|
Items in cart [1] |
Questions? Call 888-624-8373 |
| Copyright © 2009. National Academy of Sciences. All rights reserved. Terms of Use and Privacy Statement |
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 145
24~ Symposium on Naval Hydrodynamics
Fukuoka, JAPAN, 8-13 July 2002
The Use of a RANS Code in the Design and Analysis of a
Naval Combatant
Joseph J. Gorski, Henry J. Haussling, Allen S. Percival, James I. Shaughnessy, and
Gregory M. Buley (Naval Surface Warfare Center, Carderock Division)
ABSTRACT
with the entire
configuration ~ncluo~ng the propulsor and the scoop
preceding it. The overall conclusions from the
calculations are that the modifications to the bow dome
are an improvement over the original geometry.
Additionally, there is no significant separation around
the propulsor or in the scoop area leading into the
propulsor. However, because the propulsor operates
close to the hull the inflow to the propulsor is
significantly less than the free stream velocity.
Comparisons with experimental data for the final
design are also performed for validation purposes
demonstrating the accuracy of the RANS predictions.
INTRODUCTION
The computer advances in the last decade
have been a significant catalyst for improving our
ability to predict ships flows. The increased computer
power has led to Reynolds Averaged Navier-Stokes
(RANS) calculations on large grids becoming more
routine. It has also led to an increased experience base
as it became possible to do more calculations. Good
results for a variety of complicated ship flow fields
have been obtained with a number of RANS codes
(Gorski, 2001b) demonstrating the overall maturing of
the capability. Such computations provide an overall
picture of the flow field in the immediate vicinity of
the ship, which can aid in its understanding as well as a
means to do trade off studies to evaluate design
Viscous flow calculations using the Reynolds options.
Averaged Navier-Stokes (RANS) equations are
performed in support of the design effort for the Gov't-
2 geometry of the ONR Surface Combatant
Accelerated Hydrodynamics S&T Initiative. The
Gov't-2 configuration is a tumblehome geometry with
surface piercing bow and ducted propulsors. RANS
calculations are used to provide information on the
flow created by various bow dome modifications as
well as to provide information on the inflow to the
propulsor. The bow dome changes are evaluated using
bare body calculations. For the propulsor inflow,
calculations are performed
Because of their demonstrated successes,
ITTC (1996) concluded that RANS codes had matured
enough to be integrated into the design process for
addressing issues associated with resistance and
propulsion. Rood (2000) discusses how such codes are
starting to revolutionize ship hydrodynamics design
and evaluation procedures from traditional towing tank
methods to computational based methods, particularly
in support of the Navy's new land attack destroyer.
RANS codes can be used to predict differences in
resistance due to shape and Reynolds number changes
as shown by Gorski (1998) for bodies of revolution.
Additionally, RAN S codes can contribute in a marine
design process, as demonstrated by Gorski and
Coleman (2002) for a submarine sail where design
decisions were based on the predicted resistance and
flow field. For surface ships though, the free-surface
predictions with RANS codes have been more
problematic (Gorski, 2001b). Although some very
good free-surface predictions have been obtained with
various codes, and even sinkage and trim predictions
have been performed (Subrami et al, 2000), the lack of
a robust free-surface capability as part of the RANS
prediction process prevents fast and accurate resistance
predictions for a variety of speeds. It is expected that
this will be overcome. In the meantime RANS codes
can still have a significant role in providing propulsor
inflow. Depending on the particular design of interest
highly viscous and vertical flow can enter a surface
ship propeller (Gorski, 2001a). This is especially true
for integrated propulsor/hull configurations where the
propulsor may be tucked in tight to the hull. It has
already been demonstrated that RANS codes have a
role to fill in this area for tanker (Valkhof et al 1998)
and waterjet (Allison et al 2001) designs and can also
provide information on shaft effects on the flow
entering the propeller as shown for an aircraft carrier
by Gorski et al (20021.
OCR for page 146
To demonstrate the utility of RANS codes for
future Naval combatants, as well as facilitate their use,
ONR initiated the Surface Combatant Accelerated
Hydrodynamics S&T Initiative. The objective of this
effort is to apply verified, validated, and benchmarked
computational ship hydrodynamics to the exploration
of innovative propulsor/hull concepts and to develop
tools and concepts for technology options for DD(X)
and beyond. The desire is to use these computational
tools to help evaluate 'out of the box' designs so a
large experimental program is not needed. Another
purpose is to provide a careful validation against the
experimental data being generated in order to
demonstrate the extent to which the computational
codes can reproduce the experimental and actual flow
physics. To adequately test the current computational
tools a complex surface ship geometry with strong
propulsor/hull interaction is desired. To this end a hull
form known as Gov't-2 has been designed and built,
and a variety of computations and experiments have
been performed on it. The Gov't-2 hull form is a
tumblehome design with a wave piercing bow and
ducted propulsors similar in concept to that shown in
Figure 1. Since no traditional model exists for this hull
form it needed to be designed, built, and tested as part
of the effort. This paper discusses how the use of
RANS calculations contributed in the actual design
cycle where details of the hull shape were changed
based on the predicted results. Comparisons between
calculations and measurements for the final built
geometry are also shown demonstrating how well the
computations reproduced the flow field.
Figure 1: Integrated propulsor/hull concept.
GOV'T-2 GEOMETRY MODIFICATIONS
~ .
As stated previously, the design of the Gov't-2
geometry is based on a tumblehome surface ship
configuration with an integrated hull/propulsor
2
combination where the propulsor is a ducted unit
merged directly into the hull. Because of the
uniqueness of this geometry there is no historical
database to aid in the design of this new configuration.
Consequently, to aid in evaluating the modifications a
significant number of computations are performed, on
various modifications to the basic hull form, using
several flow solvers including the RANS code UNCLE
(Taylor et al, 1991, 1995) and the potential flow codes
Das Boot (Wyatt, 2000) and SWAN (Sclavounos,
1995~. The tools used to create and modify the
surfaces are FastShip from Proteus Engineering and
Mechanical Desktop from AutoDesk, Inc.
Discussed in this paper are five hull forms, the
final Gov't-2 design and four variants, Concept-2
through Concept-5, which are evaluated with RANS to
help obtain the final Gov't-2 design. The ducted
configurations discussed, Concept-3 and Gov't-2, are
somewhat similar to the hull form shown in Figure 1.
