US 2819012 A Description (OCR text may contain errors) 2 Sheets-Sheet 1 Filed Dec. 22, 1950 Gttornegs Jan, 7, 1958 R. P. ATKINSON CENTRIF'UGAL COMPRESSOR Filed Dec. 22, 1950 ENT VELOClTY A IC DENSITY 2 Sheecs--ShLaMl 2 United States Patent O 2,819,012 CENTRIFUGAL COMPRESSOR Robert P. Atkinson, Indianapolis, Ind., assignor to General Motors Corporation, Detroit, Mich., a corporation of Delaware Application December 22, 1950, Serial No. 202,273 11 Claims. (Cl. 2313-119) My invention relatos to centrifugal compressors, more particularly to compressors of aknown type' in which air enters the compressor parallel to its axis of rotation, engages an inducer portion of the compressor rotor provided with generally helical blading in which the air ows in a predominantly axial direction and is accelerated tangentially, and then proceeds through an impeller portion of the rotor formed with substantially radial vanes in which the air is accelerated tangentially and radially, leaving the rotor with a high tangential velocity. In such a compressor, the air discharged from the rotor is received in a diffuser in which the Velocity head of the air is largely converted to static head and the air is directed to the outlet or outlets of the compressor. It is to beunderstood, however, that this invention is concerned not with the diffuser as such but rather with the form of the rotor. The invention is particularly intended for and highly valuable in high performance centrifugal compressors such as are used in gas turbine engines. lt is well known that compressors for this service ordinarily must handle a large volume of air, must operate at a high compression ratio, and must be of high efficiency. The adoption of the principles of the invention in a previously known type of gas turbine engine has resulted in remarkable increases in the power and eliiciency of the engine. rThe principal objects of the invention are to provide a centrifugal compressor of high capacity, one of high compression ratio, and one of high efficiency. Further objects of the invention are to increase the flexibility and stability of centrifugal compressors and to improve the performance of such compressors and of motive power systems embodying compressors. A further object of the invention is to extend the range of usefulness of centrifugal compressors, especially single stage compressors, and thereby render more advantageous the use of the relatively simple and reliable centrifugal compressor in gas turbine engines, where heretofore the relatively temperamental, expensive, complicated, and delicate axial-110W compressor has been considered to be more suitable. Expressed in another way, the objects of the invention may be defined as to provide a centrifugal compressor in which the ow path delined by the structure of the compressor is best accommodated to the natural path of flow of the gaseous medium, in which the distribution of pressure at any section of the passages of the compressor is as uniform as practicable, in which velocity distribution is uniform across the passages to a much greater extent than hitherto, and in which the tendency to supersonic flow and choking is suppressed to a greater extent and over a wider range of operating conditions than hitherto considered possible. The principles ofthe invention and the manner in which they are applied to obtain the highly important advantages of the invention are difficult to summarize, but will be apparent to those skilled in the art from the subsequent detailed description of the principles of the invention and et a compressor embodying those principles. 2,319,912 .'lfatented Jan. 7, 195:@ Referring to the drawings: Figure 1 is a diagrammatic sectional view of a single stage single entry radial flow compressor, taken on a plane containing the axis of rotation; Figure 2 is a diagram illustrating the blade thickness; Figure 3 is a diagram illustrating the preferred form of inducer blade; Figure 4 is a vector diagram illustrating conditions at the compressor inlet; Figure 5 is a vector diagram illustrative of flow conditions within the rotor; Figure 6 is a chart illustrating the variation of static density, component velocity, and net area of the path through the rotor; Figure 7 is` a diagram expository of the principles of the invention; Figure 8 is: a diagram illustrating the construction of the mean flow path or centroid ofthe rotor; and Figure 9 is a diagram illustrating the construction of the boundaries of the flow path. Although 'compressors of the general type to which the invention relates are well known, it is believed advisable to describe briefly the structure of such a compressor in order to eliminate any possible misunderstanding as to terminology in the specification. Referring to Figure l, the compressor illustrated diagrammatically comprises a rotor 10 mounted on a shaft 11 for rotation therewith. The rotor comprises an inducer portion l?. and an impeller portion 13. While these may be integral, they are ordinarily made separately for manufacturing reasons and are fixed together. The inducer abuts the impeller at a radial plane indicated at 14 and cornmonly called the split line. The shaft 11 is mounted in bearings indicated by bushings 16 and 17 in the xed structure of the compressor. The forward