w

NASA Technical Memorandum 81668

I

lk Factors Which Influence theBehavior of Turbofan ForcedMixer Nozzles

R. H. Anderson and L. A. PovinelliLewis Research CenterCleveland, Ohio

(NASA-Td-d1UU6) ?ACTUccS anI CH INFLUENCE TdE N61-15240

bEHAViud Ur i'UEiiUt'AN FuLtCE,) illeiri NU,,4LES

(NASA) dJ N iiC AU3/[1F Avl CSC;L lolll)!iL'13S

G3/34 2`!t,9 7

Prepared for theNineteenth Aerospace Sciences 'Meeting g.^p3^ 12345sponsored by the American Institute ofAeronautics and AstronauticsSt. Louis, Missouri, January 12-15, 1981

X^

NASA v ul ^ l Z`5`

r

https://ntrs.nasa.gov/search.jsp?R=19810006725 2019-08-09T18:47:08+00:00Z

-

~-

•

I

FACTORS WHICH INFLUENCE THE BEHAVIOR OF TURBOFAN

FORCED MlXER NOZZLES

by B. H. Anderson* and l. A. Povinel11** National Aeronautics and Space AdMinistration

Lewis Research tenter Cleveland, Ohio

StJttMARY

A finite difference procedure was used to compute the mi~lng for three experimentally tested mixer geometries. Good agreeB~nt was obtained between analysis and experiment when the mechanisms responsible for secondary flow generation were properly modeled. Vorticity generation due to flow turning and vorticity generated within the centerbauy-lobe passage were found to be important. Results are presented for two oifferent temperature ratios be­tween fan and co.~ streams and for two different free-stream turbulence lev­els. It was concluded that the dominant mechanisms in turbofan mixers is associated witt the secondary flows arising within the lobe region and their d,~veloprnent within the milcJoi section. •

INTROOUCTION

Significant performance gains are achievable in turbofan engines by mixing the hot core stream ~ th the cooler fan stream prior to expansion through the exhaust nOlzle. The amount of performance gain depends on tne balance between the degree of mixing of the hot and cola streams, and the pressure losses incurred during tne mixing process. The lobe mixer has been the most successful device that has been used to promote mixing. To date, the perfonnance 9f lo~~ mixers has been Cletermined almost ~ntirely through experimentation.~1,2,3J These experiments were sufficient to detennine tne relat ive merits of on~ lobe configuration over another. However, in tne ab­sence of any clear understanding of tne mix lng process, conclusions reached with one series or tests could not De generalized to the next generation of mixer nozzles. Fluid flow analysis for the ~e~ign of turbofrn mixer nozzles was first published by Siren, Paynter, Spa1ing, and Tatchell 4) and showed encouraging results. However, the ro1e of various aerodynamic processes whictl take place within the mixer nozzles was clearly still not understood. It was first suggested in the lewis Aeropropul~ipn 1979 Conference(5) and 1 ater by Povi ne 11 i, Anderson, and Gers tenma i~r~ b) that convect i \Ie t orces arising from pressure ariven ~econdary flows playa far more important role in the mixing pro~esses than first anticipatea. These convective forces arise due to s~condary flows at entry, which are sustained and amplified in the mixer passage by transverse pressure gradients that deflect the mean flow.

*Hcdd, Aerodynamics Analysis Section; Member, AIAA. **Aerospace Engineer; Aerouynamics Analysis Section, Associate Fellow,

AIAA.

"'" _---,;0- '" ~ ~ .. & ._ r • T • - • - "-

e

These conclusions were based on the computations obtained with the vis-

cous marching procedure developed by Kriskovsky, Briley, and McDonald.(7+8). 1Subsequently, LDV measurements performed at UTkC under Lewis sponsorship,

clearly demonstrated the presence of strong radial and some tangential flowsat the lobe exit plane (ref. 9). Concurrently, flow surveys at Lewis, re-

vealed radial temperature profiles whose particular patterns could be ex-

plained by the presence of radial and tangential flows witnin the mixing

duct (ref. 6). In addition, flow angularity data were obtained which re-vealed strong radial flows at the lobe exit plane. Computations were then j

performed using a representative or generic secondary flow field as a start-ing condition for the computer program (ref. 10). Use of the generic flow

field simp lified the task of obtaining data for the initial starting profile iand led to good agreement between experiment and computation for a typical p^model mixer. However, application of the analysis to several other geome-tries suggested that nut all of the flow phenomena were accounted for in thestarting profiles. It is postulated drat otner basic geometry dependentflow structures exist at the mixer entry wnich crave an infl1jence on the flowdevelopment. These include not t1oo under the fan trough or additional vor-

tex structures. Specifically, the objective ut the present study was to

further define possiule vortex mechanisms occurring in the mixing passage

and to demonstrate the ability of (I viscous computer analysis to calcu-

late the flow in a variety of turbotan mixers.

