NACA TN ap 988 “9095900 ACO) NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS YN “aver AUVEEIT HOaL TECHNICAL NOTE 2455 A METHOD OF DESIGNING TURBOMACHINE BLADES WITH A DESIRABLE THICKNESS DISTRIBUTION FOR COMPRESSIBLE FLOW ALONG AN ARBITRARY STREAM FILAMENT OF REVOLUTION By Chung-Hua Wu and Curtis A. Brown Lewis Flight Propulsion Laboratory Cleveland, Ohio Washington September 1951 LDRST? at AOUseze ‘TECH LIBRARY KAFB, Nit AV EL NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS eee TECHNICAL NOTE 2455 A METHOD OF DESIGNING TURBOMACHINE BLADES WITH A DESIRABLE ‘THICKNESS DISTRIBUTION FOR COMPRESSIBLE FLOW ALONG AN ARBITRARY STREAM FILAMENT OF REVOLUELON By Chung-Hua Wu and Curtis A. Brown SUMMARY A vapid method for designing turbomachine blades of a given turning and a desirable blade-thickness distribution for a compressible non- viscous fluid flow along an arbitrary stream filament of revolution is presented. The method utilizes the guiding effects of the blade shape on the mean streamline shape and of the blade thickness on the specific mass flow along the mean streamline. After the flow on the mean strean- ine is determined, the extension of the solution from the mean strean- Line to the blade surfaces is accomplished by the use of a power series. A mumber of blade profiles are obtainable for the total mass-flow requirement, and one is chosen for the best velocity distribution on ‘the blade. “The results obtained in the solution can be used for a direct check on the accuracy of series approximation and, also, for ‘the more accurate determination of the velocity distribution along the leading end trailing edges of the blade. ‘The method is illustrated with the design of several turbine cas- cades of highly cambered thick bledes. The determination of the shape of the blades and the compressible flow past the blades was carried out by hand computation in 16 hours. One solution obtained by using threq, terms in the power series compared very well with an available direct solution and the blade circuletion checks closely the specified turning angle. Because the surface of revolution, on which the blades are located, is completely arbitrary, the method can be applied to axial-flow, rediel-flov, and mixed-flow turbomachines. The variation in the normal distance between the stream surfaces of revolution can be taken into account, thus incorporating into the design the principal effect of ‘three-dimensional flow. ‘he method is readily applied to the design of channels on a plane and on a general surface of revolution.2 NACA "mH 2455, ‘IFRODUCTION The increasing use of compressors and turbines in aircraft pover plants during the past 10 years has led to considerable research in the direct and inverse problems of two-dimensional potential flow past an infinite cascade of airfoils. In the inverse problen, the design of blades is often directed at control of the pressure or velocity distri ‘ution on the blade. Most of the methods ere derived for exial-type turbomachines, in which the flow is assumed to take place on cylindrical surfaces (methods for incompressible end compressible flow are discussed jm references 1 to 7 and 8 to 11, respectively). Methods for designing Dlades in a radial plane are given in references 12 and 10. In current axial- and radial-flow turbonachines, the flow surfaces are uouslly of a more general shape than either cylindrical or radial. Furthermore, the normal distance between adjacent flow eurfaces varies along the flow path. A method vas therefore developed at the Lewis laboratory for the design of blades for compressible flow along an exbitrary stream filament of revolution. Instead of the velocity distribution on the blade being the required result, the blade design is aimed at a desirable blade-thickness distribution required with respect to blade strength and Mach number in general and also the cool~ ant passage requirenent in the case of cooled turbine blades. The con- putation involved in this method is relatively simple and short, end the usual assumption of a linear pressure-volume relation for compres- sible flow is not required. For clarity and simplicity the method will first be given for compressible flow on a plane or a cylindrical surface and will be illustrated by a few examples. ‘The method will then be given for the general case of compressible flow along an arbitrary stream filament of revolution. SYMBOLS The following symbols are used in this report: B differpntiation coefficient HK total enthalpy, + dv" ae Hy relative total enthalpy, h+ 3 W e static enthalpy szee2223 NAGA ‘BI 2455, To negw - Fart L dade tength projected on turbomchine axis 1,@ orthogonal. coordinates on mean surface of revolution M mass flow P pitch or spacing P static pressure of gas r -radial distance from exis of turbonachine 5 streanline + Dade thickness in circumferential direction V absolute velocity of gas W velocity of gas relative to blade ¥ Gistance in direction of pitch for plane flow and equal to rp for flow on cylindrical surface z —@ietance along exis of turbomachine $n ange ox ctcen ett, tart HE or tart BP p ° density of ges o angle between tangent to meridional curve and axis, tan”) z 7 somal thickness of stream filanent of revolution Y ratio of specific heats V stream function © angular velocity of blade Superscript: * dimensionless value4 NACA TH 2455, Subscripts: e exit 4 inlet 1, meridional and circumferential components m mean streamline P pressure surface of blade 5 suction surface of blade