You are on page 1of 284
A Primitive Approach to Aeroacoustics Avraham Hirschberg and Christophe Schrar ca University of Technology, Dept. of App. Physies, JM. Burgerscente Postbus 513, 5600 M3 Eindhoven, The Netherlands. ‘Tel: 00(81)402473110, Fax: 00(31)402464151, Email: A.Hirschherg@tue.nl ® von Karman Institute for Fluid Dynamies, Environmental aad Applied Plaid Dynamies Dept, 72 chaussée de Waterloo, 1610 Rhode-Saint-Gentse, Bel Tel: 00(82)23500611, Fax: 00(32)23500600, Bm: sebramévki.ae-be Abstract. primitive approach to aeroacousties is presented using the aeroacous: tic analogies. This aproach is illustrated by e few examples like musical instruments, the Rijke tube, speech produetion ete 1 Introduction Due to the non-linearity of the governing equations it is very difficult to predict the sound production by fluid flows. This occurs typically at high speed flows, for which non-linear inertial terms in the equation of motion fare much larger than the viscous terms (high Reynolds numbers). Diroct simulation of such flow v difficult. As sound production represents only. fa very minute fraction of the energy in the flow, the accuracy of mumerical simulations is usually not sufficient to predict the sound generation, This is particularly dramatic in free space and at low subsonic speeds. The fact that ‘the Sound fiold is in some sense a small perturbation of the flow, ean howover mate solution be used to obtain appro Aeroacoustics provides such appraximations and at the same time a def inition of the acoustical field as an extrapolation of an ideal reference flow Which we will specify later. The difference between the aetual flow and the reference flow is identified as a source of sound. This idea was introduced by Lighthill ([21], [22)) which calls this an analogy. A second key idea of Lighthill [22] is the use of an integral equation as a formal solution. When approxi ‘mations are used in such an integral equation, random errors are smoothed out Considering confined flows, the acoustical energy can ac resonant modes. The acoustical particle displacement velocity can become of the same order of magnitude as the main flow velocity. Yet the ideas of the analogy will appear to remain useful ‘The recent publication of Fedorchenko [8] contains severe eritieisins on the analogy and the way the analogy has been used. ‘Tam [16] also provides an example of how the analogy can be used to draw erroneou nuilate into conels Y, Aug, lau, VPagious, Paton (da) LN S86, pp 1-3, 2A. Mirsehberg. © We will alse discuss some examples of such problems, ‘The key issue is that the analogy is only useful as an approximation. Such an approximation is aged on some intuition and usually empirical abservations. We could say that Lighthill nas proposed an intelligent way Lo carry out dimensional analysis. Ib isa joy for engineers to discover semi-empirical relationships between relevant parameters under circumstances where exact Solutions are not available. This kind of results are usually obtained by using rather crude appreximnations. In fact, ignoring subtleties does help, We would like to introduce the reader to this primitive approach to aeroacousties by considering a few examples: + the sound production hy a subsonic isothermal turbulent free jot (Se # Ge Chermo-acoustical oscillation of the Rijke-tube (Sect. 7 «# the aere-aconstical response of an open pipe termination (Sect. 11), and «the human whistling (Sect. 12) 1.4), Before each example we will discuss some basic theory: 4s thid dynamies (Sect. 2) @ Lighthill's analogy (Sect © therme-acoustics ( © acoustical energy (Sect. B), # vortex-sound theory (Sect. 8). 2 Fluid Dynamics We consider the motion of fluids in the continuum approximation, This means that quantities such as the velocity v and the density @ can be described by fields, smooth functions of space and time coordinates (x,t). In a non relativistic approach the mass is conserved. Hence the rate of change of the density observed while moving with the fluid velocity is equal to minus the dilatation rate ({1], [47], [20)) » a schere the Lagrangian time derivative Do/Dt is related to t! derivative O/A by Eulerian time De _ 8e Pe 8 sw. 0) ) ame ve. @ Substitution of definition (2) into (1) yields the mass conservation law applied to a fixed infinitesimal volume element: Be 2 av (on) =0 8) Ht) 3 Using Einstein summation convention we can write this for cartesian coordi nates @ = (2.2203) as Oe, Ba Bt Ox A Primitive Approach to Aeroasousties The repetition of an index indicates a summation: B00 | Bor , Bay i Oxy * On : Oy ba Please note that we do not have in a non relativistic approximation any rnass souree term. In agreement. with this approximation we apply the second nw of Newton to a fluid particle: De De where f is the density of the force field scting on the bulk of the fluid, while “VP corresponds to the forces acting on the fluid surface expressed in terms of a stress tensor P. Using the mass-conservation law (4) we obtain the momentum equation in conservation form: -V.P+f Bove , Borns —_ OPy ra a de Please note that it is rather odd that Fedorchenko [8] does introduce « rnass source Lert in & non-relativistic approach and worries about its conse zn conservation law. The isotropic part of this tensor corresponds to the effect of the hydro- static pressure py quences for the moment Py = pBy — 55 & ‘The deviation oy from the hydrostatic behavior corresponds usually ina si ple fd to the effect of viscosity. For a simple fluid , is symmetrical. Hence (3) end (7) form a sct of four oquations with eleven unknowns (@, 2.0%) Using the energy equation would add ono equation and at least one more unknown, In order to obtain a number of equations equal to the number of unknowns we have to use empirical information. This is commonly done by ising so-called constitutive equations. A widely used approximation is to assume a fluid at a state of local thermodynamic equilibrium, This implies that there exists a functional relationship between the pressure p, the den- sity o aud the specifi entropy’ s (entropy’ pet unit of mass). This is elled an equation of state p= ples) (9) In differential form we have: Dp, De, (ap) Ds : meer (ie), Be oy chere ¢ is identified as the isentropic equilibrium speed of sound: 7 -¥(2), ay 4A, Hirsehberg, C. Seam. The entropy equation ean be written as [47] Ds oP wg 40s (00 a a Vig bos (T) Qe 12) whore Tis the absolute tomporature, q the heat flue and Qu. density of tate of energy production in the bulk of the tuid, In the audio frequency ‘and amplitude rauge it appears that sound propagation in ambient air i approximately isentropic ([35} (7) 3 Lighthill's Analogy We use as starting point the exact mass and momentum equations (3) and (7). We take the time derivative of the mass equation (3) and subtract from this the divergence of the momentum equation (7) to obtain: (3) We can subtract on both sides of this equation a term ef is a velocity. ‘This provides the right hand side of the eq the wave equation of d’Alembert a Bo/Ox2) were ey tion the form of OF This is in gencral @ useless equation, The velocity cy could be the speed of light or one centimeter per century... ‘The equation would still be exact Furthermore we had four exact equations with eleven unknowns, and we are left with a single equation and as many unknowns. Clearly this seems a dead end. ‘The key idea is to compare this equation with the equation which we ‘obtain under normal andlio frequeney and amplitude conditions for the per- turbation of a uniform and stagnant fluid in the state (go, po). Under such ci cumstonces we can neglect friction losses and heat conduction for distances of propagation comparable to the wavelength. We introduce the perturbations: (15) 8) en eie (ny where in the specific case discussed v9 = 0. The linearized equations of mass and momentum become: de, Out 2 a0 a as) ot Oxy a a A Primitive Approach to Aeroacousties 5 fi (19) We have introduced here a mass source term Qm which we specify further by assuming that it doos not involve any momentum injection and that the injected fluid has the same properties as the reference Muid (do, po). Under such circumstances the reference fuid remains uniform and stagnant so that, tho equation of state (10) ean be written as: pace! (20) Due to linearization in combination with the assumption of a uniform fluid, the convective effects drop. Please note that this will not occur if gy ar pr are not uniform ((35]). Eliminating the velocity of by subtracting the divergence of (11) from the time derivative of (18) yields; ay ot When we are listening to the sound of a flute in a concert hall, the sound field around our ears is accurately described by the bormogeneous wave equa tion in which we have dropped the source terms. In the case of the equation of Lighthill (14) linearization and neglecting visco-thermal dissipation would also vield this equation if we identify eo as the speed of sound in the stagnant uniform fluid surrounding the listener. We can further define the perturba tions o = 0— a and p! = p—po as the diflerences between the local values of @ and p and the values of these quantities in the reference fli tho listener (Fig. 1). This is a generalization of (15) and (16) in which the perturbations do not need Lo be small. As a9 and py are like ey constants we can write Lighthill’s analogy as ((10),[4]) surrounding _ Powys — ay) Onde; and identify the right hand side of this equation as @ source of sound. This procedure can be considered as a generalization of Rayleigh's ap- proach to scattering theory [87]. We consider the case of the scattering of ‘8 wave on nox-homogeneities in fluid density which have a stall spatial extension compared to the wavelength of the incident wave py. When the unperturbed fluid is stagnant. and the density difference 9 ~ oy is small one ‘an assume a uniform mean pressure py. The entropy will however not be uniform so that the speed of sound e will vary. ‘The main difference between this situation and the vorlier diseussion of the acousties of a stagnant uni- form fluid is thet the convective effects in the equation of state (10) cannot be neglected. In linearized form and neglecting the pressure gradient tet Bet (8 v0 es ae (23) 6A. Mirsehberg. © Fig. 1. Sonrve and listener in the analogy of Lighthill, While in the source reaion the density fhuctuations are not necessarily small, avound the listene acoustical behavior with < go. and the corresponding mass and momentum conservation laws in the absence of souree terms (in — O and fr = 0) en and: 25) where the quantities @ and ¢ designate the time average of the density and speed of sound. They are therefore time independent. but vary in space. Eli ination of v by subtracting the divergence of (25) from the time derivati (28) yields: #18 9 (Ley) 0 a ( 8) GF ‘The fiest source term is identified as a monopole source (erm due to the ait compressibility of the reference fluid 1/go¢? surrounding the listener and the local fhuid compressibility 1/ge2 in the source region. ‘The second term is a dipole sound source due to the Archimedes foree on the non-uniform fluid particle as a result of its acceleration @v'/6t = —(Tp)/o. A further step is Lo split the acoustic field into an incident wave pl, which A Primitive Approach to Aeroasousties 7 satisfies the homogeneous wave equation and a scattered wave pi. whieh we assume to be small compared to the incident wave, An approximate solution can then be calculated by neglecting the effect of pj, on the source term (eee) -¥ (Gt ome) om Another limit case is obtained by neglecting all the source