Escolar Documentos
Profissional Documentos
Cultura Documentos
li
78-257
f .i!
A~D: LI J:;.~!ri
F4GS)~.7i .cN)1)3
10 copy
or r~pubIl5h.
1200 Avenuo
conlacllho
AmerlClln
Instllulo
01 Aeronnullcs
01 lhe Amerlcas
Now Vork. N.Y. 10019.
enci Astronaullcs.
Fort'
,j -4
II
"
()
nHI'-U.YER
mnEI.
F1.0~"S
.1.~s
trac t
An alRebraic turbulence model for t..,o-and
three-dlmenslonal
sepatated
flo"'s Is speclfled th:H
;voids the necesslty for flndlng the edge of the
poundary layer.
Propertles of the model are determlned and comparlsons cade ..,Ithexperlment
for an
Incldent shock on a flat plate. separated fIo'" over
a compression corner, and transonlc fio'" over an
alrfoil.
Separatlon and reattachment
polnts from
~al
Navler-Stokes
solutlons a~ree ..,ithexperl:nent ..,1
thin one boundar:;-layer thickness.
Use of
l.~"'-of-the-..,all
boundary condltions does not alter
the predictlons significantly.
Applications
of the
P0del to other cases are contained In companion
papers.
y~x
value of y
m.~x1mum
angle of attack
boundary-layer
thlckn~ss
boundary-layer
1nteraction
thickness
kinematic
ahead of shock
thlckness
fo6(l - u/ue)dy
9
momentum
thickncss
chord
Cf
skin-friction
Cp
pressure
f~X
m.",i:numof functie:>
free-stream
p..,
pressure
Pt
total pressure
p_
pressure
in frce stream
Re
Reynolds
number based
on chord
Rex
Reynolds
edge
number based
o_u-x/',1_
on dlstance
Ree
mornentum thickness
stress
at the ...
111
1.
length of alrfeil
coefficient
coefflclent
2(p - Pw)/o_u~
F(y)
(Eq.
(8
P~ch n~mber
at the wall
10 free
strea:
~eynolds
~_u_c/u_
number
velocity
paral1el
to solld
ue
velocity
at edge of boundary
ut
frlction
velocitv.. .~
u-
free-stream
coordinate
y+
law-of-the-wall
y~
value of
wall
lntr~duction
There is a practlcal
need in aeronautics as
...
cll as In other flelds for the capability of calculating compressible
turbulent separated flows.
Although progress ~~y have been ~~de to...
ard developing a universal (~Iltiequatlon)
turbulence model for
use in such calculations,
that ~oal has not yet been
achieved (see e.g., Ref. 1 and references thereln).
On the other hand ...
ith current n~r1cal
mcthods
2tw/o_u!
from
leadlng
~_u_e/u_
c;chl~~-
velocity
nor~l
y+
to solld
surface
coordlnate
PvUtY/~
at co:putat10n
point
closest
to
Me t hod
Research
Sc1ent1st.
Co~utat10nal
Hc::>ber
Flu1d
AlAA.
Dynarnics Branch.
CoPyrllht
(Eq. (8) 1s
for FiF\ures
a_
Chief,
Hcmber AIAA.
F(y)
of alrfoll
dlsplacemcnt
local shear
1I0cenclature
at ...
hich
..d
by
U + Ut.
replaced by
k/cp
U/Pr
18
,.
AIgebralc
Turbulence
(10)
Hodel
used
'<
Ycrossover
Ycrossover
where
Is the normal
dlstance
(1)
layers.
The prodlJct
and
~~!!:~
at
are equal.
Inner
Drlest
formulation
is used
YHAXFHAX
replaces
*ue
in the Clauser formu:atlon and the comblnation
YHAXU6IF/FHAX
replaees
\JD1F In a "a 1<.
e formulatlon.
In effect, the distrlbutlon of vorticity 15
used to determine lcngth sc:ale5 50 that the neC('5sity for' flnding the outer edge of the boundary
layer (or uake) Is rcmoved.
