You are on page 1of 11

ORIGINAL PAPER

H. S. Lee S. A. Kinnas
A BEM for the modeling of unsteady propeller sheet cavitation inside
of a cavitation tunnel
Received: 25 May 2004 / Accepted: 15 December 2004 / Published online: 14 July 2005
Springer-Verlag 2005
Abstract A boundary element method is applied to
predict the unsteady cavitating performance of marine
propellers subject to a non-axisymmetric inow with the
complete tunnel wall eect. The tunnel and propeller
problem is solved directly to predict the fully unsteady
performance of the cavitating propeller. The cavitation
on blade and wake surface is determined by applying the
dynamic and the kinematic boundary conditions on the
cavity surface. The potential on the cavity surface is
known from the dynamic boundary condition and the
relation between cavitation number and cavity velocity.
Once the boundary value problem is solved for the
unknowns- the potentials on the wetted blade surface
and the normal derivative of potentials on the cavity
surface the new cavity shape is adjusted by using the
normal derivative of the potential. The potential on the
new cavity surface is determined by using the kinematic
boundary condition on the cavity. The procedure is
repeated until the cavity shape converges and the pres-
sure on the cavity becomes constant. To validate the
present method, the eects of the number of blade
panels and the dimensions of the tunnel walls on blade
forces are presented. The predicted cavity patterns
are compared with those observed from experiment.
Introduction
Most of the cavitating characteristics of marine propel-
lers are measured or observed in cavitation tunnel since
the ow conditions can be easily controlled and adjusted
in a tunnel. However, the experiments in cavitation
tunnel inherently include the eect of tunnel walls, and
thus the experimental results have to be corrected to
predict the performance or cavitating characteristics of
propeller without tunnel wall eects. Even though
Glauerts formula [12] based on the axi-symmetric
momentum theory is used to correct tunnel wall eects,
this correction is only conned to the case of uniform
inow. Thereforfe, it is essential to develop a numerical
tool which predicts cavitating performance of propellers
subject to a general (non-axiysmmetric) inow with
tunnel wall eects.
Most of the numerical computations for the predic-
tion of cavitating performance of marine propellers have
been developed based on potential theory such as lifting
surface method and boundary element method. A lifting
surface method has a diculty in capturing the details of
ow near blade leading and trailing edge since the linear
cavity theory breaks down. On the contrary, a boundary
element method (BEM) discretizes the exact blade sur-
face and inherently include the eect of thickness load-
ing coupling. Thus a BEM predicts more accurately the
pressures and cavity ow details at the propeller leading
and trailing edge. A BEM has been successfully applied
for the analysis of cavitating performance of marine
propellers in steady and unsteady conditions. Non-linear
potential based BEM was rst applied for the analysis of
the cavitating propeller subject to a non-axisymmetric
inows by Kinnas et al. [14] and Fine et al. [11]. In their
method, the propeller back cavitation was predicted wth
prescribed cavity detachment location. In Kinnas et al.
[16] and Mueller et al. [21], the method was extended to
predict face and/or back cavitation with searched cavity
detachment.
Kinnas et al. [17] has developed a boundary element
method which can predict partial and/or super cavity on
hydrofoil inside of a square tunnel. The tunnel and
hydrofoil problems were solved iteratively to consider
the eects of each other. Their method was later
extended for the analysis of 2-D and 3-D cavitating
hydrofoils inside of a numerical wave tunnel with the
free surface by Bal et al. [1], and Bal et al. [2]. Recently,
Kinnas et al. [18] coupled cavitating 3-D hydrofoil solver
Comput Mech (2005) 37: 4151
DOI 10.1007/s00466-005-0696-z
H. S. Lee S. A. Kinnas (&)
The University of Texas at Austin, USA
E-mail: kinnas@mail.utexas.edu
with 3-D tunnel solver based on a BEM to predict 2-D
cavity patterns on both face and back sides of hydro-
foils. The propeller performance inside of a tunnel was
predicted via coupling of a lifting surface method and a
boundary element method by Choi et al. 1998 [5], and
Choi et al. 1999 [6]. In their method, the unsteady cav-
itating propeller problem and the tunnel problem were
solved separately, and the eect of one on the other was
accounted for in an iterative manner. The inuence of
propeller on the tunnel walls was considered via poten-
tial, and the tunnel walls inuence on the propeller was
considered via velocity.
In the present method, a potential based BEM is
applied to predict the unsteady cavitating performance
of marine propellers with modeling of tunnel walls. The
details of a BEM formulation for the tunnel and cavi-
tating propeller, and the corresponding boundary con-
ditions are described in the paper. Also, the method is
validated by performing the convergence studies and by
comparing with experimental measurement.
Formulation
Cavitating propeller inside of a tunnel
Consider a 3-D cavitating propeller subject to a non-
axisymmetric inow ~q
w
(x
s
; r
s
; h
s
), inside of a tunnel,
which is dened in the ship xed coordinate system
(x
s
; y
s
; z
s
), as shown in Fig. 1.
The inow ~q
w
is assumed to be an eective wake
which includes the interaction between the vorticity and
the propeller [8]. Then, the inow velocity, ~q
in
, with re-
spect to the propeller xed coordinate system (x; y; z) can
be expressed as follows:
~q
in
(x; y; z; t) =~q
w
(x; r; h xt) ~ x ~x (1)
where r =

