1、./ t .; /-, , o 2,- .,/,d-_,“,i -J/i/.,-/v,., cl-_ 2NATIONAL ADISORY COMMITTEEFOR AERONAUTICSTECIFIT/CAL NOTE 2888PERFORMANCE CHARACTERISTICS OF PLANE-WALLTWO-DIMENSIONAL DIFFUSERSBy ElliottG. ReidStanford UniversityReproduced FromBest Available CopytVashingtonFebruary 195320000504075jProvided by IH
2、SNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-K the test program includedwide variations of divergence angle and length. _Iring all tests adynamic pressure of 60 pounds per square foot was maintained at thediffuser entrance and the boundary layer there was thin
3、and fullyturbulent.The most interesting flow characteristics observed were the occa-sional appearance of steady, unseparated, asymmetric flow - which wascorrelated with the boundary-layer coalescence - and the rapid deteriora-tion of flow steadiness - whlch occurred as soon as the divergence anglefo
4、r maximum static pressure recovery was exceeded.Pressure efficiency was found to be controlled almost exclusivelyby divergence angle, whereas static pressuze recovery was markedlyinfluenced by area ratio (or length) as well as divergence angle.Volumetric efficiency diminished as area ratio increased
5、 and at agreater rate with small lengths than with large ones. Large values ofthe static-pressure-recovery coefficient were attained only w_th longdiffusers of large area ratio; under these conditions pressure effi-ciency was high and volumetric efficiency low.Auxiliary tests with asymmetric diffuse
6、rs demonstrated that longi-tudinal pressure gradient, rather than wall divergence angle, controlledflow separation. Others showed that the addition of even a short exitduct of uniform section augmented pressure recovery. Finally, it wasfound that the inst_llatlon of a thin, central, longitudinal par
7、titionsuppressed flow separation in short diffusers and thereby improved pres-sure recovery.Provided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-2 _LAC_ TN 2888I_RC_JCTIONThe exTerimenta! investigation reported herein was conceived as thefirst element of
8、a broad research program directed toward the followingobjectives: To identify the conditions upon which diffuser perforrmnceis principally dependent, to determine their influences, and to utilizethis informztion in the development of improved diffusers._nile the elevation of diffuser efficiency with
9、out regard fordimensional limitations is obviously desirable, the most welcome improve-ment from the aircraft designers viewpoint would be the reduction ofcarrent lengths without sacrifice of efficiency. Special interest istherefore attached to diffusers with large rates of divergence.Since diffuser
10、s have long been widely used, the necessity of seekingthe first of the objectives stated above may seem somewhat anomalous. Inmost technical fields, the modus operandi, capacity, and limitations ofcommonly used devices are usually well-known before they have been soused for more than a decade. Unfor
11、tunately, this is not true of dlf-f_sers - although they have been used for more than a century. 1 As amatter of fact, although the lack of fundamental information on th!ssubject has become increasingly apparent in recent years, relativelylittle new light has been shed upon diffuser performance duri
12、ng thehO years which have elapsed since Professor A. H. Gibson completed hlsnov-classlc experiments (references 1 and 2). To bring this situationinto sharp focus, a brief outline of the present state of knowledgeregarding diffusers is presented herewith.The availability of several competent digests
13、of existing diffuserliterature - notably the one by Patterson (reference 3) - makes itunnecessary to outline, here, much more than the boundaries of thatinfcrmaticn and, as implied above, thls requires but fe_ additlous to ar6sum_ of Gibsons work. In that r6sumg, however, emphasis is given toan aspe
14、ct of the work which the writer believes to have received unde-ser_edly scant attention in the past.The diffuser investigation usually associated with Gibsons nameconsisted in the testing - with water - of three families of linearlytapered diffusers which had circular, square, and rectangular crosss
15、ections, respectively. (The rectangular ones were of two-dimensionalform, i.e., they had two parallel, and t-_o divergent, walls.) Arearatios R of 9.29, _, and 9 were incorpozated in the circular andrectangular types, whereas all the models of square section had arearatios of _. In each case, models
16、 of various lengths provided coverageof the range of wall divergence angles 2_ between small values and 180 .1Uriah Boyden (180h-1871) is generally credited with _troductlon ofthe diverging discbmrge tube as an adjunct to the water tu.bine.Provided by IHSNot for ResaleNo reproduction or networking p
17、ermitted without license from IHS-,-,-_“mCAT_ 2888 3Despite some shortcsmings of technique - as seen f:om the modernviewpoint - the results of these tests indicated that the diffusers ofall three types were characterized by sharply defined minimums of headloss which occurred at divergence angles 29
