You are on page 1of 9

Why Do Mortices Generate Sound?

Alan Powell
Department of Mechanical Engineering, University of Houston, Houston, IX 77204-4792

Emphasizing physical pictures with a minimum of analysis, an introductory account is presented as to how vortices generate sound. Based on the observation that a vortex ring induces the same hydrodynamic (incompressible) flow as does a dipole sheet of the same shape, simple physical arguments for sound generation by vorticity are presented, first in terms of moving vortex rings of fixed strength and then of fixed rings of variable strength. These lead to the formal results of the theoiy of vortex sound, with the source expressed in terms of the vortex force p(u A0 and of the form introduced by Mohring in terms of the vortex moment (yA')> (p is the constant fluid density, u the flow velocity, f = VAu the vorticity and y is the flow coordinate). The simple "Contiguous Method" of finding the contiguous acoustic field surrounding an acoustically compact hydrodynamic (incompressible) field is also discussed. Some very simple vortex flows illustrate the various ideas. These are all for acoustically compact, low Mach number flows of an inviscid fluid, except that a simple argument for the effect of viscous dissipation is given and its relevance to the "dilatation"of a vortex is mentioned.

Introduction I am pleased to be invited to give this talk, as it is perhaps useful and timely to bring together the various physical arguments that bear on the mechanism by which vortices generate sound, some of them dating back to the origins of the original vortex theory and some much more recent. The intent is to present plausible physical pictures, briefly and with the minimum of mathematics (at the risk of short cuts being somewhat oversimplified in some spots), assuming of the audience only a general background in acoustics, though I hasten to add that the results do agree with those of rigorous and generally accepted theory. The interpretations given are those with which I am most familiar and are largely based on illustrations to introduce or to elucidate the various theoretical treatments in my aeroacoustics courses, originating in the late 1950's, now being updated for the current series. Consequently, the account inevitably has a personal touch and only a few pertinent references are given: there is no attempt to be all-inclusive and it is certainly not to be construed as a review of aerodynamic sound theory or even of vortex theories (for, apart from anything else, a good deal more sophistication would be unavoidable). In keeping with simplicity, the sound-generating flows are assumed to be acoustically compact (dimensions small compared to the acoustic wavelength), inviscid (except for the discussion on dissipation), low Mach number (virtually incompressible) with there being no acoustic back reaction on the flow. As is now very well known, Lighthill (1951) broke new

ground in providing the first mathematically soundand very powerfulapproach for aerodynamic sound generation. In it, the inhomogeneous wave equation for the pressure perturbation can be written to good approximation as 2P = ~ d2( pUiU^/dyidyj,
2 2

(1)

where is the d'Alembertian wave operator [d /d (l/c2)d2/dt2], c is the speed of sound, p is the fluid density and ut is the velocity in the /-direction (i = 1, 2 or 3, with the summation convention that any repeated suffix in a term is to be summed). With the source term on the right taken as for a given incompressible flow, the application of the divergence theorem twice to get rid of the derivatives in the standard solution for a free flow (no boundaries present) gives the quadrupole sound pressure in the far field as

/> =
1
ATTXC2 dt2

dt

jpuiUj

dV(y)*

fPu2xdV(y)\

(2)

where fit)* = fit*) = fit x/c) indicates the retarded time and ux = UjXi/x is the velocity in the x-direction, the last expression being Proudman's form, (1952). Then the sum of the sound pressures at three equidistant points on any set of three mutually perpendicular axes at a, b and c must be given by p(a) + p(b) + p(c)

Contributed by the Design Engineering Division for publication in the Special 50th Anniversary Design Issue. This article is based on an invited lecture delivered to the Symposium on Acoustic Radiation and Wave Propagation (see NCA-Vol. 17, pp. 1-8), Noise Control and Acoustics Division of ASME at the 1994 Winter Annual meeting in Chicago, IL. Manuscript received September 1994; revised January 1995.

1
ATTXC
2

d*
dt1

fp{u2a + u2b + u2)dV(y)*

= 0,

(3)

zero because the integral is twice the kinetic energy of the flow that must remain constant in the absence of viscous Transactions of the ASME

252/ Vol. 117, JUNE 1995

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Copyright 1995 by ASME

dissipation. As lateral quadrupoles, but not longitudinal ones, satisfy this requirement, the sources must be reducible to the former type. This is the "Three-Sound-Pressures Theorem," (Powell, 1962): it will be invoked later. An imaginary sphere drawn in the fluid centered on a lateral quadrupole undergoes oscillations with a pattern like quarters of an orange, neighboring quarters having opposite phase, there being no change of volume (zero monopole strength) and no movement of its centroid (zero dipole strength). One application is to free flows with complete symmetry about an axis, say the 1-axis. A lateral quadrupole with its axes at, say, 45 deg to the 1- and 2-axes is exactly equivalent to a pair of longitudinal quadrupoles of opposite signs (and half the amplitude) aligned along the 1- and 2-axes. Then the only permissible combination of lateral quadrupoles that is axisymmetric about the 1-axis consists of an equal pair, one with longitudinal equivalents along the 1- and 2-axes and other along 1- and 3-axes. Writing [xl] for the angle of direction x relative to the 1-axis, the far field sound pressure is then proportional to {cos [xl] - cos [x2]} + {cos [xl] - cos [x3]} = 3cos 2 [xl] - 1. Relative to unit amplitude along the 1-axis, the amplitude at angle [xl] to the 1-axis must be p(x) = ( l / 2 ) ( 3 c o s 2 [ x l ] - l ) . (4)
2 2 2 2

being to observe that Eq. (1) can be simply rewritten as a 2p = -pd/dy^u^ + d/dy^u1/!)) (8)

where f7 = dujdy^ du1/dyi is the vorticity, a relationship first used in Powell (1961b), with a different treatment in P64b. The last term of equation (8) contributes nothing to the far field, as shown rigorously by Howe (1975) in a more general development. A useful alternative in P64 is to write

a2/7 = P v-{(u A 0-v( 2 /2)}


where the vorticity is defined as usual as f = curlu = V A u.

