13 JUNI 1979
ARCHIEF
y.
Scheepsbouwkun
Technische Hogeschool
Dem
KUNGL. TEKNISKA HOGSKOLAN
I STOCKHOLM
HYD ROM E KA NIK
A VORTEX MODEL OF THE POTENTIAL FOR THE INCOMPRESSIBLE AND INVISCID FLOW
PAST A RIGID BODY
OVE SUNDSTRÖM
SUMMARY
TRITA-HYD-79-01
A VORTEX MOVEL OF TI-lE POTENTIAL FOR THE 1NCOMPRESSILE
ANV INVISCIV FLOW PAST A R1GIV BOVY
by
o. Sundström
An equation is derived for vorticity distribution over the surface of a rigid body in potential flow. An iterative technique is used for the solution of the equation. Convergence of the sequence is studied by means of the concept of kinetic energy. The theory is applied to the uniform flow past slender bodies.
INTRODUCTION
The application of potential theory to problems involving a rigid body moving in an ideal fluid is well known in hydro-dynamics. The surface of the body is usually represented by a simple distribution (sources and sinks) or double distribution (dipoles). S.J. Andersson has called the author's attention to the fact that it should sometimes be more convenient to use a vorticity distribution.
EQUATION OF THE VORTICITY DISTRIBUTION
Let
{1,2,3}
be an orthonormal basis of the Euclideanspace R . The corresponding Cartesian coordinates are x1 , x2 and x3 . We write the position vector x.é.
The physical body corresponds to a bounded and open domain D'
with a boundary S consisting of a finite number of regular
surface elements, cf. }Kellog (5). Let D denote the
complement-ary set R3-(DUS) , which is assumed to be corìriected.
Let () be the velocity potential of the flow of an ideal
incompressible fluid. We assume that 2() is harmonic in a
domain containing D'US .
Let q() be the perturbation
poten-tial for the body. Then the velocity potenpoten-tial of the disturb-ed flow becomes
+
On the surface S of D' we have the boundary condition
().Vc()
O,
where
/x.
and ñ() is the outward unit normal.i i
be two
functions harmonic and continuously differentiable in a closed regular region EUE , thenO =
J
[v1().v2())
+ aE -JdS()
This result follows by the divergence theorem.
Let
and
2Using the notations
we have
() -V,)
Set i=i , =c and apply (2.1) to the domain D-O() , where
(2.1)
and
/.
V) vQ()
+
(x()x()) dS()
,
z V2() + V() z
Q ,The perturbation velocity becomes
V)
=J+()()
dS()
where
F() =
()x()
is the vorticity distribution on the surface S
-2-(2.2)
O(i) is a small sphere with center and the radius E
Then, if
Vq()j
O uniformly as , we obtain at thelimit c=+O
dS()
xD()V)
+J+/n
dS()
where
/n ñ()V()
, and X denotes the characteristicfunction of the set E
Ti, XCE,
xE(x) z
O, E.
S+ signifies that the integral is taken over the exterior side
of the surface S which bounds D
Since AQzO
throughout DUS
, formula (2.1) applied toDO() yields
(x(ñ()xQ())
dS()
XD(X)
+
Summing these equalities, and using
D/Bn=O
on S , we deduce3. AN ITERATION PROCEDURE
We now seek a solution of equation (2.2) by a method of successive approximations. Therefore we define the iterated
functions . by
VQ()
+
ds(, ji,
with o = , .
-Q
when ±Applying the divergence theorem to the domain DO() wo
obtain
+
Js
()
DQ()/n dS()
and we note that the normal derivative is continuous across
S . The iteration procedure is repeated and it follows that
+
() .()/n dS()
J S
For the normal derivatives, we obtain
(
()/n) ±
= +ñ()(,)
.1()/n dS().
Let us introduceu.() =
-Then HenceVu.()
f
()
(Bu. 1)/BnJ dS()
J S JBy means of these functions we can write
VQ() + V.() = VQ() +
Vu.()
,V.(x)
1=1 1VQ() + V.() = Vu.() ,
The sequence of normal derivatives of the functions u.
becomes, using the notation p.()
(Bu.()/Bn),
pi+1 () + I
dS()
J
Using the notation
Tp(i) -2J
)ñ()) dS()
we can rewrite the normal derivatives
It is known that if S is a Liapunov surface, then a continuous
and bounded function p on this surface generates a Hölder
con-tinuous function Tp (cf. Günter (4) ) . Accordingly, in this
class of surfaces ñ'V4. remains continuous across S
J
4. ENERGY RELATIONS
Let p and be L'-integrable simple distributions on S
which generate the potentials
=
dS(),
=
-J
,p() dS().