bearing 16 is mounted in a body 18 which defines the inner boundary of the compressor inlet 19. `An annular casing 21 defines the outer boundary of the inlet and of the liow path through the rotor 10. The frame member 22 which supports the rear bearing 17 may be a disk closely adjacent the rear face of impeller 13. The inducer 12 is fitted with radial blades 23 which are of curved form to accelerate the air tangentially, as will be described more fully. The impeller `13 is provided with radial blades 24 which ordinarily constitute continuations of the impeller blades 23, a continuous passage 25 for flow of air from the compressor inlet 19 to the compressor outlet 26 being defined between each adjacent set of rotor and impeller blades. Forl greater clarity, one complete blade 23, 24 is illustrated as viewed at right angles to the plane of the impeller blade 24 in Figure 1. As will be apparent, `the body of the rotor. constitutes the inner boundary of the air flow path through the rotor. The outer boundary could of course be defined by a shroud fixed to the rotor but may be defined by the fixed casing structure 21 adjacent the tips of the blades, as illustrated. The air discharged from the rotor enters a diffuser 28 which may be of any suitable form, and is illustrated schematically as provided with a plurality of discharge outlets 29. Although Figure 1 illustrates la single entry compressor, the principle of the invention is equally applicable to double entry and to multi-stage compressors. The form of the inlet mayA also be varied to suit the particular installation. The impeller blades `are ordinarily radial, but may be curved ahead or back. The nature of the invention may be indicated by stating that it involves providing a ilow path in the compressor conforming to the natural flow path of the air or other liuid (referred to as air for conciseness). The air entering the compressor is considered to have a known velocity parallel to the axis of rotation of the compressor, without radial or circumferential components. In the inducer portion of the rotor, the air is accelerated circumferentially of the axis, or tangentially, in as uniform a manner'as practicable, by the inducer vanes. As the Vair acquires `ataugential velocity, centrifugal force aai/aora' 13 acts upon the air in a radially outward direction. In the impeller, the angular velocity of the `air about the axis remains constant, but Ithe centrifugal force increases with the radius. There is thus a continuously increasing centrifugal force on a lparticle of fair as it :traverses the rotor from inletto-outlet. The `inertia of the air resists `the radial acceleration, and a particle or' :air thus .tends to follow a curved path, relative to the rotor as a frame .of reference, under the action of the initial or entrance velocity and ,the centrifugal force -exertedonthe particle. The problem of ,providing an .optimum ow path is complicated by thefact that .thefaircannot 4be `treated as a free particle under `the action of .centrifugal and @inertia forces, due to the effects ,of .frictiom tot .back pressure in the rotor outlet, and .of `variations in ,the density ofthe air and the areasof-.the ilowipath. A The rotor ,ofthe invention Yis formed Iso ythat the flow path provided for .the air conforms rto lthe path .of -frec movement 'due to the entrance velocity .and centrifugal force with allowance lfor the `various .modifying `vfactors just described. As `a result, the flow vis -not forced .from its natural path bythe physical boundaries of the channel defined by the compressor casing (or shroud, if present) and the surface of therotor. The .mean ,path :is determined for a yprescribed relation between component velocity (velocity in the ,plane containing theaxis) and distance along the -meanpath The area of the passage is modified along .the .length `of Vthe passage to v.provide relation. As will be apparent, .the tangential velocity yand the radial acceleration increase as the air moves radially outward through the impeller `portion .of the rotor, .and the final path of vthe Vair -(apart from the tangential component) approaches alradial direction. The air is discharged with .a high absolute velocity, the .major component of which .is tangential, and Ithe velocity -head is largely converted to static pressure in .the diffuser, which may be of known type. The principles yot :the invention `may be most clearly explained by disclosing 4the Aprocedure involved in designing a Acompressor according uto the invention to ymeet given design conditions. Obviously, `the size .and `speciiic form of the compressor will vary with Lsuch lparameters as diameter, speed, air .flow pressure, and the like, but such variations Vdo not affect the `'design procedure. The following -basic design conditions must be established or assumed: (l) pressure and ltemperature of air at the inlet, (2) air flow desired, weight per unit time, (3') rotor speed, revolutions per unit time, (4) lassumed etiicicncy, and ('5) assumed vpressure ratio (total outlet pressure to total inlet pnessure). O'f these items, l), (2), (3), and (5) depend upon the installation for which the compressor is built. For va gas turbine engine, `pressure and temperature of `the air .may be assumed atmospheric conditions. The ,desired