terms in Light- hill’s analogy except for the force term —V. f. This is a very useful approxi- ‘mation when considering the sound production by moving objects. The force Ff represents then the reaction of the object to the force applied by the fluid flow. This corresponds to the principle of Gutin which is discussed by Gold- stein [10] 4 Jet Noise In the previous section we introduced Lighthill's analogy and related it to Royleigh’s scattering and Gutin’s principle. Lighthill's ambition was to apply this approach to the much more delicate problem of the sound production by a free turbulent flow [22]. He argued that when one considers an isothermal subsonic free jet of air in air all the sources terms in equation (22) are neg ligible except for @ ovyv;/@ar,22;. It is rather obvious that in the ab walls and of external force field f; ~ 0. The fact that the other terms vai nob as trivial as suggested by Lighthill [22]. A more careful discussion of the validity of this approximation was provided much later by Morfey ((26],[27]) and Obermeier ((32), [33], [19). Lighthill's intuition appears not to be bad AL this point we note that a source term involving a second spatial deriva- tive snch as 6%qnv,/dayén, is likely to behave as a. quadmnpole. ‘This simply is due to the fact that since there is no external force to sustain a dipole and that we have neglected monopole terms we are left with a quadrupole as leading order term. Of conrse one ean always find a pathological ease in which the leading order term in the radiated sound field is an octupole |S}, but this hhas little practical implications, If we now ignore the quadrupole character of the sound field and use a characteristic length seale Din the flow field to predict the sound strength we would find 2/8, ~ 1/D. If we assume a compact sound sonrce, we would find a radiated sound pressure scaling as the stagnation pressure of the flow p’~ ool/$ and as a consequence a sound source power scaling as Us. This is contradicted by the experiments. If we ignore the retarded time effeets and simply integrate the source taking into account the prescuce of spatial derivatives the sound production vanishes due to the limited size of the sound region (lA). ‘The most convenient way to obtain a better approximation for te effect of retarded time differences across the source region is to start from the integral 8 AL Minvehberg, C. Seam formulation: soltsalrsy) oe ayer 3 { lt elev) FE det (30) ol(x,t) = a ‘with the free space Green's fhnetion given by Bir = [2 — wi/ea) Gult-ate, 31 (atu) = ee 1) whieh has the niee symmetry property o_o @ for differentiation with respect to the source coordinate y and the observer coordinate & (see Fig, 2). By partial integration we can move the space deriva- a a ae 8G, %y aaa &, or Oey we Fig. 2. Syrnmetry of differentiation of the free space Green’s function with respect. lo cheerver’s and source coordinates. In free space, moving the source towards the observer is equivalent to moving the observer te the source. Live from the source term 2 ons; /Oys0u5 (whieh we do H0t know aeetately) towards the well known Green's function, Using the symmetry property (32) vwe can replace the derivatives of Gy with respect to the source coordinate y by derivatives with respect to the observer coordinate 2. AS the i 's performed on the source coordinate, we ean move this spatial des out of the integral, and obtain the integral formulation Splipectit eetU Jy Saree 2a SE ed sare f Poa uj dydr. (88) When the observer is in the far field, the relevant length seale there is the aconstieal wavelength. Assuming that the sound production is dorninated by the time seale D/L in the flow, we ean replace each derivative by rultiplication by a factor Uo/(apD). Assuming Uo/eg << 1 appears then to be equivalent to the assumption of a compact flow region A Primitive Approach to Aeroacousties 9 Neglecting other, more subtle, retarded time effects ([10], {23)) we find that the sound source preasure will scale as Uf and the power as U8, which is the famous law of Lighthill ‘As stressed Powell in [37] this is one of the few examples in which theory predicted something hefore it was observed experimentally. We insist again ‘on the fact that we found a quadrupole because we used the fact that we know that there are no external forees, which is used as argument to exclude any dipole term. A conceptually similar discussion will follow in Sect. 8 where an alternative formulation of Lighthill's analogy, known as vortex sound theary, will be developed. Yet, (83) is not a formal solution of our problem, since the density on its right-hand side contains the unknown (9 = go + ¢). To obtain an integral solution, Lis fuctuations of density in the source region seale with the square of the Mach umber of the flow. ‘The density in the right-hand side of (33) can then be approximated by the reference density ap, and in the low Mach amber livat, we end up with the integral solution: ill further argues that for isothermal free jets, the relative we au de mam | of la ~ yl/eo) auves dude. (34) 5 Thermo-Acousties A success of Lighthill's law wns to stress the essential difference between the sound radiated by isothermal and non-isothermal free jets. This tris damental rescarch. Initially it-was expected that heat-transfor wonldl induce a nionopole sound souree. This was not found. Morfey [26] and Obermeier [32] understood that this was a result-of the lack of difference in compressibility between the cold and the hat gas in a flow where the mean pressure is uniform (see (28). The relevant sound source appears to be the force of Archimedes dine to the acecleration of fluid particles with a density deviating from the surrounding. This results into x dipole sound source whieh in Lighthill's ansl- ogy (22) can be identified as the term é (p! cla’) 83. Superficial inspection of (22) would snggest that a monopole sound source cannot be induced by a flow. This is contradicted by Rayleigh's seattering theory (28). Furthermore ‘xe expect unsteady combustion or unsteady hheat transfer from a wall to act ‘8s 8 monopole sound source. The key issue is that the analogy is not unique. In particular we have presented an analogy in which the acoustie field expressed in terms of density flucthations 0’. It is possible to use the pressure tions 7/ as acro-acoustical variable, We simply acd to both sides of (13) the term (1/c3)(0%p'/@t2) and obtain: 10 A, Hirschberg, C. Sebran which of commse is ag exact and as useless as Lighthill's formulation ( as long as we do not introduce any approximation, I we neglect convective offects we ean write in linear approximation: (36) The entropy equation (12) provides then a link between this sound souree term and heat transfer q or heat release by combustion Qu. It appears there- fore that using p' as aero-acoustical variable is most convenient when we sindy thermo-aeoustieal phenomena (23), [4). Seeking for an optimal choice of the aero-acoustical variable for a particu lar application is one of the concerns in aeroacoustics ([6, [30)). This problem is particularly acute for supersonic flows in which the flow is not compact. 50 that convective effects on wave propagation are important. Making a choice of a seroscoustic analogy providing separation of “source” from “convective effect” is very difficult [10], [23)). 6 Acoustical Energy When considering linear perturbations of « uniform stagnant reference fluid wwe can easily derive an energy equation for the perturbations. We add the product of the mass equation (18) by p'/g to the sealar product of the momentum equation (11) with the velocity vf. We find (37) and the intensity J; is defined by: hap In integral form and averaged over a period of oscillation for periodic per- lurbations we find for Une avoustieal power < Pye > produced by the flow: eto [ctm sas [

= fi as~ f Before proceeding we should propose a model for the heat transfer from the gauze to the ambient air. We use Newton's approximation for the heat flux da = Gm, where ne is the unit normal to the surface of the ganze: an 9 = aT —T) (42) where @ is the heat transfer coefficient, Ty is the wall temperature and the ambient air temperature at some distance from the wall. We assume that Ty is constant due to the thermal inertia of the gauze. The ambient temperature T is assumed to be modulated hy the adiabatic compression corresponding to the acoustic pressure fluctuations p'. ‘The heat transfer coefficient is de- termined by the flow through the gauze. The steady component of this flow ‘with a velocity Up is due to the upwards draft of the hot air in the upper part of the Rilke tube, which acts as a chimney, In addition to this we have ‘an acoustical modulation of the velocity which isin first-order approximation related to the acoustical pressure fluctuations by the linearized momentum ‘cquution (19) We follow Rayleigh [38] by considering first the behavior of the system when the gauze is placed at the middle of the pipe where there is un arousti- cal velocity node. ‘The modulation of heat transfer is in that case due to the adiabatic temperature change 7” associated with the acoustic pressure flue- tuations p', This tends to reduce the heat transfer rate gy when the pressure pf increases. If we neglect memory effects the correspond ‘opposite in phase to the pressure fluctuations. Following (41) we expect that the gaue will only absorb acoustical eneray. This seems to be confirmed by experiment, The pipe does not sound when the gauze is placed in the middle. Tt seems logical to assume that the modulation in heat transfer ais due to @ modulation of the heat: transfer cvefficient by the acoustical velocity fluctuations u’. ‘The largest fluctuations are found near the open pipe termi- nations. As those positions correspond to pressure nodes p= 0, the source cannot transfer any energy to the acoustical resonant mode of the pipe (see (41), Hence the position haliway between the middle of the pipe and an open end is a. compromise. At this position we have both acoustical pressure and velocity luetwations Until now we have followed Rayleigh’s discussion [38]. We want now to understand why the gauze does sustain the pipe oscillation when placed at the lower part af the pipe but does not do it when it is placed at the upper part. At this point the discussion of Rayleigh is not very clear and we propose an allemative, The first observation whieh we should do is that if the heat transfer coefficient would react instantancously to velocity fluctuations the heat transfer would be 1/2 out of phase with p’, which is a direct consequence A Brimitive Approach to Aerowoonstios 13 of the momentum equation (19). If p! has a ces(2x/t) time dependence, u! wil] have a sin(2z ft) time dependence. Hence a quasi steady theory for the response of the heal. transler cannot predict any’ acoustical energy transter to the scousticel field, The average over au acoustical cecillation period vanishes RY cos(2n ft) sin(2aft)dt = 0). A plausible explanation is that the heat transfer has some memory time 7. I we assume that the heat transfer is determined by thin boundary layers between the gauze and the surrounding, an increase in velocity will reduce tho boundary layer thickness. However the remaining hot fluid from the old? boundary layer hiss to be evacuated by convection which will Last for a time + of the order of magnitude of ((A0)) rat (a3) As shown in Pig. 4 the time delay between pressure fluetuntions and velocity fluctuations for a gauze placed at x ~ L/4 is @ quarter of an oscillation period 1/(4f) so that optimal sounding