Beruflich Dokumente
Kultur Dokumente
Wilhelm Kordulla
Shear Flow in
Surface-Oriented
Coordinate
Notes on Numerical Fluid Mechanics
Volume 4
Menuscripts should be _II over 1.00 peges. As they will be reproduced foto-
mechanicelly they should be typed with utmost cere on speciel stationary which
will be supplied on request. In print, the size will be reduced linearly to approxi-
metely 75 'It.. Figurll end diagrems should be lenered accordingly so as to produce
letters not smeller then 2 mm in print. The same is valid for handwritten formulae.
Menuscripts lin English) or proposals should be sent to Prof. Dr. K. Forster, Insti-
tut fiir Aerodynemik und Gesdynemik, Pfaffenwaldring 21, 0-7000 Stungan 80.
Ernst Heinrich Hirschel
Wilhelm Kordulla
Shear Flow
in Surface-Oriented
Coordinate
With 40 Figures
The reader should not allow himself to ~e scared off'by the use of
tensorial concepts and of the index notation, both not too familiar
in classic fluid mechanics. The authors made the experience that
the use of these concepts alleviates much the work on realistic
flow problems. In the book they have emphasized the use of these
concepts and not their mathematical proofs.
The book was written while the first author was head of the Ab-
teilung fUr Theorie und Numerik (Department for Theoretical and
Numerical Fluid Mechanics) at the Institut fUr Angewandte Gasdyna-
mik der Deutschen Forschungs- und Versuchsanstalt fUr Luft- und
Raumfahrt (DFVLR) in K6ln-Porz (Institute for Applied Gasdynamics
of the German Research Establishment for Aerospace Engineering),
and later of the Abteilung fUr Grenzschichten at the new DFVLR-
-Institut fUr Theoretische Str6mungsmechanik (Dep,rtment for
Boundary-Layer Research at the DFVLR-Institute for Theoretical
Fluid Mechanics), G6ttingen. The second author worked at the same
places, and is still with the latter institute.
The authors thank their colleagues from the DFVLR for the many
v
discussions and for the contributions which went into the book.
Those who should be mentioned explicitly are K. ROBERT and his co-
-author R. GRUNDMANN, whose papers form a large part of the basis
of the book, Mrs. V. JAWTUSCH, who performed many computations in
early stages of the work, and D. SCHWAMBORN, who checked through
some of the chapters, and contributed directly to parts of chapter
12. The authors also thank numerous colleagues from outside the
DFVLR, in particular J. P. F. LINDHOUT, NLR for fruitful and en-
couraging discussions. Last not least they thank the ladies who
type-wrote the manuscript and produced the Indian-ink versions of
the figures.
VI
Contents
Preface
Page
1• In troduction .••.••.•.•.•..••.•.•••••••••.•.•••.••...•••
VII
Page
VIII
Page
IX
Page
x
1. Introduction
- 1 -
Navier-Stokes equations for high Reynolds number flows past isola-
ted components and complex configurations like wing-body-combina-
tions, realistic inlets, turbo-machinery, etc .. In any case, the
results obtained so far are very encouraging. However, in order
to exploit fully the potential of numerical methods in fluid me-
chanics, many questions need to be answered regarding especially
the phenomena of transition and turbulence, and the properties of
general three-dimensional viscous flows with and without separa-
tion. Therefore much experimental, analytical and numerical work
remains to be done, and regarding the first two disciplines, with
more emphasis as before on three-dimensional and unsteady flows.
- 2 -
methods is nearly indispensible if it has been assured that the
type of flow in question is correctly predicted. Because of the
long computer-time needed, the use of finite-difference methods
is generally accepted for accuracy checks and for basic research
purposes, e.g. investigation of transition properties, turbulence
models, and for the study of means of boundary layer control, such
as suction, blowing, cooling and the like.
- 3 -
are integrated in this form, since the transformation of results
back into physical quantities is simple. A further advantage lies
in the fact that the same tensorial concepts are employed for the
geometric approximation of body surfaces and of metric properties.
In particular transformations from and into Cartesian coordinates,
which are often used to compute inviscid flows and in which the
boundary conditions for boundary-layer predictions are thus given,
are very simple (appendix A). Also very simple is the construction
of streamlines (appendix E) .
- 4 -
ted with concave surfaces, surface-oriented locally monoclinic co-
ordinates are attractive for the use with Navier-Stokes and even
Euler equations because of their simplicity with respect to the
metric properties and in particular because of their boundary-fit-
tingcharacter. Furthermore the Navier-Stokes equations for the two
surface-tangential directions then contain naturally the boundary-
layer diffusion terms, which easily leads to a thin-layer
approximation of the Navier-Stokes equations for high Reynolds
numbers (chapter 8). To circumvent the problems associated with
concave surfaces for global flow predictions, one must replace the
surface-oriented locally monoclinic coordinates away from the sur-
face, where the x 3 -coordinate lines intersect and thus become sin-
gular, by preferably Cartesian or other orthogonal coordinates.
- 5 -
ling edge would allow to remove the intersection line B-B or to
place it at x 1 ' + 00. Note that grid-stretching or compression is
possible where wanted, see Fig. 1.1. Fig. 1.2 shows another ex-
ample, the wing-body junction in a plane x 1 ' = const .. Here the
location of the boundary of the Cartesian coordinate system de-
pends on the magnitude of the radius r of the fairing between wing
and fuselage. A sharp corner would make the concept inapplicable.
Note, that in practice both coordinate systems must overlap in or-
der to enable the interpolation of data from one mesh to the
other. The need to interpolate is certainly a disadvantage which
is enhanced further on general surfaces where the coordinate lines
X 3'
x ' =const.
vj' x 2=const.
v2'
x ' =const.
x 3=const.
2'
X
- 6 -
they exhibit definitely advantages: simplicity, boundary fitting
and very easy use on all convex shapes.
- 7 -
shifters eases considerably, for instance, order-of-magnitude con-
siderations, and other discussions of the governing equations.
- 8 -
cussed in terms of surface curvature where higher-order boundary-
layer theory applies.
- 9 -
2. First-Order Boundary-Layer Equations
- 10 -
relationships being given by the base vectors a. (x j ') of the
. -l.
xl.-system (see appendix A) .
Continuity equation
(2.1)
(2.2)
1 1 1 2 1 3 l' 1 2 1 2
P [ V't + v v'l + v v, 2 + v v, 3 + k 11 .( v) + k 1 2v v +
1
k 16 P'l + k 17 P'2 + ( I.L v, 3) ,3
1
v 'i -
- 11 -
Momentum equation for x 2 -direction
(2.3)
p [v 2 ,t + v
12
v'1 + v
22
v'2 + v
32
v'3 +
12
k21 (v) +
12
k 22 v v +
(2.4)
123
+ v T'1 + v T'2 + v T'3
1 2] [ 12 12
+ Eref ([p't + v P'1 + v P'2 + ~ k 41 (v '3) + k42 v'3 v'3 +
(2.5) o « R
min
For more detailed discussions of this and other cases see chapters
7, 8 and 9.
- 12 -
*) 1 2
In equations (2.1) - (2.4) the velocity coordinates v and v
are contravariant velocity coordinates. These are related to
the physical contravariant velocity coordinates v*1 and"v*2 by
means of the diagonal elements a 11 and a 22 of the metric
tensor of the surface 7,12]
(2.6a) a = 1,2.
Ia: (aa) / Lref
*i
The physical contravariant components v are identical with those
used generally in fluid me~hanics and which are there denoted by
vi which, however, is inconsistent with tensor notation. The
v
quantity re f is the reference velocity of the problem. Variables
without bar have been non-dimensionalize'd with appropriate refer-
ence quantities. The lower index e indicates the external in-
viscid flow field.
(2.6b) v
*3
1 2
a11(x ,x)
(2.8) (
a 21 (x 1 ,x 2 ) a 22 (x 1 ,x 2 )
(2.9) ;a 1,2;
Lref is the reference length of the problem. laaa relates the xQ-
coordinate lines locally to a physical length x*a, which, however,
in general, is not a coordinate.
- 14 -
p T
(2.10) p P T
- - 2'
P ref (v ref) Tref
k t
k ref
pT
(2.12 ) P =
Yref(Mref) 2
Continuity equation
(2.13b)
xl-momentum equation
(2.14 )
- 15 -
(2.15 )
(2.16 )
(2. 17)
(2.18 )
x 2 -momentum equation
1
(2. 19) k21 2a [a 11
(2 a 12 '1 - all' 2) - a 12 a 11 'l) ,
(2.22)
(2.23)
energy equation
(2.24)
(2.25 )
(2.26)
(2.27) g •
- 16 -
The square of the length element on the surface therefore reads
(2.28)
x3
3
(2.29) -+- co (x = 6) :
1 1
v1 ve (x , x 2 ) , v2 = v! (x 1 , x 2 ) ,
T T (x 1 , x 2 ),
e
(2.30) 0:
i = 1,2,3
although blowing or suction in the frame of the boundary-layer
assumptions
(2.31) x2,
(2.32) T (x 1 ,
- 17 -
1 2 1 2
(2.33) T'3 (x , x , 0) Tw' 3 (x ,. x )
(2.34)
v
a ( xi ) , T 1 ,2,3 a = 1,2
- 18 -
3. Boundary-Layer Parameters
(3.1)
T· 13 Lref
(3.1a) - 11 v
.1
'3 ra:t1 1'13
(3.1b) 13 1
l' - 11 v '3
(3.2)
(3.2a) 23
l'
(3.2b) 2
- 11 v '3
- 19 -
It is interesting to realize that in first-order boundary-layer
theory the influence of surface curvature is not explicitly appa-
rent in the expressions (3.1) and (3.2) for the shear stress
(/a nn is simply a length scale).
-*3 L
q ref
( 3 • 3)
'k ref'T ref ~
ref
The remarks about the use of equations (3.1) and (3.2) in turbu-
lent flows are valid for equation (3.3) in a similar fashion.
While it is sufficient to know the molecular heat conductivity in
order to determine the heat flux at the wall, which is of prime
interest, one introduces an effective heat conductivity, including
the apparent turbulent heat conductivity, away from the surface.
In most cases the effective turbulent heat conductivity is ex-
pressed by means of the effective viscosity and a constant, ef-
fective turbulent Prandtl number.
- 20 -
3.3. Displacement Thicknesses
(3.4)
- 21 -
(3.5)
xl-direction
0
61x1 IRe ref
pv
1
(3.6) o1 x 1 - (1 - ) dx 3
L ref J 1
Pe v e
0
x2~direction
6
61x2 IRe ref
pv
2
(3.7) 01 2 - (1 - ) dx 3
x L J 2
Pe v e
ref
0
a
For two-dimensional flows --- = 0, and equation (3.5) reduces to
ax 2
the two-dimensional definition of the displacement thickness (3.6),
if v 3 (x 3 = 0) = o.
- 22 -
The mass-flow displacement thickness is often used in inviscid-
viscous interaction calculations to simulate the effective (vis-
cous) shape of a body immersed into some flow, because it re-
presents the displacement of the freestream due to viscosity.
On the new body surface the boundary conditions are then applied
for the inviscid-flow prediction. An alternative is to use the
equivalent wall-outflow velocity on the geometric body as boun-
dary condition reflecting the viscous influence, LIGHTHILL [56],
PIERS et al. [69]. No partial differential equation has to be
solved, but the boundary layer profiles are still needed:
(3.8)
inv
The lower index inv abbreviates inviscid and denotes the sur-
face-normal mass-flow component to be imposed at x 3 = O. Because
Po inv is initially unknown, the use of equation (3.8) may in-
volve some iteration in compressible flows. But since the shape
of the body remains unchanged in the course of an inviscid-vis-
cous coupling procedure, the metric factors of the body remain
unchanged as well, and have to be determined only once if they
are needed for the inviscid-flow prediction.
(3.9)
62 x Q /Ref
re ,. Q = 1,2 •
Lref
- 23 -
Following the derivation of KUX [ 1] two first-order partial
differential equations result for the determination of 6 2 a in
x
zero- or first-order boundary-layer theory (see also chapter 7):
(3.10) (v 1) 2
[k01 P e e (6 2 1 1 - 6 2x1 )]'1 +
x x
+ [k01 P e v 1 v 2 (6 2 1 2 - 6 2 1)]'2+
e e
x x x
+ (v 1) 2 6 2 1) +
k01 k11 P e e (6 2 1 1 -
x x x
1 2 [ 1
+ k01 k12 P e veve 6 2 1 2 - "2 (6 2 1 + 6 2 2)] +
x x x x
+ (v 2 )2 (6 - 6 2 2) 0
k01 k13 P e e 2 2 2
x x x
and
(3.11) 1 2
[k01 P e v e v e (6 2 1 2 - 6 2 2) ]'1 +
x x x
(v 2 )2 6 22 1]'2 +
+ [k01 P e e (6 2 2 2 -
x x x
+ (v 1) 2 +
k01 k21 P e e (6 2 1 1 - 6
2 1
)
x x x
1 2 1 (6
+ k01 k22 P e v e v e - "2 2 + 6 2 2)] +
[6 2 1 2
x x x1 x
+ (v 2 )2 0
k01 k23 P e e (6 2 2 2 - 6
2 2
)
x x x
with
6
(3.12) 6 a,a 1,2 (no
x xa
2 a J
o sUmmation!) •
- 24 -
Again, because of the boundary-layer integrals (3.12) the momen-
° °
tum-flow displacement thicknesses 2 1 and 2 2 can only be ob-
x x
tained from equation (3.10) and (3.11) after the boundary-layer
equations have been integrated, or density and velocity profiles
have been measured.
The energy flux is a scalar, and therefore only one equation re-
sults as it is the case for the mass-flow displacement thickness.
Let E be the quantity under consideration: the kinetic energy
per unit mass ~(q)2, that per unit volume ~ p(q)2, or the total
enthalpy h + ~(q)2 And let the dimensionless energy-flow dis-
placement thickness °3 be defined by
(3.13 )
Then, following the concept of KUX [ 11, the form of the partial
differential equation for °
3 is the same as that for the mass-
flow (3.5) without source term if p is replaced by E:
with
(3.15 ) a = 1,2
(no summationl).
- 25 -
4. Stagnation - Point Solution
- 26 -
X3
1
X "
Fig.4.1 Nodal point of attachment on a general, curved surface.
(4.1)
o o
- 27 -
xl
(4.2)
T'l p , 1 p , 2 o .
xl = 0, x 2 = 0, x 3
-> 0:
(4.3) v* 1 = 0, v* 2 = 0,
v *1 , 1 +0, v
*2
'2 +0,
1 2 3
(4.4) p (v ,1 + v , 2) +(pv)'3 o ,
- 28 -
and thus requires the knowledge of the profiles of the gradients
v~l and v~2 in order to be integrated for v 3 as usual. Defining
new dependent variables at xl = 0, x 2 = 0, x 3 ~ 0
Aa Aa (x 3 ) [v*a(X 3 ) ]
(4.5) [v a (x 3 )]'a 'a a = 1 ,2
v'a(aa) (no summation!) ,
(4.6) o .
(4.7) a=1,2,
where the relation v a '3a = va'a3 has been used since it can be
assumed that the va are continuous at S. Furthermore, the pressure
gradient terms have been replaced by the convecitve terms at the
outer edge of the boundary layer.
(4.8)
- 29 -
Finally, the energy equation (2.4) reduces to
3
(4.9) cp p v T, 3
(4.10) a = 1,2
(no summation!) .
(4 • 11 )
+ A~ 01 2
x
(4.12) +
- 30 -
where 0, 11
x
(4. '3) 0, X
11 a = 1,2
(no summation!).
(4. '4)
2
62 1 1 + Pi.e 62 , ')
(4. , 5) x x x X""
(4. '6)
with
(4.18)
- 31 -
where
(4.19) U = 1,2
(no summation!),
After the stagnation point solution has been found, the dependent
variables in the vicinity of the stagnation point can be computed.
These values can serve as initial values for the field solution.
2
X
5
=0
,
Xb= const.
X' =0
5
S(X'5 =OX
IS
L =OX 3 ,
, J
- 32 -
For small distances between stagnation point 5 (xs1 = 0, x s2 = 0, x 3 )
and field points P(X~'X~'X3) all scalar quantities p,T,p as well
as v 3 are to first order the same in both points because of the
conditions eq. (4.2).
(4.20) a = 1,2 .
i'
Here ajs(p) are the components of the base vectors ~is of the
stagnation point coordinate system in P(X~'X~).
Here af'b are the "inverse components" of the base vectors ~jb
of the body coordinate system in P(X~'X~) following from
eq. (A. 22) •
- 33 -
5. Quasi-Two-Dimensional Boundary Layers
X 3'
\
Plane of Symmetry
X 7'
Fig. 5.1 Plane-of-symmetry flow, and (locally) orthogonal
coordinate system in the plane of symmetry at x 2 O.
- 35 -
constant chord and constant span, are orthogonal along the leading
edge which corresponds to the line of attachment imbedded in a
plane of chordwise symmetrical flow. If, however, there is an angle
of attack the line of attachment moves away from the leading edge
and becomes curved. It must then be assumed that a surface of sym-
metry approximately exists, at least locally: the above chosen
coordinates, however, are non-orthogonal there (see chapter 5.2.4).
In the case of symmetric bodies at angle of attack with a streamwise
plane of symmetry of flow, the surface coordinates usually employed
are orthogonal there.