The configurations have sonar domes, but no other
appendages are included in the calculations. Concept-3
has a simplified docking skeg which ends just forward
of the ducted propulsor. Additionally, the hull has
been 'scooped' out to provide a smooth hull surface
merging into the propulsor ducts, which are embedded
in the hull. Concept-3 is the baseline geometry from
which the RANS effort was started and Concept-2 is a
bare hull version of it. Concept-4 and Concept-5 are
attempts to improve the design over Concept-3 and
differ from it in several ways. The forebody has been
extended forward of the shoulder and the sonar dome
has been refaired to reduce the strength of the vertical
flow generated there. For the final Gov't-2 design the
skeg has been eliminated and the propulsors have been
revised and moved forward, inboard, and further into
the hull to reduce the propulsors influence on the free
surface. Concept-5 is a bare hull version of the final
Gov't-2 design.
One goal with the Gov't-2 design effort is to
keep wave drag low as well as reduce the possibility of
propeller ventilation. Potential flow solvers can provide
this type of information quite well except where
viscous effects may be important, such as in the
immediate stern region of the boat. Consequently,
most of the design effort for Gov't-2 involves the use
of the potential dow solvers because of their overall
efficiency compared to RANS codes. During the
course of the design investigation the potential flow
solvers indicated that there could be deep wave troughs
in the immediate vicinity of the propulsors. As a
consequence, a considerable effort was undertaken
with the potential flow codes to evaluate different
positions of the propulsors in order to determine ways
to improve the free-surface topography. RANS codes
could not be used to provide information on the
OCR for page 147
movement of the propulsors in a timely manner
because of the complications of resurfacing and
regrinding the geometry every time the propulsors are
repositioned.
Although much of this design effort involved
the use of the potential flow solvers the RANS codes
do have a part in the design process of such geometries
and will continue to do so. Because the propulsor is
now close to the hull, it is operating in a viscous
boundary layer, which is unlike many surface ships
where the propeller often operates in a nearly inviscid
flow. With the viscous inflow, the flow entering the
propulsor is reduced from what it normally would be
when operating outside of the boundary layer.
Additionally, any vortices or secondary flow generated
by the body upstream of the propulsors have the
potential to enter the propulsor and affect its
performance. Propulsor designers need to know the
correct flow entering a propulsor to provide a good
design. For such designs it is necessary to perform
viscous flow calculations to have any hope of
computing the flow field into the propeller accurately.
For the purposes of evaluating the design a
RANS code is used to compute double hull (no free-
surface) or linearized free surface solutions. Discussed
here are four main computations performed with
RANS during the design process along with the final
design prediction. The final design calculations are
performed with non-linear free surface modeling. One
of these design calculations is done with the hull
appended with propulsors to provide information on
how the various components interact. The remaining
three design calculations are performed on bare hull
configurations to evaluate the hullforms themselves
and to isolate the effects of the duct on the flow field.
By comparing bare hull and appended runs, an
assessment can be made of the propulsor's impact on
the flow into the scoop/duct and on the nominal wake.
The bare hull calculations provide comparisons of the
nominal wake and what impact the different bow dome
shapes have on the flow into the propulsor. The final
Gov't-2 geometry calculations are then compared with
the experimentally measured data for the free surface
along the hull, wave height behind the hull, and inflow
to the propulsor.
GRID GENERATION
It should be recognized that to perform flow
calculations for such complicated geometries is not
simply a matter of turning on a particular piece of
software. The computation of such flow fields
involves a process not unlike that of doing a model
experiment and includes: generating the geometry,
3
generating surface and volume grids, carrying out the
flow calculation and data reduction. A test of whether
a particular code can predict certain measured physics
is dependent on all pieces of this process and the grid
generation is a significant component of this process.
A prerequisite to generating computational
grids is the satisfactory specification of the actual
geometry. A day spent re-modeling and cleaning up
the geometry definition with the CAD software can
often save a week in grid generation. Details, such as
ensuring there are no gaps in the geometry and
trimming surfaces, must be taken care of before the
geometry can be used easily with current grid
generation software. An experienced grid generator
will also consider the topology of the future grid in
modeling the geometry and thereby simplify the grid
generation (i.e. combine surfaces where applicable and
eliminate unnecessary features). For the actual
definition of the geometry, a single B-spline surface for
each component is preferred. B-splines can model the
most complex shapes and provide smooth, continuous
definition with well-behaved intersections. In the
ICES format, they can be transferred between most
CAD and grid generation software packages.
When generating the computational grid, a
surface grid must first be generated on the body and all
surrounding boundaries where boundary conditions are
imposed. Then a volume grid is generated providing
discrete points in the entire flow domain where the
Navier-Stokes equations are solved. For bare hulls this
is quite straightforward. However, for appended hulls
and structured grids this can be a difficult dilemma. In
practice it is often difficult to achieve good grid
quality, a sensible amount of time spent, and a practical
grid size all at the same time. Experience is very
important because trade offs between the three areas
must often be made, particularly in a design process.
When generating the surface grid it is
important to ensure the grid conforms to the actual
geometry. Additionally, to help provide accurate
predictions, these surface grids must be clustered in
areas of high geometry gradients or where the flow is
expected to change rapidly. A computed solution can
only be as good as the grid on which it is computed If
there are high gradients in the flow it is necessary to
have enough grid points in these areas to resolve them.
If enough grid points are not present, the computation
will diffuse these high gradients. Once a flow feature
is diffused in this, or any other way, its impact and
interaction on the surrounding and downstream flow
cannot be predicted accurately.
All grids are generated with Gridgen V.13,
from Pointwise. A conventional topology is used for
all grids: 'O' transversely and 'C' longitudinally. Only
OCR for page 148
half of the flow field is calculated as port/starboard
symmetry is assumed. The far field boundary is one
body length away upstream and to the sides. The wake
extends 1.75 body lengths downstream. The minimum
spacing (initial spacing normal to the body surface) is
calculated based on a y+ value of 1, using model scale
Reynolds numbers. The grids are broken up into 24-28
approximately equally sized blocks to properly load
balance the calculations among the 24-28 processors
that are used for the calculations.
Bare Hull Configurations
The grid generation strategy for the three bare
hull grids, Concepts-2, -4, and -5, is the same in order
to provide a fair evaluation of the different hull forms
and not bias these conclusions because of differences
in results due to grids rather than differences in
geometries. No changes are needed to the inputs or
boundary conditions of the flow solver. These grids all
have the same grid dimensions and grid size with
approximately 2 million grid points total. In the
streamwise direction 257 points are used, 193 of which
are on the hull and the remainder are in the wake. In
the transverse direction 97 points wrap around the hull
and 81 points extend out from the hull in the normal
direction. Figure 2 shows the grid near the bow and
stern for Concept-2.