conditions are found for fd Ue (as we L!2,1) alot) wes LIA Post Fig. 4. Time dependence of the pressure, acoustical velocity and ofthe beat transfer a gauze placed in the lower hall of a Rijke-tube. A gauze placed in the upper part of the pipe will experience # velocity fluctuation opposite in phase to that found at the lower part of the pipe (Fix. 3). The pipe aspires and expires from both open ends. This can also be verified mathematically by caleulating the velocity field corresponding, to a pressure distribution of ~ psin(2nfz/¢o) cos(2n $1). ‘This implies # Line lag. of 3/4f. Hence for # gauze placed at the upper side of the pipe (2 = 3L/4) the fiogr velocity needed to sound the tube shonld be three times lower than for the gauze placed al the lower part. In such a case the therm boundary layer around the gauze are thicker, which reduces the heat transfer and the sound source. While it is not impossible to make a Rijke tube sound with a gauze 1A. Hirschberg, © placed in its upper part, this appears not to be possible for the particular ‘conditions considered in our experiment, Please note that in our diseussion we used @ linear theory. ‘This implies thal the acoustical power delivered by Lhe source will be proportional to Ute square of the acoustical amplitude because Qy, is proportional to u’. As the losses are also proportional to the square of the aeoustieal amplitude, such a linear theory cannot predict a finite oscillation amplitude. finite amplitude is established as a result of non-linear effects. In the ease of the Rijke tube, itis believed that when acoustical particle displacement, become of the orler of magnitude of the gauze wire diameter d the hot ait from the woke of the fuze flows back tothe gauze during part ofthe osilaton period [11]. Tl blocks the heat transfer. The amplitude ofthe parce displacement is piven by ul/(2mf). Because at the grid postion =~ L/4 we have wa (a5) ro ¢ we find fa (48) oa ‘To our groat suxprise this is verified by experiments, 8 Vortex-Sound Theory Vortex-sound theory was developed by Powell [36] for subsonic flows in free field conditions. Howe [17] realized that such an clegant. formalism was prab- ably such more general, atid proposed a generalization to internal flows and arbitrary Mach mmbers. In analogy to Prandtl’s: “Vortiees are the muscles. of the flow” we have the Miller and Obermeier [26] statement: *Vortices are the voice of the flow". The power of the “vortex-sound” theory comes ftom two facts: + the fret one is that we have an intuition for vortex dynamics because of ‘the conservation of the cizeulation of vortices in frictionless flows. The study of the movement of vortiees eorresponds to a kinematic problem (application of the Biot-Savurt induction lave [44}, which is much easier than the solution of the dynamical prablem. This is quite similar to the role of vorticity when explaining the lift of an airfoil. By using a Kutta condition for tangentiel flow separation at the trailing edge of the airfoil, the problem of calculating the flow becomes purely kinematic. The lif is thon deduced by application of the Bernoulli equation to calculate the prescure distribution on a control surface. In vortex sound theory, we also assume some existing, knowledge of the vortical distribution in the flow and deduce the aero-acoustical sound production from this knowledge A Primitive Approach to Aeroacousties 15 «the second fact is that in common flows the region over whieh the vorticity non vanishing has usually a spatial extent much smaller than does the region where Lighthill's source is non vanishing. It is therefore much easier to bound the flow regions that are significant to sound production using vortex sound theory than with the original Lighthill's formulation, ‘The derivation of vortex sound theory, as initiated by Powell ([36)) follows. tho same philosophy as tho one that Lighthill applied in establishing his analogy. It consists again in rearranging the equations of uid mechanies in ‘single inhomogeneous wave equation, the right-hand side of which being considered as the sonree of sound. The difference with respect to Lighthill's Aappronch stands in the use of the veetorinl identity (5 ww?) (w¥)v- (Tx axe any to introduce the vorticity w = Vx v. Using (47), in absence of external forces and neglecting the effect of viseesity, Uke momentuin equation (6) can be rewritten: 8 gov (2p) Hole x +6r-0 «as a Using the definition (15) and (16) of the perturbation density and pressure, and taking into aecount the fact that go and pp ate constant, we can op Pd dy’ Oa? OE Det a For an isentropic flow with uniform entropy (homentropie), the speed of sound is defined hy (20) and (49) ean be written: Ay ag oe Oat We inject the continuity equation (3) and the momentum equation (48) in the left-hand side of (50), to end up with & formulation equivalent to Lighthill's analogy (22) oe eo (ote oy)» Ze (Lote) ‘or homtentropie flows, the relative variations of the density are purely kinetic so that the four terms on the vight-band side of (51) acale as 1: 1: Sra M? ((26}), where Sr is the Stroulsal number and M is the Mach number, The 16 A, Hirschberg, C. Schram last two terms can therefore be dropped for compact flows (M <1) with a Strouhal number Sr = O(1), and we finally have. (2 ‘The contribution of the second term is negligible at low Mach number for ‘& compact source region, and because of conservation of the kinetic enersy in a nom-viscous flow ([s]). Another way to get rid of this second term is to include it in the aero-scoustical variable, which beeomes then the stagnation enthalpy {[17]). as will be shown in Sect. 10 Retaining the first source term of (52), we find a dipolar nati source A(o(w x v)}/Ax, compared to the quadrupole of Lighthi /84r,04;. We have seen that a dipole in Lighthill’s analogy (22) is induced by external forces. If there is no external force then the acoustical field produced by the first source term of (52) should have at most a quadrupole as leading order term. We will now illustrate how, using the free field Green's fmetion and taking the variation of the retarded time over the compact souree region into eccount, we can recover the expected quadrupolar acoustival field, We obtain an integral formulation of (52) using the seme free field Greens function Gp (B1) as in Sect. 4, and using its properties of symmetry {32}: and performing the integration in time gives S(t = 7 ~ |x = wien) ~aage—w ow x v)edydr (53) ele xv), | ie ote, 2am | es i whore f= ¢~ la wljep is che retarded time, Assumn and that in the fat field we can make the simplifications a compact soures |z—y|~ 2] and Where the integral vanishes, owing to the conservation of the impulse in ab- sence of non-conservative external forces ([44]). Clearly, if our aim was to A Primitive Approach to Aeroacousties AT predict the radiated sound field, something has been missed. A more sub- ‘le approach consists in keeping the first ewo terms of the Taylor expansion of the retarded time over the source region. In the far fleld, we make the approximation zy Jz — ul = ta Fel and a function g of the retarded time ¢ ~~ yl/oy can be approximated at first order by: _le=vl)_ | (,_l2!), 292 o(e-Bat) =9(-3) + Shae We have also the plane wave approximation for the far field derivative: al nae and (54) becomes dle) =~ at (Plea ty agate [lise Nae (66) chere, as before, the first integral vanishes in virtue of the conservation of the impulse (44) Following the same arguments that led to (34), we make the expression (56) explicit by approximating @ = a aud find Poseell’s integral solution oad) [stom eae ay (7) which corresponds to a quadrupolar radiation field. The diseussion above stresses that in spite of the dipolar character of the first source term in the wave equation (52) its sound radiation is quadrupo- lar because we explicitly disregarded the dipole contribution and used a suffi ciently accurate sories expansion of the retarded (ime to retain the quadrupole term. 9 Choice of the Green’s Function In the previous paragraph we limited our discussion of vortex sound to the fee field conditions, The theory can be applied to confined flows. In such ‘a case we should reconsider the choice of the Green's function. The Green's 18 A, Hirschberg, C. Schram Function is d term: + as a solution of the wave equation with an impulsive source 86 286 . Fe a a TE ev) (58) This definition should be completed by initial and boundary conditions. The initial conditions follow irom the causality condition ([28)). ‘The bou conditions can actually be chosen arbitrarily. One can for example use the free space Green's fimetion Go (31) in the presence of walls. In that case suriace terms appear in the integral equation which take reflections of the waves on the surface into account. For compact objects in free space (small compared to the wavelength). the physical interpretation of those surface integrals is quite straightforward. This corresponds to the so-called Curle’s integral formulation ({5}). This allows a.simple derivation of Gutin’s principle for compact rigid bodies, relating the foree F exerted by the flow on the body to the radiated field in far field approximation (A): Ho.) tee (1-H!) eet) ~ ae (: a ‘whore we noglestod the sound produced by the Lighthill quadrupole in the bul of the flow An alternative procedure would be to consider a Green’s function which satisfins exactly the same boundary conditions as the “acoustical fie” on the surface. This lends to a tailored Green's function. For a rigid body the nor- ral component of the flow at the surface should then match the wall velocity. In that case surface terms only appear as a result of viseo-thermal effects Friction and heat conduction are often negligible, so that when the object is compact the flow can be approximated by a locally incompressible potential flow around the object, Using the reciprocity of the Green's function, Howe {{I7)) shows thnt this Groen's function can be calculated in far field approx imation from the incompressible potential flow induced by « uniforan flow in the direction of the observer Ina confined flow such as a clarinet we can have a similar foedom in the choice of the Green's function, It is often said that a clatinet behaves as closed-open pipe. This means that the internal aconstical field is dominated by the resonant modes of such a closed-open pipe model of the clarinet ‘Thorefore it may be convenient to develop the Groen’s function in terms of these pipe modes in which se neglect friction and radiation losses. When pline wave propagation prevails in the pipe this is simply a spatial Fourier ‘Analysis of the distribution of the amplitude of the acoustical pressure along, she pipe, ‘The sources of sound in such # model are 8 monopole representing ‘time dependent air fow passing through the reed opening into the pipe tho viseo-thormal losses at the walls and a rndiation monopole at the open Fig. 5 we give a schematic representation of the elarinct in which we indicate the monopole Qy, at the reed end of the instrument which i A Primitive Approach to Acroaoonstios 19 the analogy is represented as a elosed end. OF course if the reed end would actually be elesed we could not blow into the clarinet. The monopole Qn can inelude any model of the reed, player's mouth and respiratory system: Cflien one uses a very simple model in which the player's mouth pressure is assumed to he constant. This simplistic model can lead to quite an insight 4 not be expected to explain the clarinet behavior in all playing: conditions. When this model fails, heeause the mouth acousties is