< Y
?'crQ.S.
e.r Is the smallest valu~ of
Y
value;-o-'troCl
the Inner and outer formulas
1s taken to be zero
boundary
y
UOIF
In
to turbulence
can hc;
slmulated by settlng
Ut cqual to zero everyvhcrc
In a proflle for whlch the maximum tenlatively computed value of Ut
from the foregoing relatlons Is
les5 than a speclfled valu~. that is,
reglon
(2)
1f
",here
(Ut)nax
in
<
CHlITHu~
(11)
I'rofile
(3)
Is the magnitude
Iwl
The constants appearlng in the foregolng relations have bcC'n deterrnined by requ!rlng agrectXlI!
...
ith the CebC'ciJ formulat!c.n for constant prcsslIre
of the vorticity
beundary
laycrs at transontc
deten:dned
.,rc
au
speeds.
lhe value.
(4)
~z)2
A+
26
and
(5)
CKLEB
HUTI1
C
kK
Pr
Prt
The
ratlonale
behlnd
the cholce
0.72
a 0.9
0.4
0.0168
0.25
~ 0.3
14
C"\,rK
of the follov-
(6)
"'here
tienal
K
is the Cla~~er
~onstant, and
constant,
Ccr
Is an addi-
(7)
the s:::aller
The quantitles
the function
y~\X and
FK~X
are determined
Thin-Laver
from
(8)
1n ,,a1<.es,
the exponentlal
ten:: of Eq.
(8) Is set
equal to zero.
The quantity
FK~X
is the ~ximum
value of
F(y) that oeeura in a proflle and
YHAX
The
Ia the value of y at "'hieh it oeeurs.
funet10n
FKLEB(y)
faetor given by
15 the Klebanoff
ApproxlCdtlon
inter:dttcncy
(9)
tlon
2
The devl'lq'm<'nt...,f
thr thin-loyer
evolves clrrc!ly from a reallstlc
.1pproxlm.1assessrncnt
~
.005
of .
hal is really being cO"'l'uled in a typical high
Reynolds nu~ber Navier-Slokes
simulation.
10e layer
of turbulent flol.lnear the vorllclty generating surface is so thin lhat typlcally a large amount of
computer storage capacity Is used in a highly
Rtrelched ~sh
to resolve the normal gradients of
the fIo .there.
As B result, in most, if not alI,
Navier-Stokes
solutions of hlgh Reynolds n~er
turbulent flol.lsthat have been reported to date, the
dlffusion terms lnvolving derivatives parallel to
lhe surface simply have nol been resolved~ (even
lhough an attempt may have been made to co~ute
lhem) because of the lack of computer capacity.
~tllted in another way, most, !f not a11, NavierStokes solutions that :r.akeuse of a hlghly stretched
mc~h ncar the body surface contain the thin-Iayer
aprroxi~~tion
I.Ihetheror not they are coded to take
IIdvantage of it. The principal differences
in the
various codes lie ln the l.Iaythey model the removal
of energy from the large-scale, computable structure,
and the details of thei r apprOXitMtion
to the energy
cascade in the body-normal
dlrection.
CALCULA TIO'"
3.
times
10
12
(;) -1
I1e
O 02.85
O
(;) M..
1.5
14
M_
.001
Fig. 1 Co~arison
of calculated skin f~Iction on ~
flat pl:\te with Hopkin!-Ino"ye
correlation.
XJ
10
CALCULA T 101'1
O
16
"18
-.:. 12"-
10" RI,
, .,
~
,;,
the !1u:::bers
Df :nesh roints
given in Table 1.
HOI'KINSINO\IVE
COnnElATlON
1
10
1
12
1
"
16
Is
Fi g. 2 Comt'arlson of calclllaled mornontuo thicl<.ness
on a flat pl.1le l.IithlIopl:ins-lnouye conelation.
Resdts
\'al.11e.s.-.Q.._.sk1n...
f rir tinI1-And_
velocily proflle shapes ln reaso.!::\l?..lEagreeocn,L '-'i_lh_
exrcri me-nt for equil1brium bou-;:;daryl'!!y_e_ry=.
