CHAPTER 5 T HERM ODY NA M I CS OF A I RCRA FT J ET ENGi NES
REFERENCES 4 f 4 1. Ferguson, C. Internal Combustin Engines. New Y ork: Wiley, 1968. 2. Heywood, J . B . Internal Combustin Engine Fundamentis. New Y ork: McGraw-Hill, 1988. 3. Taylor, C.F. The Internal Combustin Engine in Theory and Practice, 2nd. ed. Cambridge, Mass.: M.I .T. Press, 1985. 4. "The Jet Engine," Technical Publicalions Department, Rolls-Royce, Derby, U.K ., 1986. 5. Ciepluch, C.C., D.Y . Davis, and D.E. Gray. "Results of NASA's Energy Efficient Engine Program," AIAA J. Propulsin and Power 3 (1987):560-568. 6. Cundy, J . "Saving Fuel with theWide-Chord Fan," Rolls-Royce Magazine, June 1983, pp. 13-18. 7. Lewis, G. M . "V/STOL from thePropulsin Viewpoint," Aerospace, June/July 1983. 8. Lancaster, O. E., ed. Jet Propulsin Engines, vol. 12 of High-Speed Aerodynamics and Jet Propulsin. Princeton, N.J .: Princeton University Press, 1959. 9. Saravanamuttoo, H. I. H. "Development of Modern Turboprop Engines," CASI J. 31 (1985):131-139. 10. Morris, R. E. "The Pratt and Whitney PWlOO: Evolution of the Design Concept," CASIl 28 (1982):211-221. 11. Cook, D.L . "Development of the PWlOO Turboprop Engine," CASI l 30 (1984): 230-239. 12. Taylor, E.S. "Dimensional Analysis for Engineers," Clarendon Press, Oxford, 1974. C H A P T E R 6 A EROTHERMODY NA MI CS OF I NLETS, COMBUSTORS, A ND NOZZLES 6.1 INTRODUCTION I n Chapter 5 we used the laws of thermodynamics and fluid mechanics to explain the behavior of aircraft jet engines. We treated the several engine compo- nents as "black boxes," i n the sense that we confined discussion to the inlet and outlet conditions of the propellant, without regard to the internal mechanisms that produce its change of state. Where necessary, we related the actual perfor- mance to some easily calculated or "ideal" performance by the definition of an ap- propriate component efficiency or stagnation pressure ratio. The purpose of this and the following chapter is to examine the internal mechanisms of the various components i n order to describe the factors that impose practical limits on per- formance. We consider conditions required for high performance of components and, i n some cases, present methods for quantitative prediction of their behavior. For the ramjet, Eq. (5.34) indicated that a given percentage loss i n stagnation pressure has the same effect on engine performance wherever it occurs through the engine. For turbine engines the same conclusi n holds, though it is not so easily seen, since component performances are usually stated i n terms of adia- batic efficiencies rather than stagnation pressure ratios. Henee high performance is of equal importance for all engine components. Theattainment of high perfor- mance is generally more difficult i n those regions requiring a rise i n static pres- sure than i n those where the pressure falls. Thi s, as we pointed out i n Chapter 4, may be attributed to boundary layer behavior and the tendency for separation i n the presence of a ri si ng static pressure. Thus inlets (which generally have rising 217 ^L nv^i ncMi vi uuY NAMl Ub F I NL ETS. COMBUST ORS, A ND NOZZL E S pressure gradients even for turbojets) may be moredifficult to design for efficient operation than nozzles. Similarly, as wewi l l see in Chapter 7, compressors are moredifficuU to design (and have lower uhimate efficiency) than turbines. I n this chapter we are concerned with the major components of turbine en- gines other than the turbomachinery (which is the subject of Chapters 7, 8, and 9). The common theme is steady flow in stationary ducts, and the chief interest lies in the arrangements that are necessary for high efficiency in inlets, burners, and nozzles. 6.2 SUBSONIC INLETS A n engine installed in an aircraft must be provided with an air intake and a ducting system. Figure 5.27 shows the inlet duct for theRolls-Royce RB211-535E4 engine mounted in an underwing installation. Wewi l l see in Chapter 8 that for turbojet engines the airf low entering the compressor or fan must have low Mach number, in the range 0.4 to 0.7, with the upper part of the range suitable only for transonic compressors or fans. I f the engine is designed for subsonic cruise at, for example, M =0.85, the inlet must be designed to act as a diffuser with reason- ably gentle diffusion from M =0.85 to perhaps M =0.6. Part of this decelera- tion occurs upstreamof the inlet entrance pl ae. One can see in Fi g. 5.27 the relatively small increase in rea associated with the internal deceleration. The inlet must be designed to prevent boundary layer separation, even when the axis of the intake is not perfectly aligned with the streamline direction far upstreamof the inlet. I n other words, the performance of the inlet must not be unduly sensitive to pitch (up-and-down) and yaw (side-to-side) motions of the air- craft. It is important that the stagnation pressure loss in the inlet be small. It is even more important that the flow velocity and direction leaving the iniet be uni- form, since distortions in the velocity profile at the compressor inlet can severely upset the compressor aerodynamics and may lead to failure of the bades due to vibrations. Design of inlets that must oprate efficiently in both supersonic and subsonic flight poses special problems; we consider these in Section 6.3. Flow Patterns Depending on the flight speed and the mass flow demanded by the engine, the inlet may have to oprate with a wide range of incident streamconditions. Figure 6.1 shows the streamline patterns for two typical subsonic conditions and the corresponding thermodynamic path of an "average" fluid particle. Duri ng level cruise the streamline pattern may include some deceleration of the entering fluid external to the inlet plae (Fig. 6.1(a)]. Duri ng low-speed high-thrust operation (e.g., during takeoff and climb), the same engine wi l l demand more mass flow and the streamline pattern may resemble Fi g. 6.1(b), which illustrates external acceleration of the stream near the inlet. I n both cases there is an external change of state that is essentially isentropic, since there are no walls on which friction may act. For given air velocities at stations (a) and (2), external accelera- 6.2 SUBSONI C I NL ETS 219 tion raises the inlet velocity and lowers the inlet pressure, thereby increasing the internal pressure rise across the diffuser. I f this pressure increase is too large, the diffuser may stall because of boundary layer separation; stalling usually reduces the stagnation pressure of the streamas a whole. Conversely, external decelera- tion requires less internal pressure rise and henee a less severe loading of the boundary