(5.1 ) a
(5.2)
- 36 -
(5.3) k16
all
(5.5) k27 a 22
Continuity equation
1 2 3
(5.7) (k O1 p v ), 1 + kO 1 j) A + (k O1 p v ), 3 = 0
(5.8) 11 31 12 1
p [v v ' l + v v'3 + k 11 (v) 1 = k16 P'l + (1-1 v '3)'3
2
The momentum equation for the x -direction b~comes meaningless and
has to be differentiated with respect to x 2 in order to yield an
equation for the determination of the quantity A2 remaining in the
continuity equation (5.7). The differentiated momentum equation
(2.3) then reads at x 2 = 0:
- 37 -
(5.9)
2
+ (~L A '3)'3
with k 26 '2
Energy equation
(5.10)
1 1 2
+ Eref [v P'l + J.1 k41 (v '3) J.
(5.11 )
- 38 -
This is an ordinary differential equation for A~ which can be
integrated if the other variables are known. The initial value
for the integration of equation (5.11) is given at the stagnation
point as follows:
2
(5.12 ) 0, x - 0)
0, x 2 _ O.
(5.13) *13
t
(5.15) q
*3
- 39 -
The partial differential equations which have to be solved in
order to compute the displacement thickness discussed in chapter
3, reduce to ordinary differential equations due to the assump-
tion of symmetry.
(5.16)
with
6 1
(5.17) 61 1
x
f
0
( 1 - ~) dx 3
Pe Ve
6 2
(5.18) 61 2 =f ( 1 - ~) dx 3
x 0 Pe Ae
(5.19 )
- 40 -
(5.20) (v') 2 ( O2 - O2 )]" +
[k O' P e e
,
Pe ve A2e (0 2 , 2 -
x'x'
0
x'
) +
+ k O' 2x'
x x
(v') 2 (6 2
+ k O' k" Pe e
x'x' - 62
x'
) 0
Note that the last term disappears if a" # a" (x,) because then
k" = 0 (see eq. ( 5 . 2) ) .
(5.21) A2
[k O' Pe v e e
, (0 2 - O2 )]" +
x'x 2 x2
Pe (A2) 2 (0 - O2 ) +
+ 2 k O' e 222
,
+ - - k 2 ,'2 P e
x x
(v') 2 (6 2
x2
- 0 ) +
e 2x'
k O' x'x'
k22 Pe ve
, A2 [6
2 - 2"
, (6 2 + O2 ) ] 0
+ k O' e
x'x 2 x' x2
o
(5.22) f [1 - a,a = 1,2
o (no summation!),
where F = v' for a,a 1 and F ='A 2 for a,a = 2.The energy-flow
displacement thickness 6 3 , finally, is the solution of the fol-
lowing first-order ordinary differential equation:
(5.23) o ,
with
0 E Fa
(5.24) 63
xa
f
0
(1 - --I
E Fa
a = 1,2
e e (no summation!),
1
again with F = v for a = 1 and F = A2 for a = 2.
- 41 -
The initial data needed to start the integration of these
ordinary differential equations are given by the stagnation-point
values from chapter 4.2.
- 42 -
swept-wing flow" which allows for a chordwise change of the invis-
cid spanwise velocity component at the edge of the boundary layer
(chapter 5.2.3). In fact, the correct external boundary conditions
are used, but the spanwise derivatives in the governing equations
are neglected.
- 43 -
5.2.1 Leading-Edge Oriented Orthogonal Coordinates
x~
- 44 -
Introducing these results and neglecting the derivatives with
respect to x 2 in eqs. (2.1) to (2.4) yields the familiar equations
for steady flows (zero-order theory, see chapter 9.1) :
Continuity equation
1 3
(5.25) (p v ), 1 + (p v ), 3 o
1 1 3 1
(5.26) p [v v ' l + v v'3]
1 2 3 2 2
(5.27) p [v v'l + v v'3] (11 v '3)'3'
Energy equation
(5.28)
+ E ref {v 1 p" + 11
12
[(v '3)
22}
+ (v '3) ]
(5.29)
1 v 1 (xl), 2
v v sin !PO const.
e
T T (x'),
e
- 45 -
The boundary layer parameters in terms of the leading-edge orien-
ted orthogonal coordinate system are easily obtained from equations
(3.1) to (3.15). For the sake of completeness, they are presented
here:
Shear stress in xa-direction, a = 1,2:
(5.30)
(5.31 ) *3
q - k T'3 .
(5.32)
where
(5.33)
(5.34)
The partial differential equations (3.10) and (3.11) for the momen-
tum-flow displacement thicknesses reduce to:
- 46 -
Ii 1 a
f [1 - P v v 1 dx 3 =
e v:
(5.35) a 1,2
o P v~ (no summation!),
Ii 1
(5.36) f [1 - ~l dx 3
o E
e
v1
e
1 2 + v3 2 1 2
(5.37) P [v v '1 v '3 - p, 2 + (1.1 v '3)' 3
- 47 -
5.2.2 Leading-Edge Oriented Non-Orthogonal Coordinates
X3
1
X7
1
,
, , 3'
and parallel to the x -x plane;~O = angle of sweep
of leading edge). a c (x ) is the contour angle.
,
the upper surface for back swept wings and equal or larger than
w/2 on the lower surface. Hence a'2 (= a 2 ,) f 0 forlx I > O. At
the leading edge, not necessarily coincident with the attachment
line, a'2 = a 2 , = O. The metric tensor coordinate a'2 is only a
function of x'. For a planar wing a'2 a2' = const •• The
other metric coordinates a"and a 22 are in first-order boundary-
layer theory equal to one (chapter '0.'.2). Thus the metric fac-
tors k mn , equations (2.'3) to (2.26), are finite except for k'2'
- 48 -
k 13 , k22 and k23 which vanish due to the above assumptions.
Continuity equation
1 3
(5.38) (k 01 P v )'1 + (k 01 p v )'3 o
1 1
(5.39) p [v v , 1 + v3 v 1 , 3 + k 11 1 2
(v ) I = k 16 1 )
P '.1 + (1.1 v , 3 ' 3
(5.40) P [ V1 2 3 2 1 2 2
v , 1 + v v'3 + k21 (v ) 1= k26 P'l + (1.1 v '3)'3
Energy equation
(5.41 )
+ Eref
{v 1 P'l + 1.1 [k (v 1 '3) 2 + k42 v 1 '3 v 2 '3 +
41
- 49 -
to the orthogonal coordinates (chapter 5.2.1 and chapter 10.1.2):
(5.42)
(5.43) v! (x l )
*3
(5.45) q
(5.46)
(5.47)
- 50 -
Equations (3.10) and (3.11), which determine the momentum-flow
displacement ticknesses 02 and 02 ' reduce now to:
xl x2
(S.48)
(S.49)
6 1
(S.SO) f [1 - ~l dx 3
o E
e
v1
e
- 51 -
5.2.3 Locally-Infinite-Swept-Wing Concept
Continuity equation
1 + 2 3
(5.51) (k 01 p v ), 1 + k01 p v + (k 01 p v ), 3 o
where
- 52 -
Momentum equation for xl-direction
3 1 2
(5.52) P Iv 1 v 1 , 1 + V v 1 , 3 + k11
+
(v 1) 2 + k12 v v + k13 (v 2 )2]
1 + 1 2 (V 2 )2]
Pe Iv 1 v e 'l + k11
e
(V~)2 + k12 v e v e + k13 e
+
+ 1
(I! v , 3) , 3
where
+ 1
k12 k12 - 2a
11 all' 2
1 2 3 1 +
(5.53) P Iv V , 1 + V v 2 , 3 + k21 (v 1) 2
+ k22 v v 2 + k23 (v2) 2]
2 1 2 +
Pe Iv 1 v e'1 + k21
e (V~) + k22 ve ve + k23 (v!) 2] +
2
+ (I! v , 3) , 3
where
+ 1
k23 k23 - 2a
22 a 22 '2
Energy equation
(5.54)
{V 1 I (1) 2 1 2
+ E ref P'l + I! k41 v'3 + k42 v '3 v '3 +
- 53 -
(5.55) x, > 0, x
2 t
cons., X3 -> 00:
2
Note, again, that the relations (5.55) change with x .
*a3
(5.56) '( =- II la(nn) a = ',2
*3
(5.57) q - k T'3
(5.58)
(5.59)
- 54 -
Momentum-flow displacement thickness in xl-direction
(5.60)
1 2 1
+ k k p v v [0 - - (0 + O2 ») +
01 12 e e e 2x 1x 2 2 2x 1 x2
and in x 2 -direction
(5.61 )
where
+ kOl +++
k 01 - 2a k 01
11 a 11 '2 1
- 55 -
Energy-flow displacement thickness
(5.62)
- 56 -
turbulent transition location measured in experiment.
L /
1;7 - x'=const.
X
"
Fig. 5.4 Approximation of the attachment line for a lifting
locally-infinite-swept-wing flow at two span stations
(index inf stands for infinite-swept-wing assumption).
- 57 -
The governing boundary-layer equations are obtained from equations
(5.51) to (5.54) by taking the limit for x 1 + 0, with v 1 (x 1 = 0,
x 2 = const., x 3 ) = o.
Continuity equation
(5.63) o.
(5.64)
2
= Pe [(Ae1)2 + k+12 A1e ve2 + k 13'1 (ve )2] + (A1)
~ '3 '3·
(5.65)
+ + +
where k 01 ' k12 and k23 are defined in equations (5.51) to (5.53) .
Energy equation
A1 2 2 2 2 2
e (x ), v ve (x ), T Te (x ).
- 58 -
Shear stresses in x'- and x 2 -directions:
(5.68) '[
*'3 o ( '[ *'3 ) , , - IJ.
,
~ A '3
(5.69)
*3
(5.70) q =- k T'3
(5.71)
(5.72)
(5.73)
- 59 -
and in x 2 -direction:
(5.74) kO 1 A 1 2 (0 - O2 ) + k +++
e ve 2x 1x 2 x2 01
These are two algebraic equations which can be solved for and <5 2
x1
O2 after the boundary-layer equations (5.63) to (5.66) have been
x2
integrated. Finally, the energy-flow displacement thickness is
found from the algebraic relation:
(5.75)
For the infinite swept wing the equations for the initial data
follow from the foregoing equations if the derivatives with
respect to x 2 of all metric coefficients or factors are simply set
to zero. This holds for both the orthogonal (chapter 5.2.1) and the
non-orthogonal (chapter 5.2.2) case, for the first, beside this,
a 12 " O.
For the derivation of the above equations it has been assume.d tacit-
ly that k13 = 0 at x 1 = 0, otherwise the first momentum equation
would not have vanished completely. This means that locally the
contour is assumed to be cylindrical, which further implies, that
the attachment line is locally straight and lies on a generator of
that cylinder (see also chapter 12.3). Another point of view would
result if one would deal from the beginning with the momentum equa-
tion for the x '-direction after having it differentiated with respect
, 2 2
to the x -direction. Then one has to assume k13 v v'1 = 0 at
x 1 = O. If k'3 ~ 0, then only v~1 = 0, that is, the flow is symme-
trical with respect to the attachment line. However, this is not
true, in general. In practice therefore an incompatibility arises
between this quasi-one-dimensional attachment-line solution and
- 60 -
the locally-infinite-swept-wing solution continuing it. If the
flow or the surface is not sufficiently symmetrical no continuing
solution can be achieved at all*).
*)
This discussion is due to D. Schwamborn, for details see [78].
- 61 -
6. Initial Conditions for the Prediction of Boundary Layers on
Wings and Fuselages
- 62 -
limiting wall streamline
direction
- 63 -
leads to Figure 6.3, an illustration of RAETZ' influence principle
which simply states the following: let initial conditions be speci-
fied on a surface 51 normal to the surface of the body considered,
where the line of intersection does not coincide with the projection
of a streamline. If this surface is of finite extent, then the so-
lution of the boundary-layer equations is completely defined in
terms of those initial conditions only in a domain of influence of
finite extent. This domain is bounded by 5, and two other surfaces
52 and 53 normal to the body, see Fig. 6.3. They are erected over
a limiting wall and/or the projection of an external inviscid
streamline such that the enclosed area of the surface of the body
attains a minimum. The flow outside of this domain of influence
cannot be computed as it is affected from the flow outside of 51'
This is shown in the left part of Fig. 6.3, where the surface
55 intersects the surface 56' and 55 originates outside of sur-
face 5,.
5,
surface
of the body
- 64 -
Since the region of influence is given by the streamline directions,
to be stable, the integration scheme must always be oriented such
that locally the flow is followed, and its numerical domain of
dependence must always include the domain of dependence of the
boundary-layer equations (Courant-Friedrichs-Lewy condition). This
becomes important in three-dimensional flows when the flow lateral
to the main marching direction changes its direction. Such flows
can be handled by the second-order accurate z-shaped integration
scheme of KRAUSE et al. [44,47,481. This scheme is able to march
with conditional stability against cross flow because it uses
downstream as well as upstream information in the cross flow
direction. CEBECI et al. [201 use the characteristic box scheme
which is based on the integration of the boundary-layer equations
in local streamline coordinates within the boundary. layer.
LINDHOUT et al. [581 start their integration, based on some modi-
fied Laasonen [ 91 scheme, from one plane with initial data and
the program computes the flow in the region of determinacy which
is automatically determined when marching downstream. Both latter
methods fail to predict the flow on the outboard portion of the
upper surface of a lifting wing at angle of attack unless some
approximate solution is created along the tip where the flow is
directed inboard. Thus, on wings generally two planes with initial
data are used, along the leading edge and in chordwise direction.
- 65 -
the inviscid flow field is known, the thr.ee-dimensional stagna-
tion point can be determined, and the boundary-layer solution
discussed in chapter 4 can be computed. The construction of that
solution tacitly assumes that at least one symmetry plane exists,
along which a quasi-two-dimensional solutio~ (see chapter 5) is
calculated which can establish initial values for the circum-
ferential (lateral) integration. The integration in the main stream
direction can then be carried out with, for example, the z-shaped
scheme after initial values for that direction have been provided,
for instance near the stagnation point (see below). Often a body-
-oriented coordinate system is used which exhibits a singular be-
haviour at the nose because of the vanishing cross section. This
does not pose any problems for a symmetrical flow because the nose
coincides with the stagnation point (where the computation starts),
but for flows at an angle of attack one has to introduce a special
transformation [17] or for instance an external streamline coordi-
nate system for the nose region in order to circumvent that geo-
metrical singularity (see chapter 10.2).
Consider now the flow prediction for a finite swept wing, Fig. 6.4.
If the flow field is symmetric on the upper and lower surface
(non-lifting symmetrical wing), the flow along the geometrical
leading edge which then coincides with the attachment or "stagna-
tion" line, is described by the quasi-two-dimensional solution
- 66 -
(see chapter 5.1) denoted in chapter 5.1 with plane-of-symmetry
flow. If, however, the flow is lifting, it can be demonstrated
that there does not exist a quasi-two-dimensional solu-
tion regardless of the definition used for the attachment line
[78] (e.g. locations at which the chordwise velocity component
vanishes or where the flow locally is symmetrical). It is con-
venient then to use the approach discussed in chapter 5.2.4 which
is based on the locally-infinite-swept-wing assumption. An accurate
approach, however, is being made by SCHWAMMBORN [78] along the
leading edge of wings with chordwise symmetry plane (the existence
of the symmetry plane facilitates the generation of initial data),
where the flow in a region encompassing the attachment line is
predicted as follows. The solution is initiated in the symmetry
plane at the stagnation point and progresses in the spanwise di-
rection using the z-shaped finite-difference scheme, at the chord-
wise boundaries the ordinary four-points scheme with correct ad-
vective orientation is employed. Since the scheme requires initial
conditions for the chordwise integration direction for each span-
wise integration step, these are obtained by iteratively solving
the boundary-layer equations with two z-shaped schemes with oppo-
site marching directions for the profiles at two mesh pOints near
the location of the attachment line. Thus, second-order accuracy
is maintained also for the initializing steps in the spanwise di-
rection. Once the flow is determined in the spanwise strip, the
flow in the remaining parts of the lower and upper surface can be
predicted as usual. One of the objectives of the work is to esti-
mate the error when using locally-infinite-swept-wing solutions in
the neighbourhood of the attachment line to start the chordwise in-
tegration. Several results of the investigations by SCHWAMBORN
[78] are discussed in chapter 12.
- 67 -
x 'teading-Edge" Tip
Solution ~~
I
~
(X)
J?
V
J
"Locally-lnfinite- Swept-Wing"
Solution
x"
''plane-or-Symmetry'' Solution (Root)
Fig. 6.4 Sketch of a wing with an inviscid flow pattern
(planform view) .
- 68 -
wise direction) of the boundary-layer equations reduces the in-
fluence of the initial data chosen, anyway.
- 69 -
7. Higher-Order Boundary-Layer Equations
(7.2)
- 70 -
In order to derive his equations, ROBERT [ 3] introduces shifters
- a concept taken from shell-elasticity theory -, which express
metric quantities associated with the space above the surface in
terms of the surface metric in a very elegant manner. Shifters can
also be understood as relations which transform velocity coordi-
nates for xa-coordinates away from the surface into those given by
the directions of the coordinates on the surface (see below and
appendix A). Here, if convenient, shifters are avoided in the for-
mulation of the governing equations in order to be consistent with
the remainder of the book. Then the equations are given with the
metric factors based on the appropriate a- or g-metric. Note that
the metric factors which contain the a-metric (case I, eq. (7.1b»
are denoted by k mn = k (xl, x 2 ), and those which contain the
mn _ 1 2 3
g-metric (case II, eq.(7.1b», byk =k (x, x, x). The ve-
mn mn
locity coordinates are in either case given without overbar:
*a a 1/2
v v /(g(aa» • It is worth to note that curvature-gradient
terms are usually not regarded in second-order theory.
