Sie sind auf Seite 1von 4

986

2014,26(6):986-989
DOI: 10.1016/S1001-6058(14)60109-8

Numerical simulations of wake signatures around high-speed ships*


WANG Ben-long (), GUO Xiao-yu (), LIU Hua ()
School of Naval Architecture, Ocean and Civil Engineering and MOE Key Laboratory of Hydrodynamics,
Shanghai Jiao Tong University, Shanghai 200240, China, E-mail: benlongwang@sjtu.edu.cn
HE Chen ()
Xian Precision Engineering Institute, Xian 710075, China
(Received October 11, 2014, Revised November 18, 2014)
Abstract: Navier-Stokes equations are solved to obtain the near field flow and to reveal the generation of droplets due to the
breaking of the bow wave. With an adaptive refinement of the Octree mesh, the fine scale flow characteristics, the wave breaking and
the splashing are well simulated. For high-speed ships, the V-shape ship wave pattern with the angle smaller than the classical Kelvin
angle is confirmed by the present numerical results, which can also be applied to the catamaran. Comparisons of the vertical velocity
distributions induced by the mono-hull ship and the catamaran are made at different locations in the wake regions.
Key words: bow wave, breaking wave, volume of fluid method

Introduction
A moving ship creates foamy, aerated and turbulent wakes, as usually observed as white-water by the
naked eyes. Air droplets and bubbles in these wakes
are usually induced by the breaking of bow waves, the
air entrainment into the ships turbulent hull boundary
layer at the free surface and the propellers cavitation.
Along with the development of high-speed ships, the
production and the transportation of the bubbles in the
wake are studied for the need of the signature identification by means of visualization or acoustic. Among
various sources, the breaking of the bow waves and
the Kelvin waves are the main sources of the bubble
generation. Fast ships generate overturning bow
waves that consist of detached thin sheets of water,
which are mostly steady until they hit the water and
undergo turbulent breaking up, splashing and diffusion. Smaller bubbles are created by jet and drop impacts on the wave surface.
Due to the characteristics of multiple scales, for
the droplets of breaking waves and ship waves, nume* Project supported by the National Natural Science Foundation of China (Grant Nos. 11272210, 51379123), the National Basic Research Development Program of China (973 Program, Grant No. 2014CB046200).
Biography: WANG Ben-long (1977-), Male, Ph. D.,
Associate Professor
Corresponding author: LIU Hua, E-mail: hliu@sjtu.edu.cn

rical simulations of bow waves and near field flow require vast computational resources, which is the main
difficulty for estimating the strength of the bubble
production during wave breaking. Recently, the SPH
method was successfully used to predict the ship bow
waves[1]. Systematic experimental measurements of
overturning of bow waves are not available yet[2].
In the present work, a numerical wave tank is
built, in which both bow waves and near field flow
can be calculated simultaneously at a relative low
computation cost. The numerical results from direct
numerical simulations can be used to predict the wave
pattern in the near field around the hull, as well as the
velocity distribution and the distributions of droplets,
which may provide the main signature of the ship
wake.
Navier-Stokes equations are used to describe the
motion of the fluid

uj
+
=0
t
x j

(1)

ui u j ui
p ij
+
=
+
+ gi
t
x j
xi x j

(2)

in which xi is the coordinates, ui is the velocity


components, and p are the density and the pressure, ij is the shear stress due to viscosity and turbulence. The eddy viscosity model is widely used in

987

the free surface flows with wave breaking[3], and is


adopted in the present work.

1
3

1 ui u j
+
2 x j xi

ij kk ij = 2 t Sij , Sij =

(3)

where t is the eddy viscosity, expressed as t =

(Cs )2 2 Sij Sij , Cs = 0.15 is the Smagorinsky constant and is the mesh size. To track the free surface
at multiphase flows, e.g., the air-sea interface or the
small bubbles floating in the air, the VOF method is
employed,
F
+ (u F ) = F u
t

(4)

For multiphase flows, the properties of the mixture are


evaluated by the arithmetic average value of each
phase, specifically = F 1 + (1 F ) 2 . To study the
bow waves around a ship, four kinds of boundaries are
specified: (1) Inlet boundary at upstream. The velocity
is specified. Values of the VOF function are set to be
0 and 1 at each side of the still water level. The pressure is given following the still water distribution. (2)
Outlet boundary or the far field at downstream. The
VOF function is set to be 0 and 1 at each side of the
still water level. The pressure is given following the
hydrostatic distribution. The normal gradients of the
velocity and other scalar functions are zero. In addition, a relatively coarse mesh is also used in the far
field of the ship, which can partially dissipate the ship
waves with the numerical viscosity embedded in the
flux discretization schemes. (3) The ship surface. No
slip boundary condition is assumed. The velocity
components are all zero on the ship surface. The normal gradients of the scalar functions, e.g. the pressure
or the VOF function, are zero. (4) The sea bed and the
upper boundary of the air region. Symmetry conditions are assumed. There is no material flux across the
symmetry boundary. Therefore, the gradient of the
normal velocity is zero. The tangential velocity gradient is also zero due to zero shear stress at the sea
bed and the air boundary. The normal gradient of scalar functions, e.g., the pressure or the density, are zero
as well.
A multiphase incompressible flow solver Gerris
is employed[4], which is a second-order flow solver for
both spatial and temporal discretizations. The Octree
mesh can robustly represent complex interfaces, and
efficiently treat the flows of different spatial scales,
simultaneously. Two thresholds are used in the present work to control the Octree mesh adaption. The first
one is based on the vortex magnitude, U /