T total or stagnation state ¥-y-component z -z-component ‘DESIGNING BLADES FOR COMPRESSIBLE FLOW IN PLANE OR ON CYLINDRICAL SURFACE General. Description of Method In a recent investigation of compressible flow through a typical cascade of turbine blades (reference 15), the following resulte vere obtained: (2) The shape of the mean streamline follows approximately thet of the mean channel Line of the cascade and bas a lower curvature. (2) The variation of the ratio of the specific mass flow on the mean streamline to its inlet value follows the trend in the variation of the ratio of pitch to channel width (inside the channel the ratio of ‘the specific mass flow is about 4 percent greater than the ratio of pitch to channel width). (3) ‘The variation in fluid properties across the channel can be represented by a second-degree function for engineering accuracy. These results were used herein to develop a rapid method for designing cascade blades for either compressible or incompressible flow. ‘This method of blade design starts with the calculation of the com- pressible flow on the mean streamline. With the flow angle upstream 2223see NACA ‘BY 2455 5 and downstream of the blade determined by the velocity diagram and the pitch and the axial length of the blade given, a particular mean stream- line shape may be sketched by the designer. The designer may use the shape of the mean streamline itself as a parameter of the cascede or may specify the mean streamline shape according to figure 17 or refer- ence 15 (or even better according to any available relation of a eini- ler blading) to leed to a certain blade camber line. (If only a mmber of points of this streamline are specified, it is important that the values given are numerically suooth). In addition to the mean streamline shape the designer further specifies at a finite mmber of points along the mean streamline (such as 7... %, in fig. 1), the ratio of specific mass flow on the mean streamline to inlet value. The values of the specific mass flow on the mean streamline are determined by the blade-thickness distribution which is desireble from the consideration of blade stress and Mach number in general, and the consideration of ‘the additional requirement of coolant passage in the case of cooled turbine blades, and by a relation between blade-thickness distribution and specific mass flow on the mean streamline, such as shown in fig- ure 19 of reference 15. With these specified values, the velocity components and the density are very easily determined at the specified points on the mean stream- line. ‘The variation of velocity components and density in the pitch Girection are then obtained by using power series in that direction. ‘The derivatives in the series are determined from the fluid state on ‘the mean streamline by the use of equations of continuity and motion and the density-velocity relation for isentropic flow. A number of blade profiles and their velocity distributions are obtained by inter- preting the starting mean streamline as dividing the inlet mass flow into two slightly different amounts in the channel. ‘The velocity dis- tributions on the blades are compared and the best one is chosen. If the blade shape, the thickness distribution, or the velocity distri- ‘bution around the blade obtained requires some modification, the values apecified for the mean streamline should be modified and the process repeated. Because of the relatively short computation involved {in the illustrative exemples of turbine blades, only 16 hours were required for the compressible solution by using the first three terms in the series and 6 to 10 stations inside the channel), modifications of the solution for more desirable blade thickness or velocity distri- ‘bution is practical. Families of blade elements can be built up very quickly this way for any particular application. Basic Relations The steady two-dimensional isentropic flow of a nonviscous fluid in a plane or on a cylindrical surface is governed by the following equations of continuity and irrotational motion, and the isentropic pressure-density relation:6 NACA 1H 2455 CoH) atty) Se @ di, Bi we (2) p= Kp" (3) In these equations, the z coordinate is chosen along the machine axis and the y coordinate 4s chosen along the pitch direction (y= rg in the case of cylindrical flow with r equal to @ constant). Consider first the gas flow along a streamline somevhere in the midpart of the channel formed by two neighboring blades, such as ab in figure 1. he coordinates of the streamline and their differentials are related, respectively, by the following two equations: S(z,y) = 0 (4) Baz + Bay -0 (s) When the variation of the fluid state along the streamline is considerea, any quantity q on the streamline is a function of z only, that is, a=azy(2)] (6) ‘The total derivative of q with respect to = is aq _2a, Say G82 oy a : (1 But ¥%, x (8) Hence equation (7) may be written as $= Bs tan p 3 (9) When equation (9) is used, the continuity and irrotetionality relations can be written as e222szze NACA TH 2455 7 (20) and (a) ‘The variation of density throughout the flow region can be most conveniently expressed in terms of its inlet value through the use of equation (3) as foliows: 2yrr (#8) 0 A tabulated general relation of p/oy , im equal intervals of w/a, can first be calculated, from which either a table for p/e, im equal intervals of W* or a graph can be.easily constructed for Zach indi. vidual case and used for the evaluation of density from the