(7.3) 1 2 3 o ,
kOl P't + (k 01 P v ), 1 + (k 01 P v ), 2 + (k 01 p v ), 3
( 7 • 4) 1 2 3
kOl P't + (k 01 P v ), 1 + (k 01 p v ), 2 + (k 01 p v ), 3 +
-+ 1 -++
+ k01 P v + k01 P v 2 0
where
(7.5)
(7.6) a v1 a + v2 a + v3 a
P V't + v'1 v'2 v'3 +
- 71 -
-+ -+
(V 1 )2 + ka2 1 2 (v 2 ) 2 1
+ ka1 v v + k+
a3
a
ka6 P'1 + ka7 P'2 + ~ (\I v'3)'3
ref
where
-+
(7.7) ka1 = Aa1; 1
1;
MO { 0 1 }
-+
ka2 Aa MO { 0 1; 1 } + Aa MO { 0 l; 2}
l; 2 1; 1
-+
ka3 = Aa1;; MO
2 fo
l;
2}
a a 2 a 3 a
(7.8) p [ V't + v 1 v'1 + v v'2 + v v'3 +
+ k (v 1 )2 + ka2 v 1 v 2 + ka3 (v 2 )2
a1
1
ka6 P'1 + ka7 P'2 + Re ref (\I v~ 3) , 3
(7.10)
1 1
Pr ref Re ref
- 72 -
+ __11_
Re ref
The shifters of the first Milv and second A6E kind are defined in
appendix A.14, see also below. Note that the velocity coordinates
appearing in the above equations are those measured with respect
to the local xi-coordinates away from the surface.
Wall shear stresses and heat flux are defined as for first-order
boundary layers, see chapter 3. The situation is different for in-
tegral parameters because of the skewing of the surface-parallel
coordinates away from the surface, see appendix A. If the skewing
is expressed by the variation of the angle ~ between surface-paral-
lel coordinates along a surface normal:
(7.11) cos ~
1 2 3 1 2 3)
(7.12) Y(a) (x,x,x ) = v a (x,x,x Sa (x 1 ,x 2 ,x 3 ) =
- 73 -
1 2 3 1 2 3 (1
V~ (y) x ,x 2 ,x 3)
~y
v (x,x,x) a (x,x,x 0)
-y
i'
v ~i'
~ (3)
(7.13)
(7.14)
(7.15)
~* 1/2 ~ y
Note that v Y = (a()) v Hence, once the velocity coordinates
a YY . ~Y
v are determined, the velocity coordinates v , based on a coordi-
1 2
nate system which is constant along a surface normal x , x =
= const., can be obtained as is required for the integration of in-
tegral parameters.
f (t'" e
05
Aa ". Aa 3
(7.16) ve - tV) M dx - o
o
- 74 -
". Aa. .
Here t v ~s the boundary-layer flow quantity displacing the ex-
A An
ternal flow quantity t v (note that only the tangential fluxes
e e A A3
are being considered, the fluxes t v are still small of higher
A An
order). The quantity t v may be, see chapter 3, the density flux
(t = p), yielding the mass-flow displacement thickness 01 from eq.
A
(7.17) M
(7.19)
- 75 -
Introducing eqs. (7.19) into (7.16) yields the definition equations
for the displacement thicknesses:
Mass-flow displacement thickness
61
Ct. 1 A1
J
_ 3
(7.20) [ k01 1 Pe ve M dx ) I ,1 +
0
61
A2
+ [ k01 (d 2
1
_
f
0
Pe v e M dx 3 ) I ,2 0
where
(7.21) n = 1,2
62 1
+ [ k01
(d 12
2
_ Jx Pe ve
A1 A2
v e M dx 3 ) I , 2 +
0
62 1
+ k01 k11 [ d 11
2
- J x Pe (~1)
e
2 M dx 3 1 +
0
62 1
1 A1 A2
+ k01 k12 d 12 - '!
2
J x Pe v e v e M dx 3 +
0
'"1 '" 2 3
+ P eVe ve M dx ) 1 +
- 76 -
62 2
+ k01 k13 [
f:,22 _
2
J x Pe (~:)2 M dx3 ] 0
0
where
<5
Aa A6 Aa AB
(7.23) f:,aB = (Pe v e ve - P v v ) M dx
3
a,B 1 ,2
2 J
0
A1 A2 ·3
(7.24) PeVe ve M dx ) ] , 1 +
62 2
+ [k 01 (f:,;2 - Jx Pe (~:)2 M dx 3 ) )'2 +
o
62 1
+k01k21 [f:,~1_ JX Pe(~~)2Mdx3] +
o
+ +
- 77 -
(6 1 _
°3 E Al
ve M dx 3 ) 1 , 1 +
(7.25) [ kOl 3 f
0
°3 E vA2 M dx 3 )
[ kOl (6 23 -
+
0
f e 1 ,2 0
where
cS
Aa Aa
6 a3 ) M dx 3
(7.26)
f
0
(Ee ve - E v a 1 ,2
(7.27) it 1 ,2,
o
such that the equations are first solved for the quantity 6t • How-
ever, in contrast to the situation in first-order boundary-layer
theory, it is not evident for quantities like the momentum flux
and even the mass flux, how the reference quantities have to be
chosen such that a unique value for cSt results. Instead it is sus-
pected that different values cSt will result according to the value
of a in eq. (7.27).
- 78 -
(7.28) a 1,2 M -
There is still the question how to find the external inviscid flow
field v: (x 1 , x 2 , x 3 ), if the inviscid flow is not provided numeri-
cally by means of a field method but if only v: (x 3 = 0) is given.
In potential flow the condition of irrotationality can be used to
construct locally the necessary data by taking into account the
surface curvature of the body (see e.g. [22,29]). For supersonic
external rotational flow TASSA, RESHOTKO and ANDERSON [83] use a
coupling procedure with the field solution suggested originally by
FERRI and DASH.
- 79 -
8. Thin-Layer Approximations of the Navier-Stokes Equations
8.1 Introduction
(8. 1 ) o
(8.2)
(8.3) p c p (T , t + vi T , i )
l' i
(8.4) g J [( -p + A 0) <5 1 +
- 80 -
The covariant derivative of the stress tensor is
+ II { gmi [( v P, + v t { P }) { j }
... m m t j p -
(8.6)
In equations (8.5) and (8.6) the symbol). denotes the second vis-
cosity coefficient which is related to the bulk viscosity IL' in the
following way [ 31
2
(8.7) IL' - '3 IL
- 81 -
and the symbol 0 is
(8.8) o v~m + vk { m }
m k
Note that all (latin) indices in equations (8.1) to (8.8) run from
one to three, when the equations are written out explicitly.
(8 . 9) a
P't+(pv)'a+ a } +p v 3 { a3
Pv y { ay a } +(pv)'3
3 o
P
T3j I.J
Energy equation
( 8 • 1 2) (T 't + Va T 3 )
P cp 'a + v T'3
- 82 -
+ gaa k T, Q
'a IJ
Taj ga j 3
(8.13) (- p + A E/) + fJ. [gay (6 j v n'y + .sj V'y) +
n 3
and
3 + gC j
+ fJ. [.s Jn' vn'3 +.sj3 v'3 3 3j 3 1
v'c + g v'3
+
(8.15) Tllj gaa (- p + A e ), a +
j
- 83 -
({-e-} { B} {-3-} { B })
£Y Be + £Y 83 -
a {-3-} _ v Y
- v/3 I) B
and
(8.16) T3j I.
]
£
+ v
- 84 -
3 3
- v 'y - v, 3
(8.17) ~+
+ v3 vn n/ 3 } {yOn} + {-B }
a n {y03})
+
(v3) 2 {aa } 3
+ 3 {/3} + 2 v n va'a {yOn} + 2 v va'a {/ 3}] +
3 3 e: -3- 3
[v'a V'y + v v n {a e: l { / n} + 2 v n v 'a. { y 3 n l] +
+ gya
+
B a e:
goa [v, 3 v, 3 + v v n {/J a
{/ n l + 2 v n v, 3 {3 On l] +
-3-
+ v y v a [{ a \} { a } + {a \} {/ a} + {3 \} {a ol] +
a a
- 85 -
and e+ is given by
(8.18) e+ Y 3
v'Y + v'3 + v
O{YI
Y of + v
3{Yl
Y 3J
Due to the splitting of the latin indices into greek ones and 3,
equations (8.9) to (8.18) appear to be more voluminous than the ori-
ginal equations, but, in fact, the number of terms is considerably
reduced due to the reduced metric tensor (A.28). This is reflected
in table 8.1. Table 8.1 compares the number of terms in the gover-
ning equations for general curvilinear coordinates with those for
locally monoclinic surface-oriented coordinates, and with those
for Cartesian coordinates. These numbers result if all differen-
tiations are carried out, and if only algebraic factors are taken
as one term. Furthermore, Christoffel symbols are counted as one
quantity. From the numbers, displayed in table 8.1, it becomes
clear that a solution of the Navier-Stokes equations (including
continuity- and energy equations) is evidently impracticable for
general curvilinear or even locally monoclinic surface-oriented co-
ordinates.
- 86 -
Table 8.1: Cacparison of the number of tenns in the governing equations with respect to different coordinate systems
- 88 -
laminar or turbulent (then solving the time-averaged form of the
equations) is given. Most investigations aim at predicting steady-
-state flows although, recently, it became feasible to integrate
simplified forms of the time-dependent equations at reasonable cost
for research purposes [111,114,116,125,126). The methods cited are
based on time-accurate finite-difference schemes, and therefore
have the potential to compute unsteady flows for appropriate initial
and boundary conditions. In general, however, these methods are used
to obtain the steady-state solution for large times in an asympto-
tic manner (transient unsteady approach) because large time-steps
may then be taken since the evolution of the solution with respect
to time is of no interest unless the rate of convergence is affected.
In most cases the transient unsteady approach is still too time-
consuming to be attractive for practical purposes, and the same
holds for the alternatives for steady-state solutions, namely global
iterative methods. Therefore, to date approximate equations, which
are loosely called "parabolized" Navier-Stokes equations, are pre-
ferably employed since these result in a marching procedure for the
integration. Note, that the notion "parabolized" does not necessari-
ly mean that mathematically the equations are of parabolic type.
The marching direction must coincide with that streamwise direction
where separation is absent, and the viscous diffusion with respect
to that direction must be negligible in comparison with those in
the transverse directions. The loss of ellipticity in the stream-
wise direction can, however, lead to stability problems if the
streamwise pressure gradient is included in the solution algorithm
as is discussed in [109,128,131). The work by WALITT and TRULIO
(133), which is one of the earlier developments, is similar to the
space-marching methods, but requires a further restriction namely
that the body be slender and that the velocity component in the
direction of the axis of the body therefore remains approximately
equal to its free-stream value. Using HAYES' equivalence principle
they transform the steady three-dimensional problem into a time-
dependent two-dimensional problem where time replaces the axial
marching direction. Applications of the usually employed marching
techniques, which differ from each other essentially in the way
the pressure gradient in the marching direction is being treated,
are given, for example, by [107,109,110,113,119,120,121,124,128,
- 89 -
127,131].
- 90 -
in context with the transient unsteady approach. The form of the
steady-state thin-layer Navier-Stokes equations resembles higher-
order boundary-layer equations; the boundary conditions far away
from the surface are, however, quite different, since there the
thin-layer Navier-Stokes equations reduce essentially to the Euler
equations, and require appropriate boundary conditions.
- 91 -
oriented coordinates, the following thin-layer Navier-Stokes equa-
tions result. Note, that here the boundary-layer stretching is being
omitted, and that the pressure p is being referenced with a refer-
-
ence pressure Pref' a scaling which is usually used for computa-
tions of compressible flows. Furthermore, the metric factors k ,
mn
which are the Christoffel symbols of the second kind, contain all
metric information required for the problem in consideration. The
bar over the metric factors indicates the use of the g-metric that
- - 1 2 3 a *a 1/2 3 *3
is k mn k mn (x , x , x ), and v = v I(g(aa» , v = v
Continuity equation
1 2 3
(8.19) kOl P't + (k 01 P v ), 1 + (k 01 P v ), 2 + (k 01 P v ), 3 O.
a
Navier-Stokes equations for x -direction, a = 1 ,2
a 1 a 2 a 3 a
(8.20) P [v't + v v'l + v v'2 + v v, 3 +
1 2
+ k
al
(v 1) 2 + k
a2
v v + k a3 (v 2 )2 +
1 3 2
+ k v v + k v v3)
a4 as
1 a
(k a6 P'l + k + (I! v, 3) , 3
)2 a7 p, 2) ~
Yref (M ref ref
3 1 3 2 3 3 3
(8.21) P [v't + v v'l + v v, 2 + v v'3 +
(v 1 )2 + 1 2 (v2) 2)
+ k31 k32 v v + k33
1 3
2 P'3 + ~ (I! v, 3) , 3
(M ref ) ref
Y ref
1 v2 3)
(8.22) c p P [ T't + v T'l + T'2 + v T'3
- 92 -
+
Pr ~ (kT'3)'3
ref ref
1 2 3
P't + v P'1 + v p, 2 + v P'3
+ E +
ref
+ --~-
1 2 1 2 2 2
Re ref Ik 41 (v'3) + k42 v, 3 v, 3 + k43 (v'3) 1
The additional metric factors kmn are defined as follows (see also
append ix A. 1 3) :
(8.23)
(8.24)
1
(8.26) k25 g (g 11 g22'3 - g12 g12'3)
1
(8.27) k31 - "2 g11'3
1
(8.29) k33 - "2 g22'3
- 93 -
equations is restricted essentially to a very thin boundary-layer
sheet, e.g. in a combined Euler-Navier-Stokes computation, it might
be sufficient to use the a-metric, that is to replace the general
metric factors k- mn by k mn = k mn (x 1 , x 2 ) which depend upon the
metric tensor of the surface only. If,however, a global application
of the Navier-Stokes equations is intended, the g-metric must be
retained in general form (k mn ).
@
const.
X 1 X 2 :=
,
-+--+--1---4 x '' x 3=const.
- 94 -
prevent the locally monoclinic coordinates from converging and
intersecting (see also chapter 1). For the case of a convex corner,
Fig. 8.1, case b), the surface-normal coordinates diverge away from
the surface, and there may be flow cases which require an additional
outer coordinate system in order to provide the resolution necessary
for the flow field.
(8.30) 1 1 1
-
Re- - [(Il v'2)'2 + (Il v'3)'3]
ref
1 2
(8.311 - v'2)'2
Re- - (Il
ref
For eq. (8.21) the above remark also holds so that here the equa-
tion can remain unchanged. In the energy equation (8.i2) not only
the viscous, but also the heat-conduction terms are different now.
The viscous terms become
(8.32)
1
(8.33)
Pr ref Re ref
- 95 -
Note again, that the form of the equations given in this section
is somewhat arbitrary and represents suggestions, since the equa-
tions are derived heuristically. Depending on the flow field under
consideration, it might become necessary to include additional terms
to improve the prediction. It is also.noted that neither the poten-
tial nor the shortcomings of thin-layer approximations are thorough-
ly researched. Investigations for two and three dimensions, conduc-
ted at NASA-Ames Research Center, are, however, very encouraging
concerning the thin-layer approximation discussed in chapter 8.2.1
[116,117,118,125,126,130,131].
For general cases, which occur if the viscous effects in all three
spatial directions are of the same order of magnitude, the above
ideas can be extended further. In general such a situation will
exist only locally, e.g. at the leading edge-tip junction of a wing,
at a wing-fuselage junction or near highly skewed separation lines.
Suppose that the local characteristic lengths in all directions are
of order E, with E < 1, and that all metric properties as well as
the velocity components are of order 0 (1). Then all second-order
viscous terms are of order 0 (1) larger than first-order terms, and
E
of order 0 [(1)2] larger than second-order terms. This situation
E
can be used to derive an appropriate approximation of the Navier-
Stokes equations.
- 96 -
ticity is used to determine length scales in turbulence models, see
e.g. [108]. In the case of flows with separation the displacement
thickness with its usual definition, resulting from an integration
procedure from the surface to the boundary-layer edge, yields length
scales too large for the use in eddy viscosity turbulence models,
as was shown for two-dimensional flows [112]. These difficulties
are increased for three-dimensional flow-field predictions, since
there only the definition of the displacement thickness can be ex-
tended from two to three dimensions in a straight-forward manner,
see chapter 3 and [ 1] . Another problem is the definition of the
wall-normal distance and the boundary-layer thickness, which are
needed for eddy-viscosity models, in the case of corner flows [115].
VISCOUS
flow
strongly viscous
flow
- 97 -
In first-order boundary-layer theory, see Fig. 8.2, case a) which
is sufficient at high Reynolds numbers in most cases (see chapter 9),
the tangential viscous velocity component v*, (x 3 ) exhibits vanishing
wall-normal derivatives an v*'/(ax 3 )n = an v'/(ax 3 )n (because a" ;
a" (x 3 » at the boundary-layer edge o. The latter usually is de-
fined by the location
(8.34)
,
due to the displacement effect of the boundary layer, and there-
fore also the ve-component at the displaced surface. This effect is,
except for supercritical transonic flows, usually negligible.
Consider now case b) of Fig. 8.2 where schematically the real velo-
city profile in the neighbourhood of a surface is given. If the ex-
ternal inviscid flow is irrotational, the edge of the boundary layer
can be defined by investigating the magnitude of the tangential co-
ordinate of the vorticity vector (see appendix C, eq. (C.4) with
g-metric) :
(8.35) 0:
w
2
I-
Again only the two-dimensional case is considered.
For high Reynolds numbers the term v~3 in eq. (8.35) - v~3 = 0(')
because of the boundary-layer stretching applied in eq. (8.35) - is
the dominant term. In the frame of boundary-layer theory the boun-
dary-layer edge could be defined therefore as the point where
- 98 -
*1_
v'3 = O. The edge defined by eg. (8.35) lies above this point, be-
cause of the other terms in eg. (8.35). How far the point where
*1
v,) = 0 can be used as approximative criterion for the boundary-
-layer edge in Navier-Stokes solutions cannot yet be decided.
