29th Symposium on Naval Hydrodynamics

Gothenburg, Sweden, 26-31 August 2012

Computational Ship Hydrodynamics: Nowadays and Way

Forward
Frederick Stern, Jianming Yang, Zhaoyuan Wang,
Hamid Sadat-Hosseini, Maysam Mousaviraad, Shanti Bhushan1, Tao Xing2
(IIHR-Hydroscience & Engineering, University of Iowa,
Iowa City, IA 52242, USA)
basin facilities are needed additionally for model
ABSTRACT development and CFD validation, which requires even
more advanced measurement systems for global and
Computational fluid dynamics for ship hydrodynamics local flow variables and more stringent requirements
has made monumental progress over the last ten years, on experimental uncertainty analysis as it plays an
which is reaching the milestone of providing first- important role in validation procedures.
generation simulation-based design tools with vast In the following, the development of computa-
capabilities for model- and full-scale simulations and tional ship hydrodynamics over the past 30 years is
optimization. This is due to the enabling technologies briefed using example references idiosyncratic to the
such as free surface tracking/capturing, turbulence authors and their colleagues. In the early 1980s integral
modeling, 6DOF motion prediction, dynamic overset methods still predominated, which worked well for
grids, local/adaptive grid refinement, high performance two-dimensions but had great difficulty in extensions
computing, environmental modeling, and optimization to three-dimensions due to inability to model cross-
methods. Herein, various modeling, numerical flow velocity profiles (von Kerczek et al., 1984). Thus
methods, and high performance computing approaches three-dimensional boundary layer finite difference
for computational ship hydrodynamics are evaluated methods were soon developed, which worked well for
thereby providing a vision for the development of the thin boundary layers but had great difficulty for thick
next-generation high-fidelity simulation tools. Verifica- boundary layers and flow separation (Stern, 1986).
tion and validation procedures and their applications, Quickly partially parabolic approaches were developed
including resistance and propulsion, seakeeping, (Stern et al., 1988a) followed by full RANS solvers
maneuvering, and stability and capsize, are reviewed. with viscous-inviscid interaction approaches for
Issues, opportunities, and challenges for advancements nonzero Froude number (Tahara et al., 1992). Soon
in higher-fidelity two-phase flow are addressed. thereafter large domain RANS methods using free
Fundamental studies for two-phase flows are also surface tracking methods (Tahara et al., 1996) along
discussed. Conclusions and future directions are also with extensions for improved turbulence and propulsor
provided. modeling, multi-block, overset grids and parallel
computing were developed (Paterson et al., 2003)
1 INTRODUCTION allowing full/appended/model captive simulations for
resistance and propulsion. Next enabling technologies
In just over 30 years computational fluid dynamics of level-set free surface capturing, inertial reference
(CFD) for ship hydrodynamics has surpassed all frames, and dynamic overset grids allowed wave
expectations in reaching astronomical progress, breaking and ship motions (Carrica et al. 2007) and of
capabilities and milestone of providing the first- anisotropic URANS and DES turbulence modeling
generation simulation-based design (SBD) tools for allowed better resolved turbulence (Xing et al., 2007;
model- and full-scale simulations and optimization 2012). Extensions soon followed for semi-coupled air-
enabling innovative cost-saving designs to meet the water flows (Huang et al., 2008), 6DOF simulations
challenges of the 21st century, especially with regard to using controllers for calm-water maneuvering (Carrica
safety, energy and economy. CFD is changing the face et al., 2012b) and capsize predictions (Sadat-Hosseini
of ship hydrodynamics as the SBD approach is et al., 2011b) including rotating propellers (Carrica et
replacing the now old-fashioned build-and-test al., 2012a), exhaust plumes (Huang et al., 2012), wall
approach such that model testing is only required at the functions for full-scale DES simulations (Bhushan et
final design stage; however, towing tank and wave al., 2012c), damaged stability including motions
1 Current affiliation: Center for Advanced Vehicular Systems, Mississippi State University, Starkville, MS 39759
2 Current affiliation: Mechanical Engineering Department, University of Idaho, Moscow, ID 83844
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

(Sadat-Hosseini et al., 2012c), and high performance in the unprecedented achievements of computational
computing (HPC) (Bhushan et al., 2011a). Innovative ship hydrodynamics.
procedures not possible in towing tanks were also Progress in CFD for ship hydrodynamics has
developed for both resistance and propulsion (Xing et been well benchmarked in CFD workshops for
al., 2008) and seakeeping (Mousaviraad et al., 2010) resistance and propulsion and seakeeping (most
and CFD with system identification has shown ability recently, Larsson et al., 2011) and calm water maneu-
for improvement in system-based mathematical models vering (Stern et al., 2011a) along with the Proceedings
for maneuvering in calm water and waves (Araki et al., of the ITTC both for applications and CFD itself.
2012a, b). Optimization capabilities for ship hydrodynamics were
The next-generation high-fidelity SBD tools recently reviewed by Campana et al. (2009). Sanada et
are already under development for milestone achieve- al. (2012) provides an overview of the past captive
ment in increased capability focusing on orders of towing tank and current free running wave basin
magnitude improvements in accuracy, robustness, and experimental ship hydrodynamics for CFD validation
exascale HPC capability for fully resolved, fully as background for description of the new IIHR wave
coupled, sharp-interface, multi-scale, multi-phase, basin and trajectories and local flow field measure-
turbulent ship flow utilizing billions of grid points. ments around the ONR tumblehome in maneuvering
Current capabilities are for Cartesian grids with motion in calm water and head and following waves.
immersed boundary methods (Yang and Stern, 2009; Computational ship hydrodynamics current
Wang et al., 2009a, b), for orthogonal curvilinear grids functionality, initiation of the development of the next
(Wang et al., 2012a, b), for overset Carte- generation high-fidelity SBD tools, contributions to
sian/orthogonal curvilinear grids (Bhushan et al., V&V and CFD education, research paradigm and
2011b), and extensions in progress for non-orthogonal international collaborations, CFD workshops and ITTC
curvilinear grids. High-fidelity large eddy simulation Proceedings and optimization capabilities as demon-
(LES) simulations for plunging breaking waves and strated by the example references given above arguably
surface-piercing wedges and cylinders have resolved equals if not surpasses other external flow industrial
for the first time and identified physics of the plunging applications such as aerospace, automotive and rolling
wave breaking process (Koo et al., 2012), spray stock capabilities such that ship hydrodynamics in spite
formation (Wang et al., 2010b) and wake spreading of its relatively small size community is at the forefront
(Suh et al., 2011). Realization non-orthogonal curvilin- in computational science and technology and research
ear grids (Yang et al., 2012) will enable similarly and development.
resolved simulations for practical geometries and Herein computational ship hydrodynamics is
conditions with increased physical understanding reviewed with a different perspective and special focus
thereby revolutionizing ship design; however, consid- on the critical assessment of modeling, numerical
erable research is still needed, as high-fidelity general methods and HPC both nowadays and prognosis for
purpose solvers with the aforementioned functionality way forward. Quantitative V&V procedures and their
do not yet exist. application for evaluation of captive and free running
Quantitative verification and validation simulation capabilities along with fundamental studies
(V&V) procedures and an adequate number of well- for two-phase flows are also reviewed with the latest
trained expert users are also essential ingredients for results obtained at IIHR as selected examples. Conclu-
the successful implementation of SBD. Here again, sions and future directions are also provided.
computational ship hydrodynamics has played
leadership role in V&V (Stern et al., 2006a; Xing and 2 COMPUTATIONAL SHIP HYDRODYNAM-
Stern, 2010) and development of CFD educational ICS
interface for teaching expert users at both introductory
and intermediate levels (Stern et al., 2006b; 2012). Application areas are at the core of computational
V&V research is still needed especially for single-grid method requirements as they guide the choice of
methods and LES turbulence models. General-purpose modeling, which in return guide the grid and accuracy
CFD educational interfaces for teaching CFD are not requirements of the simulation. The grid requirements
yet available. along with HPC determine the efforts required for grid
The research paradigm of integrated code de- generation, problem setup, solution turnaround time
velopment, experiments, and uncertainty analysis along and post processing efforts. Computational methods for
with step-by-step building block approach and ship hydrodynamics include modeling, numerical
international collaborations for synergistic research methods and HPC capability as summarized in Fig. 1.
magnifying individual institute capabilities as exempli- Models required for naval applications are hydrody-
fied by IIHR (Stern et al., 2003) has been foundational namics, air flow and two-phase flow solvers, turbu-
lence models, interface models, motion solvers,
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

propulsion models, sea condition or wave models, etc. applications, since the water phase accounts for most
The numerical methods encompass the grids and resistance. However, most of these solvers are not
discretization schemes for the governing equations. capable of solving problems with wave breaking and
High performance computing encompasses the ability air entrainment, which have become more and more
to use larger grids, more parallel processors and important in ship hydrodynamics due to the develop-
speedup solution turnaround time. ment of non-conventional hull shapes and studies of
ITTC 2011 Specialist Committee on Compu- bubbly wake, among others.
tational Fluid dynamics report (ITTC, 2011) provides a
detailed review of numerical methods commonly used 3.1.2 Air Flows
for ship hydrodynamics. Most of them are also
discussed here, and readers are referred to ITTC (2011) For many problems in ship hydrodynamics, the effects
for the complete picture of CFD in ship hydrodynamics of air flow on the water flow are negligible but the air
from a different angle. The discussions herein focus on flow around the ship is still of interest. This includes
the advantages and limitations of the computational analysis of environmental conditions and air wakes
methods currently used in ship hydrodynamics, and around a ship in motion with complex superstructures,
recommendation are made for the most appropriate maneuverability and seakeeping under strong winds,
methods for a given application area. The following capsizing, exhaust plumes (Huang et al., 2012a), etc.
two sections also review upcoming computational Most CFD research of ship aero-hydrodynamics
methods focusing on the multiscale issues, which may simplified the problem by neglecting the free surface
provide hints of new development directions of high deformation and velocities, which restricted the range
fidelity solvers for ship hydrodynamics. The upcoming of problems that could be considered. A semi-coupled
numerical methods include higher-order discretization approach was developed by Huang et al. (2008) where
schemes and novel interface tracking schemes, and the water flow is solved first and the air flow is solved
HPC challenges of exascale computing. with the unsteady free-surface water flow as boundary
conditions. The limitation of the semi-coupled
3 MATHEMATICAL MODELING approach is its inabilities to deal with air entrainment,
wind-driven wave generation, cavitation, etc., as the
3.1 Ship flows water flow is only affected by the air flow through ship
motion driven by air flow load.
The fluids involved in ship hydrodynamics are water
and air (vapor phase in cavitation can be treated as a 3.1.3 Two-phase flows
gas phase as the air in the solvers). In general, they can
be considered as Newtonian fluids. The flow phenom-
In the two-phase solvers, both the air and water phase
ena can also be considered as incompressible due to
are solved in a coupled manner, which requires
usually very low Mach numbers. Therefore, the
treatment of the density and viscosity jump at the
governing equations for ship flows are the incompress-
interface (Huang et al., 2007; Yang et al., 2009). The
ible Navier-Stokes equations. Solvers for ship flows
two-phase solvers are more common in commercial
are categorized based on the solution methods for the
codes such as FLUENT, CFX, STAR-CCM+
two different fluids involved in as: (a) free-surface
(COMET) and open-source CFD solver OpenFOAM,
flow; (b) air flow; and (c) two-phase flow solvers.
as they are more general tools for a wide range of
applications. However, air flows including air entrain-
3.1.1 Free-surface hydrodynamics
ment were seldom shown in ship flow applications
performed with these solvers, due to high total grid
In free-surface flow solvers, only the water phase is resolution requirements for resolving the air flow
solved using atmospheric pressure boundary condition besides the water flow. On the other hand, two-phase
at the free-surface. Many ship hydrodynamics solvers models are slowly being implemented in upcoming
have adopted mathematical models for free-surface ship hydrodynamics research codes such as CFDShip-
models, for example, CFDShip-Iowa versions 3 Iowa version 6 (Yang et al., 2009) from IIHR, ISIS-
(Tahara et al., 1996) and 4 (Carrica et al., 2007) from CFD (Queutey & Visonneau, 2007) from ECN/CNRS,
IIHR, ship (Di Mascio et al., 2007) from INSEAN, FreSCo+ (Rung et al., 2009) from HAS/TUHH, and
SURF (Hino et al., 2010) from NMRI, PARNASSOS WAVIS (Park & Chun, 1999) from MOERI. Two-
(Hoekstra, 1999) from MARIN, ICARE (Ferrant et al., phase flow simulations are of interest in many applica-
2008) from ECN/HOE, WISDAM (Orihara & Miyata, tions, in particular, wind generated waves, breaking
2003) from the University of Tokyo, among others. waves, air entrainment, and bubbly wakes, among
These solvers are applicable in a wide range of others. Theoretically, it is possible to solve each phase
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

separately and couple the solutions at the interface. also be treated in both sharp and diffusive interface
However, this approach is only feasible for cases with manners, even though the specific treatments are not
mild, non-breaking waves or a very limited number of directly tied to the mathematical approximation of
non-breaking bubbles/droplets. Most solvers for jumps in the fluid physical properties. Detailed
practical applications adopt a one-field formulation in discussion of interface tracking is given in the numeri-
which a single set of governing equations is used for cal method section.
the description of fluid motion of both phases. In a one-
field formulation, it is necessary to identify each phase 3.2.3 Sea conditions and wave models
using a marker or indicator function; also, surface
tension at the interface becomes a singular field force Wave models are required to simulate flow fields with
in the flow field instead of a boundary condition in the incident waves or sea environments. Wave generation
phase-separated approaches. These issues are discussed can be achieved by imposing proper boundary
in the following air-water interface modeling section. conditions on the inlet boundaries. The boundary
conditions can be imposed by emulating the wave
3.2 Air-water interface modelling makers used in actual wave tanks or by imposing
velocity and wave height following the theories of
3.2.1 Interface conditions ocean waves. Ambient waves for the reproduction of
actual sea environments can be achieved by imposing
Air-water interface modeling must satisfy kinematic waves with a given spectrum (Mousaviraad, 2010). For
and dynamic constraints. The kinematic constraint deep water calculations, waves are considered as a
imposes that the particles on the interface remain on Gaussian random process and are modeled by linear
the interface, whereas the dynamic conditions impose superposition of an arbitrary number of elementary
continuous stress across the interface. The stresses on waves. The initial and boundary conditions (free
the interface are due to viscous stresses and surface surface elevations, velocity components and pressure)
tension. The latter is usually neglected for many ship are defined from the superposition of exact potential
hydrodynamics applications. solutions of the wave components. Sea spectra for
ordinary storms such as Pierson Moskowitz,
3.2.2 Interface representation Bretschneider, and JONSWAP, or for hurricane-
generated seas with special directional spreading may
One fundamental question for interface modeling is the be implemented. Linear superposition of waves can
indication and description of the interface. Smoothed also be used to create deterministic wave groups for
particle hydrodynamics (SPH) method uses particles of special purposes. Examples include especially designed
specified physical properties to identify phase infor- wave groups for single-run RAOs (Mousaviraad et al.,
mation without the need of tracking the interface 2010) and ship in three sisters rogue waves simulations
explicitly (e.g., Oger et al., 2006). The particle density (Mousaviraad, 2010). Figure 2 shows the exact
can be used as an indicator function to give the potential solution for a linear wave component and
interface position for specifying surface tension. Of generated random waves inside the computational
course, Lagrangian interface tracking methods such as domain as well as snapshots of the ship in three sisters
front tracking or marker point tracking can give simulations. For shallow water calculations, where the
accurate interface position for adding surface tension. nonlinearities are significant, regular nonlinear waves
However, it is still required to obtain a field function to may be generated using for example the Stokes second-
identify the phase information at each location within order perturbation theory. Numerical issues associated
the flow field. Eulerian methods such as volume-of- with application of such conditions include achieving
fluid, level set, and phase field methods directly give progression of waves without damping and the non-
the indicator functions at each point, but the interface reflecting outflow boundary conditions.
position is embedded in the Eulerian field and is not
explicitly specified. Another important issue of air- 3.3 Motions
water interface modeling is the treatment of the air-
water interface, i.e., is it a transition zone with a finite 3.3.1 Prescribed and predicted ship motions
thickness or a sharp interface with zero thickness?
Different answers determine different mathematical As evident from G2010 test cases, most ship motion
formulations and the numerical methods to the computations are for up to 3 degree of freedom (DoF):
solution. In general, this concerns the variations of roll decay; sinkage and trim or pitch and heave in
physical properties such as density and viscosity across waves; maneuvering trajectories constrained from
the interface. On the other hand, surface tension can pitch, heave and roll; and PMM predicting pitch, heave
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

and roll. There are limited computations for 6DoF (2012) developed a simple and efficient approach for
motions under varied seakeeping and maneuvering strongly coupled fluid-structure interactions using an
conditions. The motions are computed by solving the immersed boundary method developed by Yang and
rigid body dynamics equations due to the forces and Balaras (2006) with great simplification. The fluid-
moments acting on the ship (Fossen, 1994). The forces structure coupling scheme of Yang et al. (2008a) was
and moments are generally obtained by integrating the also significantly expedited by moving the fluid solver
contribution of pressure and viscous forces on the hull. out of the predictor-corrector iterative loop without
This approach is accurate, but its implementation may altering the strong coupling property. This approach
be complicated for immersed boundary and overset can be extended to gas-liquid-solid system similarly to
methods. An alternative approach is to balance linear the method in Yang & Stern (2009) for strongly
and angular momentum over a large control volume coupled simulations of wave-structure interactions.
containing the body. This approach is easier to
implement, but is prone to inaccuracies associated with 3.4 Propulsor modelling
numerical errors.
The influence of motion on the fluid flow Fully discretized rotating propellers have the ability to
governing equations can be either accounted as body provide a complete description of the interaction
forces in the ship system (Sato et al. 1999) or the between a ship hull and its propeller(s), but the
governing equations can be solved in the inertial approach is generally too computationally expensive
coordinates for which the grids move following the (Lübke, 2005). Simplification such as use of single
body (Carrica et al. 2007). For the first approach, the blade with periodic boundary conditions in the
grids do not need to be deformed or moved during the circumferential direction (Tahara, et al., 2005) can help
computation but important features such as the free ease the computational expense, but are still expensive
surface may shift to poor quality grid region. The for general purpose applications. Discretized propellers
second approach, although more expensive than the along with periodic conditions to define the interaction
former, is more appropriate as it allows not only proper between the blades are mostly used for open water
grid resolution during the simulation but also allows propeller simulations.
multi-body simulation. In the second approach,
deformable, regenerated or overset grids should be 3.4.1 Body force and fully discretized propellers
used to move the objects. Grid deformation and
regeneration methods are used mostly for finite volume Most commonly used propulsor model is the body
solvers, and their application is limited to small force method. This approach does not require discreti-
amplitude motions. The dynamic overset grids provide zation of the propeller, but body forces are applied on
huge flexibility in capturing motions and have been propeller location grid points. The body forces are
successfully applied for wide range of problems such defined so that they integrate numerically to the thrust
as broaching, parametric roll, ship-ship interaction to and torque of the propulsor. One of the most common
name a few (Sadat-Hosseini et al., 2011b). techniques is to prescribe an analytic or polynomial
distribution of the body forces. The distributions range
3.3.2 One-field formulations for motion prediction from a constant distribution to complex functions
defining transient, radially and circumferentially
The body domain can be included in the computational varying distribution. Stern et al. (1988b) derived
domain and the whole system can be represented as a axisymmetric body force with axial and tangential
gas-liquid-solid three-phase system, and solved using a components. The radial distribution of forces was
one-field formulation. Although the structural defor- based the Hough and Ordway circulation distribution
mation can be considered by including the structural (Hough and Ordway, 1965) which has zero loading at
constitutive models, rigid body motions are usually the root and tip. More sophisticated methods can use a
adequate for many applications. There is a large body propeller performance code in an interactive fashion
of research for incorporated structural motion predic- with the RANS solver to capture propeller-hull
tion in the flow solvers. Recently, several studies interaction and to distribute the body force according to
discussed monolithic fluid structure interaction on the actual blade loading. Stern et al. (1994) presented a
Cartesian grids (Robinson-Mosher et al, 2011; Gibou viscous-flow method for the computation of propeller-
and Min, 2012). These methods require the modifica- hull interaction in which the RANS method was
tions of the linear systems for consideration of solid coupled with a propeller-performance program in an
motion coupled with fluid motion in a single step. On interactive and iterative manner to predict the ship
the other hand, partitioned approaches allow the wake flow including the propeller effects. The strength
solutions of solid motion and fluid flow using most of the body forces were computed using unsteady
suitable algorithms for each phase. Yang & Stern program PUF-2 (Kerwin et al., 1978) and field point
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

velocity. The unsteady wake field input to PUF-2 was of the rotor, and the total number of grid points is 28
computed by subtracting estimates of the propeller- million. It is expected to see more applications of this
induced velocities from the total velocities calculated type of simple approaches in propulsor modeling.
by the RANS code. The estimates of induced velocities
were confirmed by field point velocity calculations 3.5 Turbulence modelling
done using the circulation from PUF-2. Simonsen and
Stern (2005) used simplified potential theory-based The grid requirements for direct numerical simulation
infinite-bladed propeller model (Yamazaki, 1968) (DNS) of the Navier-Stokes equation for turbulent
coupled with the RANS code to give a model that flows increases with Reynolds number, i.e, O(Re9/4)
interactively determines propeller-hull-rudder interac- (Piomelli and Balaras, 2002). Model scale Re ~ 106 and
tion without requiring detailed modeling of the full scale 109 ship calculations would require 1013 and
propeller geometry. Fully discretized CFD computa- 1020 grid points, respectively. However, the current
tions of propellers in the presence of the ship hull have high performance computing capability allows ~10 9
been performed in several studies. Abdel-Abdel- grid points (Wang et al., 2012d). The alternative is to
Maksoud et al. (1998) used multi-block technique to use turbulence modeling, which has been an important
simulate the rotating propeller blades and shaft behind research topic over the last decades. A large number of
the ship for propeller-hull interaction investigation. models have been proposed, tested and applied, but no
Zhang (2010) simulated the rotating propeller using ‘universal’ model has been developed. In turbulence
sliding mesh technique for the propeller behind a modeling, the turbulent velocity field is decomposed
tanker. Carrica et al. (2010a, 2012b) included the actual into resolved ̂ and fluctuating ( ) scales of motion
propellers in the simulations by using dynamic overset using a suitable filter function (Pope, 2000), which
grid. Muscari et al. (2010) also simulated the real results in an additional turbulent stress term ( ), which
propeller geometry using dynamic overlapping grids can be expressed using a generalized central moment 
approach. as:

3.4.2 Waterjet propulsion (̂ ̂ ) ̂ ̂ ̂. (1)

There is a growing interest in waterjet propulsion The main contribution of the above stresses is to
because it has benefits over conventional screw transfer energy between the resolved and turbulent
propellers such as for shallow draft design, smooth scales. The physics associated with the transfer
engine load, less vibration, lower water borne noise, no depends on the choice of filter function, thus different
appendage drag, better efficiency at high speeds and turbulence modeling approaches focus on different
good maneuverability. The waterjet systems can be aspects.
modeled in CFD by applying axial and vertical reaction The most commonly used turbulence model is
forces and pitching reaction moment, and by represent- the Unsteady Reynolds Averaged Navier-Stokes
ing the waterjet/hull interaction using a vertical stern (URANS) approach. In this approach only the large
force (Kandasamy et al., 2010). Real waterjet flow scales of motion are resolved and the entire turbulence
computations are carried out including optimization for scale is modeled. An emerging approach is Large Eddy
the waterjet inlet by detailed simulation of the duct Simulations (LES) (Hanjalic, 2005; Fureby, 2008). In
flow (Kandasamy et al., 2011). Figure 3 shows the LES the solution relies less on modeling and more on
waterjet flow computation results for the two waterjet numerical methods, and provides more detailed
propelled high-speed ships studied, i.e. JHSS and Delft description of the turbulent flow than URANS. The
catamaran. grid requirements for LES are still large especially in
the near-wall region, and cannot be applied for next
3.4.3 Propulsor modelling on Cartesian grids couple of decades (Spalart, 2009). Hybrid RANS-LES
(HRL) models combines the best of both approaches,
where URANS is used in the boundary layer and LES
Simulations with discretized propellers are increasingly
in the free-shear layer region (Spalart, 2009; Bhushan
becoming common practice in ship hydrodynamics.
and Walters, 2012). Full scale simulations require
Immersed boundary methods can be used for greatly
extremely fine grid resolution near the wall, which
simplified grid generation in this type of applications.
leads to both numerical as well as grid resolution
Posa et al. (2011) performed LES of mixed-flow
issues. Wall-functions are commonly used for full scale
pumps using a direct forcing immersed boundary
to alleviate these limitations, and they also allow the
method and obtained good agreement with experi-
modeling of surface roughness (Bhushan et al., 2009).
mental data. The Reynolds number is , based
on the average inflow velocity and the external radius
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

