Analysis of supercavitating and surface-piercing propeller flows via BEM

Y. L. Young, S. A. Kinnas
Abstract A low-order potential based 3-D boundary
element method (BEM) is presented for the analysis of
unsteady sheet cavitation on supercavitating and surface-
piercing propellers. The method has been developed in the
past for the prediction of unsteady sheet cavitation for
conventional propellers. To allow for the treatment of
supercavitating propellers, the method is extended to
model the separated ow behind trailing edge with non-
zero thickness. For surface-piercing propellers, the nega-
tive image method is used, which applies the linearized
free surface boundary condition with the innite Froude
number assumption. The method is shown to converge
quickly with grid size and time step size. The predicted
cavity planforms and propeller loadings also compare well
with experimental observations and measurements.
Keywords BEM, Cavitation, Super cavitating propeller,
Surface-piercing propeller
1
Introduction
Hydrodynamic cavitation is dened as the formation and
collapse of partial vacuums in a liquid by a swiftly moving
solid body. It can occur in any hydrodynamic devices that
operate in liquid when pressure drops below the saturated
vapor pressure. In the past, the goal in the design of
hydrodynamic devices was to avoid cavitation because of
its undesirable effects, which include blade surface erosion,
increased hull pressure uctuation and vibration, acoustic
energy radiation, and blade vibration. However, few pro-
pellers in practice can operate entirely without cavitation
due to the non-axisymmetric inow or unsteady body
motion. Furthermore, propellers without cavitation would
need to be larger and slower than necessary.
The objective of this work is to extend a 3-D boundary
element method, which has been developed in the past for
the prediction of unsteady sheet cavitation on conven-
tional fully submerged propellers, to predict the perfor-
mance of supercavitating and surface-piercing propellers.
1.1
Supercavitating propellers
Supercavitating propellers are often believed to be the
most fuel efcient propulsive device for high speed vessels.
The term supercavity refer to a cavity that is longer than
the blade. It tends to have smaller volume change and
produce bubbles which collapse downstream of the blade
trailing edge, which results in reduced noise and blade
surface erosion.
The development of numerical methods for the analysis
and design of supercavitating propellers has been slow
compared to conventional propellers. The main difculty
arises from the unknown physics in the highly turbulent
region behind blunt trailing edges, which is characteristic
of supercavitating propeller sections. The rst theoretical
design method was developed by (Tachnimdji and Mor-
gan, 1958), and followed by (Tulin, 1962; Cox, 1968; Barr,
1970; Yim, 1976). However, these methods were based on
2-D studies, and required many approximations and
empirical corrections. Recently, more rigorous methods
were developed by (Kamirisa and Aoki, 1994; Kikuchi
et al., 1994; Vorus and Mitchell, 1994; Ukon et al., 1995).
Nevertheless, these methods were still based on the opti-
mization of 2-D cavitating blade sections to yield minimal
drag for a given lift and cavitation number.
A 3-D vortex-lattice method was developed by (Kudo
and Ukon, 1994) to predict the steady performance of
supercavitating propellers. Their model assumed the
pressure over the separated zone to be constant and equal
to the vapor pressure. A variable length separated zone
model using a similar vortex-lattice method was presented
in (Kudo and Kinnas, 1995).
However, all of the above mentioned lifting surface
methods cannot capture accurately the ow details at the
blade leading and trailing edge due to the breakdown of
linear cavity theory. In addition, the applicability of the
thickness-loading coupling introduced by (Kinnas, 1992)
Computational Mechanics 32 (2003) 269280 Springer-Verlag 2003
DOI 10.1007/s00466-003-0484-6
269
Y. L. Young
Department of Civil and Environmental Engineering,
E-326, Princeton University, Princeton, NJ, 08544
This work was completed while as a doctoral student at The
University of Texas at Austin
S. A. Kinnas (&)
Ocean Engineering Group, Department of Civil Engineering,
C1786, The University of Texas at Austin,
1 University Station, Austin, TX 78712
E-mail: kinnas@mail.utexas.edu
Support for this research was provided by Phase III of the
Consortium on Cavitation Performance of High Speed
Propulsors with the following members: AB Volvo Penta,
American Bureau of Shipping, El Pardo Model Basin, Hyundai
Maritime Research Institute, Kamewa AB, Michigan Wheel
Corporation, Naval Surface Warfare Center Carderock Division,
Ofce of Naval Research (Contract N000140110225), Ulstein
Propeller AS, VA Tech Escher Wyss GMBH, and Wartsila
Propulsion.
in the analysis of supercavitating propellers is still under
investigation.
1.2
Surface-piercing propellers
A surface-piercing (also called partially submerged) pro-
peller is a special type of supercavitating propeller which
operates at partially submerged conditions. Surface-
piercing propellers are more efcient than submerged
supercavitating propellers because of (1) the reduction of
appendage drag, and (2) the possibility of larger propeller
diameter.
The rst theoretical effort to model partially submerged
propeller was carried out by (Oberembt, 1968) using a
lifting line approach. He assumed the propeller to be
lightly loaded such that no natural ventilation of the
propeller and its vortex wake occurs. Later, (Furuya, 1985)
also used a lifting-line approach, but he included the effect
of propeller ventilation. The method was combined with a
2-D water entry-and-exit theory developed by (Wang,
1979) to determine thrust and torque coefcients.
An unsteady lifting surface method was employed by
(Wang et al., 1990) for the analysis of 3-D fully ventilated
thin foils entering into initially calm water. The method
was later extended by (Wang et al., 1992) to predict the
performance of fully ventilated partially submerged pro-
peller with its shaft above the water surface. Similar to
(Furuya, 1985), the negative image method was used to
account for free surface effect, and the blade was assumed
to be fully ventilated on the suction side starting from the
blade leading edge. The effect of the blade thickness was
also neglected in the computation.
The 3-D vortex-lattice lifting surface method developed
by (Kudo and Kinnas, 1955) for the analysis of supercav-
itating propellers has also been extended for the analysis of
surface-piercing propellers. However, the method per-
forms all the calculations assuming the propeller is fully
submerged, then multiplies the resulting forces with the
propeller submergence ratio. As a result, only the mean
forces can be predicted while the complicated phenomena
of blades entry to, and exit from, the water surface are
completely ignored.
A 2-D time-marching BEM was developed by (Savineau
and Kinnas, 1995) for the analysis of the ow eld around
a fully ventilated partially submerged hydrofoil. However,
this method only accounts for the hydrofoils entry to, but
not exit from, the water surface.
1.3
A low-order potential based BEM
In the present method, a low-order (piecewise constant
dipole and source distributions) potential-based boundary
element method is used to predict the performance of
supercavitating and surface-piercing propellers.
The low-order potential based BEM was rst applied for
the analysis of marine propeller in steady ow by (Lee,
1987; Kerwin et al., 1987) and unsteady ow by (Hsin, 1990;
Kinnas and Hsin, 1992). The method was then extended for
the analysis of ow around 2-D partially and supercavi-
tating hydrofoils (Kinnas and Fine, 1991) and 3-D partially
cavitating hydrofoils (Fine and Kinnas, 1993a). In (Kinnas
and Fine, 1992), the method was named PROPCAV
(PROPeller CAVitation) for its added ability to analyze 3-D
unsteady ow around cavitating propellers. Later, (Muller
and Kinnas, 1999) modied the method to search for
midchord cavitation on either the back or the face of
propeller blades. Most recently, (Young and Kinnas, 2001)
extended the method to predict alternating or simultaneous
face and back cavitation on conventional propeller blades
subjected to non-axisymmetric inow. The boundary ele-
ment method inherently includes the effect of non-linear
thickness-loading coupling by discretizing the blade sur-
face instead of the mean camber surface. Thus, it requires
more central processing unit (CPU) time and memory than
the lifting surface method. However, it offers a better pre-
diction of the ow details at the propeller leading edge and
tip than the lifting surface method.
2
Formulation
The formulation for supercavitating propellers is similar to
that for conventional cavitating propellers. It is given in
(Kinnas and Fine, 1992; Young and kinnas, 2001), and is
summarized here for the sake of completeness.
Consider a cavitating propeller subjected to a general
inow wake ~qq
w
(x
s
; y
s
; z
s
)
1
, as shown in Fig. 1. The inow
velocity, ~qq
in
, with respect to the propeller xed coordinates
(x; y; z), can be expressed as the sum of the inow wake
velocity, ~qq
w
, and the propellers angular velocity ~ xx, at a
given location ~xx:
~qq
in
(x; y; z; t) =~qq
w
(x; r; h
B
xt) ~ xx ~xx (1)
where r =