.00
(10)

The divergent source term results in a basic dipole field. Here is a very fundamental result: No vorticityno sound! To emphasize that the vorticity can be considered to "induce" both the hydrodynamic (incompressible) flow field and the acoustic far field, the original theory in P64b was developed in terms of the velocity perturbations instead of the pressure perturbations of the more easily handled Eq. (9). But because for incompressible flow the velocity is much easier to handle mathematically, and to visualize, than is the pressure, in this account the velocity will be mainly used again. A totally different point of view, not at all dependent on LighthiU's approach, arose from imagining the deformation of an imaginary surface drawn in the fluid (and moving with it) enclosing the sound-generating flow. The perturbations of this envelope could be considered to force the contiguous acoustic field external to it. Then, in principle, the Kirchhoff-Helmholtz integral theorem could be used to find the contiguous acoustic field, needing the normal velocity and pressure at all points on the envelope. This approach seems manageable only if the enclosing surface is acoustically compact, when it has been used (Powell, 1963). But if the order of the multipole source is known, for example by its directionality, or that it is a lateral quadrupole by the ThreeSound-Pressures Theorem, then the appropriate point source formula enables one to write down immediately the extension into the contiguous acoustic field, only the normal velocity or the pressure needing to be known (Powell, 1963, 1964a). For convenience we dub this the "Contiguous Method." It was first used to obtain the same result as for the original vortex theory in terms of (u A f ) , (Powell, 1964c). A more rigorous method of the matching "inner" and "outer" fields is the much more mathematical Method of Asymptotic Expansions that followed, Obermeier (1967) applying the (same) result to a two-dimensional pair of spinning vortices. Starting from the same inhomogeneous wave Eq. (9) in terms of (u A f ) , Mohring (1978) used a sophisticated technique of switching the Green's function from scalar to vector formthat is unfortunately hard to interpret physicallyto yield results in which ( y A f ) replaces ( u A f ) , where ( - ) ' = d( )/dt. Here we obtain these versions in a particularly simple and straightforward manner, using the Contiguous Method for the quadrupole form. Finally, there is a simple physical argument for the effect of dissipation (Powell, 1991) that agrees with Kambe's more rigorous development (1984). A recent reformulation (Powell, 1993) of LighthiU's source term is also given. Together they cast some light on how the formation of real (viscous) vortices produce sound. 2 Vortex Rings and Dipoles The key to the theory of vortex generated sound is the recognition that hydrodynamic theory shows that the flow

Any axisymmetric inviscid free flow must have such axisymmetric quadrupoles and those alone (Powell, 1991): we make use of this result later. The "dilatation" or "simple source"concept was suggested, with somewhat different derivations, by Meecham and Ford (1958) and Corcos (1958) and by Ribner (1959). One way is to write the counterpart to Eq. (1) for the pressure in the incompressible flow V2/>inc= -d\puiu})/dyidyj, and substitute back in Eq. (1) to get n2P = V2Pic (6) (5)

where pinc is assumed prescribed by the given incompressible flow. The result is, formally,

P(X

^^^fP-*dV-

(7)

This has a tantalizing physical interpretation, namely if given pinc and then a slight compressibility of the fluid is admitted, it can be imagined to compress the fluid by the amount Pinc/pc2- This volume change must then act as a local simple source. Ribner (1964) suggests that regions of opposite signs in a flow results in quadrupole radiation. Unfortunately, while formally justifiable, on closer examination the concept turns out to have some serious problems, including poor convergence so that the integration has to be carried out to an non-compact distance where pinc is no longer a reasonable approximation to the real pressure because of the absence of time retardation in pinc, (Powell, 1964a), and so we do not consider it further here. All the same, the notion certainly has appeal physically, Legendre (1993) recently reiterating that the low pressure core of a vortex must result in some dilatation a la Ribner. Stimulated by the physical argument discussed in Section 2, it was suggested that vortex rings, or more generally vorticity, act as dipole sources in the flow and a mathematical basis was provided (Powell, 1961a, and Powell, 1964b, later referred to as P64b. This theory is entirely consistent with LighthiU's development, perhaps the most simple connection

Special 50th Anniversary Design Issue

JUNE 1995, Vol. 117 /253

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 1

induced by a small vortex ring is the same as that of a dipole, all in compressible flow. First, therefore, we need some basic hydrodynamics to establish this. If a "particle" of an unsteady flow could be imagined frozen and all the surrounding fluid vanquished, then it would have some angular velocity. Twice this is the vorticity , known if the velocity u is defined everywhere. The inverse situation is a result of classical hydrodynamics (see e.g., Lamb, 1932), that given the vorticity, the velocity can be found. Assume that u in an unbounded incompressible flow is given by the curl of some vector potential B: u = VAB.
>

Fig. 2

(11)

The continuity equation V u = V-(V A B) = 0is automatically satisfied. We find B by first taking the curl of Eq. (11), VAu = =V A (V A B) = V ( V - B ) - V 2 B . Now take V B = 0 (as can be shown post facto to be so), and solve the resulting Poisson's equation, V B = : B(x) = - / - dV(y), where r = x - y, (12)

and the volume integral covers all the vorticity. Then with Vx operating on x but not on y, u(x) = V,AB(x) = ^ - / v , A ( ^ ) r f F ( v )