3.5.1 URANS models are more difficult to implement and often less
robust than the conventional eddy-viscosity models.
In URANS the filter function represents an ensemble For this reason they are far less common than linear
average, which is typically interpreted as an infinite- models, despite potential for increased accuracy. In
time average in stationary flows, a phase-average in G2010, there were limited submissions using such
periodic unsteady flows, and/or averaging along a models, and they reproduced the measured structure of
dimension of statistical homogeneity if one is availa- the turbulence better than linear models (Visonneau,
ble. For such averaging, the entire turbulence spectrum Chapter 3 - G2010 proceedings). Stern et al. (Chapter 7
is modeled and the resolved scales are assumed above - G2010 proceedings) performed calculations for
the inertial subrange. URANS models should account straight ahead 5415 using CFDShip-Iowa V4 on up to
for: (a) appropriate amount of turbulent dissipation; 50M grids using k- based anisotropic (ARS) and
and (b) momentum and energy transfer by turbulent linear model (BKW). ARS showed significantly better
diffusion, which affects flow separation and vortex velocity, turbulent kinetic energy and stress profiles at
generation (Gatski and Jongen, 2000). the nominal wake plane than the linear model, as
The most theoretically accurate approach for shown in Fig. 4. However, the turbulent kinetic energy
URANS is the differential Reynolds stress modeling. and normal stresses were over predicted by 60% even
However, solutions of at least seven additional on 50M grid. Further, the ARS model does not show
equations are expensive. The Reynolds stress equations good predictions for the stress anisotropy.
also tend to be numerically stiff and often suffer from URANS simulations with anisotropic models
lack of robustness. on 10s to 100s million grids are desirable to obtain
At the other extreme lie the linear eddy- benchmark URANS predictions. But improved mean
viscosity models based on Boussinesq hypothesis, vortical and turbulent structure predictions require
which are calibrated to produce an appropriate amount further improvements in the models, such as ability to
of dissipation. These do not account for the stress account for rotation/curvature effects or structure-based
anisotropy as the three-dimensionality of the turbulent non-linear effects (Kassinos, 2006).
diffusion terms is not retained. The linear equation
models have evolved from zero-equation, where eddy 3.5.2 LES
viscosity is computed from the mean flow, to most
successful two equation models, where two additional In LES, the filtering scale is assumed to lie within the
equations are solved to compute the eddy viscosity. inertial subrange, such that the organized coherent
The k- model performs quite well in the boundary turbulent structures are resolved and small-scale quasi-
layer region, and k-  in the free-shear regions. Menter isotropic turbulent fluctuations are modeled. Key
(1994) introduces blended k-/ k- (BKW) model to aspects for LES modeling include: (a) resolution of
take advantage of both these models. This is the most energy transfers between the coherent and fluctuating
commonly used model for ship hydrodynamics turbulent scales, which involves both forward and
community. The one equation SA (Spalart, 2009) backscatter of energy; and (b) the requirement of initial
model solves for only one additional equation of the background fluctuation energy to instigate coherent
eddy viscosity. This model is more common in the turbulence fluctuations via the production term (Batten
aerospace community, probably due to the availability et al., 2004).
of a transition option. The most commonly used LES models are the
An intermediate class of models is the non- eddy-viscosity type model. These models are similar to
linear eddy viscosity or algebraic stress models (ASM). the linear URANS models, except that the length scale
The algebraic models are derived by applying weak- is defined explicitly as the grid size. These models can
equilibrium assumptions to the stress transport only account for the forward transfer of energy, unless
equations, which provides a simplified but implicit dynamic coefficients are used to allow backscatter in
anisotropic stress equation. The solution of the an averaged sense (Lilly, 1992). Backscatter of energy
equations can be obtained by inserting a general form is identified to be a very important aspect for atmos-
of the anisotropy which results in a system of linear pheric flows, which involves both 2D and 3D turbu-
equations for the anisotropy term coefficients. These lence (Kraichnan, 1976). Studies in this community
models have similar computational cost as the linear have incorporated backscatter explicitly via an
models, but provide higher level of physical descrip- additional stochastic forcing term (Schumann, 1995).
tion by retaining many of the features of the Reynolds The second most common class of LES models are the
stress transport equations. Several notable models in variants of the scale-similarity model (Bardina et al.,
this category have been presented (Wallin and 1983), which are developed based on the assumption
Johansson, 2002). It must be noted that algebraic that the flow in the subgrid scale copy the turbulence
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

scales an octave above. These models have been found For example, smaller scale fluctuations required as
to be under dissipative, and are often combined with inlet conditions for LES region are not predicted by the
the eddy viscosity model to obtain nonlinear mixed URANS solution. Several approaches have been
models (Meneveau and Katz, 2000). These models published to artificially introduce small-scale forcing,
have also been extended to include dynamic model either by a backscatter term, isotropic turbulence, or an
coefficient evaluation to account for backscatter in an unsteady coefficient to blend the total stress or
averaged sense (Horiuti, 1997). Another class of model turbulent viscosity across the interface (Batten et al.,
which has gained popularity for applications is the 2004).
Implicit LES (ILES) models, where the numerical Non-zonal approaches can be loosely classi-
dissipation from the 2nd or 3rd order upwind schemes fied as adopting either a grid-based or physics-based
is of the same order as the subgrid-scale dissipation approach to define the transition region. The most
(Boris et al, 1992). common grid-based approach is detached eddy
One of the major issues with the use of LES is simulation (DES). In DES, a single grid system is used
the extremely fine grid requirements in the boundary and the model transitions from URANS to LES and
layer, i.e., the grids have to be almost cubical, whereas vice versa, based on the ratio of URANS to grid length
URANS can accommodate high aspect ratio grids. scale (Spalart and Allmaras, 1992). This approach
Piomelli and Balaras (2002) estimate that grid resolu- provides transition in a simpler manner than the zonal
tion required resolving inner boundary layer (or 10% of approach, and the need for artificial boundary condi-
the boundary layer thickness) requires ~ Re1.8 points tions at the interface is avoided. The DES approach
which gives, 1011 and 1016 points for model and full assumes that: the adjustment of the dissipation allows
scale, respectively. development of the coherent turbulent scales in the
Fureby (2008) reviewed the status of LES LES mode; and that the LES regions have sufficient
models for ship hydrodynamics, and concluded that the resolved turbulence to maintain the same level of
increases in computational power in the past decade are turbulence production across the transition region.
making possible LES of ships, submarines and marine However, these criteria are seldom satisfied and errors
propulsors. However, the LES resolution of the inner manifest as grid/numerical sensitivity issues, e.g., LES
part of the hull boundary layer won’t be possible for convergence to an under dissipated URANS result due
another one- or two-decades. To meet the current to insufficient resolved fluctuations, modeled stress
demand of the accurate predictions of turbulent and depletion in the boundary layer, or delayed separated
vortical structures, modeling efforts should focus on shear layer breakdown (Xing et al., 2010a). Delayed
development/assessment of wall-modeled LES or DES (DDES) models and other variants have been
hybrid RANS-LES models. introduced to avoid the stress depletion issue in the
boundary layer (Shur et al., 2008). But these modifica-
3.5.3 Hybrid RANS-LES tions do not address the inherent limitations of the
method, which is identification of the transition region
From a broad perspective the only theoretical differ- primarily based on grid scale.
ence between the URANS and LES formulations is the Several studies have introduced transition re-
definition of the filter function. HRL models can be gion identification based on flow physics (Menter and
viewed as operating in different “modes” (LES or Egorov, 2010). Girimaji (2006) introduced partially
URANS) in different regions of the flow-field, with averaged Navier-Stokes (PANS) modeling approach
either an interface or transition zone in between. HRL based on the hypothesis that a model should approach
models are judged based on their ability to: (a) blend URANS for large scales and DNS for smaller scales.
URANS and LES regions and (b) maintain accuracy in These models have been applied for various applica-
either mode and in the transition zone (or interface). tions with varying levels of success, but have not
The HRL models available in the literature undergone the same level of validation as LES models.
can be divided into either zonal or non-zonal approach- Hence their predictive capability in pure LES mode
es. In the zonal approach, a suitable grid interface is cannot be accurately ascertained (Sagaut and Deck,
specified to separate the URANS and LES solution 2009). Ideally, a hybrid RANS-LES model should
regions, where typically the former is applied in the readily incorporate advances made in URANS and LES
near wall region and the latter away from the wall community, rather than representing an entirely new
(Piomelli et al., 2003). This approach provides class of model.
flexibility in the choice of URANS and LES models, Recently, Bhushan and Walters (2012) intro-
enabling accurate predictions in either mode duced a dynamic hybrid RANS-LES framework
(Temmerman et al., 2005). However, there are (DHRL), wherein the URANS and LES stresses are
unresolved issues with regard to the specification of the blended as below:
interface location and the coupling of the two modes.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

(2) like, horseshoe vortex, shear layer, flapping and helical

vortex instabilities as summarized in Table 1.
The blending function  is solved to blend the The Karman-like instabilities were observed
turbulent kinetic energy (TKE) production in the for wave induced separation for surface piercing
URANS and LES regions as below: NACA 0024 airfoil, for transom vortex shedding for
wetted transom Athena and transom model flows, due
̅̅̅̅̅̅̅̅̅ ̅ to the interaction of hull and tip vortices in Wigley
̅ ̅̅̅̅̅̅̅ ̅ (3) hull, due to the interaction of bow vortices for
KVLCC2, and interaction of vortices on the leeward
sonar dome. These instabilities are caused by the
The model to operate in a pure LES mode only if the
interaction of two opposite vortices initiated by shear
resolved scale production is equal to or greater than the
layer instability, and are scaled using half wake width
predicted URANS production, otherwise the model
H and shear layer velocity (US). Sigurdson (1995)
behaves in a transitional mode where an additional
reported a universal Strouhal number StH = fH/US range
URANS stress compensates for the reduced LES
of 0.07 – 0.09. For surface-piercing NACA 0024
content. Likewise, in regions of the flow with no
simulation, StH ~ 0.067, and it was found that free-
resolved fluctuations (zero LES content), the SGS
surface reduces both the strength and frequency of the
stress is zero and the model operates in a pure URANS
vortex shedding resulting in lower StH. The ship
mode. The advantage of the DHRL model includes: (a)
geometries show averaged StH ~ 0.088, which is
it provides the flexibility of merging completely
towards the higher end of the expected range.
different URANS and LES formulations; and (b)
Horseshoe vortices were predicted for the ap-
allows a seamless coupling between URANS and LES
pended Athena simulations at rudder-hull, strut-hull
zones by imposing smooth variation of turbulence
and strut-propeller-shaft interactions. Simpson (2001)
production, instead of defining the interface based on
reviewed horseshoe vortex separations, and identified
predefined grid scale.
that they occur at junction flows when a boundary layer
Non-zonal DES approach has been used to
encounters an obstacle. These instabilities are associat-
study the vortical and turbulent structures and associat-
ed with two vortex system, or dual peak in frequency.
ed instability for flows of ship hydrodynamics interest
The secondary peak amplitude decreases with the
on up to large 300M grids using CFDShip-Iowa V4.
increase in the angle of attack and sweep angle. These
Simulations have been performed for surface-piercing
structures are scaled using the thickness of the obstacle
NACA 0024 airfoil (Xing et al., 2007), Wigley hull at
T and largest dominant frequency, and show StT =
 = 45 and 60 (Heredero et al., 2010), wetted
fT/U0 = 0.17 - 0.28. Athena simulations predicted StT =
transom flow for model and full-scale bare hull and
0.146±3.9% at rudder-hull intersection and StT =
appended Athena (Bhushan et al., 2012c), wet and dry
0.053±2% at strut-hull interaction.
transom-model (Drazen et al., 2010), 5415 at straight
Shear layer instabilities, which are associated
ahead conditions, 5415 with bilge keels at  = 20 with the boundary layer separation, were predicted for
(Bhushan et al., 2011a), and KVLCC2 at =0, 12 and free-surface separation and inside the separation bubble
30 (Xing et al., 2012). for surface piercing NACA 0024 studies; boundary
Surface-piercing NACA 0024 airfoil simula- layer separation close to the appendages for appended
tions help study the effect of free-surface on flow Athena; on the leeward side for the static drift cases, in
separation and turbulence structures in the separation particular at hull bow and keel for Wigley hull, at the
region. Wetted transom bare hull and appended Athena bow for KVLCC2, and sonar dome separation bubble
and transom-model simulations help identify the for 5415. Such instability is scaled using boundary
transom free-surface unsteadiness due to the transom layer at separation () and US and shows St =
vortex shedding as shown in Fig. 5. The straight ahead 0.0056±2% for airfoil boundary layer separation
5415 simulation provided a detailed resolution of the (Ripley and Pauley, 1993). For surface-piercing NACA
evolution and interaction of the vortical structures, and 0024, St = 0.00384±0.5% for free-surface separation,
provided a plausible description of the sparse experi- and varied inversely with the non-dimensional adverse
mental data as shown in Figs. 4 and 6. The static drift pressure gradient at separation. The boundary layer
simulations were performed to analyze the flow
separation for appended Athena showed St =
features and guide the ongoing experiments. The
0.0067±3%, and for leeward side flow separation for
vortical structures predicted for KVLCC2 at  = 30
static cases St ~ 0.001 ~ 0.003, and in some cases even
are shown in Fig. 7, and those for 5415 with bilge keels
lower.
at  = 20 including preliminary comparison with Flapping instability was predicted for the free-
experiments in Fig. 8. Studies have shown Karman- surface separation bubble in surface-piercing NACA
0024 simulations, and transom wake for bare hull
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Athena simulations. Such instability occurs when a modeled stress depletion. Recently, Bhushan et al.
recirculation region exhibits a periodic enlargement (2012b) applied DHRL and DES models for straight
and shrinkage, and is scaled using the reattachment ahead 5415 in single phase using commercial software
length XR and free stream velocity U0. For canonical Fluent. The DHRL model was able to trigger resolved
cases, StR =fXR/U0 ~ 0.073 – 0.12 (Kandasamy et al., turbulence, whereas DES failed to do so.
2009). The surface-piercing NACA 0024 and Athena Hybrid RANS-LES simulations on 100s mil-
simulations showed StR = 0.28 and 0.144, respectively. lions to billions of grid points for model-scale are
Static drift simulations show helical vortices. required to enable resolution of small-scale physics,
For the Wigley hull at  = 60 and KVLCC2 at  = 30 improve understanding of turbulence and vortical
such vortices were generated on the leeward side, and structures, two-phase flow and air entrainment. Such
for 5415 at  = 20 from the bilge keel tip and fore- simulations will help in explaining the observation in
body keel. For both KVLCC2 and 5415, the vortex sparse experimental data and guide experiments, and
core frequency decreased downstream such that the StX provide benchmark datasets to develop better URANS
based on the distance from the separation point remains models. However, the existing Hybrid RANS-LES
constant, similar to the tip vortices formed over a delta models have not been previously applied for similar
wing. Overall, for most of the vortices StX ~ 0.095 – simulations, hence detailed verification and validation
1.45 compares well with the Delta wing tip vortex needs to be performed. Further, the available grid
range of 0.75 – 1.35. However further analysis is verification methodologies were developed for
required to confirm the identity of these instability URANS (Stern et al, 2006a; Xing and Stern, 2010),
mechanisms, including comparison with slender and cannot be applied straightforwardly to hybrid
fuselage vortices. RANS-LES due to the coupling of modeling and
The transom flow pattern for dry transom- numerical errors. Thus, new verification methods need
model flow shows shoulder waves emanating from the to be developed.
transom edge, which moves towards the center-plane,
overturns and breaks. A similar breaking wave pattern 3.5.4 Wall-functions
was also predicted for bare hull Athena URANS
simulation (Wilson, 2005). This instability causes The boundary layer thickness decreases with the
unsteady wave elevation pattern in the rooster tail Reynolds number, thus near wall grid resolution (y+ ~
region. Instability mechanism associated with such 1) for full-scale ship computations require very high
unsteadiness has not been identified. grid density. A rough estimate suggests that the number
Fully appended Athena wetted transom flow of grids required in the wall normal direction to resolve
shows unsteady pitch and heave motion, whereas the the inner boundary layer is ~ Re0.6 points, i.e., around
dry transom simulations show steady motions. The 250K grid points in the wall normal direction (Piomelli
motion unsteadiness were attributed to the Karman-like and Balaras, 2002). The extremely fine grid spacing
transom vortex shedding, as both show the same may also lead to numerical issues, such as increases the
dominant frequency. This instability was called errors of computing mass and momentum fluxes in
“vortex-induced-motion” and scaled using ship length high aspect ratio cells. The use of “wall-functions”
L and U0 which resulted in StL = 2.19. avoids the numerical limitations of the near-wall
Studies have shown good predictions for the turbulence model and significantly reduces the
resolved turbulence levels around 80% to 95% for computational cost. In wall-function approach the
NACA 0024, bare hull and appended Athena, and solution in the inner boundary-layer is circumvented
static drift cases, when the flow separation was dictated using flat-plate boundary layer assumptions, i.e., the
by the geometry. However, for the straight ahead 5415 flow is governed by the pressure gradients outside the
case the resolved turbulence was not triggered, which boundary layer and the velocity profile follows the
resulted in stronger, under dissipated vortices. Stern et universal sub- and log-layer. The boundary conditions
al. (Chapter 7 - G2010 proceedings) identified that the are applied at the first grid point away from the wall,
under resolved turbulence is due to the limitations of called matching point. The accuracy of such models
DES model in triggering resolved turbulence, and not depends on their ability to: (a) account for the variation
due to numerical dissipation issues. For the KVLCC2 of the grid resolution y+ on the hull, (b) prediction of
simulations on 305M grid, Xing et al. (2011) observed the flow separation point, and (b) robustness of the
that the model over-predicted the velocity near the implementation.
symmetry plane, Reynolds stresses at the propeller The most commonly used wall-function is the
plane and showed grid induced separation and standard wall-function. This approach is based on the
modeled-stress depletion in the boundary layer. The stringent criteria that the matching point lies in the log-
delayed DES (DDES) version of the model was able to layer (one-layer only). However, variation of the
resolve the induced separation issue, but not the boundary layer thickness along the ship hull makes it
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

difficult to always place the matching point in the log- roughness effects. The obvious limitations of the wall-
layer. This deficiency has been addressed by introduc- functions are in accurately predicting separated flows
ing multi-layer models, where the boundary conditions and 3-D boundary layers with significant cross-flow.
for the velocity and turbulent quantities switch Nevertheless, near-wall turbulence models also suffer
smoothly between the sub- and log-layer profiles from the same deficiency as the model constants are
depending upon the local y+ value (Bhushan et al., derived under similar turbulent boundary layer
2009). Some studies have implemented pressure assumptions (So and Lai, 1988). Multi-layer models
gradient effect in wall-function formulation to improve have performed well for ship flows including re-
predictions for separated flows (Kim and Chaudhury, sistance, propulsion, seakeeping, and maneuvering.
1995). But often the pressure gradient magnitude needs However, further research is required to develop
to be clipped to avoid numerical instability. Thus the improved pressure gradient models for accurate flow
benefit of including pressure gradient effect is ques- separation predictions, and better relation of the
tionable (Kalitzin et al., 2005). downshift of log-law with roughness length.
Implementation of wall-function models re-
quires evaluation of the friction velocity to provide 3.5.5 Two-phase turbulence modeling
boundary conditions for velocity and turbulence
variables. A one-point approach proposed by Kim and In ship hydrodynamics, the wall boundary and the air-
Chaudhary (1995) uses the flow variables at the wall water interface are the two major sources of difficulties
neighboring cells only, and allows solutions of the of resolving turbulence at high Reynolds numbers. The
momentum equations up to the matching point. This former has been the sole theme of many research topics
approach can be implemented easily for finite-volume for many years; the investigations of the latter have
schemes, but introduces additional challenges for been limited to DNS and highly-resolved LES, and
finite-difference schemes. An alternative two-point modeling means like RANS turbulence models for the
approach (Tahara et al., 2002) uses the velocity former, which are more or less mature, though
magnitude and direction at the second grid point away imperfect, are not reached yet. Droplet/bubble-laden
from the wall to obtain the boundary conditions at the turbulent flows are even less understood, especially,
matching point. Implementation of this approach is when interacting with the boundary layer near a solid
straightforward for finite-difference scheme. However, wall. Due to the extremely high computational cost,
the one-point approach is expected to be more accurate DNS is limited to low Re number turbulent flows.
than the two-point approach, as the former does not Some large-eddy simulation (LES) studies (Sreedhar
restrict the flow streamline at the matching point. and Stern, 1998a, b; Broglia et al., 2003) have been
Effect of surface roughness is more important conducted at very low Fr numbers with the air effect
for full-scale computations than for model-scale. The neglected. For two-phase interfacial flows, the eddy
most commonly used model for surface roughness is viscosity is often over-predicted if the single phase
based on downshift of the log-layer profile (Patel, based LES and RANS models are used (Liovic and
1998). Several studies have validated the existence of Lakehal, 2007). Liu et al. (2009) investigated the
downshift of log-law in the transitional roughness coupled air-water turbulent boundary layers using
regime, this provides some confidence in such direct numerical simulations. In Toutant et al. (2009a),
modeling (Jimenez, 2004). However, the amount of the two-phase LES concept was developed at a given
shift based on roughness length is still an area of active level of description that the filter is much smaller than
research. the bubbles/drops. Away from the two-phase mixture
Applicability of wall-function for ship flows region, the single-phase LES concept still applies. In
has been demonstrated by several researches for both general, turbulence modeling of two-phase interfacial
model scale and full scale (Oh and Kang 1992). flows is in its early stage. High-resolution DNS studies
Bhushan et al. (2009) implemented multi-layer wall and detailed experimental measurements are required
function using with wall roughness and pressure for the development, improvement, and validation of
gradient effects using two-point approach in CFDShip- two-phase turbulence modeling techniques. It is
Iowa V4 and performed verification and validation for expected the new models are built on top of the
smooth and rough wall Athena resistance, propulsion corresponding single phase models.
and seakeeping, and 5415 maneuvering simulations.
The results (selected results shown in Fig. 9) were
compared with model scale predictions and with
limited full scale data, for which the predictions were
encouraging.
Wall-functions are a viable option for full-
scale ship simulations and implementation of wall
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

4 NUMERICAL METHODS derived from the level set function. These advantages
have greatly increased its applications in many CFD
4.1 Reference frames fields including ship hydrodynamics (e.g., Carrica et
al., 2007; Yang & Stern, 2009). However, there is no
The governing equations for ship hydrodynamics are volume constraint in the course of level set evolution
the incompressible Navier-Stokes equations which are through the level set advection equation, which makes
solved in an absolute inertial earth-fixed reference the mass conservation a serious issue in level set
frame for resistance, pitch, heave and roll simulation, methods. For example, in CFDShip-Iowa version 5
or a relative inertial reference frame for an arbitrary (Huang et al., 2007), two-phase ship flows were solved
non-deforming control volume involving surge, yaw on multi-block structured grids with the level set
and sway motions (Xing et al., 2008). It is common method for interface capturing. A major issue to extend
practice to have a ship-fixed non-inertial reference this solver to dynamic overset grids was the discontin-
frame for solving the ship motions. uous interface (i.e., the level set function) across the
overlapping grids due to different rates of mass loss on
4.2 Interface tracking/capturing grids of different resolution. In Huang et al. (2012b) a
geometry-based approach was proposed to fix this
In pure Lagrangian, meshless flow solvers, such as issue for overset grids, in which the level set advection
SPH (Oger et al., 2006) and MPS (Moving Particle and reinitialization equations were discretized along
Semi-implicit, Shibata et al., 2009), different fluids are the upwind streamline and level set gradient directions,
represented by particles of different densities. As a respectively. It was essentially an unstructured
result, there is no need to track the interface between approach disregarding the resolution differences
different phases. On the other hand, Lagrangian between overset grids, although the discretization was
particles can also be used for interface tracking in implemented in a finite differences approach.
Eulerian grid-based flow solvers. In this type of Many schemes have been developed for pos-
methods, such as front tracking or point set methods, sible improvements, such as the particle level set
connected or unconnected marker particles are placed methods (Enright et al., 2002; Wang et al., 2009a),
on the interface and moved to new positions according coupled level set and volume of fluid methods
to the local fluid velocity. In theory, they are the most (Sussman & Puckett, 2000; Wang et al., 2009b).
straightforward methods that can provide high Recently, Sussman’s group developed a level set
accuracy. However, rapid topological changes of the method with volume constraint (Wang et al., 2012).
interface may make the operations on the marker Some studies chose different definitions of the level set
particles very tedious and difficult. Mass conservation function, for instance in Olsson & Kreiss (2005) a
during the interface evolution is not explicitly enforced smoothed heaviside function was used with value 0~1
and an indicator function is required to be obtained across the interface at iso-level 0.5, instead of a signed
from the geometric information of the interface for distance function. Although they were called level set
phase identification. methods since a reinitializtion step was still involved,
Many free-surface flow solvers adopted sur- in some sense they are closer to other methods such as
face-fitting methods, in which the grids, structured or phase field, constrained interpolated propagation (CIP,
unstructured, were iteratively updated to conform with Hu and Kashiwagi, 2010), and color function methods
the free surfaces (e.g., Tahara et al., 1996; Starke et al. that define a smoothed transition band between
2010). However, the approach has limitations for large different phases.
free surface deformations, such as for steep or breaking There is a large body of studies on interface
waves; may have singular solution at the transom capturing schemes using the volume fraction as a
corner for wet-dry transition Fr range (Li and Matusi- conservatively advected scalar. Usually their schemes
ak, 2001); and grid deformation is numerically were also named volume of fluid (VOF) methods, the
expensive. Therefore, these methods are more suitable focus of research was on the design of compressive
for steady flow computations with mild waves. advection scheme to reduce the numerical diffusion
Wackers et al. (2011) described three ship flow solvers and restrict the interface represented by the volume
with different interface tracking/capturing schemes, fraction within a narrow band. The algebraic VOF
i.e., surface fitting, level set, and volume of fluid schemes implemented in OpenFOAM are particular
(VOF) methods. representative examples. In Wackers et al. (2011) such
The level set function can be treated as a gen- a VOF scheme was discussed for unstructured grids. A
eral scalar and its advection equation can be solved major problem of these schemes is the blurred inter-
using temporal and spatial discretization schemes face, which requires very high resolution for capturing
similar to those of the fluid flow. Geometrical infor- small-scale interfacial phenomena such as droplets and
mation such as interface normal and curvature is easily bubbles.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

