Escolar Documentos
Profissional Documentos
Cultura Documentos
113-119
This study demonstrates that a bounded, physically relevant solution does exist at the so-called 7-
= UW/K = 1/4 resonance in the linear seakeeping problem for a realistic ship with forward speed,
U, frequency of encounter, ~J, and gravitational acceleration, g. The solution of the seakeeping
problem by a linear, three dimensional, time-domain Rankine panel method, validated through
numerical analysis, testing, and comparison to physical experiments, supports this claim. The
solution can also be obtained with equal validity through frequencies both above and below the
critical frequency.
ship's mean velocity, 14~ = (U, V, 0), assumed to be steady in where O ( i ' ; ~) is the Rankine source potential,
time, and parallel to the mean free surface.
Within the fluid, under the assumption of inviscid, irrota- G(~'; Z) -
tional flow, the Laplace equation governs the total disturbance J ~ -1i ' l (7)
velocity potential. The gradients of the potential are assumed
to decay to zero infinitely far from the ship, and a radiation so that both the free surface, Sp, and hull surface, SB, must
condition is imposed to prevent the return of body generated be explicitly considered.
waves. As the discrete formulation is detailed in Nakos, Kring, ~z
Decomposition of the total disturbance potential, kO, Sclavounos (1993), the spatial and temporal discretizations are
only summarized here. The unknowns on the boundaries, po-
kO(Z, t) = (I)(Z) + ~(Z, t) (i) tential, ~ ( i , t), normal derivative of the potential, ~n (i, t), and
into a basis flow component, ~, and a wave flow component, ~, wave elevation, ~(x, y, t), are related through the boundary con-
assumes that the basis flow provides the dominant contribution ditions which must be discretized in space and time. The free
to the total flow. The basis flow used in this study is the double- surface conditions, treated as two independent evolution equa-
body flow, which ensures that (I)z = 0 on z = 0. A Taylor tions, along with the body boundary condition provide the forc-
expansion from the instantaneous position of the free surface ing to equation (6) which becomes a mixed boundary value
to the mean position of the free surface at z = 0 leads to the problem. A linear system of equations results from this bound-
linearized kinematic and dynamic free surface conditions, ary integral after approximating the unknowns as a set of hi-
quadratic spline sheets over the domain boundaries.
02,:I) 0~
O<Ot (l~" - V(I)) V~ = ~-~-z2~ + ~ on z = 0 (2) The relationship between the spatial discretization, a bi-
quadratic spline sheet, and the temporal discretization of the
(o_(~_~0).v)(~)___g~+E~.V0_~v0.v01 free surface conditions was crucial for numerical stability and
accuracy. The choice for these was dictated by a rational error
on z = 0 (3) analysis of waves propagating over a discrete free surface by
Vada & Nakos (1992). This led to two time evolution schemes
where, ~(x, y, t) is the Eulerian description of the wave eleva-
that exhibit the very favorable properties of no error in wave
tion.
dissipation and error in wave dispersion no more than third or-
The body boundary condition is linearized by applying a
der in the characteristic panel length. The first scheme, a com-
Taylor expansion about the mean position of the ship. The no-
bination of explicit and implicit Euler integrations, is numeri-
flux condition results,
cally efficient for low speeds, less than about Froude number,
Fn = U/V/'(gL ) = 0.4, where L is the ship waterline length.
O-~(~+#)=~7"E+L(~tnJ+(jrn3) onSB (4) The other scheme found to be neutrally stable is a combination
j=l of Leapfrog and Trapezoidal integrations that is numerically ef-
ficient for high speeds.