= - ^ / 7 T ^ ) ,

(13)

since in the integral only r is a function of x. Now consider a vortex ring, a thin tube containing all the vorticity, see Fig. 1. The vorticity in it is always locally parallel to the tube. The circulation about it is defined by T = feda = ju-dc (14)

vortex ring may be moving or not, and it is said to induce a velocity field everywhere, including of itself (with the notable exception of in two dimensional flows) and of all other vortices. The same may be said of any part of the ring, even though parts cannot exist separately. This induction formula is believed to be due to Stokes, but because it has the same form as the formula for the magnetic field due to a current in a static conductor, as pointed out by Helmholtz, it is often called the Biot-Savart law ("analogy" better?). The fact that the circulation remains constant around any moving vortex ring enables a great conceptual simplification to be made compared to using the vorticity Eq. (3), since the vorticity generally changes with both time and position. A general flow can be considered to be made up of a continuum of vortex rings. The velocity induced at any point by each of them adds linearly: we can always back track to Eq. (13) if there is a need to do so. So without loss of generality, we may continue to consider just a single vortex ring. Lamb (1932) transforms Eq. (15) to represent a dipole sheet of uniform strength Y (per unit area) over any simple surface bounded by the vortex ring (see P64b for a more concise proof). The streamlines are therefore just the same, as suggested in Fig. 2. However, here we will need only the radial velocity at a distant point a = aa from the ring that we can get from Eq. (15). Then put r = (a y), with a y, to get a u(x) = ( a ) = - -

,
j- /a {(a - y) A s}ds(y) (16)

= 73 f i , /yAMy)-

() = TJ TA> (18> 47ra whereAa = a A is the component of A in the a-direction. This expression is exactly the same as that for a point dipole of strength D = TA in incompressible flow. Notice that Aa is the projection of the ring area as viewed F /-r.s F (.sinfrsl from the field point of observation, i.e., TAa is the vortex ring u ds = 1 ds(y 15) strength as viewed from the field point. A dipole source can be considered to be caused by a force where [rs] is the angle between the point of observation and F acting directly on the medium, in fact the local direction of the ring. F = pdD/dt = pd(TA)/dt. (19) This integral result shows that the velocity may be thought of as being "due to" the vortex ring, though it must be This implies that the momentum of the dipole fluid motion is remembered that the derivation hinges on the fact that pD = pFA. Evidently, the force is just the density times the rate f = VAu and u was presumed given in the first place. The of change of the vortex Hng strength TA.

here a is the local normal cross-sectional area and c (temporarily) is any closed circuit enclosing the cross-section, Stokes' theorem (Kelvin's transformation) having been used. The vorticity equates to the circulation per unit area of the cross-section. For example, a circular vortex core of radius a rotating as a rigid body with angular velocity co has F = aco .277a and so f = T/tra = 2co. The circulation is constant all around the ring, even if it changes shape as it moves with the flow, with the vorticity changing inversely with the cross-sectional area variations, according to Kelvin's circulation theorem, e.g., Lamb (1932). The ring must be closed, i.e., it cannot end in the flow. Then on putting T = Ts and dV = dads,

Now the vector area bounded by the ring is, from elementary vector analysis, A = /-yAds(y). It follows that (17)

- - -& / ^

" T, f - p r

>' <

254/ Vol. 117, JUNE 1995

Transactions of the ASME

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

dk = MdLds

Fig. 3

Now instead of the fluid being incompressible everywhere, take the dipole flow just considered to be that very close to a dipole source in a acoustic medium, i.e., the hydrodynamic field of the acoustic dipole source. (Strictly, we are ignoring the difference between taking kr = 0 in the former case and kr -* 0 in the latter, if k is the wavenumber for a sinusoidally varying source.) Then for this dipole we have the far field sound pressure at distance x, p(x) = pcux{x) P = d2
( T A x )

1
ATTXC

d2D*
dt

1 Arrxc

dF* dt (20)

47TXC dt2

This is equation (9) of P64b. The far field sound pressure evidently is due to the accelerating rate of change of the vortex ring strength as viewed from the field point, (FA)^. When the vorticity causes a net force to be exerted on the fluid, as is implied by Eq. (19), the vorticity must be bound, i.e., part of the (single) ring must bound to the surface of a solid body, so that the force applied to the fluid can be supported. The free field solution, Eq. (13) may still be used provided that any solid body is represented by some form of bound vortex distribution that enables the boundary condition at the solid surface (normal velocity = 0) to be satisfied, as by an image system. As this boundary condition then completely determines both the strength and position of the bound vortex system, it follows that the bound vortices are not subject to the velocity induction by the free vortices. The result embodied in Eq. (20) is the basis for two physical arguments that follow, one in Section 3 leading to the original formulations in terms of vortex force p(u A f ) , and the other in Section 5 to the expressions using ( y A 0 given by Mohring. 3 Vortex Force as the Source Here we follow Kelvin's circulation theorem and take the circulation to be constant as a vortex ring moves with the flow. Then the changes in TAX that govern the force component and the sound pressure occur because of changes in just the projected area Ax. Dipole Formulation. If the vortex is in part bound and its area A changes, Eq. (19) becomes F = pYdk/dt. (21) Now an element ds of the ring having local velocity u the increases the area of the ring in time 5; by amount d\ = (u8t)Ads, see Fig. 3, so \' = | u Ads and now we have F = Pr/uArfs = p/uA(rds) = /p(uAOdK(y). (23) (22)