significant one could consider the use of a more extended “tailored” Green's funetion. Such a Green's function would include both the pipe of the instrument and the player's respiratory system. In such an appronch the source of sound is no longer a monopole because there is no significant, change in fluid density whea the flow passes through the reed. The sound snuree can be identified as being the dipole due to the vortex shedding in the free jet formed by flow separation al the reed, The one-dimensional representation of this fluctuating, dipole ‘would he a. flretuating pressure difference between the mouth aud the pipe in an acoustical model (ignoring further any fluid flew or wall movement). @) festa state enennenSHI bout shor (ey linear oN Fig. 5. Acroacoustic analogy in which the clorinet (a) is represented as 9 elased pipe (b). The monopole Qy. cau include the modelling of the reed, player's mouth and respiratory systers, souree 0, In the choiee of the optimal Green's funetion we ean therefore use some physical intuition. In some cases Uhe most efficient. choice of the Green's function is however dictated Ly numerical considerations, The development in modes of closed-open pipe without friction is an excellent. approach in the frequency domain when for example we consider steady oscillation of a clarinet (a sustained note). In inusieal acoustics the attack transient is essential for the perception of the sound ([9}). In dynamical models used for real-time sound synthesis it appears that the optimal choice of the Green’s Function is that of a semi-infinite tube (Lertninated by the elarinet pipe). This 20 A Mirsehberg, C. Sebnram. choice avoids the multiple reflections in the Green's function which wonld give 1 very slow numerical convergence for time domain calculations ([24}) 10 Howe’s Energy Corollary We will explicit an energy corollary which has a limited range of validity, but provides an almost intuitive understanding of the interaction between vortices and an acoustic field. This energy corollary was proposed by Howe ((15},[16)) The key of Howe's approach is to define the acoustic fied by consideriny a Helmholtz. decomposition of the volocity field, v, into tho gradient of as potential, e, and the eur! of a stream funetion, ¥ vaVer Tx (60) The acoustic velocity wis the gradient of the unsteady component of the scealar potential wavy" (6) As a next step we consider Croceo's formulation of the equation of Kuler for ‘a homentropic (constant and uniform entzopy) flow 2 ope mw wrt : Bavaa—wxert (62) schere B is the total enthalpy dp jel? = [2 lee (03) a- {249 (63) ois the density, p is the pressure, w = Vx v is the vorticity and f is the density of an extemal fores. When we have a potential flow (w = 0) without external force (f = 0) we are left with an acoustic flow superposed on a steady potential flow. This indicates that the Coriolis force: fi acts as an external force on the acoustical flow field a, We also see that in a homentropic flow the (otal enthalpy, B, is natural avoustic variable. If wwe neglect wave-propagation effects in the souree region (where w #0), and _mean-flow convection in the wave region (swhere « ~ 0) we can use Kirchhoft's definition of the acoustical energy. In such a vase, the lime average

fof the power transter from the vorticity field to the acoustical field is simply the tine average of the integeal of the work performed by the Coriolis force on the acoustical field: we xv) (oy (65) A Primitive Approach to Aecoavonsties 21 Where V is the source volume (where a 7 0) and the brackets <> indicate a time average. In the case of a 1-D pipe flow (outside the source region) we can represent a compact: vortical source region as a time-dependent discontinnity AB, in the total enthalpy. This source is calculated by subtracting from the actual enthalpy difference, AB, the difference ABrer which is induced by a potential flow across the Source region. In the ealculation of AB and AByoe ‘we should use the same mess flow ® through the pipe termination. We have therefore: AB, = AB ~ AByoe (68) The corresponding energy corollary i

= < bm AB, > (67) Both the original equation of Howe and the 1-D formulation stress that vor- tices are assoviated with foree or dipole source of sound, which can only transfer energy to an oscillating acoustical flow when there is ant acoustical velocity at the source (u # 0). This corresponds to the fact that a violin string is more easily bowed in the middle of the string where there are large string displacements, than very elose to the bridge where the displacements. are much stnaller. This is the complement to the situation encountered in the Rijke-tube, In tliat cage a volume source could not perform any work when placod at a pressure node, close to the open end of the tube: 11 The Open Pipe Termination of an Unflanged Pipe Let us consider a thin-walled pipe with a mean-flow velocity Uy directed to- ‘wards an unflanged open pipe termination. The pipe eress section has a radius a. The speed of sound, ¢9, ig uniform. We will diseuss the reflection at the pipe termination of harmonic waves with a low frequency, J, travelling down the flow in the pipe. These acoustical plane waves induce (just) upstream of the pipe termination a uniiorm acoustical velocity & = (ut.0,0) along the pipe axis. As we eonsider low Helmholtz numbers, He = (af/eg) 1, tho acoustical flow at the pipe exit is a locally incompressible potential flow as sketched in Fig. 6. Near the sharp edges of the open pipe termination the potential flow is singular, which implies very large ratios of the local aconsti- cal velocity Ju| and the acoustical velocity ue al some distence upstream of the pipe termination. At a distance of the order of the pipe radius « (dewn- stream of the pipe outlet) the acoustical flow is alteady close to a spheri symmetrical flow with a velocity decreasing quadratically with the distance from the origin (locally incompressible flow). In the absence of mean flow, the inertia ofthis acoustical ow just outside the pipe induces a difference in tho scalar potential, which is often represented 2s an end correction (35). In the presence of a mean flow Up, leaving the pipe, # Tree jet is formed. ‘The shear layer separating the jet flaw from the environment is a region of 2 ©. Schram ef o 2 uv vo © Fig.6. Acoustically forced flow at an unfianged open pipe exit. a) actual vorticity distribution in the flow (PIV measurements) b) modelled flew ¢) potential flow high vorticity, w. At low Stroubal numbers, Ste ~ (fa/l’e) 1, we ean assume & quasicstationary response of the fee jet to the acoustie flaw. All the kinetic-energy flux of the flow, Py. }(Uo + te)” is dissipated by turbulence in the jet while the pressure in the jet is equal to the stationary pressure of Ue surrounding, (as for any stationary subsonic free jet). Comparison of this flow with a potential flow implies that AB, — —4(Ug + ua)?. In a linear appreximation we therefore find

= —ellona® < uf >. Using this restil to calculate the reflection coefficient. Rp — |B /B' [for an incident ‘wave of amplitude B* yields exectlent results (see Fig. 7). ‘This provides a vortex-sound interpretation of a wel It from the literature: Rp = (co ~ Uo}/ (co + Uo). Mt higher frequencies, when Sry = fa/L’a = O(1), it is not obvions that acoustical energy will be absorbed upon reflection a. an open pipe termina- tion. ‘The key to understanding sound produetion is the fact that the acons- tical velocity w changes sign from pipe outwards to pipe inwards each period of oscillation. We new focus on the vorticity perturbation generated hy the A Primitive Approach to Aerowoonstios 23 ' Teo} —— penne 6 095 08. 08 on 07s or 0 oc 00s 00s oo a; oi ow M Fig. 7. Reflection cocllicient Ruy = |B /B*| ot an unflanged open pipe termination 8 # function of the mean flese Mach number IM = Coe. acoustic flow at. certain moment of the period in the shear layer of the free jet. This vorticity perturbation w” is generuted at the edge of the pipe exit ‘where flow separation occurs. ‘The vortieity perturbation is conveeted away with the fluid in the shear layer at about, half the mean flow velocity Uo, As illustrated in Fig. 8, the sign of the vorticity is such that it: induces. « flow at the edge that: opposes the driving acoustic flow around the edge. ‘The amount of vorticity shed is such that taagential flow separation is maintained (Kutta condition). Inspection of the sign of w, w and v indieates that at the moment the vorticity is generated (f.,.1u) = —o(w! x v).a < 0 which im plies following Howe"s forms (65) that there is a transfer of energy from the acoustical flow to the vortical flow. We eecept this result easily. ‘The acousti- cal flow generates « vortical perturbation, and should pay a price for it. We also see that the interaction is strong beeanse the convective velocity w is normal to the local acoustical velocity’ w. Furthermore, due to the singularity of the acoustical (potential) flow we have locally relatively large acoustical velocities at the sharp edge. ‘The big surprise comes from the fact that both w’ and v do not change sign while w does change sign. Hence, afier half n period (Y= 1/f) of a cillation the perturbation of vorticity which we have focused on produces acoustical energy because (f/.2) > 0. As we travel away from the edge, v retains approximately constant while w and the angle between w and & both decrease with the distance from the pipe outlet. This implics that if! remains constant, the production of acoustical energy will not be sufficient to obtain a net produetion of acoustical energy’ after one period of ose tion. The mwan power,

, would thus be negative. Iu practiee, however, te initial vorticity perturbation initiates an instability of the shear layer ‘This results in an increase of the magnitude of the time-dependent vortic- ity perturbation. At low amplitudes, when linear theory esn be applied, the 2A, Hinsehberg, C. Sebram. Fig. 8. Vorticity perturbation yencrated at the edge at the moment iti generated ‘and half # period of oscillation Inter. perturbation ine ine). Tn may practical conditions the amplitude, u/Uy > 10 lange to use a linear theory. A reascx layer in terms of discrete vortex shedding. It appears from experiments thal 2 now vortex is shed at eseh instant for which the acoustic flew turns from pipe inwards to pipe outwards. The vortex structure accurnulates ll the vorticity shed at the sharp edge while it travels parallel to the pipe axis with a velocity of the order of L/2. "The magnitde cf the civeulation [P| of the vortex inereases about lineatly in time dl /dl = U/2. This is much weaker than the exponential growth predicted by linear theory. This limited growth of the vorticity petlurbation inuplies @ net sound absorption for an unflanged pipe termination (34) We should note here that a free vortex cannot have « time dependent eit calation ({17), [44)). Assuming such a Gime dependence introduces 2 sp force on the flow. By using, Howe's analogy we simply ignore the acoustical cficet of this spurious force (4). Furthermore, Howe's energy equation is also su integral formulation whieh will tend to smooth aut c se exponentially with the distanee from the edge ([39], is much too le alternative is to desetibe the shear A Prin ive Approach to Actoaoonstios 25 12 Whistler-Nozzle and Human Whistling The analysis of the acoustical response of an flanged open pipe termina lion has indicated that in the presence of a mean outflow, Uo, the net sound absorption is the result of a combination of an initially large sound absorp- tion upo followed by