(y
2
z
2
)
p
, and h = tan
1
(z=y). ~ x is a pro-
pellers angular velocities, and ~x = (x; y; z).
With the assumption that the ow inside of a tunnel
is inviscid, irrotational and incompressible, the ow eld
can be expressed in terms of the perturbation potential,
/(x; y; z), as follows:
~q
t
(x; y; z; t) =~q
in
(x; y; z; t) \/(x; y; z; t) (2)
where ~q
t
(x; y; z) is the total velocity vector.
The perturbation potential, /, satises the Laplaces
equation in the uid domain.
\
2
/(x; y; z; t) = 0 (3)
The perturbation potential, /, at every point p(x; y; z) on
the propeller (S
B
), its trailing wake (S
W
) and uid
domain bounded by tunnel surface (S
T
), has to satisfy
Greens third identity, and the following integral equa-
tion is obtained.
2p/(~x; t) =
Z Z
S
Bw
(t)S
C
(t)
/
q
(~x; t)
o
on
q
(t)
1
R(p; q)

1
R(p; q)
o/
q
(~x; t)
on
q
(t)
!
ds

Z Z
S
W
(t)
D/
q
(~x; t)
o
on
q
(t)
1
R(p; q)
ds

Z Z
S
T
(t)
/
q
(~x; t)
o
on
q
(t)
1
R(p; q)