On the other hand, in the geometrical VOF equation for pressure (correction). For steady flow
methods, the phase marker function is directly problems, the SIMPLE-family (SIMPLE, SIMPLER,
advected and a special interface reconstruction step is SIMPLEC) algorithms are predominant in commercial
required due to the sharp jump in the marker function solvers and ship hydrodynamics solvers. For unsteady
across the interface. The evaluation of geometric problems, these algorithms can also be used, but the
information such as interface normal and curvature is SIMPLE-based PISO method is more suitable. In all
not easy due to the discontinuous marker function. these methods, the solutions are advanced in multiple
Therefore, VOF methods combined with a level set iterations or time-steps. The momentum equations are
function can be quite useful for this purpose. Wang et first solved without pressure or with pressure from the
al. (2012a) developed a new VOF method on general previous iteration or time-step. Next, the Poisson
structured grids with a distance function constructed equation for pressure (correction) is solved. Finally, the
from the VOF function, which greatly expands the velocity field is corrected using the new pressure
applicability of the VOF method. Further improve- (correction). This segregation or decoupling of the
ments on VOF methods have been investigated through originally coupled equations often makes the projection
tracking additional information such as the material method-based solutions converge more slowly than the
centroids in the moment of fluid method (Ahn & fully-coupled solvers discussed above. Nonetheless, the
Shashkov, 2009). Recently Sussman’s group coupled majority of contributing CFD codes at the Gothenburg
the level set method with the moment of fluid method 2010 workshop adopted the projection method
(Jemison et al., 2012). It is also possible to couple front (Larsson et al., 2010).
tracking methods with the VOF methods such as in Fully coupled and SIMPLE-family methods
Aulisa et al. (2004). Of course, their methods are discussed above were developed mainly for solving
usually more complicated compared with the level set steady flows. Although techniques such as dual-
and algebraic VOF methods, but their minimized stepping and PISO can be used for unsteady problems,
interface position errors and optimized mass conserva- these schemes are inherently limited in the choices of
tion properties are highly desirable in high-fidelity different numerical schemes for temporal and spatial
simulations of ship hydrodynamics studies with small- discretization. On the other hand, fractional-step
scale interfacial phenomena. methods, also one type of projection methods, are in
general more suitable for time-dependent simulations
4.3 Velocity-pressure coupling and widely used in high-fidelity simulation methodolo-
gies such as DNS and LES. There are two types of
The incompressible Navier-Stokes equations have a fractional-step methods, depending on the collocation
mixed parabolic-elliptic character. For steady flows, of velocity components. The staggered arrangement
the equations are of elliptic type and this property can (MAC grid) is usually called exact projection as the
be used in the solution strategies. That is, the continuity velocity-pressure coupling is tight and the discrete
and momentum equations can be solved in a fully divergence is exactly zero (in practice a small value
coupled form, as implemented in CFX and depending on the solution of the pressure Poisson
PARNASSOS (Hoekestra, 1999), among others. Such equation). However, the staggered variable arrange-
methods are expected to be robust; however, the fully ment makes it inconvenient for general grids and
coupled manner results in very large systems of linear coordinate systems. With approximate projection
equations that are quite expensive to solve. They are methods the exact discrete divergence free condition is
usually called pressure-based methods. In a density- relaxed and cell-centered variable arrangement is
based method, e.g., the artificial compressibility usually used. On the other hand, Dong & Shen (2010)
method, the continuity equation is cast into a form akin developed an unconditionally stable rotational velocity-
to one that is widely used for compressible flows by correction scheme for incompressible flows, which can
adding a first-order time-derivative of pressure. be categorized as an approximate projection method.
Addition of this term leads to a hyperbolic system of They further developed their method in (Dong & Shen,
continuity and momentum equations, which can be 2012) by proposing a time-stepping scheme involving
solved in a coupled manner (Rosenfeld et al., 1991). constant coefficient matrices for phase-field simula-
SURF (Hino et al., 2010) and Tenasi (Briley et al., tions of two-phase incompressible flows with large
2006) ship hydrodynamics solvers use this method. It is density ratios.
usually required to adjust an artificial compressibility
parameter in these methods for achieving good 4.4 Semi-coupled air-water flows
performance of convergence.
Most solvers have adopted a different ap- For implementing the semi-coupled approach, a proper
proach, i.e., the projection method, in which the treatment is required for the boundary and initial
continuity equation is satisfied through a Poisson conditions for air over water (Mousaviraad, 2010). A
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

potential solution is obtained for air over water waves 4.6 Temporal discretization
which have a discontinuity since the tangential velocity
changes sign across the surface. Then a blending The majority of CFD solvers use implicit time-
function is introduced to treat the discontinuity in the marching schemes. Implicit time-marching schemes
potential solution and roughly represent the thin allow one to use much larger time-step size than
viscous layer above the water waves. For irregular explicit time-marching schemes, speeding up numeri-
waves, the same potential solution and blending cal solutions for flows with large characteristic time
function is used to define each elementary wave scales. Implicit time-marching, however, requires
component in the superposition. solutions of system of coupled non-linear equations,
The semi-coupled approach in CFDShip-Iowa which incur computational cost. Explicit time-
V4.5 is used to study the effects of head winds on ship marching, which forces much smaller time-step size, is
forces, moments, motions, and airwake flows for calm rarely used for RANS computations. Time step
water straight ahead, static drift, and dynamic PMM discretization is achieved by using first-order backward
maneuvers of the ONR tumblehome with validations Euler scheme for steady-state cases, and second-order
against wind towing tank experiments (Mousaviraad et schemes such as Crank-Nicolson and three-level
al., 2012). Figure 10 shows examples of the air flow backward schemes for time-accurate solutions (Larsson
field results for static conditions. Computations are et al., 2010). Limited studies have used 4th-order
also carried out for pitch and heave in regular head Runge-Kutta schemes. Studies often use solution
waves and 6DOF motions in irregular waves simulat- relaxation to improve stability of the solution.
ing hurricane CAMILLE (Mousaviraad et al., 2008).
Thermal and concentration transport models are 4.7 Semi-Lagrangian advection schemes
implemented in CFDShip-Iowa version 4.5 (Huang et
al., 2010) to investigate the exhaust plume around ship In general, high Reynolds number turbulent flows are
superstructures. Figure 11 shows exhaust plume of the advection dominated. Therefore, the advection schemes
ONR Tumblehome in an extreme motions condition. used for the momentum equation are critical for the
Complicated vortical structures are observed in air accuracy of simulations. Higher-order advection
including a pair of counter-rotating vortices down- schemes are difficult to be made implicit in time. Thus
stream of the stack for cross-flow, and bended bird- treatments like deferred corrections are quite popular in
plume shape in the symmetry plane and varying arc- engineering CFD solvers, because explicit Eulerian
shape in axial sections both for temperature and NOx advection schemes are subject to the CFL number for
concentration fields. time step restriction. When the flow velocity is high
and/or the grid spacing is small, the CFL restriction
4.5 Spatial discretization due to the time step from the explicit advection
schemes can be prohibitive. Semi-Lagrangian advec-
The discretization of the governing equations are tion schemes have been enjoyed for a long time in the
performed either using finite-volume (FV) or finite- numerical weather prediction community. With these
difference (FD) approach. A survey of the G2010 unconditionally stable schemes, higher-order spa-
submissions shows that the FV approach is more tial/temporal accuracy can be retained and significantly
common in ship hydrodynamics community than FD increased maximum allowable time steps with CFL
approached (Larsson et al., 2010). This is because FV numbers up to 4~5 can be safely used. In Xiu &
approach can be applied for arbitrary polyhedral grid Karniadakis (2001), a high-order semi-Lagrangian
volumes which are easier to generate than curvilinear spectral element method was developed for incom-
structured grids required for FD. However, FV pressible flows. Recently, Wang et al. (2012b)
methods are not suited for implementation of higher developed a semi-Lagrangian advection scheme for the
order schemes and mostly use second-order schemes. VOF method, which are more accurate than the
On the other hand, FD approach allows implementation corresponding Eulerian scheme and much improved
of 3rd-order and 5th-order schemes with ease (Yang & with regard to mass conservation. With the increased
Stern, 2009). The diffusion terms are mostly discre- popularity of adaptive mesh refinement techniques,
tized using central-differencing, whereas upwind- more and more studies of semi-Lagrangian schemes
biased schemes are used for convective terms. Higher- have been conducted. Min and Gibou (2006) developed
order schemes provide better accuracy, yet often at the a second-order projection method using a semi-
expense of additional numerical cost and solution Lagrangian advection scheme for incompressible flows
instability. Numerical methods often include flux- on adaptive Cartesian grids. They also reported a
limiter or slope-limiter designed to suppress unphysical second-order semi-Lagrangian level set method on
oscillations in solutions (Ismail et al., 2010). adaptive grids later (Min and Gibou, 2007). Most
recent studies have been focused on conservative semi-
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Lagrangian advection schemes, such as Lentine et al. polynomials in each element; iii) r-adaption, in which
(2011) and Qiu & Shu (2011). the grid points are allowed to move, but without
changing the grid connectivity, in the solution process
4.8 Grid generation to give better resolution to an interested area; and iv) h-
adaption, in which grids points are added or deleted by
Numerical grids are categorized into Cartesian, changing the grid connectivity. The h-adaption is the
structured and unstructured. Most solvers in ship most popular approach as there are no coupling issues
hydrodynamics community including both research and between different mathematical models and discretiza-
application solvers use unstructured grids wherein the tion schemes and it is not limited by the initial number
fluid region is sub-divided into tetrahedral cells in the of grid points and the difficulty of grid quality control
boundary layer and polyhedral cells elsewhere as in the r-adaption. Overset grid techniques can be
(Marcum, 1995). Octree subdivision, Delaunay point used to obtain h-adaption, but usually the refinement
insertion, and advancing front techniques are used to regions such as the free surface and wake are deter-
create the tetrahedral cells near the wall. These grids mined beforehand instead of adaptively during the
are relatively easy to make for complex geometry, but computations due to the extra computational overhead
it is difficult to control the grid quality (Baker et al., of grid oversetting procedure (Carrica et al., 2007). A
1989). Structured grids use body fitted hexahedral directional adaptive mesh refinement method was
cells, and are quite complicated to generate for developed in Wackers et al. (2012) in ISIS-CFD using
complex geometries (Thompson et al., 1985). Elliptical unstructured hexahedral finite volume meshes.
smoothing algorithms are often used to improve the The solution adaptation processes require the
quality of the grids. These structured grids are used evaluation of adaptation metric to identify the regions
along with multi-block overset grid techniques to ease of interest. The choice of adaptation metric depends on
the grid generation complexity. Multi-block techniques the application. The simplest techniques define the
use topological inter-connections to connect the faces solution adaptation region based on the experience and
of the blocks. This can be done using overset or knowledge of the user, for example wake planes or
overlapping techniques, where the interpolation is wave surfaces. Vorticity is often used for improved
applied across local cell volumes and faces. These vortex-core resolution. Other methods include feature-
interpolation schemes can be applied dynamically to detection where the feature of interest is associated
form transient moving and sliding grids to account for with a flow regime or specific characteristic or
the relative motions of the ship hull and the rotation of property. For example, in turbulent flows the following
the propulsion system and appendages (Noack, 2006; metrics have been used: Q-criterion, λ2, λci, and λ+
Wang and Parthasarthy, 2000). The Cartesian grid (Hunt et al., 1988; Jeong and Hussain, 1995; Horiuti,
methods and upcoming technology are discussed 2003). The Q-criterion uses the rotation rate tensor to
separately. refine the mesh around areas of high rotation. The λ2
metric seeks to capture pressure minima. λci metric
4.9 Solution adaptation refines the mesh based on the velocity gradient tensor.
Finally, the λ+ metric works by calculating the
Solution adaptation in CFD is the process in which the strainrate-vorticity correlation. Both h- and r-adaption
computational approaches are modified during the methods as well as hybrid approaches have been used
solution process of a given problem, in order to obtain successfully for improved flow feature detection, local
a more accurate description of the fluid flow or reduction in discretization error, and convergence to an
maintain a similar order of accuracy but with reduced optimal grid.
computational cost. There are four types of solution
adaptation strategies: i) m-adaption, in which different 4.10 Cartesian grid methods
mathematical models are used within different portions
of the computational domain, for example, the coupling Several Cartesian grid methods were presented for ship
of a potential solver and a viscous solver for ship wave problems in the literature. Miyata et al. (1985b)
flows, the coupling of an Euler solver and a Navier- developed a modified Marker and Cell method
Stokes solver for aerodynamic flows, also, the coupling (TUMMAC) for the finite difference solution of non-
of RANS models within the boundary layer and LES in linear wave generation in the near field of ships. A
the bulk flow in a hybrid RANS/LES simulation; ii) p- Cartesian grid approach was presented in (Sussman and
adaption, in which discretization schemes of different Dommermuth, 2000) with a coupled level-set/volume-
orders are used in different portions of the computa- of-fluid method for interface capturing and an embed-
tional domain, for instance, in finite element methods ded boundary method for the immersed geometry.
the p-adaptive strategy is frequently used with different They also reported another approach using immersed-
shape functions represented by different orders of body and volume-of-fluid methods and its recent
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

development was given in Dommermuth et al. (2007) On the other hand, usually it is not necessary
for ship wave simulations. Hu et al. (2010) developed a to pursue such level of universality, e.g., using
CIP based Cartesian grid method for numerical unstructured grids with all types of grid elements, for
simulation of strongly nonlinear wave-ship interaction. ship hydrodynamics applications, since the geometries
Yang & Stern (2009) developed a coupled immersed- are more or less confined to hull forms, appendages,
boundary/level-set method for wave-body interactions. and propellers of a limited range of variations. Once
A second-order sharp interface immersed boundary the surface topology of a geometrical object is
method was used for two-phase flows with multiple determined, a nearly-optimized surface decomposition
moving bodies on fixed Cartesian grids. A ghost-fluid can be obtained and the corresponding surface
method is used without smearing the density across the quadrilateral grid can be generated with usually good
fluid-fluid interface. This method was applied in the orthogonality, not to mention that most shapes have
study of ship model DTMB 5512 at Fr=0.41 in Yang et been given as parameterized solid modeling surfaces
al. (2008b). The instantaneous air-water interface which are readily dividable in u - v coordinates during
colored by the elevation is shown in Fig. 12. The the design process. This also explains why structured
breaking bow waves and scars induced by them are grids are still predominately used in computational ship
evident. Figure 13 shows the instantaneous vortical hydrodynamics.
structures colored by streamwise vorticity in the air In addition, high-order schemes can be easily
flow. There are many Cartesian grid methods for other implemented in structured grids. For example, in Suh
areas of applications, the reader is referred to Yang & et al. (2011) and Wang et al. (2012c), the Cartesian
Stern (2009) for related discussions and references grid solver CFDShip-Iowa V6.1 was extended into an
therein. orthogonal curvilinear solver (V6.2) with similar
On the other hand, the application of Cartesian accuracy and efficiency, where turbulent boundary
grid methods can become practically prohibitive for layers and small droplets/bubbles are resolved at once.
high Reynolds number flows due to the demanding Numerical methods and HPC components in V6.2 are
resolution requirement near the wall boundary, since about the same as in V6.1. It uses a recently developed
most of Cartesian grid methods approximate the novel volume-of-fluid method for general structured
velocity field near the wall boundary by a linear grids with a constructed level set function (Wang et al.,
distribution, which is only correct in the viscous 2012a), thus it can be used for detailed simulations of
sublayer. For example, in order to resolve the viscous interfacial phenomena such as wave breaking and
sublayer for a turbulent flow case at Re = 10 6, the near- wave-body interactions with simple geometries like
wall grid spacing has to be in the range of 10-5L (L is cylinder, sphere, wedge, foil, etc. The orthogonal
the reference length), and the total number of grid curvilinear grid solver has been used for studying the
points will be more than a few billions. Therefore, high Reynolds number, high Froude number turbulent
some forms of wall-layer models have to be included flow past an interface-piercing circular cylinder and a
such that the effects of boundary layer can be taken wedge-shaped bow. In addition, an overset grid solver,
into account properly. CFDShip-Iowa V6.2.5, was presented in Bhushan et al.
(2011b), which couples the Cartesain grid solver and
4.11 High-fidelity curvilinear grid solvers the orthogonal curvilinear grid solver via overset grid
package SUGGAR (Noack, 2006). Based on the
Cartesian grids have some desirable properties not Cartesian grid solver and the orthogonal curvilinear
available in other types of grids, especially, triangles grid solver, a general multi-block structured grid solver
(2D) and tetrahedra (3D). For example, the errors (V6.3) is currently being developed (Yang et al., 2012).
introduced in the discretization of diffusion terms can It maintains and extends the higher-order advection and
be cancelled out completely at two opposite cell faces; volume conservative interface tracking schemes, and
higher order schemes can be implemented much easier; enhanced capabilities of handling complex geometries
also, automatic generation of Cartesian grids with local in ship hydrodynamics.
refinement is much simpler than that of triangles or One additional complication besides the air-
tetrahedral. Therefore, it is preferred to use Cartesian water interface and high-Re boundary layer that has to
grids wherever possible and use other types of grids be considered is the motions of a ship and its control
elsewhere. For instance, Karman & Wilson (2008) surfaces and propellers (when discretized propellers
developed an automatic grid generation algorithm instead of models are used). Multi-blocked structured
based on octree refinement Cartesian grid with general grids, even with moving grid and/or sliding interface
cutting allowing viscous boundary layer grid genera- capabilities, are usually considered to be too restricted
tion. Their development was implemented in Tenasi, a for large displacement/angle motions, especially, when
general purpose unstructured grid solver for a wide ship-ship interactions are to be addressed. Theoretical-
range of applications. ly, unstructured grids with grid deformation and
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

regeneration capabilities can be applicable, but the 5.2 HPC in computational ship hydrodynamics
computational cost of grid operation could be high,
especially, for running on tens/hundreds or thousands Ship hydrodynamics community mostly uses MPI
processes, and the grid quality and algorithm robust- based domain decomposition, and the HPC efforts are
ness are concerns too. The overset grid techniques have geared to either run production jobs faster (strong
been very successful in ship flow computations (e.g., scalability) or to improve the capability of running
Carrica et al., 2007); it is expected to continue playing larger jobs that can resolve more physics with less
an important role in future high fidelity ship hydrody- reliance on modeling (weak scalability). Scalability
namics solvers due to its capability and flexibility. studies of free surface CFD codes are scarce and are
Therefore, it is envisioned that the next- dependent on hardware, thus conclusions are difficult
generation, high-fidelity ship hydrodynamics solvers to reach. Overall, incompressible flow codes show
will be based on Cartesian grids with local/adaptive limited strong scalability, i.e., speed up is typically 60
refinement capabilities for the bulk flow and immersed to 80% of the ideal scalability for 1000 processors
boundary wall modeling techniques or body-fitted (Kremenetsky 2008, Bhushan et al. 2011a). On the
multi-block structured grids for the high-Re boundary other hand, compressible solvers show linear scalabil-
layers. For the latter, overset grid techniques with ity for up to 10’s thousands of processors (Gicquel et
improved conservation properties will facilitate grid al. 2008). The poor scalability of the incompressible
connection and movement. solvers is due to the solution of pressure Poisson
equation. Weak scalability is usually more easily
5 HIGH-PERFORMANCE COMPUTING achieved and has been the focus of most recent
developments in ship hydrodynamics community.
5.1 HPC advances Static ship computations of hundreds of mil-
lions of grid points have been reported for curvilinear
Advances in HPC capability, i.e., as the computational and Cartesian grid solvers (Bhushan et al. 2011a),
capability doubles every two years as per Moore’s law, while dynamic moving computations up to 70 million
are enabling the use of larger grids and more proces- grid points were performed (Carrica et al. 2010). These
sors. However, as the grid size and number of proces- computations enable a degree of detail in the flow
sors increase, so does the communication time. Further, physics that cannot be achieved with coarser grids,
HPC hardware architecture is undergoing fundamental allowing the use of more accurate turbulence models
changes. Today’s systems are hybrid clusters of like hybrid RANS-LES and LES. Computations
multicore processors that have shared and distributed including ship motions are harder and limited by the
memory inside and across the nodes, respectively. Thus need of re-gridding or computation of overset connec-
a parallel algorithm should effectively utilize the tivity. Computations with motions are routine for grids
memory of such systems to achieve better performance ranging between 5 and 25 million grid points. New
(Shalf, 2009). Most ship codes use message passing promising numerical techniques and hardware
interface (MPI) parallelization, which has limitations technologies are rapidly changing the landscape of
on the hybrid networks as the processes running on high-performance computing. Super-scalable Cartesian
different nodes dictate the communication time grid solvers are breaking the 1 billion grid point limit
(Rabenseifner and Wellein, 2003). Hybrid paralleliza- with distributed memory platforms (Wang et al.
tion is well suited for hybrid networks, such as the MPI 2012d), and soon 10 and 100 billion grid points will be
for the internode parallelization and open multi- possible. The HPC challenges of such solvers are
processing (OpenMP) for the intranode parallellization. discussed below.
However, the OpenMP implementation overhead often
outweighs the reduction of communication time, but 5.3 HPC challenges
may be beneficial for codes with substantial collective
communication on high-latency inter-node connection In general, system memory, interconnection, and I/O
(Kaushik et al., 2009). Upcoming graphics processing are the major bottlenecks in high-fidelity flow simula-
units (GPUs) can achieve up to 10 times better tions. The current norm for system memory is about
performance than the central processing units (CPUs) 2GB/core with around 10% reserved for system usage.
and are being applied for CFD (Cohen and Molemaker, This imposes a hard limit with regard to the degrees of
2009). However, they are not yet ready for production freedom one core can effectively handle and thus the
level applications, as multi-GPU and double precision minimum number of cores required for a problem of a
calculations are still in the developmental stage. given size. In many solvers, some data proportional to
the global size of the problem are used in some stages
of the solution procedure, which further exacerbates
the limit. For instance, many solvers use the root
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

process to read some initial data such as input infor- nization operations. Semi-Lagrangian advection
mation and grid data, and then distribute them to other schemes could be much advantageous in this regard
processes. These data may be of large size and not because they usually only involve localized operations.
containable within the memory capacity of one Parallel I/O is another major bottleneck of
computer node. Current computer nodes are usually high-fidelity simulations. For simulations with billions
equipped with tens of processor cores, ranging from 8 of degrees of freedom, it is impossible to use old-
to 64. It is expected that a transition from multi-core to fashioned modes that one or a few nodes read/write all
many-core will be seen in more and more HPC the files. Similarly, with thousands of nodes, or
platforms. The oncoming exascale systems may be millions of cores, the approach in which each
equipped with nodes containing hundreds or even node/core reads/writes its own data files is not
thousands of cores. It is unlikely the system memory manageable. Even the current MPI I/O is severely
available to each core will be increased in a scalable limited in many systems, a major I/O operation such as
way, and actually the decrease of memory size per core writing solution and restart data files using hun-
is more likely to happen due to cost consideration. dreds/thousands of cores may cost the amount of CPU
Therefore, it is critical to minimize memory usage in time for running tens or even hundreds of time steps.
high-fidelity solvers for being able to run them on For high-fidelity simulations requiring frequent data
tomorrow’s HPC platforms. For this point of view, output this could be a major scalability problem, which
Cartesian grid solvers have some particular advantages. could be worse on larger systems as the network and
In contrast to curvilinear structured grids or unstruc- I/O bandwidths become more congested. Related data
tured grids, only one-dimensional arrays are required post-processing for high-fidelity simulations also poses
for grid coordinates and Jacobian matrix information. major difficulties. Data visualization and analysis has
Also, discretization stencils are far smaller if the same to be done on-site or on-the-fly. Data analysis and
order-of-accuracy is sought. For example, second-order reduction within the solvers will become essential
finite difference discretization of the Lapacian components for mitigating network and I/O bandwidth
operation gives a seven-point stencil on Cartesian requirement of the solvers. Again, compared with
grids, but nineteen-point stencil on general curvilinear solvers mainly using curvilinear and unstructured grids,
grids. On the other hand, pre-processing steps such as Cartesian grid methods or solvers using mostly
grid generation could become a major constraint to Cartesian grids have some advantages in terms of I/O
solvers for solving billions of degrees of freedom using when there are changes in grids such as movement,
curvilinear grids or unstructured grids. With the current deformation, and adaption.
mainstream grid generation techniques, usually done
on a workstation with interactive human input, it is 6 V&V PROCEDURES
very difficult to produce billion-point grids. Recently
Wang et al. (2012d) performed high-fidelity simula- 6.1 Background
tions of plunging breaking waves behind a bump using
2.2 billion grid points. It was possible to generate a 2D In spite of the ever-increasing need and importance for
grid using a commercial grid generator and the 3D grid standards for CFD uncertainty analysis/accuracy
was obtained through translational extrusion, thanks to estimation and code certification for industrial
the simple geometry. applications, there are currently many viewpoints
Interconnection performance between com- covering all aspects from basic concepts and defini-
puter nodes is constrained by network bandwidth and tions to detailed methodology and procedures.
communication latency. Because high-fidelity simula- Verification and validation are processes to estimate
tions may involve millions of parallel tasks, collective the numerical and modeling errors, respectively.
and synchronization operations, which are usually Although the definition of validation is fairly widely
unavoidable in many solvers, can become the deter- accepted (ASME Performance Test Codes Committee
mining factor for the overall algorithm scalability. It is PTC 61, 2009), many opinions on verification are
necessary to develop algorithms that can greatly reduce controversial. Herein, we focus on quantitative
communication. Currently, there is a clear trend to verification and validation (V&V) methodology and
change from pure MPI programming mode to hybrid procedures (Stern et al., 2006a), which are essential
MPI/OpenMP programming due to the transition of ingredients for the successful implementation of SBD.
multi-core CPU toward many-core CPU. Within one Different verification methodologies will be reviewed
node, the OpenMP mode can avoid MPI communica- and their shortcomings and criticisms will be dis-
tion latency, reduce request of network bandwidth, and cussed.
reduce system memory usage involved in storing
additional ghost cells in pure MPI mode. On the other
hand, implicit schemes usually require many synchro-
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