Hulls Appended with Propulsor
The grid for Concept-3 is similar to that for
Concept-2, but much more complicated because of the
propulsors. The same overall topology is used: 'O'
transversely, 'C' longitudinally with a singular pole
coming off the transom. Approximately 2.5 million
points are used, with the additional half million points
primarily in the propulsor region. In order to maintain
the boundary layer into and through the duct, a new
topology is incorporated into the propulsor region (see
Figures 3 and 4~. A half cylinder (180 degree) shaped
grid is embedded in the bare hull grid. It conforms to
the lower half of the inside of the duct and collapses to
a transverse line upstream and downstream. The grid
between the half cylinder and the hull conforms to the
upper half of the duct, thereby maintaining the
necessary grid resolution to carry the detailed boundary
layer flow into and through the duct. While it is time
consuming, this topology is straightforward and has
good grid quality. One problem with the current grid is
that good boundary layer resolution is not maintained
on the duct walls. This was necessary to meet design
deadlines and was felt to be a reasonable trade off as
this should not impact the computed flow entering the
propulsor significantly.
l
.
i
:
Figure 2: Concept-2 grid (every other point shown for
clarity).
4
.
Figure 3: Grid near the propulsor for Concept-3.
OCR for page 149
Number of points in k (normal) direction
Blue : 73
Green: 73
Number of points in j (girth) direction
65 circumferentia
33 radially
C
Figure 4: Grid in a transverse cut through the
propulsor of Gov't-2.
In order to do a grid resolution study a finer
grid is generated for the Gov't-2 appended hull with
approximately 9.5 million points for half the geometry.
A cross section of this grid in the propulsor region is
shown in Figure 4 with the corresponding number of
grid points. A coarser grid is obtained by removing
every other point in all three directions providing a
resolution of approximately 1.25 million points which
is somewhat coarser than that used for Concept-3. A
third coarser grid could be obtained by repeating this
procedure yielding a grid of approximately 175
thousand points.
Transom Stern Issues
The transom stern is still a major problem
when planning a grid generation strategy. Before
generating the grids, the free-surface boundary
conditionals) which will be used has to be decided.
Concepts 2, 3, 4, and 5 are run with the double-hull or
linear free surface boundary conditions. The transom
topology used is a continuation of the hull surface grid
collapsed to a singular pole at y=z=0. This is simple,
easy to generate, and has good grid quality. The non-
linear free surface boundary condition has a grid
requirement that the double-hull and linear free surface
boundary conditions do not. The grid conforms to the
changing free surface elevation by moving along the
grid's own lines of increasing/decreasing index in the
transverse direction. The pole topology cannot be used
for the non-linear free surface grids because the
singular pole does not allow the grid to move. To
alleviate this the Concept-5 bare hull grid is modified
to include a cap block extruding off the transom to the
downstream boundary. This topology poses some
problems because of the jump in grid resolution at the
boundary between the cap block and the rest of the
grid. Consequently, for the final Gov't-2 appended
hull grid a different topology is used where the hull
surface grid ends in a small segment of the hull
centerline, starting at the free surface and moving a
small distance down and forward along the hull. This
surface grid can then continue downstream, sharing the
symmetry plane with the free surface grid plane. This
topology does distort the grid somewhat, but the grid
quality is acceptable and it is not a complex topology.
The transom stem is also a problem from a
boundary standpoint when using the double-hull and
linear free surface boundary conditions. For Concepts
2, 3, and 4, the free-surface boundary is modeled by
assuming the dynamic free-surface elevation. For
Concept 5 the free-surface boundary is modeled at the
undisturbed water level. Due to the sharp transom
edge the area immediately aft of the transom is very
sensitive and a better solution can probably be obtained
by estimating the dynamic free-surface level.
FLOW SOLVER
To compute the viscous flow field the
incompressible Reynolds Averaged Navier-Stokes
equations are solved using the Mississippi State
University code UNCLE (Taylor et al, 1991, 19951.
The equations are solved using the pseudo-
compressibility approach of Chorin (1967) where an
artificial time term is added to the continuity equation
and all of the equations are marched in this artificial
time to convergence. Only steady state computations
are performed for this effort. For the present
calculations a thir~order upwind biased discretization,
based on the MUSCL approach of Van Leer et al
(1987), is used for the convective terms. The equations
are solved implicitly using a discretized Newton-
relaxation method (Whitfield and Taylor, 1991) with
multigrid techniques implemented for faster
convergence (Sheng et al, 1995~. The turbulence
model used for the present calculations is primarily the
q-ce model of Coakley (1983) for evaluating the hull
modifications, but the k-e model of Liou and Shih
(1996) has been used for the final Gov't-2 predictions
that are compared with the experimental data.
The free surface implementation uses a
surface tracking approach where the water surface is
treated like a material boundary and a kinematic
s
OCR for page 150
condition is satisfied there. This boundary then
becomes a boundary of the domain and the flow field is _
solved for the water portion of the problem with a
dynamic boundary condition applied at the water
surface for the Navier-Stokes equations. Usually only
inviscid boundary conditions are applied at the water ~l.
surface despite the use of a viscous flow solver, but this
should have little influence on the large scale waves.
Because the water surface is now a boundary to the
domain a grid must be generated in the domain using
the hull and water surface as its boundaries. Many
times one can simply use linearized boundary
conditions about the design waterline and get very
good results for practical purposes. In this way only a
single grid is needed for the problem. If one needs
more accurate results a nonlinear formulation is used
that iteratively redefines the water surface, based on the
calculation, and regenerates the volume grid based on
this new surface. Details of the implementation can be
found in Beddhu et al (2000~.
An important factor in being able to compute
and evaluate all the hull modifications of interest is the
implementation of a parallel version of the UNCLE
code (Pankajakshan et al, 20001. The code uses MPI
for message passing due to its portability. To run in
parallel the computational grid is decomposed into
various blocks, which are sent to different processors.
Load balancing is obtained by making the blocks as
equally sized as possible. For the present calculations,
24-28 processors are typically used.