ANALYSIS

The calculational procedure used in this study of forced mixer nozzle

flow fielos is an application of the approach developed by Briley, McDonald,

and Kreskuvsky( 1 , 8 ) and is designatcd PEPSIN. The procedure is based onthe decomposition of the velocity field into primary and secondary flow

velocities. Equations governing the stredmwise development of the primaryand secondary flow velocity tielos are solved by an efficient algoritnm

JON both block ana sCalal' AN nrethuus. AlLhuuyh the guverning equationsare suivec; by a torward mar•Chrny method, elliptic effects due to curvatureand area change are accounted fur a-priori through the rmpuseo pressure yra-dirrnts de tt'_rMille u tr'uni a putentidl M sulutlun for the geometry in ques-

tion. Since thc' pr - iilrar'y CunCer'n in the lobe mixer problem is thermal mixing

of the tan and core streams to achieve t!rrust augmentation, an energy equa-t ion is lntroducteu.

For turbulent t it)ws, thN turbulent viscosity wnich appears in the gov-erning equations must be Specitleu, lwo types of turbulent models dre useuin PLPSIM to determ;ne turbulent viscosity; a twc •-equation k -E turbule nCe'ITIOde1 as presented uy Launder and SpduIaIny( lIi tnd awake-turbulencenude 1. 1 n th i s paper, on l)' t ne resu I t ubta r ned w i ttr tilee wake IU u encemo e7 are presented.

III wake-turbulence model, lh,, assumption is maue

that the turbulent VC10Ctty ,Ind ierytrn scales are frozen at their initialva lue. Tnus, t y re turbu lent v i ..,cos r ty i s proport Iona l to .yens i ty.

Spei It IL at ruff ut the snit ii ( rfIt low) CMiditlons for d lobe mixer cal-culat ion in P r .PSIM may be pertaormed eItner uy specItyIny the initial values

of the ve Ioc ity (Int; temperature t ie id or by us i ny all procedurewhich constructs Lhe initial velocity and temperature tields parametri-Call _v. In ertrr^•r • Case It is unllkely t'-dL the specr' reu .L l ocity tieldwould be L01%)atrble with tn( 2 CunLlnuity ('uuatruns. lnert•ture, tills velocitytield is rrrudifrt'd rn wa) ttnut crakes It. Cumpatrble with the Curt,riulty

equatiOn while at lie satyr; trrtrC marnt.rrnrng the initial secondary flow vor-

3

ticity unaltered. The streamwise momentum equation is solved to obtain ap-

proximate values for the streamwise velocity gradients. ilie secondary flowvorticity is com^uted from the specified velocity field. A scalar potentialis then construc.Led and solved yielding the irrotational components of the

secondary flow which balance the streamwise velocity gradient in the conti-nuity equation. Finally, a vector potential is constructed using the ini-tial secondary flow vorticity and solved yieluing the irrotational compo-nents of the secondary flow. Due to the nature of the vector potential

these velocity components do not effect the continuity balance obtaineu fromthe scalar potential. Tire resulting velocity field is compatible with the

continuity equation to first order in the marching direction.

LXPERIMEI4TAL APPMATUS

The test apparatus used has been described in a previous paper( 6 ) andconsisted of two basic parts: a fixed upstream model section anu a rotating

Shroud (fig. 1), The upstream section simulated the flow path through atypical high by-pass turbofan engine. A cross-section of the mode l is shownin figure 2(a). Heated air was supplied to the core passage and flowea

through the lobe section. Unheated air was supplied to the fan passage andflowed around the lobe section which was interchangeable. In this paper,the results obtaineu with three mixer sections are presented. lne longitu-

dinal contours of the three mixers are shown in tigure 2(b). Configurations

A, B, and C has a penetration (lobe tip rautus/shroud radius) of U.822,0.776, and 0.721 anu the circumferential spacing ratio (core included

angle/fan included d fig 1e) of 0. 5, 1.U, and 1.3b. The ratio of the shruud

length to the inside shroud diameater kdt the lou exit plane/ was 0.71.Total pressure and temperature measurements were mane upstream in both

the fan anu core flows. Instrumentation rakes were also mounteu in the ro-tating shroud for probing the mixer flow fielo ksee fig. 2). Total tempera-ture rakes were located at five axial stations in the unxing region. The

first station was dt the lobe exit plane, the second was halfway to the endof the plug, the third was at the end of the plug, the tourth was midwaybetween the plug end dnu the nozz ► e exit, and the t iftil station was at thenozzle exit plane. the rakes at the lobe anu nuzzle exit stations as wellas the rotatinu 1;1echdnis 111 are shown in tigure 1. Total pressures were alsomeasured at the lobe dnu nozzle exit stations. The temperature data were