velocity. Pra pt= fh. Bip OB Bye Along the chosen (mean) streamline, where the slope is known and el, 4s given at a number of stations, the density at these stations can be obtained as follows: Rewrite equation (12) es (12a) or (12) vherea NACA TN 2455, Either equation (12a) is used to prepare a table of p* for equal inter- vale of (pfW,* sec f) by an iterative process, or equation (12) te used to compute tw," * sec B for a number of values p*, which are then plotted as a graph. After densities have been obtained from either a table or graph, the velocity components on the chosen (mean) strean- Line are readily computed. ‘The fluid state is then extended out in the pitch direction by the equations given in the next section (compare vith references 1¢ and 15). Variation of Fluid State in Pitch Direction Equations (8), (10), (11), ana (12) directly give the first-order partial derivatives of W,, Wy, and p with respect to y in terms of the known quantities on the chosen (mean) streamline es follows: OM, a, (Wz) 2 we. [Bes can pd | cos” B (a3) (a4) aK, w, Gs) Me second-order partial derivatives of W,, Wy, and p vith respect to y can be obtained as follows: Differentiating the con- {anuity equstion (1) with respect to y results in 2, 2, a (ote a FH) g a Equation (16) can be weitten through the use of relation (9) as 2, 2 2c) 2 (cig) a 2CoK) s—S +d (a7) SE - ton SE oF oF Which 1s expanded to obtain iy ap Bly (Mz ap 3) sa aap s : a(it,) oe +28 oe - Pat 8H, a. ° ay * (28) 2223eee NACA "TH 2455, 9 From equation (10) au, ay, a(oWt,) 'y oo a ea Differentiating the izrotationality equation (2) with respect to y and using equation (9) yield (a9) (20) Substituting equations (19) and (20) into equation (18) results in 2, 2, Sty 2 A(oWy) 9p (a Wy ati a (eH) Brae Fold ay - ten Boe) Ota ay Transposing and conbining terms give atu a(ou,) on, an, y 2 zo 1a Zz Eat 4 2 ~[gSS2 2a (Ree) seed Sts (a) After this equation is evaluated, the second partial derivative of W, with respect to y is obtained ty using equation (20) Siw a oe a ae mt (22) ‘A typical computation for these derivatives is presented in tables I and IL. The second-order partial derivative of p with respect to y is obtained again from equation (12): to ey “Ga (=) 1 Sten oe ( (ay z (23) ‘Tira and higher order y-derivatives, if required, can be obtained ina similar manner. The complete variation of any fluid property q10 NACA ‘BY 2455 across the channel can then be expressed by a Taylor's series in (y-¥q) fron the various derivatives at the given stremline, such as the mean streamline a) = 9g + (ory) (32) + oe 4) (2a) Determination of Blade Profile The blade profile can be obtained by a consideration of mass flow. At the chosen stations, mass flow across a constant z line from to y is computed as a function of y according to the following equa- ‘thon: y M= | pil, ay (25) The variations of masa flow M and the magnitude of resultant velocity W at each station are plotted ageinst y (fig. 2). Because the con- Gition on the suction surface is more critical than that on the pressure surface, the blade shape on the suction eide is determined first. From the plot of mss flow against y, a mumber of y,'s are chosen for a number of masa flows in the neighborhood of one-half the inlet masa flow, thus obteining a number of suction surfaces.’ The corresponding Yelocities on the suction surfaces are read from the velocity plots (fig. 2). The one with the best velocity @istribution is then chosen. After the suction surface is selected, the pressure surface and ite velocity are determined ina similar manner by the total mass-flow requirenent. If the shape or thickness of the blede or the velocity Gistribution on the blade obtained is not quite the one desired, the shape of and the flow on the mean streamline can be modified accordingly and the process repeated. Because each case takes a relatively small anount of computation, this modification is practical. Systematic ‘puilding up of families of blades for various applications is also not aisficult. he accuracy of the blede coordinates obtained depends mainly on the accuracy of series representation and the accuracy to which the partial derivations are evaluated. For high-solidity blades, such as ‘those investigated in reference 15, the first three terms in the series 2223see NACA ‘BY 2455, nu will give sufficient accuracy. For low-solidity blades, more terms may ve required. It may be noted thet accurate representation of the flow variation by the series is difficult to achieve in the neighborhood very close to the leading and trailing edges; but from a practical point of view it is satisfactory to fair in the nose and the teil according to sone staniera shape after the biede coordinates ere obtained up to a short distance evay from these regions. ‘The application of this design method will be greatly aided by the availability of detailed flow variations in typical bledings such es ‘those given in references 15 and 16. If e typical solution of the type of blading to be designed is not available, either a direct problem may be solved first, or even better, the accuracy of the inverse solution can be ascertained in the manner described in the following section. Method of Checking Solution The inverse solution obtained by this method can be very conven- jently checked and improved, if necessary, by the relaxation method utilizing the fluid state obtained in the solution. Inasmuch as the velocities are available in the solution, the equation for irrotatdonal absolute flow (equation (A9) of reference 15) is now written as The finite-difference form of equation (26) is then aa a8 a Bevis D> Beye | (wy) DY By pd - CH, Pay ae bog as where the seme notation used in reference 13 is employed. A grid system ie obtained by reteining the same z-stations used in the inverse solution and dividing the pitch distance into an appropriate number of Givisions. ‘The values of W, W, and p are most conveniently obtained by reading off the plots of integrated mass flow M, W, and p at each of the z-stations. The differentiation coefficients B's for equally spaced grid points as given in reference 17 can be applied throughout the domain for the present purpose by using the function values which are inside the blade but at equal spacing from the points in the channel (the first and lest rows in tables III and Iv). If the residuals obtained according to equation (27) require negligible change in W, the solution is entirely satisfactory. If the residuals are large enough to necessitate one cycle of relaxation, the net effect o (28) (en)2 NACA TH 2455 may be a slight change in the specified mean streamline flow and in the velocity distribution on the resultant blade. If the residuals are so large as to warrant a few cycles of relaxation, the flow variation for this type of blading is established, which makes the design of other similar bladings very simple. In general, for problems in which some knowledge of the flow is available, no relaxation should be necessary except, perhaps, near the nose and tail when accurate detailed velocity distribution in these regions is desired. Special Case of Incompressible Flow When the density is constent, the chamel width ratio P/(P-t) has @ relation to Wz,m similar to the relation it hes to (pWz)q in the compressible case (reference 13). A number of values of W, are there- fore prescribed at a mumber of chosen stations along the mean streamline to lead to a desirable thickness distribution of the blade. ‘The solu- tion of the incompressible problem continues in generally the sane mamer as it did for the compressible case with considerable simplifi- cation in the series terms and the integration process. In the incom Pressible case, of course, equations (12), (15), and (25) relating p and the velocities are umecessary. The firet- and second-order derivatives expressed by equations (13) (14), (21), and (22) for the compressible case are simplified to: Se (i 4 can p Bt) con? 9 (23) Be font) aire ena XG) ml) for the incompressible case. the equation for obtaining mass flow y vecomes M=p | W, dy. ‘m 2223zee NACA ‘EY 2455 3 ‘Tiustrative Examples ‘The procedure outlined in the preceding sections has been applied to the design of several highly cambered thick turbine blades for either compressible or incompressible flow. Firet and second examples. - In the firet exemple, the shepe of the nean streamline and the variation of axial velocity obtained in the incompressible solution of the blade given in reference 15 are taken as the specified values in order to determine vhether the original. lade will be reproduced. The shape of the mean streamline is shown in figure 1 and the thickness distribution of the blade is shovn in fig- ure 3. As a further check of the method, the z-stations chosen in this calculation correspond to every other z-station used in reference 15, so that the velocities obtained long these stations can also be com pared with the solutions obtained in reference 15. The given quantities vere inlet angle f;, 41° 18"; exit angle f,, -52° 57"; axial chord L, 1.5 inches; and pitch P, 1.017 inches. The same data are used in the second example in which a compres- sible solution for an inlet Mach number of 0.42 is obtained. The axial- velocity variation on the mean streamline used in the first example is now taken ac the specific-maas-flow variation on the mean streamline. ‘This exemple is presented mainly to illustrate the difference in incom- pressible and compressible solutions for the same mean streamline shape and the seme ratio of specific mass flow. The complete computation for the flow on the mean streamline and the determination of the first- and second-order derivatives of W,, W,, end p in the pitch direction at the mean streamline for these v two cases is given in tables I and I. Only three terms in the series are used because the direct solutions given in reference 15 indicate ‘that they will be sufficient for engineering accuracy. The central- point fourth-degree differentiation forma is-used at the regular stations 2, to %¢. Because the first and last stations inside the channel. employed in reference 15 are not close enough to the leading and trailing edges, respectively, two extra points are computed at stations 6.75 end 12.25 by the use of the unequal interval differen- tiation formula given in reference 18. ‘hese two points are so desig- nated because they are located at a distance of a quarter of the regular spacing avay from stations 7 end 12, respectively. This combination is unnecessary for other cases. The velocities and the densities used in the calculation are non- @imensionalized as follows:s2ee uw NAGA "I 2455 The mass flow UY, being divided by pi, 5, has the Aimension of length: 7 M ee a 2 For the compressible case, the ratio of H, to W,,; is equal to 25.78, which is the value used in the construction of the two density graphs involved in the calculation, the reduced versions of which are shown in figures 4 and 5. The values in columns 2, 5, 8, 9, 12, and 15 