with
The seakeeping simulation leads to a solution for the local
( - ~ l , m 2 , m 3 ) = (ft. V ) ( W - V ~ ) pressures on the hull and the wave pattern across the free sur-
face. These pressures are integrated along the hull to produce
(m4, ms, m6) = (~. v ) ( ~ x ( ~ - v ~ ) ) (5)
a time history of the forces acting on the ship which can be
and n 3 a generalized normal. The basis flow accounts for the Fourier transformed for comparison with experimental results
steady component and the wave flow accounts for the unsteady as shown in the next section. Figure 1 illustrates the wave pat-
components. If the ship displacements, ( j , and velocities, ~-t, tern after the force signal has achieved steady state. In this
were to be determined from the equations of free ship motion, a case, a Lewis form ship was forced to heave at a period corre-
further decomposition of the wave flow into instantaneous and sponding to the critical frequency, 7- = 1, at Froude number,
memory components would be necessary for numerical stability F n = 0.1. When animated and viewed in a frame of reference
as demonstrated in Kring & Sclavounos (1995). In this work, fixed to the ship, this wave pattern appears as a skewed con-
the ship displacement and velocity will be imposed as part of centric ring of standing waves with a secondary wave system
the forced motion radiation problem (Newman 1978). The na- traveling downstream. There is a compression of the ring sys-
ture of the 7 ---- 1 singularity will not directly effect the numer- tem upstream with a corresponding dilation downstream. In
ical stability properties of the equations of motion, if the wave the figure, the computational mesh is also presented to help
solution is bounded. illustrate the numerical method. A round grid was chosen since
Although the double body linearization is used here rather waves can be expected to radiate in all directions from the ship
than the Neumann-Kelvin formulation which has the free below T ---- 1.
A second wave system traveling d o w n s t r e a m from the ship to avoid overdamping the evanescant modes in the free surface
can not be easily discerned since the figure is not animated, near the ship's waterline. This periodic application of the spa-
but the simulation t h a t produced this wave p a t t e r n did achieve tial filter has proven sufficient to control the error in the small
steady state as shown in Fig. 2. In this case there was positive wavelengths as reported in Nakos, Kring, & Sclavounos (1993).
damping. T h e d a m p i n g must result from the physical propaga- The forces acting on the b o d y can be shown to be insensitive
tion of waves d o w n s t r e a m and not from numerical dissipation to this filter for a wave p a t t e r n t h a t has numerically converged.
as confirmed by t h e numerical testing in this paper. T h e second numerical issue involves the modeling of the open
It has been r e p o r t e d in Grue & P a h n (1985) and N e w m a n boundary. In order to do this an efficient and practical numer-
(1961) t h a t negative damping, or incoming energy flux, can ical beach is applied as in Nakos et al (1993). This numerical
exist at forward speed for a two-dimensional, submerged, oscil- beach dissipates outgoing waves w i t h o u t reflection by modify-
lating ellipsoid. This case has not been encountered with the ing the kinematic condition in an outer ring of the c o m p u t a -
present m e t h o d for a three-dimensional, surface piercing ship tional domain. The modified kinematic condition in the zone
but the formulation does allow for negative damping. In the of the numerical beach appears as,
case of forced or prescribed b o d y motion, negative d a m p i n g
indicates t h a t the b o d y is extracting energy from the waves. ~ 02a5 0~o ~,2
T h e t i m e history of the radiation force will simply exhibit a -(W-VE~).V4= ~Zz2 ~ + ~ - z - 2 , ~ + - - ~ onz=0 (8)
negative phase shift. For free motion, however, where the equa- 9
tions of motion are being solved in the time-domain, negative where, u, is a damping parameter. These added terms act like
d a m p i n g may lead to an u n b o u n d e d ship motion, unless some Newtonian cooling and Rayleigh viscosity to dissipate the waves
physical model such as nonlinear viscous d a m p i n g or energy while avoiding artificial dispersion. This m e t h o d was a d a p t e d
e x t r a c t i o n system is included. T h r o u g h o u t this paper, only the from the work of Israeli & Orszag (1981) and Cointe (1989).
forced m o t i o n problem is considered. If a sufficiently large t r u n c a t i o n distance is used in the com-
T h e validity of this solution will be illustrated in the next putational mesh, there will be no influence on the solution as
section, but several numerical issues critical to the m e t h o d must d e m o n s t r a t e d by the d o m a i n sensitivity study presented in the
first be introduced. These issues are as follow: next section.
