A93 *47397 AIAA-93-349541
NAVIER-STOKES PREDICTION OF A DELTA WING IN ROLL WITH VORTEX BREAKDOWN
Neal hI Chaderjian* and Lewis B. ScliiEt
N4S.A .4mcs Research Center. Moffett Field. C.4 994035
Abstract
The three-dimensional, Reynolds-averaged, Navier-
Stolces (RAKS) equations are used to numerically simu1at)e
vortical flow about a 65 degree sweep delta wing. Sub-
sonic turbulent flow computations are presented for this
dclta wing at 30 degrees angle of attack and static roll an-
glcs up to 4% degrees. This work is part of an on going
effort to validate the R.4NS approach for predicting high-
incidence vortical flows. with the eventual application to
wing rock. The flow is unsteady and includes spiral-type
vortex breakdown. The breakdown positions, mean surface
pressures, rolling moments, normal forces, and streamwise
ccnter-of-pressure locations compare reasonably well with
experiment. In some cases. the primary vortex suction
peaks are significantly underpredicted due to grid coarse-
ncss. Nevert,heless. the computations are able to predict
the sanit. nonlinear variation of rolling moment with roll
angle that appcared in the experiment,. This nonlinear-
ity includes regions of local static roll imtability, which is
attributed to vortex breakdown.
Introduction
The High-speed Civil Transport (HSCT), National
Aerospace Plane (NASP), and modern tactical fighters
all operate at high-angle-of-attack flight conditions. The
HSCT and NASP have slender bodies and highly swept
wings which are designed for high speed flight. While land-
ing. their low airspeed will require a large angle of attack
to maintain a proper glide slope. Fighters also operate at
high anglcs of attack to achieve greater maneuveraljility
and agility performance. If the angle of attack is suffi-
cient,l~- high. large regions of vortical flow can form and
* Research Scientist, Computational Aerosciences
Branch. Senior Member AIAA.
tSpecial Assistant for High Alpha Technology, Fluid
Dynamics Division. Associate Fellow AIAA.
Copyright @ 1993 by the American Institute of Aeronau-
tics and Astronautics, Inc. N o copyright is asserted in the
United States under Title 17, U.S. Code. The U.S. Govern-
ment has a royalty-free license to exercise all rights under
the copyright claimed herein for Governmental purposes.
All other rights are reserved by the copyright owner.
interact with the aircraft's nmt,inn. rc>sulting ill wing r o ~ l i . ~
a sustained periodic motion in roll and yaw.
The high-arigle-of-attack flight regilric is 7-er)- com-
plex and may inclurle streamwise and crcwflon- parati ti oil.
vortex brealidon-11. autl vortex asymmetry. These pl1e1ion1-
ena are inherently nonlinear. often time dependent. and
can be very sensitire to small disturbancw in tlir, flon-.
Experimental investigations aimed at understanding wing
rock have been carried out on delta-wing geometries ~vitli
a single-degree-of-freedom in roll. These simplified geome-
tries exhibit the pertinent flow physics associated with
wing rock. Experimental data typically consist of flow 3.i-
sualization and time-dependent forces and ~nonients.~-';
More recentl>-. Hanff and his colleag~ies7-'I 11aw carried
out higll-qualit5 experimental investigations of ii large+ scale delta-wing/hody model undergoing large-aniplitutlc
roll motions, and include measllrements of time-dependent
surface pressures. ,-2rena ant1 Xelsonl' haw also m e a s ~ ~ r r d
time-dependent surface pressures on a highly swept delta
wing. These experiments provide additional dctail to help
understand wing rock; and provide a timely opportlmit~-
to validate computational fluid dynamic (CFD) ~llethuds.
Computational fluid dynamics is just beginning to
be utilized to investigate wing rock anti can provide more
flow-field information than is otherwise possible with ex-
periment alone. However, these computations require large
amounts of computer time. In order to reduce the com-
putational cost, some investigators hare simplified the
problem by using the conical and three-dimensional Eu- ler equations.13-'I' The conical form reth~ces the t,lirex-
dimensional physical flow int,o a txvo-dililel-isioiial coiiipll-
tation using similarit!- principles. The conical approach
can not produce vortex breakdown which usually accompa-
nies wing rock. The inviscid approximation further limits
the flow sin~ulation to a single primary \wries that fornii
at sharp leading edges. Thc inviscid approxim;rtioli call-
not account for the vortices that are linown to forni on
smooth surfaces. e.g.. forebody. secondary. ant1 tcrtiar!.
vortices. Moreover, the primary vortex has a question-
able core strength which can significantly affect the vortex
breakdown mechanism and location. Murman" has re-
cently shown that there are significant viscous effects on
the F-18 leading edge extension JLEX), even though it ha5
a sharp leading edge. An improvement in the prediction of
the LEX suction peak was obtained by refining the grid in the boundary layer. The approach adopted here is t,o solve
the t1lrc.c.-tlirnensi(j11i-tl. Reynolds-averaged. Navier-Stokcs
IRASS) cqliations. Even though this requires more com- p~itcr time. tlic RXSS equations contain all of the relevant
flow phj-sics that are applicable to a complete aircraft ge- onictry.