layer. Therefore the inlet rea is often chosen so as to minimize exter- nal acceleration during takeoff, with the result that external deceleration occurs during level-cruise operation. Under these conditions the "upstream capture r ea" i s less than the inlet area/l |, and some flow is "spilled over" the inlet, accelerating as it passes over the outer surface. For high-Mach-number subsonic flight this acceleration (and subsequent deceleration) must not be too large or therewould be danger of shock-induced boundary layer separation on the outer surface and excessive nacelle drag. For supersonic flight such "spilling" action would necessarily be accompanied by a shock systemthat reduces the relative velocity at inlet to subsonic val esso that the air may sense the presence of the inlet and flow around it. Internal Flow Qualitatively the flow i n the inlet behaves as though it were in a "diffuser," which is a common element in fluid machinery. A better termmight bedecelera- tor, since the device is not primarily concerned with molecular or turbulent diffu- sion, but wewi l l retain here the traditional termdiffuser and define it to mean any section of a duct in which fluid momentumdecreases and pressure rises, no 220 CHAPTER 6 A EROT HERMODY NA MI CS OF I NL ETS, COMBUST ORS. A ND NOZZL E S work being done. Considerableexperimental and analytical work has been done oncyiindrical (especially conical [1] and annular [2]) diffusers, but littleof this is directly applicableto subsonic aircraft inlets. Thereason is that most work on dif- fusers focuses on theconditions that arerelated to mximumpressurerecovery which is usually associatedwith ahighly nonuniformexit velocity profileand per- haps evenwith someflow unsteadiness. I ntypical subsonic aircraft inlets thereis a stringent requirement that the flow velocity entering the compressor be steady and uniform. Consequently inlet design does not depend so much on the results of diffuser research as on potential flow calculations, coupled with boundary layer calculations andfoUowed by wind tunnel testing to assess inlet performance under awiderangeof test conditions. In the actual engineinlet, separation can takeplacein anyof the threezones shown in Fig. 6.2. Separation of the external flow in zone () may result from local high velocities and subsequent deceleration over the outer surface. Wewill discuss this possibility, which leads to high nacelledrag, subsequently. Separation on the internal surfaces may take place in either zone (5) or zone (3), depending on the geometry of the duct and the operating conditions. Zone (3) may be the sceneof quite large adverse pressuregradients, since the flow accelerates around the nose of the center body, then decelerates as the cur- vature decreases. In some installations it has not been possibleto make the exit rea of the intake more than about 30% greater than the inlet rea without the incidence of stall and large losses. Reynolds number effects may also be impor- tant for largeinlets and high-speed flow. A t high angles of attack, all threezones could besubjected to unusual pressure gradients. External Flow A s wealready indicated, inlet design requires acompromisebetween exter- nal and internal deceleration. Both can lead to difficuUies, and a balance is needed. To examine the effect of external deceleration on inlet design, methods areneeded for calculating both potential flow (internal and external) and bound- ary layer growth on intakesurfaces. References [3], [4], and [5] providedetails on how such methods have been applied to subsonic engineintakes. These methods arenot valid for separated flow, but they areableto warn the designer of the dan- ger of separation, and so provide guidance on design modifications that may beneeded to avoid separation. Expressed in computer codes, thesemethods can FIGURE 6.2 Possiblelocations of boundary layer separation. 6.2 SUBSONI C I NL ETS 221 determine the pressureandvelocity field at all points in the flow. They are able to predict the momentumthickness (and the associated nacelledrag) due to the boundary layer on the external surface of the nacelle. They are also ableto define the velocity distribution of the flow approaching the compressor. Wewill not go into thedetails of thesemethods herebut insteadwill focus on a qualitatively valid and much simplcr explanation of the fact that external decel- eration must be limited to prevent excessivenacelledrag. Following the method of Kchemann and Weber [6), we here ignore compressibility and suppose that the only effect of the boundary layer isflow separation (if the deceleration pres- surecoefficient becomes too large). Figure6.3 shows atypical streamline pattern for largeexternal deceleration. I n flowing over the lip of the inlet, the external flow is accelerated to highveloc- ity, much as the flow is accelerated over the suction surface of an airfoil. This high velocity and the accompanying low pressurecan adversely affect the bound- ary layer flow in two ways: For entirely subsonic flow, the low-pressure regin must be foUowed by aregin of rising pressureinwhich the boundary layer may seprate. Henee one might expect a limiting low pressurep^in or, equivalently, a mximum local velocity Wn,ax, beyond which boundary layer separation can be expected downstream. For higher flight velocities (or higher local accelerations), partially supersonic flow can occur. L ocal supersonic regions usually end abruptly in a shock, and the shock-waI l intersection may cause boundary layer separation. One might expect a limiting local Mach number that should not be exceeded. Whatever the cause, boundary layer separation is to be avoided, since it results in poor pressurerecovery in the flow over the after portions of the air- craft or enginehousing. This, of course, results in anet rearward forc or drag on the body. "a FIGURE 6.3 Control volumefor the calculationof thrust on inlet surface. 222 CHAPTER 6 A EROT HERM ODY NA M I CS OF I NL ETS, COMBUST ORS. A ND NOZZL E S To illustratethe major features of the external flow near the inlet, consider the simpified problemof an inlet on a semi-infinite body.^Kchemann and Weber haveshownhow to relatetheexternal flow over suchabody to the extent of external decelerationof theflow enteringthe inlet. Suppose (seeFig. 