vortex force (per unit length or volume) acting on the surrounding fluid. This notion is supported by the fact that the divergence of the latter appears in the inhomogeneous wave Eq. (9) just as would an externally body force f per unit volume applied to the fluid. But it is important to note that these are really integral relationships, and this interpretation does not mean that a physical force of the amount p(u A Tds) is necessarily applied locally to all parts of a vortex ring, for the external force can only be applied to the fluid where the vortex is bound to a solid surface. In fact, the total vortex force really equates to the rate of change of the impulse of the fluid motion, that we have taken to be the same as the rate of change of momentum, as implied by Eq. (19). Lifting Wing. This last result is the basis of the theory of lift of wings that is useful in demonstrating the concept of bound vorticity. In the simplest representation, there is bound vorticity with total circulation T about the wing. Helmholtz' theorem that the vortex ring cannot end in the flow is satisfied by the trailing wing-tip vortices connecting with the starting vortex which is left far behind, being "cast off" as the lift force was first generated, see Fig. 4. Take the wing of span b to have a constant velocity U through stationary fluid. A (downwards) vortex force pU A (Tb) must be applied to the bound vortex to move it through the fluid; its reaction is the lift force L = pUTb acting upwards on the airfoil. This is Zhukovskiy's famous formula for steady (and therefore silent) lift. In this elementary case u = 0 everywhere except for the bound part (on ignoring the vertical component for simplicity). Alternatively, take a wind tunnel frame of reference, with the airfoil stationary and the flow moving past it [a constant velocity being added to Eq. (13)]. In this case, the area of the vortex ring increases because of the drift downstream of the starting vortex, so the vortex force is pU A (Fb) downwards as before. While it gives the rate of change of momentum, it is obviously not a physical force applied to the starting vortex. But the bound vortex would have drifted downstream with velocity U as in a free flow had the vortex force of magnitude p U A ( f b ) not been applied to it by the airfoil. Now'consider the wing moving through a still atmosphere with its velocity slowly varying, with its inclination to the flow also varying so as to keep the circulation constant, then Eqs. (20) and (23) yield
P

TTc Jp(ll'*Tds>>**

(24)

Because p(u A Yds) and p(u A ) in these equations are local flow properties it seems appropriate to call them the Special 50th Anniversary Design Issue

Here u = 0 again for all but the bound part of the vortex. Here the far field sound pressure is due to those parts of the vortex ring that accelerateso causing its area and the total vortex force, as viewed from the field point, to change. On the other hand, it the wing moves at constant velocity with its angle of incidence varying slowly, its lift again also varies slowly with time. Then a series of elemental vortex rings 8T are continuously generated, all of them being stationary, except for the bound parts of them that sum to a varying total circulation about the wing EST = T(t), see Fig. 5. JUNE 1995, Vol. 117 /255

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 5 Fig.6

Then in this case we have ing force applied to the central image vortex 4T has to be included else the induced velocity there would take it away from the center. The total bound vortex force is easily found Equations (24) and (25) can be combined and generalized: to be the same as that just found, as may be expected. This last example is interesting as it demonstrates that a vortex in the proximity of a cylinder causes the cylinder to exert a force on the fluid and so there is sound radiation, (26) even though there is no circulation about the cylinder. When there is no bound vortex, or more strictly, no This agrees with the solution to Eq. (9), (P64b). It might be boundaries at all are present to support a force, the precedlooked upon as each vortex force element p(u AT)'ds (effec- ing equations give a null result. Specifically, the integrals of Eq. (26) and the time derivative of (23) may be zero, even tively) radiating like a dipole. Note that the changing vorticity in Eq. (25) is not to be though their integrands are not zero everywhere. Then the interpreted as concerning a vortex ring of varying strength: as radiation must be of quadrupole type, neglected in these is evident in the derivation, V is the changing total strength approximations. of just the bound vortex elements. Cylinder and a Vortex. If the airfoil of Fig. 5 were to be Quadrupole Formulation. Consider a solitary vortex ring replaced by a circular cylinder, and the shed vortices cast off of varying shape in a free flow. As there is no net vortex from it had alternating sign, a simple model for the aeolian force, its (vector) area A must be constant. The rate of tone due to alternating lift forces would result (P64b). Perti- change of area due to one element, u A 6s of one sign must nent to that, we consider a circular cylinder of radius b with be exactly cancelled by an equal one of opposite sign elsecirculation F about it with a single vortex T at distance R where, see Fig. 6. The same is true of the associated element of vortex force, p(u AT)ds. For the radiation from each to from its center, but for simplicity there is no mean flow. reach the point x in the far field simultaneously, that from The circular boundary of the cylinder is a streamline, and the more distant element must emit at a time y /c before the x the streamline remains if a bound image vortex T is placed other one, so exact cancellation fails by the amount of the at the inverse point b2/R on the same radial line with the solid rate of change multiplied by this time difference. Hence the cylinder removed. The real vortex has an induced velocity quadrupole radiation must be, from Eqs. (23) and (26), U = -{T/2ir)/(R - b2/R) and the whole vortex system (and its vortex forces) rotate with angular velocity w = U/R. The p(x) = ^^fyx(nATYxds\ (28) total (inwards) vortex force (per unit depth) has magnitude
Pr(o>R

P(x) =

^tp(u*r*)ds*

(25)

- wb2/R) = -QY2/2TTR

(27)

or alternatively,

that reflects the rate of change in momentum. This would be used in the two-dimensional form of Eq. (20) for the sound radiation (see, e.g., Morse, 1976). Now instead consider the force applied to the bound vortex - T to constrain it to move with its image point, that rotates about the center. Without the constraint of being bound, it would move at the velocity u = {Y/2ir)/{R - b2/R) induced by the real vortex. To force it to move with the inverse point at the velocity wb2/R an externally applied vortex force is needed of magnitude p(-T)(u)b2/R - u) = T2/2irR, in agreement with that just found. This must be the same as the force applied to the fluid by the real cylinder. A related situation is when a vortex is simply near the cylinder and there is no circulation about it (like a vortex being swept past a cylinder, except that here the mean flow velocity is zero). In this case an additional image vortex + F is needed at the cylinder's center. The angular velocity of the whole system is now found to be OJ = -(r/2ir)(b2/R2)/ {R2 b2) and the total vortex force, including that of the two images, turns out to be p (r/2ir)(b2/R3). In considering the total bound vortex force, the constrain256/ Vol. 117, JUNE 1995