erhtarned over a 54-degree Seg;rnent in ,-ueyree increments at 14 rddial posi-tions. Flow angularity measurements using the fixed probe technique(12)were made at tour circumferential locations and at the two axial locationsshown in tigure c. lire first measuring station was at the ivbe exit plane;

the 'Second at the end of the centerbooy. A photograph of the angularityrakes are shown in figure .i• Each rake lids six probes dnu each probe has

three tubes. The center tube is a chambered total pressure probe anu the

two side 1 upper anti lower) tubes have a 45-degree sweepback. The pres,.uredifference between the two slue probes and the inoicaced total pressl;re were

used alon(i with d calibration curve in order to obtd`n flow directirns (ra-uidl and dzlniutnal). Ldcn probe was individually ca^rbrateu in an open ,letat d Mach i umber of 0.45, wherein the rau io i prone WLS set A t va r ious pitchangles to the flow anu the azimutildl probe at var • luus yaw angles. Tile indi-

cated total pressure and Lne uifterences in sine tube (upper-lower) pres-sures were used to eStdblrSh d calibration.

4

Wall static pressure measurements completed the infonnation neeaea to

compute d resultant velocity. Tne resultant vulocity was then used with the

radial dlid dZIMULlldl f low angles to c:ompuLu One three veluOLy curllpuneuLs.The fan and core streams were operated with a total pressure ratio of

one and a total temperature ratio of (fan/core) of 0.14 or U.4. The Mach

number of the fan and core streams at the mixing plane (lobe exit) was ap-

proximately 0.45 and the by-pass ratio was about 4.

COMPUTATIONAL PROCEDURE

A. Computational Mesh

In this section, specific details arising from the application of theforegoing analysis to the turuofan mixer configuration described earlier is

given. The cross-section of this1r i xer geometry is presented in figure 4.

The area immediately downstream ;+ the nozzle plug tip is faired in with an

assi gned streamline to wodel the ,eparated flow region expected in this mixernozzle. Since the flow area excluded from consideration is small, this

treatment is not believed to introduce significant error. The curvilinearcoordinate system shown in tigure 4 was constructed to fit the flow passage

boundaries and has 21 streamwise nodal points, 40 radial r.odes and 11 azi-

muthal modes. In planes of constant azimuth, orthogonal streamlines and

velocity potential lin 3 were constructed from a two-dimensional plane in-

comp ressible analysis.`^ 3 ) This x-y coordinate system was then rotatesabout the mixer axis to form the axisymnletric coordinate system. Five ref-erence stations are identified in figure 4 and these correspond to the fiveexperimental survey stations mentioned in the previous section. These arelabeled 1, b, 13, 0, and 21 and correspond to the computation nodal pointnearest to the preuing stations. Station number 1 corresponds to the lobe

exit station while station number 21 is the mixer exit station.Althuugh the mixer geometry is axisymmetric, the tlow is three dimen-

sional due to the aximuthal variation of the not and cold streams. However,due to observed symmetry, only a 112 lobed pie-shaped segment of the trans-

verse coordinate surface was considered. Tne shape of this segment and theextent to typical not and cold streams at the lobe exit station is shown infigure 5. A comparison between the computatiunal and experimental lobe

shape are also shown in fiyure 5.

B. Flow Angularity Measurements

The three velocity components near the exit popane of the lobes weremeasured using the flow anyularit y prubes described in the Apparatus sec-tion. Tne flow anyularity data were obtained in order to provide intorma-tion auout the mixer inflow conditions Tne data were measured in a planeparallel to the exit plane of the loues (see fig. 2). Data were obtained at

six radial locations and four circumterential positions within the measure-ment uomain shown in figure ti for the lib anu leL lobe geometries. Toe vec-tors snown in fiyure n are the resultant of the measured radial and azimu-

thal velocities in a plane transverse to the mean flow. Strung radial flowsare evident with outflow in the core and inflow in the fan regions. Data of

this type were discussed previously in reference b and it was suggested thata vortex-type flow was present at the fan-core interfacial regions. In ref-

erence u, the flow anyularity dat was "enriched," using a tour-point linearinterpolation scheme. The enrichment was carried out in order to obtain a

5

more complete representation of the secondary flow structure at the lobeexit plane.

Computations, using the generic starting data yielded good agreementwith the data for configuration 12b (see ref, IU). However, comparisonsbetween analyses and experimental data for configurations 12A and 12C indi-cated that perhaps the initial starting conditions for the computer codewere not correct. The comparisons prompted an evaluation of the mechanismsresponsible for the generation of secondary flows.