y
2
z
2
p
, h
B
= arctan(z=y), and ~xx = (x; y; z).
The resulting ow is assumed to be incompressible and
inviscid. Hence, the total velocity, ~qq, can be expressed in
terms of ~qq
in
and the perturbation potential /:
~qq(x; y; z; t) =~qq
in
(x; y; z; t) \/(x; y; z; t) (2)
Fig. 1. Propeller subjected to a general inow wake. The pro-
peller xed (x; y; z) and ship xed (x
s
; y
s
; z
s
) coordinate systems
are shown. From (Young and Kinnas, 2003)
1
~qq
w
is assumed to be the effective wake, i.e. it includes the
interaction between the vorticity in the inow and the propeller
(Kinnas et al., 2000; Choi, 2000).
270
where / satises the Laplaces equation in the uid
domain (i.e. \
2
/ = 0). Note that the propeller xed
coordinates system is used in analyzing the ow.
2.1
The boundary integral equation
The perturbation potential, /, at every point p on the
combined wetted blade and cavity surface, S
WB
(t) ' S
C
(t),
must satisfy Greens third identity:
2p/
p
(t)
=
Z Z
S
WB
(t)'S
C
(t)
/
q
(t)
oG(p; q)
on
q
(t)
G(p; q)
o/
q
(t)
on
q
(t)
!
dS

Z Z
S
W
(t)
D/(r
q
; h
q
; t)
oG(p; q)
on
q
(t)
dS;
p S
WB
(t) ' S
C
(t) (3)
where the subscript q corresponds to the variable point in
the integration. G(p; q) = 1=R(p; q) is Greens function in
an unbounded 3-D uid domain, with R(p; q) being the
distance between points p and q. ~nn
q
is the unit vector
normal to the integration surface, with the positive
direction pointing into the uid domain. S
WB
(t) denotes
the wetted blade and hub surfaces, and S
C
(t) denotes the
cavitating surfaces.
The wake surface, S
W
(t), is assumed to have zero
thickness. The geometry of the wake surface is determined
by satisfying the force-free wake condition, which requires
zero pressure jump across the wake sheet. In this work, the
wake is aligned with the circumferentially averaged inow
using a iterative lifting surface method developed by
(Greeley and Kerwin, 1982). As stated in (Greeley and
Kerwin, 1982), this method articially suppresses the
wake roll-up. Recently, a fully unsteady wake alignment
method, including wake roll-up and developed tip vortex
cavity, is developed for propellers in non-axisymmetric
inows (Lee and Kinnas, 2001; Lee and Kinnas, 2002). The
formulation and results using this unsteady wake align-
ment method is presented in (Lee and Kinnas, 2001; Lee
and Kinnas, 2002; Lee, 2002).
The dipole strength D/(r; h; t) in the wake is convected
along the assumed wake model with angular speed x:
D/(r; h; t) = D/
T
r
T
; t
h h
T
x

; t _
h h
T
x
D/(r; h; t) = D/
S
(r
T
); t <
h h
T
x
(4)
where (r; h) are the cylindrical coordinates at any point in
the trailing wake surface (S
W
), and (r
T
; h
T
) are trailing
edge coordinates of the corresponding streamline. D/
S
is
the steady ow potential jump in the wake when
the propeller is subject to the circumferentially averaged
ow.
The value of the dipole strength, D/
T
(r
T
; t), at the
trailing edge of the blade at radius r
T
and time t, is given
by Morinos Kutta condition (Morino and Kuo, 1974):
D/
T
(r
T
; t) = /