FLOW SOLUTIONS
Three aspects of the flow solutions are
discussed. The first involves the bare hull calculations,
which provide information on the different bow dome
geometries. The second deals with the calculation of
the ducted hull configuration, Concept-3, which
provides information on the inflow to the propulsor.
Both of these efforts are important parts of the design
process using the RANS codes. The third part
discusses the calculation of the final Gov't-2 geometry
and comparisons with the experimental data.
Bare Hull Configurations
Concept-2 is a bare hull version of the
Concept-3 configuration. Although it is called a 'bare'
hull it contains the docking skeg as well as the scoop
leading into the propulsor. Calculations are performed
using the double model approximation with
port/starboard symmetry at a Reynolds number of 13.7
million, based on body length. The computed axial
velocity contours at several longitudinal locations
along the hull are shown in Figure 5.
6
-
-
-
."
T- ~
~—
_:
~ _
Figure 5: Computed axial velocity contours for
Concept-2.
As seen at the forward location, vortices are formed
over the bow dome. These vortices spread and move
outboard as they travel downstream. These vortices
produce secondary flow outboard along the hull as seen
in the cross-sectional plot of Figure 6, which shows
one of the axial velocity contour plots with the
secondary flow vectors superimposed. Here the
port/starboard symmetry plane is at Z=0. These
vortices rotate in the same direction as would a forward
bilge vortex generated by the hull. Consequently, it is
not clear if the vortex system traveling down the length
of the hull is entirely from the bow dome or is
enhanced by the natural curvature of the hull. In any
case, the vortex system moves downstream along the
path of the scoops which influence the cross sectional
extent of this vertical wake structure. As this vortex
structure is accentuated in the scoops, it is seen that
low velocity flow is pulled out from the boundary layer
and the overall size of the vortex structure appears to
increase. Here, green is low axial velocity and red is
close to free stream velocity. Additionally, the
boundary layer and wake created by the docking skeg
is seen in the computation, but does not appear to
interact strongly with the flow in the scoops. Although
there is low axial velocity flow in the scoops it does not
appear to lead to flow separation as shown in Figure 7
via the computed surface streamlines. These surface
streamlines appear to be drawn into the scoop area, but
there is no indication of flow reversal. The streamlines
also flow smoothly around the skeg. It should be noted
that there is no propulsion modeling for this
calculation. The calculation indicates a vertical flow
will enter the propulsor, which may be an area of
concern for propulsor design. The vertical flow in and
of itself can have some influence on propeller
performance due to the local angle of attack imposed
on the blades. Of more significance may be the
velocity deficit associated with this flow, since the
OCR for page 151
axial flow entering the propulsor is significantly slower
than the free stream value.
.01
.02
.03
.05
Ins'
_
0.025 0.05
Figure 6: Axial velocity contours and cross flow
velocity vectors for Concept-2.
Figure 7: Computed surface streamlines for Concept-
2.
As part of the design phase a new bow dome
was developed in an attempt to reduce the strength of
the vortices created there since they flow downstream
directly into the propulsor. To save time only a bare
hull computation is performed for this new hull, which
also has a lengthened forebody, and is referred to as
Concept4. In addition, Concept4 does not contain the
scoops or skeg so some insights of their effect on the
flow field can also be obtained. Computed axial
velocity contours for this configuration at several
longitudinal locations are shown in Figure 8.
-
a__
by_
Bare Hull Ccacopt 4_~
A:;
Figure 8: Computed axial velocity contours for
Concept4.
It should be mentioned that a grid very similar to that
used for Concept-2 is used here so that the computed
differences should be due to geometry changes and not
grid changes. Because the length of the ship has
changed, primarily in the section forward of the
shoulder, it is difficult to do a one-to-one comparison
of the vortices formed by the two bow domes.
However, it is evident that with Concept-4 the vortices
stay closer inboard, which is a possible indication of
less strength as there is less convective spreading of the
vortex pair. This is evident all the way to the stern of
the hull, although it these downstream locations the
lack of a skeg can be keeping the vortices closer
together than in Concept-2 where the skeg may push
them apart. Additionally, it appears the scoops have a
significant influence on the extent of the low velocity
flow that would enter the propulsor. For Concept-4,
where there are no scoops, the vortices become
perturbations to the hull boundary layer at the
downstream locations where the propulsors would be.
As had been seen in Figure 5, the vortices in the scoop
become the dominant feature that enter the propulsor
and stand out clearly from the hull boundary layer.
Computed surface streamlines over both bows
are shown in Figure 9. There is more turning of the
bow dome streamlines and subsequent collapse to a
limiting streamline for Concept-2 than Concept-4.
This again indicates the bow dome vortices generated
by Concept4 are an improvement over those computed
for Concept-2. A comparison of the axial velocity
contours, between the two configurations, near the
forward shoulder of the ship is shown in Figure 10.
From this location downstream, the lengthening of the
ship has no further impact and provides a good
comparison location. It is seen that the Concept4
vortex system is inboard of that predicted for Concept-
2. The axial velocity deficit is similar between the two
configurations. However, the secondary flow
7
OCR for page 152
generated by Concept4 is much weaker than that
generated by Concept-2. Hence, it appears the bow
dome vortices generated by Concept4 are an
improvement over those generated by Concept-2. It is
not clear whether the axial velocity deficit entering the
propulsor is dominated by the bow dome vortices or
the scoops. Since the propulsor is embedded in the hull
it is felt that the scoops are necessary to provide clean
inflow to the propulsor. Consequently, pursuit of
improvements to the inflow by minimizing the bow
dome vortices seems worthwhile.
Conng~6on 2 SduUen
it_
Cordiguatlon4 Sodden
Figure 9: Computed surface streamlines over the
bows of Concept-2 (top) and Concept4 (bottom).
CordIguadon 2
Con11gula~don 4
Figure 10: Cross flow velocity vectors and axial
velocity contours for Concept-2 (left) and Concept-4
(right).