6.2 Overview of V&V U SN . The most important numerical errors and

uncertainties are due to use of iterative solution
The approximation used in numerical simulations will methods and specification of various input parameters
result in error  S , which is the difference between a such as spatial and time-step sizes and other parameters
simulation value S and the truth T . However, the true (e.g., artificial dissipation). The errors and uncertainties
values of simulation quantities are rarely known. Thus, are highly dependent on the specific application
errors must be estimated. An uncertainty U is an (geometry and conditions). The errors due to specifica-
estimate of an error such that the interval U contains tion of input parameters are decomposed into error
the true value of  S 95 times out of 100, i.e., at the contributions from iteration number  I , grid size  G ,
95% confidence level. An uncertainty interval thus time step  T , and other parameters  P , which gives
indicates the range of likely magnitudes of  S but no the following expressions for the simulation numerical
information about its sign. error and uncertainty.
J
Sources of errors and uncertainties in
simulation results can be divided into two distinct
 SN   I   G  T   P    j (6)
j 1
sources: modeling and numerical. Modeling errors and J
uncertainties are due to assumptions and approxima- 2
U SN  U I2  U G2  UT2  U P2  U I2  U 2j (7)
tions in the mathematical representation of the physical j 1

problem (such as geometry, mathematical equation, Validation is a process for assessing simulation
coordinate transformation, boundary conditions, air- modeling uncertainty U SM by using benchmark
water interface, and turbulence models) and incorpora- experimental data D and, when conditions permit,
tion of previous data (such as fluid properties) into the estimating the sign and magnitude of the modeling
model. Numerical errors and uncertainties are due to
error  SM itself.
numerical solution of the mathematical equations (such
as discretization, artificial dissipation, incomplete
iterative and grid convergence, lack of conservation of E  D  S   D  ( SM   SN ) (8)
mass, momentum, and energy, internal and external U  U U
2
V
2
D
2
SN (9)
boundary non-continuity, and computer round-off). In
considering the development and execution of a CFD
where E is the comparison error,  D  D  T is the
code, it is assumed that  S is composed of additive
difference between an experimental data and the truth
modeling and numerical errors.
and UV is the validation uncertainty. When
 S  S  T   SM   SN (4)
| E | UV (10)
The simulation uncertainty equation follows directly by
considering equation (4) as a data reduction equation, the combination of all the errors in D and S is smaller
as per EFD uncertainty analysis than UV and validation is achieved at the UV interval.
If UV E , the sign and magnitude of E   SM can
U S2  U SM
2
 U SN
2
(5) be used to make modeling improvements..

where U S is the uncertainty in the simulation and U SM 6.3 Factor of safety method for solution verifica-
and U SN are the simulation modeling and numerical tion
uncertainties. It should be noted that correlations Xing and Stern (2010) proposed four steps for solution
between modeling and numerical errors are also verification: (a) convergence studies with four
possible and should be considered in the future. possibilities: monotonic convergence, oscillatory
There are two types of verification. Code veri- convergence; monotonic divergence, and oscillatory
fication is a procedure to find coding mistakes that
divergence; (b) error estimate  SN with magnitude and
affect the numerical discretization using, for example,
the method of manufactured solutions (MMS) (Roache, sign; (c) uncertainty estimate U that indicates the
2002; Knupp and Salari, 2003). Solution verification is range of likely magnitudes of  SN , but no information
a process for assessing simulation numerical uncertain- about its sign; and (d) statistical analysis to establish
ties U SN and, when conditions permit, estimating the that the interval of U at a 95% confidence level
sign and magnitude  SN of the simulation numerical bounds the comparison error E .
error itself and the uncertainties in that error estimate
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Most solution verification methods are de- pRE  pth are unreasonably small in comparison to
rived for monotonic convergence when the general those with the same distance to the asymptotic range
Richardson extrapolation (RE) method can be used to for pRE  pth . This is due to the fact that the error
estimate the observed order of accuracy pRE , error  RE ,
estimate  RE for the former is much smaller than that
and the numerical benchmark SC using three systemat- of the latter. The second is that there is no statistical
ically refined grids. The uncertainty U is defined as an evidence for what confidence level the GCI and CF
estimate of an error such that the interval of U , U , methods can actually achieve.
bounds the true value of  SN at a specified level of Two other recent studies (Eça and Hoekstra,
confidence, which is usually 95% for experimental 2006; Rumsey and Thomas, 2008) considered the use
fluid dynamics and CFD. Uncertainty estimates can be of different uncertainty estimates for different ranges of
written in the general form of pRE for solutions that show monotonic convergence.
These two verification methods were demonstrated for
 21 a manufactured solution (Eça and Hoekstra, 2006) and
U  FS (11) the flow over a backward facing step (Eça and
r p 1
Hoekstra, 2006; Rumsey and Thomas, 2008) without
detailed derivation and validation. Statistical samples
Various solution verification methods differ in the or analyses were not reported in either of these studies.
choice of factor of safety FS and p. The GCI derived Recently, Xing and Stern developed the factor
by Roache (1998) is currently used and recommended of safety method (Xing and Stern, 2010). It removes
by ASME (Celik, 2008) and AIAA (Cosner et al., the two deficiencies previously discussed for the GCI
2006). However, as discussed by Xing and Stern and CF methods. The best error estimate is used to
(2010; 2011), there are different variants of the GCI construct the uncertainty, which is
method such as the original GCI, GCI1, GCI2, GCI3,
and GCIOR, etc. Thus the choice of FS and p in the   FS1 P  FS0 1  P   RE
 0  P 1
U FS    (14)
GCI method requires user judgment calls, for which no   FS1 P  FS2  P  1   RE
 P 1
single guideline is currently available. The uncertainty
estimate for the original GCI is
FS0  2.45 , FS1  1.6 , and FS2  14.8 are recom-
mended based on statistical analysis.
 21
U GCI  FS  1.25 |  RE | (12)
r pRE  1 6.4 Discussion of the factor of safety method

The correction factor method (Stern et al., The factor of safety method is the only solution
2001; Wilson et al., 2004) uses a variable factor of verification method that was validated using statistical
safety and was validated for correction factor less than analysis, which consists of 25 samples with different
one using a few analytical benchmarks. The factor of sizes based on 17 studies covering fluids, thermal, and
safety for correction factor larger than one is obtained structure disciplines. Only the factor of safety method,
by assuming that the factor of safety is symmetric with compared with the GCI and correction factor methods,
respect to the asymptotic range where the observed provides a reliability larger than 95% and a lower
order of accuracy pRE is equal to the theoretical order confidence limit greater than or equal to 1.2 at the 95%
confidence level for the true mean of the parent
of accuracy of the numerical pth ,
scheme population of the actual factor of safety. This conclu-
P  pRE pth  1 . The uncertainty for the correction sion is true for different studies, variables, ranges of P
factor method is estimated by the sum of the absolute values, and single P values where multiple actual
value of the improved error estimate CF |  RE | and the factors of safety are available. The number of samples
absolute value of the amount of the correction. is large and the range of P values is wide such that the
factor of safety method is also valid for other applica-
tions including results not in the asymptotic range,
 9.6 1  CF 2  1.1  0.875  CF  1.125

U CF     RE which is typical in industrial and fluid engineering
  2 1  CF  1  RE 0  CF  0.875 or CF  1.125 applications. The factor of safety method has been used
to estimate CFD uncertainties in ship hydrodynamics
(13)
such as the recent CFD Workshop Gothenburg 2010
(Larsson et al., 2010).
The GCI and CF methods have two deficien-
Roache presented ten items of discussion of
cies. The first is that the uncertainty estimates for
the factor of safety method (Roache, 2011), which
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

were responded item-by-item by Xing and Stern latest workshop on ship hydrodynamics, Gothenburg
(2011). Additionally, Xing and Stern (2011) evaluated 2010, a total of 89 submissions of resistance prediction
two new variants of the GCI method including GCI OR are documented, which is the largest number in the
(Oberkampf and Roy, 2010) and GCI3 methods and workshop series and the analysis of the results shows
one new variant of the factor of safety method (FS1 that the statistical variance of all the predictions are
method). Except the original GCI method, all variants substantially smaller than the previous workshops in
of the GCI method have jumps of FS versus P, which 2000 and 2005 (2.1% in 2010 compared to 4.7% in
cannot be justified. The FS1 method is the same as the 2005) (Larsson & Zou, 2010). An overview of the
FS method for P  1 but uses pth instead of pRE in computational methods including modeling, numerical
methods, and HPC, and the applications studied are
the error estimate for P  1 . The FS1 method may have provided herein, followed by a detailed analysis of
an advantage for uncertainty estimates when P  2 V&V efforts.
where the FS and other verification methods likely The dominate modeling approaches for re-
predict unreasonably small uncertainties due to small sistance simulations are RANS for turbulence, while
error estimates. However, since the current dataset is some simulations are based on LES or DES, surface
restricted to P  2 , the pros/cons of using the FS or capturing methods such as VOF and level-set for free
FS1 method cannot be validated. Thus, until additional surface, and unstructured or structured multi-block or
data is available for P  2 , all verification methods overset grids (Resistance Committee, ITTC 2011;
should be used with caution for such conditions and, if Larsson & Zou, 2010). Two equation based, isotropic
possible, additional grid-triplet studies conducted to eddy viscosity turbulence models are most popular,
obtain P  2 . especially the family of k-ω models. Wall functions for
turbulence modeling are used in some studies, while
7 V&V FOR CAPTIVE SIMULATIONS many simulations resolve the near wall region includ-
ing the viscous sublayer. The majority of the numerical
Captive testing in the towing tank is the standard methods are based on second order schemes, while
approach in ship design. The CFD ship hydrodynamic higher order discretization schemes are seldom used.
efforts started with and mostly focused on captive Gridding techniques include adaptive mesh refinement
simulations both to replace the towing tank model and overset grids for moving-body and multi-body
testing and to carry out validations against EFD data. applications as well as local refinement by embedding
The CFD community has developed new procedures blocks of finer grids. The HPC methods have been
for single run captive computations for resistance and vastly used in most recent simulations enabling parallel
for seakeeping in head waves. The resistance procedure computing for increased grid size of a few to tens of
is not possible or highly difficult using a physical million points to improve the spatial resolution of
towing tank suggesting a potential of using CFD to aid turbulent boundary layer and wake.
the design process (Xing et al., 2008). The single run The resistance prediction simulations are car-
seakeeping procedure is also applicable for experi- ried out for a wide range of applications and condi-
ments or potential flow and can significantly reduce the tions. Other than drag, sinkage and trim, local flow
costs in all methods (Mousaviraad et al., 2010a). Full- fields such as boundary layer and wake, and wave
scale computations are carried out with CFD which are patterns are also predicted by many simulations.
rarely done in experiments. Appendages are easily Different geometries including tankers, container ships,
added/removed in CFD simulations and can be used to surface combatants, and small vessels are studied at a
study the interactive effects between different append- range of very small to large Froude numbers. Uncon-
ages and the hull. CFD simulations are carried out for ventional ships such as multi hulls, planing hulls, and
unconventional ships and high speed crafts including new concept hulls are included, e.g. Bhushan et al.
real waterjet flow computations. (2012b) simulations for a SES/ACV ship in deep and
This section focuses on CFD captive V&V ef- shallow water (Fig. 14). Fully appended ships with
forts including resistance, sinkage and trim in calm rudders, bilge keels, shafts, struts, and propellers are
water, seakeeping in regular head waves, captive also included in the literature. Full-scale resistance
maneuvering including static and dynamic simulations, predictions including the effects of surface roughness
and stability studies. have become of interest in the recent years (e.g.
Bhushan et al., 2012c). Mousaviraad et al. (2012)
7.1 Resistance and propulsion included the effects of hurricane strength head winds in
their CFD simulations.
Prediction of resistance is the oldest application of Innovative methods are introduced in CFD
CFD in ship hydrodynamics and its accuracy has been studies including a procedure to obtain resistance and
significantly improved over the last 20 years. In the propulsion curves for a wide range of velocities in a
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

single run (Xing et al., 2008). The method is based on are provided, along with a detailed analysis of the
solving the fluid flow equations using an inertial earth- V&V results.
fixed reference frame, and ramping up the ship speed The majority of the simulations are URANS,
slowly such that the time derivatives become negligible while limited simulations are based on LES or DES.
and the local solution corresponds to a quasi steady- Surface capturing methods such as VOF and level-set
state. Fig. 15 shows the resistance and propulsion are dominant for free surface. Incoming waves are
curves for Athena. mainly assumed linear and are imposed at the domain
Self-propulsion computations with rotating boundaries. Structured or unstructured multi-block or
propellers to find the propeller rotational speed (RPS) overset grids are used for motions. Numerical methods
for a given ship speed are carried out recently. Lubke mainly use second order discretization schemes for
(2005) used prescribed propeller RPS and ship speed spatial and temporal terms. The HPC methods are used
for self-propulsion computations of KCS. Carrica et al. in almost all simulations allowing for small grid sizes
(2010a) used a controller to obtain the self-propulsion at the free surface and boundary layer and small time
model point for the KVLCC1 tanker and the ONR steps to capture the motions in waves accurately.
Tumblehome surface combatant free to sink and trim, The applications for seakeeping predictions
and for self-propulsion at full scale for the KCS include a wide range of ship types and geometries,
containership at even keel. Carrica et al. (2010c) wave conditions, Froude numbers, and motion
performed self-propulsion point computations of the restrictions. Grid sizes ranging from 0.4 M to 71 M
KCS containership in model scale free to sink and trim points are used with a clear trend toward increasing
with the rotating discretized propeller. A methodology accuracy with grid size. In G2010 computations were
to compute most of the self-propulsion factors using a contributed by five groups for KCS pitch and heave in
simple prescribed body force model for the propeller is regular head waves under three different conditions
presented by Fu et al. (2010), including a method to using FreSCo, Comet, Open Foam, Wisdam and
obtain the advance velocity at the self-propulsion point CFDShip-Iowa. Five groups also contributed for
and application to the KCS containership at full-scale. KVLCC2 using Comet, Open Foam, CFDShip-Iowa,
CFD calm water V&V results are summarized Isis, Icare and RIAM-CMEN. A case for pitch and
in Table 2. The overall average errors are 3.3% for heave fee to surge was include in G2010 for KVLCC2
resistance, 22.2% for sinkage, and 32.8% for trim. For with three wavelengths with two contributions using
motions, the errors are larger for lower Fr. This could CFDShip-Iowa and Comet. Seakeeping for side by side
be both due to the measurement uncertainties at low ship-ship interactions for regular head and oblique
speed model test and small absolute D values. The waves are carried out, as shown in Fig. 16, for Hope
average errors for Fr<0.2 are 34.7% for sinkage and and Bobo geometries (Mousaviraad et al., 2011).
54.7% for trim, while for Fr≥0.2 the errors are as small Since pitch and heave computations of RAOs
as 9.7% for sinkage and 11% for trim. It must be noted can be expensive due to the large number of runs
that some studies (e.g. Xing et al., 2008 and Sadat- needed for every Froude number at different encounter
Hosseini et al., 2010) used dynamic range of sinkage frequencies, Mousaviraad et al. (2010a) developed an
and trim to evaluate errors, resulting in smaller errors innovative procedure to compute the RAOs for a
for lower Fr. For the limited verification studies the Froude number in a single run. A harmonic wave group
average USN=2.5%, 6.9% and 6.7%S for resistance, single run procedure is developed and validated for
sinkage and trim, respectively, and the average URANS, although the procedure can be implemented
validation uncertainty levels are 2.9%, 14.1% and using experiments or potential flow. Incoming waves
8.1%D. The average errors are comparable to average are deterministic wave groups defined by linear
validation uncertainty levels. superposition of a number of component waves. While
the regular wave procedure requires multiple runs, the
7.2 Seakeeping proposed single run procedure obtains the response
amplitude operators (RAO) for a range of frequencies
CFD computations of seakeeping has been rapidly at a fixed speed, assuming linear ship response. This
increasing since 2005 when the seakeeping committee greatly reduces the computational time and expenses,
of ITTC stated “seakeeping computations are still far while the results are shown for the presented cases to
from a state of mature engineering science.” While have the same order of errors as the standard single
there was only one forward speed diffraction case wave methods if the response is linear.
which involves no motions in the Tokyo 2005 CFD Following G2010, assessment of CFD predic-
Workshop, several heave and pitch in regular head tions for seakeeping in regular head waves separate
waves were included in the G2010 Workshop with capability for 1st order vs. 2nd order terms (Stern et al.,
numerous contributions for each case. Herein, an 2010). Both steady calm water resistance, sinkage, and
overview of computational methods and applications trim and unsteady head waves resistance, heave and
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

pitch are analyzed to include added resistance studies. Most of the maneuvering simulations are
For calm water, resistance is considered 1st order, URANS using 1- and 2-equation isotropic/anisotropic
whereas sinkage and trim 2nd order. Resistance 0th models and Reynolds-stress transport model for
harmonic is 1st order and 0th harmonic heave and pitch turbulence modeling. Free surface is mostly modeled
are second order. 1st harmonic heave and pitch are by a surface capturing method (e.g., level-set, volume
considered 1st order while 1st harmonic resistance is 2nd of fluid), while a few simulations use surface tracking
order. approaches. For numerical methods, spatial discretiza-
CFD seakeeping V&V results for heave and tion is done by finite difference and finite volume
pitch in regular head waves are summarized in Table 3. methods with structured/unstructured grids. The order
The average of the fine grid points is 15 M. Time step of accuracy in time integration is mostly second-order
studies were included by 3 studies. Resistance was or higher. The divergence-free condition is satisfied
included in only one UT study using a relatively coarse either by velocity-pressure correction or an artificial
grid (0.3M) with relatively high uncertainty level compressibility approach. Analytical weighted re-
(UT=21%S) compared to average uncertainty of gridding, mesh morphing, and dynamic overset
motions (UT=2.5%S). Comparing motions, heave had approaches are used to handle dynamic ship motions.
generally higher UT than pitch for almost all studies. For the high-performance computing, message passing
For grid studies, three out of five studies considered interface (MPI) or open MP is used in many cases as a
resistance with average resistance uncertainty of parallel computing technique. Multigrid technique is
UG=2%S1, which is smaller than the average time step also used in some simulations for speeding up compu-
uncertainty for resistance. Pitch has slightly higher grid tations.
uncertainty than heave, with average heave and pitch The application of CFD captive maneuvering
uncertainty of UG=4.3%S1, which is slightly larger simulations focus on PMM-type forced motions such
than that for time step studies. Overall, average as static rudder, static drift, pure sway, pure yaw, and
simulation numerical uncertainties for seakeeping yaw & drift conditions. SIMMAN 2008 included
verification studies are USN=6.7%D for resistance and different geometries, i.e. a tanker (KVLCC), a
USN=4.9%D for motions. In the view of the average container ship (KCS), and a surface combatant (DTMB
grid size for USN studies being only about 7M, the 5415). For the KVLCC test case, two stern shape
solutions are likely far from the asymptotic range and variants named KVLCC1 and KVLCC2 giving
therefore USN values are optimistic. UV values are different instability loops were included. CFD-based
generally dominated by UD, being 15% for resistance methods were used to simulate forced motions and
and 10% for motions. Average error values are very were compared with PMM/CMT model test data. A
large for 2nd order terms (44%D) while for 1st order total of 16 submissions were received for the forced
terms the average is less than 15%. On the average motion simulations, comprising different CFD-based
level, validation is achieved only for the 1st harmonic methods such as RANS, URANS, and DES. It was
amplitude of heave at 11%D interval. concluded that finer grids were needed especially for
the rudder and appendages and in regions of large
7.3 Maneuvering vortices, as well as more advanced turbulence and
propeller models for improvements in the CFD
Captive maneuvering experiments and simulations are predictions of static and dynamic PMM maneuvers.
carried out to obtain coefficients used in system-based Recent results (Bhushan et al., 2011a) for 5415 bare
models to predict actual 6DOF maneuvers. The hull 20 deg static drift using 10 to 250 M grid points
simulations include static maneuvering simulations and DES turbulence modeling have shown that errors
such as pure drift and steady turn, and dynamic planar in X, Y, and N can be reduced to less than 5%D, which
motion mechanism (PMM) simulations which provide is comparable to errors for straight ahead resistance at
a wide range of derivatives needed in system-based the Gothenburg 2000 and Tokyo 2005 CFD work-
methods than those provided by static computations. shops. Static and dynamic PMM simulations are also
Both static and dynamic captive maneuvering simula- carried out for other geometries, i.e. HTC, MARIN
tions, as well as some free model cases were included LNG, and KVLCC2M (Toxopeus, 2009; Jacquin et al.,
in the SIMMAN 2008 workshop (Stern et al., 2011a) 2006). Mousaviraad et al. (2012b) included the effects
which was the first workshop on verification and of head wind on static and dynamic PMM maneuvers
validation of ship maneuvering simulation methods. in their CFD simulations (e.g. Fig. 17). It has been
Herein, an overview of computational methods and shown that multiple-run CFD/EFD curve fitting
applications are provided for CFD captive maneuver- methods can provide better estimates for nonlinear
ing simulations, along with a detailed analysis of the maneuvering derivatives than the single-run method
V&V results. used here, but statistical convergence of the higher-
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

order Fourier series components is an issue (Sakamoto side shells and engine rooms. The grid size was
et al., 2012). between 0.9-1.3M. Also different inlet shapes were
CFD V&V results for captive maneuvering used for the compartments. They compared the results
simulations are summarized in Table 4. Following with the experimental data and showed that CFD
Stern et al. (2011), assessment of CFD predictions for predicts the loads on the compartment fairly well
captive maneuvering separate capability for original during the flooding. The results showed that it is
values of X, Y, and N, slope of forces and moments necessary to consider the compressibility of air in
versus dynamic variables referred to as linear hydrody- FLOW3D to predict the flooding rate very well. Gao
namic derivatives, and higher-order terms of the slope and Vassalos (2011) used Fluent to predict the roll
referred to as nonlinear hydrodynamic derivatives. The decay of a damaged ship free to sway and roll. They
average of the fine grid points is 6 M. The overall showed that sway motion has large effect on hydrody-
average simulation numerical uncertainties are namic coefficients in damage condition. Added
USN=10%D for X, USN=12%D for Y, and USN=4%D moment of inertia and damping from roll decay
for N. For X, the largest numerical uncertainty is for without sway are significantly larger than those from
static rudder, while for Y and N largest uncertainties roll decay with sway. However, they did not validate
are observed for pure yaw simulations. UV values are the results with any experimental data. Gao et al.
generally in the same order as U SN, being 12% for X, (2011) developed a NS solver with VOF model to
14% for Y, and 5% for N. Average error values are study the motions and flooding for a barge with
15%D for X, 15%D for Y, and 11%D for N. The damaged compartment. The grid size was about 677K.
largest error values are generally observed for pure The ship was free to roll, heave and pitch. The
yaw and static rudder simulations. Overall, the average comparison with EFD data were conducted for both
error for X, Y, and N predictions is 13.6%D. For linear motions and the water heights inside the compartment.
derivatives, the average error is much larger for Nʹ The results showed good agreement with EFD data.
(40%D) than Yʹ (15%D). For nonlinear derivatives, the The effect of flooding on roll motion was not large due
average error value is about 40%D. to the strong stability of the barge.