1
R(p; q)
o/
q
(~x; t)
on
q
(t)
!
ds (4)
where R(p; q) is the distance between the eld point p
and the variable point q. ~n
q
(t) is the unit vector normal
to the integration surface being positive when pointing
into uid domain. S
Bw
(t) denotes the wetted blade and
hub surfaces, and S
C
(t) denotes the cavity surface on the
blade and/or wake surface. The Greens formula given in
Eq. 4 implies that the potentials on the uid domain can
be represented by dipole and source distributions. The
potential on propeller blade, hub, tunnel walls, and
cavity surfaces can be computed as a linear superposi-
tion of the potentials induced by continuous source and
dipole distributions. On the other hand, the potential on
the wake surface can be expressed by continuous dipole
distribution. The solution of integral equation is un-
iquely determined by applying appropriate boundary
conditions on the body and cavity boundaries.
Boundary conditions on tunnel walls
The inow,
~
U
in
, to the tunnel is assumed as undisturbed
and non-axisymmetric ow and comes from far
upstream [17]. Since the inow to tunnel is tangent to the
tunnel wall, inow velocity satises the following equa-
tion on the wall.
~
U
in
~n = 0 (5)
v Flow tangency condition on solid tunnel walls : the
ow normal to the tunnel wall is equal to zero.
o/
on
= 0 (6)
Fig. 1 Propeller inside of a tunnel subject to a general inow.
Propeller xed (x; y; z) and ship xed (x
s
; y
s
; z
s
) coordinates are shown
together
42
v Inow and outow boundary conditions: assume that
the tunnel inlet/outlet boundaries apart far from the
propeller location, and the total velocity normal to the
inlet/outlet boundaries is equal to uniform inow,
~
U
in
.
~
U
in
~n
o/
on
=
~
U
in
~n or
o/
on
= 0 (7)
Boundary conditions on cavitating propeller
The boundary conditions applied on the propeller
blades, cavity, hub and trailing wake surfaces are as
follows:
v Kinematic boundary condition on wetted portion of
blades and hub surfaces implies that the uid ow has
to be tangent to the wetted surfaces.
o/
on
= ~q
in
~n (8)
v Kutta condition implies that the velocity at the blade
trailing edge has to be nite.
[\/[ < at blade trailing edge (9)
v Cavity is closed at the cavity end. Since the extent of
the cavity is unknown, the unsteady cavity planform
has to be determined as a part of solutions. In the
present method, the Newton-Raphson iterative meth-
od is applied to nd the cavity planform which satises
cavity closure condition [15]. For given cavitation
number, r
n
,
d(l(r; t); r
n
) = 0 (10)
where d is the cavity thickness at the cavity trailing
edge, and l(r; t) is the cavity extent. The pressure
recovery termination model is applied near the trail-
ing edge of the cavity, and the total cavity velocity
over the transition region is dened as an algebraic
function [15].
v Kinematic boundary condition on cavity surface
implies that the total normal velocity on the cavity
surface has to be zero. In a non-orthogonal local
coordinates system, (s; v; n), shown in Fig. 2, the
cavity height of any points on the blade surface can
be dened by a function, n = h(s; v; t). Therefore, the
exact kinematic boundary condition can be derived
by requiring that the substantial derivative of a
function n = h(s; v; t) vanishes on the cavity surface.
D
Dt
(n h) =
o
ot
(n h) ~q
t
(x; y; z; t) \(n h) = 0
(11)
The total velocity on each local panel can be expressed
in terms of local velocities as follows [19],
~q
t
=
V
s
~s (~s ~v)~v [ [ V
v
~v (~s ~v)~s [ [
[[~s ~v[[
2
V
n
~n (12)
with~s and ~v being the unit vectors corresponding to the
coordinates s (chordwise) and v (spanwise), respectively,
and with ~n being the unit normal vector to the cavity.
Also, V
s
, V
v
and V
n
are the total velocities of (s; v; n)
directions.
V
s
=
o/
os
~q
in
~s
V
v
=
o/
ov
~q
in
~v (13)
V
n
=
o/
on
~q
in
~n
By combining Eqs.11 and 12, the following partial dif-
ferential equation for the cavity height calculation can
be derived [15].
oh
os
V
s
cos wV
v
[ [
oh
ov
V
v
cos wV
s
[ [
= V
n

oh
ot
!
sin
2
w (14)
where h is the cavity thickness normal to the blade
surface. Therefore, the thickness of the cavity surface
over the blade surface is determined by solving the
partial dierential Eq. 14. Since the normal derivative,
o/
on
, is known after the integral Eq. 4 is solved for the
assumed cavity planform, cavity thickness can be
updated by solving Eq. 14.
Similarly, the partial dierential equation for the
cavity height on wake surface, (h
w
), i.e., in the case of
supercavity, can be derived as follows:
o/

on

o/

on
=
oh
w
ot
[~q
c
[
oh
w
os
(15)
where ~q
c
is the cavity velocity corresponding to the
cavitation number, and superscripts and denote the
upper and lower surface of trailing wake sheet.
Fig. 2 Local coordinates on the blade and hub panel: orthogonal
(u; w; n), and non-orthogonal (u; v; n) curvilinear coordinates are
shown with local total velocity components
43
v Dynamic boundary condition on cavity surface
requires the pressure everywhere on the cavity surface
to be constant and equal to the cavitating pressure,
P
c
. From the unsteady Bernoullis equation dened in
the propeller xed coordinates system, the pressure
on the cavity surface has to satisfy the followings:
P
o
q