Thcre i5 an effort at S.4S.S -4nles Research Center to validate the R l l N S approach to simulate high-incidence vortical flows about delta wings. with the eventual applica- tion to wing rock. Chaderjian18 recently used the Kavier-
Stokes Simulation code (SSS) to simulate time-dependent. \-ortical flow about R 65' sweep delta wing undergoing large-amplitudt= roll oscillations. This case was investi-
gated esperimentally by Hanff and his colleague^.^-'^ Fig- ure 1 summarizes the computational and experimental rolling niornent coefficients obtained at 15 degrees angle of attack, and static roll angles up through 42 degrees. .?llso shown in Fig. 1 is a large amplitude (40 degrees), high rate of roll (7 Hz), forced dynamic motion. .4t this
anglr of attack, no vortex breakdown was observed in the experiment or the computations. The static and dynamic
computatio~ls are in excellent agreement with the exper-
imental results. In addition, computed surface pressures,
~iornlal forces, and streamwise center-of-pressure locations were also in excellent agreement with the experiment.
,,L,
.04 + Static Exp, Ref. 9 o Static CFD, NSS
.03 - Dyn Exp. Ref. 9
.02
- .o 1 U 0
-.O 1 -.02 -.03 -.04 -.05
-50 -40 -30 -20 -10 0 10 20 30 40 50
Fig. 1 Static and dynamic rolling moment coefficients
from Ref. 18. = 0.27, a = 15"> Re = 3.67 million,
and k = 0.14 for the dynamic case.
This paper builds upon the previous study and in- yestigates flows at 30 degrees angle of attack. xhere vortex 1~rralidow-11 does occur. Experimental1O static and dynamic rolling moment coefficients obtained at 30 degrees angle of attacli are shown in Fig. 2. Kotice the nonlinear behavior
of the static rolling moment coefficient for roll angles of -10" < Q < 10". This is very different from the near linear behavior seen at 15 degrees angle of attack. -4lso notice
that the dynamic case does not follow the static rolling
moment curve. This discrepancy has been attributed''
to time lags occurring in the vortex breakdown locations. Thus. the goal of this study is to verify that the RAXS approach can predict the nonlinearity of the static rolling
moment curve shown in Fig. 2. The effects of large time
lags observed in the dj-na~nic case mill be inrwtigated conl- pntat,ionall?- at a later timc.
Static
Fig. 2 Static and dynamic experimental rolling moment
coefficients from Ref. 10. l f m = 0.27, a = 30°, Rt= = 3.67
million, and k = 0.2 for the dynamic case.
A desc~iption of the governing equations. turbulence
model. numerical algorithm, colnputational grid. and nu-
merical boundary conditions is given below. This is fol-
lowed by a discussion of the compl~tational reiults and
concluding remarks.
Numerical Approach
Governing Equations
For the high Reynolds number flows considered in this paper, the thin-layer approximation'g to the RA4NS equations is employed together with body-fitted curvilin- ear coordinates. Assuming a general coordinate transfor-
mation from physical space (x,y,z) to computational space ([,v,C), and that viscous terms are only important in the body-normal (<-coordinate) direction, the governing equa- tions can be expressed in the following strong-conservation-
law form:''
where Q is the vector of conserved dependent variahlcs, E. A A
F, and G' are the inviscid flux rectors in the (, 7 ; itnd ( A
directions, respectively. and S is the thin-layer viscous flus vector. These equations have been ~lo~idilrlensioilalizetl by
C , the wing root chord, arid (L,; the free5tream speed of
sound. The coordinate transformation metrics allow for
a general moving and deforming grid system. The perfect gas law, Sutherland's viscosity law, and a turbulence lnodel
completes the RANS system of equations. The turbulence
~rlodcl will 1jv clrscril~ctl in the nest section. Further clct,ails
aljol~t the o;nverning equations can be fo1111c1 in Ref. 19.
Turbulence Model
The higli-incidence. higli-Reynolds-nunlhrr flows con-
sidered in this paper have significant crossflow sc3para-
tion on t l ~ r leeward side of the delta wing and there-
fore rcquirei a suitable t~trbulence model. The Bi~ldwili-
Lonlas algebraic nod el^^ together with the Degani-Schiff
~ n o d i f i c a t i o i l ~ ~ - ~ ~ is a n efficient isotropic eddy viscosity
rrlodel that properly accounts for the crossflow separation
in the prwrncc of strong vorticcs above the wing.