sound produetion half » pe If we could reduce the initial absorption and increase the later produetion, ‘we can expect that a net sound production is possible (in order to reduce sclE-sustained cseillations we can use spoilers to inerease the initinl sound ab- sorption, [8)). ‘The whistler-nozzle configuration shown in Fig. 9 satisfies both ‘these requirements, and does produce sound at specifi Strouhal numbers [12]. ‘The sharp edge of the unfianged pipe termination of Fig. 6 is replaced by vortex shed od Inter Fig. 9. Whistler-nozzle configuration, stepwise increase in pipe radius from a to «4 ft at a distance w from the end of the pipe. We call this pipe segment (of length w) the “cellar” of the nozae. Vortex shedding is now initinted atthe edge of the step, where the acoustical flow presents « 90° sharp Lend rather then the eriginal 180° shaxp bend at, te edge of the open pipe termination. his already considerably reduces the singularity of the sooustical flow at the separation point. Furthermore, the collar bends the aconstieal streamlines back in the direction of the pipe axis ‘which corresponds to the direction of convection of the vortices. This implies that within the collar, the i vortices is weak (f¢-u = 0). The absorption is reduced to # small interaction rofion near the separation point. A\ strong interaction between the vortices and the acoustical flow ovcurs again at the end of the collar beeause the diverging, acoustics] streamlines ut through the path of the vortex. If the travel time of vortices along the eollar is of the order of one half eseillation, od, we do expect # net produetion. This corresponds to a critical value Sr¢ = wf Up of the Strouhal number. Increasing the step sie h reduees the interaction of the vortices with the raction betwoen the acoustical field and the mages in the wall of the collar. As these 6 , ©. Schram images slow down the vortex convection, we observe that. the ertical Strouhal number Sre increases with h ([18},{13) Ieappears that also higher-order hydrodynamic modes are possible, cor- responding to larger Lravel times of the vortices in the collar, This implies the presence of more than one vortex at the time inthe collar. Up to four vortices (n= 4) inve been obwerved. F es (01 = 5) soon fo be an upper limit for the number of vortices observed in many self-sustained oscillations ((42} (a3), 2) Th the original papers on the whistler-noazle it was argued that the sound production was due to the impingement of the oscillating shear layers ofthe free jet on the edges ofthe pipe termination at the end ofthe collar ({i8) (43). Vortox-scnnd theory indieates that such an impingement és not necessary fo sound production, ‘This is confirmned by the fact that rounding the edges a the end of the collir increases the sound production (20%) rather then reducing it [13]. This leads us to the idea that we muy not need any sharp crlge at all, as long as flow separation occurs at a reasonably fixed potat Removing all edges leeds us to a hom, as shown in Fig. 10. Experiments by Powell [2], Peters [24] and Hofmans [14] do indeed confirm that # horn can produce sound nt critical Stroubal uumbers. ‘This is illustrated by the experimental data on the nconstieal power produced by a hom as n faction of the Strouhial nuraber shown in Fig, 11, Strong selFsustained oscillations of Fig. 10, Horn at the end of a pipe. the flow in a whistler-nozzle appear when the acoustical energy accumulates in standing waves of a resonator attached to the nozzle. In such eases the acoustienl resonator imposes (he oscillation frequency, J, and the pulsations: show # maximum when the flow is adjusted to meet the critical Strouhal ion (s0e the previous section). Such self'sustained eseillations ave been olserved when a pipe seument (length L) terminated by a whistler- A Primitive Approach to Aerowoonstios 27 nowzle is placed downstream of a settling ehamber ({12}, [18], [13)). When the first pipe mode is exeited (f — ¢/2L) one observes acoustical velocities at the nozzle of the order of magnitude w/Liy = O(0.1). When a resonant: closed side branch is placed upstream of the pipe segment pulsation amplitudes of the order of magnitude u/Up = O(0.4) are observed |13}. In this ease, the acoustical energy is produced both at the Whistler nozzle and at the junction between the clesed side branch and the main pipe. The mest common example 0015 oot Aye ons * a 00s 001 02 03 03 04 O45 05 Sr Fig. 11. Acoustical power produced by a horn as a function of the Strouball number. of such self-sustained eseillations is human whistling. Following Wilson [48] human whistling is due to a coupling between vortex shedding, at our lips and the Helmboltz resonance frequency of our mouth. This i ihustrated in Fig. 12. The Stronhal condition for optimal whistling is easily observed by first fixing the geometrical configuration (lip and tongue positions) and then giadually increasing the volume flow, A certain critical flow velocity has to be reached before whistling appears. Above a certain flow velocity, whistling disappears again. Young children often blow too hard when they tiny to whistle... ‘This disappearance is due to the fact that at high flow rates the vortices are convected far outside the lips (horn) before they ean produce sound. When performing this experiment we also observe that the system does not oscillate exactly at constant frequeney. ‘The system tries instead to keep the Stroubal number as elese as possible to the optimal Stroubol mimber by increasing gradually the oscillation frequeney as we inerease the ow velocity. Large variations in oseillation frequeney ate only: possible when the acoustical mode has a low quality factor, Q (Q/2n is the number of oscillation periodsafler which, in the absence of « sound souree, the amplitude of s mode has decayed by 2 factor exp(0.5)) 13 Conclusion We hope you enjoyed and that you will use the aeroacoustieal anslogies. 28 A. Hinvehberg, C. Seluram Fig. 12. Human whistling is induced by the coupling of vortex shedding st our ips, along with econstical oscillations of our month. The lowest resonance frequency corresponds to the Helmholtz resonance, at shich the mouth volume acts as an acoustical spring and the norrowing at the lips acts as an acoust References 1. CK, Batehelor: An Introduction fo Fluid Dynamics ( Press, Cambridge 1967) 2. W.K. Blake, A. Powell: “Phe development of contemporary views of flow-tone generation”. Ta: Recent Advanees in Acroncousties, ol. by A. Krothapalli and CA. Smith (Springer-Verlag, NY 1986) pp. 248-25 3. J.C. Bruggeman, A. Hirschberg, M-E-H. van Dongen, A.P.1. Wijnnndss J. Sound Vib. 150(3), 871-398 (1991) 4. D.G. Crighton, A.P. Dowling, JI, Ploses Willisins, M. Heckl, F.C. Leppington: Modern Methods in Anakyticel Acoustics (Springer Verlag, London 1992) 5. N. Curle: Proc. Roy: Soe. (Lenéon) Ser. A 264, 821-842 (1969) 6. PE, Doak: Acoust Phys 44, 677-685 (1995) 7. AP. Dowling: and LB. Flows Williams: Sexnd and Sources ef Sond (Mis Horwood, Chichester 1983) 8 ALT, Fatorchenko: J. Sound Vib. 282(4), 719-782 (2000) umbyidge University A Primitive Approach to Aeroacousties 29 9. NH. Floteher and T-D. Rossing: The Physies of Musical Tastruments (Springer, New York 1908) 10. ME. Goldstein: Aeroacoustics (MeGraw Hill, New York 1970) IL MAA, Heekl: J. Sound Vib, 124, 117-183 (1088) 12, W.G, Hill Jr., PLR, Greene: Trans, ASME J, Fluids Eng, 99(8), 520-525 (1977) 13. A. Himchborg 1. Braggeman, A.P.J. Wijnands, N. Stnits: Acustica 68, (57 160 (1989) 1M. G.C1. Hofimans: Vortex Sound in Confined Plows, PhD ‘Thesis, Technische Universiteit Eindhoven, Eindhoven (1008) 15. M.S, Howe: J, Sound Vib. 70, 407-411 (1980) 16. MS. Howe: LMA. J. Appl. Math, $2, 187-208 (L081) 7 MSS. Howe: Acoustics of Fluid-Structure Interactions (C Pres, Cambridge 198) 18. A.KMLF, Hussain ond M.A. Hasan: J, Fluid Mech, 184, 431 (1983) 19. A. Tofrati, G. Riccardi: J. Sound Vib. 196 129-1:6 (1996) 20. LD. Landau and EM, Lifehita: Course of Theoretical Physics Volume 6: Flvid Mechanics, 2nd edn. (Pergamon, Oxford 1987) 21. MA. Lightill: Proe. Roy. Soe. A211, 561-587 (1952) 22, M.J. Lighthill: Proe. Roy. Soc. A222, 1-02 (1954) 28. G.M. Lilley: “Jet Noise Classical Theory and Experiments’. In: Aeronconstirs of Flight Vehicles, Theory and Practice, Volume 1: Noise Sources eal. by HH. Hub- bard (Acoustical Society of Amerioa, New York 1995) pp. 211-280) 24. MAE, Melntyze, RT. Selumacher, J. Woodhouse: J. Acoust. Soo, Am. 74, 1925- 1545 (1983) 25. HJ. Merk: Applied Sciontifie Research AG, 102-120 (1956-19 26, CLL, Morley: J. Sound Vib, 81, 391-397 (1973) 27. CLL. Morley: J. Sound Vib. 48, 95-111 (1984) 28. PM. Morse, H. Foshbach: Methods of Theoretical Physics (MacCraw Hill, New ‘York 1953) 20. EA, Miller, P. Obermsict: Fluid Dyn, Res, 8, 43-51 (1088) 30. RE. Masafir:‘A Discussion on the Structure of Acroacoustic Wave Equations’ In Proceedings of the 4th Congress on Acoustics, Marseille, France April, 1997 (Tokanca, Toulouse 1997) pp. 023.926 SI. MLK. Myers: J. Sound Vib. 109, 277-284 (1986) 582. F. Obermeior: J. Sound Vib. A, 463-172 (1975) 38. F. Obermeier: J. Sound Vib. 99, 111-120 (1985) 81 M.C.AM Peters, A. Hirschberg, A.J. Reijnen, A.P.J. Wijnands: J. Fluid Mech, 256, 499-534 (1903) 35. A. Pieroe: Acoustics, An Introduction to its Physical Principles and Applica tions (Acoustical Society of America, New Yark 1980) 36. A. Powell: J. Acoust. Soe. Am, $6(1), 177-195 (1964) ST. A. Powell: Tyans. ASME, J. Vib, Ac. 112, 115-159 (1000) 38. J.WS. Royleigh: The Theory of Sound (Daver, New York 1945) 89. SAW. Rienstra: J. Fluid Mech, 108, 443-460 (1981) AD. SAW. Ricnstra: An Tnoduction to Acoustics (Report TWDE 01-08, ‘Technische Universiteit Binghoven, 2001) AL, P.L, Rijke: Annalon der Physile 107, 850-545 (L850) 42. D, Rockwell, B. Naudasher: Trans. ASME, J. Fluids Eng. 100, 152-165 (1978) 15. D. Rockwell: ATAA Journal 21(5), 615-664 (1083) bridge University 30 A Mirsehberg, C. Sebnrama AM. P.J. Salfman: Vorter Dynamics (Cambridge Monographs on Mochanies and Appliod Mathematics, Cambridge 1992) 45, C. Schatn, A. Hirschberg: to be submitted AS. C.K.W. Tami ‘On the failure of the acoustic analogy theory to identify the correct ise sources’. In: Proc Th AIMA/CEAS Aeronc. Conf., Maastricht, 28-0 Moy 20001 AT. P.A, Thompson: Compressible Fluid Dynaries (MacCraw-Hill, New York 1972) 48, T.A. Wilson, G.S. Beavers, M.A, de Coster, D.K. Holger, D.Regenfust: Acoust, Soe. Am, 80, 358-72 (1971) Lectures on the Theory of Vortex-Sound Michael S. Howe Boston University, College of Engineering 110 Curnmington Street, Boston MA. 02215, USA. Abstract, ‘The theory of vortex sound is introduced, From Lighthill's aeonstic analogy, it is shoxn how vorticity and entropy fluctuations ean be seen as sousoes. ‘of sound. The use of the compact Green's functions is introduced to compute the vortex sound, As an example of the method presented, Uhis theory is applied to pressure transionts generated by high-speed trains. 