1
2
[~q
w
[
2
=
o/
ot

P
c
q

1
2
[~q
t
[
2

1
2
x
2
r
2
gy (16)
where P
o
is the pressure far upstream on the shaft
axis. g and q are the gravitational constant and uid
density, respectively. y is the vertical distance from the
axis of rotation, and r is the distance from the axis of
rotation.
Therefore, the magnitude of the total velocity on the
cavity surface is obtained from Eq. 16 in terms of the
cavitation number,
[~q
t
[ =

n
2
D
2
r
n
[~q
w
[
2
x
2
r
2
2gy 2
o/
ot
r
; (17)
and
r
n
=
P
o
P
c
1
2
qn
2
D
2
(18)
where n =
x
2p
is the propeller revolution per unit time,
and D is the propeller diameter.
The quadratic equation in terms of unknowns,
o/
os
and
o/
ov
, can be obtained by combining Eqs. 12 and 17, and
nally the following expression can be derived by solving
the quadratic equation.
o/
os
= ~q
in
~s
o/
ov
~q
in
~v

cos h
sin h

[~q
t
[
2

o/
ov
~q
in
~v

2
s
(19)
Note that the total normal velocity term of V
n
is elimi-
nated from Eq.19. If s; v, and n are located on the correct
cavity surface, the ow has to be tangent to the cavity
surface to satisfy the kinematic boundary condition, and
consequently the total normal velocity would vanish on
cavity surface. Even though the prescribed cavity surface
is not the converged, leaving the normal velocity term
out of the dynamic condition has a small eect on the
solution, as shown in [10]. The potential on the cavity
surface is determined by integrating Eq. 19 with respect
to s over cavity surface.
o/
os
on supercavity surface can be derived in a similar
manner.
o/

os
= ~q
in
~s

[~q
c
[
2
(V

u
)
2
q
(20)
where V

u
denotes the total cross ow velocity on the
upper wake surface.
v The cavity detachment location is iteratively deter-
mined to satisfy the smooth detachment conditions,
described in [22].
1. The cavity thickness is non-negative at its leading
edge.
2. The pressures on the wetted parts of blades and hub
should be greater than the vapor pressure.
Initially, the detachment location of the cavity on the
blade is guessed from the fully wetted pressure, in which
the C
P
is greater than the given cavitation number.
During the iteration step to nd the correct cavity
length, the detachment location is also updated based on
the blade pressure computed at each iteration step.
Eventhough the cavity detachment and the extend of
cavity are aected by the viscosity, the present method
does not consider the eect of viscosity since the
potential method is used.
Integral equation
The integral equation for solving the propeller and
tunnel problem can be directly derived from Eq. 4 by
applying the boundary conditions.
2p/(~x; t) =
Z Z
S
Bw
(t)S
C
(t)
/
q
(~x; t)
o
on
q
(t)
1
R(p; q)

1
R(p; q)
~q
in
~n ( )
!
ds

Z Z
S
W
(t)
D/
q
(~x; t)
o
on
q
(t)
1
R(p; q)
ds (21)

