>
(29)
where ( ) denotes time averaging over one period of harmonic motion. N,, denotes the x component of the definition of g,,. The condition of continuity of particle gsplacement at the boundary y = b takes the form vb = { + UC,, where < = c (x, t) is the wall displacement field. Equation (29) can be written as
Jbd 0
dx
W,,J dy = -
u
- &,rf>*
(30)
One can define a new axial power, which includes the hardwall axial power P,, =
J
b (N,,J
dy,
0
as
&,x= p,,=+ u
J
P&,xdx,
(31)
and then equation (30) becomes -d dx
+uJ
(32)
0
which is a time averaged continuity equation with a source term (actually a power sink). @,,is defined as the axial power consistent with the Type II energy. In the case of a Type I energy formulation the additional energy related to the lining is appended in the same way: I?, =
JJJ
$pv+ddz
+
JJJ$p2/pcz) JJJ dz +
(l/~“)(~~ * 7) p dz +
JJ JJ$
$p,[; ds
+
kbc2 ds.
(33)
This is justified because E, was identified as the Hamiltonian density and it would be desired to add addition%1 terms to the Hamiltonian. In particle mechanics the Hamiltonian coincides with the total energy and as equation (25) is in effect a particle mechanics equation of motion it is appropriate to append total energy terms to the Hamiltonian. When this is done, the development of a new conservation equation and the definition of a new axial
ACOUSTIC
ENERGY
527
IN DUCTS
power proceeds as above and yields
k,* = P,x +
U
(pbSx) dx + PU
(ubub) dx. s
s
It can be shown by use of the axial momentum equation that this is equivalent to
p,x =PIx + w$>
+ PW,i”)l.
134)
The axial power balance is then
P, is defined as the axial power consistent with the Type I energy. Equation (34) for p, is perhaps more appealing than equation (31) for $,, in that the integration required in the extra terms can be carried out explicitly and results in an expression in the local pressure, velocity and boundary displacement. Equations (32) and (35) show that if uniform mean flow is present there are additional terms to be added to the definitions of flux and power. When these terms are included the * axial rate of change of power is negative definite pb > 0) and terms with p, or P,, positive decay in the positive x direction and those with P, or f’,, negative decay in the negative Y direction. From equation (32) and equation (35) it follows that d&dx
= d&,,;dx:
that is, 6, and @,,could at most differ by a constant. independent of x, which would be irrelevant. Whether p, and p,, actually are equal depends on the constants which are included in the extra integral terms which add to P, and P,,. An appropriate choice, adopted here, is to set the constants equal to zero, or alternatively, view 8,X and P,,= as those parts of the definitions which vary with x. For any useful application it is only this part which is of any significance. 4. COMPUTATIONAL
FORMULAE
Computational formulae for intensity and power in the two dimensional case have appeared in various forms in the literature [2, 4-61. In this section formulae for P, and P,, * * are given in the present notation as well as the modified P, and P,,. While in previous sections dimensional equations were used, the computational formulae given here are in terms of non-dimensional acoustic perturbations. In two dimensional ducts the solution for acoustic propagation in terms of pressure and axial and transverse particle velocity is, in vector-matrix form. p(x. J‘,t) = e’“”Ien cos 7cn4.]{p}, u(x, y, t) = e”” [ena,, cos K~JJ jpl, u(x, y, t) = e”‘l [e,fl,I sin
K~J‘]
{p).
(M--38)
where row matrices are defined by fP)?‘=
p,.
pp.
.
&j.
[en COS K,,.v]
=
[e,
COS Xl)‘.
, 02y
COS K2,v~j
and e,, = exp( - ik_x), cl,, = (kJk)/(l
- MkJk),
/j,, = - i(ic,Jk)/(1 - Mkx,,: kj
The modes of propagation could conveniently be ordered so that modes 1 to N are decaying for positive x and modes N + 1 to 2N are decaying for negative x. In equations
528
W. EVERSMAN
(36H38) the acoustic perturbations are non-dimensional, p with respect to pc2 and the velocities with respect to c. The mean flow Mach number is denoted by M. From the boundary kinematic relation (ur,is the non-dimensional fluid transverse particle velocity at L’= b) cub = [* + ciV& It foil ows that in each mode the boundary displacement and velocity are related to the fluid velocity at the boundary by i,,(x, t) = (llik)~bn/(l - MkJk), Note that [, the boundary lead to
i,,(x, t) = -@,,v~,,,i,,(x, 0 = c~J1
- MkJk).