In an effort to obtain some additional flow
enhancements, the final configuration calculated with
RANS as part of the design process is a modified
version of Concept4, referred to as Concept-5. The
modifications of Concept-5 are considerably simpler
than those performed when going from Concept-2 to
Concept-4. Consequently, it is expected the flow
changes will also be smaller. The primary
enhancements consist of further fairing the bow dome,
softening the shoulder in the sectional area curve, and
softening the bilge radius approaching the transom
knuckle. Very little difference is seen between the
viscous boundary layer flow on a bare hull for this new
configuration and that of Concept4. A comparison of
the axial velocity contours and secondary flow
generated by Concepts-4 and -5 just downstream of the
bow dome is shown in Figure 11. The flows are quite
similar with no detrimental effect on the bow dome
vortex from these latest modifications. It should be
noted that only the double model solution is shown
here. The grid for Concept-5 is similar to that for the
previous two bare hulls. With the Concept-5 bare hull
calculations, both linear and nonlinear free surface
solutions have been obtained with UNCLE v1.2.6. The
presence of the free surface has little effect on the bow
dome vortices or the viscous flow that would enter the
propulsor. Consequently, it is believed that the double
model calculations provide useful information on the
flow field.
0 962795
0 913421
0 Be4047
0 s, 4672
0 76s2s8
0.7' Ss24
0 66 665
o61 7176
0567802
0s'8428
0 469054
041969
0370306
0320932
0 27 1557
0 222183
0 172809
0 123435
0 07 40511
0~24~87
Concept 4
Figure 11: Comparison of Concept-4 (right) and
Concept-5 (left).
Concept-3 with Propulsor
The bare hull calculations give an indication
of some of the flow issues involved with these complex
configurations, but do not truly provide the details of
8
OCR for page 153
the flow into the propulsor which is needed for ithe
propulsor design. The Concept-3 calculations provide
some of this pertinent information. For Concept-3, the
propulsor is embedded in the hull and there is also a
hub inside the duct. This ducted geometry provides
significant blockage to the flow field and the flow can
no longer flow smoothly down the hull as observed for
Concept-2. The duct and hub are included in the
RANS calculations of Concept-3, but it is necessary to
model the thrust through the propulsor as part of the
RANS calculation. This is done by using an actuator
disk model, which imposes a constant axial force over
a plane inside the duct passage. The force is applied at
a single axial plane and provides the net effect and not
the individual effects of the internal components. The
duct and hub are part of the RANS calculation and do
not need any additional modeling. The axial force
supplies a thrust to the model, but no radial or
tangential force components are imposed. The
computed axial velocity contours at several
longitudinal positions along the body are shown in
Figure 12. The flow upstream of the propulsor is
nearly identical to that computed for Concept-2 as it
should be, since Concept-3 is a ducted version of
Concept-2. The Concept4 and -5 bare hull
configurations were pursued concurrently with the
Concept-3 calculations. It is clearly evident that the
vertical structures, and associated velocity defects,
flow along the hull and enter directly into the
propulsors. In these figures, red is near free stream
velocity and blue is low speed flow. This figure
indicates that the flow in the entrance region to the
propulsor is slower than that farther upstream. This is
because of the blockage induced by the propulsor
despite the flow being propelled via the actuator disk
model.
_
_
_
_
~ _
a_._
a_ _
wen_~
_h
Figure 12:
Concept-3
Axial velocity contours just downstream of
the leading edge of the duct are shown in Figure 13.
The flow inside the duct is significantly slower than the
free stream with a mean value of roughly 70 percent of
ft. This is unlike many conventional surface ships
where the propeller is operating in a nearly inviscid
free stream flow field. Noticeable in this and the
previous figure is the low velocity flow which has been
pulled out from the boundary layer due to the vertical
nature of the flow entering the propulsor. Also seen in
Figure 13 is the outline of the large vertical structure
which propagates along the hull with the propulsor
clearly in the center of it. The boundary layer on the
skeg is also seen, but this seems to have little impact on
the propulsor inflow. The axial velocity contour at
roughly 70 percent of the propulsor length is shown in
Figure 14. Here the flow is somewhat faster than that
at the entrance. There is also an interaction between
the two propulsors producing the higher velocity flow
on the inboard side of the propulsors. The flow at the
exit of the propulsor, Figure 15, indicates that the flow
velocity is increasing, but is still less than the free
stream, which may indicate under propulsion with the
actuator disk model. A velocity contour plot further
downstream, Figure 16, indicates there is still some
wake structure associated with the propulsor, but it is
not the significant vertical type structure seen entering
the propulsor. Another view of how the axial flow
changes through the propulsor is shown in Figure 17,
which is a cut along the centerline lengthwise through
the propulsor. The axial velocity keeps decreasing as it
approaches the propulsor. This is due to a combination
of the increasing boundary layer thickness as well as
the decreasing depth of the hull. Additionally, it is
believed much of the axial velocity decrease is due to
the blockage effects of the propulsor. Of significance
Computed axial velocity contours for
is the low velocity near the hull, which enters the
propulsor and could be a cause for concern if it were to
separate. As one progresses longitudinally along the
propulsor the flow velocity eventually increases as it
exits at nearly the free stream velocity. It should be
noted that the change in flow from the inlet to the exit
of the duct is controlled by mass conservation and the
changing duct area. The actuator disk model
influences the magnitude of the flow entering the
propulsor. Without this actuator disk model there
would be significantly lower velocity at the entrance to
the propulsor and some possible flow separation.
9
OCR for page 154
r`L
-n no
-n nd
NSWCCD (6199)
Computed Propu~or Inflow
(AXEII Velocity Contours)
Concept 3
0 0.025 of. ~ _
Z/L
Figure 13: Axial velocity at the inflow to the
Concept-3 propulsor.
no
non
-0.04
NSWCCD (6/99)
Ux
1.10
1.05
1.00
0.95
0.90
0.85
0.80
0.75
0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
PropulsorThrough Flow
(Axial Velocity Con~um)
Concept 3
0 0.025 0.05 0.075
Z/L
Figure 14: Axial velocity contours at 70% of the
propulsor length.
-0.04
NSWCCD (6199)
Computed Propulsor Outflow
(Axial Velocity Contours)
Concepts
0 0.025 0.05 0.075
Z/L
Figure 15: Axial velocity contours at the exit of the
Concept-3 propulsor.
1 Ox
1.10
1.05 -0.02
1.00
1 0.95
1 o.so _,
0.85
0.80
0 75 o 04
0.65
0.60
0.55
0.50
0.45
0 40 -0.06
0.35
0.30
10
Ux
1 .10
1.05
1.00
0.95
0.90
0.85
0.80
0.75
0.70
0.65
0.60
0.55
0.50
0.45
0.40
0.35
, 0.30
O.025
Z/L
n rye
0.075
Figure 16: Axial velocity contours downstream of the
Concept-3 propulsor.