_0 f lh 1s pape r produres
Flat Plate
Boundary
Laver
Shock-Wave
Inou)'e correlation.
The
eorr.p:Jrlsonsin Figs.
~gebraic
convergence
turbule_n_ce mode!
Table
time 0:1
COmputllt
36
6900
1980
1800sec 100
1400
5200
Flow1000
over
1980
time compression
steps to
Hesh
points
on Flat rIate
Roundarv
Laver
Tn FIg. 3 calculaled
"udace
pressure and skin
friellon are compared wllh the measure~nts
of Reda
and HurphyG for the intera,:tion of a shock wave "'ith
the turbulent boundary lay"r on a flat plate.
Accordlng
to Rose7 addittonal
measurements for this
fIo,,'are bein" made to det"rmine the extent to .htch
three-dimcnsion.11 and unst'!Ady effrcts influence the
lneasuremenls.
Recause of :he possihle existence of
such effects, Il is not kn,,1.IT1
ho'" i~ortant
the
modifications
of the turbulence models, whlch have
IH lh' the cons tsnt CHUTH = l~ in Eq. (lI) the foregoing turbulence codel predicls transition froe
lamlnar to turbulenl flol.l(as indlcated by rhe rlse
in skin friction) at a ReY!101ds nucber based on distance fro~ the leadln~ edge of about 300,000.
In
Fig. 2 the calculated C>O::1Cntumthickness Reynolds
number is co%?ared '-'ilhresftlts from the Hopkinsand 2 lndicat~
lncldent
and computation
77
64 v 36
CDe 7600,
corner
Nurnber of
tioes
Code
veTsloo
Jlln. 1977
J:I11.1977
Jnn.
1977
July
1977
-------
(y;-!01
EXI'ERIMEHT (REF . I
.1t1
of the lltV-of-ti1c-.all3'l'roximation
would increasc ,.ith lar~l' values of
.10
Separated
""'1
:C
,,-
NE
and
leal
40
1201-
TV
200
~.
120
110
T+Y2'
l\
,,/
,
,(
,',/
,,'-
calculation
has not been
carried
out althout:h
as yet.
son vith exper1ment
in this
problem,
the~
o
ti
FiR. )
- --
EXHRIMEHT
-
6 -----
IREF. 101
CALCULA TICN
CALCULATIC-fo/
, ,
"B
IREF.I:n
O
O
layer.
been cansideredB-11
to accomrnodate these data, may
be.
ln any case, the present predictions
of the
positions of separation and reattachment
are ~ithin
about one boundary-layer
thickness of the
measurements.
O O <XtIoo
O O
. 2.85 SHOCY
-~O.24'
III,
"
at a distance corresponding
to y! - 50
The flov equations vere solved 8nalyti-
for an adiabatic
cally in the interval
O < y < Y2
valI (vith the conditiona
3p/ay - O
and
T - Tv
imposed) to obtain an expresaion
for
Tv
1n the
teres of the values of the flCN quantities
at
y - Y2'
Slip values of velocity and internal energy
~ere then imposed at the valI such that the 3nalytiTv
and the adiabatic
cally determined value of
~all condition vould result from the finite difference solution.
In this procedure the expression
for
Tv is icplicit and involves a universal quadrature
Fig. 4
4
Separat~d
flov ovc~
a co~'ressI0n
corner.
Transonic
Airfoils
Fly)
Ao- 21 x,oa
.1
'-="""
.04
- O.1M
08
.02o
.08
- 2.M
LO'II'ER SURFACE
a velocity profile
u+
u/~T
versus
y+ (Eq. (5
at a station near the nose of the airfoil.
The edge
~f the boundary laye, is ~t y+ - 500.
The middle
plot shova the values of the product
*ue
that
'~ould be obtalned from this profile if the edge of
the boundary layer ~ere taken to be at an arbitrary
value of y (corresponding
to y+
in the scale at
the bottom) rather than at the actual edge of the
boundary layer.
This sho~s the inaccuracies
that
can occur in the evaluation of eddy viscosity
according to the Clauser formulation if the edge of
the boundary layer is not nccurately determined.