of table I and 7, 8, 15, 14, 22, and 25 of table II are used. to compute the variation of VW, and W, in the pitch direction by the power series. The calculation 7. of density follows directly as does the integration across the channel for mass flow, the mass flow being determined numerically. Because the specific mass flow pil, was made dimensionless with its inlet value, ‘the numerical value for the mess flow is equal to the pitch, the height eing considered unity. This indicates that the integration for mass flow along the pitch direction starting at the mean streamline was con- sidered to be complete in either the plus or minus direction when a value equal to one-half the pitch distance or one-half the total mass flow was reached. Thus the channel flow and the blade coordinates were obtained, This calculation at one station, 10, is shown in tables ITI and IV for the two cases, respectively. The blede profile obteined by interpolating y, and yp for one-half of the inlet mass-flow value is shown in figure 6. Because of the sane mean streamline shape and the sane variation of specific mass flow on the mean streamline pregcribed in the incompressible and com- preseible cases, the blades obtained for these tvo cases look quite similar except that the compressible one is somewhat thicker and the suction surface of the compresaible solution is situated farther from ‘the given mean streamline; these results are consistent with the direct solutions given in reference 15. The velocities obtained in the two cases (fig. 7), hovever, are quite different. Those in the compressiblesee ‘MAGA ‘TW 2455 15 solution are, in general, higher than those in the incompressible case, principally because of the high velocity on the mean streanline result- ing from the use of the sane (plJ,), anda decreasing p along the mean streanline in the compressible case. The blade coordinates obtained in the incompressible solution are compared with those of the original blade in figure 8. The velocities at the six regular stations in the channel are compared (figs. 9 to 11) with the values obtained in the relaxation solution of the original blade reported in reference 15. These four figures show that, in general, the present solution ie satisfactory. ‘he relatively large differences near the leading and trailing edges are partly due to the inaccuracy in the second-degree polynomial approximation in the present calculation and partly due to the inaccuracy of the mnerical solution obtained in reference 13 caused by the relatively coarse grid used in these regions. As a check of the consistency of the solution, an integration of the velocity along the blade profile obtained in the sane example was made and compared with the circulation value computed from the inlet and exit tangential velocities and the pitch. ‘he two are in agreement within 1 percent. Third and fourth exemples. - In the previous two examples, the mean streamlines and specific-mass-flow distributions prescribed were not entirely arbitrary, having been obtained from the results of a direct problem for incompressible flow. ‘he possibility of obtaining an unrealistic blade shape was, for this reason, largely eliminated. Con- sequently, in order to give the method a still more rigorous test, the thickness distribution (fig. 12) end mean blade line (fig. 15) for the bub section of an experimental cooled turbine blade were arbitrarily chosen from which a mean streamline and specific miss flow were estima- ted by means discussed in the section "General Description of Method." In addition to the mean blade line end thickness distribution, the following data were used: inlet Mach mmber, 0.42; inlet angle, 36.2°; exit angle, 42.7°; axial chord, 1.8 inches; and pitch, 1.176 inches. The mean blade line vas faired into the inlet and exit directions (fig. 14) with modification to obtain a mean streamline according to the information obteined for a typical turbine blade in reference 13, (see fig. 8), The thickness distribution vas, used to obtain the specific-mass-flow distribution along the mean streamline (dashed curve in fig, 18) according to results obtained in reference 13. After the flow on the mean streamline was calculated, it was then extended out from the mean streamline across the channel by means of power series as vefore. Both compressible and incompressible solutions were completed, the incompressible case being considered first because of its relative16 NAGA TN 2455 simplicity. ‘The results of the incompressible case served as a guide to obtaining better prescribed velues for the compressible case. the blade obtained in the incompressible solution has a slightly higher thickness @istribution than vas wanted and, because results obtained in the first two examples indicate that the compressible blade vill be thicker than the incompressible for the sane specific-mass-flow distribution, this Gistribution Was depressed by a Linear proportioning (fig. 15) in order to obtain better starting values for the compressible case and conse- quently e thinner blade (fig. 14). he same uean streamline mmerically snoothea to give small fourth aifferences wes used. in both cases. fis @ check of the accuracy of the lest solution, a grid having the seme spacing in the 2-Girection used in the inverse solution (0.18) and a grid spacing of 0.147 in the y-direction 1s chosen, and the residuals at the grid points are computed according to equation (27) using the central point second-degree differentiation formula. Aa shown in