air liow would 'be determined by the engine air requirements. The rotor speed is usually that of the turbine, ,and is limited by the necessity to avoid excessive tangential velocities, Yand thereby by the diameter of the rotor. The value assumed for eiliciency is based upon experience', and 'for vthe `compressor of the invention experience indicates a'value oi 80% to be suitable for designputposcs. The vassumed pressure ratio is determined bythe rotortipspeed'niof course, must not exceed whatfcan be realized, 'The values assumed must be compatible, `and the selection -of design parameters may involve trial and error 'to some extent. Certain dimensional values must also b'e' assumed: (l) impeller radius, ('2) outer and inner radiioff the annular inlet, (3) impeller blade width at the outlet, Y(4,) axial length of the inducer portion, (5) number of `blades, and (6) blade thickness, normally f expressed in terms lof thickness at the 4tip and at the rootof the blade. The impeller radius isordinarilyflimited by the dimen` vthe aibovementioned prescribed 4 sions of `the engine. It is related to speed, as excessive centrifugal forces in the rotor must be avoided. -'Pressure ratio is determined primarily by the speed and diameter of the rotor. The inlet diameter is based on a compromise between having a large diameter to maintain a reasonably low inlet velocity and a small diameter to reduce `the relative Mach number at :the outer diameter ofthe inlet eye. The outlet width of the impeller is based in the usual manner on the dow conditions .at the outlet, that is, .upon the area necessary lfor iiow Iof the assumed weightof air and lassumed outlet conditions. it is not critical, and may be varied over a rather -wide vrange to suit .installation requirements. The length of the `induceiumay `be limited by the allowable overall length of the rotor and machine tool require ments. in some installations, especially double-sided compressors and cases i-n which a .compressor is substituted in a previously existing engine, it .may not `be possible to malte the inducer of optimum length. 'In the `inducer, the tangential velocity of the .air is increased at roughly constant radius. In the impeller, the .tangential velocity is increased as the air flows outward, tangential velocity being proportional to the radius. There is a gradual 'transition from axial iiow at the inducer entrance to nearly .radial ilow at the rimpeller outlet. As indicated -above, installation conditions may dictate `an inducer shorter than optimum, but, in general, an etiort should be made to provide an 'inducer portion of such length that tangential acceleration Vin the inducer approximately equals that in the impeller. The number of blades is a matter of informed choice vbased upon 'test and experience. With too few blades, there is excessive turbulence; with too many blades, excessive friction. A small number of blades will give rise to Igreater pressure liuctuations The use of a prime number of blades may reduce resonance. The blade thickness Tis a matter of structural design to obtain suliicient mechanical strength combined with the required vibration characteristics .and a structure adapted to manufacturing processes. Figure 2 illustrates blade thickness in the Vimpeller, the thickness at the tip of the lblade being indicated by aand at the root of the blade by b. As Willlbe apparent 'to those skilledin the art, the above discussion of design procedure largely relates to background material, as these or equivalent assumptions are necessary for the design of compressors ingeneral .of the type to Vwhich the invention relates. vFor this reason, the computations and the underlying theory .are not discussed in detail. From the design conditions and physical dimensions assumed, the area, velocity of ow, static density, total pressure, and total temperature, at the rotor inlet and outlet, may be computed in'known manner. The `preferred inducer blading, upon which the subsequent discussion is based, is ysuch that the .line of intersection of a blade with any plane perpendicular to the axis is radial, and the intersection of a blade by a cylinder concentric with the axis is a parabolic curve inthe cylin der. vOtherwise expressed, the blade is of the form, in cylindrical coordinates, T is proportional to X2, where X is the distance measured along the axis from vthe discharge end of the inducer and T is the angular displacement of a radial 'blade element relative to the discharge end of 'the blade. At any given radius, we may state Vthis as Y=KX2, where Y is the length of arc through which the blade is displaced, .K is a constant to be determined, and X is as before. This relation is illustrated in Figure 3. The first step in the determination of the constant K is the determination of the angle of attack of the inducer blade at the centroid ofthe inlet passage. VIn this specilication, the term centroid is defined as the mean path; that is, Athe radius of the centroid is the radius -which divides the ow path into two portions of equal area. asienta The radius Roof the centroid at the inlet may be determined for this purpose from the dimensions of an inducer passage, fixed by the inner and outer radii of the inlet and the number and thickness of the blades, by graphical