T h c Baldwin-Lornax model often chooses a length
scale associated with the vortical flow structure outside
the viscous boundary layer. This results in a n eddy vis-
cosity which can he one or two orders of magnitude too
large. and eff'ectively suppresses all but the most dominant
vortex structure. The modified model of Degani and Schiff
restricts the choice of length scale to the boundary layer
regioii in a ratiorlal ma~lner . and therefore gives correct
values for the eddy viscosity.
A11 viscous c o m p ~ ~ t a t i o n s presented in this paper as-
silnlr: f d l y tiirhident flow l~cginr~ing at the apex of the delta
wil~g. i.e.. transitiorial effects are ignored.
Numerical Algorithm
There is a choice of two implicit, approxinlately-
fx torcd . central clifferencr algorithms in the KSS code
to integrate Ey. (1). T h e first algorithnl is t l ~ e Beanl-
Warming a l g ~ r i t h m , ' ~ and the second is a diagonal ver-
sion of thc Beam-LVarrrling algorithni due to Pullianl and
C I i a u ~ s e c . ~ ~ Both algorithms use second-order-accurate
central differencing throughout, with added artificial dissi-
pation t o darnp out high frequency errors. Explicit artifi-
cial dis~ipat ion terms consist of blended second-order and
fourth-order differences. The terms have been modified to
reduce thc dissipation within the bolmdary laJ-er.Z"he
Beam-JYarnling algorithm uses second-order implicit arti-
ficial dissipation and thereforcx requires the solution of a
5 x 5 b l ~ c l i tridiagonal system of equations. The diagonal
algorithm uses an iniplicit dissipation that is idelltical to
the explicit seco~id/fortl~-order dissipation. and therefore
requires the solution of scalar pent adiagonal equations.
The diagonal algorithm requires fewer computational op-
erations per time s t ~ p than the Beam-Warming algorithm.
and is therefore more computationally efficient. However.
Levyz6 has shown that the diagonal algorithm has a n in-
herent directional hins. If the angle of attack is too large.
this ma?- lead to a no~iphysical vortex asymmetry in the
flow. The choice of algorithm is case dependent. Both
algoritlims use Eulcr implicit timc tliff~sc~~icillg. n-hicli pro-
vides first-order tinie ;LCCII~B(-J- . Both algcritl~nis in the
XSS code treat the viscous tcrnis in ail csplicit maliner.
The KSS codc can also treat conlplrs gconictries.
such as a complete F-16-4 fighter aircraft." Thi.; is accom-
plished b y using a zonal grid approacl~. The zonal gritl ca-
pal~ility will facilitate the cveutual application of this cmdr
to a maneuvering aircraft e q x x i e ~ ~ c i n g \x-ing rock. -4 zollal
grid approach can also h e used to reduct, ill-core inelnosy
requirements on computers with ext,endetl s~conclary rntm1-
ory. A brief sulnnlary of the zonal interface procecl~~rr is
given in the numerical boundary condition sc:ct,icm.
Computational Grid
Extensive wind-tunnel investigations were carried
out by Hanff and his colleague^^-^^ for the 65' sweep delta
wing shoxvn in Fig. 3. This wing is symmetric frolli tol)
t o bottom and side to side. with double bevels running
along all its edges. A fuselage/sting houses instrumen-
tation and pro\-ides for a olie-tlcsrc-1.-of-frcecloni in roll.
Time-dependent surfact. presslxrei arc, o h t a i ~ ~ e t l acrois the
span of the wing at thr. 73% rnot chord positiol~.
b = 22.83 in.
Fig. 3 Thrw-vien dra~virig of the n-intl-tunnel model.
A three-tlil~iensiorinl hyperl~olic grit1 gcmeratos'%\-as
used to gcnerate a spherical gritl for this clrlta n-inq ge-
ometry. A perspecti\-e vicv of the viing surface gricl ant1 a
portion of tlic s t i~ ig is i11on.n in Fig. 4. A4 porrioli of the vis-
cous gricl clustering liormal t o thc n-ing at tllcl trailing edge
is also showli ill this figure. The ~l111c.rical axis t.steuds up-
wind from the wing apes. The far-field l~m~~lt lar ic . ; arc not
shown in the figure, but extend two root chord lengths lip-
wind and downwind of the wing l~ody , and five root chord
lengths in the body-normal direction. The grid consists of
67 points in the streamwise direction, 208 points in the cir-
cumferential direction, and 49 points in the body-normal
direction. totaling about 700,000 grid points. In order to
resolve the leeward-side vortices, there are more grid points
on the leeward side of the wing (in the circumferential di-
rection) than on the windward side.
Fig. 3 Perspective view of the computational grid.