P = T-^T A>,0"xdV(y),*
4-TTXC
J

(29)

the latter in agreement with the solution to Eq. (8), P64b. The physical argument leading to Eq. (28) for a moving vortex ring was oversimplified in that the changing yx is also important. As can be appreciated from the Contiguous Method discussed in Section 3, the time differentials really apply to the whole integral so that the final result is that p(x) = j^fyx(nAr)xds"*. (30)

[The same results more formally by the incorporation of a moving delta function within the bracket of Eq. (28).] Spinning Vortices. The most simple case is that of a pair of spinning straight line vortices t of infinite length and distance 2b apart. We consider the three-dimensional radiation from a short length I of them, Fig. 7, using Eq. (30) as in P64b. Here the (two-dimensional) vortex velocity is U = T/Airb (or F = AnrbU) and the whole system rotates about their center with angular velocity u> = U/b. The vector Transactions of the ASME

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

u=y A B \x~-V(f>
Fig. 7 Fig. 8

u A T d s = AiraUh is directed outwards from the center. The time derivative rotates it to the direction of U and as this is perpendicular to b, we anticipate a rotating lateral quadrupole. Then Eq. (30), with yx = bx = b cos [bx], gives 2p d2 p(x) = - 4 b2U2l, 2 cos 2 [b,x] xc2 dt Now cos [b,x] = cos [b,l] cos [l,x] + cos [b,2] cos [2,x] = cos a>t cos [l,x] + sin a>t cos [2,x] and cos 2 [b,x] = l/2(cos 2 [lx] - cos2[2x])cos2o)f + cos [lx] cos [2x] sm2u>t. The first term, a pair of opposite longitudinal quadrupoles, may be written as a lateral quadrupole with axes 1', 2' at 45 deg. to 1, 2. The result then is p(x) = xc + cos [lx] cos [2x] sm2u>t} (31) The two fixed lateral quadrupoles in quadrature are equivalent to the single rotating quadrupole to be expected. This was the first aerodynamic sound result to be given in closed form, (Powell 1961a, P64b). "Leapfrogging" Vortex Rings. A pair of vortex rings of nearly the same radius, average radius a, of the same strength (and sign) and very close together, rotate about each other, described as "threading through" or "leapfrogging" each other, maintaining symmetry about the 3-axis. As they are very close together, a small length of the pair can be likened to the length i of the spinning pair distance 2b apart just considered. Its (1,2) quadrupoles evidently cancel with the diametrically opposite like element, while the (l',2') quadrupoles are additive. With axial symmetry about the common axis, the far field sound pressure on it must be given by Eq. (30) with I replaced by 2TTR while cos [l'x]cos[2'x] = 1/2, so
87r ?

The Contiguous Method Here a surface, conveniently spherical of radius a, is imagined drawn in the fluid surrounding the sound generating flow, see Fig. 8. Its perturbations due to the unsteady free flow contained within can be found by any method, though here vortex methods are being emphasized, and the surface is taken far enough away that the flow region is relatively very small (approaching a point). These perturbations drive the contiguous acoustic field. The far field sound pressure can then be found from the Helmholtz-Kirchhoff integral equation taken over the spherical surface, in its quadrupole approximation, 1 l
2

/>

ATTXC

dt'

iyx"x

:y* dn dS(a)

(34)

where n is the outwards normal. The pressure for a free incompressible vortical flow, given by the incompressible version of Eq. (9), turns out to be (P64c) p(a) = - ^
47T(l
T

/ 4 {cos[l'x]cos[2'x]cos2o^ / 3 y >
J
A

0 ^ n y ) ,

(35)

for use in Eq. (34). This method was used for the spinning vortex problem, finally yielding the same Eq. (31), (Powell, 1963). But this frontal attack is not necessary if the source is acoustically compact, as the pressure perturbations can be viewed as those in the hydrodynamic field of any lateral point quadrupole driving the contiguous acoustic field. These fields are simply related to each other as follows. The velocity potential for a general lateral point quadrupole can be written as

*w =
3
ATT

'

3
2 X C

d
dt

d2

XC2 dt2

g, 7 *cos[xi]cos[xj],

(36)

so the pressure (pd<t>/dt) in the hydrodynamic and far acoustic fields are P 3 K a ) = -JZ a" Q'ucos [ a i ] -j 4-JT
cos

p(x) =

P RU

cos2a)t

(32)

[*>]>

P The sound field is axisymmetric and, as discussed in the />(*) = ATTXC2 Ql'* cos[xi]cos[xj], Introduction, Eq. (4), it must have the directionality 1/2(3 cos 2 [xl] 1) times the amplitude in the axial direction. , respectively. Hence there is the simple unique relationship Hence 1 aJ AirpR (37) 4 2 P(X) = p(x) = / (3cos [xl] - l ) c o s 2 ( (33) 3 x7P(a) This agrees with Mohring's result (1978). In this case a vortex force (or, the rate of change of momentum) can be associated with each vortex because of its area change. But together there is zero net force, consistent with them forming a free flow of quadrupole nature. In the present case, on using Eq. (35), (x) =
FK !