- 99 -
9. On the Influence of Surface Curvature
- 100 -
face to be computed at every location (x',x 2 ). There exist quite a
few cases where zero-order theory predicts reasonable results, see
e.g. [42,43,591. For flows past bodies like fuselages, etc. zero-
order theory cannot be applied, of course, since at least the trans-
verse curvature must be considered.
(9.1) 0 (e) «
,
0 ('6) ,
a-metric with Qc (X',X 2 ) - 0
(9.2) 0 (e) « 0
a-metric with Qc
together with
- '0' -
(9.4) I) o (1) < 0 (D) < 0 (1) ,
IS
and
(9.5)
(9.6) 0 (e) 0 (1 )
IS
0 (D) 0 (1)
IS
g-metric with u Uc (x 1 , x 2 )
c
- 102 -
9.2 Boundary-Layer Flows at the Leading Edge of
Swept Wings
3. Infinite-swept-wing flow
In Fig. 9.1 the planform and the cross section of the wing under
consideration are given together with the generally used non-ortho-
gonal (a) and additionally with orthogonal (b) surface-oriented co-
ordinates.The representation of the surface of the wing in the Car-
tesian reference coordinate system in terms of the surface para-
meters is for case a) (all lenghts have been non-dimensionalized
with Lref ) :
- 103 -
l'
X
I
_2_2
X,X
LX,..,,2-L
- X2-5
-
1'
X
Lx' =C05 ~ C
Fig. 9.1 Schematic of a swept wing with non-orthogonal (aI,
and orthogonal (bl surface-oriented coordinates.
- 104 -
x1
(9.7) x1 ' tg "'0 s x2 + c
0
J cos a c (f; 1 ) d f;1
2' 2
x s x
x1
x 3'
cJ
0
sin a
c
( f; 1 ) d f; 1
x- 1
(9.8) x1' tg "'0 s ;C2 + cos 2"'0 cf cos a c (~1) d ~1
o
-1
J
x
x
2'
s x-2 - sin "'0 cos "'0 c cos a c
o
-1
J
x
x3 ' cos "'0 c sin a c (~1) d ~1
o
where the tilde is used here and in the remainder of this chapter
to distinguish between the two sets of coordinates, and not, as
usual, to indicate boundary-layer stretching.
The transformation matrices for case a) and b) follow from the dif-
ferentiation of eqs. (9.7) and (9.8) (see chapter A.4 and also
A.3, appendix A). For case a) one obtains:
(9.9) sin a c (x 1 )
cos a c (x 1 ) c tg "'0 s
N
tg '" sin a (x 1 )
B 0 s o c
N
cos a (x 1 )
sin a (x 1 ) c 0
c
c N
- 105 -
and
(9.10)
cos a
c
N N
where
(9.11) N [1 + t 9 2 ~o Sln
.2a (x 1 )]1/2
c
(9.12 )
cos
2
~o c cos a
c
(~1 ) tg ~o s - cos ~o sin a
c (~') \
c cos a sin a (~, )
- sin c (x, ) s
B ~o cos ~o sin ~o c
and
- 106 -
(9. '3)
-cos 11>0 sin ac (~') sin 11>0 sin ac (;,) cos ac (;,)
Consider next the relation between the nose radii rN and rN with
respect to the two different surface-oriented coordinate systems,
which requires the relation between two corresponding contour angles
a c and ac • This can only be achieved in an approximate way, the re-
sult, however, turns out be exact for special configurations, in
particular for the leading edge of a wing.
and in xl-direction by
- 107 -
x 2'
x
"
(9.17) Qc'1
-108 -
(9.18) «c'l
and one obtains further with the help of the corresponding elements
of the transformation matrices (9.9) and (9.12), after use has been
made of equation (9.16):
cos ljIo
(9.19) «c'l ------~--------~~----~--~~-
(cos «c + cos 2
2~ . 2«~) 3/2 aC'l~
ljIo s~n .c
Taking the limit «c ~ n/2 at the leading edge, and introducing the
physical lengths of the coordinates yields the relation of the cur-
vatures at the leading edge
(9.20) «c'l*
(9.21) rN
With the geometric properties of the wing given the well-known re-
sult for the thickness of the laminar incompressible boundary-layer
at the attachment line of an infinite swept circular cylinder
(see e.g. ROSENHEAD [10] or SCHLICHTING [11]) is employed for the
derivation of the criterion:
(9.22)
- 109 -
(9.23) ;5
\! \! \!ref
,,*1 -1
x x L-1 L
x ref
rN rN L ref
(9.24)
-1
In order to express the velocity gradient ve'l in terms of the gra-
dient v:'l in the non-orthogonal xa-coordinate system, Fig. 9.1 a),
the wing is considered as having no thickness, since the transfor-
mation of the derivative for a real wing is very tedious. With
ac a c - 0 in eqs. (9.9) to (9.13) and using the infinite-swept-
wing assumption a/a x 2 = a/a x
2 = 0 as well as the condition that
the leading edge is a line of symmetry, one obtains:
-1 1
(9.25) v v
e'l e'l
c, 1
(9.26) <5
IRe
ref Iv!, 1
where the constant c 2 takes care also of the error associated with
the assumption of flatness of the wing.
- 110 -
(9.27)
(9.28)
rN ~ (!)2
c c '
(9.29) 6
- -
rN
Eq. (9.29) contains the usual requirement for fixed geometry, name-
ly that the reference Reynolds number be large, since the boundary-
layer thickness decreases with increasing Reynolds number. Further-
more, the thicker and the larger the wing, the larger the nose ra-
dius and the acceleration at the leading edge, the smaller there-
fore the ratio 6/ r N . The effect of the angle of attack is also in-
corporated in formula (9.29) via /v~'1. The appearance of the sweep
is due to the effective chord c cos ~o for infinite swept wings.
Fig. 9.3 shows, with c 2 cos ~o/c assumed to be unity, that for Rey-
nolds numbers larger than 10 6 first-order boundary-layer theory
for laminar flows is valid for nearly any value of the remaining
parameters.
- 111 -
arises if the flow is already turbulent at the leading edge due to
leading-edge contamination [34,70]. No solution for the turbulent
boundary-layer thickness equivalent to that for laminar flows, eq.
(9.22), is known to the authors. Therefore no estimation canbemade
for turbulent flows as to when second-order theory should be con-
sidered, and, more important even, when curvature effects have to
be incorporated into the turbulence model for such three-dimensio-
nal flows.
1
tic
"e~,
,
,,-, D 50 0.05
"""
+ 100 aDS
,~ ~ <> 150 aDs
0.1 t:. 50 0.10
, , 0 100 0.10
"""
.,""'" 150 0.10
~ " '"
I>
"""
I'.. "
50 a15
tom ~"
)(
~ ~ ~~ ~ a 100 0.15
V 150 0.15
~
~
,
'"
~""
, , .....
l"..'
-- r-- ~-
0.001
""" ~"
~ ~ ~~""" ~
..... ..... ,,,...... ~ .......
-~~
"'"'" ~"
~~
"'
"""
.
I I ~~
0.0001
-. ~~. ---t- - ~
- ~
.~
Reref::----..
~
- 112 -
9.2.2 Results for Zero- and First-Order Theory
2
(9.31) (cos 1P0 c) o
aae
0
For the flat wing with a c - 0, the metric tensor becomes for case a)
- 113 -
For the laminar flow near the leading edge of a swept tapered wing
results for zero- and first-order theory are compared with each
other in Fig. 9.4. For a cross section near mid span the location
of separation on the suction side is given for zero-order (a = 0)
1 c
and first-order (a c = a c (x )) theory for several angles of attack
[37]. The experimental results [74] show the existence of a sepa-
ration bubble for angles of attack larger than 5 degrees. The zero-
order-theory predictions locate separation somewhat more downstream
which is in accordance with experience in two-dimensional boundary
layers [29,31]. These results are in remarkable agreement with
those for first-order theory although the flow in the experiment
is very complicated (compressible; flow unstable at xl ~ 0,02;
three-dimensional bubble separation). For a = 15 0 larger separation
zones exist already near the leading edge resulting in a wrong pre-
diction of the inviscid flow field which is documented by the dif-
ference of the predicted and experimental locations of the attach-
ment line in the left portion of Fig. 9.4 [37]. For practical pur-
poses the difference of the predicted separation locations is neg-
ligible. Since bubble transition occurs, the question of bubble
length and initial conditions for the boundary-layer computation
after turbulent re-attachment is much more important [23]. These
initial conditions influence strongly the prediction of the now in
general decelerated turbulent boundary layer and may govern comple-
tely a calculated vortex-sheet separation location near the trai-
ling edge of the wing.
Further away from the leading edge the influence of surface curva-
ture is negligible in laminar-flow calculations [36,37]. A study
of two-dimensional flows by GRUNDMANN and ROBERT [31] indicates
that second-order effects are small even if the stagnation point
on the lower surface of the airfoil is markedly away from the
leading edge.
- 114 -
lower- upper side of wing
laminar separation
(sep. bubb/~J
<:> experiment
o
5
o attachment line
experiment
..
0 0
-QO' -0.02 0 0.02 0.0' 0.06
x,
J' 2' 2
X x I X x' = const.
- 115 -
9.3 Some Remarks on General Cases
- 116 -
10. Samples of Surface-Oriented Coordinate Systems
- 117 -
10.1 Surface-Oriented Wing Coordinate Systems
x2 =const.
Trailing edge (T E)
t---------s
x"
- 118 -
2 2'
that the x -distances are measured along x and not along the
2 1
x -coordinate lines (lines for which x =const.). If the percent
lines are straight lines, x 2 divides equal portions on s and on the
percent lines*).
(10.1)
(10.2)
X 3'
l'
XL.E. X 7'
- 119 -
on the lower side (Fig. 10.2). These lengths are taken to be the
local normalizing lengths for xl (lengths in chordwise direction).
It is, of course, possible to measure these lengths, including the
xl-coordinate, from the attachment line, which divides the flow on
the upper surface from that on the lower surface, and which
therefore naturally separates the corresponding domains of inte-
gration for the flow equations (see discussions in chapter 5).
This, however, would require to compute the metric properties anew
whenever the angle of attack changes.
2
(10.4) 0, x ) ,
3I 1 2
(10.5) x LE (x oI X )
(10.6)
J
xl
(10.7) Xul' (x 1 ,x)
2 = x l' (O/X)
LE
2 + Lx1 (x)
2 cosa c (~ 1 ,x 2 )df; 1 I
u 0 U
(10.8)
- 120 -
3' 1 2
(10.9) Xu (x ,x )
l'
(10.10) x TE (x 1 = 1,x 2 ) l'
- x LE (x 1 = O,x 2 )
In the following the geometry for only one side of the wing, the
upper side, will be considered. The index u is therefore omitted.
ax 1 ' 1 2
(10.11 ) 15 11 ' -1- (x 1 ,X 2 ) Lx 1 (X 2 ) cosac(x ,x )
ax
2' ax2' (x 1 ,x 2 )
(10.12) 15 1 - ·0 ,
ax 1
3' 3' 1 2
(10.13) 15 1
ax (x 1 ,x 2 ) Lx 1 (x 2 ) sinac(x ,x )
~
- 121 -
1 2 1
cosa c (~ ,x ) d~
ax 2 I 1 2
(10.15) - - (x ,x ) -
ax 2
3I 2
(10.16) (x LE (o,x )) +
ax 2
The coordinates B~' of the third base vector can now be found
following appendix A.3. Then all metric properties and trans-
formation matrices can be calculated as shown in appendix A.
(10.17)
cR -
x1
(10.18 ) X
1I
[c R- (cR-c T ) x 2 ) J
o
2I
(10.19) X
- 122 -
xl
3'
(10.20 ) x [c R-(c R-c T )X21! sl.·na(~l) d~l
<, <,
.
o
The coordinates of the base vectors ~1 and ~2' eqs. (10.11) to
( 10. 16) , become
(10.21)
2'
(10.22) 13 1 0 ,
3' 2
(10.23) 13 1 [c R - (cR-cT)x 1 sina (x 1)
c
1
x
1'
(10.24) 13 2 =tg"'LE s - (cR-c T ) ! cos a (~l)d~l
c
0
2'
(10.25 ) 13 2 = s ,
x':: 0
v'
")(1=1+£
x ::const.
X, =const.
~----------------s----------------~
X l'
- 123 -
xl
('0.26) (cR-c T ) J sina c (E; 1) dE; 1
o
(10.27)
1
x
(10.28) = (s) 2 { 1 + [tgCPLE -
1
---
s
o
J
+ [-
cR-c T
s
Jx
o
(10.29)
(10.30) CR:c T J
x
3I 3I
x R -x T
1
(10.31) tg E = tgE (x )
s/coscp
P
(10.32)
- 124 -
s 2
(10.33) a 22 = (COSf/l casE) •
p
The latter expression is the square of the length of the coordi-
nate line x 1 = canst. on the wing surface. This length relates
the x 2-coordinate lines to physical lengths (see chapter 2).
Keeping in mind that
(10.35) s
Lx 1 cos lP CaSE cos3 ,
p
and
(10.37)
The components of the third base vector ~3 resul ts from eqs.. (A. 1 J)
to (A.15):
(10.38)
- 125 -
tg ~LE sin a c (xl)
(10.39)
N
(10.40)
N
with N [ 1 + tg 2 ~LE
.2 a
S1n (x1»)l/2.
c
o
(10.41) b
at!
o
From eqs. (A.41 and (A.42) one gets the two principal curvatures
of the surface:
(10.42)
+ tg 2 ~LE da c (xl)
(N)3 dx 1
The directions of the two principle curvatures are with eq. (A.46)
(10.43) dx' o ,
- s
- '26 -
The contravariant metric tensor (A.48) of the surface of the swept
cylindrical wing becomes:
tg ~LE cos ac (x 1 )
s (N)2
(10.44) aa s
/
tg ~LE cos a c (x 1 )
s (N)2 (s N)2
2
1 + tg ~LE
(N) 3
(10.45)
b1 0
2 '
b 22 0 •
X
3 1 da c (x 1 )
(10.46) 1 - 2 - -~"7"-- +
N dx1
- 127 -
(10.47)
2
(10.48) (s/cos Cj) LE) •
The g-metric thus differs from the a-metric only due to g11.
For x 3 ~ 0, a 11 is recovered.
2' 2 1,
(10.49) x cos Cj)LE x , Lx 1 L 2
x
The metric tensor for the infinite swept wing surface follows
then from eq. (10.37) with
- 128 -
(10.50)
cosa c (x 1 )
(10.51)
( 0 :1
Note that in this case the surface properties no lonqer appear
in the metric tensor. The surface curvature tensor b ae of course
does not become trivial, and the metric mixed variant curvature
tensor and the g-metric are non-trivial as well (see also chapter
9) •
- 129 -
In the third sub-chapter the coordinate system used by BLaTTNER
and ELLIS [15] is briefly sketched which does not need a
separate coordinate system for the nose region, but which needs
much computational work similar to the external streamline
coordinate system. For a body of revolution at an angle of attack
of zero degree this coordinate system coincides with the second
monoclinic system mentioned in chapter 10.2.2. STOCK [82]
employs a coordinate system for the prediction of the flow past
an ellipsoid, which has also the advantage that geometrical and
aerodynamic singularity coincide at the stagnation point. In all
these cases, however, the entire coordinate systems has to be
defined anew for changing angles of attack. ,
3'
X
- 130 -
frame such that the two poles are linked by the xl '-coordinate
axis. The x 2 -coordinate line on the surface of body, i.e. lines
where xl const., are defined by transverse-frame cuts (cross
sections parallel to the x 2' , x 3' -plane) .
(10.52) x1' L 1
xl Lx 1 = xl
x 2
x
(10.53) x
2' 2' 1
Xo (x ) + Lx2 (x 1 ) J
0 2
cosa c (x 1 ,E;2)dE;2
2
J
x 2
3' x~' (x l ) sina c (x 1 ,E;2)dE;
(10.54) x Lx2 (x 1)
2
0
l' 3'
For a body symmetrical with respect to the x , x -plane, of
course, Xo2' 2 1 ) can be chosen as half of t h e Clrcum-
0, and Lx(X .
The coordinates 8!' of the base vectors ~a(a 1,2) follow from
eq s . ( 1 0 . 52) to (1 0 . 54) :
(10.55) 8 11 ' L,
2' d 2' 1 d
(10.56) 8 = - 1 (x O (x II + - -
1 dx dx 1
x2
cosa 1 I; 2 ) dE; 2 +L 2 (1)J
(x, x _a1 (COSa (x1,C2»dC2
<, <,
c2 x 0 ax c2
- 131 -
A
X
3'
~=const.
x =
2 II
x'= 1
x'=O
~
L-
® tx 3
'
x ' ::const.
x2;; OJ 1
view A-A
2'
Xo
L 2 (x')=circumferentia[ arc length
x , 2
0c
2
=Q'c (x X )
2
J
- 132 -
2
x
(10.57) 3' (L x 2 (x 1 » [ .
s~na 1 E; 2 ) dE; 2 -
(x,
B1 c2
(10.58) 1' 0 ,
B2
2' L 2 (x 1 ) cosa (x 1 ,x 2 )
(10.59) B2 x c2
In the same manner as in the case of the wing (chapter 10.1) all
metric properties can now be calculated following appendix A. Note
that, in general, the coordinate system will not be orthogonal.
(10.61)
(10.62) wx 2
- 133 -
X3' x T =const
X2 = "
xT=O x T= 1
~f------- L -------~
X 3' 2
X=O
view A-A
/ xo2 =O
2' xJ'(x T) =r(x~
X o
L 2 (XT) =Trr(x~
)(
- 134 -
reads
1
(10.63) x 1 ' = Lx 1 x1 = Lx
(10.66) 1' L
111
1 2
(10.67) 2'
81
(0)
"2 r (x 1)
(1 - 2x 1 ) sin (lTX 2 )
3' 1 2
( 1 _ 2x 1 ) cos (lTX 2 )
(0)
(10.68) 81 ,
"2 r (x 1 )
(10.69 ) 1'
82 0 ,
(10.71) 3'
82 _ lTr (x 1) sin (lTX 2 ) .