In
a machine code it is difficult to automate the
Q
,,
.1
.03
.08
~_'6'y
ca..
.04
cj>
.02
o
30
o
Fly)
t:D
.2,
(b R
p\ I1
u'
O-
cI>?
20
10
o
1
Fig. 6
P6
cf
o
R. - 2\ 10'
I
I
M_ - 0.7S6
0-
2.156
~ 0.01108
c
JO
o
1
10
101 16!
10'
It vas
Fig. 5
Profiles
cf Garabedian-
Cound vorth\JhiJe
to u~c a thrcc-point
quadratic
Cit in thr detenltnation
FHAX
from the functton
FI)') (s,('('
5
,.
edge of Garabed1an-
y+ - 30,000.
Again widely different values of *ue
(middle plot) can be obtained if the edge of the
boundary layer ts incorrec:ly determined, whcreas
the maximum of
F(y) in th.! top plot ia relatively
vell defined.
40
near trailing
Korn ahfoil.
,
b
I
10"
Ul'!'ER SURFACE
O
'\
10
I
IaS
R
I
Profiles
I
10'
YMAX and
Flg!l. 5-7 3nd
oC
o:>
O \I~I
c--::-
.011
-.3
2.8&
.S
.4
.06
.2.1 R.-21.,0'
.10
.Ot
Ul'f'ER
SURFACE
100&..M...
o .011
0.1M
yF
-c-
.02
.03
.02
y~
Fb
.01
, I
.02
9 I
.01
'
.8
l' '
, ,
16'u,1 h-y
""-
10
~
urPER
SURFACE
-10
-20
y'
7
Separated
prafile
aidoil.
fo11s.
tlons
on Garabedian-Kcrn
o.~
-o
.003
.C02
.020
.COI
.018
F __
.018
'0...
.014
~.
UI'f'~ER SURFACE
,-
~nd lcngth
of attaek.
sealcs
on
In the foregoing
treatment
af transonlc
8irtransitian
"'as n,t included
in, the caleula(CMUTH - O in Eq. (I 1.
Figure
11 contains
.07
R. 21 X 10"
M...
vc10city
at anglc
..
"'-
-- -.9
.03
.06
1!r1; CIRCUI.AR
R
104_
11 X
ARC
o"
0.783
y ... F ___
.Ot
.OC!
~IJ
,cxb.
L-'tbcfd
.012
y_.OI0
-C .coe
.coe
.004
.002
o
o
.06.1
.2
.3 .4
.5
.S .7
.8
.OS.I.
.3
,4
.5
.1.
Flg.
8 Turbulenee
vclaeity
Garabedlan-Korn
and length
alrfoil.
scales
on
Fig.
"
...
af
In Figs.
8 and 9 thc varlations
of
YHAX and
FMAX along the uppcr sllrfaee
are sho\"'11 for the
Garabedlan-Korn
:l1rfo11
at angles
of attaek
of zero
and 2.66'.
respeetively.
There is a s=11
amount of
scatter
of the calcul~tpd
values
~bollt a meen line
fa red thraur,h
the po! n1:5.
Ollr previous
experience
indicates
that calculatpd
values
of
6*ue
show much
greater
sc.H tcr so that srnoothng
ove, adjaccnt
..
alues of
x Is nccc!:sary
as \"'cl1 as' the Imposi tion
of arbitraI")'
restrlctions
on the arnount by whiclo 6"
is al.~()'Jed to vary from polnt-to-point.
Thus. much
of the arbltrariness
In thc calculated
values
of
eddy viscos i ty in the ollter part
of thc boundary
laycr
Is rcrnoved by adoption
of the prescnt
turbulence medel in place of that of Ccbcci.3
Figure 10
shO\o/s values
of
YHAX and FHAX
on a biconvex
airfoi1.
Fig.
Turbulence
airfoil
Eq. (8.
Th1s proccdutc
leads
to smoother
variaY~\X
and FHAX
from onc prpfile
to the
tons
of
next than are obtained
~~en the diseretized
peaks
F(y) are used directly.
f~?