There are two Cray C-90 supercomputers and one
Cray Y-YIP superco~nputer available at NASA Anles Re-
search Center. Each of these machines is configured with
different internal memory and extended secondary mem-
ory (SSD) capacities. In order to take full advantage of
each machine, the single delta wing grid was split into four
zones in the streamwise direction. This reduced the inter-
nal memory requirements so that each machine could be
fully utilized.
Numerical Boundary Conditions
The no-slip condition (zero velocity relative to a solid
surface) is imposed on the wing and fuselage/sting surfaces
while density and pressure are found by extrapolation.
The total energy per unit volume is then computed from
the perfect gas law. Uniform flow is imposed at the far
field. while a zero-gradient condition is used at the outflow
boundary. Flow variables.are averaged across the wake
cut, which extends from the wing trailing edge downwind
to the outflow boundary. Boundary conditions are imposed
on the spherical axis by averaging flow variables that are
one point off the axis, and located circumferentially around
the axis.
Zonal boundary conditions are updated sequentially
with the most recent data available. .Adjacent zones over-
lap by three grid cells so that centrally differenced fourth-
order dissipation can be used across zonal bourldaries.
Since the zonal grids were originally constructed from a
single grid, zonal interfaces have coincident surfaces with
identical grid points. Data transfer from one grid to an-
other is accomplished by direct injection. A more complete
description of the zonal interface bou~iclary conditions is
given in Ref. 27.
Results
The time-dependent, RAYS equations are ilu~rit~ri-
cally integrated to simulate viscous flow al~out the 63 dc-
gree sweep delta wing shown in Fig. 3. Each computa-
tion had a fre~streain Mach nuinher JI, = 0.27. angle
of attack a = 30°, and Reynolds numbrr l~a.;ed on thr
wing root chord Re=3.67 million. These flow conditions
correspond to experimental data provided by Hanff and
his colleagues.7-" Six static (fixed) roll angles are inves-
tigated, i.e., q3 = 0. -5, -7, -14, -28, and -42 degrees. The
direction of positive roll and the coordinate axis orienta-
tion are shown in Fig. 4. -4 positive roll angle corrcsponcli;
to the right wing down and the left wing up, from a pi-
lot's perspective. The flow evolution is solved ill a time-
accurate manner with a constant nondimensional time strp
of AT = 0.005. This value provided good time-accurntc rc-
sults in Ref. 15.
Initially the diagonal algorithm was used to compute
the flow about the delta wing at zero roll angle. This
resulted in a vortex asymmetry that was not present in
the experiment. Levyz6 has shown analytically that there
is a directional bias in the diagonal algoritlm~. Furthcr.
he has also demonstrated computationally that this bias
can lead to spurious vortex asymmetry at sufficirntly high
angles of attack. The Beam-Warming algorithm prod~lcecl
a symmetric vortical flow, consistent with the experiinent.
and has therefore been used for all of the cases presented
below.
Instantaneous particle traces are used to visualizr thc
leeward side vortices. A planform view of these vortices are
shown in Fig. 5 for GJ = 0 and -5 clegrecs of roll. The vortes
cores are indicated in the figure. E ~ p e r i m e n t a l ~ - ~ ~ surface
presure data is available at the 75% root chord location.
which is also indicated in Fig. 5. .4t zero roll, a spiral-
type vortex breakdo~vil is indicated by a rapid espasssiosl
of the vortex diameter and the ..corliscreiv" path of thc
vortex core. At d = -5 degrees, an asyrrimetric vortex
breakdown occurs. In this case the left wing is down, re-
sulting in a reduction of the effective leading-edge sweep.
while the effective leading-edge sweep of the right ning is
increased. Consequently thc left x-ilig I-ortcs is stroi1gc.r.
than the right one and breaks down sooner. It is intewst-
ing to note that even though there is vortes asynm~etr!-.
the computed rolling rnoment for this roll angle is nearly
zero.
It can be difficult to precisely quantify the location of
vortex breakdown, because there is no univrrsally accepted
(b) 4 - -5O Fig. 5 1,-isualization of leeward-side vortices with install-
taneous particle traces. -11, = 0.27. a = 30'. Re = 3.67
n.~illion.
definition for breakdown. Breakdown is usually visualized
in an rxperirrlent by natural condensation occurring within
the vortices. or by injecting smoke into the vortex cores.
The shear layer lcaving the leading edge of thc delta \\-i~lg
rolls up and entrains the smoke. rendering the vortex visi-
blr. -4 b u l ~ l j l c - t j - ~ ~ 11rc.akdon-11 i i irrinir~cli;\ti-l~- c)i)~-ioui I ) , , -
c;lilse of the tlisprriior~ of i lnu l r tlon-wtic~;rlt~ of'Ij~c~akc!i~:\-t~.
