}=0.03.$ }
\label{bvp33}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gx_ug003ng_9.eps,height=
3in,width=5in}}
\caption{Downstream evolution
of real parts of gust vortical velocity and pressure amplitudes, at
radial strips: --, $s=0.25$, -, $s=0.5$, '.', $s=0.75$. $k_{g}=9,
n_{g}=45$. $M_0=0.03, \bar{M}_{\Gamma}=0.03.$ }
\label{bvp34}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gx_ugo003ng_1.eps,height
=3in,width=5in}}
\caption{Downstream evolution
of real parts of gust vortical velocity and pressure amplitudes, at
radial strips: --, $s=0.25$, -, $s=0.5$, '.', $s=0.75$. Left: total
values; right: amplifying components. $k_{g}=1, n_{g}=5$. $M_0=0.03,
\bar{M}_{\Gamma}=0.01, \bar{M}_{\Omega}=0.02.$ }
\label{bvp35}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/gx_ugo003ng_9.eps,height
=3in,width=5in}}
\caption{Downstream evolution
of real parts of gust vortical velocity and pressure amplitudes, at
radial strips: --, $s=0.25$, -, $s=0.5$, '.', $s=0.75$. Left: total
values; right: amplifying components. $k_{g}=9, n_{g}=45$. $M_0=0.03,
\bar{M}_{\Gamma}=0.01, \bar{M}_{\Omega}=0.02.$ }
\label{bvp36}
\end{figure}
In the vortical mean flow, the radial non-axisymmetric vortical velocity
component is non-zero at the duct walls. The impermeability condition is
provided by a corresponding component of the potential gust velocity (not
shown) which cancels the vortical counterpart. All the vortical velocity
components have comparable non-zero amplitudes at the mid-chord position,
so
that an initially 2-D vortical perturbation develops into a 3-D
acoustic-vorticity wave interacting with the blades. The\ resulting
complex
interference of phases and amplitudes of the gust components along the
blade
surfaces may induce an unsteady aerodynamic response which significantly
diverges from the linear cascade results, which was observed above.
\subsubsection{\bf High Mach numbers.}
In the last section of this dissertation, we examine the unsteady
aerodynamic response of a 3-D cascade interacting with axisymmetric 2-D
and
non-axisymmetric 3-D acoustic-vortcity waves in a high-speed swirling
flow
typical of turbomachinery applications. As before, two cases of potential
and vortical swirls are considered, with $M_0=\overline{M}_\Gamma =0.6$
and $%
M_0=0.6,\overline{M}_\Gamma =\overline{M}_\Omega =0.3,$ respectively.
For an axisymmetric impinging gust ($n_g=0$), the unsteady lift
coefficients
$|C_L^{0.5}|$ are compared with the uniform flow and 2-D cascade analyses
in
Figure \ref{bvp37}. All results show good agreement for low reduced
frequencies, up to $k_g\simeq 4.$ In the resonance region of frequencies,
the curves diverge significantly, but some unexpected trends can be
noticed.
The results for the potential swirl in fact remain close to those
obtained
for the uniform flow, and follow the single-peak behavior of the latter
over
the set of $k_g.$ Similarly, the case of the vortical flow happens to
match
the results of the integral method, but misses one peak. Moreover, both
curves occur to be in phase (compare with Figure \ref{bvp7}). Note that
for
higher reduced frequencies, all the analyses again show good comparison.
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/cls06.eps,height=3in,wid
th=5in}}
\caption{Unsteady lift coefficient vs. reduced frequency. Present
results: -, $M_0=0.6$ ($\chi =0$); $'+'$, $M_0=\bar{M}_{\Gamma}=0.6$
($\bar{\chi} =45^{o}$); 'x',
$M_0=0.6, \bar{M}_{\Gamma}=0.3,\bar{M}_{\Omega}=0.3 (\bar{\chi} =45^{o})
$. 2-D cascade analysis: 'o', $M_0=0.6, \chi =0$. $n_{g}=0.$ }
\label{bvp37}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/eig_ug06.eps,height=3in,
width=5in}}
\caption{Eigenvalues for upstream/downstream unsteady response. $k_{g}=1$
(a), 3 (b), 6 (c), 9 (d); $n_{g}=0$. $M_0=0.6, \bar{M}_{\Gamma}=0.6,
\bar{\chi}=45^{o}.$ }
\label{bvp38}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/eig_ugo06.eps,height=3in
,width=5in}}
\caption{Eigenvalues for upstream/downstream unsteady response. $k_{g}=1$
(a), 3 (b), 6 (c), 9 (d); $n_{g}=0$. $M_0=0.6, \bar{M}_{\Gamma}=0.3,
\bar{M}_{\Omega}=0.3, \bar{\chi}=45^{o}.$ }
\label{bvp39}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/cp_ug06.eps,height=3in,w
idth=5in}}
\caption{Aerodynamic response at blade strips: --,$s=0.25$;
-,$s=0.5$; -.-,$s=0.75$. $k_{g}=1$ (a), 3 (b), 6 (c), 9 (d); $n_{g}=0$.