table V, the residuals are rather smell when they are compared with the magnitude of the coefficient at the points (-154.3). As en indication of the percentage error in the velocity, these residuals are first divided by -154.5, resulting in en approximate change in the y value at each point. ‘Then the probable error in a¥/oy or pl, is calcula- ted. ‘The result is shown in table VI, which indicates thet the solution obtained is sufficiently accurate for ordinary purposes. DESIGNING BLADES FOR COMPRESSIBLE FLOW ALONG ARBITRARY SUREAM FILAMENT OF REVOLUTLON Basic Relations ‘The blede design method presented in the section "Designing Blades for Compressible Flow in Plane or on Cylinfrical Surface" can be very easily extended to the more general case of flow along an arbitrary stream filament of revolution having a varying normal thickness (4g. 16). When only en average value in the strean filenent of xevolu- tion, as represented ty the flow on the mean strean surface of revolu- ‘tion described by @ set of orthogonal coordinates 1 and © (fig. 16), ip considered, the equation of continnity for steady relative flow end ‘the equation of rotational absolute flow are given in reference 15 as follows: altom,Z) | B¢TeM) z+ GQ) 1, Wosine - Zap tp tw sino =o (2") 2225eee NACA TH 2455 aq where primes in equation mumbers indicate equations similar to those of ‘the cylindrical case. When the fluid flow along a streamline 9(2,®) = 0 on the mean surface of revolution is considered, it is convenient to write any quantity q on S as afunction of 1 only. Then the total derive- tive of q with respect to 1, following the streamline, is a da a, 341% dq, ten B da : oh 3+ 3 P- s 2 ee (") hen wy, woe Berg When the preceding relation is used, the continuity and irrotationality conditions can be written as (tow, r) o Gah) A(T eM) Te - ten B ee (aor) ena at, oH, 3a mee bat Meee (2 +m) ein oxo qa) ‘The variation of density throughout the flow region, in general, ie obteined from the velocity by using the following equation: taketh? Lat wae (st ae) Ge) Tn order to obtain the density on the sterting (mean) streamline, equa- tion (12') 4s written in the following form (compare with equation (7a) of reference 13): (128') whereaa ACA ow 2455 vl a 1922 +e, t+ Bat @ = (oti * see 0? (Be Vee ) Once the general relation between Z and © 4s available, the evalua- tion of p* for certain given values of p*W," sec B along the chosen (mean) streamline is made simple if auxiliary tables or graphs giving 2 eae Ao a g (aaa) (ated) Sunn Sone and { =-2— ]\————__ as functions of r ay ot ay are first obtained. ‘The determination of the flow along the starting (mean) streamline (tig. 17) proceeds very mich the sane as in the case of plane flow or flow on @ cylindrical surface. ‘The shape of the streamline gives sec B. The variation of pMi,* is obtained from the blade-thickness variation along the mean surface of revolution as follows: If the value of oH on the mean streamline represents its average value in ‘the circumferential direction T(E) (Pt) = (TONDP) 4 (32) wet Pas EF (38) TOW Im F_ oP wana, i Although there is always some deviation from this simple relation, especially around the leading and trailing edges, a relation between the two terms in equation (54) similar to that between (ptW,*), and 2228NACA ‘BY 2455, 19 P/(P:t), im the previous cylindrical case can be expected. ‘Then, from - THoMt,*), 2% (p%i,*),, ie calculated end combined with sec ® from which 9," is obtained by equation (12e'). After p," is determined, igm* ae easily calculated. Variation of Fluid State in Pitch Direction The first-order partial derivatives of W,, Wy and p with respect to @ are readily obtained from equations (10'), (11'), and (12") as: . s2ze tan p &(70H,7) OW, aM, Mr [Oe sane Ome), (iy + 2) sin a]enes (as) a ai, al 2 Be {sans [BOs (ig a ata J] BO Jats Gat) 1p 1 ( ao *) 2300 _ set iyse) 0s) 039°” Gay @ ad ee Ew ae * "eae Ina manner similar to that of the case of plane or cylindrical flow ‘the second-order partial derivatives of W,, Wy» and p are obtained as follows: 3 2d Aomyr) oa a ai 32) eee eae *% 3p) |* oy, w. tan p («2 So + ata “H)\ae e (2) my a, a Sgt Farag to 8 Ses om o SB (22) 13% #z(3e) - Page ge 8920 NACA IN 2455 Atter the variation of fluid state in the circunferential direction is determined, the mass flow across a constant 1 line from 9 to is computed by 7 ‘? uert| pW, ap (28) Fa, from which the blade coordinates are determined as in the previous case. If checking is desired, the following equations are to be used: ay a ay 1 o& ( do a By. (see. 2a) (Ge Be ae an elu (26") sy fat, (emo dint\tailys \y BE nc [ish (22 - ety niles > ae 0 oF) an ‘he Sai of (3) So ek os ate om rt] 0 (a7) where the same notation used in reference 15 is employed. SUMMARY OF RESULTS A rapid method for designing turbamachine blades of a given turning and a desirable blade-thickness distribution for a compressible non- viscous fluid flow along an arbitrary stream filament of revolution is Bresented. The method utilizes the guiding effects of the blade shape on the mean streamline shape and of the blade thickness on the specific mass flow along the meen streamline. After the flow on the mean stream- Line is determined, the extension of the solution from the mean strean- line to the blade surfaces is accomplished by the use of a power series. A number of blade profiles are obtainable for the total mass-flow requirenent, and one is chosen for the best velocity distribution on the Dlade. ‘he results obtained in the solution can be used for a direct check on the accuracy of series approximation and, also, for the more accurate determination of the velocity distribution along the leading and trailing edges of the blade. 