The core v c l o ~ i t ~ will also go to zero ir t l j r (~ ; i l ;~ l i~n .~~ . C ' O ~ I I ~
putationally. n l~~~l j l> le - ty~)c . Lri~alirlr~n-11 , . ; l l r i~lw) 1i1, c,;t-il>- . .
idrntificd by siniilar tw1111i(~uc< ~15111; l l l ~ t i l l l r ; t l l ( ~ o i 1 . I ) ~ I t i -
cle trac.c2s a n d e s a n i i ~ ~ a t i o n of t11~ core ~-r ' lo( . i t~- .
Figure G shows tllc axial location of the esljc~rillici~t;rl
and computational vortex brealidow11 p o s i t i o ~ ~ s oii tlit. lcft
wing at variolw roll angles. T l lv : i i l a l o ~ o ~ ~ ~ in fo~i l~ i \ t i t )~ l
for the right wi11g can be f11jtaitte:tl 11)- l i , f l ( ~ , t i ~ ~ ~ thi> (lata
a110iit d = 0'. The c o ~ n p a r i w i ~ l>rtn-cen t11r r ~ ~ ~ t i i ~ ) i ~ r a t i i ) t ~ ; i l
and experinlcutal reslllts i i good n-hcli l ~ l t ~ a l i i l o i ~ x occnus
~lpstrearli of thc n1i~11~1ti)rcl pi15itio11. ;III(I i - on1~- f1~i;~lit:i t i~-c
x l i m l ~ r ~ ~ a l i ~ l o ~ w oc.c,i~ri 011 I aft 1 i , 111 of ~ l t c \xvi~:c. . .
It should be noted t h a t the l~rc~;rlitlon~i~ pc~-lt~c>:i i- \-('rj-
sensitive to roll angle over tlir range - 10" ( o 5 3'. n l ~ e r c
the breakdown position moves rnpitll!- froin a ryqicm \-cry
ncar the v ing apcx (nose) t o thi' wing ti.iiiling etl;c. T11~t.r
1.2 1 Left Wing
Fig. 6 Cornpalison of cor~~~)l l tecl and r q , c ~ i n i c ~ ~ ~ t a l ~ o r r i ~
hrealitlown location for differmt btatic rill1 ailglc~s. .\I, =
0.2'7. a = 30". Rc. = 3.67 million.
.. ' N C = 0.75 2,5 ( a ) @ = O 9
I
.~.. Exp. Ref. 10 . . , . .-. ,
2.0 , , . Max. NSS , .. 8 ,
.:' , ',., Mcm. NSS ,:' . o '.., I ',
.... A I , n ', ... Min.NSS ,.:" o o ' ~
iq alio :: riiljitl clin~lgc. of s loIx~ in the. experimental ljreak-
(1i1m1 c i ~ r ~ - ( > :I~;II c) = -.To. v:lli(:l~ is &I e\-irlcnt in tll t
c(>1il1iilt~~iO11;d cilrr-c.
At zcro roll angle. the c o m p u t ~ t l mrali pressiu.ci
agree only cli~:ilitativc.ly wit11 the eslii~ri:neut:rl 1-a1ur.q. Tlita
conlputat,io~ls do not picli 1-11) tliv large mca11 ilictic~~i peal;
of thc grirnnrj- vortcs. The grid qhon-11 ill Fig. 4. ant1 iisecl
liert. was constriicted to resolvc ~ o r t c x streiigt,lls that rc-
sult in si~ction pea l s ncnr 1.5. Tbi.; g;r\-r3 1.cl.y acci1r;rtc
results a t 15 tlegrecs angle cif att;~cli." Howe~-r'r for t h ~
present case. thc rr~asilrnull trnllmral C',, ii alxn1t 2.2. :I fi11c.r grid is nreded to sustain tlie very lligli temporal SIIC-
tion pealis recjiiiretl t o captin-? the 11l(~ali presllrei. Tlli'
same t r m d can 1)e noted on the wing's left sit!e for 0 = -5"
ant! the wing's right sidc for o = - 7". Each of tlirse c~;c,c~s
correspond to a side in wliicli t , l~erc are largc meail C',,. On the uther lland. x-hm thc suction pralis arca smallrr.
i.c.. the right side fur o = -3'. tlia left side for o = -7". and l~o t l i sirlei f i r L? = -14. -2s. iilltl - 42 t1egrcc.s. the
aejreerr~ent is good. For roll ariglcs of 7 . -14. '7s. ant1 -42 degrees. tlie flat portion of the llleilll CI, c1lrvt.s arc2 cap-
tured. The flat surface pressure regious are due to the:
.i comparison of mean computation;d arid rsperi-
111ental rolling nlonitmt rocfflcients ( C ' ( ) are slion-n in Fig. 8.