T
(r
T
; t) /

T
(r
T
; t) (5)
where /

T
(r
T
; t) and /

T
(r
T
; t) are the values of the
potential at the upper (suction side) and the lower (pres-
sure side) blade trailing edge, respectively, at time t.
Note that Eq. (3) is a Fredholm singular integral
equation of the second kind. It should be applied on the
exact cavity surface S
C
, as shown on the top of Fig. 2.
However, the cavity surface is not known and has to be
determined as part of the solution. In this work, an
approximated cavity surface, shown on the bottom of
Fig. 2, is used. The approximated cavity surface is com-
prised of the blade surface underneath the cavity on the
blade, S
CB
(t), and the portion of the wake surface which is
overlapped by the cavity, S
CW
(t). The justication for
making this approximation, as well as a measure of its
effect on the cavity solution can be found in (Kinnas and
Fine, 1993; Fine, 1992).
Using the approximated cavity surface, Eq. (3) may be
decomposed into a summation of integrals over the blade
surface, S
B
(= S
CB
' S
WB
), and the portion of the wake
surface which is overlapped by the cavity, S
CW
.
2.2
Boundary conditions
Equation 3 implies that the perturbation potential (/
p
)
can be expressed as: (1) continuous source (G) and dipole
(oG=on) distributions on the wetted blade (S
WB
) and cavity
(S
CB
' S
CW
) surfaces, and (2) continuous dipole distribu-
tion on the wake surface S
W
. Thus, /
p
can be uniquely
determined by satisfying the following boundary
conditions:
Kinematic boundary condition on wetted
blade and hub surfaces
The kinematic boundary condition requires the ow to be
tangent to the wetted blade and hub surface, which forms a
Neumann-type boundary condition for o/=on:
Fig. 2. Top: Denition of the exact surface. Bottom: Denition
of the approximated cavity surface. From(Young andKinnas, 2001)
271
o/
on
= ~qq
in
~nn (6)
Dynamic boundary condition on cavitating surfaces
The dynamic boundary condition on the cavitating blade
and wake surfaces requires the pressure everywhere on the
cavity to be constant and equal to the vapor pressure, P
v
.
By applying Bernoullis equation, the total cavity velocity,
~qq
c
, can be expressed as follows:
~qq
c

2
= n
2
D
2
r
n
~qq
w

2
x
2
r
2
2gy
s
2
o/
ot
(7)
where r
n
= (P
o
P
v
)=(
q
2
n
2
D
2
) is the cavitation number; q
is the uid density and r is the distance from the axis of
rotation. P
o
is the pressure far upstream on the shaft axis;
g is the acceleration of gravity and y
s
is the ship xed
coordinate, shown in Fig. 1. n = x=2p and D are the
propeller rotational frequency and diameter, respectively.
The total cavity velocity can also be expressed in terms
of the local derivatives along the s (chordwise), v
(spanwise), and n (normal) grid directions:
~qq
c
=
V
s
~ss (~ss ~vv)~vv [ [ V
v
~vv (~ss ~vv)~ss [ [
|~ss ~vv|
2
(V
n
)~nn (8)
where ~ss, ~vv, and ~nn denote the unit vectors along the non-
orthogonal curvilinear coordinates s, v, and n, respectively.
The total velocities on the local coordinates (V
s
; V
v
; V
n
) are
dened as follows:
V
s
=
o/
os
~qq
in
~ss; V
v
=
o/
ov
~qq
in
~vv;
V
n
=
o/
on
~qq
in
~nn (9)
Note that if s, v, and n were located on the exact cavity
surface, then the total normal velocity, V
n
, would be zero.
However, this is not the case since the cavity surface is
approximated with the blade surface beneath the cavity
and the wake surface overlapped by the cavity. Although
V
n
may not be exactly zero on the approximated cavity
surface, it is small enough to be neglected in the dynamic
boundary condition (Fine, 1992).
Equations 7 and 8 can be integrated to form a quadratic
equation in terms of the unknown chordwise perturbation
velocity o/=os. By selecting the root which corresponds to
the cavity velocity vectors that point downstream, the
following expression can be derived:
o/
os
= ~qq
in
~ss V
v
cos w sin w

[~qq
c
[
2
V
2
v
q
(10)
where w is the angle between s and v directions, as shown
in Fig. 2. Equation 10 can then integrated to form a
Dirichlet type boundary condition for /. It should be
noted that the terms o/=ot and o/=ov inside [~qq
c
[ and V
v
in
Eq. 10 are also unknown and are determined in an
iterative manner.
On the cavitating wake surface, the coordinate s is
assumed to follow the streamlines. It was found that the
crossow term (o=ov) in the cavitating wake region had a
very small effect on the solution (Fine, 1992; Fine and
Kinnas, 1993b). Thus, the total cross ow velocity is as-
sumed to be small, which renders the following expression
on the cavitating wake surface:
o/
os
= ~qq
in
~ss [~qq
c
[ (11)
Kinematic boundary condition on cavitating surfaces
The kinematic boundary condition requires that the total
velocity normal to the cavity surface to be zero:
0 =
D
Dt
(n h(s; v; t)) =
o
ot
~qq
c
(x; y; z; t)
~
\\
!
(n h(s; v; t)) (12)
where n and h are the curvilinear coordinate and cavity
thickness normal to the blade surface, respectively.
Substituting Eq. (8) into Eq. (12) yields the following
partial differential equation for h on the blade (Kinnas and
Fine, 1992):
oh
os
V
s
cos wV
v
[ [
oh
ov
V
v
cos wV
s
[ [
= sin
2
w V
n

oh
ot

(13)
Assuming again that the spanwise crossow velocity on
the wake surface is small, the kinematic boundary condi-
tion reduces to the following equation for the cavity
thickness (h
w
) in the wake:
o/

on

o/

on
!