Z Z
S
T
(t)
/
q
(~x; t)
o
on
q
(t)
1
R(p; q)
ds
Notice that
o/
on
termonthe tunnel boundary is not shownin
Eq. 21 due to boundary conditions given in Eqs. 6 and 7.
As shown in Eq. 21, the propeller and tunnel inter-
action is solved directly without involving iterations, and
it results in increasing computing time to solve this
problem. However, the inclusion of circular tunnel wall
shape allows to solve the problem only in between two
Fig. 3 Modeling of N3745 propeller and tunnel walls
44
adjacent blades row because of its repeatability of panel
shapes, and helps reducing CPU time. In addition, the
solving procedure is accelerated by splitting the inuence
of other components with several blocks and solving
iteratively [9]. This scheme can applied to the arbitrary
tunnel shape, but the full tunnel wall and propeller
blades have to be solved together. In this case, the iter-
ative method would be better method for solving the
propeller and tunnel problem.
Numerical Results
In order to validate the present method, the convergence
studies are rst performed for N3745 propeller subject to
an uniform inow inside of a circular tunnel. Figure 3
shows the geometries of N3745 propeller, its wake, hub
and tunnel. The tunnel geometry shown in Fig. 3 is the
only part of tunnel between two adjacent blades. Notice
that the panels on tunnel walls are aligned to follow the
pitch angle of the blade tip. The axial spacing on tunnel
wall where the blade is located is same as the chordwise
spacing of the blade, and a cosine spacing is used for the
last of the tunnel parts. The tunnel length upstream and
downstream of the propeller are taken equal to 3 and 6
which are normalized by propeller radius, respectively.
Cavitation number, r
n
= 5:0, Froude number, F
n
= 2:0,
and an advance coecient, J
s
= 0:6 are used for the
convergence study. The eect of grid types on tunnel
walls is rst investigated by comparing the circulation
calculated at the fully wetted conditions. Two types of
grid on tunnel walls, the straight panel (Fig. 4) and the
hyperboloidal panel aligned with blade pitch angle (Fig.
3), are used. As shown in Fig. 5, the circulation distri-
butions from two dierent panel types are well com-
pared when the gap size increases. As decreasing the gap
size, the dierence of circulation distribution increases,
and it shows clear dierence at the zero gap condition.
When the blade tip is attached to the tunnel wall, it is
well known that the circulation is nite and close to zero
slope at the blade tip, and the panel type aligned with
blade pitch angle well predicts this trend at the zero gap.
Eventhough the straight panel predict circulation very
well except for the zero gap case, the calculation is
performed with the aligned panel type in order to avoid
the problem at the zero gap case. The convergence of the
present numerical method is tested by changing the
number of panels along chordwise and spanwise direc-
tion on the blade. The dependency of cavitating circu-
lation on the chordwise number of panel is shown in
Fig. 6, and the results converge to the nest result as
increasing number of chordwise panels. Figures 7 and 8
show the dependence of fully wetted and cavitating cir-
culation distributions on the spanwise grid size. The gap
between blade tip and tunnel wall is equal to zero, and
60 panels along chordwise direction are used on the
blade. The predicted steady wetted circulation distribu-
tion shows very quick convergence with increasing the
number of spanwise panels. The overall convergence of
cavitating circulation is slower than that of wetted one,
however, the converged circulation distributions with
the spanwise number of panels are well compared each
Fig. 4 Paneling on tunnel walls using straight panel
Fig. 5 Comparison of circulation distribution between the straight
panel and the aligned panel with blade pitch angle at fully wetted
condition: J
s
= 0:6. Wake length = 2.5R
Fig. 6 Convergence of steady cavitating circulation distributions with
chordwise number panels: gap size = 0.0. J
s
= 0:6, r
n
= 5:0, and
F
n
= 2:0. Wake length = 2.5R. 40 panels along blade spanwise
direction
45
other as shown in Fig. 8. In addition, the predicted
cavity patterns show nearly identical shapes as depicted
in Fig. 9. The eects of upstream and downstream
tunnel length on the propeller forces per blade are shown
in Figs. 10 and 11. The downstream tunnel length is
taken equal to 6 in Fig. 10, and the upstream tunnel
length is 3 in Fig. 11 The number of circumferential
panels between two adjacent blades is 10, and the panel
size for each tunnel length is kept same by proportion-
ally increasing the number of panels along longitudinal
direction. Even though the predicted K
T
and K
Q
are not
much sensitive to upstream and downstream tunnel
length, the predicted forces converges to the constant
value as increasing tunnel length. Figure 12 shows the
eect of transition wake length on the circulation dis-
tribution at the fully wetted condition. As increasing the
wake length, the circulation distribution converges to
the longest wake. The radial distributions of steady
wetted and cavitating circulation on the blades for zero,
5%, 10%, 50%, and 100% gap of the propeller radius
are shown in Figs. 13 and 14. The circulation distribu-
tion without tunnel wall eect is shown as well. As
expected, decreasing the gap size from 100% to zero
increases the maximum circulation, and the circulation
at the tip does not change much except for the case of
zero gap. In the case that gap between tunnel wall and
blade tip is small, the strong tip vortex which the viscous
eect is dominant would develop from the leading edge
of the blade and aects the performance of propeller [3,
4]. In order to capture the viscous ow eect, the
method of the orice equation model which has been
successfully applied to the panel method by [13] and [20]
has to to be coupled with the present method. However,
the gap model is not included in the present method, and
the result for small gap such as 0:5% would be less
reliable. The cavity planforms for each gap size are
shown in Fig. 15. Even though the gap size decreases
from 100 to 5%, the cavity patterns do not change much
comparing with zero gap. Since the circulation for zero
gap are lower than those of others up to 0.7 radius, the
cavity length in that region is smaller than those of
others.
The method is applied to determine the cavity shapes
and force predictions on N4148 propeller. The geome-
tries of N4148 propeller and tunnel walls are shown in
Fig. 16. N4148 propeller is subject to a non-axisym-
metric inow which is shown in Fig. 17, and the test
conditions are J
s
= 0:954, r
n
= 2:576, and F
n
= 9:159.
Fig. 8 Convergence of steady cavitating circulation distributions with
spanwise number panels: gap size = 0.0. J
s
= 0:6, r
n
= 5:0, and
F
n
= 2:0. Wake length = 2.5R. 60 panels along blade chordwise
direction
Fig. 9 Convergence of cavity shapes with spanwise number of panels:
gap size = 0.0. J
s
= 0:6, r
n
= 5:0, and F
n
= 2:0. Wake length =
2.5R. 60 panels along blade chordwise direction
Fig. 7 Convergence of steady wetted circulation distributions with
spanwise number panels: gap size = 0.0. J
s
= 0:6, r
n
= 5:0, and
F
n
= 2:0. Wake length=2.5R. 60 panels along blade chordwise
direction
46
The eective wake used in this computation includes the
eect of tunnel walls and the vortical inow and pro-
peller interactions, which is taken into account using
Euler solver [7]. The harmonics of propeller forces,
which are evaluated for time step size, Dh, of 12