methods. The tangential velocity Uc of the impeller at the centroid is the product of Rc in feet by the angular velocity of the rotor in radians per second. The air velocity, assumed parallel to the axis, equals the air tlow in weight units divided by the product of the air density and the total area of the inlet annulus (the inlet 19) and is represented by V1. Given these values, the relative velocity of the air with respect to the rotor, Vr, may be determined as indicated by the Vector diagram of Figure 4. Experience has shown that an angle of attack of approximately 10 degrees at the inducer entrance will give peak eiiiciency, due presumably to the resulting increase in eective inlet area. Too great an angle of attack will result in stalling at the entrance. The angle Q1 (Figs. 3 and 4) of the inducer blades at entrance is therefore taken as ten degrees less than the angle of Vr at the centroid to the axis. Since the blade equation is where XT is the known length of the inducer. This equation may be solved for K, which determines the form of the inducer blading. It may be noted that, since the blades are radial, the angle Q and the angle of attack vary with the radius. So much having been done, it is now in order to begin the determination of the mean path through the rotor and from the mean path the inner and outer boundaries of the path of air through the rotor. The first stage in this operation is to prepare a plot of component velocity against path length, as illustrated in Figure 6. The path length is the length of the projection on a plane containing the rotor axis of the path through the entire rotor of a particle of air entering the compressor at the centroid of the inlet, and must be estimated from the dimensions of the rotor. The component velocity Vc is velocity in a plane containing the rotor axis. In other words, component velocity is the resultant of the axial and radial components of the absolute velocity of a particle. This velocity at inlet and outlet is likewise computed from the stated design conditions. Component velocity at Zero path length is that due to total area of the inlet. Component velocity at one halt` inch from the inlet is increased by the decrease in path area due to the inducer blades, and may be computed from total llow and net area of the path (static density assumed to equal that at the inlet). From the one half inch point, component velocity is assumed to increase linearly with path length. Y Figure 5 illustrates the relationship between U, tangential velocity of the impeller, Vr, velocity of the air relative to the impeller, and V, absolute velocity of the air, which is the resultant of U and V1. It also illustrates the rectangular components Vc and V1, of V taken respectively in a plane containing the axis of the impeller and normal to this plane. Vc is the component velocity and V1 the tangential Velocity of the air. It will be noted that the design assumption of linear increase in component velocity is based upon the desired result of smooth continuous changes in Vc, a condition believed to promote efficiency of compression. The ultimate form of the rotor will be such as to justify this assumption. The principle underlying the construction of the optimum centroid of the mean air liow path is illustrated in Figure 7, which shows diagrammatically a single-sided impeller with axis of rotation X-X. The centroid of the air path, to be determined, is C-C. Assuming an elementary particle of air at a point P on the centroid, this particle will be subject to a centrifugal force F1 due t0 its rotation about the rotor axis. F1 forunit mass will equal tangential velocity VT squared divided by the radius R1. The radius is defined by the position of P and equals O1P. The tangential velocity may be calculated. As indicated in Figure 5, the tangential velocity equals U Vc tan Q. V is the product of R1 and the angular velocity of the rotor in radians per second. Vc may be taken from the chart (Fig. 6). Q is determined by the form of the inducer blading and the radial and axial coordinates of point P. The air iiow path obviously curves outwardly, the problem being to define the path of natural iiow of the air so as to form the impeller to avoid dellecting the air from this path. Also, the curve C-C will clearly be a continuous curve. It will, therefore, have a center of curvature corresponding to the point P, indicated on the diagram at O2. The location of O2 is xed by the form of the curve C C and the location of P. The elementary particle of air at P is rotating about the instantaneous center O2 at a radius R2 equal to the radius of curvature of C-C at point P. The velocity of P in this path is Vc, the component velocity (Figs. 5 and 6). The particle will therefore be subject to a centrifugal force F2 directed away from O2, the magnitude of which, for unit mass, equals VGZ/R2. The resultant of F1 and F2. is indicated by F. If F is tangent to C-C at P, the resultant force on the particle acts along the centroid and there is no tendency for the particle to `depart from the assumed centroid. It the curvature of the centroid is too great, the resultant F will be directed inwardly from the tangent to C-C, and the actual air ow in the rotor will crowd against the inner boundary of the flow