oh
w
ot
= [~qq
c
[
oh
w
os
(14)
Note that h
w
in Eq. 14 is dened normal to the wake
surface. In addition, the quantity h
w
at the blade trailing
edge is determined by interpolating the upper and/or
lower cavity surface over the blade and computing its
normal offset from the wake sheet.
Cavity closure condition
The extent of the unsteady cavity is unknown and has to
be determined as part of the solution. The cavity length at
each radius r and time t is given by the function l(r; t). For
a given cavitation number, r
n
, the cavity planform, l(r; t),
must satisfy the following condition:
d l(r; t); r; r
n
( ) = 0 (15)
where d is the thickness of the cavity trailing edge.
Equation 15 requires that the cavity closes at its trailing
edge. This requirement is the basis of a Newton-Raphson
iterative method that is used to nd the cavity planform
(Kinnas and Fine, 1993).
Cavity detachment condition
The cavity detachment locations are determined iteratively
by satisfying the Villat-Brillouin smooth detachment
condition (Brillouin, 1911; Villat, 1914). This condition
requires that the cavities do not intersect the blade at its
leading edge, and the pressure upstream of the cavities to
be greater than the vapor pressure. It should be noted that
the current work assumes the ow to be inviscid. It is
272
widely known that viscosity affects the cavity detachment,
as well as the extent and thickness of the cavity. However,
investigations by (Kinnas et al., 1994) concluded that the
effect of viscosity on the predicted cavity extent and
volume is negligible for the case of supercavitation.
2.3
Solution algorithm
The unsteady cavity problem is solved by inverting Eq. (3)
subject to Eqs. (4), (5), (6), (10), (11), (15), and the cavity
detachment condition. The integral surfaces are approxi-
mated with hyperboloidal panels (Kinnas and Hsin, 1992)
on which constant strength dipoles and sources are dis-
tributed. Eq. (3) is applied at the panel centroids in order
to determine the potentials at each panel. The problem is
solved in the time domain with constant time step size Dt.
The numerical implementation is described in detail in
(Kinnas and Fine, 1992). In brief, for a given cavity plan-
form, Greens formula is solved with respect to unknown /
on the wetted blade and hub surfaces, and unknown
o/=on on the cavitating surfaces. The cavity heights on the
blade and the wake are computed by differentiating
Eqs. (13) and (14) with a second order central nite dif-
ference method. The extent of cavities at each time step is
obtained iteratively using a Newton-Raphson technique
which requires Eq. (15) to be satised everywhere on the
blade at each time step. Finally, the correct cavity planform
is obtained by adjusting the cavity detachment locations
until the smooth detachment condition is satised. It
should be noted that the split-panel method developed by
(Kinnas and Fine, 1993) is used to treat panels which are
intersected by the cavity trailing edge. In addition, at each
time step, the solution is only obtained for the key blade.
The inuence of each of the other blades is accounted for
in a progressive manner by using the solution from an
earlier time step when the key blade was in the position of
that blade.
3
Supercavitating propellers
Experimental evidence showed that the separated zone
behind the thick blade trailing edge forms a closed cavity
that separates from the practically ideal irrotational ow
around a supercavitating blade section (Russel, 1958). In
addition, the pressure within the separated zone (also
called the base pressure) can be assumed to be uniform
(Riabouchinsky, 1926; Tulin, 1953). However, a turbulent
dissipation model (such as the one used in (Vorus and
Chen, 1987)) is necessary to determine the the mean base
pressure and the extent of the separated zone, which is not
practical for engineering purposes.
In the present method, the base pressure is assumed to
be constant and equal to the vapor pressure, similar to that
assumed in (Kudo and Ukon, 1994). Hence, the separated
zone can be solved for like an additional cavitation bubble.
To avoid openness at the blade trailing edge, a small
initial closing zone, shown in Fig. 3, is introduced. The
precise geometry of the initial closing zone is not impor-
tant, as long as it is inside the separated region and its
trailing edge lies on the aligned wake sheet. The method is
modied so that it treats the original blade and the initial
closing zone as one solid body. It should be noted that the
initial closing zone is used in the entire analysis, and its
geometry does not change with time. However, the size
and the extent of the cavities and the separated regions are
allowed to change with time, and are solved for as part of
the solution.
The solution method is the same as that for fully sub-
mergedconventional propellers. However, additional care is
needed to ensure the potential to be continuous between the
wetted portions of the blade, the cavity surfaces, and the
closing zones. An additional condition which requires the
cavities to detach prior to the actual blade trailing edge is
also needed. Furthermore, the pressure acting on the thick
blade trailing edge must also be included in the force cal-
culation. This is accomplished by multiplying the separated
region pressure acting normal to the blade trailing edge with
the trailing edge area. Details of the numerical algorithmand
numerical validations of the method are presented in
(Young, 2002; Young and Kinnas, 2003).
As depicted in Fig. 3, this scheme is applicable to fully
wetted, partially cavitating, and supercavitating conditions
in steady and unsteady ows assuming P = P
v
in the
Fig. 3. Treatment of supercavitating blade sections. From (Young
and Kinnas, 2003)
Fig. 4. Cavitation patterns on supercavitating propellers that can
be predicted by the present method. From (Young and Kinnas,
2003)
273
separated region. Cavitation patterns on supercavitating
propellers that can be predicted by the present method are
shown in Fig. 4.
3.1
Numerical validations
In order to validate the method, parametric studies are
conducted for a rectangular hydrofoil at an angle of attack
a = 1