deflection remains dimensional, as does time. These relations
i(x. t) = eiof[enmPn sin ic,bJ {P>,5,(x, t) = eiorr - enaJn sin ic,bJ {p}, i,(x, t) = c eiwf[e,dJn sin icnbJ {p},
(3941)
where cS~= l/(1 - MkX,/k), y, = G,,/ik. Expressions for the wall displacement are also available through the impedance condition itn = c&b,,? where pb is the fluid pressure at the wall. This yields the alternate expressions 5(x, t) = (A/ik) eiot[e,, cos ~,b] {p}, c,(x, t) = A eiotr - (kX,/k) e,, cos IC,,~](p}, [,(x,
t)
CA eiot[en cos IC,~] {p}.
=
(42-44)
The eigenvalues K,,and kXnare defined by Kb tan rcb = ikbA(l
- MkX/k)2, k,/k = [l/(1 - M’)] [ - M + 41 - (1 - M2)(rc/k)2]. (45746)
A is the wall admittance, non-dimensional with respect to l/pc. Equations (42)-(44) are consistent with equations (39)-(41) through the eigenvalue equation equation (45). By using the real parts of the acoustic disturbances in equations (2) and (4) there results, in terms of non-dimensional perturbations,
4(N,J/pc3
= (1 + M2)(pu*
4(NIIX)/pc3
+ p*u) + 2M[uu* + pp*],
= pu* + p*u + M [w*
(47)
+ vu* + pp*].
(48)
The superscripted asterisk denotes the complex conjugate. If equations (36H38) are used in equations (47) and (48) the following relationships result: 4P,lpc3 = {pjT[(l 4P,,lpc3
+ M’)[Al
+ 2M PI]
(P*>,
(49)
= {P)~[CAI + M Ccl] {P*>,
(50)
where b An,,,
=
(a,,
+
a:)&,
exp
(-@,,
-
k:Jx),
Inm
cos
=
xny
cos
FC;L’
dy,
s0 Bm
=
(1
+
a,&)Z,, exp( - i(kX, - kzm)x),
[C] = [B] + [D], D,,,,, = /?np*J m nmexp( -i(kX, - kfQx), J,,, =
b sin K,J sin lczq’dy. s0 It can be verified that Tester’s expression [6] for the individual powers of Type I in each mode, P,,,, results from equation (49): that is,
k12
“n = 2 11 - MkX,/k12
k M + (1 - M2) Re 2
(
)I
I,,,.
(51)
ACOUSTIC
ENERGY
529
IN DUCTS
The equivalent Type II result is, from equation (50)
+ M
II2 ’ 1. J,!,,!.
(52)
I The exponential term is suppressed in equations (51) and (52) or, alternatively. included in the definition of the amplitude, p,. Note that P, and P,,. the acoustic powers, when divided by pc3/4 are not wholly non-dimensional as there is no scaling of lengths in the analysis. The resulting dimension is that of duct area (a unit depth is assumed). The additional terms in the power flux are now expanded. Define in non-dimensional terms
s
k,, = % +PC3M@,i,>dx = P,,x + $ &x = P,
+ pc3M
(p,(,)
M
dx + pc3M
s =Plx + f
(p,(:
(u~II,,)
+ p,*[,) d.y.
dx
s
M
(p&I,* + p,*[,)dx +$ M (u,,u; + u,*u,) dx. s s From equations (36) and (40) it can be determined that
I
+ p,*i,)dx = (P&x*
where Hnm =
--l-
kx, - kx*,
{p}'[H]{p)
(M,/?,sin rc,,bcos KEb + cczfl: cos i$,b sin IcEb)exp ( - i(kX,,- k,*,,)u).