18、between 3.9 and ll, thatthe bead losses increased rapidly toward the theoretical values corre-sponding to sudden enlargement of section as the divergence anglesexceeded their optimum values, and that the losses in comparable dif-fusers were least for the circular, and the greatest for the rectangula
19、r,cross sections. These general characteristics have been repeatedlyverified by others and no significant errors in Gibsons quantitativedata have yet come to light.Upon completion of this outstanding - but, nonetheles: essentiallyroutine - exploratory study, Gibson embarked upon an investigation ofm
20、ore f_undamental character. Unable to deduce, a priori, the optimumlongitudinal distribution of cross-sectional area for a diffuser, heinvestigated the characteristics of the three curved-wall types whichappeared to him most promLslng. The first was so designed that, if theflow were frictionless, th
21、e retardation dV/dt would be constant through-out the length of the diffuser; the resulting form is best described as“trumpet-shaped.“ ThL. second, which had a less-proncanced flare, wascharacterized by const%ncy of the ideal value of dV/dx. The third wasdesigned by an empirical method 2 intended to
22、 provide uniform loss ofhead per unit length; the wall curvature of this type was the least ofthe three.Only three models of the first two types were tested because nosignificant improvement was effected. However, 13 models of the unlform-head-loss type - 6 of circular section and 7 rectan_alar, two
23、-dimensionalones - were built and tested and all of them proved superior to thecomparable linearly tapered diffusers. It Is unfortunate that the effec-tive divergence angles of these curved-wall diffusers were greater thanthose a_ which minimum head loss occurred in their _linearly taperedcounterpar
24、ts because this precludes the direct comparison of relativemerits under optimum conditions. However, the measured reductions ofhead loss ranged from 16 to more than 90 percent and conservative extra-polation of the corresponding experimentally determined curves leaveslittle doubt of the superiority
25、of the uniform-head-loss type even underoptimum conditions. 32Based on the experimentally determined relaticzship between headloss and divergence angle for linearly tapered diffusers; for details,see pp. i06-108, reference 2.3Ackeret (reference h) tested two very similar carved-wall diff_sersand obt
26、ained results com_istent with those of Gibson.Provided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-4 _ACA T_ 2888The fundamental importance of this phase of Gibsons work is foundin neither the development of an optimum diffuser form o lot there isno evide
27、nce that this was accomplished nor the considerable improve-ment of efficiency which was achieved, but rather in the demonstrationthat the efficiency of a diffuser of given length and area ratio issubstantially influenced by variations in the longitudinal distributionof cross-sectional area. It also
28、 seems worth noting, specifically, thatthe foregoing results clearly show the linearly tapered type of diffuserto be endowed with no special virtue except simplicity of form.At this point, attention is drawn to the striking analogy betweenthe diffuser of fixed length and area ratio and the airfoil o
29、f specifiedcamber and maximum thickness. Recognizing the fact that Gibsons studyof diffuser profiles was a preliminary one which has never been system-atically extended, it appears not unfair to appraise the present stateof knowledge regarding difi“users as no better than that which prevailedin the
30、case of airfoils Just prior to the investigations which yieldedthe low-drag and high-critical-speed profiles now in common use. Thusthe principal necessity cf the first undertaking of the present programis found in the fact that, as of today, the effects upon performance ofvarying the longitudinal d
31、istribution _f cross-sectional area in adiffuser of fixed over-all proportions are neither comprehensively knownnor thoroughly understood.While the foregoing co_,ents do not imply tDmt there has been littleprogress in diffuser research since Gibsons work was published, it doesappear that attention h
32、as been largely diverted from the properties ofsimple diffusers and concentrated upon auxiliary devices intended toovercome their apparent deficiencies. Of these auxiliaries, boundary-layer control, entry guide vanes, and rotation vanes appear to deserveindividual comments here.Perhaps the most infl
33、uential deterrent to further research on plaindiffusers is the success with which suction boundary-layer control hasbeen applied to the suppression of flow separation in short, wide-anglediffusers. The effectiveness of this arrangement, originally suggestedby Prandtl in 1904 (reference 5): has been