6.3) that theexternal crosssectionof theinlet growsto amximumarea^max and that the body remains cyiindrical fromthis point downstream. A control surface is indi- cated that extends far fromthe inlet onthe sides and upstreamend, crosses the inlet at its mnimumreaA, passes over the inlet surface, and extends down- streamfar enough for the external fluid velocity to return essentially to the up- stream or flight velocity Ua (neglecting boundary layer effects). Thus all the external flow entersand leaves thecontrol volumewith an axial velocity compo- nent Ua, if one assumes that the sides of the control volumeare sufficiently re- moved fromthe inlet. The internal flow enters the control volumewithvelocity Ua and leaves with velocity t/ (assuming, for simplicity, one-dimensional flow in the inlet). The net momentumflux out of the control volumeis then, ignoring changes inthe air density, m,Ua + puf Ai - pU^Afnax- Fromcontinuity, the side flow rate is = pu^A^ax - puAj, so that the net momentumflux can be expressed as pAi{u - uMO). The net forc inthe axial direction onthe control volumeis PuAmax ~ piA Fx, where is the axial component of the forc on the control volumedue to the forces onthe external surface of the inlet. I f weneglect friction, pdA^, A, where p = pressure on surface, X =unit vector alongaxis inthe flow direction, n =outward (fromthe inlet surface) pointingunit vector, dA = increment of external surface rea, dA:, = increment of external surface rea normal to x {lirrdr for an axisymmetric inlet). Combining theseexpressions, the momentumequation requires /?fl/lmax ~ PiAi - pdA, = pAi{uf - UMa) JA, or {pa - p)dA, = pA,(uj ~ u,Uu) +{p, - Pa)Ai. (6.1) -4, * A ninlet followed by acyiindrical portion several diameters inlengthwould behave similarly, but few practical enginehousings are actually cyiindrical over any appreciablelength. 6.2 SUBSONI C I NL ETS 223 The equation is arranged in this formbecausewe can consider the integral a component of thrust AST,, which actsonthefront external surface of the inlet duc to reductionof local surface pressure. Applying Bernoulli's equation to the exter- nal deceleration of internal flow, weobtain Pa = P / 2 2 'Ua - Ui Thus or (Pa - p)dA^ = pA,{uf - UMa) + pAi ^ul Ui 2\ 1 2 ^ \ / (6-2) ipUaAi \ This shows that the greater external deceleration (i.e., the smaller the ratio Ui/ua), the larger must bethe "thrust" increment: ^% = \"^^\pa- p)dA,, On the outer surface of the nacelle, the pressure must risefromsome mnimum valu/?n,in(at thepoint wherethelocal free-streamvelocity isw^ax) to theambienl valupa associated inthis simpified casewith straight parallel flow downstream. Hereweneglect the boundary layer except to say that the pressure coefficient Pa ~ Pm'\n 2 Po^ max must not be too large or the boundary layer will seprate. If, stili for the outer surface of the nacelle, wedefine an average pressure difference (Pa - p)dAx A, where5isafactor between O and 1, wecan write A2r =S{pa ~ Pmin)iAm3x - Ai), so that Eq. (6.2) becomes S(pa - Pm\n) (^max " A,) kpulAi 1- ^1. We can rewritethis as ~A~ = 1 + SCr Ua j 224 CHAPTER 6 A E ROT HE RM ODY NA M I CS OF I NL ETS, COM BUST ORS, A ND NOZZL E S or Ai = 1 + 1 - ^ (1 ^pmax) sC (6.3) pmax The valu of s wi ll depend on the shapeof the nacelle. Taki ng s = 0.5 for pur- posesof i llustrati on, wecan show, as i n Fi g. 6.4, the dependence of the sizeof the external surface necessary to prevent external boundary layer separation, for any givenvalu of Ui/Ua. The mai n point here is that the larger the external deceleration (i.e., the smaller the valu of ,/), the larger must be the sizeof the nacelle i f one is to prevent excessivedrag. Even i n the absenceof separation, the larger the nacelle, the larger the aerodynamic drag on i t. But i f the external deceleration is modest (e.g., u/ua > 0.8), its effect on m ni mumnacellesize isquite small. The useof parti al internal deceleration is, of course, doubly effective i n reducing mxi mum diameter becauseit permi ts a reduction i n both.^^ anA^^jA,. Wi th an aircraft flight Mach number of 0.85, an enginewi th a transonic fan wi l l require li ttledeceleration of the i ncomi ng ai r, since the allowable absolute Mach number at entry to the fan may be as high as 0.6 or even higher. Sti ll the inlet must becarefully designed for thiscase. Wi th a flight Mach number of 0.85, the mxi mumvelocity near the external surface could easily be supersonic, so that there is the possi bi li ty of shock-induced boundary layer separation. Ful l allowance for compressibility effects istherefore necessary for inlet design. To summarize: Thi sanalysispertains to a si mpi fi ed pictureof the real flow around inlets. Nevertheless, it shows that the performance of an inlet dependson the pressuregradient on both internal and external surfaces. The external pres- sureriseisfixed by the external compression and the rati o,^^ax/^i of mxi mum A. .04=C^,, 0.2 0.4 0.6 0.8 LO IDeceleratin ratio. ^ "ir FIGURE 6.4 Mni mumfrontal rea ratio. Vari ouspressure coeffi ci ents, Cp, (Eq. 63, witti s = 0.5). 6.2 SUBSONI C I NL ETS 225 rea to inlet rea. The internal pressurerisedependson the reduction of velocity between entry to the inlet diffuser and entry to the compressor (or burner, for a ramjet). Nacellesizerequired for low drag can be quite strongly dependent on the degreeof external deceleration. I n realistic analyses one must consider com- pressibility effects. Inlet Performance Criterlon A s Chapter 5 showed, one may characterize the differences between actual and ideal performance of aircraft engine inletsby a "diffuser efficiency" or by a stagnation pressureratio. Wedefine theseas follows: a. Isentropic efficiency. Referri ng to Fi g. 6.5, we can define the isentropic efficiency of a diffuser i n thi s form: _ h[)2s ~ hg _ 7O2J ~ Ta h^~- ha T^a - Ta ' StateC@) isdefined asthe statethat would bereached by isentropic compression to the actual outlet stagnation pressure. Since Ta the diffuser efficiency TJ ^isalso given by [(y ^ 1)/2]M^ ^ b. Stagnation pressure ratio, r^. The stagnation pressureratio. 226 CHAPTER 6 A EROT HERM ODY NA M I CS OF I NL ETS. COM BUST ORS. AND NOZZLES 1.00 0.95 0.90 0.85 i 1 n 0.2 0.4 0.6 M 0,8 1.0 FK5URE 6.6 Typical subsonic diffuser performance; 7 = 1.4. is widely used as a measure of diffuser performance. D.ffuser efficiency and stag- nation pressure ratio are, of course, related. I n general, P^ Pa P.^a\^ ^ ~ i r ^ ' ) ' 02 ^ 02 Poa _ and, with Eqs. (6.4) and (6.5), y - 1 MM( r ^) ( r -i ) / r _ 1 [(y - 1)/2]M (6.6) BecauseTJ ^ wi l l beprimarily affected by theinternal deceleration ("diffusion"), it is unfortunate that thesecriteria are based onoverall deceleration rather than on internal deceleration. The relationship between internal and external decel- eration depends on engine mass flow rateas well as flight Mach number M . But for illustrativepurposes Fig. 6.6 gives typical vales of stagnation pressure ratio Kd. The diffuser efficiency 7}d was calculatedfromr^, with the use of Eq. (6.6). 