INVERSE PROBLEMS IN UNSTEADY AERODYNAMICS AND AEROACOUSTICS
y z
Sheryl M. Patrick and Ha z M. Atassi
Aerospace and Mechanical Engineering Department
University of Notre Dame
Notre Dame, Indiana
and
x
William K. Blake
David Taylor Mo del Basin
C.D./N.S.W.C.
Bethesda, Maryland
ABSTRACT duce uctuating pressure forces on the structural com-
p onents which leads to vibration and noise.
The feasibility of determining an unknown vortical
disturbance in an approach ow to a streamlined solid In order to either reduce or eliminate suchunwanted
b o dy from the radiated sound is studied. The problem e ects, the aero dynamics and aeroacoustics of the prob-
is treated in two separate parts. First, the inverse aero- lem need to be understo o d. This entails determining
dynamic problem, wherein the upstream disturbance which are the key parameters resulting in ecient or
is determined from unsteady surface pressure, is con- nonecient sources of vibration and noise and optimiz-
sidered. It is shown that the transverse comp onentof ing them for vibration and noise control.
an incoming vortical disturbance can b e uniquely pre-
To this end, the study of the aero dynamic prob-
dicted from the surface unsteady pressure. This result
lem asso ciated with impinging ow nonuniformities on
can b e extended for incoming turbulence in terms of the
a streamlined b o dy such as wings or blades, known as
surface pressure density sp ectrum. The second problem
the gust problem, has b een the fo cus of considerable in-
treated is the inverse acoustic problem, wherein the un-
terest in unsteady aero dynamics. The direct gust prob-
steady surface pressure is determined in terms of the
lem is to calculate the unsteady pressure distribution
radiated sound. It is shown that for a single frequency,
along the airfoil/blade surface knowing the upstream
the governing equation reduces to the Helmholtz equa-
disturbances and the shap e of the b o dy. The rst treat-
tion, and the solution to the inverse acoustic problem is
ment of this problem in incompressible owwas given
unique. However, the problem remains ill-p osed and ex-
by Sears 1941. A recent review on the sub ject was
hibits extreme sensitivity to noise in the far- eld data.
given byAtassi 1994.
The singular value decomp osition metho d and a simple
Another e ect asso ciated with the interaction of a
regularization metho d are used to solve the discretized
nonuniform ow and a solid body is the radiation of
system of equations resulting from the inverse acoustic
acoustic sound. Atlow Machnumb er, the direct acous-
problem. The simple regularization metho d is, however,
tic radiation problem can be treated using Lighthill's
more accurate.
acoustic analogy Lighthill, 1952 and the Ffowcs Wil-
liams-Hawkings equations Ffowcs Williams and Hawk-
INTRODUCTION
ings, 1969. In this treatment the unsteady pressure
Aircraft wings and propulsive system structural com-
along the airfoil/blade surface is equivalent to a dip ole
p onents often op erate in nonuniform ow conditions.
distribution. More recent analyses by Amiet 1976 and
The nonuniformities arise from atmospheric and inlet
Atassi et al. 1990 derive the far- eld sound directly
turbulence, viscous wakes, secondary ows, and rotor/
from the unsteady aero dynamic solutions. These treat-
stator interaction. Upstream ow nonuniformities pro-
ments consider dip ole and quadrup ole e ects. The lat-
ter b ecome signi cant as the mean- ow Machnumber
y
Fellow, Center for Applied Mathematics
and the reduced frequency increase and as the mean
z
Professor
airfoil/blade loading increases Atassi et al., 1993.
x
Research Scientist 1
ularization scheme. The Tikhonovscheme can b e used Another imp ortant asp ect of active control of vibra-
alone to obtain a solution or in conjunction with the tion and noise is sensing and analyzing data to predict
single value decomp osition to alleviate the dep endence their sources. This is the fo cus of the work presented
on very small singular values. in this pap er. Here, the feasibility of inverse aero dy-
namic and aeroacoustic problems is considered. The
The inverse problem discussed in this pap er, dif-
inverse problem allows for new metho ds of determining
fers from the inverse scattering problem investigated
the origins of unwanted vibration and noise. For the
by those mentioned ab ove. Here the incident eld is
aero dynamic problem, the pap er studies the feasibil-
unknown, although it is assumed to b e either vortical
ity of determining the upstream disturbances from the
or acoustic. The shap e of the b o dy is known, however
uctuating unsteady pressure on a moving streamlined
the unsteady pressure along the b o dy is not given. De-
body. The aeroacoustic problem concerns the feasibil-
termining the b oundary condition on the b o dy, i.e. the
ity of determining the surface unsteady pressure from
unsteady pressure, from far- eld measurements will b e
the radiated far eld. Ultimately b oth inverse problems
referred to as the inverse acoustic problem, while de-
can b e combined to analyze the upstream disturbances
termining the incident disturbance from the b oundary
from the radiated sound.
conditions on the b o dy will b e termed the inverse gust
problem. In spite of the imp ortance of these inverse problems
to applications and, in particular, to control and sens-
The problem of determining unknown b oundary con-
ing, they have received much less attention than the
ditions for a given ob ject from far- eld measurements
direct problem. Like most inverse problems, these are
has b een extensively investigated in conjunction with
ill-p osed in the Hadamard sense; that is, a problem is
acoustic holographyVeronesi and Maynard, 1989, Kim
ill-p osed if one of the following criteria fails: a solution
and Lee, 1990, Sarkissian et al., 1993. These problems
exists, it is unique, it depends continuously on the data
di er from the present problem in that they do not con-
Hadamard, 1923. These issues will be considered in
sider a mean owvelo city.
regard to the present application.