7.4 Intact and damaged ship instability 8 VALIDATION FOR FREE RUNNING SIMU-
LATIONS AND SYSTEM IDENTIFICATION
There are very limited simulations conducted to
investigate the ship instability. The intact ship stability Numerical simulation of free running test cases
is conducted by Sadat-Hosseini et al. (2010) to (maneuvering, course keeping and etc.) is a challenge,
investigate parametric rolling using CFDShip-Iowa due to both the complexity of the physical phenomena
(e.g. Fig. 18). The computations were conducted for involved and the level of capability and resources
the ship free to heave, roll and pitch and with and needed to perform the computations. This type of
without bilge keels and for different ship speed and simulation requires self-propulsion, moving appendag-
loading in head waves. The results were compared with es such as rudders, controllers, and in general full
the experimental data and the predictions from 6DOF capabilities in a free surface environment. Thus,
nonlinear dynamic approaches which solve 1DOF roll there have been a few computations of this type so far.
equation using empirical or experimental values for In this section, the computational methods, the
damping and restoring moment coefficients. The applications and the V&V efforts for the free running
results showed that CFD can predict the parametric simulations are reviewed.
rolling for the ship with no bilge keels, similar to EFD. The computational methods applied for free
The parameter rolling was observed for the ship running simulations have employed different tech-
without bilge keels when the encounter frequency was niques for modeling, numerical methods and HPC as
about twice of the natural roll frequency of the ship. shown in Table 5 and 6. The free surface is mostly
The large damping coefficient of the ship with bilge predicted by interface capturing schemes such as
keels resulted in no instability. The comparison with single/two phase level set method and volume of
nonlinear dynamic approaches showed that the fraction technique. For turbulence modeling, the
nonlinear dynamic methods could predict parametric majority of the CFD codes solve two additional
rolling fairly well but their results were sensitive to the equations (i.e. in addition to the momentum and mass
accuracy of the coefficients. Cho et al. (2006) studied conservation equations) for the eddy viscosity, one for
the flooding of a compartment with no motion using the turbulence kinetic energy (k), and one for its
FLOW3D solver, a RANSE solver based on finite dissipation rate (typically ε or ω). Only one CFD code
volume method. The computations were performed for (RANS code developed by INSEAN) solves one
two compartments, the simplified compartment with no additional equation based on Spalart-Allmaras. Also,
internal complexity and the compartment with all the some employed the turbulence models based on
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Reynolds-Stress by solving the equations for the six damaged ship stability in calm water and waves. In
Reynolds stress components directly. Only DES is these simulations the damaged ship is floated and no
applied in this category in which the computational rudders are used for course keeping.
time reduces by employing unsteady RANS equation in
the boundary layer and applying LES in the rest of the 8.1 Maneuvering and system identification in
domain. For most computations, the propulsion is calm water and waves
usually implemented as body force and thus the
computational grid does not conform to the geometry Muscari et al. (2008a,b) used the RANSE code
of the rotating components. The body force is deter- developed at INSEAN (Italian Ship Basin) to simulate
mined by propeller open water curves and applied as a the turning maneuvering of KVLCC2 tanker. The free
source term to the underlying grid. To solve the rigid surface in the simulation was neglected and treated as
body equations it is necessary to obtain the instantane- symmetric plane. Thus the vertical motions including
ous forces and moments acting on the object. This is heave, pitch and roll were not included in the computa-
done by integrating the contribution of pressure and tions. The simulations were conducted for two different
viscous forces on each cell on the solid body in all grid sizes of 424K and 3.4M. The results for fine grid
solvers. In most CFD codes the flow is assumed to be showed the average error of 4.8%D for turning
incompressible and thus the continuity is enforced maneuver parameters which increases to 8.2%D for
using projection method to solve Poisson equation as coarse grid. The turning maneuver parameters include
the governing equation for pressure. The codes mainly the advance (A), transfer (TR), tactical diameter (TD)
adopt finite volume discretization and only one uses and turning diameter (D) as identified by IMO
finite difference scheme. High order schemes are used (International maritime Organization). They showed
for spatial and temporal discretization in all of the that the main features of the flow, including the
solvers. Most of the codes use the structure body fitted formation of the bilge vortices, were captured even on
grids where the cells are distorted to fit around a the coarse mesh. The same RANS code is used by
complex shape. The majority of the solvers use overset Durante et al. (2010) for turning circle simulation of a
or overlapping grid and the flow parameters is tanker with single rudder and twin propellers. The
interpolated from one grid to another. The interpolation simulations were conducted in full scale and compared
is done dynamically to form transient moving and with the EFD data which were repeated three times.
sliding grids to account for the relative motions of the The body force propeller model was modified to take
ship hull and the appendages. One solver (ICARE) uses into account both axial flow reduction at the propeller
structured grid with dynamic overset technique but the disk and the side force developed by the propeller. The
moving appendages such as rudders are modeled by side force was estimated to be proportional to the thrust
external forces. Few codes use unstructured re-gridding and the instantaneous angle between the propeller axis
for unsteady simulations. The HPC method is used in and the ship velocity. Two solutions were computed by
most of the solvers enabling parallel computing. A few using fine (6.2M) and medium (0.77M) grids and the
computations are conducted for a fairly large grid size numerical results were compared with the experiments.
with more than 15M grid points. However, the The prediction errors for turning circle were 4.95%D
computations of hundreds of millions of grid points are for fine grid and 9.31%D for coarse grid. Also, the
not yet conducted for free simulations while these EFD and CFD time history comparisons of ship speed,
types of simulations are reported for captive tests. drift angle and yaw rate showed that the speed was
The free running computations are conducted predicted better by medium grid while the drift angel
for different types of applications including maneuver- and yaw rate were predicted well by fine grid. The flow
ing, course keeping and stability. The majority of the field was analyzed in terms of wave pattern, surface
computations are for maneuvering predictions in calm pressure and the velocity field. The flow field was
water. Among the maneuvering computations, mostly characterized by a strong cross flow which generates
are conducted for turning simulations and few are large vortices including the vortical structures generat-
performed for zigzag. Two maneuvering simulations ed at the bilge keels, the keel and from the propeller
are conducted not only to validate the CFD results but apparatus. Dubbioso et al. (2012) extended the analysis
also to estimate the maneuvering coefficients using in Durante et al. (2010) and focused on the stern
system identification technique. The course keeping appendages forces/moments and their contribution to
computations are mostly carried out to study the ship the maneuvering. However, the forces/moments on the
stability in regular or irregular following waves and appendages were not validated. They showed that the
only few computations are conducted for course hydrodynamic loads of the appendages reduce the
keeping in calm water. One computation includes the turning quality of the vessel. This is due to the position
wind effects in the simulation to predict ship course of the center of pressure of the hull force, which is
keeping in hurricane. Some simulations are for shifted towards the stern, proving a stabilizing moment
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

which counteracts the vessel motion. It was also shown results were obtained using both constant propeller rate
that the appendages on the external side of turning and constant torque condition for propellers. The total
develop a higher lateral force with respect to the inner grid points were 7M for calm water increasing to 12M
side ones as the appendages in inner side are in wake of for waves. The simulations were conducted for both
the hull. Even though the lateral force of inner side model and full scale and no validation was reported.
appendages were small, the induced moment was in the The details of turbulence modeling for the model scale
same order as that developed by the bare hull. This and full scale simulations were discussed later in
stresses the extreme importance of the stern appendag- Bhushan et al. (2009). For model scale simulations,
es on the maneuvering behavior of the vessels. blended k-w was used while for full scale computations
Jacquin et al. (2006) employed ICARE, the the multilayer wall-function models were employed.
RANS code developed by ECN (Ecole Centrale de For zigzag simulations, the full scale case reached the
Nantes) in France. They performed turning maneuver rudder check points faster than that of model scale,
of a containership (Series 60). They mentioned that the indicating a slightly more efficient rudder action in full
computations were conducted with actual moving scale. Since the boundary layer was thicker in the
rudders and also with simplified condition where a side model scale case, the lower velocities at the propeller
force was used to mimic the existence of a deflected plane was expected compared to that of full scale. This
rudder behind the ship. However, only the free running provided higher velocity reaching the rudders for full
results of the model with side force were reported. The scale and consequently better steering capacity. For
computations were only conducted to demonstrate the turning simulations in waves, with constant RPM, the
capability of ICARE in free running maneuvering and ship was locked in the wave when the wave was
no validations were shown. The solver ICARE later following waves providing surf-riding. In the case of
was used by Ferrant et al. (2008) and Drouet et al. constant torque the ship did not have enough power
(2008). Ferrant et al. (2008) demonstrated the capabil- and the following waves could overcome the ship. The
ity of ICARE for turning circle simulation in regular validation of calm water turning and zigzag simulations
waves for 5415M, a surface combatant. The rudder was for 5415M was reported in Carrica et al. (2012b).
approximated by a time varying side force applied at Figure 19 shows the predicted free surface and vortical
the rudder location. The Wave is treated using structures during turning circle reported in Carrica et
SWENSE (Spectral Wave Explicit Navier-Stokes al. (2012b). The average error of turning and zigzag
Equations) approach by combining the description of maneuvers predictions were 5.74%D and 6.83%D,
undisturbed incident waves introduced by a non-linear respectively. Mousaviraad et al. (2012a) used CFD-
spectral scheme in potential flow theory and the Ship-Iowa for turning and zigzag maneuvering of a
computation of the non-linear viscous diffracted flow surface effect ship (SES), as shown in Figure 20. The
using the free surface RANS solver. Since there is no body force propeller model was used in the simulation.
direct impose of wave condition on free surface, there The simulations were conducted in both deep and
would be no need to have a fine grid to ensure a good shallow water and in both calm water and waves with
propagation of the waves. Drouet et al. (2008) con- the grid size of 8.6-9.3M. It was shown that shallow
ducted turning maneuver simulations for Humburg test water increases transfer and tactical diameter in turning
case and 5415M using ICARE solver. For both maneuvers. However, radius and yaw rate showed no
geometries, the rudders were modeled by an external change in shallow and deep water. Also, shallow water
side force. For Humburg test case, the simulations were reduced the 1st and 2nd overshoot for zigzag. For
conducted for the model free to horizontal motions turning, wave effect is significant on the maneuver of
(surge, sway and yaw) and also free to all motions. The the ship in shallow water. For zigzag, the waves reduce
comparison with EFD data showed better agreement the overshoot angle. Sadat-Hosseini et al. (2011a) and
for 3DOF simulations with error of 0.3%D for tactical Araki et al. (2012a, b) employed CFDShip-Iowa for a
diameter compared to 5%D for free simulations. For surface combatant (ONR Tumblehome) maneuvering
5415M, the turning maneuver results were reported in in calm water and waves. The model included all the
calm water and waves but no validation was conducted. appendages with grid size of 12.1M in total but the
For calm water, a model with different bulb forms were propellers were simulated using body force. The
considered and showed that the tactical diameter average error for calm water simulations were 1.62%D
reduces 5% for a bulb form shorter and more immerged and 1.88%D for turning and zigzag. The results for
than the initial form. waves also showed very good agreement with the EFD
Carrica et al. (2008a) employed CFDShip- data. They also used system identification technique
Iowa for turning and zigzag maneuver simulations in for first time to improve the maneuvering coefficients
calm water and waves for 5415M. The body force and their variations induced by waves, rudder forces
propeller model was used for the propulsion and the and wave forces from CFD free running outputs. They
actual rudders were included in the simulations. The used two system identification methods including
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

extended Kalman filter (EKF) and constrained least using both body force and actual propeller are shown in
square (CLS). The results in calm water showed the Figures. 22 a and b.
average system based prediction errors for maneuver- The validation of forces and moments on the
ing simulations drop from 16% to 8% by using the appendages in free running simulations were per-
maneuvering coefficients and rudder forces found from formed by Stern et al. (2011b) and Sadat-Hosseini et al.
CFD free running instead of those from captive (2012a). They employed CFDShip-Iowa and ISIS-CFD
experiments (see Figure 21). Also, the system based to study the predictions of course keeping for 5415M in
results in waves were significantly improved by tuning calm water and regular waves. The CFDShip-Iowa
the maneuvering coefficients and wave forces in the simulations included course keeping under either roll
mathematical model using CFD outputs. decay or forced roll in calm water and course keeping
in head and beam waves. For ISIS-CFD computations,
8.2 Course keeping and stability in calm water the roll decay in calm water was only performed. The
and waves moving rudders and stabilizers were included in
CFDShip-Iowa while both were fixed in ISIS-CFD.
Carrica et al. (2008b) and Huang et al. (2008) demon- The propeller was modeled as a body force in both
strated the capability of CFDShip-Iowa for course solvers. The grid size for CFDShip-Iowa and ISIS-
keeping and the instability of ONR Tumblehome in CFD was 6.3-18.6M and 5.9M, respectively. The
irregular following waves. The actual rudder was results showed very good agreement with EFD data
included in the simulation and the propeller was using both CFD simulations. Also, the CFD results
modeled through a body force. Broaching was were compared with potential flow and system based
predicted during the course keeping but the results method predictions and confirmed that CFD provides
were not validated. Mousaviraad et al. (2008) demon- better results than the other methods. Figure 23 shows
strated the capability of CFDShip-Iowa for course the grid topology and the instant view of CFDShip-
keeping in waves in present of significant winds. They Iowa solution during course keeping in beam waves.
simulated course keeping in CAMILLE hurricane for Sadat-Hosseini et al. (2012b) extended the validation
ship in quartering waves with winds hitting the ship at of forces and moments to irregular wave condition, as
45 and 225 deg. The results showed the wind has shown in Figure 24. The irregular wave was generated
strong effects on ship forces and moments. The case by summation of many regular waves with amplitudes
with wind from 45 deg loses control at some instants computed from the desired wave spectrum and with
but regains control while the ship turns to port and random phases. A very fine grid was used to resolve all
cannot be controlled for the case with wind at 225 deg. the components of the irregular waves from very small
Sadat-Hosseini et al. (2011b) validated the course to very large amplitudes. The grid size was about 24M,
keeping and instability of ONR Tumblehome for decomposed into 249 CPUs for parallel processing.
various speed and heading in regular following waves. The computations were conducted for course keeping
The CFD simulation could remarkably predict the of a surface combatant in irregular beam waves for
instability map. CFD predicted the boundary between JONSWAP wave spectrum and the results were
periodic motion and surf-riding/broaching at Fr=0.3 for compared with the experimental data. The results were
heading less than 30 degrees. Periodic motion was validated against the experiments not only for the ship
predicted below the boundary (Fr<0.3) whereas surf- motions but also for the loads on the appendages. The
riding/broaching was observed above the boundary, in correlation between ship motions and input irregular
agreement with EFD. However, there were discrepan- waves were also studied. Comparing the irregular wave
cies in the time histories due to missing EFD initial results with the results computed from regular wave
condition for ship speed and wave phase. The authors simulations at several discrete wave length conditions
later included the actual propeller in the simulation and showed that the ship has fairly similar motion in both
showed improvement for the time histories of all regular and irregular waves with same wave length
motions (Carrica et al., 2012a). Also direct integration condition.
of forces and moments on all appendages were Dreyer and Boger (2010) used OVER-
conducted to investigate the contribution of the REL_TCURS solver to simulate the overtaking of a
appendages on the instability of the ship during the submarine and a ship. The OVER-REL_TCURS solver
course keeping. The yaw moments showed that the is developed at ARL (Applied Research Laboratory,
turning moments produced by the appendages were Pennsylvania State University). The free surface is
dominated by the rudders, with values one order of ignored in the simulation. The body force propeller
magnitude larger than any other appendage. Also, both model was used in the simulation for both the ship and
propellers contributed a net positive turning, helping the submarine. The grid size of the ship and submarine
prevent broaching. The CFD solutions for broaching was 2.3M and 4.1M, respectively. The comparisons
were performed for depth and pitch. Good agreement
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

with EFD was obtained for locked fin cases but large
discrepancies for those cases with moving fins were 9.1 Flow around surface-piercing object
observed. A few simulations were conducted for the
ship with damaged compartment to investigate the Experimental and computational studies have been
stability of the ship under damaged condition. conducted for the flows of a surface-piercing flat plate
Strasser et al. (2009) conducted CFD simula- with focus on the wave induced effects on the flat plate
tion of floating damaged barge. They included the boundary layer below the juncture region (Stern, 1986;
compressibility of the air trapped in the damaged Stern et al., 1989; Stern et al., 1993). More recent
compartment. The time histories of sinkage, trim, heel experimental and computational studies focused on the
and water height inside the compartment were juncture region and turbulence modeling using towed,
compared with EFD data and the results showed very two-dimensional laser-doppler velocimetry (LDV) and
good agreement. Sadat-Hosseini et al. (2012c) RANS and large-eddy simulation (LES) methods for
employed CFDShip-Iowa for a damaged ship and flat free surface (Longo et al., 1998; Sreedhar and
investigated roll decay and motions with flooding in Stern, 1998a, b). The juncture region experiment
calm water and waves. The grid size was in range of (Longo et al.,1998) is extended to include wave effects
6.3-28M, depending on the case. The results were using servomechanism wave gauges (Kang et al.,
compared with experimental data not only for ship 2008). It is observed that the interface appears always
motions but also for the water height inside the broken in the contact line region even for the case with
compartment to evaluate the flooding rate prediction. a smooth wave field. Air is entrained into the boundary
The verification study was also conducted, showing layer at the contact line region due to the interaction of
that the results were not very sensitive to the numerical the contact line with the solid surface. A string of
uncertainty. Overall CFD showed good agreement with bubbles can be generated near the solid wall usually
experimental results for both ship motions and water with small bubble sizes. For flows past a blunt leading
heights inside the compartment for all cases in calm edge, such as NACA0024 foil, the bow wave breaks
water and waves. CFD results could capture the and wraps around the bow with similar features to
complex free surface shape inside the compartment spilling breakers (Metcalf et al., 2006) at a Fr number
generated by the flooded water and the reflected water of 0.37. Wave induced flow separation occurs and
by the walls inside the compartment (see Figure 25). reattaches to the foil surface resulting in a wall-
For roll decay cases, CFD showed that the damping bounded separation bubble. The complexity of
coefficient is larger and roll frequency is smaller for unsteady, wave-induced, boundary layer separation
the roll decay under damaged condition. makes the experimental measurements difficult. With
the combined CFD studies (Kandasamy et al., 2009
9 FUNDAMENTAL STUDIES FOR TWO- and Xing et al., 2007), more detailed flow description
PHASE FLOWS is provided. Three main instabilities, namely the initial
shear-layer instability, Karman-like shedding and
Breaking waves, spray formation, and air entrainment flapping of the separation bubble, were found in the
around ships are one of the main sources of the unsteady Reynolds-averaged Navier-Stokes (URANS)
underwater sounds and white-water wakes, which are study (Kandasamy et al., 2009). These instabilities are
of great importance for signature of ships. These flow difficult to be identified in both the DES study (Xing et
phenomena occur on a large range of temporal and al., 2007) and experimental study (Metcalf et al., 2006)
spatial scales. At small scales, spray formation and air due to the numerous small-scale vortices in the
entrainment in the contact line region depend on the separation region. There are limitations in both of the
length scales which are orders of magnitude smaller two CFD studies. The complicated flow structures at
than the length of the ship with correspondingly small the interface as shown in the experimental images, such
time scale. For the large scales, the overall structures as splashing, breaking waves, air entrainment and free
include wave breakings along the hull and the induced surface induced turbulence observed in the experi-
vortices, instabilities, separations and scars, etc. ments, cannot be resolved since a single phase flow
Understanding the small scale physics and capturing solver (air effect is neglected) and relatively less
their effects on the large scale features are of primary accurate interface modeling methods were used. These
importance for ship hydrodynamics. Skin-friction drag unresolved flow structures at the interface might affect
reduction is of great significance in ship hydrodynam- the global flow patterns and turbulence. In Suh et al.
ics. One of the important techniques for the drag (2011), flows past an interface piercing circular
reduction is to inject gas into the liquid turbulent cylinder is investigated using LES with a focus on the
boundary layer to form bubbles or air layer. In this effect of air–water interface on the vortex shedding
section, fundamental studies for two-phase flows in from the vertical circular cylinder (selected results
ship hydrodynamics are reviewed. shown in Fig. 26).
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Previous studies are mainly focused on the for a wedge-shaped bow wave based on the experi-
global structure of ship flows, such as the wave mental measurements and elementary fundamental
elevation, scars, and vortices. The small scale details of considerations, which define the main characteristics of
interface breaking, spray formation, and air entrain- a ship bow (wave height, wave crest location and
ment are not well understood. profile, and flow steadiness or unsteadiness) in terms of
ship speed U, draught D and waterline entrance angle
9.2 Bow wave breaking and spray formation 2θ. A computational study has been carried out by
Broglia et al. (2004) with the results compared with the
Ship bow waves exhibit both large and small scale experimental data (Waniewski et al., 2002). Since a
features. The most prominent large scale feature is the single phase level set method was used for the free
Kelvin bow wave pattern scaled by Fr; however, for surface tracking, the small scale interface structures
sufficiently large Fr and depending on bow shape, were not captured. The mechanism of the liquid sheet
spilling and/or plunging breaking occurs and induces disturbance, fingering, pinching off drops and spray
vortices and scars additionally scaled by Re and We. formation has not been thoroughly studied.
Small Re and We inhibit breaking, i.e., in general In Wang et al. (2010b), flows around a wedge-
model scale flows exhibit reduced breaking compared shaped bow are numerically simulated with the aim of
to full scale, and smaller models display reduced investigating the wave breaking mechanism and small
breaking compared to larger models. A recent com- scale features of ship bow waves. The study (Waniew-
plementary EFD and CFD study has documented bow ski et al., 2002) was selected as test cases to validate
wave breaking and induced vortices and scars for the capability of the code of CFDShip-Iowa Version 6
model 5415 for Fr = 0.35 using CFDShip-Iowa Version (Yang and Stern, 2009; Wang et al., 2009b) for small
4 (Olivieri et al., 2007; Wilson et al., 2007a). The most scale features of ship bow waves. The simulations are
prominent small scale feature is the bow wave crest carried out using a Cartesian grid solver first with the
formation of thin overturning sheets which break up sharp interface, coupled level set and volume-of-fluid
into spray. In the study by Beale et al. (2010), recent (CLSVOF) and immersed boundaries methods; and an
in-situ measurement efforts were reviewed and spray orthogonal curvilinear grid solver (Wang et al., 2012a)
formation generated by full-scale naval platforms were is also used in order to increase the grid resolutions
examined. Re and We scale effects are large such that near the wall.
replication of full scale phenomena of the small scale The wedge geometry is similar to the large
features of ship bow waves is difficult even with large wedge model used by Waniewski et al. (2002). The
models. The extent of the thin sheets is drastically side length of the wedge is L = 0.75 m, and the height
reduced and remains attached, as shown by Stern et al. of the wedge is H = 1 m. The half wedge angle is θ =
(1996) for the Series 60. 26º and the flare angle φ = 0º. The sharp edge corners
However, studies for wedge flows by are rounded with an arc of a small radius in order to
Waniewski et al. (2002) and Karion et al. (2004) make the grid orthogonal at the two corners. The grid is
display and document the structure of bow waves (thin 1536×768×848 (one billion) which is refined near the
water sheet formation, overturning sheet with surface solid surface.
disturbance, fingering and breaking up into spray, For the case considered here, the water depth
plunging and splashing, and air entrainment) with is d = 0.0745 m and the upstream velocity is U = 2.5
valuable experimental data provided such as wave m/s, the corresponding Reynolds number,
elevation, extent of wave breaking, spray droplets size Re  Ud /  = 1.64×105, and the Froude number,
and number. These two studies are summarized in
Table 7. In Waniewski et al. (2002) the bow waves are Fr  U / gd = 2.93. The domain boundary is given
created by a deflecting plate mounted at an angle in a by x = [5.33 m, 4.55 m], y = [0.0745 m, 0.6 m], and z =
flume, the towing tank experiments are also conducted [0 m, 5 m]. Uniform inflow and convective outflow
with two wedge models. Typical bow wave profile boundary conditions are used. Slip-wall boundary
obtained in the towing tank experiment shows a thin conditions are imposed at all the other boundaries. A
liquid sheet is created at the leading edge of the wedge, uniform velocity field same as the upstream velocity is
and it continues to ride up on the side wall. This thin prescribed to the entire computational domain at t = 0.
liquid sheet starts to separate from the side wall as it Figure 27 shows the computed bow wave profile
reaches its maximum height. Once the crest reaches its compared with the experimental video image. The
maximum height, an overturning jet is formed and overall wave structure is very similar to the experi-
plunges back onto the undisturbed free surface. A large mental observation (Waniewski et al., 2002) as shown
area of splash is generated at the wake of the wedge in figure, such as the thin liquid sheet at the leading
due to wave plunge and air entrainment. Noblesse et al. edge of the bow, overturning jet, jet plunging onto the
(2008) proposed a series of simple analytical relations free surface, and splashes at the wake. As the liquid
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

sheet overturns, the sheet is stretched and fingered up, Lubin et al., 2006; Iafrati, 2010), simulations are
and some ‘‘cylindrical drops’’ then pinch off from the conducted with the air entrainment, 2-D and 3-D vortex
liquid sheet, when the detached drops impact onto the structures, and energy dissipation discussed.
water surface, a spray region is created. The close-up It should be noted that most previous studies
view of the bow sheet breakup is shown in Fig. 28. In on plunging wave breaking are for deep water or
the experimental study conducted by Deane and Stokes sloping beach for which wave plunges forward in the
(2002), the diameters of most observed bubbles due to same direction of the mean flow. Yao and Wu (2005)
the fragmentation process are greater than 2 mm. Mean experimentally investigated the shear currents effects
drop size is 2.3 mm as observed in experiments on unsteady waves but with a focus on incipient
(Karion et al., 2004). The effective diameters measured breaking. Moreover, the geometry and conditions in
in the experiments by Beale et al. (2010) for full-scale most cases of CFD differ from the experiments even
ship bow spray are mainly in the range of 1 mm to 2.5 though the experiments are usually used to guide the
mm. For small size bubbles/droplets, surface tension analysis of CFD. Present interest is ship hydrodynam-
force is dominant and further fragmentation is difficult. ics for which body-wave interactions are important and
In the present study, the grid spacing is 0.125 mm near the direction of wave breaking is opposite or at an
the wedge and 1 mm in the plunging region. With the angle to the mean flow. Previous research used model
current grid, the minimum drop size is 0.8 mm near the ships in towing tanks focused on scars, vortices and
wedge. The droplets and bubbles near the wedge can mean and root mean square (rms) wave elevation
be effectively captured. Further grid refinement (3 to 4 induced by ship bow and shoulder wave breaking
billion grid points) is needed to increase the resolution (Miyata and Inui, 1984; Dong et al., 1997; Olivieri et
in the wake region. al., 2007), which suggests the presence of underlying
coherent structures. A complementary CFD study to
9.3 Plunging wave breaking the latter study was carried out by Wilson et al. (2007).
However, typical plunging wave breaking can hardly
Plunging wave breaking is one of the most violent be obtained using model ships in towing tanks, and
phenomena of air-water interface interactions, produc- detailed measurements of the wave breaking processes
ing strong turbulence with large amounts of air are difficult. Recently, Shakeri et al. (2009) provide
bubbles, water droplets, jets and sprays. These detailed measurements and analysis of divergent bow
phenomena commonly occur in ship flows and are one waves using a unique wave maker for simulating 2D +
of the main sources of the underwater sounds and t flow. A numerical study using a 2D + t model has
white-water wakes, which are of great importance for been reported by Marrone et al. (2011) for high speed
signature of ships. Many experimental and computa- slender ships. For slender bow ships, 2D + t wave
tional studies for the plunging wave breaking have breaking process is similar to deep water and sloping
been done in the past few decades. Early experimental beach studies, i.e., plunges with forward splash-ups. In
studies are focused on wave geometric properties the early experimental studies by Duncan (1981; 1983),
(Bonmarin, 1989), wave breaking process (Tallent et a fully submerged, two-dimensional hydrofoil was
al., 1990), energy dissipation (Melville, 1994), jet towed horizontally to produce breaking waves. These
characteristics and air entrainment (Chanson and Fang, studies are focused on spilling breakers. Greco et al.
1997), and turbulence (Chang and Liu, 1999). With the (2004) investigated the impact flows on ship-deck
Particle Image Velocimetry (PIV) techniques, more structures due to head-incoming waves.
detailed velocity field, turbulence, and void fraction In the experimental fluid dynamics (EFD)
data and analysis have been investigated in recent study by Kang et al. (2012), a quadratic profile bump
studies (Melville et al., 2002; Deane and Stokes, 2002; mounted in a shallow water flume is used to create
Grue and Jensen, 2006; Kimmoun and Branger, 2007; impulsive sub critical flow conditions where plunging
Blenkinsopp and Chaplin, 2007; Drazen and Melville, wave breakers are successfully obtained. Ensemble-
2009). Due to the technical difficulties, the experi- averaged measurements (relative to the time tb at which
mental measurements can only be done in the water the maximum wave height is reached just before the
region, and detailed description of the flow field in the first plunge) are conducted, including the overall flume
energetic wave breaking region is not available. With flow and 2-D PIV center-plane velocities and turbu-
the development of the computational fluid dynamics lence inside the plunging breaking wave and bottom
(CFD) technology, detailed wave breaking process and pressures under the breaking wave. The plunging wave
velocity profile can be obtained in both water and air breaking that is triggered by the flow over a submerged
phases (Chen et al., 1999; Watanabe and Saeki, 2002). bump is of relevance to ship hydrodynamics since it
The early CFD studies are usually 2-D due to the includes wave-body interactions and the wave breaking
prohibitive computational cost for the 3-D simulations. direction is opposite to the mean flow as discussed in
In the more recent CFD studies (Watanabe et al., 2005; Kang et al. (2012). The idea and approach of creating
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