The coniputcd rcsults for positive roll angles are ohtair~etl
by rc'flc~ctiol~ of tlir rcsillts f r u ~ n negative roll a~lgles. In
genc~al . co~nputed C'! are in clualit,ativi~ agreement with
the esperirnent and exhibit the same nonlinear behavior
for - 10' 5 o < 10'. The corrlputations indicate that there
are tlnec stable static trim points presmt, a t o = -13. 0, anti 1.5 d e g r ~ e s . There are alio rr~gions of posirivc slol)c;
whcrc tlics wing is statically unstable. Even thnilgll thc
q i r l is crx+l+l. tlir R.AXS ecltlations arc nl,le t o moclcl the
l)h~.hical ~ i i e c l ~ i ~ ~ ~ i s ~ i ~ s r ( , ~ p o n ~ i l ~ l r for t11<> ncmli~i~~al i ty .
Fig. S Cornparison of rlieaii co~nputecl and eslwrinle~ltal
rolling n~onierlt coc,fficients for diffcsent static roll angles.
Jrl, = 0.27, (i = 30". Re = 3.67 r~rillion.
Recirll that at 15 tlegrwi angle of attack. Fig. 1. tlie
rollirig moment varies alniost lil~early wit,li roll aiiglc arid is
.itatically stable over the entire roll angle range of --42" < o < 42". 111 this case there is no vortex br.c~alitlown. iil~tl t h ~
static stability can be explaincJ in the following manner.
For positive roll angles. thc low right wing has a rcduced
rffectivt: bvwt'p xvllile tlrt high lcft wing has a n increasrd
eff'ertix-e .sn-pep. This causcs the right wing vortcs to 11e
itrongrr tliari the left wing vortex. rcsdt ing ill a nrgative
l s t a l ~ i l i ~ i r l ~ ) rolling ~noincnt . Similar arguriients can h r
give11 for r~cgativc, roll ;triglcs. A more cletailcd description
call h~ found in Ref. 18.
The situation is very different a t 30 degrees angle of
attack. due to vortex breakdown. For small positive roll
angles (less than 10 degrees) the right wing should have
a stronger x-nrtex than the left wing due t o sweep effects.
This would result in a negative (stabilizing) rolling mo-
nlent. However. tile right wing vortex is observed com-
putationally to break down closer to the delta-wing apex
than the lcft wing. see Fig. 6. This causes a loss of lift
on the right wing, resulting in a positive (unstable) rolling
r n o ~ n e ~ i t . This accounts for the positive slope of the rolling
The mean coinputational rlormal force cot$ficir~its.
C,y, are coniparccl -.\-it11 cxperirilcmtal v;ill~t~; : I n o l i ~ i l o i ~ ~
the roll angle range, -42" 5 o 5 1'7". ill Fig. 9 , Tlie
;malogous cmmpariioli for thc mrari s t r e i i ~ ~ l ~ v i w r.1 n t ~ r - o f -
p r e s a ~ c location. -y,Ij. is 511on.11 in Fig. 10. Oi-(~rall . tlr<>
comparisoris are good. Tile computc~d C'.\ c.orlll);irc
c ~ p t i o ~ ~ a l l y well with esperirr~ent for large roll ; in~les . For
low roll angles; CV is undcrprcdictetl tliie to ~irrrc~wlr-ctl
vortex suction peaks. The streamwise ccnter-of-l~rcssiirc
locations are rrlativelj- constant and agrcr wirli thc, c.xl,<'l-
imcntal values within 3%
Fig. 9 Co~riparison of mean cornputetl ant1 cspr,rimeli-
ta l normal force coefficiel~ts for different static roll angles.
r\f, = 0.27. or = 30". Re = 3.67 million.
T h e Beam-Warming algorithrrl. with a block periodic
solver in the circumferential direction, requires 26 p s l g r i d
point/timc step on a Cray C O O superconlputcr. Each
static case was solved in a time-accurate nlannt3r. Aipprox-
imately 8000 time steps were required to insure the initial
start-up transients were 110 longer p r r smt . This rcqnirctl
46 hours of C-90 compilt?r time.
Fi5. 10 C o m ~ > ; \ r i s o ~ ~ of mean computrd and ~ s p c r i ~ n e n t a l
streamwise cc.11ter-of-pressi~rf: locations for different static
roll ariglcs. -11, = 0.27. cr = 30'. Re = 3.67 million.