path. Conversely, if the curvature is too small, the air liow will trend toward the outer boundary of the rotor. The centroid must therefore be determined in accordance with the principle illustrated by Figure 7. Conceivably, this problem might be solved by deriving an equation deiining the locus C-C. However, the mathematical difliculties appear to be insuperable in View of the complication of the problem by such factors as friction, changing density of the air, and varying area of the liow path. Various step-bystep methods of computation of the centroid may be employed. I presently prefer a partly graphical method, illustrated in Figure 8, in which the centroid is constructed by drawing short circular arcs about assumed centers. The resulting values for each increment are checked, and the assumed center varied, until the results are consistent. By repeating this process, the entire curve is constructed. Referring again to Figure 7, since F2 is normal to the centroid and F should be tangent to the centroid, the desired conditions will result only when F is normal to F2. The component of F1 normal to F must therefore equal F2. This component equals F1 `cos B, where B is the angle between F1 and R2, which obviously equals the angle between the tangent to `C-C at P and the axis X-X. Substituting in the equation F1 cos B=F2 the values of F1 and F2 previously stated, and solving for R2, we have R2=Vc2R1/VT2 COS B. Referring to Figure 8, the boundaries of the inlet, the centroid of the inlet, the axis of rotation X-X, and the split line between `the inducer and impeller are indicated. These are laid out, to any suitable scale, for the solution of the problem. The radius of the inlet centroid may be calculated by graphical methods :from the physical dimensions and lform of the impeller. The ilow is assumed to have no radial component for the iirst half inch of the path (in a particular large impeller), an assumption which introduces no significant error. Thus, the first half inch is laid out as a straight line from the centroid of the inlet parallel to the axis to the point A. At this point, R1 is known, angle B is zero, Vc may be taken from the chart (Fig. 5), and VT is derived aan state 7 from We land -the values -of U and .Q for the coordinates of point A, as indicated in .Figure .5. Aichart showing thevalues oftan 1Q in terms of X and R throughout 4the -induoer `may be 'prepared for -ready reference. Tan Q-is kunity inthe -imp'ellen ilor the specific inducer described, tan Q is propontional to the'product of X .and The nvalue ot R2 'may thus be computed, and is laid out as line .AG normal to the axis in Figure 8. With lO `as center, an arc is drawn to A. The extent of .this must be small, 'as Athe .computation `'proceeds by small increments. arc is exaggerated .in Figure v8 for clarity. An arc of the order yof 2 degrees might Vbe used. Ol vionsly, .the arc mustapproximate.thesegment of'ftheinon- `circular locus which it represents. Tlhe Acomputed 'rva'lue :of R2 eis .checked tor the interval AA by calculating R2 ffor the .midpoint P of the arc. The .radial and axiad coordinates of .and the value `ot angle B at .P are'determined in the obvious manner, yand the basic equation diz: VCZRI/:VTZ cosiB is solved for the 'valuesof Vc and 'VT for thefactual coordinates of?. This value -for radius R2, indicated by :the dotted line OP, should .closely .approach :the radius 0A. df .the discrepancy is ,greater than 'one percent, .successive approximations may be made. Thus, if OP is greater than OA, a greater value of R220/l is assumed, a Anew arc is drawn, andthe valueoilz is recomputed 4for the neu/coordinates of P. When the equation :for the midpoint of the arc is balanced within one percent, the arc AA .may be iconsidered a sufficiently exact approximation to lthe .segment ot the desired centroid. The vprocess yis repeated, computing R2=0A from the coordinates .of A land the :value of angle B, Vc, and VT at A. The arc AA is drawn with center 0 on .OA' and radius equal to the new value of R2. The'midpoint of this arc is checked and OA moditiedif necessary, thus .deter` ymining successive radii R2, R2', R2, and so on. In this Way, a close approximation to the mathematically exact centroid vis built up segment by segment, and the nal curve may appear substantially .as vindicated in Figure 7 by C-C. When this has been done, Lthe length of -the centroid C-C is measured. l-f this varies signiticantlyzfrom the -value assumed, which is .thevztbscissa of the chart (Fig. 6) -of component velocity, a new .assumption as to length .of C-C is made, .the .chart is corrected, and the centroid is recomputed. A discrepancy 'of one percent in the path ylengthis acceptable. The iinal operation in dening the zforrn ofthe ,rotor consists of determining the inner and outer :boundaries of the flow path, based upon the centroid :as determined. lf, as illustrated in Figure y9, we draw a normal .to .the .centroid at any point yP, and rotate the centroid and the normal about the axis of rotation .of the vimpeller, the centroid will generatea complex .surface .ofsrevolution and .the normal will generate a conical surface. The area 4of .the conical surface between the inner and outer boundaries of the air path through the rotor is taken as the total area AT or" .the tow