OCR for page 155
us
0.8 0.85 o.9
X/L
Figure 17: Axial velocity contours on a vertical plane
through the centerline of the Concept-3 propulsor.
The original version of Concept-3 had an
abrupt step from the propulsor to the hull at the exit to
the propulsor. This is because of the decreasing draft
of the hull at the stern. This step led to significant
flow separation at the exit to the propulsor as seen in
Figure 18. Consequently, ramps were added to the
propulsion exits to provide a smooth transition from
the propuslor to the hull. Computed surface
streamlines for this configuration are shown in Figure
19. The streamlines clearly enter and exit the
propulsor with no apparent separation so exit ramps
were also included in the final Gov't-2 design. Also of
. .
Interest Is the strong turning of the flow around the
propulsors both inboard and outboard of them. The
strong turning to the outboard side of the propulsors
helps give an indication of why the propulsors can have
a strong effect on the free surface waves generated near
them. The inboard turning of the flow, and its
interaction with the skeg, provides a high axial velocity
between the propulsors which was indicated in the
axial velocity contour of Figure 14.
Figure 18: Computed surface streamlines without
ramps at the propulsor exit.
Figure 19: Computed surface streamlines for
Concept-3.
The computation for Concept-3 gives some
indication of the complexity of the flow field generated
by the propulsor/hull interaction. The flow upstream of
the propulsor is highly viscous. The subsequent inflow
to the propulsor is significantly different than what is
found on conventional surface ship configurations.
Similarly, the propulsor has a strong affect on the
surrounding flow field. The ability to dissect such flow
field computations to help understand the underlying
flow physics is one of the advantages of RANS
calculations where the entire flow field is available for
. . .
Investigation.
GOV'T-2 CALCULATIONS
The final Gov-t-2 design consists of the
extended forebody of Concept4 with the sonar dome
of Concept-5. Additionally, the skeg has been
eliminated and the propulsors have been revised and
moved forward, inboard, and further into the hull than
used on Concept-3. Overall, the computed flow field
appears similar to that for Concept-3 already shown in
Figure 12 through Figure 19. This hull form was tested
at NSWCCD to provide a validation database for the
computations. Me asurements consisted of
longitudinal wave cuts, stern topography, and velocity
measurements of the propulsor inflow.
Calculations are performed at model scale
corresponding to a Reynolds number, based on model
speed and length, of 15.44e+06 and a Froude number
of 0.232. Like Concept-3, body force terms are
included for propulsion approximately halfway through
the propulsor in the streamwise direction. The body
force terms model the total thrust of all the blades
combined and do not include forces generated by the
hub or duct since these are computed directly from the
RANS calculation. This is a total blade force and is
11
OCR for page 156
represented in the code at an axial location,
corresponding to a single axial grid plane. The total
torque of the unit is negligible so no torque is modeled
via body forces. Additionally, the body force is
applied uniformly across the disk plane with no attempt
to model radial variations in the force model.
To better estimate the quality of a solution
there have been efforts to develop uncertainty estimates
and validation procedures for computations (e.g.
ALLA, 1998; Roache, 1998~. This is an area of
significant importance as the computational community
tries to provide metrics for how good a computation is
and more work needs to be done in this area. To this
end though it is becoming largely recognized that one
cannot declare a particular code validated because a
particular solution depends on many factors including:
geometry definition, grid quality, turbulence modeling
and user experience among other variables. In the
current effort an attempt was made to apply the
verification and validation procedure of Stern et al
(1999~. This requires solutions on three different grids.
As already discussed solutions are obtained for the
Gov't-2 hull form on three different grids consisting of
a fine or large grid containing almost 9.5 million points
(after blocking for parallel processing), a medium grid
containing about 1.25 million points, and a coarse grid
of about 175 thousand points. To implement the
procedure correctly all three grids must be in the
asymptotic limit of a final grid independent solution.
This is not the case for the coarse grid here and
uncertainty estimates obtained with it can be erroneous
as discussed by Ebert and Gorski (2001~.
Consequently, uncertainty estimates are not provided
here, but it is still very instructive to look at the grid
dependence of the computed solutions.
The first comparison shown is for the wave
profile on the hull. As shown in Figure 20 the fine grid
provides a very good comparison with the measured
profile on the hull. Here the bow of the hull is at X/L =
O and the stern at X/L = 1. The medium grid captures
some aspects of the measured data, but the coarse grid
loses many of the details. Longitudinal wave cut data
at various locations away from the hull is also available
for comparison. At Y/B = 0.66, Figure 21, a
comparison similar to that shown for the hull profile is
obtained with the fine grid capturing the wave heights
well, but the coarse grid has significant damping.
Downstream of the hull the computations quickly start
to dissipate the computed wave field as the
computational grid expands. This dissipation of the
computed waw field away from the hull is illustrated
further in Figure 22, which provides a comparison of
the longitudinal wave cut at Y/B = 2.5. Here it is seen
that even the fine grid eventually damps out the wave
away from the hull and the coarser the grid the more
quickly the waves are dissipated with distance from the
hull.
new
0.004
n on'
Hi
o
n on;
-0.004
— Coarse Grid
Medium Grid
. . - - Fine Grid
0 Measured
a
a
0 09 0.4 0.6
XJL
0.8 1
Figure 20: Measured and computed wave height on
the hull.
0.005
0.004
0.003
0.002
0.001
O
-0.001
-0.002
-0.003
-0.004
0 Tank Y/F = 0~6
- Coarse Grid
Medium Grid
Fine Grid
-0.5 0
XIL
Figure 21: Computed and measured wave height at
Y/B =0.66.
OCR for page 157
-
0.004 _ _
0.003
0.001
-0.00
-0.002
-0.003
1
o Tank YE ~ 2~50 |
Coar" land
Medium Grid
— Fine Grid
Coarso Grld
n In
.O nS
-0.5 0 0.5 1 1.5 ~ in
XJL
Figure 22: Computed and measured longitudinal wave
height at Y/B = 2.5. a) Coarse and
Comparisons with free surface heights
measured immediately downstream of the hull are
shown in Figure 23 for all three grids. Here the
computations from the three different grids have been
interpolated onto the same "grid" the measurements
were obtained on to provide a truer comparison. Each
figure has the computed wave heights on top and the
measured wave heights on the bottom. The data shows
a rooster tail forming immediately aft of the hull with a
trough downstream of that. The coarse grid does a
very poor job of representing the measured data and
significantly under predicts the rooster tail height md
following trough depth. The medium grid comparison b) Medium and
is better containing much of the correct behavior of the
flow, but over predicts the height of the rooster tail and
the subsequent trough behind the rooster tail is not as
deep as it should be. The fine grid solution is really
excellent capturing the flow extremely well. n Ace,
The measured axial and secondary flow
velocity at X/L = 0.77 is shown in Figure 24, which is
in the scooped out area just upstream of the propulsor.