Conclusions
C'omln~tations fi)r a 6.5' sn-eep clc'lta wing at 30 tle-
grer-s anglr of a t t ~ c l i and 3.67 million Reynolds n u m h r ner r ~ x e ~ ~ i t ( d i ~ ~ i n g the time-dependent . three-dirrlellsional
RrlTS ecji~iitiol~s. Static (fixed) roll angles were invcs-
tigated up tlirollgli 42 drgrees of roll. The flon- field
was fount1 to 11e ~ ~ n s t e a d y at a11 roll angles. and included
;I sI>iral-tYIw vortex hrcakdown. The vortex breakdown
position. nlc~ari surface pressures. rolling moment coefi-
cicmts. normal force cocfficirnts. ant1 strearnwise center-of-
1)reisurc locations coniparecl reasonal~1~- well with rsprr i -
~ i lcn t . In some casc>s. the corilputational and experimental
I-orri1j;irisorii wcle only qldi tat ive1~- good. Finer grids are ~ieet l (d t o hr t tcr rc-solve. tlir c~ff-surface vortices and their c o r r I o i ~ l i f a r s r ti a l s Ncvcrtlle-
1 ~ s i . the ~lolllinc~nr variutioli of rolling n lo~r l r~ l t with roll
a~lgl(. n-;ls nr l l pretlictc~tl. Details xvrre prt'sc:ritcd on lion-
tllc vor tc ,~ l j r t : t~ l i< l<) \~~~ r a ~ i ~ d the: 11o11linear region5 of local
~ t a t i c roll k t ;hilit!-.
T l ~ r author.: would like to t h ~ t ~ l l i Drs. Hanff and
Himnq of the Canadian Ir~st i tutr of Aerospace Research for thcir 11i.lpfid disrussio~ls and for graciously proriding
their cspc~rirr~ental data .
References
' Baucom. C. 11.. Clark, C. . '.Navy Departure/Spin
ant1 .Air Cornhat Ilancux-erirlg Ei-alllation of a Kational
.4eronautici antl Spxcr Xtlministr~~rion Developed Flight Control System for t h r F-14." T~wnty-Xin th Sl-mposillnl
Proccc~lings. The Society of Experimental Tcst Pilots.
ISSS #0742-3705. Sept. 1985.
' Squ>-cn. L. T.. Yip. L.. and C11arnl)crs. J . R . . "Self- I n t l ~ i c ~ l T l - i l l r : Rock of Slrritler Drltn 11-ingi." .iI.4.4 Paper
S1-1983. .411c. 19S1.
" L(.~-in. D. . arid I i a : ~ . .J.. ..Dyllalliic Lo:rtl 111~;r~l l rc~-
menti with Dcltn 1Yiligs Untlcrgoing S!.lf I l l t l~~r~r t l Roll 0.- cillntio~?~." .Toiim(i/ of Az~crriff. 1'01. '21. Yo. 1. .1:111. 19%.
p1;. 30-36.
Arena J r . . A. S.. antl Nelso~i. R . C' . . "The Effect
of Asymmetric Tortes lYakc Charactrristic~i 011 a Slcndcr
Delta lying Uudcrgoing n-ing Rocli l lotion." .iI.-\X Papcr 89-3348. -lug. 1989.
" -4re1ia J r . . -4. S.. and Nrlsou. R . C'.. .. .4i1 Ex-
pcrirnel~tal Stud!- of th(, S o ~ l l i ~ ~ t . a r D!-n;il;lir- P ~ I ( Y I ~ I L I ( W O ~ ~
I<rioxvn as Lying Rock." .-iI.4A4 Paper 90-'2Sl2. .&up. 1990.
" hIorris. S. L., and SYartl. D. 7 . . '.A4 1-itlro-Bascd
Experin~cntal Iur-t>stigation of 1Yilig Rock." AI.4.4 pap(^
59-3349. . iug. 1989.
' Hanff. E. S.. .Tcnliini. S. 13.. . ~ L i t r g c ~ ~ . - 1 i l l l j l i t ~ ~ t l ~ ~
High-Rate Roll Experinlcnts on a Delta ant1 Dcn~I~l r Delta
1Vi11g." -4IA-4 Papcr 90-0224. J m . 1990.
Hanff. E. S.. Iiapoor. I i . . rlnDtey. C. R.. and Prini.
A,. "Large-Aniplitude High-Ratc Roll Oscil1;it ion Sy5tcm
for the Measurement of Xm-Linear . i irloatl~." -41.4B Pa-
per 90-1426. June 1990.
" Hanff. E. S.. and Hilang. S. Z.. "Roll-Intluccd
Cross-Loads on a Delta IVirlg at High Incicloncr." A41A4.4
Paper 91-3223. Sept. 1991.
'' H i ~ a ~ i g . X. 2.. and Hanff. E. S.. "Prediction of
Leadiug-Edge l i l r t e s Brcaktlown on a Delta TYing Oscil-
la t i~lg in Roll." -4Li.4 Paper 9'3-2677. . J I ~ ( ' 1992.
Jc~lkin.;. J . E.. \I>-att . .J. H.. :mtl H;mff. E . S. . "Body--4sis Rollirig IIotioli C'rltical Statoi of ;r G.?-D(.grc,c,
Delta lying." =\I.Ai-\ Paper 93-06'71. .Tan. 1993.