(53)
If equation (43) is used this also can be written as Hmn =
--I A* % k& - k:_ (
cos icnbcos rczb + A $
cos Knbcos rczb exp( -i(k,,
- k:,,).x)
j (54)
It is apparent that if A = 0, the hard wall case, Hnm = 0. When real eigenvalues occur Hnm becomes singular for n = wt.A form of equation (53) or (54) can be derived to circumvent this by returning to the original integral definitions. These formulae are not given explicitly here. Alternatively, by defining the integral to have a lower limit x = 0, a constant can be introduced into equation (54) which makes evaluation of Hnm for real eigenvalues a straightforward limiting process. This would have the interesting effect of making Hqm(x = 0) = 0 and thus p,(.u = 0) = P,(x = 0), although the decay rate dp,/dx(x = 0) is not equal to dP,/dx (x = 0). The present choice of constant (zero) amounts to taking a lower limit x = f co (kX, assumed to have at least a small imaginary part) and has the salutary effect of producing a power which is of the proper sign for a given direction of propagation. This choice and a similar one in constructing the p,, terms also ensures that p, = I’,, (equation (55) below). By proceeding in the same way it is established that
s
(PJ:
+ p,*i,) dx +
s
by,*
+ ++,,
dx = {P)’ [f’l{p*l, .-
W. EVERSMAN
530
where pnnI = Hnm + k
.x.
: kzm(cc*P:cos rC,b sin Kzb + ~:a, sin rC,bcos $b)
x
x exp (- i(kX, - kzm)x).
Equation (45) allows this to be written as i p,,,,, = i
A* (, _ Mkx,,,)
A ~0s K,b ~0s C,b - c1 _ Mkznlk,
~0s ~,b ~0s Gb
1 x
x exp ( - i(kX, - kz”)x).
(55)
In the hard walled case P,,,,, = 0. Finally, the computational formulae are 4@pc3 = (P>’ [(l + M’)[A]
+ 2M[B] + M[P]] {P*},
43,J~c~ = {PI’ [L-AI f MCCI +
(56)
Wffl] @*I.
(57)
The individual powers of Type I and Type II. including the additional terms to account for wall impedance now follow from equations (56) and (57) with n = m:
4P,12
M + (1 - M’) Re
Inn
nn = (1 - MkJk(’ & Mlp,Y Re{(a,, - cr,*)/?Xcos lc,b sin rc:b), fm(kXn) lP,12
(58)
I II k
M + 2(1 - M2)Re
- M(l - M2) 2
nn = 11 - Mk,Jk12
2
Inn
MlpI2Re{G[~P~cos K,,b sin K,*b}. Wkxn)
- n
(59)
Again, the exponential term is suppressed in equations (58) and (59) or included in the definition of the amplitude p,. The axial rate of change of power dfi,/d.x or d?,,/dx is easily computed from equations (56) and (57) by noting that the x dependence is contained in the exponential factors in the coefficient matrices. Thus, for example, & [A] = [A’], A;, = -i(kXn - kzm)A,,,. Similar relations obtain for the other coefficient matrices. The dissipation term, W, is defined by W = (r,C:> = (rJ2) This leads to the vector-matrix
[i,C:l.
formulation 4WlP3
= {PI ‘cwP*L
where R,,,,, = 2?$ nmnm S*p fi* exp( - i(k,, - kzm)x) sin rC,b sin Iczb,
where P,,is the real part of Z/PC (non-dimensional).
(60)
The dissipation in the individual modes
ACOUSTIC
ENERGY
IN DUCTS
531
is thus
with the exponential suppressed or included in p,. Tables 1 through 4 show, in addition to calculations of power based on the Morfey and Ryshov and Shefter relations, the calculations of axial power in individual modes based on the new relations developed here (equations (58) and (59)). The power gradients and the dissipation term are also shown. It is apparent that PI and p,, coincide and that the sign of the axial power is consistent with the propagation direction: that is, positively propagating modes (attenuated for positive x) have positive power and the reverse for negatively propagating modes. It is also noted that the axial gradient of power equals the dissipation term. as stipulated by the time averaged continuity equation. Because of the factor k ‘, it would appear from equation (55) that at higher frequencies the correction P,,,, to P, is likely to be small. For well cut-on modes, ones for which P, is relatively large, Pnmis expected to be of only minor importance. This does in fact happen as shown in Table 4 where the reduced frequency kb = 6.0. Note, however, that for well cut-off modes moving with the flow the correction still dominates and the sign of p, and P, are different. 5. OBSERVATIONS Tester [6] used Mohring’s [2] definition of intensity, specialized to sheared flow in a two dimensional duct, to explain the deficiencies of the Morfey and Ryshov and Shefter definitions of intensity when the duct wall is soft. This was accomplished by treating the uniform flow as one with a vanishingly thin shear layer at the wall. When suitable modification is made to account for the fact that Tester’s result is dimensional, the additional power