plunging wave breakers using a submerged bump is In a 2D CFD study conducted by Iafrati
obtained collectively from the previous experimental (2010) for Stokes waves, drops and bubbles in breaking
(Cahouet, 1984; Miyata et al., 1985a) and CFD studies waves are quantitatively estimated with some limita-
(Iafrati et al., 2001; Yang and Stern, 2007; Huang et tions due to the numerical model adopted in the
al., 2007). The CFD results were used as a guide for simulations. Recently, simulation on a large grid of
the test design of the experiments (Ghosh, 2008) and 1920×1280×896 (2.2 billion) was carried out for
the initial experimental data was used for validation. bubble/droplet size distribution in the breaking waves
Subsequently, a complementary CFD study was used (Wang et al., 2012d). Figure 31 shows the close-up of
to aid in the data analysis simultaneously as the the bubble and droplets in breaking waves on the large
experimental data is used to validate a Cartesian grid, grid, where detailed small interface structures are well
immersed boundary, coupled level set and volume-of- demonstrated.
fluid CFD method (Wang et al., 2009b). Wang et al.
(2009b) indentified three repeated plunging events 9.4 Air-layer drag reduction
each with three sub-events [jet impact (plunge),
oblique splash and vertical jet]; however, they used Bubble drag reduction (BDR) is an important technique
fully impulsive initial conditions and adjusted the that injects gas into the liquid turbulent boundary layer
initial velocity and water elevation to match Ghosh's to form bubbles to obtain drag reduction. This tech-
(2008) wave breaking position, which precluded nique can substantially reduce skin friction, which has
detailed temporal validation. great potential applications in ship hydrodynamics.
In the study by Koo et al. (2012), impulsive During the past several decades, a large amount of
plunging wave breaking downstream of a bump in a research has been devoted to the BDR (Merkle and
shallow water flume is numerically simulated with the Deutsch, 1992). However, most of the studies were
aim of providing a detailed quantitative description of conducted at relatively low Reynolds numbers and
the overall plunging wave breaking process. The time- small scales. Proper scaling of BDR remains unclear.
dependent velocity and wave elevation boundary In the study by Sanders et al. (2006), a set of
conditions are specified at the inlet and outlet using the BDR experiments were conducted for a large scale flat
exact experimental data provided in Kang et al. (2012). plate turbulent boundary layer at high Reynolds
The computational results are compared with the numbers. It has shown that significant levels of BDR
experimental measurements to validate the capability could be achieved only near the air injector, and
of the code of CFDShip-Iowa Version 6 (Yang and limited persistence of BDR exists away from the air
Stern, 2009; Wang et al., 2009b) for wave breaking. injector. This short persistence distance of BDR makes
The simulations are carried out on a 2D Cartesian grid it impractical for applications. It has also shown that a
using the sharp interface, coupled level set and volume- layer of gas was formed and persisted along the entire
of-fluid (CLSVOF) and immersed boundary methods. plate at lower flow speeds and higher gas injection
In the study by Wang et al. (2012b), a 3D LES of the rates, which could lead to skin-friction reduction of
impulsive plunging wave breaking is performed on an more than 80%. Elbing et al. (2008) continued the
orthogonal curvilinear grid of 768×256×64 (stream- study of Sanders et al. (2006) in an effort to understand
wise, vertical, and spanwise) points in order to identify the mechanisms underlying the limited persistence of
the 3D structures of breaking interface. The constant the BDR and the onset conditions for the air layer drag
inlet velocity imposed at the left boundary is u = 0.87 reduction (ALDR). The experimental results indicated
m/s for water and zero for air. The initial interface that ALDR could be established once the air was
elevation is 0.2286 m and a uniform velocity field is injected beyond a critical rate, and more than 80% drag
prescribed in the water domain with the air phase at reduction could be obtained over the entire plate. Three
rest. The overall interface structure of the breaking distinct regions associated with drag reduction were
wave compared with the experimental video images is observed with air injection rate: BDR, transition and
given in Fig. 29. The major events of the first plunge ALDR. It was found that the air layer was sensitive to
wave breaking are demonstrated, i.e., maximum height, the inflow conditions. In the recent work (Ceccio,
first plunge, oblique splash, and vertical jet as identi- 2009), a 1/2 inch (12.7 mm) step was used at the inlet,
fied by Wang et al. (2009b) using 2D simulations. On and the air was injected from the base of the backward-
the video images, the entrapped air tube is marked by a facing step. This greatly enhances the stability of the
circle. As shown in the figure, the computational air layer. The ALDR is a potential alternative to BDR,
results match the experiments very well. The 3D span- however, the knowledge of ALDR mechanism is quite
wise interface structures due to the centrifugal limited and more comprehensive studies are needed.
instabilities at the curved flow as observed in the Related to ALDR, partial cavity drag reduction
experiments are also captured as shown in Fig. 30. (PCDR) is another important technique to reduce skin
friction. PCDR needs potentially lower gas flux
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

compared to ALDR, but un-optimized cavity flow can equations, and partially parabolic RANS equations, in
lead to significant form drag (Ceccio, 2009). Partial general with a flat free surface, to full RANS equations
cavities are sensitive to flow speed and perturbations for single-phase or two-phase flows around model- or
from the incoming flow (Amromin and Mizine, 2003). full-scale ships with deformed or broken interfaces,
In the study by Wang et al. (2010a), URANS 6DOF motion prediction, and motion controllers. Now
simulations of ALDR on a large scale flat plate are computational ship hydrodynamics is to embrace the
performed. The simulations are carried out using a exascale computing era with LES of multi-scale and
sharp interface Cartesian grid solver, with the interface multi-physics multi-phase ship flows on billion-point
tracked by a coupled level set and volume-of-fluid grids, resolving most scales of turbulence and bub-
(CLSVOF) method and turbulence modeled by a bles/droplets at the same time.
Spalart-Allmaras (SA) turbulence model (Spalart and In the course of this astronomical progress,
Allmaras, 1992) with a wall function (WF) approach. there is a clear line of evolution in governing equa-
The experimental data reported by Elbing et al. (2008) tions. Initial work was mainly focused on the boundary
is used to validate the simulation results. The simula- layer portion of the solution domain due to its funda-
tions are carried out on a two dimensional (2D) mental importance in ship flows, lack of efficient
computational domain. The Reynolds number is Re = gridding and solution techniques for high-Re incom-
7.37×107 which corresponds to Test 1 in Elbing et al. pressible flows with complex geometries, and limited
(2008) with a free stream velocity of 6.7 m/s. It is a computing power available at that time. Then steady
challenge for the numerical simulation of such flow RANS equations with grid-fitted free-surface condi-
since high Reynolds numbers, air-water interface, and tions became the dominant model in many solvers,
two-phase turbulence are involved. The air layer along some of which are still in use today. Nowadays most
the entire test plate is successfully achieved and the solvers solve the unsteady RANS equations for the
drag reduction is approximately 100%, which agrees water flows (and semi- or fully-coupled air flows in
with the experimental findings very well. With reduced some solvers) around ships with the air-water interface
air flow rate, BDR is also observed; the computational embedded in grids. For tomorrow’s high-fidelity
results also qualitatively match the experiments. The solvers, there is no doubt that the complete set of
transitional region from BDR to ALDR is also governing equations with no or minimum phenomeno-
observed in the present simulation. However, the logical modeling will be employed. This means the
critical air flow rate to form the ALDR is lower in the rigorous implementation of interface jump conditions
simulations than in the experiments. Several possible due to discontinuous density and viscosity and surface
reasons are likely accounting for the low critical air tension, fully coupled 6DOF ship motion, fully
flow rate in the simulations, such as SA-WF turbulence discretized propulsors, LES with wall models, and
model, three-dimensional instability and surface environmental effects. Such a list of mathematical
tension effects. The critical air flow rate does not formulations might seem to be too sophisticated and
change much with grid refinement. unnecessary for some applications well within the
In the study by Kim and Moin (2010), DNS of capabilities of many current tools, but they are indeed
the air layer drag reduction was conducted for a flat the logical consequence of the above development and
plate with a backward-facing step. In their simulations, required for understanding the temporal-spatial
the Reynolds number is 2.28×104 with a free steam correlations of a much wider spectrum in ship hydro-
velocity of water of 1.8 m/s, which is much lower than dynamics.
that used in the experimental study (Elbing et al., 2008) Numerical methods for solving the mathemat-
and CFD study (Wang et al., 2010a). This is because ical models in ship hydrodynamics are mostly finite
the Reynolds number in the experiment of Elbing et al. difference and finite volume methods, although finite
(2008) is too high for the direct numerical simulation. element and particle methods are also used by some
The simulation results show that the air layer was research groups. In general finite difference methods
stabilized with increased air injection rate, and much require complex transformed governing equations and
larger skin-friction was obtained when the air layer high quality structured grids, which are sometimes
broke up. quite difficult to generate for complex geometries.
Cartesian grid methods can overcome the grid genera-
10 CONCLUDING REMARKS tion problems, but further development is still required
for the accurate imposition of boundary conditions,
10.1 Conclusions especially for high-Re turbulent boundary layers, at the
immersed boundaries. Finite volume methods have
The last thirty or so years have witnessed monumental gained wide popularity because of their accommoda-
progress in CFD for ship hydrodynamics: started from tion to polyhedra of arbitrary shape as grid cells and
solving momentum integral equations, boundary layer straightforward derivation of discretization schemes
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

from conservation laws. Computational ship hydrody- popular one for its conformability with different types
namics research codes mostly use multi-block struc- of grids and discretization schemes.
tured grids because of a higher level of accuracy they For computational ship hydrodynamics, the
can provide, although unstructured grids are increas- most commonly used application area is resistance and
ingly applied, especially with commercial or open- detailed local flow field predictions around the hull,
source general-purpose application solvers, to ship followed by seakeeping, manoeuvring and self-
flows. Most solvers use an inertial reference frame for propulsion. For seakeeping simulations, wave models
the flow field and a ship-fixed non-inertial reference are useful for simulating incident waves or sea
frame for ship motion; the latter is also used in some environments. Winds and air flows are important in
solvers for the flow field but the applications are applications such as landing-of-aircraft, high speed
limited to single ship cases. In general geometry ships, and wind force applications. Propulsion model-
motions in ship flows are treated using deforming, ing is performed using body force approach based on
sliding, regenerated, and/or dynamic overset grids; the radial distribution body force for finite-bladed
only the last approach provides a good compromise of propeller. Waterjet propulsion systems have also been
applicability and efficiency, although conversation modeled using body force method for the pump and by
property between overlapping grids is still an outstand- applying axial and vertical reaction forces and pitching
ing issue. Level set and algebraic VOF methods are reaction moment.
widely used in many solvers for tracking the air-water Linear URANS models, in particular two-
interface; but the former are long known of a flawed equation blended k-/k- model, are the most com-
mass conservation property and the latter are also monly used turbulence closure for ship hydrodynamics
criticized for giving a diffusive, blurred interface. applications. The model performs reasonably well for
Geometric VOF methods with various enhancements, the model-scale resistance prediction on up to 10M
such as the constructed distance function, centroids of grids. However, they fail to accurately predict the mean
volume fractions, and tracked surface markers, can vortical and turbulent structures. Currently used
provide highly accurate interface position as well as anisotropic models show some improvements over the
strict mass conservation, and facilitate sharp interface linear models in predicting these features, but are not
treatment in Navier-Stokes equations; they will be seen sufficiently accurate. Wall-functions are a viable option
in more and more high-fidelity simulations in ship for full-scale ship simulations including wall roughness
hydrodynamics. For solving the incompressible effects, and perform reasonably well for resistance
Navier-Stokes equations, pressure-based and density- predictions when compared with ITTC correlation.
based (artificial compressibility) fully coupled methods Thus far, limited LES or hybrid RANS/LES simula-
can provide robust and fast solution, but they demand tions have been performed for ship hydrodynamics.
too much computer memory and are mostly used in Nonetheless, the simulations emphasize the advantages
steady flow problems. The SIMPLE-family algorithms of these methods over URANS as they reduce depend-
were initially designed for steady problems too; and ency on modeling, resolve small scale-scale physics,
their tightly coupled iterative solution procedure for improve understanding of turbulence and vortical
both velocity and pressure greatly restricts the applica- structures, and two-phase flow and air entrainment.
tions of many novel numerical schemes within them They also help in explaining the observation in sparse
including PISO for unsteady problems. On the other experimental data and guide experiments. The
hand, fractional-step methods have been applied in advances in high performance computing will soon
high-fidelity simulations for a long time; they also enable calculation on hundreds of millions to billions
allow different optimized strategies for different terms. of grid points on a regular basis. Simulations on tens to
High-order schemes including semi-Lagrangian hundreds of millions of grid points can help us in
advection schemes can be implemented with ease. On obtaining benchmark URANS predictions using the
the contrary, most current RANS solvers use first- or existing model to identify their limitations. Hybrid
second-order temporary and spatial discretization RANS/LES on hundreds of millions to billions of grid
schemes; and applications of higher-order schemes are point can help in resolving 90-95% of the turbulence,
usually cautioned with robustness issues. Therefore, and can be used to develop accurate anisotropic
fractional-step methods will be seen in more and more URANS models for general purpose applications.
solvers for ship hydrodynamics applications due to For quantitative V&V, there are several prob-
their flexibility in using various higher-order schemes lems in using the Richardson extrapolation method. It
for improving overall accuracy of the solution. In requires that all solutions should be sufficiently close
addition, solution adaption strategies can be used to to the asymptotic range, i.e., within about six percent of
obtain the required level of accuracy with minimized the pth of the numerical method. When solutions are
computational cost, in which the h-adaption is the most not in the asymptotic range, multiple grid-triplet
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

studies often show oscillatory convergence. In such maneuvering simulations are carried out to obtain
cases, the estimated order of accuracy pRE approaches coefficients used in system-based models to predict
pth with oscillations and a wide range of values (Celik actual 6DOF maneuvers, mostly as part of the
SIMMAN 2008 workshop. The average of the fine grid
2008). The Richardson extrapolation method requires points is 6M with average error value of 13.6%D for
at least three systematic high-quality grids, which may forces and moments predictions. In the view of the
be too expensive for industrial applications. The progress on free running simulations, it is expected that
oscillatory convergence with grid refinement may be the future challenges and method development efforts
treated using the least-square (Eça and Hoekstra 2002) for modelling, numerical methods, and HPC will focus
or response-surface (Logan and Nitta 2006) methods, on free running rather than captive simulations.
which requires solutions for more than three grids. However, captive CFD will remain an important tool
There are some issues in using these two methods. The both for fundamental issues and local flow studies and
relationship between their estimates for the order of uncertainty analysis/optimization efforts.
accuracy, error estimate, and numerical benchmark and Due to the significant amount of information
those for individual grid-triplet studies is not estab- provided by CFD free running, the free running
lished. They do not discriminate between converging simulation is the future in computational ship hydrody-
and diverging grid studies and the use of diverging grid namics and captive simulations will be used only for
studies is not well founded. The requirement of at least limited purposes. Despite the progress in free running
four solutions is often too expensive for industrial simulations, there are still significant challenges ahead
applications. All the solutions are required to be in the that have to be addressed. The verification study is not
asymptotic range, which is contradictory to the use of performed systematically yet for any computations due
solutions that show oscillatory and non-monotonic to the complex grids structures of the hull and the
convergence. They also introduce additional uncertain- appendages. The validation studied showed fairly
ties due to the least-square fit. The difficulty and good agreement with EFD data for motions
computational cost associated with the Richardson (E=8.46%D). However, the validation for local flow is
extrapolation method may be resolved by the single- not conducted yet due to the complexity in the local
grid method (Celik and Hu 2004; Cavallo and Sinha flow measurement for free running ships. For more
2007). However, the sensitivity of the solutions to grid computer-intensive applications such as seakeeping
spacing and time step is not provided and control of the and route simulation, an extremely long solution time
spatial discretization error as the simulation progresses and a very large operating condition should be covered.
needs to be further investigated. Additionally, the For these applications, the speed of current CFD
applicability of the single-grid method to different solutions is still far too slow. Thus, using much faster
discretization schemes and turbulence models needs to method such as system based method could be
be validated. The available grid verification methodol- considered. However, the mathematical models for
ogies were developed for URANS (Stern et al. 2006; these methods should to be improved using high
Xing and Stern 2010; Xing and Stern 2011), and fidelity CFD solutions along with system identification
cannot be applied straightforwardly to LES and hybrid techniques. Furthermore, innovating numerical
RANS-LES models due to the coupling of modeling methods for easier and faster CFD solution is essential.
and numerical errors. Thus, new verification methods Moreover, taking advantage of faster computers such
need to be developed. Very recently, the method of as next generation of massively parallel multi-core
manufactured solutions is used to study the conver- machines should be considered.
gence characteristics of numerical solutions of highly Breaking waves, spray formation, and air en-
non-linear systems of partial differential equations such trainment around ships are of great importance to ship
as the RANS models (Eça et al., 2012). hydrodynamics. Previous studies are mainly focused on
Captive V&V simulations are carried out for a the global structure of ship flows, such as the wave
wide range of applications and have shown improve- elevation, scars, and vortices. The small scale details of
ment over the years. Calm water resistance, sinkage interface breaking, spray formation, and air entrain-
and trim simulations have achieved relatively small ment are not well understood. With the development of
error values for resistance, averaged at 3.3%D. For the CFD technology, detailed studies of the two-phase
sinkage and trim, the error values are larger for smaller region become possible. In the simulation of the wedge
Fr<0.2 (44.7%D), while for larger Fr>0.2 the average flow, the predicted overall bow wave profile (thin
error value is 10.3%D. The number of seakeeping liquid sheet at the leading edge, overturning jet and
simulations has increased since 2005, with average fine plunging, and splashes at the wake), plunging jet shape,
grid size of 15M. The average error value is relatively and surface disturbances on the wave crest are similar
small (15%) for 1st order terms while for 2nd order to the experimental observations. With 1.0 billion grid
terms the average error is very large (44%D). Captive points, the droplets and bubbles near the wedge can be
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

effectively captured. Further grid refinement (3 to 4 to be developed with focus on improvements of orders
billion grid points) is needed to increase the resolution of magnitude in accuracy, robustness, and perfor-
in the wake region. In CFD study of the plunging wave mance.
breaking over a submerged bump, the overall interface Current mainstream RANS solvers for ship
structure of the breaking wave match the experimental hydrodynamics are expected to continue performing
results very well, and the major events of the first well for even larger grids of up to a few hundreds of
plunge wave breaking, i.e., maximum height, first millions of points. However, there will be a threshold
plunge, oblique splash, and vertical jet are also that further increase of grid size cannot improve the
demonstrated. With a large grid of 2.2 billion points, results anymore because of the inherently limited
the experimental observed mean droplet/bubble size RANS turbulence models. Of course, it is possible to
can be effectively captured. Further grid refinement (up switch to second moment closure or hybrid RANS/LES
to 16 billion points) is needed in order to capture the such as DES; again, there will be a threshold of
minimum drop/bubble size observed in the experi- accuracy improvements from better turbulence
ments. 3D structure of the plunging wave breaking modeling due to the limited orders of accuracy of the
over the bump will also be investigated in the future discretization schemes. These solvers will replace
work. In the study of air-layer drag reduction, the air potential solvers to a large extent in the routine practice
layer along the entire test plate is successfully achieved of industrial ship design process. And just like the
and the drag reduction is approximately 100%, which potential solvers are run on today’s desktops with one
agrees with the experimental findings very well. With or a few processor cores, these RANS solvers will be
reduced air flow rate, BDR is also observed; the very likely to be run on desktops or workstations with a
computational results also qualitatively match the few hundreds of processors cores, including the GPU
experiments. The critical air flow rate to form the cores. GPU acceleration capability can be easily
ALDR is lower in the simulations than in the experi- obtained by using OpenMP-like compiler directives,
ments, which needs further investigations in the future such as OpenACC, with a compiler supporting them.
work. This step may not require major changes of the code.
However, it is almost sure that computer memory
10.2 Future directions capacity won’t increase in a similar scale as the
computing power from a many-core processor. In order
The oncoming exascale HPC era is to change our to achieve the same level of acceptance of potential
approaches to grand scientific and engineering solvers in ship design process using commodity
challenges and to transform modeling and simulation computers, it will be essential to greatly reduce
into a specified discipline of predictive science. Central memory usage in these RANS solvers. Unfortunately,
to achieving a predictive science in ship hydrodynam- the solution algorithms and discretization schemes in
ics is the development and application of verified and many of these solvers are well-known to require very
validated computational methodologies, capable of large amounts of memory, which may become the
utilizing exascale HPC platforms and predicting ship major issue hindering their daily applications in ship
resistance and propulsion, seakeeping, and maneuver- design process. Furthermore, RANS solvers are
ing, as well as breaking waves, turbulence, fluid- generally not accurate enough for tackling many
structure interactions, and signature, with quantified fundamental problems related to interfacial phenomena
uncertainty. High-fidelity, first-principles-based in ship hydrodynamics such as wave breaking,
simulations with unprecedented resolution can reveal turbulent contact line, bubbly wake, and cavitation,
vast unknown temporal-spatial correlations in multi- many of which have long been addressed separately
scale and multi-physics phenomena that are beyond using totally different tools without a ship flow context.
today’s computing capabilities. It is to revolutionize The modeling of these phenomena in many RANS
ship hydrodynamics research and along with optimiza- solvers is usually overly simplified and could not
tion techniques ship design process. On the other hand, provide much insight into the mechanism behind the
exascale platforms will be dramatically different from problems.
current mainstream supercomputers in terms of The next-generation, high-fidelity ship hydro-
computing power and parallel architectures; and few dynamics solvers have to be developed aiming at the
codes could make full use of their potentials without oncoming exascale computing platforms, and address-
major overhauls or even rewrite for minimized ing modeling issues, discretization schemes, and HPC
dependence on phenomenological/empirical models memory and scalability restraints at the same time.
and substantially improved scalability. To meet the However, it is not meant to embrace some totally
challenges of exascale computing for fundamental different, novel techniques only showing limited
studies and simulation-based design in ship hydrody- promise in some specific application areas. These new
namics, the next-generation high fidelity solvers have techniques may take several decades to reach the same
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