, 10

, 8

,
and 6

, are shown in Fig. 18 and Table 1 for the fully


wetted condition, and Fig. 19 and Table 2 for the cavi-
tating condition.
The eect of time step size on the unsteady forces
are negligible in both wetted and cavitating cases. The
cavity shapes at zero blade angle position are shown
in Fig. 20, and the cavity shapes are nearly identical
each other, which is consistent with unsteady propeller
forces shown in Fig. 19. In order to validate the
present method, the results from the present method is
compared with that of the experiment. Even though
the tunnel section from the experiment and eective
wake calculation was square, the present method uses
circular section which keeps same sectional area as
that of square tunnel. The radius of circular tunnel
used in the present method is 1.88 times of propeller
radius. The predicted cavity shapes compared with
those observed in the experiment are shown in Fig. 21.
The observed blade angles are 30

, 6

, and 30

. The
Fig. 10 Comparison of K
T
and K
Q
per blade with tunnel
length upstream of propel-
ler: gap size = 0.0. J
s
= 0:6,
r
n
= 5:0, and F
n
= 2:0. Tun-
nel length downstream of
propeller is x
d
= 6:0. 60
panels along blade chord-
wise direction
Fig. 11 Comparison of K
T
and K
Q
with tunnel length
downstream of propeller:
gap size = 0.0. J
s
= 0:6,
r
n
= 5:0, and F
n
= 2:0.
Wake length = 2.5R. Tun-
nel length upstream of pro-
peller is x
u
= 3:0
47
predicted cavity shapes from the present method agree
well with those observed from experiment. The eects
of gap sizes on the predicted cavity volumes per blade
are shown in Fig. 22. As decreasing gap sizes, the
drastical increase in cavity volumes is shown in the
gure. Figure 22 shows the eects of gap sizes on
unsteady cavitating thrust and torque coecient per
blade. The unsteady forces predicted in the wetted
Fig. 12 Comparison of fully wetted circulation distribution with wake
length downstream of propeller: gap size = 0.0, and J
s
= 0:6. Tunnel
length: 3 _ x _ 6. 60 40 panels on blade
Fig. 13 Comparison of steady wetted circulation distributions with
gap size: J
s
= 0:6, r
n
= 5:0, and F
n
= 2:0. Wake length = 2.5R.
Tunnel length: 3 _ x _ 6. 60 40 panels on blade
Table 1 Fully wetted force harmonics per blade
K
T
10K
Q
n=0 n=1 n=2 n=0 n=1 n=2
6 0.03726 0.01767 0.00589 0.0611 0.0254 0.0082
8 0.03702 0.01791 0.00593 0.0607 0.0257 0.0082
10 0.03677 0.01731 0.00567 0.0602 0.0248 0.0078
12 0.03642 0.01703 0.00551 0.0596 0.0244 0.0076
Table 2 Cavitating force harmonics per blade
K
T
10K
Q
n=0 n=1 n=2 n=0 n=1 n=2
6 0.03632 0.01573 0.00451 0.0596 0.0222 0.0060
8 0.03612 0.01603 0.00461 0.0593 0.0226 0.0061
10 0.03586 0.01551 0.00453 0.0588 0.0218 0.0060
12 0.03561 0.01552 0.00474 0.0584 0.0218 0.0063
Fig. 14 Comparison of steady cavitating circulation distributions with
gap size. J
s
= 0:6, r
n
= 5:0, and F
n
= 2:0. Wake length = 2.5R.
Tunnel length: 3 _ x _ 6. 60 40 panels on blade
Fig. 15 Comparison of cavity planforms with gap size: J
s
= 0:6,
r
n
= 5:0, and F
n
= 2:0. Tunnel length: 3 _ x _ 6. 60 40 panels on
blade
48
region (between 60