Rayleigh, in 1896, showed that a source of sound
THE DIRECT PROBLEM
could not be determined uniquely from the far- eld
In order to gain a fundamental understanding of is-
sound Rayleigh, 1896. This statement is true for a
sues involved with our inverse problems, we restrict our-
general sound source whose propagation is governed by
selves to a at-plate airfoil in subsonic ow with vortical
the wave equation. For sound which is pro duced by
disturbances imp osed upstream. For the at-plate air-
a disturbance which is harmonic in time, however, the
foil, the mean ow remains uniform. The formulation
governing equation reduces to the Helmholtz equation
for the direct problem of a at-plate airfoil in a subsonic
and inverse solutions to the Helmholtz equation may
ow with convected vortical gusts is given in Atassi et
b e unique. The inversion of problems governed by the
al., 1990, where the unsteady pressure jump across the
Helmholtz equation has b een studied quite extensively
at plate is calculated using an integral equation and
in recent years. This recent interest can be traced to
then the scattered eld is calculated using Green's the-
an article by Kac 1966 entitled, \Can One Hear the
orem. The formulation is summarized here for clarity
Shap e of a Drum?". The emphasis in inverse acous-
and for reference.
tic scattering problems has b een on surface reconstruc-
tion. Many researchers have attempted to use the far-
The Gust Direct Problem
eld acoustics, pro duced by the interaction of a known
Assume the owisinviscid, and non-heat conduct-
acoustic wave and an unknown body, to reconstruct the
ing. When the upstream disturbance is small, i.e. the
b o dy surface This research has b een motivated by ap-
owisweakly rotational, the velo city, pressure and den-
plications in tomography, geophysics, optics and sonar.
sity can b e linearized ab out their mean values Atassi,
It has led to the establishmentofseveral theorems p er-
1994.
taining to the existence and uniqueness of solutions to
b oth direct and inverse scattering problems governed by
the Helmholtz equation. Researchers studying inverse
~ ~
acoustic scattering have proven that a radiating solu-
U ~x; t = U + ~u~x; t 1
1
tion to the Helmholtz equation can b e obtained uniquely
0
p~x; t = p ~x+p ~x; t 2
from the far- eld data, provided a radiating solution ex-
0
ists Colton and Kress, 1992.
0
~x; t = ~x+ ~x; t 3
0
Even though the theorems guarantee uniqueness
when a solution exists, the diculty remains in prac-
In the absence of upstream incident acoustic waves,
tice to nd this solution. In many cases, the formulation
the upstream ow can b e written as
leads to a Fredholm integral equation of the rst kind.
To solve this integral equation, usually, one has to invert
~
~ ~
an ill-conditioned matrix. Hence, existing metho ds for
U ~x; t=U i + ~u ~x i U t 4
1 1 1 1 1 1
solving very ill-conditioned matrices have b een further
where ~u represents the upstream rotational distur-
explored and new techniques have b een found Kress,
1
~
1989. The most widely used metho d found in the
bance in the ow and i is a unit vector. The axis and
1
literature is the singular value decomp osition metho d.
ow direction are shown in Figure 1.
Another metho d commonly used is the Tikhonov reg- 2 a
longitudinal 1
~
ik t k ~x
1
x ; 0;x = a e for 1 2 u a2 1 3 2 1 k 2 Along the vortex sheet extending downstream from the trailing edge, u is discontinuous, but 1 x a 2 2 0 p = 0 k 4 1 x a 1 u = 0 x > 1 and x =0 10 2 4 a2 1 2 x 3 transverse 0 0 0 where 4p signi es p p . + U 8 In the direct problem, the vortical disturbance is known, i.e. 9 is given, and then the unsteady pressure |a| flat plate wake on the surface is found by solving the b oundary value oblique problem 6-10. or a transverse gust the pressure jump on a at- c F wwas given by Sears 2 plate airfoil in incompressible o k 3 as: r 1 x 1 Figure 1: Flat plate in ow with imp osed vortical gusts ik t 0 1 a k S k e 4p x ;k =2 U 2 1 1 1 1 o 1 1+ x 1 11 From the splitting theorem Goldstein and Atassi, where 1976, Atassi, 1994 the total unsteady disturbance ve- lo city is split into an acoustic, irrotational part ~u , and 1 a 12 S k = 1 a vortical, rotational part ~u which is convected by 1 2 2 k k iH k H 1 1 1 1 0 the mean ow. Since the problem is linear, ~u can b e 2 1 broken down into its Fourier comp onents and a single is the Sears function Fourier comp onent can be considered without loss of generality. So that The Acoustic Direct Problem ~ i!t k ~x Once the unsteady pressure on the at-plate is known, ~ ~ i + ~ae + ~u ~x; t 5 U ~x; t=U 1 a 1 the scattered eld can be determined using a direct calculation based on Green's theorem. The governing ~ where ~a =a ;a ;a is the amplitude vector and k = 1 2 3 equation for the far- eld pressure radiated from an air- k ;k ;k is the wave number vector describing ~u . 1 2 3 1 foil in compressible ow with a three-dimensional gust Since ~u is convected by the mean ow, ! = k U . In 1 1 1 ik x 3 3 such that the dep endence on x is given by e is 3 this formulation, lengths are normalized with resp ect develop ed as follows: to the half chord, c=2, velo cities with resp ect to U , 1 and time with resp ect to c=2U . Thus the reduced Taking the material derivative of 6, subtracting 1 frequency is k = !c=2U . ~u is the p otential ow dis- the divergence of 7, and using the isentropic relation- 1 1 a ~ 0 2 ik t k ~x 1 , c is the sp eed of sound, the governing ship p = c 0 0 turbance that results from the interaction of ~ae 0 equation b ecomes with the airfoil. 2 The linearized Euler's equations b ecome D 1 0 2 0