This data was taken with a LDV system and axial
velocity contours are shown with secondary flow
vectors. The dominant feature is the vortex, which is
generated at the bow dome, as was shown with the
previous computations. This flow is atypical of
propeller inflow for conventional ships. There is a
large region of low velocity created by the vortex
pulling boundary layer flow out away from the hull.
Additionally, some high velocity flow is pulled in
toward the hull by the vortex.
1.0 1.1 1.2
XJL
Medium Grld
0.10
0.05
0.00
-0.10
-0.05
-0.10
c) Fine grid
1.0 1.1 1.2
XIL
Figure 23: Computed and measured free surface
height in the stern region: a) Coarse grid, b) Medium
grid, c) Fine grid.
13
OCR for page 158
i:)
D
-~0 ~ ~
-any
by:
'0.~
-~
Figure 24: Measured velocities at X/L = 0.77.
Rae
1 ~:.~
(>sand
do ~ ~
1
raspy
i
~ In,
@ t>~
i ~~!
1 ~~!
1 'by '
rem '.
-~.r~
I..
. . ...... i. .... . , ~
..~.. .
~ 5.~ ~~ y';:r$ 0~ *~ a) Coarse grid
i.it
i.
.^
-~.O It:
The computed flow with all three grids at this
location is shown in Figure 25. The solutions from all
three grids have been interpolated onto the
experimental "grid." The very low velocity predicted
next to the hull in the computations is not present in the
measurements since the flow could not be measured
that close to the hull. The most apparent difference
between the computed and measured data is a slight
shift of the vortex location. This shift is probably due
to the computed vortex not being as strong as the actual b) Medium grid
vortex, which can be inferred from the secondary flow
vectors. However, the agreement is very good and the
predictions are accurate enough to provide cavitation
estimates for the propulsor. The solutions just
discussed are with the k-£ model. A second
calculation performed with the Ace model on the
medium grid showed a slightly stronger vortex and
shifted the center slightly more toward the
experimental data, but the changes are not significant
enough to show here. One significant difference
between these comparisons and the wave height
predictions is the grid dependence of the solution.
While the predicted wave heights are very grid
dependent the computed axial and secondary flow
fields seem more grid insensitive. However, the coarse
grid is overly dissipative as seen from the overly thick
boundary layer on the hull and the larger low velocity c) Fine grid
area due to the vortex as compared to the finer grid
solutions and measured data.
14
-.
A,
_~1i']4
~ C'£-i ~ `- t40~ t.1,~ 1~.~ Alit
· yea
. ~ ~ . . .
`~;wO 1 (
~~ ~1
Figure 25: Computed velocities at XIL = 0.77: a)
Coarse grid, b) Medium grid, C) fine grid.
OCR for page 159
CONCLUSIONS
The RANS calculations reported here are done
as part of the design process for the Gov't-2 geometry
of the ONR Surface Combatant Accelerated
Hydrodynamics S&T Initiative. The effort
demonstrates that RANS computations can be done
quickly and accurately enough to provide ship design
guidance. Although the free surface is not included in
the RANS calculations in the early design phase of this
study, the RANS calculations do provide useful insight
to aid in the design. Specifically, the RANS
calculations provide information on inflow to the
propulsor as well as a means to evaluate the different
bow dome geometries identified. For these purposes
double model calculations provide nearly as much
information as a computation with the free surface.
One thing that needs to be considered when using
RANS for design work is the ability to quickly re-grid
the geometry as it changes. The current geometry with
its ducted propulsors could not be re-gridded quickly
enough when the propulsors are moved. Consequently,
there was an emphasis on bare hull calculations. To
make RANS a viable design tool the re-gridding of
complicated geometries to reflect changes must
become more automated or the use of RANS solvers,
and their associated flow information, will be limited.
An accurate flow prediction method is
necessary to provide input for novel surface ship
concepts more quickly and more cheaply than possible
with conventional build and test procedures. Because
of increasing requirements on surface ship performance
high fidelity is needed in these computations. To meet
these requirements there is a push for integrated
hull/propulsor designs which require viscous codes
with free surface effects for their prediction which this
effort addresses. The comparisons done here with the
measured experimental data for the Gov't-2 hull form
demonstrate that RANS codes can provide accurate
representation of the propulsor inflow as well as the
near-field wave heights. Such capability should
improve as experience is gained in performing
calculations for such hull forms. Consequently,
RANS is a viable tool to aide in the design of surface
ships.
It is important to note that the RANS
computations do not provide designs. The
computations only provide information on the flow
field and a means to evaluate design options.
Individual designs are developed based on the
experience of the team as well as iterations on various
options based on information provided by the
computations. It is felt that the best means of
evaluating designs currently is the combination of
calculations and experiments. The computations
provide an overall picture of the flow field and are a
good way to gain understanding of the flow physics
involved. Additionally, the computations provide a
way to evaluate design options more quickly than is
possible with physical model testing. However, the
computations have some limitations because of grid
dependencies and turbulence modeling deficiencies.
Experiments on the other hand provide highly detailed
flow information, but typically only in specific areas.
Consequently, it appears a good approach is to use the
computations to help understand the flow physics and
interactions of the various components such as the hull
and propulsors, as well as to rank order individual
modifications. Experiments for the final designs need
to be done to both verify that the computational based
designs meet the desired goals as well as to get details
of the flow in specific areas to validate the
computations at these locations.
ACKNOWLEDGEMENTS
This effort was funded by the Office of Naval
Research as part of the ONR Surface Co mbatant
Accelerated Hydrodynamics S&T Initiative under the
direction of Dr. Edward Rood. The authors would like
to thank Ian Mutnick, Toby Ratcliffe and Christopher
Chesnakas of NSWCCD for providing the
experimental data for comparison. Much useful
advice' in the use of the UNCLE code, was also
provided by Murali Beddhu, Ramesh Pankajakshan,
and Lafette Taylor of MSU. In addition, this work
was supported in part by a grant of computer time from
the U.S. Army Research Laboratory at Aberdeen as
part of the DOD High Performance Computing
Challenge Project on Time Domain Computational
Ship Hydrodynamics. The authors would also like to
thank the U.S. Navy Hydrodynamic/Hydroacoustic
Technology Center for the use of its facilities and
software licenses.