Small wavelengths and the resolution of the grid scale,
Long wavelengths and the modeling of the open boundary, ...... t~t (g/L)'~= 0.04
T e m p o r a l and spatial convergence and the accuracy of the 10 ........ At (g/L)~= 0.02
discrete model. _ _ At (g/L)'~= 0.01
FJpgV
T h e first numerical issue has been examined by the error
analysis of waves propagating over a discrete free surface. As 5
t i m e step and panel size become infinitesimal, the m e t h o d is
consistent and reduces to the continuous formulation, but for
practical discretization sizes, the presence of waves shorter t h a n
a few panel lengths can introduce some error.
T h e stability analysis has shown t h a t the waves are neutrally
stable in time, but the numerical dispersion relation reveals
-5
some error in the group velocity of very short waves. In order
to correct this numerical problem a low-pass spatial filter is . . . . i . . . . i . . . .
applied at periodic time steps in order to prevent the spurious 0 20 40 6'0 8'0 1C)O
t (g/L) v~
waves from c o n t a m i n a t i n g the solution. T h e filter has been de-
signed to m a t c h the properties of the error uncovered by the Fig. 2 H i s t o r y o f vertical force f o r a Lewis hull in heave (T = 1 / 4 , Fn
numerical analysis. It is not applied at every t i m e step in order = 0.1) f o r three t i m e - s t e p sizes
SWAN
2.0 - A3.~pV ........ r/L= 6. 2.0 - B ~ / p V ( g / L ) '~
......... r / L = 9.
r / L = 12.
",, m Exp. (Maruo)
1.5 1.5
1.0 1.0
Q//
0.5 0.5
, , i . . . . I . . . . i . . . . ~ . . . . i , , , , I . . . . i . . . . i . . . . I . . . . ] , .
1.0 1.5 2.0 2.5 3.0 1.0 1.5 2.0 2.5 3.0
o~(L/g)
Fig. 3 A d d e d mass a n d d a m p i n g coefficients for a Lewis hull in heave at zero speed for three domain sizes
SWAN
A33/p v ......... r/L= 5.
2.5 ! ,, ......... r/1_=9.
'~ r/L= 12. i B~pV(g/L) '~
D Exp. (Maruo)
2.0 =.
2.0
1,5
1.5
1.0
O
0,5
1.0
0.0 I , , i , I . . . . L . . . . m . . i . . . . q . . . . I . . . . i , ,
Fig. 4 Added mass and d a m p i n g coefficients for a Lewis hull in heave at Fn = 0.1 and Fn = 0.2 for
three d o m a i n sizes
SWAN Ass/PVL2
AJpV ........ 2~ ~ s
2.5 ~ ................. 3700pImels - 0.15
\ - - 4290 panels
Exp. (Maruo)
2.0
0.10
..j
1.5 Fn=0.2
\ W
1.0 :
1 0.05
0.5 ~
0.0 t . . . . I . . . . I . . . . I . . ; 0 . 0 0 I i i i i I i i i i I i i i i I i
BJpV(g/L) 1~ Bss/pV(gL)1~
,,',~ 0.10
2.0 Fn=0.2
!/':'~Fn=0.2
1.S / ~ ~ ~ " ~ m 0.05
o I
1.0
i .... i .... i .... i , , 0.00
0.15 0.25 0.35 0.45 1.15 0.25 0.35 0.45
~= UoYg ~= Um/g
Fig. 5 Added mass and d a m p i n g coefficients for a Lewis hull in heave and pitch at Fn = 0.1 and Fn = 0.2 for
three grid densities
0.3.5 0
SWAN f
AsJPVL .
.......
. . . 2~opan=,
3700
0.3 42g0
panels
panels
-0.05
Fn= 0.1
0.25 -0.1
0.2 -0.15
0.15 -0.2
0.1 -0.25
0.1
Ss3/PVL(g/L)~
0
-0.1
-0.2
0.4
r
A Fn=0.2
-0.3
.2F F:0.x \
-0.4
-0.5
I . . . . I , , , , I . . . . i ,