C. Secondary Flow Generation

As mentioned previously in reference 5, streamwise vorticity is genera-ted at the entrance to the mixer and sustained and amplified in the mixerpassage by transverse pressure gradients that deflect the mean flow. Threeadditional mechanisms appear to be responsible for the generation of second-ary flow in lobe mixer geometries. Tile most important one is due to thebasic turning of the fan and core streams in oppsite radial directions, asshown in figure 7(a). The secondary flow generation here is basically anireviscid phenomenon ano results in outward radial core flow adjacent to in-ward radial fan flow as shown in figure 7(u). It is believed that thismechanism is properly represented in the starting conditions via the flowangularity data obtained for each lobe geoinetry. The second mechanism re-ponsible tor • secondary flow is due to the interaction of upstream ductboundary layers with the lobes, which in this case represent flow obstruc-tions, as shown conceptually in f i gure H. The vorticity witnin the boundarylayers encounters tine lobes and vortex filaments wrap around the lobe in anorseshoe-life pattern. A set of vortices are subsequently set up both inthe fan troughs and a set in the core flow. However, inspection of the ex-perimental raolal and tangential velocities at the loue exit plane did notindicate that any significant effects were caused uy this seconu mechanism.In addition, the Ico w penetration lobes would tend to minimize the formationof strong horseshoe vortices. See reference 14 for further information.The third mecnanisin responsiule tur Secondary flows is a passage vortexwhich occurs as the core flow approaches the loue exit anu encounters thenarrow yap between the centeruody and the bottom of the fan trough. Asshown in f Ig ire 9, t y re vortex forms as flow washes up around the side of the.fan troughs. Tuft photographs taken of the inside surface of the lobes re-vealed a strong upward racial velocity component near the bottom of the lobeas shown In figure 10. Subsequent turning of the flow in the duwnstreamdirection occurs at a slightly larger radial position as inuicateo by thetufts in figure W. The uehavior of the tufts strongly suggests the pres-ence of a passage vortex. It was decided, therefore, to model the vortex inorder Co obtain the proper startiny conditions for the computer code. Inthe absence ct any measurements of the passaye vortex strength, a nominalvalue of the raulai outflow was asSUlned for the initial computation. Subse-quent estinra':es of the vertex strength were made for the tinal computationswhich art , p;-eserited In this Daper.

It snouId be noted that Lire cumputat1onaI model, shuwri In f Igure 5,does riot 1 oc I uue a riot core layer uetween the Cente; Uody and tile bottom ofthe Ian trough. Addltiinal grid resolution in the cuue Is requireo beforethe etfeet of tills layer un the computations can be assessed.

r,

D. Representation of the Inlet Flow Field

The secondary flow, that is, the flow that is generated transverse to

the streamwise direction, is highly complex after passing through the curves

lobe section. This highly complex nature of the real flow field procludesnumerical simulation of all the large scale and sma7 scale structure simplybecause this information is not available. However, the large radial veloc-

ities of the secondary inflow field can be measured experimentally and simu-

lated. This class of secondary inflow representation is i,ibeled qenericflow fielus, that is, they attempt to simulate the large scale secondary

flow field structure entering the mixer section by a parametric representa-tiun, as well as modeling the passage vortices. For comparative purposes, abaseline generic flow field is defined which is the generic flow without

passage vortex modeling. The baseline flow was used in earlier results

(ref. 10). In contrast to real and eneric inflow representations, idealinflow conditions incoproraTe—no sec o tlows entering the mixer pas-

sage. The flow in this case is parallel to the strearnwise coorcinate of themesh and was used ilt the earlier results presented by Povinelli, Anderson,and Gerstenmdier.(6

A generic representation of the large scale flow transverse to the

streamwise direction can be conceived as being composed of basically radialoutflow in the core passage and radial inflow in the fan passage (fig. 5).

The experimental data for the mixer nozzle 126 configuracion under study

suggests that the radial velocities in the core and fan lobe passages are 25and 20 percent, respectively, of the stredrimise velocity. For the 12A and12C geometries, the core Clow was 14 percent and the fan tiow was M per-cent. The streanrwise velocity at each mesh point in the two respectivestreams was assigned its nondimentional reference valu- which was then cor-

rected to account for nonnal pressure gradients at the initial plane as de-termined from the axisyammetric potential flow. To account fur boundarylayers on the lobe, plug and shroud surface, the streamwise velocity pro-

files were turther scaled in accordance with an assumed turbulent boundary

layer prof i le anu urstatiCe fr'uhr the lobe surface. The temperature field wasconstructed in the core and fan streams by assuming a total temperature

ratio T fah^ T core = 0.74 and 0. 4, which corresponds to the experimental tem-perature ratios. In aduitiun, an entrance Mach number of 0.45 was assumed.