We introduce the symmetric bilinear form
1515
Let
(u/n)
and (u/an) denote the normal derivatives ofu at interior and exterior points of S respectively. Then
(u/n)
(1-T)p ,(u/nJ
-(1+T)p
By the divergence theorem it follows that
fVu.Vv dV
J
y(u/n)
dS =J:vusvv dV
-J
y (3u/n) dS z and henceJVuVv dV
(p,)
R3Because of symmetry we obtain ((1±T)p,v) =
p,(1±T))
, henceWe write
(Tp,v) = To(p,v)
It follows immediately that
(1±T)op,p) O ,
Ki,i)
OHence
Further, for every real number o. , +o.v,p+o.) O . Hence
the Schwartz inequality
<
If f(T) is a polynomial in T , we can form the simple distribution f(T)p , and obtain the inequalities
(1±T)o(f(T,f(T))
= (1±T)f(T)2o(,p) O.We have the identity 1-T2 ((1+T)(1-T)2 + (1-T)(1+T)2) and
it follows immediately that (1-T2)o(i,p) O or
KTp,Tì-')
Applying the energy integral to the recurrence formula
(1-T)p. , Q/n , we obtain
(1-T)2o(p.,1).
The inequalities (Tp,p)j(p,p) and (Tp,Tp)(p,i) give
and
(1_T)op11,i11)
(1-T)o(i,i1), (1+T)o(p.1,p.41) (1+T)o(p.,i.),(1-T)o(p.,p.),
24(1+T)o(p.1,.1)
(1-T)o(.,p.),
2(1+T)o(p.
j+1
,p.1)
(1-T)o((p.,p.)
-Hence 2(1+T)o(p.1,.1)
1=1 (24.1) (24 . 6)Since
(l-T)1p1()
,i1
, (4.7)
it is easily seen that
(p.,p.) O , (1±T)o(p.,p.) 1 0
. (4.8)
The energy of the perturbation velocity outside the body becomes
(Vj dv
= (1+T)o( p. , p. ) = D i=1 i=1 =(p,pi
-1=1 Since, by (4.7), (1+T) p. p -i=1 JAs a matter of fact, the kinetic energy is p times this
expression, p is the density of the fluid. For the two cases
¿=22. even and j=22.+1. odd, we obtain by (4.7),
I
(V.)2 dV
((l-T)oKp.,p.) + 1=1 - ((1-T)o(p.,p.) + (p1,p1)) , r(V)2 dV
((1-T)o(p.,p.) + J 1=1 ((1-T)o(p.,p.) + (p.,p.)) + (l+T)o(p2.1,p2.1).Theorem 1: If D - is a sphere, then To(p,p) 1 0
Proof: Let the radius of the sphere be R . We have
2Tp()
-J ñ((-) -3p() dS()
where ñ()
/R and 2 == R2 , if the center is origin
of coordinates, and we find
4RTp()
= JSdS()
=dS()
Hence 2 R(Tp,p) =dS()
,P()
dS()]
o .I
-6
(4.9)
co tH H (D
Z
(o Ç)0
ZH
HM
-
(JJ -o
H i-HO -- -o t'i H. H00H
H-)te subdomains of D and D such
. <
H us d about any point
cD1UD lies + C-»
0
-a H + H k-OJjt'1
HZ Q
OH H
H F-i.
tends to zero uniformly
in
D1(d).O 'i
u0
H
H n lIN.
t of D1(d) , then as harmonic
ithmetic means throughout spheres oy are harmonic Cn C')
()/
dV()
t-1 - £ L'I inequality gives __i'__' i)ji1IU
JK(u.()/)2
dV()
(3/4îd3) (1+T)oi.,p.) ó (3/rd3) If E24rd3/3 , then < e for ilN(6) .I
J-By (.8) it follows forthe perturbation velocity
I
(Vj2 dV
J dV J(Vj2 dV
when ij
. (5.i) 1 D ' D Set M 11mf
D(V)2 dV
sphere,
the inequality (1.5) yields(1-T)o(.,p.)
, and it follows that
harmonic functions we infer
imme-ary positive number. The conditions
to each 6>0 there is a N(6) such
Theorem 3: If M<,
then the sequenceV. converges
uniformlyin D1(d) to a harmonic function V
Proof: Given >0 . By (5.1) it follows that to each 5>0
there is a N(5) such that M-<...
J
(Vj2 dV
J
(V.)2 dV
M D 1 Dwhen jiN(6) .
Let be any point in D1(d) . By the meanvalue theorem for harmonic functions, we have
-
.()/x
(3/4d3)J (./
-./) dV()
where 2=1,2,3 . The Schwarz inequality and (5.1) give
-
.(/x)2
(3I4d)J (v-
vJ2
dV < (3I4d3)
If -241Td3/3 , then
-
.(/x
< whenj,iN(&)
Hence
a./3x-
uniformly in D1(d) .By Harnacks
theorem it follows that the limit function is harmonic.