in equation (56) with [P] defined by equation (55) can be shown to yield the identical diagonal terms defined by his equation (22). This is done by using the conjugate of the expression in brackets in equation (55) and by introducing the magnitude squared of the term (1 - MkJk) in the denominator. Equation (31) with the extra power term appearing in integral form, may be open to criticism as depending on an acoustic fluctuation product integrated with respect to the axial co-ordinate. The extra terms in equation (34) also were originally arrived at in the same way but were representable in terms of an explicit integration yielding results in terms of local pressure, velocity, and lining displacement. In either case, when the computational expressions of equations (56) and (57) are obtained the power is in terms of local pressure fluctuations. The acoustic power considerations addressed here verify the results of Tester for the diagonal terms and extend the implications to include power conservation laws which show the effect of the wall as a power sink. The results in this analysis can be viewed as having been derived from relatively conventional energy conservation concepts with no necessity to consider a wall shear layer. In the Tester approach, via Mohring’s intensity definition, the shear layer is essential. Miihring’s [2] equation (15) does not reduce to the uniform flow result (with soft walls) unless the proper limiting behaviour across the shear layer is accounted for. 6. CONCLUSION It
has been shown that both of the common acoustic energy density-energy
flu\- relation-
532
W. EVERSMAN
ships, of which Morfey [l] and Ryshov and Shefter [3] are the principal proponents, are related to the governing field equation for irrotational acoustic disturbances on a uniform duct flow by a variational principle. The Ryshov and Shefter energy density can be viewed as the sum of a potential energy density and a kinetic energy density from which the Lagrangian density can be formed. Hamilton’s Principle then yields the convected wave equation. The Morfey form of the energy density has been shown to be the Hamiltonian density, related to the Lagrangian density by a Legendre transformation. The modified form of Hamilton’s Principle then yields the convected wave equation. The concept of energy flux in the Morfey representation follows naturally from the definition of the Hamiltonian density. The problem asociated with the use of either the Morfey or Ryshov and Shefter energy formulations in the case of ducts with soft walls has been addressed and a new energy equation including an energy sink due to the wall dissipation has been derived. The acoustic flux, and hence the acoustic power, associated with the modified energy equation is consistent in sign with the direction of propagation of acoustic disturbances as deduced from attenuation arguments. ACKNOWLEDGMENT
The research reported here was supported by the National Aeronautics and Space Administration Lewis Research Center under Grant NSG 7192. The author wishes to thank the reviewers for particularly useful comments. Several significant shortcomings in presentation and content were brought to light and these have been modified, in certain cases by using ideas proposed by the reviewers. REFERENCES
1. C. L. MORFEY1971 Journal of Sound and Vibration 14, 1.59-170. Acoustic energy in non-uniform flows. 2. W. M~HRING 1971 Journal of Sound and Vibration 18, 101-109. Energy flux in duct flow. 3. 0. S. RYSHOVand G. M. SHEFTJZR1962 Journal of Applied Mathematics and Mechanics 26, 1293-1309. On the energy of acoustic waves propagating in moving media. 4. W. EVERSMAN1970 Journal of the Acoustical Society of America 49, 1717-1721. Energy flow criteria for acoustic propagation in ducts with flow. 5. S. H. Ko 1971 Journal of the Acoustical Society of America SO, 14181432. Sound attenuation in lined rectangular ducts with flow and its application to the reduction of aircraft noise. 6. B. J. TESTER1973 Journal of Soundand Vibration 28,205-215. Acoustic energy Bow in lined ducts containing uniform or “plug” flow. 7. S. M. CANDEL 1975 Journal of Sound and Vibration 41, 207-232. Acoustic conservation principles and an application to plane and modal propagation in nozzles and diffusers. 8. W. EVERSMAN1970 Boeing Company Document 03-8152. The propagation of acoustic energy in a flow duct. 9. H. GOLDSTEIN1959 Classical Mechanics. New York: Addison Wesley Publishing Company. Chapters 7 and 11. 10. P. M. MORSEand K. U. INGARD1968 Theoretical Acoustics. New York: McGraw-Hill Book Co. See pages 100-106, 196-200 and 248-250. 11. W. M~~HRING1976 Fortschritte der Akustik, DAGA 76, VDI-Verlag, Dusseldorf, 543-546. Uber Schallwellen in Scherstromungen.