level of maturity of some widely accepted algorithms Amromin, E., and Mizine, I., "Partial Cavitation as
such as MAC grids for variable arrangement, projec- Drag Reduction Technique and Problem of Ac-
tion methods for velocity-pressure coupling, and VOF tive Flow Control," Marine Technology, Vol. 40,
methods for interface (volume) tracking. The problem 2003, pp. 181-188.
is, the first exascale computer has been projected to Araki, M., Sadat-Hosseini, H., Sanada, Y. Tanimoto,
become reality before 2020 and it is very likely that K., Umeda, N., and Stern, F., “Estimating Ma-
these new techniques won’t be ready to tackle real neuvering Coefficients Using System identifica-
world ship hydrodynamics problems by that time. tion Methods with Experimental, System-based,
Therefore, it is more sensible to adopt the advanced and CFD Free-running Trial Data”, Journal of
developments of long-tested modeling formulations Ocean Engineering, Vol. 51,2012a, pp. 63-84.
and numerical algorithms  this won’t be simple tasks Araki, M., Sadat-Hosseini, H., Sanada, Y., Umeda, N.,
by any means  from various research areas. For and Stern, F, “Study of System -based Mathemat-
example, the level set based sharp interface (ghost ical Model using System Identification Method
fluid) method for two phase flows has been around for with Experimental, CFD, and System-Based Free-
a while, but it has been seldom implemented and Running Trials in Wave”, Proceedings of the 11th
applied in ship flow related studies; geometrical VOF International Conference on the Stability of Ships
methods with very high accuracy of mass conservation and Ocean Vehicles, Athens, Greece, Sept 23-28,
have been rarely used either; similarly, semi- 2012b.
Lagrangian advection schemes have been developed to ASME Performance Test Codes Committee PTC 61,
a very sophisticated level coupling high order accura- "V&V 20: Standard for Verification and
cy, strongly conservative advection, and unconditional Validation in Computational Fluid Dynamics and
stability in time domain, but expensive, low-order, Heat Transfer," 2009.
implicit schemes have been predominated in ship Aulisa, E., Manservisi, S., Scardovelli, R., “A surface
hydrodynamics solvers including commercial and marker algorithm coupled to an area-preserving
open-source solvers. It is expected that the incorpora- marker redistribution method for three-
tion of these new algorithms in the high-fidelity solvers dimensional interface tracking,” Journal of Com-
to be developed for exascale computers will elevate the putational Physics, Vol. 197, Issue 2, 2004, pp.
state of the art of computational ship hydrodynamics to 555-584.
an unprecedented level, even each numerical technique Baker, T. J., “Automatic Mesh Generation for Complex
or solution algorithm is not completely newly invented. Three-Dimensional Regions using a Constrained
In addition, many critical choices can be made during Delaunay Triangulation,” Engineering with Com-
the solver envisagement and design stage by taking puters, Vol. 5, 1989, pp. 161–175.
into account many characteristic exascale issues such Balay, S., Buschelman, K., Gropp, W., Kaushik, D.,
as many-core processors, GPU accelerators, memory Knepley, M., Curfman, L., Smith, B. and Zhang,
constraints, etc. All these considerations are essential H. “PETSc User Manual”, ANL-95/11-Revision
for the realization of the next generation, high-fidelity 2.1.5, 2002, Argonne National Laboratory.
solvers for solving ship hydrodynamics problems on Bardina, J., Ferziger, J., Reynolds, W., “Improved
exascale computers. turbulence models based on large eddy simulation
of homogeneous, incompressible, turbulent
ACKNOWLEDGMENTS flows,” Technical Report TF-19. Department of
Mechanical Engineering, Stanford University,
This research was sponsored by the Office of Naval Stanford, California, 1983.
Research under Grant N000141-01-00-1-7, under the Batten, P., Goldberg, U., Chakravarthy, S., “Interfacing
administration of Dr. Patrick Purtell. statistical turbulence closures with large eddy
simulation,” AIAA J., Vol. 42(3), 2004, pp. 485–
REFERENCES 492.
Beale, K.L.C., Fu, T.C., Fullerton, A.M., Drazen, D.,
Ahn, H.T., Shashkov, M., “Adaptive moment-of-fluid Wyatt, D.C., “An Experimental Characterization
method,” Journal of Computational Physics, Vol. of Full-Scale Ship Generated Bow Spray,” 28th
228, Issue 8, 2009, pp. 2792-2821. Symposium on Naval Hydrodynamics, Pasadena,
Abdel-Maksoud, M., Menter, F. R., and Wuttke, H., California, 12–17 September 2010.
“Numerical Computation of the Viscous Flow Bhushan, S., Xing, T., Carrica, P., Stern, F., “Model-
around Series 60 CB =0.6 Ship with Rotating and Full-Scale URANS Simulations of Athena
Propeller”, Proc. 3rd Osaka Colloquium Ad- Resistance, Powering, Seakeeping, and 5415 Ma-
vanced CFD Applications to Ship Flow And Hull neuvering,” Journal of Ship Research, Vol. 53,
Form Design, Osaka, 1998, pp.25-50. No. 4, 2009, pp. 179–198.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Bhushan, S., Carrica, P.M., Yang, J., and Stern, F., Cahouet, J., “Etude numerique er experimentale du
“Scalability studies and large grid computations problem bidimensionnel de la resistance de
for surface combatant using CFDShip-Iowa,” In- vaques non-lineaire,” Ph.D. Thesis, ENSTA,
ternational Journal of High Performance Compu- 1984, Paris, (in French).
ting Applications, Vol. 25, No. 4, 2011a, pp. 466- Campana, E., Peri, D., Tahara, Y., Kandasamy, M., and
287. Stern, F., “Optimal ship design algorithms and
Bhushan, S., Hanaoka, A., Yang, J. and Stern, F., their application to industrial problems,” Transac-
“Wall-Layer Modeling for Cartesian Grid Solver tions SNAME, Vol. 117, 2009, awarded ABS-
Using an Overset Boundary Layer Orthogonal Captain Joseph H. Linnard Prize for best paper
Curvilinear Grid,” 49th AIAA Aerospace Scienc- 2009.
es Meeting, 2011b. Carrica, P.M., Drew, D., Bonetto, F. and Lahey, R.T.,
Bhushan, S., Carrica, P.M., Yang, J., and Stern, F., Jr., "A polydisperse model for bubbly two-phase
“Scalability studies and large grid computations flow around a surface ship," Int. J. Multiphase
for surface combatant using CFDShip-Iowa,” In- Flow, Vol. 25, 1999, pp. 257–305.
ternational Journal of High Performance Compu- Carrica, P., Wilson, R., Noack, R., and Stern, F., “Ship
ting Applications, Vol. 25, No. 4, 2011, pp. 466- Motions using Single-Phase Level Set with Dy-
287. namic Overset Grids,” Computers & Fluids, Vol.
Bhushan, S, Alam, M.F., Walters DK. “Application of 36, 2007, pp. 1415-1433.
dynamic hybrid RANS/LES model for straight Carrica, P.M., Ismail, F., Hyman, M., Shanti Bhushan,
ahead surface combatant vortical and turbulent S., Stern, F., “Turn and Zig-Zag manoeuvre of a
structure predictions,” ParCFD 2012, 22-26 May, surface combatant using a URANS approach with
2012a, Atlanta, US. dynamic overset grids”, Proceedings of SIMMAN
Bhushan, S., Mousaviraad, S.M., Stern, F., Doctors, 2008, 2008a.
L.J., “T-Craft Resistance, Seakeeping and Ma- Carrica PM, Paik K, Hosseini H, Stern F, “URANS
neuvering – Part I: Single-Phase Modeling and Analysis of a Broaching Event in Irregular Quar-
Verification and Validation for Captive Cases,” tering Seas”, J. Marine Sci. Tech., 2008b,13, 395-
Submitted to Applied Ocean Research, 2012b. 407.
Bhushan, S., Xing, T., Stern, F., “Vortical Structures Carrica, P.M., Stern, F., “DES Simulations of the
and Instability Analysis for Athena Wetted Tran- KVLCC1 in Turn and Zigzag Maneuvers with
som Flow with Full-Scale Validation,” Journal of Moving Propeller and Rudder”, Proceedings of
Fluids Engineering, Vol. 134, 2012c, 031201-1 – SIMMAN 2008, 2008.
18. Carrica PM, Castro A, Stern F, “Self-Propulsion
Bhushan, S., Walters DK. “A dynamic hybrid Computations Using Speed Controller and Dis-
RANS/LES modeling framework,” Physics of cretized Propeller with Dynamic Overset Grids”,
Fluids, Vol. 24, 015103, 2012. J. Marine Sci. Tech. 15, 2010a, 316-330.
Blenkinsopp, C. E., and Chaplin, J. R., “Void fraction Carrica, P. M. Huang, J., Noack, R., Kaushik, D.,
measurements in breaking waves,” Proceedings Smith, B., Stern, F., “Large-scale DES computa-
of the Royal Society A 463, 2007, pp. 3151-3170. tions of the forward speed diffraction and pitch
Bonmarin, P., “Geometric properties of deep-water and heave problems for a surface combatant,”
breaking waves”, J. Fluid Mech.,Vol. 209, 1989, Computers and Fluids, Vo. 39(7), 2010b, pp.
pp.405-433. 1095-1111.
Boris, J.P., Grinstein, F.F., Oran, E.S., Kolbe, R.L., Carrica, P.M., Fu, H., and Stern, F., “Self-Propulsion
“New Insights into Large Eddy Simulation,” Flu- Free to Sink and Trim and Pitch and Heave in
id Dyn. Res., Vol. 10, 1992, pp. 199-209. Head Waves of a Kcs Model”, G2010: A Work-
Briley, W., Arabshahi, A, Webster, R.S., and Whit- shop on CFD in Ship Hydrodynamics, Chalmers,
field, D. L. "Computational Analysis and Ad- Gothenburg, December 8-10, 2010c.
vancement of Solid Rocket Motor (SRM) Model- Carrica, P., Sadat-Hosseini, H., and Stern, F., “CFD
ing and Simulation Tools," Monthly Technical Analysis of Broaching for a Model Surface Com-
Report for period February 2005-July 2006, batant with Explicit Simulation of Moving Rud-
SPARTA, INC. 25531 CA 92630. ders and Rotating Propellers,” Computers & Flu-
Broglia, R., Di Mascio, A., Muscari, R., “Numerical ids, Volume 53, 2012a, pp. 117-132.
simulations of breaking wave around a wedge,” Carrica, P.M., Ismail, F., Hyman, M., Bhushan, S.,
Proceedings of 25th symposium on naval hydro- Stern, F., “Turn and Zigzag Maneuvers of a Sur-
dynamics, St. John’s Newfoundland and Labra- face Combatant Using a URANS Approach with
dor, Canada, 2004. Dynamic Overset Grids,” Journal of Marine Sci-
ence and Technology, 2012b.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Castiglione, T., Stern, F., Bova, S., Kandasamy, M., Dommermuth, D.G., O’Shea, T.T. , Wyatt, D.C., et al.,
2011, “Numerical investigation of the seakeeping An application of Cartesian-grid and volume-of-
behavior of a catamaran advancing in regular fluid methods to numerical ship hydrodynamics,
head waves,” Ocean Engineering, Volume 38, in: Proc. Nineth Int. Conf. Numer. Ship Hydrody-
Issue 16, Pages 1806–1822. namics, Ann Arbor, MI, 2007.
Cavallo, P. A., and Sinha, N., "Error quantification for Dong, R. R., Katz, J. and Huang, T. T., “On the
computational aerodynamics using an error structure of bow waves on a ship model,” Journal
transport equation," Journal of Aircraft, 44(6), of Fluid Mechanics, 346, 1997, pp.77-115.
2007, pp. 1954-1963. Dong, S., Shen, J., “A time-stepping scheme involving
Ceccio, S.L, "Evaluation of Air Cavity Lubrication at constant coefficient matrices for phase-field simu-
High Reynolds Numbers," Air Layer Drag Reduc- lations of two-phase incompressible flows with
tion Program Review Meeting, Stanford Universi- large density ratios,” Journal of Computational
ty, Stanford, California, September 15, 2009. Physics, Vol. 231, Issue 17, 2012, pp. 5788-5804.
Celik, I. B., Ghia, U., Roache, P. J., Freitas, C.J., Dong, S., Shen, J., “An unconditionally stable
Coleman, H., Raad, P. E. , "Procedure for Estima- rotational velocity-correction scheme for incom-
tion and Reporting of Uncertainty Due to Dis- pressible flows,” Journal of Computational Phys-
cretization in CFD Applications," Journal of Flu- ics, Vol. 229, Issue 19, 2010, pp.7013-7029.
ids Engineering, 130(7), 2008, p. 078001. Drazen, D. A. and Melville, W. K., “Turbulence and
Celik, I., and Hu, G., "Single Grid Error Estimation mixing in unsteady breaking surface waves,”
Using Error Transport Equation," Journal of Flu- Journal of Fluid Mechanics, 628, 2009, pp. 85-
ids Engineering, 126(5), 2004, p. 778. 119.
Chang, K. A. and Liu, P. L. F., “Experimental investi- Drazen, Fulletron, Fu, Beale, O'Shea, Bruker, Dom-
gation of turbulence generated by breaking waves mermuth, Wyatt, Bhushan, Carrica and Stern, "A
in water of intermediate depth,” Physics of Fluids comparison of model-scale experimental meas-
11, 1999, pp. 3390-3400. urements and computational predictions for a
Chanson, H., Fang L., “Plunging jet characteristics of large transom-stern wave," 28th Symposium on
plunging breakers,” Coastal Engineering, 31, Naval Hydrodynamics, Pasadena, California, 12-
1997, pp. 125-141. 17 September 2010.
Chen, G., Kharif, C., Zaleski, S., Li J., “Two- Dreyer, J.J. and Boger, D.A., “Validation of a Free-
dimensional Navier-Stokes simulation of break- Swimming, Guided Multibody URANS Simula-
ing waves,” Physics of Fluids ,11, 1999, pp. 121- tion Tools”, 28th Symposium on Naval Hydrody-
133. namics, Pasadena (Ca), 2010.
Cho, S., Hong, S., Kim, Y., “ Investigation of dynamic Drouet, A., Jacquin, E. et al., ''Simulation of unsteady
characteristics of the flooding water of the dam- ship maneuvering on calm water and in waves
aged compartment of an ITTC RoRo-Passenger”, using free-surface RANS solver'', 27th Symposi-
J. of the Soc. of Naval Architects of Korea, Vol. um on Naval Hydrodynamics, Seoul, South Ko-
43, Issue 4, 2006, pp.451-459 rea, 2008.
Choi, H. and Moin, P., “Effects of the computational Dubbioso, G., Durante, D., Broglia, R., Mauro, S.,
time step on numerical solutions of turbulent “Comparison of experimental and CFD results for
flow”, J. Comput. Phys., Vol. 113, 1994, pp. 1-4. a tanker-like vessel”, Proceeding of MARSIM
Cohen, J., Molemaker, J., “A Fast Double Precision 2012, 2012.
CFD Code using CUDA,” 21st Parallel CFD Ducan, J.H., “An experimental investigation of
Conference, May 18-22, Moffett Field, Califor- breaking waves produced by a towed hydrofoil,”
nia, USA, 2009, pp. 252-256. Proc. R. Soc. London Ser. A, 377, 1981, pp. 331-
Cosner, R. R., Oberkampf, W. L., Rumsey, C. L., 348.
Rahaim, C., and Shih, T., "AIAA Committee on Ducan, J.H., “The breaking and nonbreaking wave
standards for computational fluid dynamics: sta- resistance of a two-dimensional hydrofoil,” Jour-
tus and plans," AIAA paper, 889, 2006. nal of Fluid Mechanics, 126, 1983, pp. 507-520.
Deane, G.B., Stokes, M.D., “Scale dependency of Durante, D., Broglia, R., Muscari, R. and Di Mascio,
bubble creation mechanisms in breaking waves,” A., “Numerical simulations of a turning circle
Nature, Vol. 418, 2002, pp. 839–844. maneuver or a fully appended hull”, 28th Sympo-
Di Mascio, A., Broglia, R., Muscari, R., On the sium on Naval Hydrodynamics, Pasadena (Ca),
application of the single-phase level set method to 2010.
naval hydrodynamic flows, Computers and Flu- Eça, L., and Hoekstra, M., "An Evaluation of Verifica-
ids, 36 (5), 2007, pp. 868-886. tion Procedures for CFD Applications," 24th
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Symposium on Naval Hydrodynamics, Fukuoka, Gatski, T., Jongen, T., “Nonlinear eddy viscosity and
Japan, 2002. algebraic stress models for solving complex tur-
Eça, L., and Hoekstra, M., "Discretization uncertainty bulent flows,” Progress in Aerospace Sciences,
estimation based on a least squares version of the Vol. 36(8), 2000, pp. 655–682.
grid convergence index," Proc. Proceedings of the Ghosh, S., “Free surface instabilities and plunging
Second Workshop on CFD Uncertainty Analysis, breaking wave downstream of a bump in shallow
Instituto Superior Tecnico, 2006. water open channel flume,” Ph.D. Thesis, 2008,
Eça, L., Vaz, G., Hoekstra, M., "Assessing Conver- The University of Iowa, USA.
gence Properties of RANS Solvers with Manufac- Gibou, F., Min, C., “Efficient symmetric positive
tured Solutions," European Congress on Compu- definite second-order accurate monolithic solver
tational Methods in Applied Sciences and Engi- for fluid/solid interactions,” Journal of Computa-
neering (ECCOMAS 2012), Vienna, Austria, tional Physics, Vol. 231, Issue 8, 2012, pp. 3246-
2012. 3263.
Elbing, B.R., Winkel, E.S., Lay, K.A., Ceccio, S.L., Gicquel, L.Y.M., Staffelbach, G., Cuenot, B., Poinsot,
Dowling, D.R., and Perlin, M., "Bubble-Induced T., “Large eddy simulations of turbulent reacting
Skin-Friction Drag Reduction and the Abrupt flows in real burners: the status and challenges,”
Transition to Air-Layer Drag Reduction," J. Fluid Journal of Physics: Conference Series, Vol. 125,
Mech., Vol. 612, 2008, pp. 201-236. 2008, 012029.
Enright, D., Fedkiw, R., Ferziger, J., Mitchell, I., “A Gilmanov, A., Sotiropoulos, F., and Balaras, E., “A
Hybrid Particle Level Set Method for Improved general reconstruction algorithm for simulating
Interface Capturing,” Journal of Computational flows with complex 3d immersed boundaries on
Physics, Vol. 183, Issue 1, 2002, pp. 83-116. Cartesian grids”, J. Comput. Phys., Vol. 191,
2003, pp. 660-669.
Falgout, R.D., Jones, J.E., and Yang, U.M., “The Girimaji, S., Jeong, E., Srinivasan, R., 2006. “Partially
design and implementation of HYPRE, a library averaged Navier-Stokes method for turbulence:
of parallel high performance preconditioners”, in Fixed point analysis and comparison with un-
Numerical Solution of Partial Differential Equa- steady partially averaged Navier-Stokes,” Journal
tions on Parallel Computers, A.M. Bruaset and A. of Applied Mechanics, Vol. 73, 2006, pp. 422-
Tveito, eds., Springer-Verlag, Vol. 51, 2006, pp. 429.
267-294. Greco, M., Landrini, M., Faltinsen, O.M., “Impact
Ferrant P., Gentaz L., Monroy C., Luquet R., Ducrozet flows and loads on ship-deck structures,” Journal
G., Alessandrini B., Jacquin E. and Drouet A., of Fluids and Structures, 19 (3), 2004, pp.251–
“Recent Advances Towards the Viscous Flow 275.
Simulation of Ships Manoeuvring in Waves”, Grue, J., Jensen, A., “Experimental velocities and
Proc. of 23rd International Workshop on Water accelerations in very steep wave events in deep
Waves and Floating Bodies, Jeju, Korea, April water,” European Journal of Mechanics B/Fluids
2008. 25, 2006, pp. 554-564.
Fossen, T. I., “Guidance and Control of Ocean Hanjalic, K., “Will RANS Survive LES? A view of
Vehicles,” (John Wiley & Sons Ltd.), 1994. perspective,” Journal of fluids engineering, Vol.
Fu, H., Michael, T., Carrica, P.M., “A Method to 127, 2005, pp. 831-839.
Perform Self-Propulsion Computations with a Heredero, P. A., Xing, T., and Stern, F., “URANS and
Simplified Body-Force Propeller Model”, G2010: DES for Wigley hull at extreme drift angles,”
A Workshop on CFD in Ship Hydrodynamics, Journal of Marine Science and Technology,
Chalmers, Gothenburg, December 8-10, 2010. 15(4), 2010, pp. 295-315.
Fureby, C. Towards the use of large eddy simulation in Hino, T. Ohashi, K, Kobayashi, H. “Flow Simulations
engineering, Progress in Aerospace Sciences, Using Navier-Stokes Solver Surf”, Proc. of
Vol. 44, 2008, pp. 381–396. G2010 Workshop, 2010.
G2010, Gothenburg 2010 A Workshop on CFD in Ship Hoekstra M., “Numerical Simulation of Ship Stern
Hydrodynamics, Chalmers, Gothenburg, Decem- Flows with a Space-marching Navier-Stokes
ber 8-10 2010. Method,” PhD Thesis, Delft 1999.
Gao, Q., Vassalos, D., “ Numerical Study of the Roll Horiuti K., “A new dynamic two-parameter mixed
Decay of Intact and Damaged Ships”, 12th Inter- model for large-eddy simulation,” Physics of Flu-
national Ship Stability Workshop, 2011. ids, Vol. 9(11), 1997, pp. 3443– 3464.
Gao, Z., Gao, Q., Vassalos, D., “Numerical simulation Horiuti, K., “Roles of non-aligned eigenvectors of
of flooding of a damaged ship”, Ocean Engineer- strain-rate and subgrid-scale stress tensors in tur-
ing, Vol, 38, 2011, pp. 1649-1662.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

bulence generation”, Journal of Fluid Mechanics, grids with applications to KVLCC2 tanker,” Int.
Vol. 491, 2003, pp. 65-100. J. Numer. Meth. Fluids, Vol. 64, pp. 850–886.
Hough, G. and Ordway, D., “The generalized actuator ITTC 2011, The specialist Committee on computation-
disk”, Developments in Theoretical and Applied al fluid dynamics, Proceedings of 26th Interna-
Mechanics, 1965, 2:317-336. tional Towing Tank Conference, Rio de Janeiro,
Hu, C. and Kashiwagi, M., “Numerical and experi- Brazil, 28 August – 3 September, 2011.
mental studies on three-dimensional water on Jacquin, E., Guillerm, P.E., Drouet, A., Perdon, P., and
deck with a modified Wigley model”, Proc. 9th Alessandrini, B., 2006, “Simulation of unsteady
Inter. Conf. Numer. Ship Hydrodynamics, Ann ship maneuvering using free-surface RANS solv-
Arbor, Michigan, 2007. er,” 26th Symposium on Naval Hydrodynamics,
Hu, C., Kashiwagi, M., “CIP Based Cartesian Grid Rome, Italy.
Method for Prediction of Nonlinear Ship Mo- Jemison, M., Loch, E., Sussman, M., Shashkov, M.,
tions”, Proc. of G2010 Workshop, 2010a. Arienti, M., Ohta, M., Wang, Y., “A Coupled
Hu, C-H., Sueyoshi, M., Kashiwagi, M., “Numerical Level Set-Moment of Fluid Method for Incom-
Simulation of Strongly Nonlinear Wave-Ship In- pressible Two-Phase Flows,” Journal of Scientific
teraction by CIP based Cartesian Grid Method,” Computing, in press, 2012, 10.1007/s10915-012-
International Journal of Offshore and Polar Engi- 9614-7.
neering, 2010b (June), Vol. 20, No.2, pp.81-87 Jeong, J. Hussain, F., “On the identification of a
Huang J., Carrica P., Stern F., “Semi-coupled air/water vortex”, Journal of Fluid Mechanics, Vol. 285,
immersed boundary approach for curvilinear dy- 1995, pp. 69–94.
namic overset grids with application to ship hy- Jiang, G.-S. and Peng, D., “Weighted ENO schemes
drodynamics”, International Journal Numerical for Hamilton-Jacobi equations”, SIAM J. Sci.
Methods Fluids, Vol. 58, 2008, pp. 591-624. Comp., Vol. 21, 2000, pp. 2126-2143.
Huang, J., Carrica, P., Stern, F., “Coupled ghost Jiang, G.-S. and Shu, C.-W., “Efficient implementation
fluid/two-phase level set method for curvilinear of weighted ENO schemes”, J. Comput. Phys.,
body-fitted grids,” International Journal Numeri- Vol. 126, 1996, pp. 202-228.
cal Methods Fluids, Vol. 55, Issue 9, November Jimenez, J. “Turbulent Flows over Rough Walls,”
2007, pp. 867-897. Annual Review of Fluid Mechanics, Vol. 36,
Huang, J., Carrica, P., Stern, F., “Numerical Study of a 2004, pp. 173-196.
Ship Exhaust Plume with Waves and Wind,” In- Kalitzin, G., Medic, G., Iaccarino, G., Durbin, P.
ternational Journal Numerical Methods Fluids, “Near-wall behavior of RANS turbulence models
Vol. 68, Issue 2, January 2012a, pp. 160-180. and implications for wall functions,” Journal of
Huang, J., Carrica, P. M. and Stern, F. A geometry- Computational Physics, Vol. 204, 2005, pp. 265-
based level set method for curvilinear overset 291.
grids with application to ship hydrodynamics. In- Kandasamy, M., He, W., Takai, T., Tahara, Y., Peri,
ternational Journal Numerical Methods Fluids, D., Campana, E., Wilson, W., and Stern, F., “Op-
Vol. 68, 2012b, pp. 494-521. timization of Waterjet Propelled High Speed
Hunt, J.C.R., Wray, A. A., Moin, P. “Eddies, streams, Ships -Jhss and Delft Catamatran,” 11th Interna-
and convergence zones in turbulent flows, in: Its tional Conference on Fast Sea Transportation,
Studying Turbulence Using Numerical Simula- FAST 2011, Honolulu, Ha wa ii, USA,
tion Databases”, Vol. 2, Proceedings of the 1988 September 2011.
Summer Program (SEE N89-24538 18-34), 1988, Kandasamy, M., Ooi, S.K., Carrica, P., & Stern, F.,
pp. 193–208. “Integral force/moment water-jet model for CFD
Iafrati, A., “Air-water interaction in breaking wave simulations,” Journal of Fluids Engineering, 132,
events: quantitative estimates of drops and bub- 2010, 101103-1 - 9.
bles,” In: Proceedings of 28th Symposium on Na- Kandasamy, M., Xing, T., Stern, F., "Unsteady Free-
val Hydrodynamics, Pasadena, California, 12-17 Surface Wave-Induced Separation: Vortical
September 2010. Structures and Instabilities", Journal of Fluids and
Iafrati, A., Di Mascio, A., Campana, E. F., “A level set Structures, Vol. 25, Issue 2, 2009, pp. 343-363.
technique applied to unsteady free surface flows,” Kang, D.H., Longo, J., Marquardt, M., and Stern, F.,
International Journal for Numerical Method and “Solid/free-surface juncture boundary layer and
Fluids, 35, 2001, 281-297. wake with waves”, Proc. 27th ONR Symposium
Ismail, F., Carrica, P.M., Xing, T., Stern, F., 2010, on Naval Hydrodynamics, Seoul, Korea, 2008.
“Evaluation of linear and nonlinear convection Kang, D., Ghosh, S., Reins, G., Koo, B., Wang, Z., and
schemes on multidimensional non-orthogonal Stern, F., “Impulsive plunging wave breaking
downstream of a bump in a shallow water
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

flume—Part I: Experimental observations,” Jour- ciplinary Analysis and Optimization (MDO).”