1 Aerodynamic Sound ‘The sound generated by vortieit hounded fluid is called aerodynamic sound. Most unsteady flows of technological interest are of high Reynolds number and turbulent, and the acoustic radiation is a very small by-product. of the motion. The turbulence is usually produced by Quid motion relative to solid boundaries or by the instability of free shear layers separating a high speed flow (such as jet) from a stationary environment. In this lee- ture the influener of boundaries on the production of sound as apposed to the production of vorticity will be ignored. ‘The aerodynamic sound prob- Jem then reduces to the study of mechanisms that convert kinetic energy of rotational motions inte acoustic waves involving longitudinal vibrations of fluid particles. ‘There are two prineipal source types in free vortical flows: a quadrupole, whose strength is determined by the unsteady Reynolds stress, and a dipole, which is important when mean mass density vatiations occur within the souree region, 1.1 Lighthill's Acoustic Analogy (1952) The theory of aerodynamic sound was developed by Lighthill, who reforma- lated the Navier-Stokes equation into an exact, inhomogencous wave equation ‘whose source termes are important only within the turbulent (vortieal) region. Sound is expected to be such @ very sinall component of Uhe whole motion that, once generated, its baek-reaction on the main How is usually negligi- ble. Tu a first approximation the motion in the source resion may then be determined by neglecting the production and propagation of the sound. This ‘would be inappropriate if the Mach number 7 is large enough for compress- ibility to be important in the source flow, when the source flow is coupled to ‘9 resonator, such as an organ pipe, or when bubbles are present in the ease of liquids, Y, Aumgan, A. Maur, Pagnoux, J.-F. Pinton (ide): LNP 388, pp. 81-123, 2068, © Spring Verlag Dorlp ilsdelbone 2002 32 MS. Howe Consider the sound generated by a finite region of rotational flow in an unbounded fluid at rest at infinity, Let us compare the equations for the density fluctuations in the real flow with those for an ideal, linear acoustic suedium which coincides with Ue real fluid at large distances from the sources. ‘The difference between these equations will be shown to be equivalent to a. distribution of sources in the ideal acoustie medium, whose radiation field is the same as that in the real flow and may therefore be calculated by the methods of linear asoustics. ‘To do this, body forces are neglected, and the ‘momentum equation is written in the form. png) {Ot = ~Omag/ ery, ay where my is (he morentur flux tensor ij = privy + (P — Pol bes — 15, 2) and p. Ws Py os a at respostively the dons pressure at infinity; and the visoous stress tensor By integrating over a fixed region V, it can be seen that (1) equates the rate of change of momentum in V to the action of tho pressure and vis cous stresses on lary and fo the convection of momentum seress the boundary at 8 rate determined by the Reynolds stress ovv». In an ideal, line ear acoustic medium, momentum transfer is produced solely by the pressure. ie Hay > 42, = (p— pedi = Alo — po)Oiz, where py and Co are the mean densi nd sound speed, By eliminating pry between (1) (with 4; replaced by m4) and the continuity equation y, velocity, pressure, uniform bou A= po) , Hoes) — 0 a OF, wwe obtain the equation of linear acoustics for the perturbation density 2— po. (0° /c0t* — 0") |e2(o — pe)| = 0. 8) bir the absence of externally applied forces or moving boundaries, this equ tion has only the trivial solution p— p, ~ 0, bocause the radiation condition ensures that sound waves earinot enter from infinity ‘The sound generated in the real fluid may now be seen to be exactly equivalent to that produced in an ideal, stationary acoustic medium that 1s Fotced by Ihe stress distribution Tiy = my ~ af. Tyy is the Lighthill stress tonsor ‘(8 — Po)) Bij ~ O45 ao ‘The Reynolds stress poy; is nonlinear, and is significant only within the rotational source region. The serond termn is the excess of momentum transfer by the pressure over that: in an ideal (“linear”) fluid of density p, and sound speed ¢,, This is caused by wave amplitude nonlinearity, and by mean density Tig = poiey + ((P Po) — Theory of Vortex-Sound 33 variations in the source flow. The viscous stress tensor aj is linear in the perturbation quantities, and properly accounts for the attenuation of the sound; in most applications the Reynolds number in the source region is sulliciently large that oy, can be neglected, and attenuation in the radiation zone is usually ignored in a first approximation, Lighthill's acoustic analogy equation for the production of aerodynamic sound is obtained by first: re-writing (1) as the momentum equation for an ideal, stationary fluid of density p, and sound speed ¢, subject to the exter- nally applied stress T,; (pf + 8 (cS = po}) 10a, = ~OTy fOr Blimiuation of the momentum density yo; between this and the continuity equation yields Lighthill's equation, which is the exact, nonlinoar counter- part of (3) ) (Sle nal Re ‘The problem of calculating the aerodynamic sound is therefore formally equivalent to solving this equattion for the radiation into a stationary, ideal finid produced by a distribution of quadrupole sources whose strength pet unit volume is the Lighthill stress tensor T,j The solution of Lighthill's equation should strietly be regarded as an ak ternative (integral equation) formulation of the Navier-Stokes equation, that provides « useful representation of the sound when Tij is known, Ty; ac counts not only for the generation of sound, but also for selémodutation due to acoustic nonlincarity, convection by the flow, refraction due to sound spoed variations, and attenuntion due ta thermal and viscous actions. Non- linear effeets on propagation and dissipation are usually sufficiently weak to bo neglected within the source region, although they may affect: propagation 10.2 distant-observer. Convection and refraction of sound within and near the souree flow can be important, for example when the sources are contained in a turbulent shear layer, o are adjacent to a large, quiescent region of fluid whose mean thermodynamic properties differ from those in the radin- tiow zone. Effects of this kind are accounted for by contributions to Ti, that are linear in the perturbation quantities relative to a mean background ow. ‘Thus, with the exception of flows amenable to special treatment, uhe practical utility: of Lighthill's equation rests on the hypothesis that. all of these effects, which actually depend on the compressibility of the source How. can be ignored, and that adequate predictions of the aerodynamic sound axe obtained by taking for 1, an estimate based on the equations of mation of fan incompressible fluid. This approximation is likely to be acceptable when M? «eI, and when the wavelength of the sound is much larger than the dimension of the source region. The remoinder of this section is devoted to a consideration of such eases Bo MS. Howe 1.2 Aerodynamic Sound from Low-Mach-Number Turbulence of Uniform Mean Density When the mean density and sound speed are uniform, the variations in produced by low Mach number, high Reynolds number velocity fluctuation are of order pyM?, and peiey = pot. Similarly, p — p, ~ e&(p ~ p.) (P= p.) = 2/2) ~ O(pev® MP) = pce? Thus, Ty % potvoy, When viscous stresses are neglected, and the solution of Lighthill’s equation becomes wht) 5 2 fesinloe = beled gy 85 tae ¥ ; zit eae ee i Teateram | Paveniit be vileo) @y, o chore p(x,t) = ip ps) is the perturbation pressure in the far field, Quan- titative predictions can be made from this equation provided the behavior of the Reynolds stress is known, [AERODYNAMIC SOUND 1. Acrodynamie sound in an unbounded fluid. Tho order of magnitude of p can be estimated in terms of the character- istic velocity » and length seale € (of the energy containing eddies) in the source region (Fig. 1). f is determined by the scale of the mechanise re- sponsible for turbulence production, stich a8 the width of « jet mixing layer Fluctuations in vey occurring in different regions of the turbulent flow sop- arated by distances > O(0) will tend to be statistically independent, and the sound may therefore be considered to be aenerated by a collection of Theory of Vortex-Sound 35 V,/€ independent erldies, where V, is the volume occupied by the turbu- lence. The dominant frequency of the motion ~ v/€, so that the wavelength of the radiated sound ~ £/M > € (M ~ vfep). and each eddy is there- fore acoustically compact. Hence, the acoustic pressure generated by & sin- ale eddy is of order p ~ (é/lx|}poe?M?, and the acoustic power it radiates ~ rx PP? /pace % Epak jel = Ep," M®. This is Lighthills “eighth power” law. The total acoustic power is I, © (Vo/E)(Cpoe'M®) = pov MVo/t Dimensional arguments and experiment indicate that the rate of decay of the turbulence kinetic encray ~ O(Vep.0*/0). In a statistically stendy state this must equal the rate Ho, say, at which energy is supplied to the flow by the action of external forces. ‘The efficéency Ha/Hl, with which this energy converted into sound is therefore proportional to -M®, confirming Lighthill’s hypothesis thatthe flow generated sound is an infinitesimal by-prod the motion. 1.8. Aerodynamic Sound from Low-Mach-Number Turbulence of Variable Mean Density When the mean density in the source region is not constant, the Reynolds stress quadmpoles are augmented by a. dis strength being equal to the hydrodynamie force experienced by a fluid par- ticle of density p relative to the force the same particle would experience had its density been equal to p.. This dipole is ‘hidden’ within the Lighthill quadrupole Ty and might therefore be expected to have the same radiation elliciency as the Reynolds stresses, However, the component of Tis responsi- ble for the dipole is e2(p— po), which is larger than the Reynolds stress pet by @ fector of order O(1/M®) provided the mean value of p ~ go # 0 in the "To obtain a precise estimate of the dipole radiation, the solution of (5) is written ution of dipotes, the dipole plx.t) = Are} BP fe i where the square brackets [] indicate evaluation at the retarded time f— Ix=yl/eo. ‘The density of a fluid particle changes by (p ~p.)/e? when isentropically compressed by departures of p from the ambient pressure po, where ¢ is the local spoed of sonnd. Let us assume that the motion of each fluid particle is isentropic, but that the mean density, which will be denoted by Gx. 0) varics from point to point in the source region, where (— po)/ps need not necessarily be smnall. This means that (p—p.)/e® = p~ p and Dp Dt = 0. 35 MS. Howe The Reynolds stress on the right-of (7) must he replaced by fev, and the integrand of the second term in (7) becomes aie ~( ~ ad} # {o-m( ‘The first term on the right is already in « form suitable for estimating the acoustic radiation, However, the second term would yield zero when retarded time variations are neglected over a coherent source region of constant mass. ‘To deal with this the relation D(p— p,)/DE ~ 0 is used to show that 7 Dey acre EEE (0-0 BB) + soe naive) @ (0)(P — Poldiv v) @ ‘This is further simplified by substitution from the momentum equation Dvy{Dt= (-La)ep/Oery, and from the continuity equation div v= (-1/pe!)Dp/ Dt = —DI(p — p.)/pe!|/ Dt ‘The amplitude of sound produced by the final term on the right of (8) fs O(M} relative to the preceding one and is neglected. Inserting these results into (7), and disearding the overbar on p, we obtain ran i ol: (2-3) lo fo ft 1L\) a, '9) iam | (Ga me )ie- ray aoe - shore quenttiein square braces ate oxluatod at time t= fe—yl/ee The now terms in this formula. ination tothe Reyolds stres radiation of (6, are See aE eM Reece tee then poet # po, and are the soon falda of pelo and monoyae sour “hilo ete source types ove in the turbulent ising sean of hot gos jet exhouting into cold nr: shot spots “entropy inhomogeneities behave Bree ee eee reer directly into sound. The dipole source sti th is proportional to (1/~— Var, ies tothe diffrence beteen the acceleration of fd of censty p in the ot and thet which Bud of ambient mean density p, Would cxpereace in the some pressure gradient Foran ideal gas with p ~ pe Olpae?) in the Jet, the order of magnitude of the dipole sound from an eddy of seale € is pe (USDC — pofplpoe®M = (t/be(AT/T)p,v™ M, Theory of Vortex-Sound 37 where 7 denotes temperature, and AT/1' = (p~ p.)