34、demcnstrated by Schrenk(reference 6), Ackeret (reference h), and more recently by Biebel (refer-ence 7). While it has almost unlimited possibilities, the use of boundary-layer control involves the provision of auxiliary ducting and either ablower or some other suction-producing device of adequate ca
35、pacity. Theseare complications which aircraft designers have, thus far, been unwillingto accept.Some promising work with entry guide vanes .has been done by Frey(reference 8), but its scope was so limited that the results are notgenerally useful. However, the attainment of pres_ule efficiencies ofPr
36、ovided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-DNACA TN 2888 552 and 47 percent with dif_asers of R = 3 and divergence angles 2eof 90 and 180 , respectively, demonstrates that moderate pressurerecovery csn be had even in very short diffusers. _ile the
37、 effective-ness o_ such vanes in diffusers of moderate divergence is conjectural,investigation of this question appears well-warranted by Freys results.The idea of using fixed vanes to produce helical flow in a diffuserprobably stemmed from earlier efforts to design efficient draft tubesfor water tu
38、rbines from which water is discharged with a vortexlikedistribution of tangential velocity. Peters (reference 9) has shownthat if a substantiallyuniform, that is, rigid body, rotation is super-posed on the axial inflow of a conical diffuser, the pressure efficiencyis considerably greater that that f
39、or simple translatory flow. A con-siderable part of this improvement may, of course, be ascribed to thefact that, since the spiral path is longer than the rectilinear one andthe pressure rise per unit of path length correspondingly smaller, theintroduction of the tangential velocity is equivalent to
40、 increasing thelength and thus reducing the effective divergence angle of the diffuser.However_ the _emonstrated improvement of the efficiency of a diffusercharacterized by the optimum divergence angle for translatory flow can-not be thus explained and Patterson has suggested that it may arise fromt
41、he radial pressure gradient which is peculiar to the spiral flow.The practical significance of this work has been at least ambiguously,if not erroneously, interpreted by Patterson who concludes, in reference 3,that, “In a conical diffuser having an angle of expansion in the range15 deg. g 2e _ 50 do
42、g. an efficiency of 80 per cent can be obtained bysuperposing a rigid body rotation on the axial flow.“ Since the vanesused by Peters were installed well upstream from the diffuser entranceand the efficiencies computed from data obtained at the entrance and ata station in the exit duct, these effici
43、encies are based on the existenceof helical flow at the entrance and take no account of the energy lost inthe production of the tangential viocity. Thus Peters experimentsdemonstrate only that, if appropriate spiral flow exists at the entranceof a conical diffuser, the efficiencies cited by Patterso
44、n may be obtainedand they do not prove that the efficiency of a given diffuser may beaugmented by installing within it rotation-producing vanes. This pos-sibility is, however, one worth investigation and a bas_s for the expecta-tion of some improvement is seen in the high efficlencles obtained byPet
45、ers with diffusers having large angles of divergence. An additionalpossibility which deserves consideration is that of recovering energyfrom the tangential motion by the use of counterrotation vanes at thediffuser exit.Because of their bearing upon the character of the present experi-ments, two addi
46、tional items must be included in this resumg; they concernthe influences which the entrance boundary layer and the exit duct exertupon the efficiency of a diffuser.Provided by IHSNot for ResaleNo reproduction or networking permitted without license from IHS-,-,-6 NACA TN _=_88TLe former can be descr
47、ibed quite simply: It -_as been demons:rate_S -perhaps most thoroughly by Peters (reference 9) - that the pressureefficiency of a diffuser diminishes as the thickness ef its entranceboundary layer increases. The effect is most prone:riced when the i=_yeris very thin and tends to disappear as the thi
48、ckness becomes large.These findings have been verified at high subsonic steeds by the workof Copp and Klevatt at the Langley Aeronautical iab_ratory of theNational Advisory Committee for Aeronautics; however, the results cfthis work ace not yet generally available.The character and origin of exit-du
49、ct influence have long beenkncwn. Gibson, for example, reported in reference i that, when adiffuser discharged into a uniform duct having the same cross sectionas the exit, maximum static pressure occurred not at the exit sectlcnbut at some distance downstream in the duct and this fact has beenverified by numerous others. Redu:tion in the duct of the nonunifcrmityof velocity with which