Tne rditial turbulence quantities were initialized through specifica-

tion of a lenytn sidle .jnd tree stream turbulence intensities. These turbu-lence quantities are assumed to be constant across the shear layer but tovary with urstance iron y the wall. For the calculations presented in tnisp,3per, trio initial lenyth scale was set at 0.UUo of the outer shrcud radiusand t.urT,lent Intensity ut but.h the core and fan streams was set at either uor 1,' pore ent .

RESULTS

A. int luence of Secondary Flow Field

126 Mixer

1 11 '^tarti F ig VelOCItV ^ectu ► • t field constructed accordiny to the proce-dure' described in trio previous section is snown in tigure « for the 1213mixer. Construction of tnc' secondary veiocity vector' tielu wds based randveraye radial secunudr^ t ruws of (-:) and W percent ut the streamwise corn

1

and fan velocities, coirresponding to the experimental measurements. Al-

though the computational segment included only one half the lobe segment and

one half the core passage, the computational results in figure 11 were re-flected to represent one lobe and two fan regions. The secondary velocitiespresented are normal to the streamwrse mesh coordinate and are shown only in

the region near- the plug (centerbody) surface. Figure 11(a) shows the base-

line flow field, that is, with no passage vortices. Figure 11(b) shows thegeneric starting flow field where a set of passage vortices have been incor-porated in the core flow field. These passage vortices are in the region

between the ,flug surface and the lower part of the lobe wall as describedpreviously and shown in figures 9 and 10. The complete secondary flow fieldat the starting point, which is the lobe exit plane (reter to fig. 4), is

shown in figure 12(a). The computational results were reflected to in-clude two lobe and three fan regions as shown. A strong vortex pattern canbe observed which is aligned with the interface region baetween the fan and

core streams, as well as the small passage vortices near the plug surface.At station number 8, which is located half way along the plug, the vortex

p attern begins to condense into a more circular pattern and moves radiallyoutward. As indicated by the disappearance of radial inward flow near the

plug surface the passage vortices disappear by station number 13. Tne mainvortex pattern continues its outward movement (figs. 12(c) to (e)). At the

mixer exit plane (station number 21) the vortex is still relatively strong.

A comparison between the measured and computed total temperature con-

tours at the mixer exit plane (station number 21) is presented in fig-

ure )3. Tire development of the hurseshoe-shaped total temperature contour

first identified in reference b is clearly evident. This temperature signa-ture is the result of the secondary flow field which sustains itself by thenormal static pressure gradients within the mixer nozzle passage.

A comparison between the measured and computed total temperature dis-

tributions at the mixer nozzle exit are presented in figure 14 for six rakepositions. The centerline of the fan flow is at a theta value of 0 anu thecore flow centerline is at a theta value of 15 degrees (see fig. D). Compu-

tations were made assuming ideal inflow conditions (no secondary flow),

baseline inflow conditions secondary flow without passage vortex), andgeneric inflow conurlions (secondary flow with passage vortex). A compari-

son be weer the calculations using t y re generic seconoary inflow pattern and

the Treasured data show excellent agreement. The baseline inflow conditionsyielded good agreements except over the first 20 percent of the radius. Thestrong influence of t yre secondary flow structure is evident by comparing the

solid with the cashed ines which represent Ure generic and ideal inflow con-

uitions. I, is apparent that the characteristic hurseshoe-shaped tempera-ture signature identified in reference b resulted from the inflow secondary

flow vortex structure established by the initial core dnd fan streams.

12C Mixer

The corresponding data and computations for the 1,/L mixer are presentedin tigures 15 to 1!. the starting velocity vector field is shown in fig-ure i5 and IS based on an dverage radial velocity ut 13.1 and 8 percent ofthe streamwise cure dnu tan veiulcrties, based un the> experimental measure-ments. Figure ib(a ) snows the baseline flow field. Figure 15(b) shows the

generic starting flow tieiu wnerein the passdye vortices are located neart y re plug surface. the complete secondary flow field at the lobe ex;t planeis p resented in t rgure 10( a) . Tile suuseyuer t cndnges in Use flow field

8

downstream are shown in figures :b(b) to (e). The qualitative features ofthe secondary flow pattern are similar to those shown previously for the 128mixer, however, the magnitude of the radial flows is smaller and results inless radial displacement of the main vortices.

A comparison of the measured and computed total temperature distribu-

tions at the mixer nozzle exit are c .wwn in figure 17 for six angular loca-tions. In this comparison it is seen that the baseline inflow condition as

well as the ideal inflow condition yield poor agreement with the experimen-tal data, whereas the generic condition leads to excellent agreement. It isnoted that the 12C mixer yields a radial temperature distribution distinctly

different from that shov:n for the 126 in figure 14.