. The cross section of the hydrofoil is modied from

that of a NACA66 thickness distribution with NACA a = 0.8
mean camber line. The maximum thickness to chord ratio
(T
max
=C) is 0.05. The thickness to chord ratio at the foil
trailing edge (T
TE
=C) is 0.02. The maximum camber to
chord ratio (f
max
=C) is 0.02. The predicted cavity planform
and cavitating pressure distribution are shown in Fig. 5.
Notice that there are midchord supercavitation on the back
side of the hydrofoil, and leading edge partial cavitation on
the face side of the hydrofoil. In addition, notice that the
smooth detachment condition and the cavity closure
condition are satised everywhere on the hydrofoil.
The inuence of the initial closing zone length on the
pressure and cavity shape is shown in Fig. 6. Since the
entire initial closing zone is inside the separated region,
the length of the initial closing zone should have negligible
inuence on the solution, which is conrmed by Fig. 6.
In order to assess the uniqueness of the solution, the
sensitivity of the total cavity volume to the initial guess of
cavity lengths on the suction side (l

i
) and on the pressure
side (l

i
) is shown in Fig. 7. The quantity NTSTEP denotes
the step number. Within each step, a maximum of 30
iterations are allowed to determine the cavity lengths which
satisfy Eq. (15) for the given guess of cavity detachment
locations. At the end of every step, the cavity detachment
locations are adjusted via the smooth detachment condi-
tion. Note that the initial guess of cavity lengths are non-
dimensionalized by the local chord length at each radial
strip. As shown in Fig. 7, the converged solution is inde-
pendent of the initial guess of cavity lengths.
The dependence of the 3-D cavity planform on the
tolerance (d
tol
) for cavity trailing edge openness [d as
dened in Eq. (15)] is shown in Fig. 8. Note that both d
tol
and d are non-dimensionalized by the local chord length at
each radial strip. As shown in Fig. 8, the solution is
insensitive to d
tol
for d
tol
_ 0:001.
The convergence of the method with number of panels
is shown in Fig. 9. The error shown in Fig. 9 is computed
as follows:
err = C
N
mid
C
NM
mid

(16)
where C
N
mid
is the circulation at midspan for N number of
panels on the foil surface. NM is number of panels on the
foil surface for the nest discretization, which is 7200. As
shown in Fig. 9, the convergence rate is approximately 1.4.
3.2
Validation with experiments
To validate the treatment of supercavitating propellers, the
predicted force coefcients are compared with
Fig. 5. 3-D hydrofoil. r
v
= 0:11, f
max
=C = 0:02, T
max
=C = 0:05,
a = 1

. Uniform inow
Fig. 6. Dependence on the initial closing zone length. XSR is the
length ratio of the initial closing zone to the chord. r
v
= 0:11,
f
max
=C = 0:02, T
max
=C = 0:05, a = 1

. Uniform inow
Fig. 7. Dependence on the initial guess of cavity lengths on
the suction side (l

i
) and pressure side (l

i
). r
v
= 0:11,
f
max
=C = 0:02, T
max
=C = 0:05, a = 1

. Uniform inow
274
experimental measurements (Matsuda et al., 1994) for a
supercavitating propeller. The test geometry is M.P.No.345
(SRI), which is designed using SSPA charts under the
following conditions: J
A
= V
A
=nD = 1:10,
r
v
= P
o
P
v
=0:5qV
2
A
= 0:40, and K
T
= T=qn
2
D
4
= 0:160.
V
A
is the propeller advance speed. T and Q are the pro-
peller thrust and torque. The discretized propeller geom-
etry is shown in Fig. 10. Comparisons of the predicted and
measured thrust (K
T
), torque (K
Q
= Q=qn
2
D
5
), and ef-
ciency [g
o
= (K
T
=K
Q
) (J
A
=2p)[ are shown in Fig. 11. Note
that the kink at J
A
= 1:2 in Fig. 11 is due to the shift
from leading edge supercavitation to midchord superca-
vitation, as depicted in Fig. 10. It is worth noting that the
smooth detachment condition is satised for all J
A
s. In
addition, the method converged quickly with number of
panels, as demonstrated in Fig. 12. The error, ERR, in
Fig. 12 is dened with respect to the non-dimensional
circulation at r=R = 0:7, C
0:7
, for the nest discretization:
ERR = C
N
0:7
C
NM
0:7

(17)
where N is the number of panels on the blade surface. NM
is number of panels on the blade surface for the nest
discretization, which is 5600. It should be noted that
Fig. 8. Dependence on the
tolerance for openness at cavity
trailing edge (d
tol
). r
v
= 0:11,
f
max
=C = 0:02,
T
max
=C = 0:05, a = 1

.
Uniform inow
Fig. 9. The rate of convergence with respect to number of panels
on the foil surface (N). r
v
= 0:11, f
max
=C = 0:02,
T
max
=C = 0:05, a = 1

. Uniform inow
Fig. 10. Discretized geometry of propeller SRI and predicted
cavity planform for J
A
= 1:2 and 1.3
Fig. 11. Comparison of the predicted and versus measured K
T
,
K
Q
, and g
o
275
extensive parametric studies of the method were also
performed for this propeller, and are presented in (Young
and Kinnas, 2003).
4
Surface-piercing propellers
Since surface-piercing propellers are partially submerged,
the computational boundary in Eq. (3) must also include
the free surface. As a rst step to model partially sub-
merged propellers in 3-D, the linearized free surface
boundary condition is applied to account for the effect of
the free surface:
o
2
/
ot
2
(x; y; z; t) g
o/
oy
s
(x; y; z; t) = 0
at y
s
= R h (i.e. free surface) (18)
where h and R are the blade tip immersion and blade
radius, respectively, as dened in Fig. 13. y
s
is the vertical
ship-xed coordinate, dened in Fig. 1.
Assuming that the innite Froude number condition
(i.e. F
r
= n
2
D=g ) applies, Eq. (18) reduce to:
/(x; y; z; t) = 0 at y
s
= R h (19)
The assumption that the Froude number grows without
bounds is valid because partially submerged propellers
usually operate at very high speeds. Studies by (Shiba,
1953; Brandt, 1973; Olofsson, 1996) have also shown that
the effect of Froude number is negligible for
F
nd
= V=

gD
_
> 4 or F
r
= n
2
D=g > 2 in the fully venti-
lated regime.
Equation 19 implies that the negative image method can
be used to account for the effect of the free surface. Con-
sequently, only vertical motions are allowed on the free
surface. This is accomplished by distributing sources and
dipoles of equal strengths but with negative signs on the
loaction of the mirror image with respect to the free sur-
face. A schematic example of the negative image method
on a blade section is shown in Fig. 14. By summing the
image inuence coefcients with the real inuence coef-
cients for each submerged panel, the discretized form of
Greens equation can be written as:
2p/
i
=
X
NK
k=1
X
M
k=1
X
NB
s
(m;k;t)
n=1

A
i;m;n;k
(t)/
m;n;k
(t)
(
:
B
i;m;n;k
(t)
o/
on
m;n;k
(t)
!