r
ATTXC2

fyx(uAn"JV(y)*

(38)

exactly the same as Eq. (29), JUNE 1995, Vol. 117 1257

Special 50th Anniversary Design Issue

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

In terms of velocity ratios instead we have 9 "(a) = " * ( a ) = TT^T Qij c o s f a '] 47Tfl 4 1
M cos

source, driving the contiguous acoustic field external to the cylindrical surface. Then take Eq. (41) and rewrite it as NL a3 U5^ (44)

,(x)

,3 ATTXC

dt3 1

cos fxi]cos[xj],

Then the use of Eq. (43) yields the far field sound pressure: \p(x) (39) l = pC \UX(X) 1 = pC.tyTT. x l^ c 3 / 2 ' (45)

so
u

*00 = 9 ^1 " "

while the sound power (per unit length) is P = 1 \p(x)\\ 2 pc


2pUi 2irx = 16-rr j - . c

that will be used later. Such simple relationships cannot be written for the longitudinal quadrupole, since its hydrodynamic and acoustic fields have different directionalities. Similar relationships can be readily written for dipole and monopole source types (Powell, 1964c). In two dimensions, for the sinusoidally varying lateral quadrupole we have

(46)

<Kx) =
l)

H?\z)e'",

(40)

where H^ is the Hankel function of second order of the first kind, and z = kx with the wavenumber k = 2co/c. The radial velocity has the asymptotic amplitudes -J(2/TTZ) as z = be -> oo and 8/(irz 3 ) as z = ka -> 0 at the surface of the enclosing circular cylinder. Then

The seventh power law typifies aerodynamic sound production in two-dimensional free flow. This result was first given by Obermeier (1967) on introducing the Method of Asymptotic Expansions to aerodynamic noise problems, followed by Stiiber (1970) who solved the inhomogeneous wave Eq. (8) by using Green's functions in the Fourier transform plane. Elliptic Vortex (2-D). In two dimensions, if a vortex of circular cross-section (radius a) and of uniform velocity is distorted into an ellipse, the elliptical shape remains, but it rotates at the rate /4, Lamb (1932). Its boundary has radius R(t) = a + e c o s 2 | 0 - - f |, (47)

ux(x) I = y

r(fea)
jrkx

w(a) I for a small eccentricity e, where 8 = [al]. This perturbation clearly has the same form as a point lateral quadrupole, so the boundary of the vortex may be considered to drive the whole acoustic field contiguous to it. Its radial velocity is a (a) =dR/dt= (42) - e s i n 2 J 0 - - t\. (48)

= A xl/2c5/2 I u f l ( a ) | .
For the dipole we have 4>(x) = that ( TT\ o)3/2a2

(41)

H\1\z)e'a with the result

As for any two-dimensional inviscid free flow, apply Eq. (41), in which now k = /2c, so we can immediately write p(x) = pcux(\)

This Contiguous Method is distinguished the fact that the perturbations on the compact enclosing surface can be readily associated with the appropriate type of point source for the contiguous acoustic field, since the directional characteristics of the monopole, dipole and lateral quadrupole are unambiguous and orthogonal to each other. There is the further requirement that dependence on the radius a must vanish when the substitution for p(a) or fl(a) is made. Spinning Vortices (2-D). We illustrate this last result by considering a pair of spinning line vortices in two dimensions. The notation is the same as in Fig. 7. Here the vortex velocity U = T/Arrb, so T = A-nUb and co = U/b as before. Then the induced velocity at distance r from each vortex is f AT/2rrr, so for together on the sphere of radius a u(a) = -

fr
64
p6fl

3/-7/2 a%

x^c3^2

sin2|6- - f r ) .

(49)

In terms of the velocity U = at,/A of the bulges travelling around the periphery of the vortex, this is
f/7/2

p(x) = 2 > . p - ,1/2

1/2 1/2.3/2

sin 2

1 -t\.

(50)

This gives Howe's (1975) result, obtained by solving a generalized form of Eq. (8). 4 Vorticity-Moment as the Source

r
27rr

/(a-b), r
b|'

(a + b ) A r
la + b | -

The radial velocity a (a) = a u(a) is needed, and with 2 o > i ) w e have | a - b I = (a - ba) and {a - ba)- 2 _ (1 + 2ba/a), hence (a) = . v ( v
77 a

Dipole Form. Here we consider the flow to be made up of vortices again, but for the moment instead of moving they are fixed, with their strength varying. Equation (20) can be rewritten for a single ring as p /> -A, ATTXC d2T* dt' (51)

bAr).
/

so by using Eq. (17) instead of Eq. (22) we can write more generally

The last factor is sin [ab] while we can write [ab] = (cot 9), so b3 u a (a) = 4 Usm2(a>t - 9). (43)

pw = ^j^(y-n*dv(y)*

(52)

This unambiguously has the directionality of a quadrupole 2 5 8 / Vol. 117, JUNE 1995

This is the dipole formulation given by Mohring (1978), apart from a notational difference, found by solving the inhomogeneous wave Eq. (9) (without the last term) with the Transactions of the ASME

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

be defined by a continuum for fixed elemental vortex rings, their strength varying in such a manner as to represent moving rings with no change in their total strength. To use the Contiguous Method, take Eq. (13) and find the radial velocity at the surface of an imaginary sphere of radius a, more precisely than for the dipole strength of Eq. (16), u(x) 1 1
4TT

Q-y)A /
la-y|
3

dV(y)

4-77- J

(y A 0
| a

yI I a y I = (a - ya) and

Fig. 9

Now with a : y we have - 3 (1 + 3)ya/a\ so ia-ya)-3>

aid of a vector Green's function. It evidently has a very simple physical interpretation in view of the entirely equivalent Eq. (51): The far field sound pressure is again due to the accelerating change of total vortex strength as viewed from the field point, due here to changing circulation or vorticity. But for a bound vortex, the (vector) area A likely changes too, so we have the more general form from Eq. (21): 1 P(*) = 7- I j (y*T)*ds(yy* ATTXC (53)