12A Mixer

Radial temperature distributions at the mixer exit plane for the 12A

mixer are shown in figure 18. As with the previous two geometries, the

generic inflow condition yields the itst agreement with the data. The anal-ysis appears to mess the spread of secondary flow and to interpredict tem-perature mixing oiver t y re inner W percent of the ra,ius and overpredicts

near the o0-percent point based on all the data points. However, it is

noted that the two experimental data sets have significant scatter at theb0-percent. radial position.

B. Int1uence of hot Lore Temperature

The results presented in the previous section were based on

the warm

flow testing where the ratio of fan-to-cure strean total temperature wos0.740. In this section, the results from d tully simulated ur hot flow con-

dition with a temperature ratio of U.40 are presented.Computations were mace using the generic, baseline, and ideal inflow

conditions for the 126 mixer having a high temperature primary stream. Tlieratio of core-to-fan total temperature of U.4U matcneu the experimental val-

ues used err the test prt^gram. Tlrt seconuary velocity vector field was con-structeu using the same average radial flows as used in the previous calcu-lations (tiy. 14). The primary velocity was increaseG assuming matchea Mach

numbers and total pressures in the two streams at the lobe exit plane. Acurnparisun of the computeu results with the experimental uata is shown in

figure l y . The comparison is at the mixer exit plane, that is, station num-ber :I, and incluues six anyuldr locations frum the centerline of the fanflow, tnet.d = 0, to the centerline of the core flow, thetd - 15. As in the

previous comparioos, the generic inflow condition yields the pour -=st agree-ment. It is rioted that tr,e experimental data at the fdn centerbody (fig.19(a)) shows a small tem p eratuare rise at a normalized radius of U.7, which

was not p resent in the warm flow testing (see fig., 14(a)). This rise inui-w.ates spreduiny of the pr,mdry stream into the core of the tan stream. Thecomputer cuue underpreuicteu either the secundary fIuws u the temperature

mixing in this region.

L. lnt luence of Turuule,:cv

AuuitTonal computations were pertormed with a h,yher level of free-stream turuulence in troth the fan ano cure streams for the it6 mixer geome-try. The computations were made for only the generic, irf low condition. Thecons tru(-trun of the sturtirig secondary flow y ield was ruent Ica I to that de-

9

scribed in section A. Toe temperature ratio, TF/TF, used in the computations

was 0.40. The turbulence intensity (fluctuating component of the mainstreamvelocity/mean streamwise velocity) was increaseU to 12 percent and the re-

sult is shown in tigure 20. Also shorn in figure 2U is the previous generic

result for a turbulence intensity of 4 percent (from fig. 19). The exper-imental data are also from the previous figure. No attempt was made to runthe experimental test with higher turbulence. The results in figure 20 show

that the higher intensity (dasned lines) lead to greater temperature mixingnear the shroud, toat is, normalized radius of about 0.85, over the tneta

range of b to 15 degrees (figs. 20(c) to (f)). At the same radial position,no additional mixing is predicted witri the higher intensity near the fanlobe centerline (figs. 20(a) and (b)). A slight shift is observed in the

radial position of the maximum temperature, as seen in figures 20(d)to (f). The higher turoulence intensity moves the peak to a lower radialposition.

At the intermediate radial positions, the larger intensity value yieldstemperature ratios which are closer to the ideally mixed value of 0.5. Andfinally, at the lower radial locations (-0.2), the ditference in temperatureratio results from the two intensities begin to approach zero.

W NCLUUING kLMAkKS

A finite dirterence pruceaure has been used to compute the mixing forthree different lobe geometries which were experimentally tested. The toi-lowing conclusions were made based on a comparison of the analyses aria thedata:

1. The dominant mechanisms within modern turbofan turced mixers is as-

sociatea with the pressure driven secondary flows arising within the lobe

region upstream of the mixer- and tneir development in the mixing region.

2. Secondary flow generation at the lobe exit is caused by three prin-

ciple mechanisms, that is, vorticity due to turning (tlap vorticity), pas-sage vorticity, and norsesnoe vorticity.

3. Tne flap and passage vorticity appear to be adequate flow mechanisms

in initial flow conditions to obtain goon agreement with experimental mixerexit profiles.

4. A generic representation of the vorticity mechanisms yield goodagreement between analyses anu experiments for three different lobe geome-tries and for two difterent temperature ratius of operation.

b. Tnere is a crrtarcal need to outarn detailed experimental infornd-tion regarding the nature of the secondary flows generated within the loberegion in order to pruperly analyze mixer nuzzle. Empirical data relatinglobe geometry ettects un seconaary flow generation is nighly desirable as adesign standpoint.

kEFEkENt;ES

i. Kumar, A. P. and Chamberlin, k., "Scale Model Performance Test Investi-gation of Exhaust System Mixers for an Energy Etticient Engine (E3)Propulsion System," AIAA Paper 60-02[9, Jan. 1980.