path. The net area AN oit .the 4ilow path equals AT-./l, where Av is :the .portion .of AT .occupied by the vanos. The desired valueof AN for any point on the centroid is derived as follows: A curve offstatic density of the air against path length -is plotted, as illustrated in Figure 6. Static y.density at vthe `inlet Vand outlet are computed from the stated design conditions. Static density is assumed to remain constant for the tirst .half inh of the path (no compression as the air enters the spaces between the inducer blades) and then to increase .linearly .to the outlet. y The desired net area AN of the patlrat anypoint is equal to the air flow (pounds per second) dividedbythe product of static density and .component velocity. AN .is computed and plotted against ,path length, Yas .indicated .in 'Figure 6. tof .the blades in the inducer. cite calculations based upon specilic values The flocation of the inner and outer envelopes o 'the flow path is determined so `that lthe portions -o'f the ne't flow -path outside 1and inside the `centroid C--C are vvof equal area. The desired area Imay be obtained by properl-y flo'cating theseenyelopes. Since AT, AN, and Av-do not vary in the same Iratio with changes inthe depthof the air passage, and no simple relation exists l'between AT, AN, and Av, a method Aof successive approximations is used. Referring to Figure 9, the ycentroid C--C has beenfestablished. To determine the inner and outer boundaries of the l'low path, shown in broken lines, a series of points on leach boundary are found, each point corresponding to `a'point on A`the-centroid. With a number of points thus iocated, curves may be aired through these points. For any Vpoint "P, Ifor example, 'the desired =value of AN at Pis determinedfrom fthe curve (Fig. 6) for `'the distance of point TP from theinlet along the line C-C. The value of AT A'sufficient -to give the l'required AN is estimated. 'The fangle TB between the 'radius OP l'and the normal to the centroid is measured. The values of L1=PP and L2=PP corresponding Lto the assumed value of AT are determined. The areas of the conical surfaces `generated by rotation of these line segments about the axis X-X must each equal one-half AT. Therefore, from the .geometry :of yFigure 9, 1,2 .AT cos B The -total width of the passage, which `equals the depth .ofthe Iblade, 4is LVI-L2, which we ymay call L. Then AT=frL(R3i-R4f), which relation may be used ato check .the computation. Since AN=ATAW Av must be determined. Av maybe .found with sufficient accuracy `from `the formula A=NLt/sin Q where .N .is .the number of vanes, .t is Ithe .average thick- -ncss of the vaues, .and Q .is taken .as the value .at the centroid. The term -sin Q corrects for the inclination ln the impeller, sin Q equals unity. The value of AN=AT-A obtained by subtraction is checked against lthe value .of AN lorginally assumed. If the .discrepancyis greater than one percent, the assumed value of AT .is adjusted, and L1, L2, Av, and AN are re* computed until a satisfactory Yagreement is reached. A series of points .on the inner and outer boundaries -ot the blades being thus determined, .the form of therotor 'is completely established. The casing .'21 is formed 4rfor a small clearance from the rotor. The .rotormush of course, be Aanalyzed .for centrifugal vand other stresses, but this procedure is .not material to my `invention and need not be .explained lherein. The .foregoing explanation and description ,will enable those .skilled .in the art to practice the invention. It is unnecessary to an understanding of the invention to reof design parameters or to specify the resulting forms and dimensions; each compressor according to the invention must be designed for its` particular environment. It will be seen from .the foregoing that the .salient feature .of compressors accordingto my invention is ,that .the radial component .of velocity oftheair `tlow is'developed 9 or, in other words, the airis dee'ct'ed* from axial flow to iiow with a substantial radial component, by the free action of centrifugal force. This distinguishes from previous compressors in which the deflection of the air flow from substantially axial to more or less preponderantly radial is either forced by the inner boundary of the iiow path or restrained by the outer boundary. In such prior compressors, the forced deiiection or forced restraint -of deflection results in unequal pressure and velocity distribution across the flow path, reducing the total iiow, pressure ratio, and etiiciency of the compressor. ' The compressor of the invention may be distinguished from prior art compressors by the term free deflection compressor, and the principle of the invention may be called the free deflection principle. An example illustrating the benefits of `the invention may be cited. The compressor of a previously existing gas turbine engine has been redesigned in accordance with this invention. The new compressor rotor is equal in diameter to the previous rotor. Notwithstanding this, striking improvements in the performance of the engine have been realized. The installation comprises a rotor of fifteen inch radius directly coupled to a turbine. The air flow has been increased twenty-two percent, the pressure ratio has been raised from 4.4 to 4.9, and the