and 300

) are not sensitive to the


gap size except for the zero gap case.
conclusions
A potential based boundary element method was
applied to predict the unsteady performance of the
propellers subject to a non-axisymmetric inow with the
eect of tunnel walls. The propeller/tunnel problems
were formulated with one integral equation to solve
them simultaneously and account for the fully unsteady
tunnel eect.
The present method was applied to the N3745 pro-
peller subject to an uniform inow, the convergence of
the circulation distribution and forces with number of
Fig. 16 Modeling of N4148 propeller and tunnel walls
Fig. 17 Non-axisymmetric inow wake for N4148 propeller
Fig. 18 Eects of time step size
on the unsteady wetted forces:
gap size = 0. J = 0:954,
r
n
= 2:576, and F
n
= 9:159.
Tunnel length: 3 _ x _ 6
Fig. 19 Eects of time step size
on the unsteady cavitating
forces: gap size = 0, J = 0:954,
r
n
= 2:576, and F
n
= 9:159.
Tunnel length: 3 _ x _ 6
49
panels was studied, and the eect of gap size was
investigated. The method was also applied in the case of
a propeller subject to a non-axisymmetric inow, and
validated the present method by comparing the cavity
patterns with those observed in the experiment. In order
to improve cavity prediction compared with the experi-
mental results, however, it is essential to predict the
eective wake more accurately by coupling the present
method with the Euler solver which solves ow eld
Fig. 20 Eects of time step size on cavity shapes at blade position
h = 0