6.3 SUPERSONICINLETS Even for supersonic flight it remains necessary, at least for present designs, that theflow leaving the inlet systembesubsonic. Compressors capableof ingest- ing a supersonic airstreamcould providevery high mass flow per unit rea and, theoretically at least, very high pressure ratio per stage. However, the difficulty of passing afully supersonic streamthrough thecompressor without excessiveshock losses (especially at off-design conditions) has so far made the development of fully supersonic compressors a possibility that is somewhat remote. A s we wi l l see in Chapter 8, the Mach number of the axial flow approaching a subsonic 6.3 SUPERSONI C I NL ETS 227 compressor should not be much higher than 0.4; for a transonic stageit can be about 0.6. Thetermtransonic refers hereto theflow velocity relativeto the blade tip, not to the absolute entrance velocity. For a ramjet thereis no such limitation based on compressor aerodynamics. Al so it is possible to have combustin in a supersonic streamwithout prohibi- tive aerodynamic losses [7]. The acronymSC R A M J E T denotes the supersonic- combustion ramjet, a concept that has long been under study for flight Mach numbers so high that the ramjet should besuperior to the turbojet in propulsin efficiency. The SC R A M J E T concept has not yet found application in a flight vehicle; ramjets developed to daterequire subsonic airstreams to provide stable combustin without excessive aerodynamic losses. Wedo, however, briefly dis- cuss supersonic combustin in Section 6.5. Section 6.6wi l l givesomeinsight into why a typical Mach number at inlet to a subsonic combustor is 0.4. Here wefocus attention on means of decelerating a supersonic flow to sub- sonic speeds tolerableby existing compressors or fans (or ramjet combustors). Reverse Nozzle Diffuser Figure 3.8 shows that deceleration fromsupersonic to subsonic flow speeds can be done by a simple normal shock with small stagnation pressure loss if the upstream Mach number is quite cise to 1. For high Mach number the loss across a singlenormal shock would beexcessive; in this caseit would bebet- ter to use a combination of obiiqueshocks. I sentropic decelerationwould be still better. Froma simpleone-dimensional analysis, it might appear that a supersonic (converging-diverging) nozzleoperated in reverse, would be the ideal device to produce nearly isentropic deceleration. There are formidable difficulties, how- ever, in implementing such a concept. These havechiefly to do with theproblem of succcssful operation of a reversed-nozzle inlet over a range of flight Mach numbers andwith the serious problemof the interactions of internal shocks with boundary layers. Thereis also the problemof flow instability; under certain con- ditions the flow can become quiteviolently oscillatory. The Startng Problem I nternal supersonic deceleration in a converging passage(of nonporous walls) is not easy to establish. I n fact, as wewi l l now show, design conditions cannot be achieved without momentarily overspeeding the inlet air or varying the diffuser geometry. Thi s difficulty is due to shocks that arise during the deceleration pro- cess, and it need not be related to boundary layer behavior. Therefore let us neglect boundary layer effects for the moment, whileweexamine the starting be- havior of a converging-diverging diffuser that is one-dimensional and isentropic except for losses that occur in whatever (normal) shocks may be present. Thi s simpified analysis contains the essential features of the phenomena, and it could be avalid representation of a real flow fromwhich the wall boundary layer fluid was carefully removed by suction through porous walls. a CHAPTER 6 A EROT HERMODY NA MI CS OF I NL ETS, COM BUST ORS, AND NOZZLES Figure6.7 illustrates successive stepsin the acceleration of a fixed-geometry converging-diverging inlet. To isolate the inlet behavior fromthat of the rest of the propulsin device, weassumethat whatever is attached to the diffuser exit is always capable of ingesting the entire diffuser massflow. Thus massflow rateis limited only by choking at the mnimumdiffuser area>l/. Condition (a) illustrates low subsonic speed operation, for which the inlet is not choked. In this case the airflow through the inlet, and henee the upstream capture area/l^, is determined by conditions downstreamof the inlet. In condi- tion (b), though the flight velocity isstill subsonic, the flow is assumed to be ac- celerated to sonic velocity at the mnimumarea^,, andthe inlet massflow rateis limited by the choking condition at.4,. Sincethe flow is assumed isentropic, then A, A* and the upstreamcapture reaAa is given by ^ ^ ^ _ J _ _ 2 / y 1 \)/2(-y-l) A* A , ~ u[yT~ iy''~ Y~ ^)\ (315) For sufficiently high subsonic vales of M [see point (b) in Fig. 6.8], we have A a AQ A i ^~ A','^A,' Thus for theseconditions thecapture rea.4^must be less than the inlet area^,, and therefore spillagewill occur around the inlet. For sonic or supersonic flight speeds the spillage mechanismis necessarily non isentropic. That is, in order to "sense" the presence of the inlet and flow around it, the spilled air must be reduced to subsonic velocity upstreamof the M <! M <1 M <I M <1 (a) H -w^^*u ^Weak shock (b) M =I M <1 M =l M <1 Spillage (c) M >I \I : : : : : V xi M =I M <1 "Strong shock (d) M <1 M =l M <I (e, M =Mo-)-iM M =M M >1 M >] (eO *Weak ' M <] M >1 M =J shock (O 6.3 SUPERSONI C I NL ETS 229 (Detached shock) 0.1 0.2 0.4 0.6 1 Mz) M 6 10 Flight Mach number, M FK3URE 6.8 Accelerationand overspeeding of a one-dimensional supersonic inlet; A,Ax = 1.5. inlet plae. The mechanismfor this deceleration is adetached "bow wave" that standssufficiently far upstreamto allow the required spillage. One can imagine the process of establishing the detached shock wave as follows: Suppose that when the air first reached supersonic velocity there were no shock. Then flow would have to enter, without deviation, the entire inlet rea, in effect making^, act as the capture rea Aa. But for low supersonic Mach numbers (see Fig. 6.8, where >AalA"), the allowable capture rea as limited by chokingaty , is less thany4,. Henee therewould bean accumulation of mass andarisein pressure in the inlet. This pressurerisewould build uprapidly until ashock of sufficient strength moved upstreamagainst the supersonic flow and became established at a position that would allow the required spillage. Once the shock is established, the flow entering the inlet is no longer isen- tropic. Henee when the design Mach number of the aircraft is first reached, as at (d) in Fig. 6.7, the "reversed isentropic nozzle" mass flow cannot pass through the throat area>,. This follows fromEq. (3.14), which indicates that the choked massflow through agivenrea (A,) is proportional to po, and fromthe fact that the fluid suffers astagnation pressureloss in traversing the shock. For the flow that does enter the inlet (assuming isentropic flow fromapoint just downstream of the shock to the throat), Eq. (3.15) gives A2^ (A \ A, \A*)2 M2 _ y + 1 1 + y - 1 M2' (T+l)/2(y-l) (6.7) where the subscript 2refers to conditions just downstreamof anormal shock. We may assumethat the slightly curvedshock may beapproximated over the capture rea by a normal shock, so that M 2 may be expressed as a function of the up- streamMach number M by the normal shock relation M 2 = [2/(y - 1)] +M ' (6.8) 230 CHAPTER 6 A E R OT H E R M ODY N A M I C S OF I NL ETS, C OM B U ST OR S, A N D N OZZL E S Thesetwo expressionsmay becombinedto givetheratio of capturereato throat QTca,AjA, as afunctionof flight Mach number for the flow that includesade- tached normal shock. Theresult, plottedinFig. 