'' .Arena J r . . -4. S.. and Xrlson. R . C'.. ..Unitendy
Surface Prrssure 3leasuren~ents on a Sl(wcler Delta TShg
Undergoing Limit Cycle SVing Rock." -4Ll.4 P a l m 91- 0434. Jan . 1991.
l3 Lee. .E. 11.. and Batina. J . T.. "Conic.al Elllt'r
Methodology For Unsteady \;ortical Flon-s .About, Rolling
Delta brings." .lIAA Paper 91-0730, Jan. 1991.
l4 Kandil. 0. A , . and Salman, A. A.. "Effects of
Leading-Edge Flap Oscillation on L~is teady Drlta lYing
Flow and Rock Control." .iI.SA Paper 91-1796. June 1991.
'' Iiandil, 0. A. . and Salmnn. A. A.. ..Three-
Dimensional Si~lllllation of Slrntler Delta TI-i lq Rock a11(1
Divergence," -4IX.i Paper 92-0'780. .Tan. 1992.
"' Lcc-Ra~~ic l i . E. 11.. and I3ati11;l. .T . T . . ..C'onic;il
Euler -4nalysis and .4cti~-r. Roll Snlpprrs-ioli f o ~ i.-liitc.atl!-
Vortical Flows About Rolling Delta Wings," NASA T P -
3239. Liar. 1993.
hlurman. S. IT.. Schiff, L. B. . and Rizk. Y. M.. '.S11111erical Sirrnilation of the Unsteady Flows about a n
F-18 Aircraft in the High-.4lplla Regime." AIAA Paper 93-3405. Aug. 1993.
'"lladerjian, N. 51.. "Navier-Stokes Prediction of
Large-Amplitude Delta-SVing Roll Oscillations Character- izing SYing Rock." AIAA Paper 92-4428, ,4ug. 1992.
'"ulliam. T. H. and Steger. J . L., "Implicit Finite
Difference, Simulations of Three-Dimensional Compressible Flow." AIAA Joumad, Vol. 18. No. 2. Feb. 1980. pp. 159-
167.
' O Baldwin. B . S., a.nd Lomax, H., "Thin Layer Ap- proximation and Algebraic Model for Separated T~lrbulent Flow." AIAA Paper 76-257, Jan . 1978.
21Degani, D. , and Schiff, L. B., "Computation of Su- personic Viscous Flows Around Pointed Bodies at Large Incidence." AIAA Paper 33.0034. Jan. 1983.
' I 1 --Degani. D.. and Schiff. L. B., "Computation of Tin-
bulent Supersonic Flows Around Pointed Bodies Having
Crossflolv Separation." Journal of Computationml Physics.
l b l . 66. S o . 1. Sept. 1983, pp. 173.196.
2 3 Beam, R . LI. and Warming. R. F . , "An Im-
plicit Finite-Difference Algorithm for Hyperbolic Systems
in Conserva t io~~ Law Form," Journal o f Com.putc~tionnl
PhYszcs. \hl . 22. 1976. pp. 87-110.
24Pulliam. T. H. and C1la11.sr.c.. D. S. . -.A Diago-
nal Form of An Implicit A111,rvxil1iateFtictori/;1tioi1 .4l<o- rithrn." Journal of Co~np~utatio7~(~1 physic.^. \r,l. 30. So. '2.
1981. pp. 347-363.
2 5 Chadcrjian. N. 11.. "Xmnerical .4lgoritlim COI~I-
parison for the =Iccuratc, and EfFici(mt Conlp l i t i i t i~~l of'
High-Incidence I i ~ r t i c a l Flow.'' AIAI-1 Paper 91-0173. .Jan.
1991.
26 Levy, Y.. Private con~iminication. 1993.
" Flores, J . , and Chaclerjian. S. 11.. ..Zolial Sarier-
Stokes SIethodology for Flow Sim~ilation al~ollt a C'omplete
-4ircraft." J o ? I . T ~ of Aircmft. \,hl. 27. So. 7, Ji~l!- 1990. pp. 583-590.
28Steger. J . L.. and Rizk. Y. 11.. " G m ~ r a t i o ~ ~ uf
Three Dimensional Body Fitted Coordinates Using H y perbolic Partial Differential Equations." S A S A T I 1 SG733.
June 1985.
2 9 Levy, Y., Degani, D.. and Seginer. A , . '.Graplii-
cal Visualization of Three-Dimensional ITortical Flo~vs 1,.
Means of Helicit!." AIAA Jotirnal. Tb1. 28. No. S. -411~.
1990. pp. 1347-1352.
30 Robinson, B.. Barnett. R . . and *4gran-al. S.. "-4 Simple Numerical Crit,erion For Tortex ~~~~~~~~~11." ilI.4d Paper 92-0057. .Ja.n. 1992.
31 Hanff, E. S.. and Ericsson. L. E.. . . l l~lltiple Roll -ittractors of a Delta Tying at High Iiicicl~nce." ,IG.ARD CP-494, Oct. 1990.