: gap size = 0. J = 0:954, r


n
= 2:576, and F
n
= 9:159. Tunnel
length: 3 _ x _ 6
Fig. 21 Comparison of cavity shapes observed in experiment (top),
and computed from present method (bottom): J = 0:954, r
n
= 2:576,
and F
n
= 9:159. Tunnel length: 3 _ x _ 6
Fig. 23 Comparison of cavitating K
T
and K
Q
per blade with gap size:
J = 0:954, r
n
= 2:576, and F
n
= 9:159. Tunnel length: 3 _ x _ 6
Fig. 22 Comparison of cavity volumes with gap size: J = 0:954,
r
n
= 2:576, and F
n
= 9:159. Tunnel length: 3 _ x _ 6
50
inside of a tunnel with body force terms representing
propeller forces.
Acknowledgement Support for this research was provided by Phase
IV of the Consortium on Cavitation Performance of High Speed
Propulsors with the following members: AB Volvo Penta,
American Bureau of Shipping, Daewoo Shipbuilding and Marine
Engineering Co. Ltd., Kawasaki Heavy Industry, Naval Surface
Warfare Center Carderock Division, Oce of Naval Research
(Contract N000140110225), Rolls-Royce Marine AB, Rolls-Royce
Marine AS, VA Tech Escher Wyss GMBH, Wartsila Propulsion
Netherlands B.V. and Wartsila Propulsion Norway AS.
References
1. Bal S, Kinnas SA, Lee HS (2001) Numerical analysis of 2-D and
3-D cavitating hydrofoils under a free surface. J Ship Res
45:3449
2. Bal S, Kinnas SA (2003) A numerical wave tank for cavitating
hydrofoils. J Comput Mech 32:56
3. Brewer, W. H., Newman, J. C., Burgreen, G. W. & Burg, C. O.
E. (2003), A design method for investigating cavitation delay, in
The 8th International Conference on Numerical Ship Hydro-
dynamics, Busan, Korea
4. Chesnakas, C. & Jessup, S. D. (2003), Tip-vortex induced cavi-
tation on a ducted propulsor, in International Symposium on
Cavitation Inception, ASME FED Summer Meeting, Honolulu
5. Choi J.-K, Kinnas SA (1998) Numerical water tunnel in two
and three dimensions. J Ship Res 43(2):8698
6. Choi J.-K, Kinnas SA (1999) Numerical modeling of cavitating
propeller inside of a tunnel. J Fluids Eng 121:297304
7. Choi, J.-K. & Kinnas, S. A. (2000), An unsteady three-dimen-
sional euler solver coupled with a cavitating propeller analysis
method, in The 23rd Symposium on Naval Hydrodynamics,
Val de Reuil, France
8. Choi, J.-K. & Kinnas, S. A. (2001), Prediction of non-ax-
isymmetric eective wake by a 3-D Euler solver, Journal of
Ship Research 45(1), 1333
9. Clark, R. (1985), A new iterative matrix solution procedure for
three dimensional panel methods, in 23rd Aerospace Science
Meeting, AIAA, Reno, Nevada
10. Fine NE (1992) Nonlinear analysis of cavitating propellers in
nonuniform ow. Department of Ocean Engineering, Massa-
chusetts Institute of Technology
11. Fine NE, Kinnas SA (1993) The nonlinear numerical prediction
of unsteady sheet cavitation for propellers. In: Sixth Int Conf
On Numer Ship Hydrodynamics University of Iowa, Iowa, pp
531544
12. Glauert, H. (1947), The Elements of Aerofoil and Airscrew
Theory, 2nd Edition, Cambridge University Press, New York
13. Hughes MJ (1997) Impementation of a Special procedure for
modeling the tip clearance Flow in a panel method for ducted
propulsors. In: Propellers/Shafting 97 Symposium, Virginia
Beach
14. Kinnas SA, Fine NE (1992) A nonlinear boundary element
method for the analysis of unsteady propeller sheet cavitation.
In: Nineteenth Symposium on Naval Hydrodynamics. Seoul,
pp 717737
15. Kinnas, S. A. & Fine, N. E. (1993), A numerical non-linear
analysis of the ow around 2-d and 3-d partially cavitating
hydrofoils, Journal of Fluid Mechanics 254, 151181
16. Kinnas SA, Grin PE, Mueller AC (1997) Computational tools
for the analysis and design of high speed propulsors. The
International CFD Conference, Ulsteinvik
17. Kinnas SA, Lee HS, Mueller AC (1998) Prediction of propeller
blade sheet and developed tip vortex cavitation. 22nd Sympo-
sium on Naval Hydrodynamics. August, 914 Washington,
D.C. 182198
18. Kinnas SA, Hong S, Lee HS (2003) Numerical analysis of ow
around the cavitating CAV2003 hydrofoil. CAV2003: Fifth Int
Symp cavitation. November, 14 Osaka, Japan, pp 14
19. Lee, J.-T. (1987), A Potential Based Panel Method for The
Analysis of Marine Propellers in Steady Flow, PhD thesis,
Department of Ocean Engineering, Masachusetts Institute of
Technology
20. Moon IS, Kim KS, Lee C.-S, (2002) Blade tip gap ow model
for performance analysis of waterjetpropulsors. Int Association
for Bound Elem Meth. Austin TX, May 2830
21. Mueller AC, Kinnas SA(1999) Propeller sheet cavitation pre-
dictions using a panel method. J Fluids Eng June 121:282288
22. Young, Y. L. (2002), Numerical Modeling of Supercavitating
and Surface-Piercing Propellers, PhD thesis, Department of
Civil Engineering, The University of Texas at Austin
51

You might also like