X
NCW
s
(m;k;t)
n=1
C
i;m;n;k
(t)Q
m;n;k
(t)

X
NCW
s
(m;k;t)
n=1
W
i;m;n;k
(t)\/
m;n;k
(t)
)
for i=1;...;MNB
s
(m;k;t)NCW
s
(m;k;t)
(20)
where
A
i;m;n;k
(t) = A
s
i;m;n;k
A
i
i;m;n;k
(t)
B
i;m;n;k
(t) = B
s
i;m;n;k
B
i
i;m;n;k
(t)
C
i;m;n;k
(t) = C
s
i;m;n;k
C
i
i;m;n;k
(t)
W
i;m;n;k
(t) = W
s
i;m;n;k
W
i
i;m;n;k
(t) (21)
The superscript s and i in Eq. (21) denote the sub-
merged panel and its image, respectively. A
i;m;n;k
represent
the potential induced at the ith submerged control point
Fig. 12. Convergence rate of the cavitating circulation distribu-
tion with respect to number of panels on each blade
Fig. 13. Denition of ship-xed (x
s
; y
s
; z
s
) and blade-xed
(x; y; z) coordinate systems
Fig. 14. Schematic example of the negative image method on a
partially submerged blade section
276
on the key blade by unit strength diploes at the real and
imaged nth panel on the mth strip of the kth blade. Note
that k = 1 refers to the key blade. Similarly, B
i;m;n;k
; C
i;m;n;k
,
and W
i;m;n;k
represent the sum of the real and image
inuence coefcients due to unit strength source on the
blade, unit strength source on the cavitating wake, and
unit strength dipole on the wake, respectively. Q
m;n;k
rep-
resents the cavity source strength on the nth panel of the
mth strip of the kth blade.
Note that the sign in front to the image coefcients in Eq.
(21) are negative due to the equal and opposite strengths of
the image singularities compared to the real singularities. In
addition, since the problem is solved with respect to the
propeller xed coordinates which rotates with the key blade,
the locationof the nth image onthe mthstripof the kthblade
changes with time t. The quantities NB
s
; NCE
s
, and NW
s
represent the number of submerged panels on the blade,
cavitating wake, and wake surface, respectively, on the mth
strip of the kth blade. The quantity NKin Eq. (20) represents
the total number blades, while M represents the total num-
ber of radial strips per blade.
4.1
Solution algorithm
For surface-piercing propellers, Greens formula is only
solved for the total number of sub-merged panels on the
key blade and the cavitating portion of the key wake. The
values of / and o/=on ( ) are set equal to zero on the blade
and wake panels that are above the free surface. Note that
the current algorithm does not re-panel the blades and
wakes at very time step in order to maintain computation
efciency. As a result, there are some panels that are
partially cut by the free surface. In the present algorithm,
the strengths of the singularities are also set equal to zero
for the partially submerged panels. Nevertheless, a method
similar to the split-panel technique (Kinnas and Fine,
1993) can be applied to account for the effects of these
panels.
The solution algorithm for partially submerged pro-
pellers is similar to that explained earlier for fully sub-
merged supercavitating propellers. However, iterations to
determine the correct cavity lengths are no longer neces-
sary since the ventilated cavities are assumed to vent to the
atmosphere, as observed in experiments.
Ventilated cavity detachment search algorithm
Depending on the ow conditions and the blade section
geometry, ventilated cavities may detach aft of the blade
leading edge. Thus, the cavity detachment locations on the
suction side of the blade are searched for iteratively at each
time step until the smooth detachment condition is satis-
ed. In addition, due to the interruption of the free sur-
face, the following detachment conditions must also be
satised for partially submerged propellers:
v The ventilated cavities must detach at or prior to the
blade trailing edge; and
v During the exit phase (i.e. when part of the blade is
departing the free surface), the ventilated cavities must
detach at or aft of the intersection between the blade
section and the free surface.
It should be noted that the ventilated cavities on the
pressure side of the blade are always assumed to detach
from the blade trailing edge. It is possible to also search for
cavity detachment locations on the pressure side. How-
ever, such occurrence is unlikely due to the high-speed
operation of partially submerged propellers.
4.2
Validation with experiments
In order to validate the treatment of partially submerged
propellers, numerical predictions for propeller model 841-
B are compared with experimental measurements collected
by (Olofsson, 1996). A photograph of the partially sub-
merged propeller and the corresponding BEM model are
shown in Fig. 15. The experiments were conducted at the
free-surface cavitation tunnel at KaMeWa of Sweden.
Details of th experiments are given in (Olofsson, 1996).
The current method assumes the blade to be rigid, the
cavities to be fully ventilated, and the Froude number to be
innite. Thus, the following combination of test conditions
were selected to minimize the effect of Froude number,
cavitation number, and blade vibration:
shaft yaw angle: w = 0