$M_0=0.6, \bar{M}_{\Gamma}=0.6, \bar{\chi} =45^{o}.$ }
\label{bvp40}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/cp_ugo06.eps,height=3in,
width=5in}}
\caption{Aerodynamic response at blade strips: --,$s=0.25$;
-,$s=0.5$; -.-,$s=0.75$. $k_{g}=1$ (a), 3 (b), 6 (c), 9 (d); $n_{g}=0$.
$M_0=0.6, \bar{M}_{\Gamma}=0.3, \bar{M}_{\Omega}=0.3, \bar{\chi} =45^{o}.
$ }
\label{bvp41}
\end{figure}
damped since very few acoustic modes are excited in the duct; as a
result,
the unsteady aerodynamic loading concentrates at the leading edge as seen
in
Figures \ref{bvp12},\ref{bvp13}. In the present cases, the leading edge
response dominates only for small reduced frequencies. This also
corresponds
to the results for the 2-D cascade. In addition, since the impinging gust
in
a swirling flow exhibits large variations of axial and circumferential
phases of complex amplitudes over the span and chord of a blade, the
unsteady loading shows the corresponding pronounced variations in Figures
\ref{bvp40} and \ref{bvp41}.
If an incident non-axisymmetric acoustic-vorticity wave propagates with $
%
n_g=r_mk_g$, its circumferential phase remains almost constant (exactly
constant for a free vortex swirl) along the surface of a blade aligned
with
turning streamlines of the mean swirling flow. The unsteady response
predictions, however, are additionally affected by an axial gust
component,
and effects of cascade quasiperiodicity on blades interference and
inflow/outflow conditions.
Figure \ref{bvp42} compares the results for the unsteady lift predictions
in
the cases of swirling flows and 2-D cascade model. A significant
difference
from the corresponding low Mach number calculations (in Figure
\ref{bvp24})
is easily noticed. The resonance effect of cascade quasiperiodicity is
now
superimposed on the effects of multiple acoustic resonances in the duct.
As
before, the cases of potential and vortical swirling flows diverge for
low
reduced frequencies but show better agreement for higher $k_g.$
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/cls06ngnew.eps,height=3i
n,width=5in}}
\caption{Unsteady lift coefficient vs. reduced frequency. Present
results: $'+'$, $M_0=\bar{M}_{\Gamma}=0.6$; 'x',
$M_0=0.6, \bar{M}_{\Gamma}=0.3,\bar{M}_{\Omega}=0.3$.
2-D cascade analysis: 'o'. $\bar{\chi} =45^{o}, n_{g}=k_{g}r_{m}.$ }
\label{bvp42}
\end{figure}
\begin{figure}
\
centerline{\psfig{figure=../Gvan/FORT/BLADES/OUT/eig_ug06ng.eps,height=3i
n,width=5in}}
acoustic modes cuts on for the potential swirl but yet does not appear
for
the vortical. On the other hand, a total pair of modes may propagate
upstream in one case but splits in both directions, in another. These
indicated differences may cause the divergence of unsteady loading
characteristics in the two cases of mean swirl.
The gust evolution calculations in both cases are very similar to the
corresponding low-speed results presented in Figures
\ref{bvp29}-\ref{bvp32}
and \ref{bvp33}-\ref{bvp36}, by the reasons given above. The effect of
different $\overline{M}_\Omega /\overline{M}_\Gamma $ ratio on the gust
pressure response is slightly more pronounced in the present example but
still is not very significant.
Similar to the previous case of an axisymmetric disturbance, the same
incident wave produced a drastically different distribution of unsteady
loading along the chords and span of the blades (Figures \ref{bvp45},
\ref
{bvp46}). In general, the responses occured to be close, in the shape of
amplitude oscillations, to the previous example, but the phases and
amplitudes are clearly different as one would expected. For the free
vortex
swirl in Figure \ref{bvp45}, a sudden change of loading oscillations at
the
mid-span strip is notable.