2225wo ACA TN 2455, a1 The method is illustrated with the design of several cascades of airfoils. In each case either some indication of the accuracy of the solution is given or the relation between the incompressible and com- pressible problems are shown. In the first case, the results obtained in an incompressible direct solution vere used as prescribed values for the inverse problem to show that the original blade can be relatively easily reproduced with adequate accuracy. The varistion of the velocity and its components across the channel vas shown to compare favorably with those obtained in the direct solution and an integration around the blade for the circulation checked with that obtained from inlet and exit values within 1 percent. ‘In the second problem, the same mean streamline and specific-mss—flow distribution along it were used; but this time the density was alloved to vary as in compressible flow and a comparison was made between the blades obtained in the first two cases. In the third and last problem, an arbitrary mean streamline and specific mass-flow distribution were chosen, the blade obtained, and the solution checked by calculation of the residuals as in a direct relaxation solu- tion, Because the residuals were small the solution was considered to de reasonably accurate. A computation designed to give some indication of the accuracy of the velocities by dividing the derivative of the error in the stream function Y by the derivative of W itself showed them to be accurate generally within 1 percent except near the blade Doundaries vhere an end-point forma ves necessarily used to obtain derivatives. Because the surface of revolution, on which the blades are located, is completely arbitrary, the method can be applied to axial-flow, radial~ flow, and mixed-flow turbomachines. The variation in the normal, distance between the stream surfaces of revolution can be taken into account, thus incorporating into the design the principal effect of ‘tthree-dimensional flow. The method is readily applied to the design of channels on a plane and on a general surface of revolution. Lewis Flight Propulsion Laboratory, National Advisory Committee for Aeronautics, Cleveland, Ohio, May 31, 1951.22 NAGA 1B 2455, ‘REFERENCES 1, Weinig, F.: Die Strammg um die Schaufeln von Turbomaschinen. Jchann Anbrosius Barth (Leipzig), 1935. 2, Ackeret, J.: The Design of Closely Spaced Blade Grids. Schweiz. Bauzeitung, vol. 120, no. 9, Aug. 29, 1942, pp. 103-108. 5, Mutterperl, William: A Solution of the Direct and Inverse Potential Problems for Arbitrary Cascades of Airfoils. NACA ARR LéK2zb, 1944. 4. Lighthill, M. J.: A Mathematical Method of Cascade Design. R.& M. No, #104, British A.R.C., June 1945, 5, Diesendruck, Leo: Iterative Interference Methods in the Design of ‘Thin Cascade Blades. MAGA TM 1254, 1947. 6. Goldstein, Arthur W., and Jerison, Meyer: Isolated and Cascade Air- fotls with Presorited Velocity Distritution. NACA Rep. 869, 1947. (Formerly WACA TN 1308.) 7, Hansen, Arthur G,, and Yohner, Peggy L.: A Numerical Procedure for Designing Cascade Blades with Prescribed Velocity Distributions in Incompressible Potential Flow. NACA TN 2101, 1950. 8. Spurr, Robert A., end Allen, H. Julien: A Theory of Unstaggered Air~ foil Cascades in Compressible Flow. NACA Rep. 888, 1947. (For- merly NACA FM A729, 1947.) 9. Alpert, Sumer: Design Method for Two-Dimensional Channels for Com- pressible Flow with Application to High-Solidity Cascades. NACA TH 1931, 1949. 10. Tin, C, C.: On the Subsonic Flow through Circular and Straight Lattices of Airfoils, Jour. Math. and Phys., vol. XXVIII, no. 2, duly 1949, pp. 117-130. 11. Costello, George R.: Methods of Designing Cascade Blades with Prescribed Velocity Distributions in Compressible Potential Flows. NACA TN 1970, 1949, 12, Betz, A., and Filigge-Lotz, I.: Design of Centrifugal Impeller Blades, WACA TM 902, 1939. 13, Wu, Chung-Hus, and Brom, Curtis A.: Method of Analysis for Com- pressible Flow Past Arbitrary Turbomchine Blades on General Sur- face of Revolution. NACA TM 2407, 1951. 2223see ACA ‘Tm 2455 23 14, Dreyfus, bs A.: A Three-dimensional Theory of Turbine Flow and its ‘Application to the Design of Wheel Vanes for Francis and Propeller Turbines. ACTA Polytechnica, Mech. Eng. Series, vol. 1, ur. 1 (Stockholm), 1947. Reissner, Hans: lade Systems of Circular Arrangenent in Steady, Compressible Flow. Studies and Essays. Courant Anniversary Volume, Interscience Publishers, Inc., (New York), 1948, pp. 307- 327. 16. Stentz, Join D., and Ellis, Gaylord 0.: Two-Dimensional Compres- sible Flow in Centrifugal Compressors with Straight Blades. NACA Rep. 954, 1950. (Formerly NACA BN 1932.) 17. Bickley, W. G.: Formilae for Numerical Differentiation. The Math. Gazette, vol. XV, no. 263, Feb. 1941, pp. 19-27. 18. Wu, Chung-Hua: Formulas and Tables of Coefficients for Numerical Differentiation with Function Values Given at Unequally Spaced. Points end Application to Solution of Partial Differential Equa- tdons. NACA TN 2214, 1950,. . 