I
6. SOLUTION OF THE PROBLEM FOR THE SPHERE We have shown that
V() + V.()
VS)
+
Vu.()
,VÇ2() + V.()
Vu.(x) , XEDThe conditions (4.9) and (4.10) give
(v)2
dV 2(u1)
3 1=1
This inequality holds whenever To(p,p) 0 . In this special case we may also derive that the perturbation velocity is
integrabel over the outside and inside of the surface
S . Here
the sequence of normal derivatives becomes
2R.1()
Rp.() - u.()
and we find
r p dS R p dS -
((1_T)o(1,)
+ i+1 is Thu sil
1J dB ii dS 1+1By means of the divergence theorem the following equalities
can be shown
1-T)o(p.,p.) ((Vu.)2
- 2(Vu.)
û.Vu.) dS()
(1+T)o(p.,p.)
f( 2(.Vu.)
Vu. -
(Vu.)2 ñ.)
dS()
13+ J_
i
Hence, since =
Rñ()
4' 1 (Vu )2 dS
r (Vu )2 dS 4 p aS
Js_
)3+
sWe infer immediately that V
Vu. - VO
is integrabel
forall
j.
Using Schwarz inequality, we find
( dS (
J(vu
dS
and it follows that 1
(V.(fl2 dS()
S
(p())2 dS()
is+
i
Then,
by Schwarz inequality, the perturbatiion velocity is integrabel over the sphere.Since (i.,pi) - O as i -- , it follows that (1-T)o(p.,p.) O
Theorem 2 can therefore be extended to the domain D1(d) Finally, the condition M<
in theorem
3is
fulfilled.7. CONVERGENCE IN THE CASE OF A REGULAR REGION
Let DUS be a regular region (cf. Kellog (5) ). We shall prove the following theorem,
Theorem 4: Let LY be a regular region and D2UD a sphere. If AQ=O in the closed sphere, then the iteration procedure is convergent.
Proof: Set D3
D-(D2UD)
and let be the surface of thesphere.
According to previous result, the problem for the sphere
D2UD has a solution
+ vq() ,
v() =
o+ v() ,
cD2UD' with J (V )2dV<
R3Now, we form a new sequence W. , defined by
Vw.()
- , W.(cx) = QSince V O in the sphere, we obtain for xD2UD
Vw.()
v)
+dS()
For points D' we have
VW.() = Vu.()
Obviously, the normal derivative is continuous across and
the jump of W. is independent of its index j . Therefore,
(VW.)2 dV
ñ().VW.() (w7() -w)) aS(r)
+ D3 R3-1
ñVW. W
dS + IñVW.
(wT-w
)as
j-1 Hence f (VW.)2 dV = f I)We now obtain the inequality
2f
(7W.)2 dV = (1_r)o(p.,p.) Dj-1
-
(VW)2 ) dV - 1_o -+).Vw.() (wT() -W;()) dS()
Thu s
r (i-T)o(p.,.)
J(VW)2 dV
- limJ (VW.)2 dV j=2 Dj--D
whereJ
(VW1)2 dV2f
(V)2 dV
+ J (VQ)2 dV < D D Using (4.»4), we obtain by (4.9),r
+ (i-T)o(p.,p.) 1=1 + (Pl,Pl) ++ r
+ 'D(2
<Hence, lim(p.,p.) = O and limJ
(V.)2 dV
M <j-D
The theorem now follows from the theorem 3.
8.
APPLICATION TO THIN OR SLENDER BODIESLet D represent a body in a homogeneous flow wfth the
velocity VQ = Uë1 . If the surface normal ñ is mainly almost
orthogonal to ,
i.e. ñ
o(i) ,
we may consider as agood approximation of . We obtain for ED
v) v()
U1 +
UJ(x(ñ()x1)
dS()
= UV(x1
- JD
(ci-Xi)
i-- dv())
and as
Ux1 +
(U/1)V(D)
x1For a slender body, we set = and obtain an
approxima-tion for the volume integral. Let the area of the
cross-section of the body be A(i) where -LF1L .
AssumingA(L) = A(-L) O , we have, on integrating by parts
Ux1 - (U/4)j
((1-x1)2
-
x+ x)
dA(1)
-L
(i-T)o(. ,p1)
I
ID
When there is a solution, we can write
+ VR(,ct)
with V V +
x(x) dS
If the remainder R is minimized by ci. in the following way
J(VR)2
dVwe obtain, since
Vc0
in DJ((1-)V
+ aVu1)2 dVIn that sense we obtain a still better approximation by using
VÇ instead of
V1
ACKNOWLEDGEMENT
Professor B.J.
Andersson
has given many valuable suggestions. In particular, he demonstrated theorem 1 which made it possible to realize this theory.LITERATURE
ANDERSSON, B. J "On the Mathematical Model of a Ship,'t International seminar on wave resistance. The Society of Naval Architects of Japan.
(1976)
BATCHELOR, G. K. "An Introduction to Fluid Dynamics,"
Cambridge University Press
(1967)
COURANT, R. "DirichletTh Principle,"
Interscience
Publishers, New York(1950)
. GUNTER, N. M. "Die Potentialtbeorie," Teubner, Leipzig
(1957)
KELLOG, O. D. "Foundation of Potential Theory," Dover Publications, New York
(1953)
LAMB, H. "Hydrodynamics," Sixth edition.Cambridge University Press