max U , where is a small quantity and takes


a value of 0.01. Another adaptive strategy is the

gradient of the function F , which will refine the


mesh size of cells across the free surface. The octree
mesh can adaptively refine the flow region where a
small flow structure exists. Therefore the total computational cost can be dramatically reduced, which
makes possible a systematic study of overturning and
breaking of bow waves, as shown in the upper panel
of Fig.1. The minimum grid size min = L / 2048 , and
the total grid number is around 8M. Taking L = 1 m ,
the droplets of a few millimeter in diameter can be directly resolved, accompanied with the ship waves of
meter scales, as shown in Fig.1. The computations are
performed using 32E5-2670 Intel Xeon cores connected by IB network. The typical runtime is about
300 h, corresponding to 104 CPU hours.

Fig.1 Illustrations of mesh refinement of ship waves and floating droplets due to splashing of breaking waves. The top
panel is a zoom view of a small region near the central
line in the bottom panel. The small water droplets are
resolved by the adaptive Octree mesh, which strongly
depends on the mesh size. In the bottom figure, the
discrete droplets are shown in dark color. Fr = 1.04 ,

Re = UL / = 3 106

The validations of the present solver and the setup of the numerical wave tank are carried out for a
slow ship, e.g., with the hull Froude number Fr = U /
gL = 0.316 . The drag force and the free surface elevations are compared with experimental data available.
Good agreements are achieved. The details are not
given here due to the space limitations. In the following computations, a standard S60 hull of length L
and draft 0.056L is considered. The catamaran is

988

composed of two side-by-side S60 hulls of length


0.794L , with the same tonnage as the mono-hull ship.
The distance between the two S60 hulls centerlines is
0.1588L .
In the traditional hydrodynamics, Lord Kelvin
showed that the angle of the V-shape ship waves is
19.5o , irrespective of the speed of the ship. However,
in a real world the appearance of the near field waves
seems to be influenced by the ship speed and the nonlinear effects. More like the Mach cone in supersonic
flows, the angle of the wake is governed by the ship
speed, i.e., the faster it goes, the more its wake stretches and narrows. Rabaud and Moisy[5] analyzed
images from Google Earth, using the measurements of
the boats hull lengths and the wake angles, as well as
the calculations of their velocities, to help them build
a new mathematical model to describe the narrow
wakes as follows:

= tan 1

= tan 1

19.47o ( Fr FrC )

(5a)

2Fr 2 1
( Fr FrC )
4Fr 2 1

(5b)

where FrC = 3/ 4 0.49 . In their work, the numerical results of the linear theory were in good agreement with their image analyses. However, their observations and explanations were revisited recently by a
few scholars for a clearer physical explanation[6,7].

Fig.2 Wave patterns around the ships. Left, mono-hull ship,


corresponding hull Fr = 1.04 , tan = 0.17 , right, catamaran, L = 39.7 m , Fr = 1.17 , tan = 0.19 . The
dash line marks the maximum displacement of divergent
waves, excluding the breaking bow waves

Through the direct simulation of the fully nonlinear flow problems with Navier-Stokes equations, the
pattern of the near field waves is obtained. The amplitude of the diverging wave is found to increase with
the Froude number. The wavelength of the transverse
waves increases faster as well due to k = g / U 2 , but
the amplitude decreases rapidly and tends to vanish.
Therefore, only a V-shape pattern composed of divergent waves can be observed, whose samples are
shown in Fig.2.
In the present work, the apparent angle of the V-

shape pattern is determined from the distributions of


the local maximum surface wave displacement (usually recognized as the breaking point) inside the wake
range extending to 6L behind the ship, which is different from the original definition of the Kelvin angle.
In realistic situations, the wave breaking (e.g., as seen
from the top panel in Fig.1) also changes the positions
of the maximum surface elevation as compared with
that predicted by the linear theory. It should be noted
that even for a slow ship, the apparent angle of the Vshape pattern can be significantly smaller than the
Kelvin angle due to nonlinear effects, e.g., 10.9o was
observed for a field ship at Fr = 0.48 [8]. The knowledge of high-speed ship wave patterns may help analyzing the dark and bright stripes in the SAR image[9].