2. Snumpert, P. K., "„n Experimental Muuel Investryation of Turbufan EngineInternal Exhaust bas Mixer Configurations,” AIAA Paper 50-UZZ8, jan.i9su.

10

3. Kozlowski, H. and Kraft, G., "Experimental Evaluatiun of Exhaust Mixersfor an Energy Efficient Engine," A1AA Paper 8U- 1088, June 1980.

4. Birch, S. F., Paynter, G. C., Spalding, U. B., and Tatclrell, U. G.,"Numerical Modeling of Three-Uimonsional Flows in Turbofan Engine Ex-haust Nozzles," Juurrrd l of Al rLrdt t, Vol.r 15, Aug. 1978, pp. 469 -1490.

5. Bowditch, D. N., McNally, W. nde-son, B. H., AdamcZyk, J. J., and

Sockol, P. M., "Computational Fluid Mechanics of Internal Flow," Aero-propulsion 1979, NASA CP-"[092, 1979, pp. 181-?JU.

6/ Povinelli, L. A., Anderson, B. 11 ., and Gerstenmaier, W., "Computation of

Three-Dimensional Flow in Turbotan Mixers and Comparison with Experi-mental Data," AIAA Paper 60-0227, Jan. 1980.

7. Kreskovsky, J. P., Briley, W. it., and MLDunalrt H., "Development of aMethod for Computing Three-Uimensiondl Subsonic Turbulent Flows in Tur-bo`an Lote Mixers," Scientific Research associates, Inc., 61astonuury,LT., k79i-:00006-F, Nov, 1979.

8. Briley, W. R.f, and McDonald, H., "Analysis and Computation of ViscousSubsonic Primary anu Secunuary Flows," Can utati,^nal r luid DynamicsConference, American Institute of Aeronatitics and Astronautics, nc.,19/9, AAA Lonf. Paper /9-14bj, pp. 74-88.

9. Patterson, k., and Werle, ^1. J., "lurbotan Forced M yer Flow Field,"United fechnoloyres Research Ccriter, East Hdrtford, LT, UTkL /R79 -912924 - [4, July i9/9.

10. Anderson, B. h., PUvine111, L. A., ,ind Gersteni^aier, W. G., "Int1uence

of Pressure Driven Sed:undary Flaws use the behavior of lurbotan Forced

Mixer's," AIAA Paper W-1196, July 1966.11. Launder, B. E., anU Spaloiny, U. ti., "The numerical Computation of Tur-

bulent Flows," Com titer McthoUS in A 1 rea Mechanics and En ineerin ,Vol. 3, Mar. 19/4, pp. 209- Z89.

12. Uuozinski, T. J., anu ►,rause, L. N., "flow-Direction iledsurement withfixed - P , sition Prooes," NASA T .1 X-1904, iyb^.

13. Anderson, U. L., "Finite - Urtterenct- ;,Olutrun for Turbulent Swirling^umpressiule Flow rn Axrsyr,unetrt(_ UUCt, with Strut S," NASA LR-23b5,1974.

14. Anderson, 0. L., "Tur-L)otan FUrceii vIxer NullIe Internal Flowtreldi,"NASk, i.ontraCtO F'S Rep,rr't tU tie putrlrShVU, lyol.

%0

l,••2

rj 6

^ 1

V

FtL

m ^

6

N

N

O

N

Ik

i

4

OC

.l

s

m

N

0

t

c pE

u ^,

O

Q ^V

w ,^

c0

1

1 \.

IEIE

0

OZ^ N

m u^

za =

^ u

J N

` K

J

STATION N0. 1 (1001 EXITI

SNP.OUD _ I e

LORI h .

RUG

1121I IMIXf R I XITI

1^

1

I qure i. I Ic, anquler t^ rdtr^.

r

t

SHROUD SURFACE

I /i

EXPERIMENTAL

^ I LOBE SHAPE

VFµ VCORE

/ COMPUTATIONALLOBE SHAPE

PLUC SURFACE

Figure 5. - Transverse computational segment

SHROUD SURFACESHROUD SURFACE

f-I 'fff

1

PLUG SURFACEPLUG SURFACE

UI CONFIGURATION 12B lbl CONFI SURATION 12C

Figure 6. - Measured secondary velocity vectors, lobe exit station.

CORE FLOW

r

FAN FLOW

RADIAL FLOW

is) BASIC FLOW TURFING

ILL RADIAL FLOW ORONTATION.

Figure T. - Secondary flow generation, turning fMapl vorticity.

W-01107

pnl FAN PASSAGE_

V

tDl CORE PASSAGE.