6.8, indicates that the inlet rea Ai will remain too largeand spillagewill continu evenbeyondthe design Mach number M/), unless the inlet canbe overspeededto aMach number Mo. A t this Mach number [or just below it, as at (e)], the inlet iscapableof ingestingthe en- tireincident mass flow without spillage. Theshock positionwill bejust onthelip of theinlet, as inFig. 6.7(e), andaslight increment inspeed, as to {e'), will cause theshock to enter the convergence. Sinceashock cannot attain astableposition within the convergence,^it will movequickly downstreamto cometo rest within the divergence, at apositiondetermined by downstreamconditions. Theflow to the throat is now isentropic, and the rea xaXxo A^/A* (now greater than ^/y ,, sinceA, > A*) would begiven by (e') of Fig. 6.8. Theincoming flow is deceler- atedfrom^, tOv4 whereuponit isreaccelerated supersonically inthedivergence. Having thus attained isentropic flow in the inlet, the Mach number may be re- duced from Mo to M D , as at (f). A t exactly the designspeed, the throat Mach ^Suppose that conditionsare such that astationary shock exists someplacewithin the convergence. Twoconditionsare possiblefor the flow downstreamof the shock: Either it is choked at A, or it isnot choked. I f it ischoked, the mass flow through the throat is eivenbv E q. (3.14), m, po A, VR :V~y ( 2 \ < T+ 1 V2 ( T- 1 ) ' '^PQ2, where, for constant inlet velocity (flight speed), ^is aconstant andpo2 sthe stagnation pressure downstream of the shock. Nowsuppose that asmail disturbance moves the shock slightly upstreamwhere the shock M ach number is slightly higher. A greater stagnation pressure losswi l l occur across (he shock, lowering the downstreamstagnation pressure and the throat mass flow. Sinceconditionsupstreamof the shock are not affected, the mass flow through the shock remains constant and there resuhs an accumulationof mass and arisein pressure behind the shock. A sthe static pressure ratio across the shock increases, il s propagation speed relativeto the fluid increases and the shock moves farther upstream. T his further increases the stagnation pressure loss, of course. so that the mass flow imbalance increases and the shock contines to moveupstreamand out of the inlet. T he final positionof the shock is far enough outside the inlet rea to allow the correct "spil l ing" of air. By similar argumenis one can see that slight motiondownstreamby the shock iscontinued until the shock is carried through the throat. I f the flow downstreamo Ai is not choked, an appropriate downstreamboundary condition might be that at some exit area/l ^aconstant static pressure is maintained. I n this case, for steady flow, rh, le = PcUeAt A, pe RZ V2C,(To-Z). A ssuming isentropicflow downstreamof the shock, we can express this as 1 2y p,2'\ (r-i)/r y - ^ \ p j [ \ p j Thus it can be seen that, sincem, is dependent only onpoj, changes inshock positionresult in a mass flow imbalancesimilar to that for the chokedcase, and the same conclusionshold: Since any shock would exist inthe convergence by moving into it, it follows that no shock can reach a stable positionwithin the convergence. 6.3 SUPERSONI C I NL ETS 231 number would bejust unity and isentropic decelerationfromsupersonic to sub- sonic flow would exist. Even for this simpified model, however, thiscondition would beunstable. A slight decrease of flight speed or increaseof back pressure would requirespillage, and ashock would moverapidly out through the inlet to reestablish condition(d). Thus it might be better to maintain the throat Mach number slightly greater than unity while reachingsubsonic velocitiesthrough a very weak shock just downstreamof the throat. This simpified description contains the essential feature of the starting problemassociatedwith aninternally contractingpassage. That is, aninlet having Ai/A, greater than 1will always require spillageuponreachingsupersonic flight velocities, s'mcQ Aa/At will always passthrougham nimumof 1just assonic flight velocity isattained. It isnecessary to performsomeoperationother than simply acceleratingto the designspeed inorder to "swallow" the startingshock and es- tablish isentropicflow. Overspeedingisonesuchoperation, but thereare others. I f overspeedingisnot feasible(notethat, except for very modest designMach numbers, substantial overspeedwould berequired), it might bepossibleto swal- low theshock by avariationof geometry at constant flight speed. Theprincipieis easily seen in terms of simple one-dimensional analysis. Suppose the inlet is accelerated to thedesignMach number M with thestartingshock present, as at point (d) inFigs. 6.8and6.9. Now, if the actual rea ratio canbe decreasedfrom AJA, to the valu that can ingest the entire inlet flow behind the shock, the shock will be swallowed to take up a positiondownstreamof the throat. This variation would normally involve amomcntary increase of throat rea from^, to a newvalu that wewill cal M / (see Fig. 6.9). Having thus achievedisentropic flowwithin the convergence, the throat Mach number M ^isgreater than 1, and a relatively strongshock occurs farther downstream. Completely isentropic flow M >l \ A/A* M >1 M =M' M <1 (dO FIGURE 6.9 Shock svi/allowing by areavariation. 