The metric tensor then reads (underlined result for general axi-
symmetric body):
(L)2 + lQl2 (1_2x1)2
4 x 1 (1-x 1 ) o
= (L)2 + (r(x 1 ) ,1)2
(10.72)
( IT 0) 2x 1 ( 1-x 1 )
o = (J~x2 (x 1) ) 2
- 135 -
The components of the third base vector ~3' which is normal to
the surface, are determined with eqs. (A.13) to (A.1S):
with
o
(10.76 )
- 2 !1T) 2 Lx 1 ( 1--x 1 )
M
(10.77) -2L
and
(10.78)
-2L
M(D)2
20 2
and
- (L)2 o
- 136 -
and for xl o and xl 1:
2L 2L
and
- (0) 2 - (0)2
(10.79) aafl
o
and these cf the mixed variant curvature tensor, eg. (A.49), read
o
(10.80)
-2L
o (0)2 M
31 2
(l-x b 1 ) all
(10.81)
o
- 137 -
v1 +0, while the coordinate a 22 remains finite, but with
a 22 + 0 inferringv 2 + Note, however, that these singularities
00.
- 138 -
streamlines in surfaces parallel to the surface of the considered
body ("characteristic box scheme")). Then an interpolation scheme
has, in general, to be employed at the common boundary of the two
different computational domains in order to provide initial data
for the prediction of the flow on the remainder of the body. Only
in the case where the stagnation point moves to large xl, it
might be useful to employ flow dependent inviscid-streamline
coordinates for the entire body, as has been done by e.g.
GEISSLER [28) , SCH(5NAUER et al. [77) and SCHNEIDER [76).
- 139 -
®
r
X3' A
l'
X
L
A
... 2'
X
(Xc (x 2) = 7Tx 2
2
view A-A
Fig. 10.7 Definition of pure surface-oriented coordinates on a
body of revolution:
a) side view, b) cross section x' = constant.
- 140 -
(x' = 0) to the leeward pole (x' ,). The local radius r(x')
is given by
x'
(, 0.83) x1 ' Lx' J cosa
c,
' ,
IF; )dF;
0
x2
('0.84) x2 ' Lx 2 (x') J
cosa c (F; 2 )dF; 2
2
0
x' x2
1'0.85) x 3' Lx' J sina
c, "
(F; )dF; - L 21x )
x ,J sina
c2
1F;2)dF;2
0 0
('0.86)
,
trr(x ) .
('0.87) trX 2
- 141 -
2
so that the integrations with respect to x can be easily per-
formed in eqs. (10.84) and (10.85).
1I 1
(10.88) 13 1 = Lx 1 cos a (x),
c1
2I 1 2
(10.89) 13 1 = Lx1 sina c (x) sina c (x)
1 2
3I 1 2
(10.90) 13 = L sina (x) cos a (x)
1 x1 c1 c2
1I
(10.91) 13 2 =0 ,
2I 1 2
(10.92) 13 = L 2 (x ) cosa (x) ,
2 x c2
3I 1 2
(10.93) 13 = -L 2(x ) sina (x)
2 x c2
(10.94)
(10.95)
- 142 -
2' (x 1) sina (x 2 )
(10.96) 63 cosa
c1 c2
3'
(10.97 ) 63 cosa (x 1) cosa (x 2 )
c1 c2
1
da (x)
c1
L 1
x o
(10.98) baB
o cos
(10.99)
o
- 143 -
paragraph is the singularity with respect to v 2 and x 2 at the
windward pole of the body. This can be removed for flow at an
angle of attack if the origin of the coordinate system is moved
with the stagnation pOint, see Fig. 10.8.
X'=o
s X "
- '44 -
nose region of a body since the lines x 2 = const. resemble
streamlines close to the stagnation points. Since the origin of
the coordinate system moves with the stagnation point it has the
same disadvantage as the streamline-coordinate system: for each
angle of attack the coordinates and metrics have to be computed
anew. Note that since x 2 , which is an angle, is not defined at
the stagnation point, BLOTTNER and ELLIS assume that the surface
is locally planar there, in order to obtain the stagnation point
flow.
- 145 -
11. Conditions of Compatibility at the Body Surface
i
(11.1) 0: v o i = 1,2,3.
i i i
(11.2) v 't 0, v 'a 0, v , aa 0, etc.
3
( 11 .3) P't + P v , 3 o.
3
(11.4) v ,3 o.
- 146 -
Because it is needed later the continuity equation (8.9) is used
to derive a condition of compatibility for the second derivative
v 3 '33 by differentiating it with respect to x 3
(11 .5)
a=1,2.
3 1 2 1 1 1
(11. 6) v '33 + v '13 + v '23 + v '3 [p P'l + 2g .g'l] +
(11.8)
a, fl,y 1,2.
- 147 -
'j
And for T3 J j one obtains
(11. 9)
II 3
+ ~'II v '3 + ~ v '33 = 0 ,
xl-momentum equation
+ .!! v 1 , 3 (- 3 a 22 b 11 + 4 a 12 b 12 - all b 22 ) +
a
2
+ 2 .!!
a v ' 3 (a 12 b 22 - a 22 b 12 )·
x 2 -momentum equation
(11.11)
x 3 -momentum equation
(11.12) o
1 1 2 1 1
+ v '3 kOl k 01 '1 + v '3 kOl k 01 '2 f -
- 148 -
where the second viscosity coefficient A has been replaced by the
bulk viscosity ~' = A + i
~ which, in general, is set to zero.
Equation (11.6) can be used to eliminate v 3 '33 in equation (11.12):
1 2
(11.13) o ~' 1
p (P'1 v '3 + P'2 v '3)
1 1 1 21 1
IL [v , 13 + v , 3 ( ko 1 kO 1 ' 1 + 3" p p, 1 + it ~, 1 ) +
2 2 1 2 1 + 1
+ v '23 + v '3 k"- kO 1 ' 2 + 3" p p, 2 ~ ~, 2) ]
01
(11.14) o
(11.15) 1 2 1 2 2 2
~ [k 41 (v '3) + k42 v '3 v '3 + k43 (v '3) ] +
- 149 -
- .la k {T ' 1
Equations (11.6), (11. 10) to (11. 13) and (11.15) represent the
compatibility conditions at the wall as derived from the governing
partial differential equations. They are given in terms of "primi-
i
tive variables" p,p,T,v . Often it is useful to replace the gradi-
ents of velocity components by terms of wall shear stresses since
these can be measured directly while it is more difficult to mea-
sure velocity profiles near walls.
The complete stress tensor T ij for laminar flows eqs. (8.13) and
(8.14) reduces at the wall to the following expression:
1
(11.16) k16 P k17 P IJ. v '3
2
k26 P k27 P IJ. v ,3
1 2
IJ. v ,3 IJ. v ,3 - P
3 a3 3a a
(11.17) x 0: T T - IJ. v '3
(11.18) a a3
v '3 T
- 150 -
and therefore
(11.19) a a3 _.1. a3
v , 313 ~ (T '13 ~ ~'13 T )
a,13 1 ,2.
~3 ~~ ~13 ~23
(11.20 ) I.1V'33- T '1- T '2
_ ~13 (.1. 1 k
p P'l + kOl 01'1
(11.21)
~ ~13 1 ~~
0 k16 P'l + k17 P'2 T + ~ v '33 +
1.1 ~'3
-1/2 [~13
+ Re ref (3 a 22 b 11 - 4 a 12 b 12 + all b 22 ) +
a
~23
+ 2 T (a 22 b 12 - a 12 b 22 )]
(11.22)
- 151 -
x 3 -momentum equation (11.12)
{"'13 ",23
(11.24) 0 P'3'" - Re -1
ref T , 1 + T ,2 +
4 j,l'
+ "'13 [_1_
p P , 1 (3 + ~) j,l j,l, II +
T +
kOl kO l' 1
+
"'23
T
1
[r k 01 '2 + -p P , 2 (i3 + ~')
j,l j,l j,l, 2l }
01
(11.25) o
+ a pr -1
ref
{ -1
Re ref [a 22 (k T'l)'l + all (k T'2)'2 -
- 152 -
The form with stretching of the conditions of compatibility at
the surface clearly exhibits the relative orders of magnitude of
the different terms for viscous shear flows. The compatibility
conditions can, certainly, be formulated in different ways but
the given formulations seem to be most convenient for further
investigations (see, for example, the next section).
3'
X X 3'
X l'
Fig. 11.1 Definition of coordinates and of
the geometry of a cylinder •
The components of the base vectors on the surface are (see appendix
A.3) :
1 2' 3' 1
(11.26) 13 11 ' rO cos x 13 1 0 , 13 1 - rO sin x
2' 3'
13 21 ' 0 13 2 1 , 13 2 0
- 153 -
The metric tensor of the surface a reads (see appendix A.G)
as
(11.27)
(11.28 )
:1
The metric factors k .. reduce to the following non-zero terms:
1)
(11.29 ) - 1
",23 ",13 )
(11.30) o ",3 "''''
~v'33-T
",13
'1 - T '2 - T (1p P'1 - ~ ~'1
3 ",13
Re -1/2
ref T
rO
",23 2 -1 ",23
(11.32) 0 - p, 2 - ~
~'3 '" T + ~ v "'''' -
'33
Re ref
rO
T
- 154 -
o 1 !!:. ') ["'13 ",23
(11.33) P'3'" ref {(-3 + Il
+ Re -1 1: '1 + 1: '2
-1
p, '"3 - Re ref {"'13
(11.34 ) 0 "'23
'[ ,1 + 1: '2 +
+
"'13
1: [ (i +
Il')
- Il, 11 +
3 Il P P'1 Il
"'23 [ (i !!:.') 1
+ 1:
3
+
Il P P'2 - i1 Il, 21 }
(11.35) 0
+ Re -1/2
ref
- 155 -
12. Local Topology of Three-Dimensional Separation and
Attachment Lines
- 156 -
The converse of separation lines are attachment lines, which occur,
for example, near or at the leading edge of wings, or in the sym-
metry planes of bodies, and which originate from the stagnation
point, a nodal point of attachment in general. It is typical for
such attachment lines that the density of both, external inviscid-
-flow streamlines and limiting wall streamlines, is extremely high,
and that the flows exhibit a strong divergence. In contrast to such
simultaneous behaviour the external inviscid flow at separation
lines does not usually reflect immediately the presence of three-
-dimensional separation. An important special case occurs if the
flow is symmetrical with respect to the attachment line. Then the
attachment-line boundary-layer is a quasi-two-dimensional one (see
chapter 5.1). In general, however, the leading-edge attachment line
on wings is slightly curved, and therefore the inviscid and the vis-
cous attachment lines differ only slightly as well. Note that cur-
vature means here and in the following the local curvature within
the plane tangential to the body at the point under consideration.
- 157 -
12.2 Local Topology of Separation Lines [ 92]
3'
X
S1=const., £3= 0
2'
X
X l'
Fig. 12.1 Schematic of coordinates used for the study of separa-
tion lines (~i: surface-oriented non-orthogonal; xi:
surface-oriented locally orthogonal; xi: local Carte-
sian coordinates) •
- 158 -
series expansion can be performed using the surface-compatibility
conditions of chapter 11 and the transformation and series expan-
sion rules provided in the appendix. Because of the large effort
involved if the curvilinear, locally orthogonal xi-system is used,
a Cartesian xi-system is introduced (see Fig. 12.1), which has the
same properties as defined above. Its xa-plane, however, is the
plane which is tangential to the surface at Po' Therefore the vali-
dity of the results of the investigation is restricted to the im-
mediate neighbourhood of the separation line. It is then consis-
tent to take into consideration only the linear terms of the series
expansion. Hence the series expansion for the velocity reads (note
that because of the Cartesian coordinates the velocity and wall
shear stress coordinates have physically correct dimensions) :
(12.1) (x i ) = - {1- [, a3 x 3 +
J.1
a3 J.1, 1
+ ,a3) x 1 x 3 +
h'1 J.1
a3 J.1'2 ,a3) 2 3
+ (, , 2 x x
J.1
1 13
+
23
"2 +
p, 1
(- -)
J.1'1 , 13 +
2J.1 £t'1 p J.1
+
P'2
(-
P
- J.1'2
-)
J.1
,23] (x3) 2 + ••• }p
i 1 ,2, 3
0
a 1 ,2 •
The wall shear stress coordinates and the derivatives of p, p and
J.1 are obtained by transforming the values measured or computed in
the original surface-oriented ~i-coordinates. Included in eqs.
(12.1) and (12.2) are variations due to compressibility or to a
given wall-temperature distribution. Because of the appearance of
J.1'3 in eq. (12.1) knowledge of turbulence properties is required
for turbulent flows.
- 159 -
Eqs. (12.1) and (12.2) were used first by OSWATITSCH [100) in order
to study three-dimensional separation at singular points on the
surface, where the total wall shear stress vanishes. For two-dimen-
sional separating flows the separating streamline was shown to
leave the surface at a finite angle A [100]:
13
3 1:'1
(12.3) tan A
P'1
(12.4) tan A2
streamline assumed
to leave the surface
separation line
- 160 -
From egs. (12.1) and (12.2) with x 1 o the following relation is
obtained:
1
(12.5) tan >'2 - 2" 23 + 23 2 1 x3
T T'2 x - 2" P'2 P
o
13 T23
+ T'1 + 2 iT
(12.6) tan >'2
23
T
(12.7) tan 1/I w
""""i3
T
- 161 -
Fig. 12.3 Sketch of a separation line and the separating stream
surface emanating thereof.
(12.8)
2 " 131
tan A = - - -
I
1 P'l Po
(12.9) , a3 (x 1 ,x 2 ,x 3 - 0) =
a3 a3 Il, 1
+ - ,a3) x 1 +
T
IT'l Il
a3 1l'2 ,a3) 2
+ ("2 Il x + •.• lp a = 1,2
o
- 162 -
In Fig. 12.4 both, the pattern of the skin-friction lines as ob-
served in experiment, and the distribution of the skin friction in
the neighbourhood of the separation line are given. In part a) the
region to the left of the separation line is called forward flow
region, that to the right of it backward flow region.
@ 1-r:I-minimum line
ge
Such designations apply locally only because the flow on both sides
may originate from the forward stagnation point. Note that the di-
rection of the external inviscid streamline (ge) may differ con-
siderably from that of the separation line. In part b) typical dis-
tributions of the wall shear stress near a separation line are
- 163 -
sketched. It is seen that the magnitude of the total wall shear
stress 1,1 exhibits a - relative - minimum which is observed in ex-
periments, see e.g. 51], and often interpreted as location of se-
paration. In order to prove that the distribution of 1,1 has an ex-
tremum not at, but near a separation line, consider the derivative
of 1,1 with respect to xl at Po (note that it is tacitly assumed
that separation and 1,1- minimum line are parallel to each other
and only slightly curved [ 92].):
(12.10)
, 13
Now, according to the assumptions made, ,13 = 0 at Po' but not ,23
in general. Furthermore, in general, ,~~ is finite and not equal to
zero at separation, except for the flow in a plane of symmetry
x l = constant(then the flow is directed completely in x 2 -direction).
Separation occurs only if ,23 O. In three-dimensional flow this
indicates - together with ,13 0 - singular separation points
which exist e.g. in lee- and windward symmetry planes of a body at
angle of attack. Thus the derivative 1,1'1 does not vanish at Po'
in general.
(12.11)
13 1
("1 13 2) 13 + (~23 23 1 23 2) 23
x + "2 x T'1 ' + T'1 x + T'2 x T'1
P
o
- 164 -
( T 13)2 + T 13 13 23 2 23 23
(12.12) I T I , 11 Ipm ,1 T , 11 + (T, 1 )
IT I
+ T T , 11
Ipm -
2
Ipm
( ITI , 1)
IT I
13 2 23 2
(T'1) + (T'l)
(12.13)
23 23
T T , 1
13 13
T , 1 T'2 +
and the inclination of the ITI- minimum line with respect to the
xl-axis is:
13 2 23 2
(T'l) + (T'l)
(12.15) tg 1/I m Ix2 o p
o
- 165 -
in these cases, according to eq. (12.14). Because the inclination
of a skin-friction line at P reads:
m
23 23 1 23 x2
(12.16 ) tan 1jJw
• +"1 x +"2
13 1 13 x2
T , 1 x + T'2
(12.17) -
T
13
---,-:3
T , 1
I
m
- 166 -
The curvature of the separation and of any skin-friction line is
defined by (.a: local Cartesian coordinates) :
(12.18) Ww,s
a ax a a a a
(12.19 )
as as ax a
= cos 1/Iw
;;r + sin 1/I w
~
=
13 23
• ;;r
a •
+ --
a
I• I I• I ~
( 13) 2 23 + 13 23 13 _ (~23)2 13
(12.20)
• "1" - "1) , "2
13
Therefore the curvature of the separation line at Po is with • 0:
13
"2
(12.21)
-~ P
o
- 167 -
fore each skin-friction line in the forward-flow region must ex-
hibit a point of inflexion near the separation line. If the sepa-
ration line is positively curved, the points of inflexion lie in
the backward-flow region.