Figure & - Secondary flow generation, horseshoe warts&

r- 108E)FIT ICES

CENTERDODY

Figure a, - Secondary flow generation, passage w bx.

Figure lo. - lull I tehdVi or in core Ilam

BASFIIfi IFLAP ORTICITY^

PLUG SURFAct

P1, 1141 ACI

RIC IFIAP PII1S PASSArj VORTICITI trC'► Nu t t, IL Initial xundary flow held, lobe exit plane.

1:9 Mixer, TeT F . q 1461

Z^O

G

c

to

X

y8x

v

V

1. •', • ^ '^• Y•

. . . . . . . .. . ..... 9 I

cz :•

< . . . • . . . . . . . I

N

Jd

i

x

^

WW

w

d

z2O

t

0

f

i

t

......^\

+^.^^^, i 1 ^^'• SHROUD SURFACE

1 ^

(e) MIXER EXIT. STATION NO. 2L

Figure 12. - Concluded

NNJQ

Q

ccAD

`CI

N

E

0c6Vv

A

ta

0000N^O

OU

dU

2 ^t

axT

I INFLOW

CONDITIONS

O EXPERIMENTAL

DATA

GENERICI — BASELINE

1.OF— --- IDEAL

.B

^ a —L-- I I I I

(a) THETA • Q 0.C; L0toe Br

6 1 I I I I1DI THETA • 10.

0 1.0—

.6^0 .2 .4 .6 .8 1.0

NORMALIZED DISTANCE

Icl THETA • 6.4

Figure 14. - Comparison of computed temper-ature profiles with experimental data 12BMixer. TF1T C • 0.14 Station 21.

}}

...rol

vi

. c

.Cur

to CD

cywZy3 Y

lip! I '^̀ 1 .o n

W

o IIIit ^ ZM

AL J^—^a o .o .0

L_ ^l0 0 .o

oo ^e

i3alu '011" RfUVa3dw31 lviol

SHROUD SURFACE

PLUG SURFACE

(a) LOBE EXIT. STATION NO. L

....... ,\,, SHROUD SURFACE

.. ..Hf f+r, ..o^.

PLUG SURFACE^

(b( STATION N0. 8

Fqure 14 Computed secondary velocity vectors,

.....^frrttrrr '•. ••.; ••

.....rti11 '

r

SHROUD SURFACE

PLUG SAFA^

Icl STATION Na 11

............

.......n.rrr .

.......Htftr^SHROUD SURFACE

ldl STATION NO. ll.

Fqurn 16. - Contlnur4

o

aIA ,d^ fi t

d

C C

o '^a

°U W O O

)DI

__D--° so 'd!c e a o s

AM I 1 ours 3MfllVb3dW31 lViol

N0KSN

N

N l^

bt(tl

1

f;

§^ _

2 &df o ° § p

§1§ f

§^K o g o § § §.° ^^ n ^ -

^ |^^ ^ n ^\.

^------^ I^

^2mn@&n3d"r&

^

2$$ n ^'§K ^^

f

. X^

\

.

^,^^/] f

^° ^^ k \ ^^ ^^ kI

\| \j §^

^

1 k

^$ ^ ^ nK

^o ^0- _7 q IL

63

co

^

/ L

= } i

ky#

. . ^ . .

Ir Al , IV!!o

n § $§$

^^o

_^^- n ^I'

^

•- ^ ^ of ^ • . ^

13bm ton lb & n Nw£ IV 2!

a

O

Io co

go

O I \\ \\

I

I ®II /\ /p. \ 1 ~

1

I

\I

J l ^o w a ^ o ^ e ^°

3111 111 'ours Unlra3dw 31 rrioi

s

cQ

4

N Y ° °

Z0 3 S2 w I d co

0 I I i / I/ I ^

I I,

Ha lo 'Dora lb ruYb ldWb riiO,

TURBULENTINTENSITY

U"UREFO EXPERIMENTAL

10^ DATA

0.04

8

—_ —

0.12

i

6o ^^

O

W 4 0! q

(a) THETA • 0.0.

c B^

6L

q `^ 00

0 .2 6 B 1.0NORMALIZED DISTANCE

lb) THETA • 10.

Figure 20. - Comparison of computed and ex-perimenta' temperature drofiles. 128 Mixer,TFTC•0.40.

1 ^ W ^

Z^ _^ ap

^^z^^ O a d l dO

W o d d <OoW

t

I

HO J

'-

oI 2

®i

(1NI '

^ IO .,O

l I I i.p V O aD

I I.D VO

M ill 'ollV H ]NnlV836w111Mlol

0

0 0' t ^,

z^wrN

I

d

' Pnc l-"' oc < ^ ^ Nnp

I 'N=

o w ,c v o oe .c d°

~ t3Nlil 'ollvm lbili i3dMI 3Vl0l