232 CHAPTER 6 A EROT HERMODY NA MI CS OF I NL ETS, COM BUST ORS. A ND NOZZL E S can then be achieved by returning the rea ratio to its original valu, while the operating conditions movefrom(d) to (f). A geometric variation such as that shown schematically in Fig. 6.9would be difficult, mechanically, for axisymmetric flow. However, geometries that permit the axial motion of acenter plug between nonparallel walls can be used. A some- what similar effect can be had through the use of porous convergent walls. With the shock external, the high static pressurewithin the convergencecauses consid- erable "leakage" through the walls, thus effectively increasing the throat area to permit shock swallowing. Once this occurs, the lower static pressurewithin the convergence somewhat decreases the flow loss. But since the total porosity re- quired is even greater than the throat area [8], thereremains a high mass flow loss under operating conditions. This is desirable to the extent that it removes boundary layer fluid, but the lost flow is more than is needed for this purpose. Flow that is decelerated but not used internally contributes substantially to the drag of the propulsin device. Fromthis discussion it would seemthat onecould avoid the starting problem altogether by using a simple divergent inlet, that is, one for which A,/At = 1, as shown in Fig. 6.10. For supersonic flight speeds and sufficiently low back pres- sure, it is possibleto accelerate the internal flow within the divergencebefore de- celerating it in a shock. To reduce stagnation pressure losses, it is desirable to have the shock occur at the m nimumpossibleMach number, which, for this ge- ometry, is the flight Mach number. One can achieve this condition by adjusting the back pressure (by varying the engine exhaust area, for example) so that the shock is positioned just on the inlet lip. A slight improvement, called the Kantrowitz-Donaldson inlet by Foa [9], is also illustrated in Fig. 6.10. This configuration uses the mximuminternal con- vergence that will just permit shock swallowing at the designflight Mach number. Referring to Fig. 6.9, wecan seethat this '\s]usiAi/A,- for adesign Mach number Simple diverging inlet M =M M>My> L ow back pressure M <] M =M / , "Best" back pressure Kantrowitz-Donaldson inlet M <l M =M; j M =M' M < i L ow back pressure M =M / ; M =M ' M <l "Best" back pressure FIGURE 6.10 Fixed-geometry diffuser with intentional normal shocks at the design Mach number M o (Refer to Fig. 6.9.) 6.3 SUPERSONI C I NL ETS 233 of M D - A S in the simpledivergent inlet, it is necessary to adjust theback pressure to assurethat the shock occurs at the m nimumpossible Mach number. The ad- vantage of the internal convergence is, of course, that this m nimumMach num- ber is less than the free-stream valu. A disadvantage is that thereis an abrupt change in performance just at the design Mach number, since the shock wil l not be swallowed below this valu and it will be immediately disgorged if the speed falls off slightly fromthe designvalu. Further, sinceshocks of any origin are un- stable within the convergence, such an inlet would be quite sensitive to changes in angleof attack. Thecondition inwhich ashock just hangs on the inlet lip is called the critical condition. Operation with the shock swallowed is calledsupercritical, whereas that with adetached shock and spillageis calledsubcritical. Notethat subcritical operation may occur as the result of choking in the inlet, as discussed here for inlets alone; or, for a complete engine, as the result of any downstream flow reslriction that cannot accept the entire massflow p^UaA,. The Shock-Boundary Layer Problem A s wenoted earlier, sketches such as Figs. 6.9(d') and 6.10 are realistic only if the boundary layer fluid is removed fromthe walls between which the shock is located. Across ashock waveof appreciablestrength, theboundary layer seprales, and this separation may have a large effect on the structure of the shock. We might suppose, in accord with the boundary layer concepts discussed in Chap- ter 4, that aturbulent boundary layer will sepratewhen the shock-induced veloc- ity reduction is between 20%and 30%(0.36<Cp< 0.51). Using thenormal shock relationships of Chapter 3, we can see that this would correspond to upstream Mach numbers of 1.15 (20%reduction in velocity) or 1.25 (30%reduction). To begin, weconsider the interaction of aweak shock with aboundary layer. Figure6.11 is aseries of schlieren photographs of the interaction of ashock wave and aboundary layer for upstreamfree-stream Mach numbers ranging from 1.30 to 1.55. Schlieren photographs show regions of high-density gradient [7], so they reveal sharply the presence of stationary shock waves as well as the presence of boundary layers in acompressible flow. In Fig. 6.11 (a) wecan seewhat is nearly aplae normal shock wave(with an upstreamMach number M of 1.3) and awall boundary layer whose thickness is growing perceptably in the flow direction. I n Fig. 6.1 l(d), for M = 1.4, the shock wavenear the boundary layer has taken up a *'lambda" shape, with an obiiqueshock reaching ahead of the main shock loca- tion. Because alargefraction of the boundary layer is subsonic, the pressurerise dueto the shock is sensed (near the wall) somedistance ahead of the main shock wave. This upstreampressuregradient causes the boundary layer to grow rapidly, as onecan seefromciseinspection of Fig. 6.11(d). Theincrease in displacement thickness affects the flow in the same way as would a wedge in inviscid flow; heneean obiiqueshock is formed ahead of the main wave. A s thefree-stream Mach number increases aboveperhaps 1.25, the boundary layer thickens very rapidly under this lambdashock system(see, e.g., Fig. 6.11(i)) and causes the boundary layer to seprate. The pressure gradient near the wall ^ c n u i n t HMUUY NA I V i Ub Uh INLb F, i . COMBUST ORS A N D N OZZL E S 6.3 SUPERSONI C I NL ETS J O has becometoo largefor the slow-movi ngflui d near thewall to conti nu movi ng i n the mai n flow di recti on. Fi gure 6.12 shows the effect of astrong shock i n adi vergi ng duct. Here the mai n stream-flow adjustment i s very far fromthesi mplenormal shock pi cturei n Fi g. 6.11. Instead the i nteracti on between the "shock" and the boundary layer results i n at least fi vc major pressure adjustments (shown by theA'-li kc dark re- gi ons i nFi g. 6.12). Largeseparati on zones causeahi ghly di storted, and probably unsteady, flow fi eld that may requi rean axi al di stanceof 10 duct wi dths or more to return to reasonable uni formty of flow. Theflow fi eld di sturbances and di s- torti ons shown i nFi g. 6.12would haveseri ously harmful effects on thebehavi or of acompressor or combustor placed i mmedi ately downstream. The chi ef lesson herei s that unless onemakes astrenuous effort to remove thewall boundary layer, strong shocks may havedi sastrous effects on duct flow. If ashock wavemust beplaced i n asupersoni c streamof gi ven Mach number, then: a. A n obi i queshock i s much better than anormal onebecause the pressure ri sei s less; b. Theshock should i nteract wi th thewall al the poi nt wherethe boundary layer i s thi nnestperferably at the leadi ng edgefor the si mpledi vergi ng i nlet of Fi g. 