efciency of the compressor from the former seventy-four percent to eighty percent. This improvement could not have been achieved by applying previously known principles of compressor design. As a result, the specific thrust of the engine in pounds thrust per pound of air has Abeen increased fteen percent and the fuel consumption for unit thrust has been reduced by twelve percent. It may be noted that these gains vinvolved an increase of 1e`ss than one half of one percent in compressor R. P. M. The turbine of the engine was. redesigned to lit the increased air ilow, but the improvement of the engine is not based upon increase in turbine eiciency. Since this degree of improvement was made in a compressor and an engine which had been highly developed for some years prior to the invention, the importance of the invention will be obvious. The above `description relates to a compressor in which component velocity increases linearly with path length. Obviously, other assumptions as to component velocity could be made, and the principles described are applicable to the design and construction of compressors based on any `desired relation between component velocity and path length. It will be apparent to -those skilled in the art that my invention may be practiced in a variety of forms and that the principles thereof are readily capable of wide application. The invention is not to be considered as limited by the detailed description of an illustrative embodiment thereof. I claim: 1. A centrifugal compressor comprising a vaned rotor rotatable about an axis, means deining an inlet to the rotor for entrance of air axially of the rotor, and means dening a diffuser for air discharged from the outlet of the rotor, the rotor being formed to discharge air radially relative to the rotor, the vanes of the rotor defining an air ow path from inlet to outlet the area of which varies from inlet to outlet to provide an even variation in air velocity along the path, the intercept of the mean surface of the path by a plane containing the axis of rotation being tangent to the resultant of the centrifugal force vectors due to tangential velocity of the air and to velocity of the air in the path defined by the said intercept of the mean surface at each point along the said mean surface. 2. A centrifugal compressor comprising a vaned rotor rotatable about an axis, means defining an inlet to the rotor for entrance of air axially of the rotor, and means defining a diluser for air discharged 'froml the outlet o'f the rotor, the rotor being formed to discharge air radially relative to the rotor, the vanes of the rotor dening an air flow path from inlet to outlet the intercept of the mean surface of which by a plane containing the axis of rotation is tangent to the resultant of the centrifugal force vectors due to tangential velocity of the air and to velocity of the air in the path defined by the said intercept of the mean surface at each point along the said mean surface. 3. A centrifugal compressor of the axialto-radialflow type comprising, in combination, means defining an annular inlet and a vaned rotor comprising an inducer portion adapted to accelerate a gas stream circumferentially relative to the rotor axis `and an impeller portion adapted to accelerate the gas stream radially and tangentially, the gas iiow path between the rotor vanes varying from inlet to outlet of the rotor so as to provide substantially linear variation in the resultant of the axial and radialfco-mponents of gas velocity, the inner and outer boundaries of the gas iiow path through the rotor being so disposed that the centroid of the path substantially coincides with the free deflection path of gas ow through the rotor under design conditions. 4. A centrifugal compressor comprising a vaned rotor with an inducer portion and an impeller portion7 the mean ow path through the rotor conforming substantially to that defined, for design conditions of operation, by the following procedure: Assumption of substantially linear increase through the rotor of the component of velocity of the air in a plane containing the rotor axis; and derivation of a ilow path originating at the centroid of the inlet and corresponding to the free path of flow of a particle under the effect of centrifugal force and with the component of velocity varying as stated. 5. A centrifugal compressor comprising a vaned rotor with an inducer portion and an impeller portion, the boundaries of the ilow path through the rotor conform ing substantially to those defined, for design conditions of operation, by the following procedure: assumption of substantially linear increase through the rotor of the component of velocity of the air in a plane containing the rotor axis; derivation of a centroid of the flow path originating at the centroid of the inlet and corresponding to the free path of flow of a particle under the effect of centrifugal force and with the component of velocity varying as stated; assumption of substantially linear increase through the rotor of air density; determination of the required net area of the path as a function of path length for component velocity and density varying as stated; and location of the inner and outer boundaries of the flow path so that the path is divided into two equal parts by the centroid and the net area of the path conforms to that determined as stated above. 