/p is the fractional tem- perature difference between the hot spot and its environment, whieh ean be large. ‘The sound power ~ €2(AT/T)29,0° M? exceeds that from a Reynolds stress quadrupole by factor of order (AT/T)?/M?. Thus, ‘entropy noise’ may be an important component of the noise of @ hot gas jet at very low Mach numbers The final term on the right of (9) is monopole whose strength is deter- mined by the difference between the adiabatic compressibilities (K’, ~ 1//pe?} in the source region and in the ambient medium, In an ideal gas FC, = 1/7. so that this source is generally small, although it may be significant during the turbulent mixing of gases with different values of the specific heat ratios +, Itean be very important in multiphase flows where, for example, the pres- cence of small air bubbles in water often leads to an immense imerease in the turbulence generated noise, 2 Vorticity and Entropy Fluctuations as Sources of Sound 21 The Réle of Vorticity in Lighthill’s Theory Al low Mach numbers the velocity defining the Reynolds stress quadrupole in Lighthill's equation can be determined by regarding the source flaw as incompressible, and using the Biot-Savart induction law of hydrodynamics to express v in terms of the vorticity w. Consider an acoustically compact, homentropic vorticity distribution w of seale f centered on the coordinate origin in a medium of density p, and sound spoad e,. The velocity v = wt Vy, where m is the incompressible velocity field defined in terms of w by the Biot-Savart induction formula, which implies that a ~ O(L/Ix[?) as fc] = 00. Since div m=, the value of y is determined by the compressibility of the fluid, in the souree region p—po ~ pot? and the characteristic frequency of the source flow ~ u/?. Thus, Dp/Dt ~ ppi/t, and the continuity equation Vp + Do/pDt Ve + (Ifp.c2)Dpf Dt = 0 implies that Vp = OuM?), M = ufe, for |x| ~E (19) Let us now write usa) {Ors0rs = div (wo Au) +0? (w 2) an 38 MS. Howe and express the solution (6) of Lighthill's equation in the form rift) + pats aay where [] denotes evaluation at time ¢— jx ~yl/eo. When retarded time variations aeross tho souro region aro neglected the identity (11) and the divergenee theorem imply that. fw A uld'y = 0, since u ~ O(1/ly[*) as ly| oc. To estimate the value of the fist integral in (12) it is therefore nnooassary to expand (w 4 u)(t — [x —yl/eo) in powers of the retarded time element x¥/ccfx. The fh tbe expansion yields, wary piste eae [ nlwanyty.t =e) ey ~ Ol6/lxD pn? M? F (13) whore a/08-~ ae ‘The order of magnitude of p(x, 4) is estimated by first writing the mo- anys (feo sti2s ott) ~-wnn-wary actor ae (xf nips Le + ot) ase sand integrate over the whole af space. The contribution from the divergence vanishes becouse u(f dp/p + 40? + 0/21) tends to zero at least as fast a I/iy?? as [y| + 20. Hence, using (10) of Litiy nity =— finer deity thy ~ een? H/ ponty=- frwrverynty~eew, ay where the final estimate is really a crude upper bound, that takes no ac count of the details of the interactions between the vorticity and inrotational velocity hs, pales t) ~ 0 ((E/Ix))pou? M*) , and by comparison with (18) we see that p» ~ O(M®)p1 <¢ py when M <1 ‘The component divipew A v) of O(povin4)/Oeiézy is therefore the principal source of sound at low Mach numbers. Theory of Vortex-Sound 99 2.2 Acoustic Analogy in Terms of the Total Enthalpy ‘The analysis in §2.1 and the discussion of §1.3 show that the dominant acous- tic sources at low Mach numbers are vorticity and entropy fluctuations. When Lighthill's equation is east in a form that emphasizes vorticity and entropy 1s the sound sources, it will is necessary to select an independent acoust variable that reduces to Ag/At in irrotational regions, where gs, f) is the ve- locity potential governing the motion in the irrotational regions of the flui. Bernoulli's equation suggests that the total enthalpy B= wt appropriate choice, where w is the specific enthalpy (= f dp/p(p) in bomen tropic flow). B is constant in steady irrotational flow, and at large distances irom the acoustic sources perturbations B = —Ag/It represents acoustic ‘To reformulate Lighthill's equation in terms of B for a homogeneous fluid, whose chemical composition is the same everywhere, we start Irom Croceo’s equation Oval VB =—wAviTVst a (as) a1 = (Ifph8oj Bey, (18) where 3 is the specific entropy, and rearrange the continuity equation to read az where Dp/ Dt has boon eliminated by means of the relations = #2, (22) av ona (22) ~ (#2) (2) = 2H to B+ (3) a me (HE), - Gt), (ar) aa Band ey being respectively the coelicient of expansion and the specific heat ft constant pressure. For an ideal gas 9 = 1/7. uburact the divergence of (15) from the time derivative of (17) sound, because they are unchanged when the compressibility K, = L/ec? ‘The first term on the left of (18) is expanded as follows: 40 MS. Howe peta eects gt oi Ga a oes compressibility K, = 1/pe is a function of the pressure alone, Le., when the 1/rp and 7 = cp/c, = constant. Variations of y in a real gas are significant I etic eye tempi eae snes constant; acoustic sources involving variations of K, are important only when bra et Tha et rue a Gay ee ects ae ae 1a 1D oy Pe vg om ae and second, from (15). ee 8 vp (WR +wAv-TVs-o) ey ng from (19) into (18), we arrive atthe acoustic analogy equa- Lion for the total enthalpy LB) 8 in ) (wAv-1Vs~o) ala(rn)} 2.3 Vorticity and Entropy Sources The nonlinear operator on the left of this equation is identical with that gov- cerning the propagation of sound in irrotational, homentropie flow. All terms on the right hand side vanish in irrotational regions, and in the absence of such terms and of boundary motions, B = constant, The radiation condition ensures that the terms on the right may be identified as acoustic sources. ‘They are confined to the region in which « / 0 and where Vs f 0. The wave operator on the left describes propagation of the sound through the non-uniform flow although, as for Lighthill's equation, it will not usvally bbe permissible to neglect the interaction of the acrodynamically generated sound with the vortivity and entropy gradients when the souree flow is very extensive, The following special cases should be noted: (i) High Reynolds number, homentropic flow. When dissipation is neglected and s ~ constant (22) becomes b(ip) 1 1 ; (B(EB) Aeon) ahaa. =D, Theory of Vortex-Sound AL AL low Mach numbers when the flow is at rest at infinity (where p= p, and ¢ = ¢) further simplifieation is possible by (a) neglecting nonlinear effects. of propagation and the scattering of sound by the vorticity, and (b) taking. © cp, and p= po. Then 1F : Ban 7V') B= ave ay) en ‘and in the far field the acoustic pressure is givon hy the linearized approsi- mation Pl) po BUX, (25) (ii) Non-homentropic source flow. When the source region is not hemo: tropic. it is important to distinguish becween terms on both sides of (22) that account prineipally for seatfering of the sound. and those (on the right) that can unequivocally be recognized as sources. For compact sourees we can argue that the correct source terms are those remaining when the fuid is temporarily taken to be incompressible. Those are the sources on the right of (18). When dissipation is ignored, (22) then reduces to (E09) B@re 9) dr eo}e- viene too 1 ( PDs SEP SY | (26) ro) or ‘This approximation is useful when the background mean flow may be r- garded as invotational (steady flow past a streamlined body, for example) at nniea velocity U(x) and density and sound speed p(x) and o(x). At very small Mach numbors it simplifies further to Le oy veay 28) 4 2 (EB! ie (EE -¥) p-atenvren +8 (EE). ox ‘The final entropy souree on the right of (26) and (27) represents the pro- duction of sound by unsteady heating of the fluid. According to the equation of continnity (27) this somtee is equivalent to a volume monopole af strength, a(x,t) = Be es) Spinning vortex pair (Fig.2): Two parallel line vortices each of circulation 1 and distance 2¢ apart rotate abont the x-axis midway between them at angular velocity 2 in fluid of mean density pp. At time ¢ their positions in the zysrz-plane are x ~ s(t) = ¢(cos 2, sin $2). Then r divi Av) - (eild)sy(0800)) +205 = (ease) BrDey 42 MS. Howe sound WF | oO : (et “Se Pig. 2. Spinning vortices. where the terms omitted are higher order multipoles. If I/eot «1 (compact sourve condition), the aeoustie pressure is given by (t— rho) + pe 1/2 pu a ae (29-20 ), Prfey 20 7 where U = P/4xe, M = Ufe, and (7,9) are polar coordinates of (x1,-2) ‘The power radiated per unit length of vortex ~ fg,U7Af*, whieh is the same 1s for ‘two-dimensional’ turbulence Oy) 2 sound Fig. 3. Kirchholl's vortex Kirchhotf’s spinning vortex (Pig.3): Consider the sound produced by a spinning, columnar vortex of elliptic cross-section defined by the polar equa- tion r— (1-4 evos(20 ~ 24/2), whore ¢-<: 1 and £2 the uniform vorticity in the core. The ellipse rotates al angular velocity 182, and the velocity dis- tribution within the core is given by, v= (v.09) = (sino + esin(@ — 4/2), ~c0s0 + evos(@ — 24/2). Theory of Vortex-Sound 43 The vortex is equivalent to the two-dimensional quadrupole div(w 4 v: 8 (ty 6(a)8(ay)) (Or,0e;, where __ Pa f cos 242) sin( 24/2) ae (Sits) Sean) and that the acoustie pressure is, © [Oma aqs32 $f 29,0200? cos (29 — By ( 2 rete r/o 4. : 142). ary 2) and M =Uje,, The acoustic frequency is (wice the rotation frequency of the where 27 is the maximum mean flow speed (at the edge of the core cllipse, and the directivity has a characteri quadrupole, ic fourlobe shape of a “lateral” 3 Fundamental Solutions of the Wave Equation ‘To investigate the production of sound it is convenient to consider first sources whose strengths vary periodically with time, This is because the wave equa- tion is linear, and solutions corresponding to the different frequency compo- nents of the source may be superposed, permitting the full solution to be expressed as @ Fourier series or integral 3.1 The Helmholtz Equation ‘Consider the sound radiated into an unbounded, stationary fluid from a time- harmonic volume source g{X.t) = 9(x.w)e “ of radian frequency . The ¥ potential p(x.4) = pix,we "evidently: oscillates atthe same: fre= and ¢p(X, ) satisfies the inhomogencous Helmholtz equation (Weer a (29) where hy = w/ea (30) is the acoustic wavenumber. We adopt the shorthand notation that (except ‘where confusion arises) ye and ¢ represent the frequency-domain quantities (%,.2) and g(x, 2). In an unbounded fluid the solution of (29) raust satisy the vadiation condition that energy delivered to the fluid by the source rauliates away from the source, ie. the solution must exhibit oulgoing wave behavior The solution for a unit point souree q(x.) — S(x—y) = 62, ~ yp )B(e2— 1¥2)5(25 ~ ys) is called the frequency-domain Green's fumnetion G(x.) which satisfies (wv KIIG = Oxy), (1 MMS. Howe The identity g(x.) = fy gly,w)6(x—y)d2y, where the integeation is taken over a volume V containing the sourees, and the principle of superposition permits the solution of the general problem (29) to be written eau) = [ Cla riwayaty (32) Let us solve (31) by the method of Fourier transforms. The n-dimensional Fourier space-transform /(k) of a function f(x) of x = (r1,:22... at) sate isfios the reciprocal relations 100 = paige [ soow mere, 100 ~ J