23 13 23 13
(12.23) a1 1: , 1 h'1 1:'2 - 1:'2 1: 23
, 1) }
P
0
23 13 23 13 13
(12.24) a2 1: [1: , 1 h'2 - 1:, 1) - 2 1:'2 1: 23
, 1] }
P
0
(12.25) a3 - 13
{ 1:'2 13 23 13 23 }
(t'1 1:'2 - 1:'2 1: , 1 )
Po
(12.26) a4 - { 1: 23 1:'2
13 23
(1:'2 + 1: 13
, 1) }
P
0
(12.27)
(12.28) 13
1:'2
- 168 -
For the special case of separation at an infinite swept cylinder
only one solution remains:
(12.29) - { 13)2 23
( T'l o
T
(12.30) 1
x ip 1,2 ={ 23
2 T'l 13 23
(T'l T'2 - 13 23) }-1
T'2 T'l
Po
.{ T
23 13 13 23 13 23
[T'1 (T'l - T, 2) + 2 T'2 T, 1 ] :!:
23 13 13 23)2 + 4 13 23]1/2}
:!: T T'l [IT'l T'2 T'2 T'l
P
0
*1 *1
(12.31) ~xm « L and ~xip « L
- 169 -
*1
where 6x denotes the physical length increment and where L is the
characteristic body length.
Swept wing s:
~2
= 0.60833
I.. I "ft~
me f-POin,' of infl~xion Ii~e
I~I-mmlmum
x"~
I I I
I--'inal separation line
0.002
provisional separation line ~XA
I'
I -,ax~
·x"
• -I
- '1'." -
1
1 Po Ps
a
r-t--. I'
2 r2
t1x;p, I e"
X,5
-0.002
t1x~" ~J
.o1X:p, I
~-'
:.I,J ~
-0.002 0.002
- 170 -
line. Taking recourse to the assumption of parallelism of all lines,
aI
a new separation line was constructed (x -system). Thus a final
approximate location of the separation line can be found via eg.
12.17. Note the very small distance between the lines. Fig. 12.6
shows the characteristic pattern of the skin-friction vectors in the
, 2
5wt'pt wing So =0..60.8; So =D.82j e = 0"
"'",,= 0.8; Rt' 7.0.9· 70 1 M~ =-0.0.0.70.7
~LE' 27.5"; ex = ,. .I1x! = 0.000.32/3.93
'P1!2
- 171 -
12.3 Local Topology of Attachment Lines
In general the attachment line of the inviscid flow near the leading
edge of a finite wing, in particular at an angle of attack, is
slightly curved, see Fig. 12.7. Here the inviscid flow, whose
ONERA M6 Wing
ex =15°
x 2'
x 2'
(ower upper
surface surface
3'
x
" X
Fig. 12.7 The attachment line at the leading edge of the finite
swept and tapered ONERA M6 wing (from [ 781).
- 172 -
streamlines are shown, was predicted by means of a panel method.
The streamlines are obtained by integrating backwards against the
flow direction, starting from points far downstream. This approach
was chosen since it is difficult to pick the streamlines in the
neighbourhood of the attachment line such that a clear spanwise dis-
tribution of streamlines is achieved. The pattern of the stream-
lines near the leading edge attachment line, and the fact that the
line is, in general, only slightly curved, suggests an investi-
gation like that for the separation line. In addition, there is a
line like the 1.1- minimum line, where the pressure attains a re-
lative maximum close to the leading-edge attachment line. For the
special case of the non-lifting flow past a symmetrical wing both
lines coincide.
1 2 3
(12.33) p P'l (x ,x,x - 0) ,
- 173 -
1 2
(v'l) + 2
2 v
(12.34) x
z 1 1 1
v'2 (v'l v'l
p p
o o
1
(12.35) x
z 1 2 1 2
(v'l) + v'2 v'l
p
o
1 2 2
(v'l) + v'2 v'l
(12.36) tg ljJz 1 1
v'2 (v'l
p
o
Thus, in general, the pressure maximum does not occur at the attach-
ment line. Assuming irrotational flow with v~2 = v~l' eqs. (12.34)
to (12.36) can be cast into the same form as eqs. (12.13) to (12.15).
For the special case of the flow past an infinite swept cylinder
with --"- =0 the location of the pressure maximum is given by the
dX 2
attachment line, according to eq. (12.35), and the inclination be-
comes ljJz = n/2, eq. (12.36). The pressure gradient for the x 2 -di-
rection vanishes, and the quasi-two-dimensional flow, discussed in
chapter 5.2, occurs.
(12.37) P'll z - p
1 2
(v'l) Ip
o
- 174 -
The curvature of the attachment line in Po is, similar to that of
the separation line:
(12.38) 1/Iz, s IP P
o o
- 175 -
quasi-two-dimensional solution of the boundary-layer equations can
be derived (see chapter 4 and 5). If the inviscid attachment line
is curved strongly and deviates from a geodesic line [ 78], the
limiting wall attachment line will be curved even stronger, and in
the same sense. The reason is the pressure gradient normal to the
inviscid attachment line which is fully transmitted throughout the
boundary layer. In regions where the external streamline exhibits a
point of inflexion, the flow within the boundary-layer follows with
a certain lag which results in the well-known s-shaped cross-flow
profiles.
The skin-friction lines near the limiting wall attachment line are
governed by the same laws as those near the separation line although
this flow is diverging and the other converging. 1.1- minimum line,
curvature and the points-of-inflexion line can be determined from
the equations given in chapter 12.2. Note again that the curva-
ture of the limiting wall attachment line will usually be of the
same kind as that of the external attachment line because sign
(.~~) = sign (V~2)' in general.
Fig. 12.8 shows the attachment lines and a selected number of exter-
nal and limiting wall streamlines for the case of a very strongly
flattened triaxial ellipsoid at large angle of attack resembling a
wing with varying leading edge sweep (from SCHWAMBORN [ 78]). Be-
cause of the scale of the drawing the small distance between invis-
cid-flow and limiting wall attachment lines is not distinguishable.
Because of the positive angle of attack both lines are located on
the lower surface. Again, in order to produce a clear picture of
the streamlines diverging from the attachment lines, points on the
surface were picked some distance downstream on the lower surface,
and the streamlines were integrated backwards towards the attach-
ment lines. Thus inviscid and limiting wall streamlines can be com-
pared which pass through the same point downstream. The integra-
tion of the streamlines of the upper-surface flow started arbitrari-
ly at the leading edge, but at the same spanwise station. For the
span station far outboard two starting points for the streamline
integration were chosen for the lower-surface flow to show that
the difference between viscous and inviscid streamlines decreases
- 176 -
the closer to the attachment lines the common point is chosen in
spite of the strong three-dimensionality of the flow. Note the re-
duction of the distance between the inviscid and limiting wall
attachment
line
B
leading
edge
---
upper
surface
-----
-----
limiting wall} "
" ""d streamlmes on lower
mVlscl surface
3'
stagnation point 5 x" A X
- 177 -
oj
30r 11111 c)
T:* 13 ~..;( I IIIII I bJ
X' :conS f. T:* 23 /T:* I
8
-...J
co "1
Il"·_M 24f- /I I I f I I I I )(7: const.
20~ 1111111 VI
.l:con st
mag nitu de
and b) ,'2 3). and of the
sica l com pon ents ( a) ,"3
Fig . 12. 9 Dis trib uti on of the phY l elli psO id
(c) iT* I) nea r the lea din g edg e of the tria xia
of the wal l she ar str ess :;-*(13 IRe ref / Pre f(V ref} 2}.
Fig . 12. 8, [78 J fT* a3 =
the flow, looses its validity if the very tip region is approached
since the nose radius becomes smaller and smaller (see chapter 9),
while the boundary-layer thickens more and more.
Fig. 12.9 shows and compares the distributions of the wall shear
stress components and of the total wall shear stress itself in the
neighbourhood of the attachment line depicted in Fig. 12.8. The
distributions are given versus the chordwise (xl) surface parameter
for several spanwise stations (x 2 ). To indicate the variation in
spanwise direction x 2 , the values along selected lines xl = const.
are presented as solid lines. In addition the shear stress along
the attachment line, see Fig. 12.8, is displayed. Note the large
variation of the wall shear in xl-direction towards the leading
edge (+ xl) in comparison to that in negative x 1 -dir-ection towards
the trailing edge on the lower surface. The h*l-plot shows clearly
the IT I -minimum line, and also that there is apparently a differ-
ence in location between the IT I -minimum line and the attachment
line, although they are quite close to each other, see [ 92].
- 179 -
13. Similarity-Type Transformation of Boundary-Layer Equations
in Contravariant Form
NASH and SCRUGGS [66,67] do not scale the dependent variables with
the corresponding values at the "edge" of the boundary layer.
Instead they use apprdpriate constant reference quantities (see
chapter 2). The scaling of the boundary-layer thickening is
achieved by simply adopting suitably varying distributions of a
fixed number of node points across the boundary layer. For laminar
flows a uniform distribution is chosen with
(13 • 1 )
-3
x ~ 0 m/M
- 180 -
across the layer of thickness n &,
while m denotes the local dis-
tance in the computational domain x 3 = m ox 3 . For turbulent flows
a non-uniform distribution is used which provides an increased
density of the mesh points in the region of large gradients close
to the surface. According to the various regions of the turbulent
boundary layer - viscous sublayer, logarithmic region and wake
region - the expression in terms of physical quantities is
(1 3.2)
-3
x (1
<I>
+
&
<p) M [c 1 (m - M) + n Ml ,
where <p = <p (m) is such that <p becomes large as m ~ M. Thus in
the outer part of the boundary layer a uniform distribution of
mesh points is obtained. In the inner part equation (13.2) is
controlled by
(1 3.3)
- 181 -
(13.4) r. x-3 /6-
f'
w + a x3
(13.5)
where x3 -3 -
x /6, a br/6 and b (0 f'w - 1)/(1 - r).
- 182 -
turbulent flows:
(13.6)
(13.7)
- 183 -
the direction of the major axis for the flow about a body at a
not too large angle of attack. If external streamline coordinates
were used, one would choose xl to coincide with the external stream-
lines. Note, however, that the definition of external-streamline
coordinates is, in general, a difficult task in particular if
the inviscid velocity components result from a numerical predic-
tion. Note also that if the similarity transformation which was
originally devised for laminar flows is used in turbulent-flow cal-
culations, it is necessary to use a non-equidistant wall-normal mesh
with geometric progression to cope with the growth of the boundary-
layer thickness. CEBECI et al. report a maximum number of about
30 points across the boundary layer for the incompressible flow
about wings at moderate Reynolds number
1
(13.8) kOl P v ""3
2
kOl P v .", 3
0 = T/Te ,
(13.9) d~
- 184 -
Re e ,x 1 = Pe v~ xl/~e' with the chosen main coordinate xl and with
the density ratio piPe' Note that, since constant reference
quantities are used for all variables (see chapter 2), the usual
boundary-layer stretching with (Re ref ) 1/2 of the coordinate x 3
evolves naturally, and that Re e,x 1 is actually referenced with
Re ref . It is assumed that non-dimensionless quantities are
used, so that the surface metric tensor coordinates and
therefore the metric factors are non-dimensional as well. The
functions W and ~ are assumed to have the form
(13.10)
Q = 1,2,
(13.11 )
13.12) f' v1 u
v e1
g' v2 v
2"
vr
(13.13 )
- 185 -
with
(13.14) ljJ , 1
<1>'2
c
r
where
(13.16 ) w = ~'1
V e1 U + ~'2
v2
r v + ~'3 v
3
- 186 -
(13.17)
Since, as usual, only first and second derivatives of the new va-
riables are to be discretised, this approach is taken here.
(13.18)
1 1 1 1
K14 c
r
(k 12 + v e '2/ v e) x , K15 k13 (C r )2 x ,
(13.20)
with
- 187 -
1 1
(13.21) K21 x Kll K22 C x K12
I
r
1 2 2 1
K23 (k 21 /c r ) x K24 = (k 22 + v r'1 /v r ) x
2 2 1
K25 c (k 23 + v r'2 /v r ) x K26 x 1 /v 1 K16
r e
Energy equation
(13.22)
V,
r,
with
1 1
(13.23) K41 (Te ' 1/Te) x K42 x K11
1 1
K43 cr (Te ' 2/Te) x K44 c x K12
r
-1
K45 X 1 /v 1 K16 K46 (Il e Pel K18
e I
K47 x 1 /T K48 c x 1 /T
e r e
1 2
K49 k41 (v e ) /(Il e P e Te)
- 188 -
Note that the quantities at the "edge" of the boundary layer are
referenced with the constant reference quantities defined in chapter
2. If CEBECI's approach would be followed equation (13.13) would be
needed in the form :
(13.24)
(13.25)
- 189 -
but the two-dimensional displacement thicknesses change as follows:
1
I,( x 1/2 0 P e
(_e_ _ )
(13.26) 01 1 1 I (P- - u) d~
x Pe v e 0
1
I,( x 1/2 0 P e
(_e_ _ )
(13.27) 01 2 1 I ( - - ~)
v
d~
x P e ve 0 P e
one obtains the equations given e.g. by LIBBY [ 53) if the limiting
process for xl ~ 0 and x 2 ~ 0 is applied to the equations derived
in the preceding chapter 13.2. The transformation relation (13.9)
yields then
- 190 -
and the dependent variables u and v, equations (13.12), change
according to equations (13.28) and (13.29):
(13.32) g' v
2 2 2 2
The notation vr and v r '2 is kept here. Although v r '2 v e '2 is a
good choice at the stagnation point since v 2 '2 = v 2 '2 results in
r 2 e 2
Ve 1, this is not so in the symmetry plane x = 0 where v = 0,
since v!'2 generally changes its sign with increasing xl while v~2
does generally not at the same time. However, one could use there
a constant, e.g. v;'2 = A; = v!'2S = const., where the subscript S
denotes the stagnation point. This choice can result in large va-
riations of V along both, surface normal and the coordinate x 2 = o.
Therefore it would also be appropriate to use a constant value v;
(the same as for the general field calculation), and to consider
~ 2 2
v = v'2/vr as new dependent variable (see equation (13.12». Here
the definition (13.32) with A; = A! as reference quantity is used
since it is then easy to arrive at the usual similarity-type form
of the equations if wanted. The dependent variables are then U, V,
wand B according to equations (13.31), (13.32), (13.16) and (13.8).
Continuity equation
(13.33) o
with
(13.34) V
(13.35 )
with
- 191 -
(13.36)
(13.37)
where
(13.38)
Energy equation:
(13.39)
where
(13.40)
(13.42)
- 192 -
13.4 Plane-of-Symmetry Flow
1
(13.43) f' --,
v
v
= u
e
2
2
v, 2 A2
v
g' 2-> -2- V
A2
vr v r '2 r
With
(13.44)
(13.45) o
where
- 193 -
(13.46)
2
( 13.47) K11 u u'1 + K13 (u) + K16 w u,~
with
(13.48)
Momentum equation for x 2-direction (see eq. (13.20), and note that
the equation has to be differentiated with respect to x 2 )
(13.49)
where
(13.50)
(13.51)
- 194 -
component is given by equation (13.26), that one with respect to
the v 2 -component by equation (13.27) if v and ve are replaced by
V and Ve. The remaining displacement thicknesses follow in a simi-
lar fashion.
(13.52)
where
(13.53)
(13.54) 2 + 2
Kll U u'l + K13 (u) + K14 U v + K15 (v) +
where
(13.55)
Pe
[ K 13 + K+
14 ve + K 15 (ve ) 2
(13.56)
- 195 -
where
(13.57)
Energy equation
(13.58)
g' v 2 /vr2 = v
- 196 -
and w and e are defined as usual (see e.g. chapter 13.2). The trans-
formation relation for the wall-normal coordinate is given again by
equation (13.30) assuming a linear variation of v 1 to hold near
1 2 2
x = 0 along x = constant. The reference quantity vr in equation
(13.59) should be consistent with the choice made for the equations
2 2
in chapter 13.5. However, in general, the choice v - v would be
1 r e
possible at x = 0, and would result in ve - 1. Note that in the
derivation of the equations in chapter 5.2.5 as well as in that of
*2
the following equations the assumption of v'1 = 0 has to be made in
order to arrive at a closed set of quasi-one-dimensional equations.
The governing differential equations read as follows:
Continuity equation
(13.60)
where
(13.61)
(13.62) (U)
2 + U v + K+ (v) 2 +
+ K14 15
where
(13.63)
and with K+
16 and K18 given by equations (13.36) and (13.19).
(13.64) K+++ 2 +
25 (v) + K26 w V't =
- 197 -
where
and K+
26 and K28 are given in equations (13.38) and (13.21).
Energy equation
( 13 • 66 )
where
(13.67)
- 198 -
Appendix A: Basic Geometrical Relations
i,j 1,2,3 .
are known (note that x j may be parameters only, and not necessarily
lenghts measured along the coordinate lines). Then the coordinates
8~ of the covariant base vectors g. of the xj-coordinate system
J -J
are determined as follows:
(A.2) a
.... J
0 = -R', J 0
~i'
i,j 1,2,3.
-i'
where Bj are the partial derivatives of the coordinates of the
position vector R'
(A.3)
- 199 -
The covariant metric tensor which is symmetric then follows as
3
(A.4) E j ,k 1,2,3,
i'=l
3
(A.5) (dS)2 -i'
j ,k 1,2,3.
i'=1 '\
(A.G)
i,j 1,2,3,
a = 1,2,
- 200 -
x ]'
X =const.
2
l'
X
i'
(A.7) a
-3 = [~' (xi' ) I , 3 = ~' , i ' x ,3 ~i'
- 201 -
surface) is uniquely defined if the corresponding point P s on the
-i'
surface and the coordinates 6 3 are known. The problem of defini-
tion of coordinates is then reduced to the definition of surface
coordinates. Where physically appropriate, the neglection of the
dependence on x 3 further eases the formulation of viscous-flow
problems as has been shown for boundary-layer problems in the main
chapters of this book.