2223 oy cor Dy wns] = | ae pote [een fomtnl ope | ee [P= (SL fae ae ® a s 1 ater asi 2 faase 3 |- .ose0 70408 4 | e466 1.81627 5 |--ssu z.orse7 | 1.01246 o.oensr 8 am Lueresa | 1.01657 e.7s |” loess Legose | *lesess Talbgeno v)| aia itzsera | Srreee “sass oases | .z4eas = sons 9 | 6288 osose |- .emas sas. | ere same |= 19077 340 za} asasea | see a.ar20 |~ eee |2.60000 |-2.a00%e aoems a2_| x.s00¢ | ose |2.2080 |-a.asove [o.ses | a. ross vse. a5" | E6026 | “emo [Tame | “ress | loon |-2tsre0 ie su | oes | - soca | a.senso = -asee2 as eam |- 2.06n |-a.sesn oe ae | esos | s.ausre |-a.s4007 eeceoee | enmeeee ar | evs | a.sase ae ae | ecerse | 2.4008 seve Mb vovN szI a @ Snes an ® lots Tyg ea soe es es [> feels Hist | eee ee rec eens oem | ees ges ee eee | 8 = Sou] 8517 | .971 | 1.05ms| 2.07887 | 1,800 1.08687 (25487 0.25763 ej. .asma| .a7s| 1.2500] .oses|i-arese | 2,408 1.12980 Lezse 2.08268 el iesee | see | eels ese |p ee pane a a alleen tes | cen | ee |p Eee fae 7s Seca gees |e |e ie cess tee ane | reer (arse area | aces |e 5 Sooke os epee S| tet recat tees eter |ateces | eras | eee tee | Sear ou ee am eee eres | | ema |e | eee |e Hns|Eitt| fe] tam fcummelteme | sme | as eee | tee Be) reBs] mie) rage) eme rem) gee | eee |e | reece |= ose | ee mes cee |e | eres | eed toe |e Be a ans | ene ns nes ter | eer | ted | ee |e | ees es tes at Uae ee | ec | cee | eese ra ee oe |e ee [pe ||) szee 92 sore Mi VOVE[P-UERIVASIVES AE MEAN OTREAGING GF BECOND REALE (oamssris P0¥) 2223 ol_e |e] es [a Teta Or re, Manin ty | mete ta] Bevin of seven ©) o@-00| *@ | 8 |:e0-8 o.sreat sea, = Tier caone0s -2ast6 -a.sr002 = .sssrs 0.13504 = enisas alkene eroes = seu 45100 fags eaons 2.44308 2.60780 coun 28855 2.00764 ams -a.e8790 -a.orase aisaue cove Nb VOW uzfoue THE - COPLMieaEN of VEDI CORONER 45D MASS FLOY AZ OROION 10 OF HE PE RUMI (smMERERETE TI) (io = 0.070) vg = 1.0080) ep = 0305851 fy =-0,0085] a a vol BBE ee PLEA |= I” RAAy » wr | Eine jo-o0+ eeo-o0 |e 40000 | 08000 =. | aaene el mi = 3 | yz 7 Ir male oad been he pz ae a see 82 sez NL YOvaL2223 [Ae 57 = CCURATDN oF YRLECIDY COMERS AD WI FLOW RATED 10 GF ER GUND EMER (OOPSLA VL) [to = 2.48 ay = 2.008) ay = 0.886 Oy, 0.0035] pigeveide eal Tete oTe, se Te fete tere 5 he pag hee Ln wr [Foot 00-9007 [s.2 log [nO [iro sop2 NE OWE 6230 NACA TH 2455 1-376] 0.2065) 0.7716} -0-086i| 0.2726) 0.0585 1oz0|-2.20n2|-"“ae10| °s0sa|”-s007|-aeis| 227] 0.0782|-0.4s16| ‘as2| 1.0122) “20a ‘aeaa| ‘oou1) ‘asva}‘oz7s|~“seoe) — -22s0) ‘rss|-“seae] ‘20sa]- “ooes|- ‘aaoe]- “ia97|- ‘a1ge|- “zes2|- “ier ‘58a “ioae| “2eca|” “ozia|- “ozss}” ‘zro6]” “osas| “si2s| “x2eal “aaa|- ‘sr09]- “sees] ‘ez0a|~ ‘iors| ‘asoc) ‘ozzs] - ‘ouso|- ‘9737 “esa|> “strs| = Yoo05| 2esi| ‘SABLE VE - ESPRAGED PERCEAGE ERROR IN B¢/By OF Felons [ose [ose [ove | 0.90 | 2.00 | 126 | a0 | ae2 | 260 f..a76| ~0,01008)-0.012i7| ©.c1z23) 0.01s4a) 0.001 f1:029} -0.00022|~ corm} -o111s| -co08| -00773|- -coxzs| 0.01248} 0.02031 |-0.00s87| ‘eea| 01599) “eosta| “oasaa|- “oosso} - “oosea|= ‘ooss2|~“o0sss| ° ooise|- -00z26) o.oxses Zr5s} “oisa3|“oovas|- ‘ooz2a)- “ooom| “oonse|” “ooos|” “oisan|~ ‘oo177 |” Zoossn) °-onse8| ‘s8a| - “ooexo| - “o085e|” “oosan|” “ooxza|- “ooa4s] Zooao| Zooses|- Corse | Jcos96|- £00190 ear] = “oo107|= “oxsea) “oxzz0)- “ooesa)- “oiszi|- ‘oo1s9|~ L00785| Lo0zs6|- ‘o0sso|- ‘0is08] “z94|> “oo015) Tooas7| ‘ososs | ‘oios2)- Loossa| nay [° Soosse|” Lonsva lo 00555] e2222.0) as 2223 seve ML VOvE. Be = _ n ee , 1 45 es an is as es a7 he svation i . L 1 : 7 © ergs one .0 28 Figure 1. = Hoan stranaline, inst and exit angles, and ¢: 1s32 Ya NACA BH 2455, 4 W Figure 2, - Determination of blade coordinate and velocity on blade. 2225 WACA IN 2455 2223 1.04, 1.00 +96 e +92 +88 1.6 2.0 Bah 2.8 3.2 3.6 (otug sec p)? ‘ Figure 4. - Variation of p* with (p*w3 sec 8)? in accordance with equation (12b).34 ae 2.08 1.04 2.00 96 o 98 86 84 76 272, Te 8 2, “2 Wz, 4 Figure 5, - Variation of p* with equation (12) NACA TW 2455, 6 2.0 24 2.8 Wte,1 tm accordance with eee2223 NAGA TIT 2455 Flgwe 6. = Inverse solutions for compressible end incompressible flovs. 35Figure 7. - Velocities on blade surfaces Solution jo Compress sbie (0 ~ — ~ Inccapressible and noan etreanline obtained in inverse colutins. sae oe soya Mb VOWNACA W 2455 37 2223 Figure 8. - Comparison of incompressible solution with original blade.38 ACA BN 2455, 2.0 1a saa 2.2] Relaxation solution (Feference 13) © Present value of YoHaye Ee Pigure 9. - Comparicon of axial velocities. Vertical Lines are stations. Using Seale, read horizontal distance from station to corresponding curve to dbtain velocity.2228 ACA ‘IH 2455 39 2.0~ Relaxation solution ah (reference 13) © resent value ¥, — fay o 12 <6 78 1.0 she Figure 10. - Comparison of tangential velocities. Vertical Mnes are stations. Using scale, read horizontal distance fron station to corresponding eurve to obtain velocity.40 NAGA TN 2455 18 sz2e 16 La Le 1.0 ——— Felaxation solution ah (zeference 15) © Present value of — wh, Figure 11. - Compariaon of resultant velocities. Vertical lines are stations. Using scale, read horizontal distance from station to corresponding curve £6 obtain velocity.NACA ‘BY 2455 aL A 8 1.2 1.6 2.0 2, in. Figure 12. - Blade-thickness distribution. ° 8 1.2 1.6 2.0 z, in. Figure 15. - Desirable mean blade line and corresponding Dlade shape.42 ACA ‘BY 2455WAGA TH 2455 43lly gy) —aA a g a ay « BO ag pW, La a . \Frid2 ein o ey of revolution (a) Stream f11ament of revolution. (>) Mean strean surface of revolution. (c) Element. (a) Applications to axial-flow, mixed-flow, and radial-flow turbomachines. Figure 16, - Flow on arbitrary eurface of revolution and stream filament of revolution, sov2 ih VOVIENACA TN 2455, ' go e a 3 |} Lit Toho hw23 angus Se tshatt = constant Se Figure 17. - Blade section on arbitrary surface of revolution.