,....
2.66
I-
I30
~ 0.8146
I
o
40
Fig.
R, - 21 X 10"
104_ O. ~
!,
p
O
10
Vf loelty
Tllrbll1enc('
clreul~r
are
nnd 1ength
8irfoi1.
!'ca1rs
on
4.
-11
C,)nclusions
- ..
c~
\
o
~Of
.
..
.~
'-1_ - 0.756
R.-21Xlo"
a - O
c,
equation
turbu1ence
(Thc present mcthod
for
that IJas present
in models.
prevlous Navier-Stokes
solutions
separated
turbulent
flo\ls based
on used
zero in
or oneonedefining
1cngth
scales cnu1d
also bc
equation models.)
Comparisons IJith mcasure~nts
from two experlc>ents sho\l agreemcnt ln the predlcion of separation andrctt:Cl1ment polntS \llthin
-;bt one boundary-layer
thickncss, as \11'11 as reasonable levcls of skin friction aft of reattach~nt .
Preliminary
calculations
of the f10IJ over an airroil
indicate that the mode1 is capablc of dea1ing IJith
relatively
large shock-illduced and tral1ing-edge
separations.
Coc:parison: IJith additional experi::>cntsand establ1shed km'IJ1edge of turbulent bounda r)' 1a)'ers a rI' needed.
These 1M)' ind ica te a need
for variation IJith Reynolds number and Hach nU::1ber
of the va1ues of the par
""ters that h3ve bec-n
determlned.
The extent 10 Io'hichthe thin-1ayer
at'proxim..Hion is appllca!le In the prcsence of large
shock-induced
separatlon: over airfolls needs
further investigation.
~pp1ications
to threeUse of the
dimensional
f10\Js '1ft> being pursued.
t~in-layer
approxim.,tlon racilitates 5uch app1ications.
Possible use of lo1\J-of-thc-~~11 boundary
ccndltlons
to avold res01ution of the viscous sub-
.002
.001
layer
;-:i:)'
o
.6
RefE~
x/c
Fig.
11
plots of pressure
froe calculations
and skin-friction
coefficients
in which transition was consldered
curves
reach
Uncertainties
in ~~ch number and ang1e-ofattack corre1ations
in the experiment ha~er
comparisons IJith the ca1culations.
The expert~nta1
pressure dlstributions
~ver a range of ~~ch n~hers
and
ang1es of attack indlcate the presence of a shock
IJave, cn the upper surface near' x/c
0.3, that ts
nct fo~d
ln the calculation.
Refinecent of the
mesh spacing in the X direction near the posttion
of the shock Is probab1y needed in the calculation.
Although additional calcu1ations
are needed for
IJorthIJh11e cor.parisons w1th the exp~rimental
pressure distribut10ns,
Fig. 11 1s inc1uded in this
paper to il1ustrate the properties of the turbu1ence
oodel.
The experi~nta1
drag coefficient
is 0.0120
over a range of ~~ch numbers and an~les of attack
near zero and is 0.0091 from the calculation.
The
computed
11ft coefficient
is 0.533; the coefficient
that Kacprzynski
et a1.13 believe corresponds
to
zero angle of attack at a ~~ch nucber of 0.739 is
0.53.
T . "Calculation
of Co"rressib1e
Tur-
f'
r'Reda, r. C. and Hurphy. J. O., "Shock '-J,Jve
Turbulent
Boundary La)' r Interaetlon
In Rectangular
Chann ls. Pare 11: Th Influence
of Slde\ol:1l1 BOllndSeparatlon
and Srnlr
of
ar)' Layer~ 00 tnrlplrnt
Intrr.lcllon,"
AIAA P.1per 7}-23~.
prt'senlcd
.1t AlAA
11th Aprospace
Scienees
Heetlng
In ~ashlngton,
D.e.,
Jan. 10-12, 1973.
7Rose,
~.
C.,
private
eo~unleatlon.
13K.acprzynskl,
P. R., and Korn. D.
a Shockless
Llftlng
Condltlons,"
National
Aeron:l\Jtleal
Repore