advance coefficient: J
A
=
V
A
nD
= 1:2
shaft inclination angle: c = 0

Froude number: F
nD
=
V
s

gD
_ = 6:0
blade tip immersion: h=D = 0:33
cavitation number: r
v
=
P
o
P
v
0:5qV
2
s
= 0:25
Note that P
o
is the pressure far upstream on the shaft axis.
Comparisons of the observed and predicted ventilated
cavity patterns are shown in Fig. 16. Comparisons of the
measured and predicted individual blade force and mo-
ment coefcients are shown in Fig. 17. The solid lines and
the symbols in Fig. 17 represent the load coefcients pre-
dicted by the present method and measured in experi-
ments, respectively. (K
F
X
; K
F
Y
; K
F
Z
; K
M
X
; K
M
Y
; K
M
Z
) are the
six components of the individual blade force and moment
coefcients dened in the coordinate system shown in
Fig. 1. As shown in Figs. 16 to 17, the predicted ventilated
Fig. 15. Photograph of propeller model 841-B shown in (Olofsson
1996), with corresponding BEM model on the right
277
cavity patterns and per blade force agree well with
experiments. In addition, the method also converged
quickly with time step size and grid size, as shown in
Figs. 18 and 19.
5
Conclusions
A 3-D boundary element method has been extended to
predict complex types of cavitation patterns on the back
and the face of supercavitating propellers in steady and
unsteady ow. The current algorithm assumes the pressure
to be constant and equal to the vapor pressure on the
separated region behind non-zero thickness blade trailing
edge. Based on this assumption, the method is able to
predict the extent and thickness of the separated region.
The numerical predictions compared well with experi-
mental measurements for steady inow. The solution also
converged with number of panels. Numerical validation
studies using a 3-D hydrofoil seemed very reasonable. In
the case of supercavitation, the initial closing zone length
has no effect on the solution as long as it is inside the
supercavity bubble. The method was also shown to be
independent of the initial guess of cavity lengths and tol-
erance of cavity trailing edge openness for d
tol
_ 0:001.
The convergence rate for the case of 3-D hydrofoil
appeared to be slightly better than linear. However,
additional studies are needed to determine the effect of
Fig. 16. Comparison of the observed (top) and predicted (bot-
tom) ventilated cavity patterns for J
A
= 1:2. Propeller model 841-
B. 4 Blades. h=D = 0:33. 60 20 panels. Dh = 6

. From (Young
and Kinnas, 2002)
Fig. 17. Comparison of predicted (P)
and measured (E) blade forces for
J
A
= 1:2. Propeller model 841-B. 4
Blades. h=D = 0:33. 60 20 panels.
Dh = 6

. From (Young and Kinnas,

2002)
278
prescribed base pressure on the predicted blade forces in
the case of fully wetted and partially cavitating ow.
For surface-piercing propellers, the negative image
method is used to account for the effect of the free surface.
The method is also able to search for detachment locations
of ventilated cavities. In general, the predicted cavity
planforms and propeller loadings compared well with
experimental measurements and observations. The meth-
od also appeared to converge quickly with grid size and
time step size. Currently, the authors are in the process of
studying the added hydrodynamic forces associated with
jet sprays during the entry phase. This is accomplished by
performing a systematic 2-D study using the exact free
surface boundary conditions. Initial results of the 2-D
studies are presented in (Young, 2002; Young and Kinnas,
2002). Current efforts include:
v Determine the effect of prescribed pressure on the
separated zone behind non-zero trailing edge sections
in fully wetted and partially cavitating conditions.
v Complete the 2-D nonlinear study of surface-piercing
hydrofoils, and nd a simplied approach to incorpo-
rate the results into the 3-D model.
v Couple the hydrodynamics with a structural analysis
model to include the effect of blade vibration.
v Validate the results with experimental measurements,
and perform convergence studies with time and space
discretizations.
References
Barr RA (1970) Supercavitating and superventilated propellers.
Transactions of SNAME 78: 417450
Brandt H (1973) Modellversuche mit schiffspropellern an der
wasseroberache. Schiff und Hafen 25(5): 415422
Brillouin M (1911) Les surfaces de glissement de Helmholtz et la
resistance des uides. Annales de Chimie and de Physique 23:
145230
Choi J (2000) Vortical Inow Propeller Interaction Using Un-
steady Three-Dimensional Euler Solver. PhD thesis, Depart-
ment of Civil Engineering, The University of Texas at Austin
Cox G (1968) Supercavitating propeller theory the derivations
of induced velocity. In The 7th Symposium on Naval
Hydrodynamics, Rome
Fine N, Kinnas S (1993a) A boundary element method for the
analysis of the ow around 3-d cavitating hydrofoils. J Ship
Res. 37: 213224
Fine N, Kinnas S (1993b) The nonlinear numerical prediction of
unsteady sheet cavitation for propellers of extreme geometry.
In Sixth International Conference On Numerical Ship
Hydrodynamics, University of Iowa, Iowa, pp. 531544
Fine NE (1992) Nonlinear Analysis of Cavitating Propellers in
Nonuniform Flow. PhD thesis, Department of Ocean Engi-
neering, Massachusetts Institute of Technology
Furuya O (1985) A performance prediction theory for partially
submerged ventilated propellers. J Fluid Mech. 151: 311355
Greeley D, Kerwin J (1982) Numerical methods for propeller
design and analysis in steady ow. Trans. SNAME 90
Hsin C-Y (1990) Development and Analysis of Panel Method for
Propellers in Unsteady Flow. PhD thesis, Department of
Ocean Engineering, Massachusetts Institute of Technology
Kamiirisa H, Aoki D (1994) Development of supercavitating
propeller for outboard motors. In Second International
Symposium on Cavitation, Tokyo, Japan
Kerwin J, Kinnas S, Lee J-T, Shih W-Z (1987) A surface panel
method for the hydrodynamic analysis of ducted propellers.
Trans. SNAME 95
Kikuchi Y, Kato H, Yamaguchi H, Maeda M (1994) Study on a
supercavitating foil. In: Second International Symposium on
Cavitation, Tokyo, Japan, pp. 127132
Fig. 18. Convergence of thrust (K
T
) and torque (K
Q
) coefcients,
as well as cavity planform with time step size. Propeller model
841-B. J
A
= 1:2. 60 20 panels
Fig. 19. Convergence of thrust (K
T
) and torque (K
Q
) coefcients,
as well as cavity planform with grid size. Propeller model 841-B.
J
A
= 1:2. Dh = 6