Fig.3 Comparison of near field apparent V-shape angle, dashed


line indicates the decay trend of V-shape angle for monohull ship

As can be observed in Fig.3, the asymptotic formulation (5) can reasonably be used to predict the apparent angle of the V-shape wave pattern in the near
field with different Froude numbers. However, Eq.(5)
predicts the near field wave pattern of the catamaran
with a fair accuracy. Based on the interference analysis, Noblesse et al.[10] concluded that the wake angle
decreases as 1/ Fr as Fr increases for a catamaran,
but decreases as 1/ Fr 2 for a mono-hull ship. In the
present numerical results, the decay of the wake angle
for a mono-hull ship appears faster than that of a catamaran when Fr 1 .
In a certain circumstance, the fraction of air bubbles in the water is needed, which is generally predicted by the Boltzmann type population balance equation and sub-grid air entrainment models. The generation and the transport of the bubble flow in the ship
wake region are caused by the strong nonlinear flows
with wave overturning and breaking. The development of the sub-grid air entrainment models depends heavily on the experiments and high-accuracy CFD, specifically the entrained air flux due to the wave breaking is related to the downward velocity[11] or the surface normal velocity gradient[12].
Next, we study the velocity distribution around
the high-speed ships using the Navier-Stokes solver
associated with a good resolution of bow waves and
breaking. The vertical velocities are plotted, which

989

zones.
So far, the main flow structures, e.g., the velocity
distribution and the apparent ship wave pattern, can be
well captured using the adaptive mesh refinement,
nearly independent of the geometry scales in numerical simulations. However, it is still a long way to go to
fully capture the wave breaking process and the motion of droplets in a realistic length scale. The bow
waves and the near field flow may be significantly affected by the ship profile, which is beyond the scope
of the present work. Massive parallel computations
should be employed to study these practical influences
on the local velocity field in the further work.
Acknowledgement
We appreciate the Center of High Performance
Computing in Shanghai Jiao Tong University for providing the HPC resources and technical assistance.
References
[1]

[2]
[3]

[4]

[5]
[6]
[7]
Fig.4 Distributions of vertical velocity at different longitudinal
sections behind the ship. From top to bottom: x = 3L,
6L . Two typical contours are plotted for w = 0.02 m/s
and 0.1 m/s by taking L = 52 m and with Froude
number similarity

can be used to predict the air entrapment region[11,12].


The corresponding velocity distribution is shown in
Fig.4. To give a clearer physical picture of practical
applications, we present the velocity field in a dimensional scale. Comparing the vertical velocities induced
by the mon-hull and the catamaran of the same tonnage, there is some difference at x = 3L . However, little difference can be observed when X 6 L at these
specified Froude numbers. For even larger Froude
number, the apparent angle of the V-shape pattern
may affect the distribution of downward velocity

[8]

[9]
[10]

[11]

[12]

MARRONE S., BOUSCASSE B. and COLAGROSSI


A. et al. Study of ship wave breaking patterns using 3D
parallel SPH simulations[J]. Computers and Fluids,
2012, 69: 54-66.
NOBLESSE Francis, DELHOMMEAU Gerard and LIU
Hua et al. Ship bow waves[J]. Journal of Hydrodynamics, 2013, 25(4): 491-501.
GUO X., WANG B. and LIU H. et al. Numerical simulation of two-dimensional regular wave overtopping
flows over the crest of trapezoidal smooth impermeable sea dike[J]. Journal of Waterway, Port, Coastal,
and Ocean Engineering, 2014, 140(3): 04014006.
FUSTER D., AGBAGLAH G. and JOSSERAND C. et
al. Numerical simulation of droplets, bubbles and waves:
State of the art[J]. Fluid Dynamics Research, 2009,
41(6): 065001.
RABAUD M., MOISY F. Ship wakes: Kelvin or Mach
angle?[J]. Physical Review Letters, 2013, 110(21):
214503.
DARMON A., BENZAQUEN M. and RAPHAEL E.
Kelvin wake pattern at large Froude numbers[J]. Journal of Fluid Mechanics, 2014, 738: R3
DIAS F. Ship waves and Kelvin[J]. Journal of Fluid
Mechanics, 2014, 746: 1-4.
BROWN E. D., BUCHSBAUM S. B. and HALL R. E.
et al. Observations of a nonlinear solitary wave packet
in the Kelvin wake of a ship[J]. Journal of Fluid Mechanics, 1989, 204: 263-293.
REED A. M., MILGRAM J. H. Ship wakes and their
radar images[J]. Annual Review of Fluid Mechanics,
2002, 34: 469-502.
NOBLESS F., HE J. and ZHU Y. et al. Why can ship
wakes appear narrower than Kelvin's angle?[J]. European Journal of Mechanics B/ Fluids, 2014, 46: 164171.
MORAGA F. J., CARRICA P. M. and DREW D. A. et
al. A sub-grid air entrainment model for breaking bow
waves and naval surface ships[J]. Computers and Fluids, 2008, 37: 281-289.
MA J., OBERAI A. A. and LAHEY R. T. et al. Modeling air entrainment and transport in a hydraulic jump
using two-fluid RANS and DES turbulence models[J].
Heat Mass Transfer, 2011, 47(8): 911-919.

Das könnte Ihnen auch gefallen