() = T " - lya(3^)adV(y), (55) Anra J the first, dipole, term being zero by hypothesis. This can be the velocity field of only a lateral quadrupole in incompressible flow, witness the a~4 and the dependency on the two axes of y and (y A f ) at right angles to each other. This hydrodynamic field drives the contiguous acoustic field external to the imaginary spherical surface according to Eq. (37), so we have the far field sound pressure directly, Pi*) = PCMX(X)
4TTCX2

fjy*(y*r)xdV(y)*

(56)

This is an integral relationship, and its integrand cannot be associated uniquely with the local flow property, for it depends on the location of the origin for the value of y. Hence the integrands cannot be reasonably interpreted physically as dipole strength (force) per unit volume or length, (though this certainty is possible in an abstract mathematical sense), nor can the integrand be cast as the source term of the inhomogeneous wave equation. Since (y A D is the moment of T about the origin, it seems appropriate to call this the vortex-moment formulation* to distinguish it from the vortex force form. Oscillating Rigid Sphere. Consider a small rigid sphere of radius a oscillating with velocity V(t). This has a known dipole acoustic field with strength D = (27ra3)U, see Fig. 9. The velocity tangential to the surface of the sphere at point a at angle 9 to U is Ue = (1/2) U sin 9, relative to the undisturbed medium; and relative to the sphere it is U sin 9 + Ue = (3/2) U sin 9. The vortex sheet at the surface of the sphere can be considered to consist of elemental circular vortex rings of area ira2sin29 and of circulation ST = (3/2) f/sin 9.ad 6 by Eq. (14). Integrating over all the elements for use in Eq. (20) immediately gives pa3 p
ATTXC

This is Mohring's (1978) quadrupole result, apart from a notational difference. For a system of vortex rings the corresponding expression follows as
P(x)

2xc the standard dipole formula. This final result was also obtained in P64b by a slightly different method that explicitly recognized the presence of the solid, moving, surface by a monopole distribution, while the classical pressure distribution was replaced by a vortex sheet. In the treatment here, both of these are represented by a (three times stronger) vortex sheet alone. Of course, this is an example of bound vorticity and there is no other vorticity: the solid surface obviously supports the force applied to the fluid. Quadrupole Form. A single vortex ring of variable strength is not permissible in a free flow, since the force applied to the fluid would be non-zero. Instead, consider a free flow to
T h e author has previously used the term vortex-alone to emphasize that the velocity is not explicitly involved in Eq. (53); however, for moving vortices this feature is lost as soon as the time derivatives are taken.

nrxc J 5 Proof of Equivalence of Forms. It may be proved directly that the pairs of far field dipole and quadrupole formulas in terms of integrals of the vortex force p(u A ) and of the vortex moment (y A '), in both two- and three-dimensional flows, are in fact exactly equal to each other (Powell, 1995b). Applications. Mohring (1978) shows how this quadrupole result may be used to estimate the radiation from a pair of vortex rings threading through each other, describing them with the help of S-functions. This was treated earlier, as the physics may be more easily appreciated there. We note that in the two-dimensional dipole applications with rotating vortices, T is constant, so if the origin is taken at the center of rotation, d(yT)dt = coyF = uT and the vortex-moment calculations immediately reduce to those of the vortex force method, and so are not repeated here. A computational fluid dynamics solution (CFD) of the Navier-Stokes equation was found by Mitchell et al. (1992) for a pair of viscous, compressible two-dimensional vortices. They found good agreement with vortex-moment formulation (using computed values in the integrals) at low Mach number and small times. At a larger time the vortices suddenly merge to resemble a rotating elliptic vortex. The latter was considered by the Contiguous Method in Section 4. Effect of Dissipation The dissipation of the kinetic energy of the flow by viscosity has two effects. One is purely dynamical, for as may be seen from Eq. (3), a change of kinetic energy ( = K) is associated with a source effect of uniform directionality. Evidently, as a Corollary to the Three-Sound-Pressures Theorem, we have the contribution p(x) 1
6-7TXC

= Pcux(x) = - ^ / \yx(yAr)xds(yy*

(57)

K"

(58)

Special 50th Anniversary Design Issue

JUNE 1995, Vol. 117 /259

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Qualitatively, this may be looked at as being due to a contraction as the low pressure in vortex cores decreases because of reduced centrifugal force ( ~ pu2) as they lose strength. The other effect is the change of volume due to the heat from the energy of the viscous action (equal to the loss of kinetic energy). This contribution is easily found to be 1 P
4TTXC

sure viscous vortex core. Hence there is in fact a rational basis for why the formation of a real vortex should result in dilatational sound radiation. But it appears to be due to a purely dissipative effect, and not the dynamical one that first comes to mind. It seems that a long-standing paradox has been resolved.