279
Kinnas S (1992) A general theory for the coupling between
thickness and loading for wings and propellers. J. Ship Res.
36(1): 5968
Kinnas S, Choi J, Lee H, Young J (2000) Numerical cavitation
tunnel. In: NCT50, International Conference on Propeller
Cavitation, Newcastle upon Tyne, England
Kinnas S, Fine N (1991) Non-linear analysis of the ow around
partially or super-cavitating hydrofoils by a potential based
panel method. In: Boundary Integral Methods. Theory and
Applications, Proceedings of the IABEM-90 Symposium,
Rome, Italy, October Heidelberg. Springer-Verlag, pp. 289
300
Kinnas S, Fine N (1992) A nonlinear boundary element method
for the analysis of unsteady propeller sheet cavitation. In:
Nineteenth Symposium on Naval Hydrodynamics, pp. 717
737, Seoul, Korea
Kinnas S, Fine N (1993) A numerical nonlinear analysis of the
ow around two- and three-dimensional partially cavitating
hydrofoils. J. Fluid Mech. 254: 151181
Kinnas S, Hsin C-Y (1992) A boundary element method for the
analysis of the unsteady ow around extreme propeller
geometries. AIAA J. 30(3): 688696
Kinnas S, Mishima S, Brewer W (1994) Nonlinear analysis of
viscous ow around cavitating hydrofoils. In: Twentieth
Symposium on Naval Hydrodynamics, University of
California, Santa Barbara, pp. 446465
Kudo T, Kinnas S (1995) Application of vortex/source lattice
method on supercavitating propellers. In 24th American
Towing Tank Conference, College Station, TX
Kudo T, Ukon Y (1994) Calculation of supercavitating propeller
performance using a vortex-lattice method. In Second Inter-
national Symposium on Cavitation, Tokyo, Japan, pp. 403408
Lee H, Kinnas S (2001) Modeling of unsteady blade sheet and
developed tip vortex cavitation. In: CAV 2001: Fourth Inter-
national Symposium on Cavitation, Pasadena, CA. California
Institute of Technology
Lee H, Kinnas S (2002) Fully unsteady wake alignment for pro-
pellers in non-axisymmetric ows. J. Ship Res (Submitted for
Publication)
Lee HS (2002) Modeling of Developed Tip Vortex Cavitation and
Unsteady Wake Alignment. PhD thesis, Department of Civil
Engineering, The University of Texas at Austin
Lee J-T (1987) A Potential Based Panel Method for The Analysis
of Marine Propellers in Steady Flow. PhD thesis, Department
of Ocean Engineering, Massachusetts Institute of Technology
Matsuda N, Kurobe Y, Ukon Y, Kudo T (1994) Experimental
investigation into the performance of supercavitating
propellers. Papers of Ship Research Institute 31(5)
Morino L, Kuo C-C (1974) Subsonic Potential Aerodynamic for
Complex Congurations: A General Theory. AIAA J. 12(2):
191197
Mueller A, Kinnas S (1999) Propeller sheet cavitation predictions
using a panel method. J. Fluids Eng. 121: 282288
Oberembt H (1968) Zur bestimmung der instationaren u-
gelkrafte bei einem propeller mit aus dem wasser hera-
usschlagenden ugeln. Technical report, Inst. fur Schiffau der
Universitat Hamburg, Bericht Nr. 247
Olofsson N (1996) Force and Flow Characteristics of a Partially
Submerged Propeller. PhD thesis, Department of Naval Ar-
chitecture and Ocean Engineering, Chalmers University of
Technology, Goteborg, Sweden
Riabouchinsky D (1962) On some cases of two-dimensional uid
motion. In Proceedings of London Math Society, Vol. 25, pp.
185194
Russel A (1958) Aerodynamics of wakes, existence of unsteady
cavities. Eng. 186: 701702
Savineau C, Kinnas S (1995) A numerical formulation applicable
to surface piercing hydrofoils and propellers. In 24th Amer-
ican Towing Tank Conference, Texas A & M University,
College Station, TX
Shiba H (1953) Air-drawing of marine propellers. Technical
Report 9, Transportation Technical Research Institute
Tachnimdji A, Morgan W (1958) The design and estimated per-
formance of a series of supercavitating propellers. In the
Second Ofce of Naval Research Symposium on Naval
Hydrodynamics, ACR-38, pp. 489532
Tulin M (1953) Steady two-dimensional cavity ows about
slender bodies. Technical Report 834, DTMB
Tulin MB (1962) Supercavitating propellers history, operating
characteristics, mechanisms of operation. In Fourth Sympo-
sium on Naval Hydrodynamics, pp. 239286
Ukon Y, Kudo T, Kurobe Y, Matsuda N, Kato H (1995) Design of
high performance supercavitating propellers based on a
vortex lattice method. In An International Conference on
Propeller Cavitation (PROPCAV 95), The University of
Newcastle Upon Tyne, England, UK
Villat H (1914) Sur la validite des solutions de certain problem d
hydrodynamique. Journal de Mathematiques 6(10): 231290
Vorus W, Mitchell K (1994) Engineering of power boat propel-
lers. In Propellers/Shafting 94 Symposium, pp. 116 (paper
No. 12), Virginia Beach, VA. Society of Naval Architects &
Marine Engineers
Vorus WS, Chen L (1987) An extension of the malkus hypoth-
esis to the turbulent base ow of blunt sections. J. Fluid
Mech. 184: 551569
Wang D (1979) Oblique water entry and exit of a fully ventilated
foil. J. Ship Res. 23: 4354
Wang G, Jia D, Sheng Z (1992) Study on propeller characteristics
near water surface. In: The 2nd Symposium on Propeller and
Cavitation, Hangzhon, China, pp. 161168
Wang G, Zhu X, Sheng Z (1990) Hydrodynamic forces of a three-
dimensional fully ventilated foil entering water. J. Hydro-
dynamics 5(2):
Yim B (1976) Optimum propellers with cavity-drag and frictional
drag effects. J. Ship Res. 20
Young Y, Kinnas S (2001) A BEM for the prediction of unsteady
midchord face and/or back propeller cavitation. J. Fluids Eng.
123: 311319
Young Y, Kinnas S (2002) A BEM technique for the modeling of
supercavitating and surface-piercing propeller ows. 24th
Symposium on Naval Hydrodynamics
Young Y, Kinnas S (2003) Numerical modeling of supercavitating
propeller ows. J. Ship Res. 47(1): 4862
Young YL (2002) Numerical Modeling of Supercavitating and
Surface-Piercing Propellers. PhD thesis, Department of Civil
Engineering, The University of Texas at Austin
280