- 202 -
thin-layer Navier-Stokes, general two-dimensional problems),
c) for surface-oriented coordinates for points P s on the surface
of the body (zero- and first-order boundary-layer problems
neglecting terms involving x 3 ). Coordinates a) are referred to as
general coordinates, coordinates b) as surface-oriented coordinates
with g-metric, and coordinates c) are called surface-oriented
with a-metric. The notation "locally monoclinic" is omitted in
the following.
a = 1,2.
Here a!' (note the difference between a!' and s!', the latter
either being partial derivatives with respect to general coordinates
(eg. (A.2)), or especially derivatives with respect to off-surface
coordinates with g-metric, eg.(A.1B)) denotes the partial deriva-
tives of the Cartesian reference coordinates with respect to the
Gaussian surface coordinates:
(A. 10)
- 203 -
The third base vector ~3 is, by definition, orthogonal to ~1 and
~2' and a vector of unit length:
(A.12)
(A.13) 6i '
3 I;i/I; ,
where
2'
1;1 61 6 32 ' - 6 31 ' 6 22 '
2'
1;3
1'
81
2'
82 - 1'
82 8,
and
(A.16) a = ',2
(A.17) ~3
- 204 -
-if
Therefore the components 8 j read as follows:
-i f
(A.19) 83 8i ' i' 1,2,3,
3
a 1,2.
1
The relations (A.18) show why the x , x 2 -coordinates get skewed
when one considers different surfaces x 3 = const. at the same
location (x 1 ,x 2 ): the cosine of the corresponding angle is given
-1/2
by cos ~ = g12 (g11 g22) , which will in general be only con-
stant for constant surface parameters xa'if ~3 = ~3 does not
depend on x a (see also page 73). This obviously holds always in
the case of flat surfaces.
If
-1 ' -1 ~ -1 '
81 82 83
- 205 -
is the transformation matrix for the transformation of the
contravariant vector coordinates eq. (A.20), one finds by inver-
sion the matrix
-3 -3 -3
13, , 13 2 , 13 3 ,
-2' -3' -3' -2' -3' -3' -1' -2' -2' -1 '
(13 2 13 3 - 13 2 13 3 ) (13 2 -"
13 3 - -1'
13 2 13 3 ) (13 2 13 3 - 13 2 13 3 )
1 (13-3' -2' -2' -3' -1' -3' -3' -1' -2' -1' -1' -2'
TBT 1 13 3 - 13 1 13 3 ) (13 1 13 3 - 13 1 13 3 ) (13, 13 3 - 13, 13 3 )
with
-, ,
+ 13 2
-2' -3'
(13 3 13 1 - -2' -3'
13 1 13 3 ) +
and likewise
(A.26)
i,j 1,2,3.
j,k 1,2,3.
- 207 -
A.6 The Covariant Metric Tensor for Surface-Oriented Coordinates
(A.28)
(A.29)
(A.30)
- 208 -
or written explicitly:
(A.3S)
(A.36)
(ds)2 1 2 1 2 2 2
(A.37) a 11 ( dx) + 2 a 1 2 dx dx + a 22 ( dx )
- 209 -
A.7 The Covariant Curvature Tensor bas of the Surface
(A.39)
12
baS (""
b 12
b
b 22
)
3
(A.40) E
k'=1
a 11 b 22 + a 22 b 11 - 2 a 12 b 12
(A.41) K1 + K2 2
a 11 a 22 - (a 12 )
2
b 11 b 22 - (b 12 )
(A.42) K1 K2 2
a 11 a 22 - (a 12 )
- 210 -
The principal curvature radii are
(A.43) R,
,
K1
(A.44) R2
,
K2
(A.45)
(A.46)
- 211 -
Here 6 u is the Kronecker tensor, that is a uB is the inverse matrix
£
of (A.30). Inversion of (A.30) yields, again, a symmetric matrix
a 11
12
( a )
IA. 48)
a
21 22
a a
21
where a
The contravariant metric tensor gaB off the surface and also the
kl
general contravariant metric tensor g are similarly defined.
(A.49) 1 ,2
11 12
IA.50) b1 b 11 a + b 12 a
1
12 22
(A.51) b2 b 11 a + b 12 a
1
11 12
(A.52) b1 b 12 a + b 22 a
2
12 22
(A.53) b2 b 12 a + b 22 a
2
- 212 -
(A.54) btl
a
A.'2 The Metric Tensor gatl and the Curvature Properties of the
Surface
(A.55)
3 2
+ (x) a
- 3 ,a • ~3,tl
a,a = ',2.
Using instead of the form (A.38) for the curvature tensor baa
another form ( 71:
(A.56) - ~a • ~3,a
- 2'3 -
allows to implement curvature properties into equations (A.55)
(b al' = b l'a ):
3 + a ya b (x 3 )2
(A.57) gas aal' - 2b as x ay b Sa
3 2
aaB - 2 baB x + b ay bYB
(x3)
a, (3, y 1 ,2.
Thus, written out explicitly, equations (A. 57) yield the desired
relations:
3
(A. 58) g 11 a 11 - 2 b 11 x + (b 11 b 1 + b 12 b~) (x 3 )2
1
3
(A.59) g12 g21 a 12 - 2 b 12 x + (b 11 b 1 + b 12 b;) (x 3 ) 2
2
3
(A.60) g22 a 22 - 2 b 22 x + (b 12 b 1 + b 22 b;) (x 3 ) 2
2
- 214 -
A.13 Christoffel Symbols and Metric Factors
(A.62)
with
(A.63)
(A.64) {~}
(A.65) {~}
(A.66) {2 1 2}
(A.67) {~}
(A.68)
(A.69) {/2}
- 215 -
(A.70) a,tl 1 ,2 ,
(A. 71)
(A.72)
(A.73)
(A.74)
(A.75) o o i 1,2,3,
(A.76)
(A.77)
- 216 -
Further the following relations hold
(A.79) -1 k-
k 01,2
01
(A.80) £2 1 _1_ k
2 2g g'3 it 01,3
01
where
_ g22
(A.82) k16 9
g'2
(A.83) k17 9
g12
(A.84) k26 9
_ g11
(A.8S) k27 9
(A.86) k31 6 11
(A.90) k42 2g 12
- 217 -
A.14 Shifters
There are two ways to derive the expressions for shifters. One way
is to consider the covariant base vectors [ 3], see eq. (A.16):
(A.93) IS Y - bY x 3
ex ex
(A.94)
(A. 95)
which yields [3
(A.96)
- 218 -
where
(A.97) M det (M a ) =
6
3
1 - bYy x + b (x 3 )2
1 - (K 1 + K2 ) x 3 + Kl K2 (x 3 )2
(A.98) v_(a) = va a
oiia
= ~Y a
-y
(A.99)
..
V 1.
and
(A. 100)
Note that ~* Y (a ) 1/2 With eqs. (A.18) and (A.19) one ob-
(yy)
tains from eq. (A.l00):
(A. 101)
- 219 -
The orthogonality relation
(A.l02)
(A.l03) - bY a
a -Y
(A.l04) 2 y • -3'a
a
are used to replace the terms within the parentheses in eq. (A.l0l):
(A. lOS)
Hence the shifter of the first kind is obtained. From eq. (A.98)
it is evident that both ways are completely equivalent: the kind
of shifter that relates contravariant components coincides with the
kind of shifter that relates covariant base vectors.
- 220 -
Appendix B: Series Expansion of Vector Quantities in
General Coordinates
(B. 1) F (~O) +
N
+ E
n=1
(B. 2) RT
(N+ 1) !
o < 0i < 1,
i 1,2,3.
k k
(B. 3) a.l-F F , i !lk + F !lk,i
(B.4) alJ-
.. F a.J (a.1-
F)
k k k Fk
F'ij !lk + F, j !lk,i + F'i !lk,j + !lk,ij
- 221 -
Higher derivatives are easily derived in the same manner. The
derivatives of the covariant base vectors ~k are obtained from
the relations [ 71:
1 nm
(B. 6) -2 g (g lm,
. k + g mk ,l. - gk'l,m ).
(B.7)
1 k g1l3
(B.9 a) aa ~ .9.1 {v~ a +
2' v (gae ,k + gek,a -gka,e)}
+
1 k 213
+ .9.2 {v~a + 2' v g (gae,k + 9 13k, a - gka, e) } +
+ ~3 {v 3 _ 1 v e gea,3} ,
'a 2
- 222 -
(B.9b)
33 ~ ~1 {V~3 + .12 vll gla
gall,3} +
2 + .12 vll g 2a +
+ ~2 {v'3 gall,3}
3
+ ~3 v'3
1 1a 3 g2a 3
(B. 10) a1 ~ (x )
-s ~1 {v~l + '2 g gla,3 v } + ~2 {~ gla,3 v },
gla 3 1 2a 3
32 ~ (~s) ~1 {~ g2a,3 v } + ~2 {v~2 + '2 g g2a,3 v },
v (x ) 3
33
-s ~3 v, 3
- 223 -
~3 ~3 ~ -1/2
+ ~3 ~x v'3 Re ref
With equation (A.61), and with the assumption that the metric
properties aal! and bal! are of order unity, one obtains for large
Reynolds numbers ~ ~ a , g 0 ~ a 0 and g 0 3 ~ - 2 b o. There-
a -a a~ a~1/2 a~, a~
fore, neglecting terms of order Re ref and less, equation (B.11)
finally yields equations (4.20) in vector form:
(B.12)
- 224 -
Appendix C: The Vorticity Vector
From the text books on tensor algebra, e.g. [ 7,12], one obtains
the following expression for the vorticity vector ROT~, after
contravariant velocity components have been introduced:
(C. 1) ROT oi
~ 9:i
9:1
(g) ~1 /2 [ (g 3 j v j ) , 2 - (g 2j v j ) , 3] +
+ 9:2
(g)-1/2 [( g V j )'3
1j - (g3jvj)'11 +
+ 9:3
(g)-1/2 [( g V j ) , 1
2j - (g 1j V j ) '2 1 ,
where 9 = det (g .. ), and where the inner brackets with the terms
1J
gkj are readily recognized as covariant velocity components v k •
The vectors 9:i are the usual covariant base vectors for general
coordinates. For locally monoclinic surface-oriented coordinates in
off-surface points equation (C.l) reduces to the following vector
equation (ga3 = 0, g33 = 1, g21 = g12):
(C.2) Rot ~ 9:1
(g)-1/2 [v 3 '2 - 1
(g12 v )'3 - 2
(g22 v ), 3 1 +
+ 9:3
(g)-1/2 [(g'2 v )"
, +
2
(g22 v )"
2
- 1
(g11 v )'2- (g12 v ) '2 1 ,
which in the frame of first-order boundary-layer theory can be
further reduced, reading in terms of the surface metric tensor
aaB (see appendix A) in surface pOints:
i
(C.3) rot ~ ~i w
V 1 2 1
- a'2 '3 - a 22 v'3 +
- 225 -
1 2
- all v '2 - a 12 v '2]
where the symmetry of the metric tensors has been taken into
7
ac ount: a 12= a 21 , b 12 = b 21 . The physical components n*i or
w*1 are obtained by multiplying the contravariant components with
the respective coordinates of the metric tensors (g .. ) 1/2 or
1/2 3 3 3 3 11
(aaa) . Note that n = n* and w = w* for locally monoclinic
coordinates since g33 = 1.
Re 1 / 2 (a 22 v 2 ,~ + a 12 v 1 , ~) } +
ref
- 226 -
(
a 12 ,1 - a 11 ,2 ) V1 (a ) 2} ,
+ + 22 ,1 - a 12 ,2 v
(c.s) -1/2
Re ref rot ~
-1/2 1 2
+ ~2 (a) [a 11 v ,~ + a 12 v ,~l
- 227 -
Appendix 0: Dimensions and Boundary-Layer Stretching
q*
(0.1) -
q
q*
q
q*/qref
c Je c
Cq q
q qref
In the table 0.1 flow variables are listed. and in table D.2
geometrical quantities are considered. Concerning geometrical
quantities one has to remember that in this book the Cartesian
reference quantities are introduced as dimensionless lenghts.
The surface parameters xa, of course, in general have no direct
physical lenght properties. Thus the coordinates of the metric
- 228 -
and curvature tensors gij' aaB and b aB etc. and derivatives
thereof are dimensionless as well because of their definitions
[21. In order to gain some insight into the dimensional proper-
ties of some geometrical quantities, it is assumed in tables
D.2 and D.3 that xii-coordinates are dimensional, with lengths
as dimension. Then the coordinates of most base vectors, and
therefore most of the metrical properties are dimensional, too,
as is, remarkably, also x 3 • Equation (A.57), which determines
gaB as a funktion of aaB and of the components of the curvature
tensor
(D.2)
(D.3) q q
- 229 -
In order to compare results obtained from integrating the boun-
dary-layer equations in contravariant form, with those obtained
e.g. in experiment, the predicted quantities have to be transformed
into correctly dimensioned quantities with the help of equations
(2.6a) and (2.7) or by using the tables given in this appendix
-*
solving for q . It is then sometimes of interest to know deriva-
tives with respect to the physical coordinates x*i without having
to perform transformations. This can be done at a general point
(x',x 2 ,x 3 ) by using the definition of a general length element
(A.5) • The general physical length element along the chosen curvi-
linear coordinate line is linked to the element of the correspond-
ing coordinate by means of the corresponding diagonal term of the
metric tensor, resulting in the derivative:
a a (x (i) ) a -1/2
(0.4) --*-.
oX ~ ox*(i) ox i (g(ii)) oxi
(0.5)
ax *.~ ax * J.
1 _a_. 1
- "2 g (ii) ax~
Then only the operand with correct dimensions for the operations
(0.4) and (0.5) needs to be expressed in terms of the contra-
variant quantity which can be done with the help of the tables
given.
- 230 -
Table D.1 Flow Variables
-* -
No quantity q q c remarks
qref q
[dimension] [dimension] I
i -
1 .1 velocity catp. v v*i[m/s] vref[m/s] Ig(ii)
-
i = 1,2,3
,
i surface-oriented locally
1.2 velocity catp. va v*a[m/s] vref[m/s] rronoclinic coordinates
I Ira-:- (aa)
(aa) , ~ a = 1,2
I
1.3 v3 v*3 [mls] 1
" " ,
velocity catp. vref[m/s] surface-normal
component
I
IV I i=1,2,3 general coordi-
w i ;;;*i [s -1] - - -1'
1.4 vorticity catp. w v ref/Lref [s ] \ Ig (ii) nates, otherwise
I like 1 .2 and 1 .3
i
I I
1.5 pressure p p[Pa] Pref[pa] 1 Navier-Stokes eqs. I
I
I
I
I I
I - - 2 I
1.6 pressure p p[Pa] Pref(vref ) [Pa] I 1 boundary layer eqs.
I
-p[kg m-3 ]
II - -3
1.7 density P I Pref[kg m ] 1
specific heat
1.9 at constant c Cp[J/kgK]
-c [J/kgK] 1
pressure p Pref !
Table D.l Flow Variables (ccmtinued)
-* -
No quantity q q c remarks
'Iref q
[dimension] [dinension]
specific heat
1.10 at ccmstant cv cv [J/kgK] -
c [J/kgK] 1
v ref
volume
IV
W
N 1.13 time t t[s] Lre/Vref [s] 1
--- --- --- -
Table D.2 Geanetrical quantities (assuming xi' ,x 3 to have lengths as dimensions)
No quantity q -
q-* ,c - C remarks
,C q q qref,c qref q
[dimension) [dimension)
cartesian
xi' -*i'
2.1 reference x [m) Lref [m) 1 i = 1,2,3
coordinate
general
2.2 dx i dX* i [m) 1 19 (ii) i = 1 ,2,3
coordinate
surface-oriented ~
(aa) ,
2.3 locally m::moclinic dx a d;{*a [m) 1 a = 1 ,2
coordinate ra-:-:
g (aa)
tv
W
W 3 -*3
II II x x [m) surface-nonnal
2.4 Lref[m) 1
coordinate
No quantity cq
-
cq -cqref remarks
[dimension] [dimension]
3.1 - -1 -1
curvature K Lref[rn ] a = 1,2
N
Ka [rn- 1 ]
a
IN
,c:..
3.2 radius of R
curvature a Ra [rn] Ere fern] a=1,2
--
Table D. 3 Georretrical quanti ties and boundary-layer stretching (cq = 1 for 3. 1 to 3. 5)
~ ~* -
No quantity q q qref Cs remarks
[dimension] [dimension]
corrponent of *) -i I -i I -1/2 .
3.2 base vector S'" S3[-] - (Reref ) l.=1,2,3
3
(E.1 ) v x ds o
where
1/2
{
g for ikl cyclic 123,
1/2
Eikl = - g for ikl cyclic 132,
0 otherwise.
(E.3)
or written explicitly
2 3
(E. 4) A1 v dx 3 - v dx 2 0
A2 v
3
dx 1 - v
1
dx 3 0
1 2 1
A3 v dx 2 - v dx 0
- 236 -
By combining equations (E.4) with each other one arrives at the
same well-known form of the equation for the streamline as for
Cartesian coordinates
(E. 5)
Note, that only two of the equations (E.4) which determine the
streamline are linearly independent. It is then obvious that, like
for Cartesian coordinates, the streamlines can be obtained as
common lines of surfaces of two different families of streamfunc-
tions. The projection of streamlines on,.for example, the surface
x 3 = const. requires the integration of the first of equations
(E.5). On the surface of the considered body, x 3 = 0, one has,
of course, to consider as usual the derivatives of v 1 and v 2 with
respect to x 3 since both velocity components vanish (concept of
limiting wall streamlines or skin-friction lines [57]). Note
further that equations (E.5) would result as well if, instead of
covariant components Ai' contravariant components Aj (covariant
base vectors) were introduced in equation (E.2).