(y-l)K"

(59) References
Corcos, G. M., 1958, "Some Measurements Bearing on the Principle of Operation of Jet Silencing Devices," Douglas Aircraft Company report SM-23114. Howe, M. S., 1975, "Contributions to the Theory of Aerodynamic Sound, with Application to Excess Jet Noise and the Theory of the Flute," Journal of Fluid Mechanics, Vol. 71, No. 4, pp. 625-673. Kambe, T., 1984, "Influence of Viscosity on Aerodynamic Sound Emission in Free Space," Journal of Sound and Vibration, Vol. 95, No. 3, pp. 351-360. Lamb, H., 1932, Hydrodynamics, 6th edition, MacMillan, Dover reprint 1945. Legendre, R., 1993, "High Velocity Acoustics," Journal of Sound and Vibration, Vol. 157, No. 2, pp. 193-203. Lighthill, M. J., 1954, " O n Sound Generated Aerodynamically, I, General Theory," Proceedings of the Royal Society (London), Vol. A211, pp. 564-587. Meecham, W. C , 1958, "Acoustic Radiation from Isotropic Turbulence," Journal of the Acoustical Society of America, Vol. 30, No. 7, pp. 318-322. Mitchell, B. E., Lele, S. K., and Moin, P., 1992, "Direct Computation of the Sound from a Compressible Co-Rotating Vortex Pair," American Institute of Aeronautics and Astronautics, Paper 92-0374. Mohring, W., 1978, " O n Vortex Sound at Low Mach Number," Journal of Fluid Mechanics, Vol. 85, Part 4, pp. 685-691. Morse, P. M., 1986, Vibration and Sound, Acoustical Society of American, 3rd printing, paperback edition. O b e r m e i e r , F., 1967, " B e r e c h n u n g aerodynamisch erzeugter Schallfelder mittels der Methode der 'Matched Asymptotic Expansions,' " Acustica, Vol. 18, pp. 238-240. Powell, Alan, 1961a, "Vortex Sound," Univ. Calif. Los Angels, Dept. Eng. Report 61-70. Powell, A., 1961b, Aeroacoustics Notes, Course 224a, Univ. Calif. Los Angeles, pp. 4.82C-4.82F (unpublished). Powell, A., 1961c, " O n Aerodynamic Sound from Dilatation and Momentum Fluctuations," Journal of the Acoustical Society of America, Vol. 33, No. 12, pp. 1798-1799. Powell, A., 1962, "Three-Sound-Pressure Theorem, and its Application, in Aerodynamically Sound," Journal of the Acoustical Society of America, Vol. 34, No. 7, pp. 902-906. Powell, A., 1963, "Mechanisms of Aerodynamic Sound Production," NATO Advisory Group for Aeronautical Research and Development (AGARD) Report 466. Powell, A., 1964a, " O n Sound Radiation in Terms of Pressure Independent of the Sound Speed," Journal of the Acoustical Acoustical Society of America, Vol. 35, No. 8, pp. 1133-1143. Powell, A., 1964b, "Theory of Vortex Sound," Journal of the Acoustical Society of America, Vol. 36, No. 4, pp. 177-195. Powell, A., 1964c, " O n Flow Fields Driving a Contiguous Acoustic Field," Journal of the Acoustical Society of America, Vol. 36, No. 1, pp. 830-832. Powell, A., 1991, "Nature of the Sound Sources in Low-Speed Jet Impingement," Journal of the Acoustical Society of America, Vol. 90, No. 6, pp. 3326-3331. Powell, A., 1995a, "Aerodynamic Noise Theoryan Alternative Monopole Source Term," Journal of the Acoustical Society of America, Vol. 97, No. 4, pp. 2144-2146. Powell, A., 1995b, "Vortex Sound Theory. Direct Proof of Equivalence of Vortex Force and Voriticity-Alone Formulations," Journal of the Acoustical Society of America., Vol. 97, No. 3, pp. 1534-1537. Proudman, I., "The Generation of Noise by Isotropic Turbulence," Proc. Roy. Soc. (London), Vol. A214, pp. 119-132. Stiiber, B., 1970, "Schallabstrahlung und Korperschallan-regung durch Wirbel," Acustica, Vol. 23, pp. 82-92. Ribner, H. S., 1959, "New Theory of Jet-Noise Generation, Directionality, and Spectra," Journal of the Acoustical Society of America, Vol. 31, No. 2, pp. 245-246. Ribner, H. S., 1964, " T h e Generation of Sound by Turbulent Jets," Advances in Applied Mechanics, Vol. 8, pp. 231-262.

The two effects are of opposite sense, so together (5 - 3 7 ) P(*)


K"*.
\2TTXC

(60)

Dimensionally, this monopole resembles three orthogonal quadrupoles. This simple physical argument (Powell, 1991) yields a fundamentally important result found by Kambe (1984). The Contiguous Method readily accommodates the added non-directional term for the monopole due to dissipation. 6 An Alternative Monopole Form The Lighthill Eq. (1) may be put in another form by use of the continuity equation for incompressible flow:
P = -pdUi/dyj.dUj/dy,.

(61)

After further manipulation, this can be written as a2p 1

r-ef

(62)

where ef is sum of the squares of the principal rates of strain (Powell, 1995a). The pressure in the flow is given by V2p

p\~r

(63)

The squares of the vorticity and strain rate are always positive, but carry opposite signs. The first one means that pressure minima are likely to occur in the center of regions of high vorticity, while pressure maxima arise where the strain rates are highest, as may be imagined to occur where vortices collide with each other. These pressure regions of opposite sign are clearly associated with the sound radiation, evident from the similar Eq. (62). The total instantaneous strength of these source distributions can be proved to be zero. The final result is that the pressure regions of opposite sign must form lateral quadrupole fields, that term in the multipole expansion giving

This picture is reminiscent of Ribner's sketch in connection with the dilatation theory (Ribner, 1964). We can now address the following enigmatic question, raised in Powell (1961c): "There is a dilatational effect, very evident in the formation of a vortex core since its low pressure core must undergo some expansion. But since no external agency can cause this dilatation, we are led to enquire of the process causing it and, further, whether is also results in sound radiation," Indeed, the monopole effects due to the rolling up of a vortex cancel, exactly. But now we see that dissipation results in a net monopole-like radiation and there is certainly dissipation in the formation of a low pres-

2 6 0 / Vol. 117, JUNE 1995

Transactions of the ASME

Downloaded 04 Feb 2010 to 131.175.12.86. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

You might also like