6. A centrifugal compressor of thef."azrial-tofradiahiiow type comprising, in combination, means defining an annular inlet and a vaned rotor comprising an inducer portion, with vanes decreasing progressively in angular relation to the axis of the rotor from a maximum angle at the inlet to substantially a zero angie at. the discharge end of the inducer, and an impeller portion with vanes substantially at a zero angle to the axis, the axial length of the inducer being such that the tangential acceleration of gas flowing through the inducer accords with the tangential acceleration of the gas in the portion of the impeller adjacent the inducer. 7. A centrifugal compressor of the axial-to-radialiiow type comprising, in combination, means defining an annular inlet and a Vaned rotor comprising an inducer portion, with vanes decreasing progressively in angular relation to the axis of the rotor 'from a maximum angle at the inlet to substantially a zero angle at the discharge end of the inducer, the vanes being curved to provide substantially uniform tangential acceleration of gas in the inducer, and an impeller portion with vanes substantially 1'1 yat aizero angle `to s-the axis, the axial lengthef the Yinducer being such that :the tangential acceleration of gas Etion/ing :through the inducer raccords with the vtangential acceleration Lof 'the .gas in fthe-portion of the impeller adjacent the inducer. `8. A compressor comprising, in combination, `a rotor rotatable about an axis, .means defining an annular inlet for directinggas into the rotor substantially parallel -to .thelaxis, and a `diuserfor receiving lgas fdischargedifrorn the rotor, the rotor comprising a body in the form lof .-a 'body yof revolution about the laxis with vanos extending from the surface tof vlthe -body Asubstantially normally to ythe surface, the spaces bet-Ween adjacent Vanes constituting ygas flow paths from the `inlet y'to the diffuser trending Asmoothly .from a direction :substantially parallel ,to the la-Xisat the inlet ends thereof to Va'direction approximately normal to the :axis at lthe discharge ends thereof, the depth 'of the 4Yanes r:varying from inlet to youtlet so that lthe area of each fpath .varies `'as =a @function of Vdistance along .the path `accordiugito the relation .that the .area .-is the quotient y,of a constant by ithe :productloftwo quantities each vary ing substantially linearly with distancetal'ongtthe path, the said quantities being component velocity yof gas `in the path and static density of gas in the path. V9. A compressor comprising, 1in combination, a rotor rotatable about an axis, tmeans defining an annular inlet for directing .gas into Vthe rotor substantially parallel -to .the axis, and a diiuser .for receiving gas discharged `from the rotor, the rotor `comprising .a Abody 4in the .'forrn of a 'body of revolution 'about the axis 4.with vanes extending from vthe :surface of the body substantially normally Vto the surface, the spaces between nadjacent vanes constituting gas flow paths -from the inlet to tthe diffuser trending smoothly from a direction substantially :parallel :to the axis at the inlet ends thereof toa directionapproximately normal to the 'axis at -the discharge ends `thereof, Vthe form of the surface ofthe body `being such and the depth of the vanes varying from :inlet to outlet so as :to :define a path the locus of 1the geometrical center/of which is tan- `l2 gent at any iPOlKlt along :the ,path to tthe resultant of the centrifugal `force `Vectors at that point due v.to .tangential ne locity of the gas about the said anis Yand to ,yeloeitytofth gas yinthe saidfpath. L0. yA -Xcompressor as recited in claim 9 in ywhich the area Yet the path varies ,with Idistance :along 4the `path ,so that .the velocity fof the gas .in .the path varies linearly with distancealong the path. l-l.. lA compressor compri-sing, 4in combination, a ,rotor rotatable about .an axis, means .defining ,an annular inlet for directing gas into the rotor .substantially parallel to the axis, and a diffuser for receiving gas discharged `from the rotor, the rotor `rcomprising a .body in .the form `of a body .ofrevolution I.about the axis with yanes extending from the surface ofthe `bodysubstantially normally @to ,the surface, the spaces vbetween .adjacent Names constituting gas 4How paths from l,the inlet tto .the `diffuser trending smoothly from .a .direction ,substantially parallel to the axis at the .inlet ends thereofto .aldireetionapproximately normal tothevaxis at ,the `discharge ends thereof, theorm of the surface lof ythe body lbeing such and the depth of the yYanes varying ,from inlet :to outlet .so .as .to define a path :the locus ofthe geometnicalfcenter `of whichlis the free deflection path foriflow ofgas :from the inlet to the outlet, References Cited 4in vthe lfile of this ypatent UNITED STIATES k,BIATENTS A1,097,729 Rice May 26, 1914 2,228,194 Birkigt Jan. 7, v1941 '2,398,203 Browne Apr. 9, 1946 2,899,852 -Campbellfet al May 57, 1946 FOREIGN vI JATENTS 279,426 :Great Britain ;Aug. 9, 51928 l460,985 'Germany June 8, 1928 Sil-2,098 AGreat Britain .of 1939 6119,617 l"Great Britain kof 1:949 Patent Citations
Referenced by
Classifications
Rotate |