6.10. External Deceleration The absence of astarti ng problemfor thenormal shock i nlet i s offset by the accompanyi ng stagnati on pressure loss at all but low fli ght Mach numbers (less than about 1.5). For example, Fi g. 3.8 i ndi cates a28%stagnati on pressure loss at M =2 (7 = 1.4). f wewi sh to obtai n reasonableperformance whi lemai ntalni ng the starti ng characteri sti cs of asi mpledi vergent i nlet, then i t i s clear that some externa! decelerati on must occur upstreamof the i nlet plae i n order to reduce the Mach number of the normal shock to asui tablevalu. The si mplest and most practi cal external decelerati on mechani smi s an ob- i i queshock or, i n somecases, aseri es of obi i queshocks. Though such shocks are FIGURE 6.12 Shock-Dounaary layer i nteracti ons i n a duct, Flow fromlef to rtght. (Courtesy M. I T Gas Turbi neLaboratory. ) 236 CHAPTER 6 A E ROT HE RM ODY NA M I CS OF I NL ETS, COM BUST ORS, A ND N OZZL E S not isentropic, the stagnation pressureloss in reaching subsonic velocity through a series of obiiqueshocks followed by anormal shock is less than that accompa- nying asinglenormal shock at the flight velocity. Thelosses decreaseas the num- ber of obiiqueshocks increases, especially at highflight Mach numbers. I n the external compression process, shocks and boundary layers may inter- act strongly, so that it is highly desirable to lcate the obiiqueshocks at points whereboundary layers are absent. This can bearranged easily if oneuses a center body (primarily for axisymmetric flow), as in Fig. 6.13. These schematics, taken fromOswatitsch [10], illustrate the typical singleobiiqueshock systemand two double obiique shock systems. Although the double shock systems theoretically givebetter performance, onecan seethat several problems may arise in their use. I n the configuration of Fig. 6.13(b), the second shock, generated by aturn of the wall, occurs at a point where the boundary layer has had time to develop, and separation may result. The configuration of 6.13(c) avoids this at the point of shock generation, but the second shock still intersects the boundary layer on the center body. With (if need be) boundary layer removal, wecan appreciate the performance gain to be expected through the use of mltiple oblique-shock deceleration by looking at a two-dimensional example. Consider the diffuser in Fig. 6.14, in which the flow is deflected through two 15 angles before entering a normal shock. Using Fig. 3.12, wecan show that the Mach numbers in regions (2), (3), and ( ) are 2.26, 1.65, and 0.67, respectively. Then, usingFig. 3.11, wecan obtain the stagnation pressure ratios as follows: H =0.895, ^ =0.945, ^ = 0.870. Poi p02 P03 Thus the overall stagnation pressureratio is approximately poVpoi =0.735. If the deceleration had been achieved by a single normal shock, the overall stagnation pressureratio would havebeen only 0.33. Remember that theseestimates do not includelosses due to boundary layer effects, which may beespecially important in the subsonic diffuser. This example does not necessarily employ the best arrangement of three shocks, of course, since a variation of their relative strengths might provide a higher overall stagnation pressureratio. For the simpletwo-dimensional case, in 6.3 SUPERSONI C I NL ETS M =3 FIGURE 6.14 Two-dimensional diffuser. which conditions downstream of each shock are uniform, Oswatitsch [10] has shown by theoretical analysis that, for a givenflight Mach number and agiven number of obiiqueshocks followed by one normal shock, the overall stagnation pressure ratio will be maximized if all the obiique shocks have equal strength, that is, if the Mach numbers of the velocity components normal to each shock, and incident to it, are equal. It follows, of course, that the stagnation pressure ratios will also be equal. This result will not seemunreasonable if werecall that for a normal shock the stagnation pressureloss rises very quickly with incident Mach number. Oswatitsch found that the Mach number of the final normal shock should be less than the normal component of the obiique shocks (about 0.94 times as great). Figure6.15 shows the best performance usingn obiiqueshocks of equal strength followed by one normal shock. For axisymmetric inlets and conical shocks, the best arrangement is not so easily determined, sincethe downstreamflow of aconical shock js not uniform. I n fact, fluid properties are constant alongconical surfaces emanating fromthe shock vrtex, and the streamlines downstreamof the shock are curved [11]. The effect of the streamline curvature is further diffusion, which can bevery nearly isentropic, so that subsequent shocks occur at reduced (but nonuniform) Mach numbers. It is even possibleto achievedeceleration to subsonic velocities behind aconical shock without subsequent shocks, as shown for arather low Mach num- ber inFig. 6.16. Theperformance advantageof mltipleconical shocks is qualitatively similar to that of the multiple-planeshocks shown in Fig. 6.15. Going to the limit (for FIGURE 6.15 Best stagnation pressure ratio for dilfuser consistingof n obiiquesfiocksof uniformstrength and one normal shock; y = 1.4. (After Oswatitsch [10].) 238 CHAPTER 6 A EROT HERM ODY NA M I CS OF I NL ETS, COMBUST ORS, A ND NOZZL E S FIGURE 6.16 Typical conical-flow streamlines. (After Shapiro [12].) either axisymmetric or two-dimensional geometries) of an infinite number of infinitesimal shocks, generated by continuouswall curvature, onecould, theoreti- cally at least, achieve isentropic external deceleration to sonic velocity. Such an inlet, indicated qualitatively in Fig. 6.17, would seemto providethe ideal geome- try to achievelow losses, whileat thesametimeavoiding thestarting problems of an internal convergence. However, several practical difficulties would beencoun- tered in theoperation of such an inlet. Thi s geometry, like that of theisentropic internal flow diffuser, would function properly at only oneMach number, and per- formance would bevery sensitiveto angleof attack. Furthermore, the boundary layer along the curved surface, unlikethat alongplae or conical surfaces, would besubject to a high adverse pressuregradient, which might causeseparation. Fi - nally, for high flight Mach numbers it would be necessary that the flow turn through largeangles beforereaching sonic velocity. Theresultant largecowl angle would at least exhibit high drag and perhaps even interfere with the inlet flow. It will be clear fromthe foregoing that variablegeometry is an almost ines- capable requirement for an engine inlet that must oprate at both subsonic and supersonic speeds. Fig. 6.18 shows the design of the two-dimensional intake adopted for the Concorde aircraft, whose design flight Mach number is 2; the Concorde is also required to cruiseover certain landreas at subsonic speeds. Figure6.18shows the intakegeometry during takeoff; herethe ramp assem- bly is raised to allow as much air as possibleto the engine. Shock waves areof FIGURE 6.17 I sentropic external diffuser. 