sonar, (83) where k is the n-dimensional wavenumber vector k= (Ub, fea...) Take the Fourier transform of (31) by applying the integral operator (1/2n)8 f° (o)e**ax, By integration by parts (which is permissible for any generalized function such as Glas,yiw)) we write if VEG. yuo Hee f Cbx, yywhe Ox where k= |kj, to obtain -—soo 34) CC 7 ‘The inverse transform (the second of equations (33)) yields +f ek 5) wR ‘To evaluate the integral introduce spherical polar coordinates (k,4, 6) for k, where the latitude @ is measured from the sourosobserver direction of x—y, and dk Adddodk. Performing the integrations with respect to @ and 9, we find Glxyi0) mel) ak. 6) ‘The remsining integral is undefined for real values of «6, because of the pole on the real axis at & = sol. The pole can be avoided by indenting the path into the complex plane to pass either above or below the singularity. ‘Two altemative solutions of equation (31) are obtained in Uhis way, but only ono, where path runa below or above the pole according as x 2 0, exhibits outgoing wave behaviour Theory of Vortex-Sound 45 sma complete Fig. 4, Path of integration in the wavenumber plane when se > 0. "Yo prove this suppose that, > 0. Indent the integration path to pass below the pole at k= 8» (Fig. 4), and consider the separate contributions to the integral in (36) from the two exponential terms in the integrand. For the first, Cauchy's theorem permits the path of integration to be rotated ‘through 90° onto the positive imaginary axis, on which the integrand decays exponentially as # + +30. The pole at k = 4, is crossed during this rotation, giving a residue contribution ize’***~¥ to the value of the integral. The path of integration for the second exponential may be similarly rotated onto the negative imaginary oxis, this time without encountering any singularities, ‘The two integrals on the positive and negative imaginary axes are equal and opposite, and therefore (a When the exponential time factor is restored Glo, yun) MS me OO tal = y |, which represents a continuous wave propagating radially outwards at speed o (osith inereasing time) from the source at y. ‘The amplitude decreases inversely with distance [x — y|, in accordance with the inverse square lave of energy conservation. When jx — y| > 00 the wave “crests” (surinces of constant. phase) become locally plane with wavelength 27 /re. Conversely, when the integration path in (36) passes above the pole at > 0, the solution is of the form (37). but with the sign of «i, reversed. This represents a spherical wave conversing, towards the source in violation of the radiation condition, “The outgoing wave solution is also obtained from the condition that dis- sipation within the fluid (which gradually transforms acoustic encrgy into haat) canses the wave to decay faster than I/Jx— y| at large distances from the source, This will be the case if, a initio, re is imagined to be assigned ‘asinall positive imaginary part which (for w > 0) shifts the pole off the real axis into the upper half-plane. 46 MS. Howe This is effectively identical to a formal method proposed independently by London and Lighthill for obtaining the causal solution of time-harmoni wave problems. ‘The frequeney w is temporarily assigned a small. positive imaginary part ie, which is equivalent Wo considering the radiation from a source that is ‘switched-on’ in the infinite past, whose amplitude increases. slowly tn proportion to et. The exponential growth of thesohution as f+ +30 is of no significance because causality ensures that it cannot influence the behavior at finite times, When «) > D the pole is displaced to k= ky + fe/ey in the upper half plane, and the integration is performed as before, leading to (87). The method is applicable generally to time harmonic wave problems governed by linear equations. However, misleading predictions can oceur for systems that are linearly unstable. For example, the linearized treatment of the motion produced by a source in a steady flow over an elastic surface exhibits linear instabilities that grow by extracting energy from the mean flow. These instabilities are not, captured by the Landau-Lighthill procedure, because it assumes that all of the perturbation energy is supplied by the source. In these circumstances a causal solution can be derived by considering, fan initial value problem. Example: In {wo-dimensions Green's function @(x,yiw) for the Heltuholtz, equation (W343) =8(x—y), x= (ey y= (uw), is given by Gx, yew) = (38) whore HY isa Hankel finetion, ‘Phe result may’ be derived by’ integrating the three-dimensional Green’s function over —o0 <3 < 00. It represents a cylindrical disturbanoe whose behavior at large distances from the source is sven by Example: In one dimension Cle, ys) = —s (39) 3.2 The Wave Equation Green’s function for the wave equation is the solution G(x,y.¢— 7) with outgoing wave behavior of (Peak — VE = bx —y)slt~ 0) (40) "Theory of Vortex-Sound 47 The right hand side is an impulsive point source which vanishes except al ter. The equation is solved by superposition, by observing that 4(¢ ~ 7) (1/2n) [%e"-7)d, and therefore that (81) is transformed into (40) by application of the integral operator (=1/2n) [®,(e)e-!"-Ydu. Hence, Cixyt—7) = f Gla yale Md ap so that in three dimensions the solution (37) of (31) supplies Glxy.t=7) weal eh evlddy ae bx — yl/co). (42) ‘This vanishes for ¢ < 7, in accordance with the resents an impulsive, spherically at y at the speed of sound. "The causal solution in unbounded fluid of the inhomogeneous wave equa tion ssality principle, and rep- symmetric wave expanding from the source V2)e= fst) (a3 is obtained by superposition, by observing that fet) = ff seo. eex— yale ey nd thoreore that olx,2) [fo G(x, yt — ry a ta be-yiles) Rv eed ‘The integral formula (45) is called a retarded potential, and represents al position x and time /-2s a linear superposition of contributions from sources at positions y which radiated at the earlier times ¢~|x—yl/co. where sx — yl/ep is the time of travel of sound waves from y to x: 48 MS. Howe Pg. 5. Solution of the wave c of a boundary 8, 4 General Solution of the Inhomogeneous Wave Equation 4.1 General Solution in the Frequeney-Domain ‘The solution of the wave equation in the presenee of arbitrary source distribu: tious and arbitrary surfaces S involves the Green's funetion for the reciprocal problem (Fig. 5). Consider first the frequency domain form of (48), ie. the inhomogeneous Helmboltz equation, = -Flx.), (48) where f(x,.0) = (1/2x) JX, Fix.0) Green's ametion Gly. xs) = G(x, y;2) is eny solution with outgoing wae behavior of the reciprocal problem in which the source is placed at the point x¢ where the solution g(x.) is to be found x, feat is the Fourier time transform of (V3 1 2) yu) = 668 —y). (ay ‘The notation implies that G(x, yj) is sought as @ function of ¥ For fixed x. Equations (46) and (47) are now combined in the following way: (i) replace x by y in (46) and multiply by Gex,yse), (il) subtract the product of (47) and ¢(y.w). Integrate wilh respect Lo y over the whole of the fluid and apply Theory of Vortex-Sound 49 the divergence theorem to obtain ew) = f few yiw) (7.0) PORE Coys «)} masts) Ou = [ Cle yw flywe)dy, (43) ‘where the unit normal m on § is directed into the fluid, ‘The radiation eondi- tiow ensures that there ave no additional contributions from a surface integral at infinity Equation (48) is applicable for any outgoing solution G of (47). The sur- face integral can be evaluated ify and O/0y, are known on S, but these ‘quantities cannot be prescribed indepenctently. 4.2. General Solution in the Time-Domain Fourier superposition may now be used to detive an integral formula for the solution y(x,f) of the wave equation (43). Application of the canvelution theorem (wet, ee ees 4 © Ale (wera = ef. Sule) folt—r)dr (49) to (48) vieds Kirebhotts formal : p(y. 7) 2S (x,y,t— 7) — L(y, 2)GGny.t—7)} n, Ir oe = f foo nee yt 9 — ey natayst 1) nyasiys ' [« x,y, — fly, nd yar, (50) where the retarded time integration with respect to 7 is taken over (—90, 0). sand G(x,¥,f—7) is an outgoing solution of (40) and is velated to G(x, y:0) by (AD). ‘The special ease in which Cf is th be nates free space Green’s function (12) should In this formula we have set ty = 2y/g = normal velocity on 8. The sur- {ace intograls respectively represent the sound produced by distributions of onopole and dipole sourees on S, of strengths n(x, £) and (x,t) per unit area, These sourees cannot he preseribed independently. 50 MS. Howe 5 Compact Green’s Functions ‘The compact Green's fnetion provides a formal and intnitive procedure for caleulating the leading orcer monopole and dipole terms in the multipole expansion of the sound produced by sourees near a solid body 5.1 Tine-Harmonic Problems Consider the particular Green's funetion ) ned by the Helm holtz, equation (47) that has vanishing normad derivative on the surface § of an acoustically compact body of diameter €. This equation is to be solved as 1 function of y for fixed value of x, which we take to be in the acoustic far field ofS. ‘The determination of G(x, ¥;w) ean be posed as a seattering problem in schich the spherical wave (37) [assumed to be generated by a point source at x, and which is just equal to the free space Green's function) is incident on S. If the coordinate origin is within S and y is clase to S, the compactness condition xf 1 permits G(x, yw) to be expanded in the form (u-#i9)) + Deore, i y ~O(6), |x| 00. The first termi in the large braces represents the incident wave (37) evaluated at y = 0. The next term is O(set) and includes a component. ~inosi4/|x| irom the incident wave plus a correction ~ix3e{(y)/[xl due to S. To this oriler of approximation Yi(y) = m—7{y) is solution of Laplace's equation satisiving BY;/6y, ~ 0 om’. Since yi {y) must decay with distance from §, Lis implies that gf is simply the velocity potential of the incompressible rotion that would be produced by translational motion of Sas a rigid body ‘a unit speed in the i-direction, The remaining terms in (52) are of order (of)? oF smaller. When they are neglected the resulting approximation for G can be used to determine the monapole and dipole terms in the multipole expansion of the solution ofa fluid-siruetuce interaction problem, ‘The potentials pf are uniquely defined by the shape of the body and satisiy =m ons (53) When the body is rigid they also determine the added mass tensor My. which is symmetric and given by ‘Theory of Vortex-Sound SL Fig. 6, Sound generation by a risiel bod ‘oscillations, ccuting small amplitude trarelotional A body in translational accelerated motion at velocity Ul) in an ideal incompressible ffuid experiences a reaction force from the fluid equal to =A(M,jU,)/2t in the iadirection. An extemal force F; acting through the center of mass of & body of mass m therefore produces accelerated motion determined by Afni + MyU,)/OU = Fi Example: Sound generation by a compact, rigid body executing small am- plitude translational oscillations (Pig, 6) Let the translational velocity be U(2) and consider first the contribution to the sound from the Fourier component U(a}e**, Use Green’s function (52) in the general solution (48), with f= 0, and va = U.non 8. Then —tesade*l 1 agg) 0% L yn asiyy y OW ) ~~ ark 3 sdSiy) — fr pASly) + OW (06 Yi w= vy) ‘The first where mio is the mass of fluid displaced by the body. When the body acoustically compact Us(w) # 0 only for ref 1. Hence, multiplying by Legral vanishes identically; the second is equal to (m2,8,+Mis)/ © and integrating over 90

You might also like