*1 *2 *3
dx dx dx
(E.6)
*1
v
~
v
*3
v
- 237 -
If, on the other hand, a curve s is given in the surface-oriented
i
x -coordinate system, the directional angles ~ and A, Fig. (E.1)
are found by constructing the true physical relations by means of
3
eg. (A.36), where dx has to be a non-stretched quantity.
X 3'
X 2'
X l'
Fig. E.1 Directional angles of a curve s in the
surface-oriented locally monoclinic
i
x -coordinate system .
1 2 3
(E.8) tg ~ (x ,x IX )
- 238 -
and the elevation angle
123
(E. 9) tg A (x ,x ,x )
r- 1 2
(E. 1 0) dS va dx dx ,
-
;g 1 2 3
(E. 10) dV dxdxdx.
- 239 -
List of Symbols
1. Latin Letters
Ai abbreviation, (236)
Aa = v~ a , partial derivative of contravariant velocity
coordinate, (29)
a determinant of covariant surface metric tensor, (15)
and (204)
ai abbreviation, (168)
- 240 -
cr velocity ratio, (185)
Cs general boundary-layer stretching parameter, (229)
Cv specific heat at constant volume, (232)
D diameter, (133)
E energy quantity, (25)
reference Eckert number, (15)
~i' unit vector of Cartesian reference coordinate system, (199)
~ general vector quantity, (221)
Fk general contravariant vector coordinate, (221)
g.
-J
covariant base vector of general coordinates, (199)
gij coordinate of general covariant metric tensor, (200)
gkl coordinate of general contravariant metric tensor, (212)
- 241 -
H abbreviation in chapter 13, (192)
H mean surface curvature, (213)
h enthalpy per unit mass, (25)
Appendix D, (229)
- 242 -
R general radius of curvature, (226)
Re ref reference Reynolds number, (15)
ROT rotation in general coordinates, (225J
Rot rotation in off-surface pOints in locally monoclinic
surface-oriented coordinates, (225)
Ra principal surface curvature radius, (211)
R' general position vector, (199)
r radius, (133)
r factor in scaling relation, (182)
rN nose radius of wing cross section or airfoil, (109)
rot rotation in locally monoclinic surface-oriented
coordinates in the frame of boundary-layer theory, (225)
- 243 -
xi Cartesian coordinate in chapter 12, (159)
-i
x orthogonal surface-oriented coordinate, (158)
i'
x Cartesian reference coordinate, (199)
xa Gaussian surface parameter, (10) and (200)
x3 surface normal locally monoclinic coordinate, (200)
position vector, (199)
2. Greek Letters
- 244 -
scaled surface-normal coordinate, (181), (183) and (184)
- 245 -
3. Indices
etc., (199 )
etc., (200)
min minimum
R wing root
- 246 -
S stagnation point
s stagnation-point coordinate system, (32)
T wing tip
TE wing trailing edge
u upper side of wing
w,wall wall or surface
i=1,2,3:
j
Ili' v
i' j
= Ill' v
1' j
+ 1l2' v
2' ... j
1l 3 , v
J'
- 247 -
References
1. Basic Papers
2. Books
- 248 -
3. Boundary Layers
- 249 -
26. FANNEr.0p, T. K.: "A method of solving the three-dimensional
laminar boundary-layer equations with application to a
lifting reentry body", AIM J., ~ (1968), pp. 1075-1084.
- 250 -
39. HIRSH, R. S., CEBECI, T.: "Calculation of three-dimensional
boun~ary layers with negative cross flow on bodies of revo-
lution", AIAA paper 77-683 (1977).
41. KITCHENS, C. W., JR., SEDNEY, R., GERBER, N.: "The role of
the zone of dependence concept in three-dimensional
boundary-layer calculations", USA Ballistic Research
Laboratories, Aberdeen Proving Ground, Maryland, R 1821
(1975).
- 251 -
52. KROGSTAD, P. A. : "Investigation of a three-dimensional
turbulent bounoary layer driven by simple two-dimensional
potential flow", doct. thesis, Norwegian Institute of
Technology, Trondheim 1979.
58. LINDHOUT, J. P. F., MOEK, G., de BOER, E., van den BERG,
B.: "A method for the calculation of 3-D boundary layers on
!)ractical wing configurations", The Joint ASME-CSME Applied
Mechanics, Fluid Engineering ano Bioengineering Conference,
Niagara Falls, New York, June 18-20, 1979 (NLR MP 79003 U).
59. LINDHOUT, J. P. F., van den BERG, B., ELSENAAR, A.: "Com-
parison of boundary-layer calculations for the root section
of a wing: the September 1979 Amsterdam Workshop test
case", NLR-report to be published.
- 252 -
66. NASH, J. F., SCRUGGS, R. M.: "An implicit method for the
calculation of three-dimensional boundary layers on finite,
thiCk winqs", SYBUCON Report, SYB-76-102, Atlanta (1976).
69. PH~H.S,w. J., SCHIl:'HOL'f, G ••J., van r1en BERG, 13.: "Calcula-
tion of the flow around a swept wing taking into account
the effect of the three-dimensional boundary layer", NLR TR
75076 (1975).
- 253 -
78. SCHWAMBORN, D.: "Boundary layers near the attachment line on
finite wings and related bodies at angle of attack", 15th
IC~AM, ~oronto, Canada, Aug. 1980 (to appear as doctoral
thesis, RWTH Aachen).
1:l5. 'ERELLA, M., LIBBY, P. A.: "Similar solutions for the hyper-
sonic laminar boundary layer near a plane of symmetry",
AIM .1., 1 (1965), pp. 75-81.
- 254 -
90. BRO\~, S. N., STEWARTSON, K.: "Laminar separation", in: An-
nual Review of Fluid Mechanics, Vol. 1 (1969). pp. 45-72.
- 256 -
116. HUNG, C. M.: "Numerical solution of supersonic laminar flow
over an inclined body of revolution", AIM paper 79-1547
(1979), also AIAA J., ~ (1980), pp. 921-928.
- 257 -
129. SHANG,.T. 5., HANKEY, ,JR., W. L.: "Numerical solution of
the Navier-Stokes equations for a three-~imensional cor-
ner", AlAI'. J.,..!..2 (1977), pp. 1575-1582.
133. WALITT, L., TRULIO, ,J. G.: "I'. numerical metho~ for compu-
ting three-dimensional viscous supersonic flow fields about
slender bodies", NASA CR-1963 (1971).
- 258 -
Subject Index
- 259 -
Characteristics, 63 f
Characteristic properties, 8, 25, 63 f, 68, 89
Chord, see wing
Chord direction, 67, 69
Christoffel symbol, 80, 86, 92, 215 ff, 218, 222
Circumferential direction, 66
Compatibility conditions, 9, 146 ff, 159
Computational domain or area, 4, 18, 69, 120, 139, 180 ff, 202
Computation method, 2, 180 ff
Continuity equation, 11, 15, 28 f, 37, 45, 49, 52, 58, 71, 80, 82,
87, 92, 146 f, 151, 185 f, 191, 193, 195 ff
Contour angle, 7, 9, 44, 48, 107, 113, 117, 120 ff
Convective terms, 29, 63, 147
Coordinate line, 4, 14, 200 ff, 230
Coordinate systems, 4, 11, 80, 87 f, 117, 199 ff, 221 ff, 225, 228,
231, 236
accelerated, 11
Cartesian, 13,44,87,94,159,173,237
Cartesian reference, 4 ff, 10, 33, 103, 107, 117, 120, 130, 133,
153, 199, 203, 219, 228, 233
external streamline, 7, 65 f, 129 f, 184
orthogonal, 22, 29, 158, 213
percent line, 7, 119
surface oriented, locally monoclinic, 4 ff, 10 ff, 17, 26 ff,
32 f, 35 ff, 44 ff, 66, 71 ff, 82, 87 ff, 103, 107,110,
117 ff, 146, 158 f, 199, 200 ff, 222, 225 f, 228, 231 ff, 238
Corner flow, 6, 94 f, 117
Courant-Friedrichs-Lewy condition, 63, 65, 69, 169
Crocco relation, 182
Cross flow, 65, 69
Cross-flow profile, 176, 180
Curvature of surface, 12, 44, 70, 73, 79, 100 ff, 109 ff, 125 ff,
209 ff, 224, 226, 234 f
direction of principal curvature, 211
Gaussian, 213
gradient, 70 f
mean, 213
principal, 70, 75, 136 f, 143, 210 f, 213, 219
principal radii, 211
radius, 4, 12, 70, 100, 110, 116, 202, 234 f
Curvature tensor of surface
covariant, 126, 136, 143, 151, 154, 210, 213,216 ff, 228 ff, 233 ff
mixed variant, 127, 137, 143, 212 f, 216 ff, 229, 233 ff
Density, 14, 231
Derivative, 11, 29, 38, 230, 233, 237
covariant, 75, 81, 83, 147
of vector quantity, 215, 221 f
partial of scalar, 207
spanwise, 42 ff, 166, 174
Displacement thickness, see Boundary-layer integral parameter
Dissipation function, 81, 85
Dividing stream surface, 34
Domain of dependence, 8, 63 ff, 169
of influence, 8, 63 ff
Drag, 3
Duct flow, 95, 128
- 260 -
Eckert number, 15
Ellipsoid, 133 ff
Energy
equation, 12, 16, 28 ff, 38, 45, 49, 53, 58, 72, 80 ff, 87, 92,
146, 149, 152, 188, 192 ff, 196 ff
flux, 25, 75
kinetic, 25
Energy-loss thickness, see Boundary-layer integral parameter
Enthalpy, 25
Entropy, 28
Equation of state, 15, 93
Equivalent inviscid source distribution, 8, 23, 31, 40, 46, 50, 54,
59, 78, 190 ff, 198
Euler equation, 1, 5, 91, 94
External inviscid flow of boundary layer, 1, 4, 13, 17, 22, 38,
42 ff, 58, 63, 65, 68,75 ff, 109, 114 ff, 138 f , 157, 172 ff,
180 ff, 196
Gasdynamic equations, 11
Gaussian surface parameter, 7, 10, 82, 103, 117 ff, 203 ff, 223
Geometrical singularity, 66, 130, 135
- 261 -
Heat flux, heat transfer, 18, 20, 26, 30, 39, 46, 50, 54, 59, 73,
95
Independence principle, 45
Infinite swept wing or cylinder (see also locally infinite swept
wing), 34, 42 ff, 90, 103, 108 ff, 128 f, 162, 166 f, 174 f
Initial conditions, also data, 18, 32, 34, 38, 42 f, 56 ff, 60,
b2 ff, 89, 93, 114, 139, 196 ff
Inlet of propulsion engine, 2
Integral method or solution, 2, 21
Integral parameter, see Boundary-layer integral parameter
Interpolation of data, 6, 139, 202
Inviscid flow, see also External inviscid flow of boundary layer,
1, 21, 23, 42, 52, 65 ff, 96, 98
Irrotational flow, 42, 79, 98, 174
Isobar, 42, 52
Inviscid-viscous coupling, 1, 23, 90
Jacobian transformation matrix, 206
Kronecker tensor, 212
Laminar flow, see Flow (boundary layer)
Leading edge, see Wing
Leading edge contamination, 100, 112, 114
Leading-edge oriented coordinates, 44 ff, 48 ff
Length element of coordinate line, 17, 200, 209, 230
Length, physical, 14, 125, 199, 223, 229
Levy-Lees transformation, 180
Limiting wall-streamline, see Skin-friction line
Locally-infinite-swept-wing approximation, 42 f, 52 ff, 56 ff, 67,
69, 195 ff
Logarithmic wall region, 181
Main-stream direction, 66
Metric factor, 15,23, 36 f, 44, 48, 53, 60, 71, 92 ff, 113, 154,
185 f, 215 ff
Metric properties, 4, 7, 70 f, 91, 96, 117, 120 ff, 133 ff, 199 ff
228 ff, 233 f
Metric tensor of locally monoclinic surface-oriented coordinate
system
contravariant, surface, 127, 137, 211 f
covariant, surface (a-metric), 13 ff, 19, 36, 44, 48, 52, 60, 71,
75, 98, 101 f, 113, 124 ff, 135 ff, 146, 154, 185 f, 203,
208 ff, 218 ff, 224, 225 f, 228 ff, 233
covariant, off surface (g-metric), 7, 71, 75, 92 ff, 101 f, 113,
127 ff, 137, 148, 203 f, 208 f, 213 ff, 218 ff, 222 ff, 228 ff, 233
Momentum equation, see also Navier-Stokes equations and Boundary-
layer equations, 71 f, 148, 151 f, 173
Momentum flux, 23, 75
Momentum-loss thickness, see Boundary-layer integral parameter
Navier-Stokes equations, 1 ff, 8 ff, 80 ff, 100, 102, 116, 146 ff,
231
thin-layer approximation, 5, 8, 80 ff, 91 ff, 203
Newtonian fluid, 11, 80
Noise, 3
- 262 -
Non-dimensionalization, 13 ff, 19, 118, 151, 199,214,226,228 ff
Normalizing length, 13, 119 ff, 131 ff
Nose radius, 107 ff, 179
No-slip condition , 1, 17, 20, 38, 146
Numerical solution, 1 f, 79, 62 ff, 88 ff, 116, 180 ff
- 263 -
Similarity transformation, 9, 180 ff
Skewing of coordinates, 6, 73, 202, 205, 216
Skin-friction line, also limiting wall streamline, 63, 65, 156 ff,
161 ff, 236
curvature, 167
points-of-inflexion line, 168 ff
Solution
steady state, 89
transient unsteady, 89
Span of wing, see Wing
Specific heat, 14 f, 231 f
Speed of sound, 15
Stagnation line, see Attachment line
Stagnation point (nodal), also point of attachment, 7, 26 ff, 39, 42,
65 ff, 114, 138 ff, 163, 175, 190 ff, 193, 222 f
axisymmetric, 30
saddle, 26, 39
Step
forward facing, 116
backward facing, 116
Stream function, 237
Streamline, 9, 236 ff
curvature, 26 f
direction, 63, 65, 180
direction angle, 233
elevation angle, 160 ff, 238 f
external inviscid, 26, 63, 64 f, 157, 163, 172 ff
length element, 236
points-of-inflexion line, 175
projection on surface, 64, 236 f
Stream surface, 27 f, 161 f
Stress tensor, 80 ff, 147, 150
Suction, see Boundary-layer suction
Surface, 2, 4 ff, 38, 61, 71 ff, 88, 97,146 ff, 159 ff, 200ff, 216,
223, 229, 236 f
concave, 4 ff, 201 f
convex, 4 ff, 39, 201
cooling, 3,18,26,28
curvature, 1, 9, 20, 100 ff
element, 239
heating, 3, 18, 26, 28
hydraulically smooth, 20
impermeable, 20, 22, 146
no-slip condition, 1, 17, 20, 38, 146
Sweep angle, see Wing
Symmetrical flow, see also Plane of symmetry, 27, 36, 57, 60 f,
66 f, 110
Symmetry of variables, 36, 40
even, 3&
odd, 36
- 264 -
Transport properties and coefficients, 3, 12, 93, 102, 149, 232
effective, 12, 149, 232
mULecular (laminar flow), 88
turbulent, 12, 102, 116, 149, 159
Turbomachine, 2
Turbulence
large eddy simulation, 88
model, 2 f, 20, 70, 97, 100, 112, 114
Turbulent flow, see Flow (boundary layer)
Turbulent Prandtl number, 20
Turbulent reattachment after separation, 114
Unit vector, 199 ff
Variables
dependent, 21, 29, 32, 36, 91, 146, 180, 191 ff, 228, 231, 237
independent, 2, 91, 228, 233 ff
primitive, 150
Vector coordinates (components)
contravariant, 3, 9, 13, 205 f, 221, 225, 228, 237
covariant, 13, 237
Velocity
chordwise, 56, 67
covariant coordinates (components), 13, 225, 237
contravariant coordinates (components) 3, 9, 13, 22, 33, 71.ff,
82, 91, 205 f, 223, 225, 229, 231, 237 f
distribution, 65
gradient, 28 f, 37 f, 110, 150, 191 f
physical, 13, 229, 231
profile, 2, 97 f, 180
spanwise coordinate (component), 42, 50, 52, 69
vector, 73 f, 205 f, 218 ff, 222 f, 236
Viscosity, 14, 232
apparent turbulent, 20
bulk, see Bulk viscosity
effective, 20, 232
molecular, 20
Viscous-inviscid interaction, 1 f, 23, 74, 98, 100, 116
Viscous sub-layer, 181
Volume element, 239
Vortex flow, 1,69,86,116,156
Vorticity vector, 98, 225 ff, 231
Wake flow, 1, 100
Wake region in turbulent boundary layer, 181
Wall-heat flux, see Heat flux
Wall temperature, 17, 45
Weingarten formula, 212, 220
Wing, 1, 5 f, 62, 66 ff, 96, 100, 103 ff, 110 ff, 116, 117 ff,
133, 157, 172 f
chord, 36, 48, 52, 56, 69, 111, 118
cross section, profile, 35, 119
finite, 34, 42 f, 52, 56, 67 f, 189
leading edge, 34 ff, 42, 44, 46, 48, 52, 67 f, 96, 103 ff, 107,
110 f, 114 f, 118 ff, 138, 157, 172 f, 176, 179, 189
lifting, 34, 56, 65
lower surface, 34, 48, 67, 121
non-lifting, 66
- 265 -
root, 42, 69, 103, 116, 118 ff
swept, 68, 103 ff, 111, 114, 122 ff, 170
sweep angle, 42, 44, 48, 52, 124, 176
tapered, 43, 52, 56, 114
tip, 42, 62, 65, 69, 103, 120
tip vortex, 69
trailing edge, 1, 5, 42, 108, 114, 118 ff, 179
upper surface, 34, 48, 67, 120 f
Wing-body combination, 2, 65
Wing-body junction, G, 62, 96, 116
- 206 -