6.3 SUPERSONI C I NL ETS 239 Boundary layer Nozzlesecondary Take-off Supersonic flight FIGURE 6.18 Cross sections of variable-geometry intake for Concordesupersonic aircraft, (Courtesy Rolls-Royce, pie.) course absent as the air enters the enginewith a Mach number of about 0.5. A s the Concorde reaches a flight Mach number of about 0.7, the auxiliary door ctoses. Above a flight Mach number of 1.3, the ramps are progressively lowered; theforward ramp controls the position of theobiiqueshock waves that decelrate the airstreamfromsupersonic to subsonic speeds at the engine intake. The Flow Stabiilty Problem Theoperation of external shock diffusers isdivided into subcritical, critical, and supercriticial modes, which depend on the external and internal shock con- figuration. Thethreenodes areshown schematically for a typical casein Fi g. 6.19. With entirely diverging internal flow such as this, the normal shock position is determined by a downstreamflow restriction rather than by the inlet geometry. Henee the operating mode is sensitive to variations in exhaust-nozzle area and fuel flow rate. Subcritical operation entails "spilling" of flow and a normal shock upstreamof the inlet. "L ow" and "high" subcritical operalions differ only jn the extent of spilling. Supercritical operation occurs at the samemassflow as critical operation, but with increased losses, since the normal shock occurs at a higher Mach number. Numerous investigators have tested the performance of experimental super- sonic diffusers, owing to the substantial importance of inlet performance, espe- cially at highMach numbers. A widevariety of geometries havebeen considered, each depending on a particular application. Special attention is paid to the off- design performance of an inlet; this is of great importance in an actual flight application but is not so amenable to analysis as the design performance. Various adjustable inlets havebeen considered to extend the favorableoperating rangeof an inlet. The data of Dailey [13] are typical of many investigations of an important instability that occurs during the subcritical operation of most supersonic inlets. This phenomenon, known as "buzz," consists of a rapid oscillation of the inlet 240 CHAPTER 6 A EROT HERMODY NA MI CS OF I NLETS, COMBUSTORS, A ND NOZZL ES Cri ti cal Supercritical FIGURE 6.19 Typical modes of inlet operation. shock and flow pattern; the resuhant internal disturbance is very detrimental to engine performance. I n a ramjet, for instance, the onsct of buzz usually ex- tinguishes combustin. Although the pulses of the shock systemare similar, the interval between pulses is not constant [13]; heneebuzz cannot be considered a periodic phenomenon. Although it is not thoroughly understood at present, buzz has been shown to be a function of conditions only at, and immediately down- streamof, the inlet. I n some cases boundary layer bleed fromthe center body can delay the onset of buzz. I n other cases the use of a length of nondivergent sub- sonic passagejust downstreamof the inlet can have a beneficial effect. This latter casecould be attributed either to the establishment of a healthy boundary layer (as compared with a separated one just downstreamof the normal shock) or to the establishment of moreuniformfree-streamconditions (see below) before sub- sonic diffusion is attempted. Figure 6.20 is a plot of typical diffuser performance as a function of mass flow, expressed as the ratio of actual to critical (or supercritical) mass flow rate rhc, or the equivalent ratio of actual to critical capture area. Ratios of m/rhc less than 1 signify subcritical operation. One can see that for mass flows slightly less than critical (the so-called high-subcritical flow) the stagnation pressure ratio in- creases slightly with decrease i n mass flow rate. This is explained by the presence of lower aerodynamic losses in the internal passages due to reduced mass flow and velocity. Overall performance, of course, does not increase with subcritical operation, since spilling is accompanied by both increased drag and decreased thrust. A s the flow is further reduced, the normal shock is pushed farther streamand the stagnation pressure ratio decreases. A portion of the air entering thecowl may travel through a single strong shock (see "low subcritical," Fi g. 6.19) 6.3 SUPERSONI C I NLETS 241 0.1 30 Cone diffuser Mo=I .96 - Cowl-I ip coordinate angle= 45.0^ Cone wave angle=48.8 O 2 h [^ote: Cowl central-body eccentricity=1.5% cowl entrance diameter.] / 0.2 0.4 0.6 m/ihc (a) 0.8 1.0 P02 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Entrance of high-entropy air ^ Constant sonic- l '^upercri ti cal throttle area. / / / / /Cri ti cal / 30 Cone diffuser Mo=1.96 Cowl -l i p coordinate angle=45.6'^ Cone wave angle=48.8 [Note: Cowl central-body eccentricity =0.5% cowl entrance diameter.] L / / o 0.2 0.4 0.6 0.8 1.0 m/ic (b) FIGURE 6.20 Typical supersonic inlet performances. (Courtesy Dailey [13].) rather than the two weaker shocks it would pass through during critical opera- tion. The entrance of this low-stagnation pressure (or high-entropy) air causes a reduction in performance. In supercritical operation the stagnation pressure ratio drops rather rapidly at constant mass flCA- rate. We can see by comparing Figs. 6.20(a) and (b) that the ingestin of l ow- stagnation pressure air may or may not correspond to the onset of buzz (ahhough it usually does). For both subsonic and supersonic intakes, the best balance of external and internal deceleration is a matter of concern because of the need to keep nacelle drag as low as possible and to supply uniformflow to the engine. For the sub- sonic intake, fixed geometry can generally provide acceptable boundary layer be- havior over all flight speeds fromtakeoff to cruise. For supersonic flight, shock waves are a practical necessity for deceleration of the inlet air to subsonic speeds; the inlet geometry required to minimize shock and boundary layer losses depends strongly on Mach number. This means that to copewith a wide range of flight Mach numbers, variable geometry is needed. The control of the intake geome- try must be done carefully to avoid flow instability. Large perturbations i n the geometry of the inlet shock pattern could disturb the engine intakeflow suffi- ciently to cause serious problems for the compressor or burner.