REFERENCES
AIAA, "Guide for the Verification and Validation of
Computational Fluid Dynamics Simulations", G077-
1998, 1998, American Institute Aeronautics and
Astronautics.
Allison, J. L., Becnel, A. J., Gorski, J. J., Hoyt, J. G.,
Purnell, J. G., Stricker, J. G., and Wilson, M. B.,
"Research in Waterjet Inlet, Hull, and Jet Interactions,"
Int. Conf. Water~et Propulsion III, Gothenburg,
Sweden, 2001.
Beddhu, M., Pankajakshan, R., Jiang, M.- Y.,
Remotigue, M., Sheng, C., Taylor, L., Briley, W.R.,
and Whitfield, D.L., "Computation of Nonlinear
15
OCR for page 160
Turbulent Free Surface Flows Using the Parallel
UNCLE Code," Proc. 23rd Symposium on Naval
Hydrodynamics, Val de Reuil, France, Sep. 2000.
Chorin, A. J., "A Numerical Method for Solving
Incompressible Viscous Flow Problems," Journal of
ComputationalPhysics,Vol.2, 1967,pp. 12-26.
Coakley, T. J., "Turbulence Methods for the
Compressible Navier-Stokes Equations," AIAA Paper
No. 83-1693, 1983.
Ebert, M. P. and Gorski, J. J., "A Verification and
Validation Procedure for Computational Fluid
Dynamics Solutions," NSWCCD-50-TR-2001/0006,
Feb. 2001, Naval Surface Warfare Center, West
Bethesda, MD.
Gorski, J. J., "Drag Calculations of Unappended
Bodies of Revolution," CRDKNSWC/HD-1362-07,
May 1998, Naval Surface Warfare Center, West
Bethesda, MD.
Gorski, J. J., "Marine Vortices and Their
Computation," Proc. NATO RTO AVT Panel
Symposium on Advanced Flow Management Loen,
Norway, May 2001a.
Gorski, J. J., "Present State of Numerical Ship
Hydrodynamics and Validation Experiments," Proc.
OMAE'01 20th Int. Conf. Offshore Mechanics and
Arctic Engineering, Rio de Janeiro, Brazil,
OMAE01/OFT-1350, June, 2001b.
Gorski, J. J. and Coleman, R. M., "Use of RANS
Calculations in the Design of a Submarine Sail," Proc.
NATO RTO AVT Symposium on Reduction of
Military Vehicle Acquisition Time and Cost Through
Advanced Modeling and Virtual Product Simulation,
Paris, France, April 2002.
Gorski, J. J., Haussling, H. J. and Coleman, R. M.,
"Model and Full Scale Predictions of a Carrier Flow
Field," NSWCCD-50-TR-2002/009, January 2002,
Naval Surface Warfare Center, West Bethesda, MD.
ITTC, "Report of the Resistance and Flow
Committee," Proc. 21st ll-lC, Trondheim, Norway,
1996.
Liou, W. and Shih, TH., "Transonic Turbulent Flow
Predictions with New Two-Equation Turbulence
Models," NASA Contractor Report 198444, Jan. 1996.
Pankajakshan, R., Taylor, L.K., Jiang, M., Remotigue,
M.G., Briley, W.R., and Whitfield, D.L., "Parallel
Simulations for Control-Surface Induced Submarine
Maneuvers," AIAA Paper 2000-0962, 38th Aerospace
Sciences Meeting, Reno NV, Jan. 2000.
16
Roache, P. J., Verif~cation and Vaii]ahOn in
Computational Science and Encineering, Hermosa
Publishers, Albuquerque, New Mexico, 1998.
Rood, E. P., "Computational Ship Hydrodynamics for
Revolutionary Naval Combatants," Proc. DoD HPC
Users' Group Conf., Albuquerque, NM, June, 2000.
Sclavounos, P. D., "Computation of Wave Ship
Interactions," appeared in Advances in Marine
Hydrodynamics, M. Ohkusu, ea., Computational
Mechanics Publications, Southampton, chap. 4, 1995,
pp. 233-278.
Sheng, C., L. Taylor, and D. Whitfield, "Multiblock
Multigrid Solution of Three-Dimensional
Incompressible Turbulent Flows About Appended
Submarine Configurations," AIAA Paper No. 95-0203,
Jan. 1995.
Stern, F., Wilson, R. V., Coleman, H. W., and
Paterson, E. G., "Verification and Validation of CFD
Simulations", IIHR Report No. 407, Sept. 1999, Iowa
Institute of Hydraulic Research, Iowa City, IA.
Subramani, A. K., Paterson, E. G., and Stern, F., "CFD
Calculation of Sinkage and Trim," J. Ship Research,
Vol. 44, No. 1, March 2000, pp.59- 82.
Taylor, L. K. and D. L. Whitfield, "Unsteady Three-
Dimensional Incompressible Euler and Navier-Stokes
Solver for Stationary and Dynamic Grids," AIAA
Paper No. 91-1650, June 1991.
Taylor, L. K., A. Arabshahi, and D. L. Whitfield,
"Unsteady Three-Dimensional Incompressible Navier-
Stokes Computations for a Prolate Spheroid
Undergoing Time-Dependent Maneuvers," AIAA
Paper No. 95-0313, Jan. 1995.
Valkhof, H. H., Hoekstra, M., and Andersen, J. E.,
"Model Tests and CFD in Hull Form Optimization,"
SNAME Trans., Vol . 106, 1998, pp. 391 - 412.
Van Leer, B., J. L. Thomas, P. L. Roe, and R. W.
Newsome, "A Comparison of Numerical Flux
Formulas for the Euler and Navier-Stokes Equations,"
AIAA Paper No. 87-1104-CP, June 1987.
Whitfield, D. L. and L. K. Taylor, "Discretized
Newton-Relaxation Solution of High Resolution Flux-
Difference Split Schemes," AIAA Paper No. 91-1539,
June 1991.
Wyatt, D. C., "Development and Assessment of a
Nonlinear Wave Prediction Methodology for Surface
Vessels," Journal of Ship Research, Vol. 44, No. 2,
June 2000, pp. 96-107.
Representative terms from entire chapter:
free surface