nal of Fluids and Structures, Vol. 32, 2012, pp. 21st Parallel CFD Conference, May 18-22, Mof-
104-120. fett Field, California, USA, 2009, pp .338-342.
Karion, A., Fu, T. C., Sur, T.W., Rice, J.R., Walker, Larsson, L. Stern F. Visonneau, M (Editors), Proceed-
D.C. and Furey, D.A., “Experiment to examine ings of G2010 Workshop on Numerical Ship Hy-
the effect of scale on a breaking bow wave”, drodynamics, 2010.
Carderock Division, Naval Surface Warfare Cen- Larsson, L., and Zou, L., 2010, “CHAPTER 2 -
ter, Hydromechanics Research and Development Evaluation of Resistance, Sinkage and Trim, Self
report, 2004, NSWCCD-50-T R-2004/060. Propulsion and Wave Pattern Predictions,”
Karman, Jr., S.L., and Wilson, R., 2008, “Hierarchical Gothenburg 2010 workshop on CFD in Ship Hy-
Unstructured Mesh Generation with General Cut- drodynamics.
ting for Free Surface Simulations,” Proceedings Larsson, L., Stern, F., and M. Visonneau, “CFD in Ship
of the 27th ONR Symposium on Naval Hydrody- Hydrodynamics – Results of the Gothenburg
namics, Seoul, Korea, 5-10 Oct. 2008. 2010 Workshop,” Proc. IV International Confer-
Kassinos, S. C., Langer, C. A., Kalitzin, G. Iaccarino, ence on Computational Methods in Marine Engi-
G., “A simplified structure-based model using neering, 28-30 September 2011, Lisbon, Portugal.
standard turbulence scale equations: computation Lentine, M., Grétarsson, J.T., Fedkiw, R., “An
of rotating wall-bounded flows,” International unconditionally stable fully conservative semi-
Journal of Heat and Fluid Flow, Vol. 27(4), 2006, Lagrangian method,” Journal of Computational
pp. 653–660 . Physics, Vol. 230, Issue 8, 2011, pp. 2857-2879.
Kaushik, D., Balay, S. Keyes, D., Smith, B., “Under- Leonard, B. P., “A stable and accurate modelling
standing the Performance of Hybrid procedure based on quadratic interpolation”,
MPI/OpenMP programming Model for Implicit Comput. Methods Appl. Mech. Engrg., Vol. 19,
CFD Codes,” 21st Parallel CFD Conference, May 1979, pp. 58-98
18-22, Moffett Field, California, USA, 2009, pp. Li, T., and Matusiak, J., “Simulation of a modern
174-177. surface ship with a wetted transom in a viscous
Kerwin, J.E., Lee, C.S., “Prediction of Steday and flow,” Proc. of the 11th International Offshore
Unsteady Marine Propeller Performance by Nu- and Polar Engineering Conference, Stavanger,
merical Lifting-Surface Theory”, SNAME Trans- Norway, June 17-22, 2001, pp. 570-576.
actions, Vol. 86, 1978, pp.218-253. Lilly, D. K., 1992. “A proposed modification of the
Kim, D, Moin, P., “Direct numerical study of air layer Germano subgrid-scale closure method,” Physics
drag reduction phenomenon over a backward- of Fluids A, Vol. 4, 1992, pp. 633-635.
facing step,” Center for Turbulence Research, Lin, S. P., Reitz, R. D., Drop and spray formation from
Annual Research Briefs, 2010, pp. 351-363. a liquid jet, Annual Review of Fluid Mechanics,
Kim, S.-E., Choudhury D., “A near-wall treatment Vol. 30, 1998, pp. 85-105.
using wall functions sensitized to pressure gradi- Liovic, P., Lakehal, D., “Multi-physics treatment in the
ent, Separated and Complex Flows,” ASME, vicinity of arbitrarily deformable gas–liquid inter-
FED-Vol. 217, 1995, 273-280. faces,” J. Comput. Phys., 222, 2007, pp. 504–
Kimmoun, O. and Branger, H., “A particle image 535.
velocimetry investigation on laboratory surf-zone Liu, S., Kermani, A., Shen, L. and Yue, D., “Investiga-
breaking waves over a sloping beach,” Journal of tion of coupled air-water turbulent boundary lay-
Fluid Mechanics, 588, 2007, pp. 353-397. ers using direct numerical simulations,” Phys.
Knupp, P. M., and Salari, K., Verification of computer Fluids, 21, 2009, 062108.
codes in computational science and engineering, Logan, R. W. and Nitta, C.K., "Comparing 10 Methods
CRC Press, 2003. for Solution Verification, and Linking to Model
Koo, B., Wang, Z., Yang, J., and Stern, F., “Impulsive Validation," Journal of Aerospace Computing,
plunging wave breaking downstream of a bump in Information, and Communication, 3(7), pp. 354-
a shallow water flume—Part II: Numerical simu- 373, 2006.
lations,” Journal of Fluids and Structures, Vol. 32, Longo, J., Huang, H.P., and Stern, F., “Solid-fluid
2012, pp. 121-134. juncture boundary layer and wake”, Experiments
Kraichnan, R.H. “Eddy viscosity in two and three in Fluids, Vol. 25(4), 1998, pp. 283-297.
dimensions,” J. Atmos. Sci., Vol. 33, 1976, pp. Longo, J., Shao, J., Irvine, M., and Stern, F., “Phase-
1521–1536. averaged PIV for the nominal wake of a surface
Kremenetsky, M. and Kodiyalam, S., “Parallel ship in regular head waves”, J. Fluids Eng.,
Performance of CFD Algorithm and the Ubiqui- Vol.129, 2007, pp.524-540.
tous Need for HPC with High Fidelity, Multidis-
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Longuet-Higgins, M., "On the Disintegration of the Jet equations on non-graded adaptive grids,” Journal
in a Plunging Breaker", Journal of Physical of Computational Physics, Vol. 219, Issue 2,
Oceanography, Vol. 25, 1995, pp. 2458–2462. 2006, pp. 912-929.
Lubin, P., Vincent, S., Abadie, S., Caltagirone, J. P., Miyata, H. and Inui, T., “Nonlinear ship waves,”
“Three-dimensional large eddy simulation of air Advances in Applied Mechanics, 24, 1984, pp.
entrainment under plunging breaking waves,” 215-288.
Coastal Engineering, 53, 2006, pp.631-655. Miyata, H., Matsukawa, C., and Kajitani, H., “Shallow
Lübke, L.O., “Numerical Simulation of the Flow water flow with separation and breaking wave,”
around the Propelled KCS”, Proc. CFD Work- Autumn meeting of Naval Architecture, 1985a,
shop Tokyo 2005, Tokyo, Japan. Japan.
Marcum, D. L., “Generation of Unstructured Grids for Miyata, H., Nishimura, S., Masuko, A., “Finite
Viscous Flow Applications,” AIAA Paper 95- difference simulation of nonlinear waves generat-
0212, 1995. ed by ships of arbitrary three-dimensional config-
Marrone, S., Colagrossi, A., Antuono, M., Lugni, C., uration,” Journal of Computational Physics, Vol.
and Tulin, M.P., “A 2D + t SPH model to study 60, Issue 3, 1985b, pp. 391-436.
the breaking wave pattern generated by fast Mousaviraad, S.M., Carrica, P.M., Huang, J., and
ships,” Journal of Fluids and Structures, 27, 2011, Stern, F., “CFD Prediction of Ship Response to
pp. 1199-1215. Severe Ocean Waves and Wind”, 27th Symposi-
Mattor, N., Williams, T.J., and Hewett, D.W., “Algo- um on Naval Hydrodynamics, Seoul, South Ko-
rithm for solving tridiagonal matrix problems in rea, 2008.
parallel,” Parallel Comput., Vol. 21, 1995, pp. Mousaviraad, S.M., Carrica, P.M., and Stern, F.,
1769-1782. “Development and Validation of Harmonic Wave
Melville, W. K., “Energy dissipation by breaking Group Single-Run Procedures for RAO with
waves,” Journal of Physical Oceanography, 24, Comparison to Regular and Transient Wave
1994, 2041-2049. Group Procedures Using URANS,” Ocean Engi-
Melville, W. K., Veron, F. and White, C. J., “The neering, Vol. 37, Issues 8-9, June 2010, pp.653-
velocity field under breaking waves: coherent 666.
structures and turbulence,” Journal of Fluid Me- Mousaviraad, S.M., “CFD prediction of ship response
chanics, 454, 2002, pp. 203-233. to extreme winds and/or waves,” PhD diss., Uni-
Meneveau, C., Katz, J., “Scale-invariance and turbu- versity of Iowa, 2010.
lence models for large-eddy simulations,” Annual Mousaviraad, S.M., Sadat-Hosseini, S.H., Carrica,
Review of Fluid Mechanics, Vol. 32, 2000, pp. 1– P.M., Stern, F., “URANS Studies of Ship-Ship
32. Interactions in Calm Water and Waves for Re-
Menter, F. R. Egorov, Y., 2010. “The scale-adaptive plenishment and Overtaking Conditions”, Second
simulation method for unsteady turbulent flow International Conference on Ship Maneuvering in
predictions. Part 1: Theory and model descrip- Shallow and Confined Water: Ship-to-Ship Inter-
tion” Flow, Turbulence and Combustion, Vol. 85, action, Trondheim, Norway, 18-20 May 2011.
2010, pp. 113-127.. Mousaviraad, S.M., Bhushan, S., Stern, F., “CFD
Menter, F.R. Two-Equation Eddy-Viscosity Turbu- prediction of free-running SES/ACV deep and
lence Models for Engineering Applications. shallow water maneuvering in calm water and
AIAA Journal, Vol. 32 (8), 1994, pp. 1598-1605. waves”, Proceeding of MARSIM2012, 2012a.
Merkle, C., and Deutsch, S., "Microbubble Drag Mousaviraad, S.M., Cook, S.S., Carrica, P.M., Toda,
Reduction in Liquid Turbulent Boundary Layers," Y., Stern, F., “Complimentary EFD and CFD on
Appl. Mech. Rev., Vol. 45, No. 3, 1992, pp. 103- Effects of Headwinds on Towing Tank Resistance
127. and PMM Tests for ONR Tumblehome,” 29th
Metcalf, B., Longo, J., Ghosh, S., Stern, F., “Unsteady Symposium on Naval Hydrodynamics, 2012b,
free-surface wave-induced boundary-layer separa- Gothenburg, Sweden.
tion for a surface-piercing NACA 0024 foil: tow- Muscari, R., Broglia, R., Di Mascio, A, “Trajectory
ing tank experiments”, J. Fluids Struct., Vol. 22, prediction of a self-propelled hull by unsteady
2006, pp. 77-98. RANS computations”, 27th Symposium on Naval
Min, C., Gibou, F., “A second order accurate level set Hydrodynamics, Seoul, Korea, 2008a.
method on non-graded adaptive Cartesian grids,” Muscari, R., Broglia, R., Di Mascio, A, “ANALYSIS
Journal of Computational Physics, Vol. 225, Issue OF THE FLOW AROUND A MANOEUVRING
1, 2007, pp.300-321. VLCC”, Proceedings of the ASME 27th Interna-
Min, C., Gibou, F., “A second order accurate projec- tional Conference on Offshore Mechanics and
tion method for the incompressible Navier–Stokes Arctic Engineering, Estoril, Portugal, 2008b.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Muscari R., Felli M., Di Mascio A., “Numerical and Peregrine, D. H., “Breaking Waves on Beaches”, Ann.
experimental analysis of the flow around a pro- Rev. of Fluid Mech., Vol. 15, 1983, pp. 149-178.
peller behind a fully appended hull”, 28th Sym- Phillips, A.B., Turnock, S.R., and Furlong, B., 2009,
posium on Naval Hydrodynamics, Pasadena, Cal- “Evaluation of manoeuvring coefficients of a self-
ifornia, 2010. propelled ship using a blade element momentum
Noack, R.W. “SUGGAR: a general capability for propeller model coupled to a Reynolds averaged
moving body overset grid assembly,” 17th AIAA Navier Stokes flow solver,” Ocean Engineering,
Computational Fluid Dynamics Conference, Vol. 36, pp. 12 17–1225.
2006, Toronto, Ontario, Canada. Piomelli, U., Balaras, E., “Wall-layer models for large-
Noblesse, F., Delhommeau G., Guilbaud, M., Hendrix, eddy simulations,” Ann. Rev. Fluid Mech., Vol.
D., Yang, C., “Simple analytical relations for ship 34, 2002, pp. 349–374 .
bow waves,” Journal of Fluid Mechanics, Vol. Piomelli, U., Balaras, E., Pasinato, H., Squaires, K.D.,
600, 2008, pp. 105-132. Spalart, P.R., “The inner-outer layer interface in
Oberkampf, W. L. and C. J. Roy. Verification and large-eddy simulations with wall-layer models,”
Validation in Scientific Computing, Cambridge Int. J. Heat Fluid Flow, Vol. 24, 2003, pp. 538–
Univ Pr., 2010. 550.
Oh, K.J., Kang, S.H., “Full Scale Reynolds Number Pope S. B., “Turbulent Flows,” Cambridge University
Effects for the Viscous Flow around the Ship Press: Cambridge, MA, 2000.
Stern,” Computational Mechanics, Vol. 9, 1992, Pope, S. B., “The calculation of turbulent recirculating
pp. 85-94. flows in general orthogonal coordinates”, J.
Oger, G., Doring, M., Alessandrini, B., Ferrant, P., Comput. Phys., Vol. 26, 1978, pp. 197-217.
“Two-dimensional SPH simulations of wedge wa- Posa, A., Lippolis, A., Verzicco, R., Balaras, E.,
ter entries,” Journal of Computational Physics, “Large-eddy simulations in mixed-flow pumps
Vol. 213, Issue 2, 2006, pp. 803-822. using an immersed-boundary method,” Comput-
Olivieri, A., Pistani, F., Wilson, R., Campana, E., and ers & Fluids, Vol. 47, Issue 1, 2011, pp. 33-43.
Stern, F., “Scars and vortices induced by ship Qiu, J.-M., Shu, C.-W., “Conservative high order semi-
bow wave breaking”, J. Fluids Eng., Vol. 129, Lagrangian finite difference WENO methods for
2007, pp. 1445-1459. advection in incompressible flow,” Journal of
Olsson, E., Kreiss, G., “A conservative level set Computational Physics, Vol. 230, Issue 4, 2011,
method for two phase flow,” Journal of Computa- pp. 863-889.
tional Physics, Vol. 210, Issue 1, 2005, pp. 225- Queutey P, Visonneau M., “An interface capturing
246. method for free-surface hydrodynamic flows,”
Orihara, H. and Miyata, H., "A Numerical Method for Computers & Fluids, Vol. 36(9), 2007, pp. 1481–
Arbitrary Ship Motions in Arbitrary Wave Condi- 1510.
tions using Overlapping Grid System," Proceed- Rabenseifner, R., Wellein, G., “Communication and
ings of the 8th International Conference on Nu- Optimization Aspects of Parallel Programming
merical Ship Hydrodynamics, Busan, KOREA, Models on Hybrid Architectures,” The Interna-
2003. tional Journal of High Performance Computing
Park I. R. and Chun H. H., “A Study on the Level-Set Applications, Vol. 17(1), 2003, pp. 49-62.
Scheme for the Analysis of the Free Surface Flow Rhee S.H., Stern F., “RANS model for spilling
by a Finite Volume Method”, Journal of the Soci- breaking waves,” J Fluid Eng, Vol. 124, 2002.
ety of Naval Architects of Korea, 36, No. 2, Ripley, M. D., and Pauley, L. L., “The unsteady
(1999), pp.40-49. structure around a surface piercing strut,” Physics
Patel, V. C., “Perspective: Flow at high Reynolds of Fluids A, 5(12), 1993, pp. 3099-3106.
number and over Rough surfaces- Achilles Heel Roache, P. J., “Verification and Validation in Compu-
of CFD,” Journal of Fluids Engineering, Vol. 120, tational Science and Engineering,” Hermosa Pub-
1998, pp. 434-444. lishers, New Mexico, 1998.
Paterson, E.G., Wilson, R.V., and Stern, F., “General- Roache, P. J., "Code Verification by the Method of
Purpose Parallel Unsteady RANS Ship Hydrody- Manufactured Solutions," Journal of Fluids Engi-
namics Code: CFDSHIP-IOWA,” Iowa Institute neering, 124(1), 2002, pp. 4-10.
of Hydraulic Research, The University of Iowa, Roache, P. J.. "Discussion:“Factors of Safety for
IIHR Report No.432, November 2003, 105 pp. Richardson Extrapolation”(Xing, T., and Stern,
Peng, D., Merriman, B., Osher, S., Zhao, H., and Kang, F., 2010, ASME J. Fluids Eng., 132, p. 061403)."
M., “A PDE-based fast local level set method”, J. Journal of Fluids Engineering 133, 2011, p.
Comput. Phys, Vol. 155, 1999, pp. 410-438. 115501.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Robinson-Mosher, A., Schroeder, C., Fedkiw, R., “A Waves”, 2012d, Under review, Ocean Engineer-
symmetric positive definite formulation for ing.
monolithic fluid structure interaction,” Journal of Sagaut, P. and Deck, S., “Large eddy simulation for
Computational Physics, Volume 230, Issue 4, aerodynamics: status and perspectives,” Phil.
2011, pp. 1547-1566. Trans. R. Soc. A., Vol. 367, 2009, pp. 2849.
Rosenfeld, M., Kwak, D., Vinokur, M., "A Fractional Sakamoto, N., Carrica, P.M., and Stern, F., (2012),
Step Solution Method for the Unsteady Incom- “URANS Simulation of Static and Dynamic Ma-
pressible Navier-Stokes Equations in Generalized neuvering for Surface Combatant: Part1-
Coordinate Systems," J. Comp. Phys., Vol. 94, Verification and Validation of Forces, Moment
1991, pp. 102-137. and Hydrodynamic Derivatives”, Journal of Ma-
Rumsey, C. L., and Thomas, J. L., "Application of rine Science and Technology, in press.
FUN3D and CFL3D to the Third Workshop on Sanders, W.C., Winkel, E.S., Dowling, D.R., Perlin,
CFD Uncertainty Analysis," NASA Report No. M., and Ceccio, S.L., "Bubble Friction Drag Re-
TM-2008-215537, 2008. duction in a High-Reynolds-Number Flat-Plate
Rung, T., Wöckner, K., Manzke, M., Brunswig, J., Turbulent Boundary Layer," J. Fluid Mech., Vol
Ulrich, C., Stück, A., Challenges and perspectives 552, 2006, pp. 353-380.
for maritime CFD applications. Jahrbuch der Sanada, Y., Tanimoto, K., Sano, M., Yeo, D-J, Gui, L.,
Schiffbautechnischen Gesellschaft, 103. Band; Toda, Y., and Stern, F., “Trajectories and Local
2009. Flow Field Measurements around ONRTH in
Sadat-Hosseini, H., Stern, F., Olivieri, A., Campana, Maneuvering Motions,” Proceedings 29th Sym-
E., Hashimoto, H., Umeda, N., Bulian, G., Fran- posium Naval Hydrodynamics, Gothenburg,
cescutto, A., “HEAD-WAVES PARAMETRIC Sweden, 26-31 August 2012.
ROLLING”, Ocean Engineering, Vol. 37, Issue Sato Y, Miyata H, Sato T, “CFD Simulation of 3D
10, 2010, Pages 859-878. Motion of a Ship in Waves: Application to an
Sadat-Hosseini, H., Araki M., Umeda N., Sano M., Advancing Ship in Regular Heading Waves,” J.
Yeo D. J., Toda Y., Stern F., “CFD, system-based Marine Sci. Tech. Vol. 4, 1999, pp. 108-116.
method, and EFD investigation of ONR tumble- Schumann, U. “Stochastic backscatter of turbulence
home instability and capsize with evaluation of energy and scalar variance by random subgrid-
the mathematical model”, 12th International Ship scale fluxes,” Proc. R. Soc. Lond. A, Vol. 451,
Stability Workshop, 2011a, pp.135-145. 1995, pp. 293–318.
Sadat-Hosseini, H., Carrica, P., Stern, F., Umeda, N., Shakeri, M., Tavakolinejad, M., and Duncan, J. H.,
Hashimoto, H., Matsuda, A., “CFD, System- “An experimental investigation of divergent bow
Based and EFD Study of Ship Dynamic Instabil- waves simulated by a two-dimensional plus tem-
ity Events: Surf-riding, Periodic Motion, and poral wave marker technique,” Journal of Fluid
Broaching,” Ocean Engineering, Vol. 38, Issue 1, Mechanics, 634, 2009, pp. 217-243.
January 2011b, pp. 88-110. Shalf, J. “Hardware Trends,” Panel discussion
Sadat-Hosseini, H., Stern F., Toxopeus S., Visonneauc presentation in 21st Parallel CFD Conference,
M., Guilmineau E., Lin W.M., and Grigoropoulos May 18-22, Moffett Field, California, USA, 2009.
G., “CFD, potential flow, and system-based simu- Shibata, K. Koshizuka, S., Tanizawa, K., “Three-
lations of course keeping in calm water and sea- dimensional numerical analysis of shipping water
keeping in regular waves for 5415m”, 2012a, In onto a moving ship using a particle method,” J.
preparation to Ocean Engineering. Marine Sci. Tech., Vol. 14(2), 2009, pp. 214-227.
Sadat-Hosseini, H., Stern F., Toxopeus S., “CFD Shirani E, Jafari A, Ashgriz N, "Turbulence models for
simulations of course keeping in irregular waves flows with free surfaces and interfaces", AIAA
for 5415M”, Report No. 3, 2012b. Journal, Vol. 44(7), 2006, pp. 1454-1462.
Sadat-Hosseini, H., Kim, D.H, Lee, S.K., Rhee, S.H., Shu, C. W. and Osher, S., “Efficient implementation of
Carrica, P., Stern, F., “CFD and EFD study of essentially non-oscillatory shock-capturing
Damaged Ship Stability in Regular Waves”, Pro- schemes,” J. Comp. Phys., Vol. 77, 1988, pp.
ceedings of the 11th International Conference on 439-471.
the Stability of Ships and Ocean Vehicles, 23-28 Shur, M., Spalart, P. R., Strelets, M. Kh., Travin, A.
September 2012c, Athens, Greece. 2008. “A hybrid RANS-LES approach with de-
Sadat-Hosseini, H., Wu P.C., Carrica P., Kim H., Toda layed-DES and wall-modeled LES capabilities,”
Y., Stern F., “CFD Simulation and Validation of International Journal of Heat and Fluid Flow, Vol.
Added Resistance of KVLCC2 with Fixed and 29, 2008, pp. 1638.
Free Surge Conditions in Short and Long Head
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Sigurdson, L. W., 1995, “The structure and control of Plate: Experiment and Theory," Journal of Ship
turbulent reattaching flow,” Journal of Fluid Me- Research, Vol. 37, No. 2, 1993, pp. 102-118.
chanics, 248, pp. 139-165. Stern, F., Kim, H.T., Zhang, D.H., Toda, Y., Kerwin,
Simpson, R. L., 2001, “Junction flows,” Annual J., and Jessup, S., “Computation of Viscous Flow
Review of Fluid Mechanics, 33, pp. 415-443. Around Propeller-Body Configurations: Series 60
Simonsen, C., and Stern, F., 2010, “CFD Simulation of CB = 0.6 Ship Model,” Journal of Ship Research,
KCS Sailing in Regular Head Waves,” Proceed- Vol. 38, No. 2, 1994, pp. 137-157.
ings of Gothenburg 2010 workshop, Report No. Stern, F., Longo, J., Zhang, Z.J., and A.K. Subramani,
R-10:122, pp. 473-478, Chalmers University of "Detailed bow-flow data and CFD of a Series 60
Technology, Sweden. CB = .6 ship model for Froude number .316", J.
Simonsen, C., and Stern, F., 2005, “RANS Maneuver- Ship Res., Vol. 40, 1996, pp. 193-199.
ing Simulation of Esso Osaka With Rudder and a Stern, F., Wilson, R. V., Coleman, H. W., and Pater-
Body-Force Propeller,” Journal of Ship Research, son, E. G., 2001, "Comprehensive Approach to
Vol. 49, No. 2, 2005, pp. 98-120. Verification and Validation of CFD Simula-
So, R.M.C. and Lai, Y.G., “Low-Reynolds-number tions—Part 1: Methodology and Procedures,"
modelling of flows over a backward-facing step,” Journal of Fluids Engineering, 123(4), pp. 793-
Journal of Applied Mathematics and Physics, 802.
Vol. 39, 1988, pp. 13-27. Stern, F., Wilson, R., Longo, J., Carrica, P., Xing, T.,
Spalart, P. R., “Detached-Eddy Simulation,” Annual Tahara, Y., Simonsen, C., Kim, J., Shao, J., Ir-
Review of Fluid Mechanics, Vol. 41, 2009, pp. vine, M., Kandasamy, K., Gosh, S., and Wey-
181. mouth, G., “Paradigm for Development of Simu-
Spalart, P. R., Allmaras, S. R., "A One-Equation lation Based Design for Naval Hydrodynamics,”
Turbulence Model for Aerodynamic Flows," 8th International Conference on Numerical Ship
AIAA Paper 92-0439, 1992. Hydrodynamics, September 22-25, 2003, Busan,
Sreedhar, M. and Stern, F., "Large Eddy Simulation of Korea.
Temporally Developing Juncture Flows," Interna- Stern, F., Wilson, R., Shao, J., “Quantitative approach
tional Journal of Numerical Methods in Fluids, to V&V of CFD Simulations and Certification of
Vol. 28, No. 1, 1998a, pp. 47-72. CFD Codes,” Int. J. Numer. Methods Fluids, Vol.
Sreedhar, M. and Stern, F., "Prediction of Solid/Free- 50, 2006a, pp. 1335–135.
Surface Juncture Boundary Layer and Wake of a Stern, F., Xing, T., Muste, M., Yarbrough, D.,
Surface-Piercing Flat Plate at Low Froude Num- Rothmayer, A., Rajagopalan, G., Caughey, D.,
ber," ASME Journal of Fluids Engineering, Vol. Bhaskaran, R. Smith. S., Hutching, B., and
120, 1998b, pp. 354-362. Moeykens, S., “Integration of Simulation Tech-
Starke, B, van der Ploeg, A. Raven?H, “Viscous free nology into Undergraduate Engineering Courses
surface flow computations for self-propulsion and Laboratories,” International Journal Learning
conditions using PARNASSOS”, Proc. of G2010 Technology, Vol. 2, No. 1, January 2006b, pp.
Workshop, 2010. 28-48.
Stern, F., "Effects of Waves on the Boundary Layer of Stern, F., Carrica, P., Kandasamy, M., Gorski, J.,
a Surface-Piercing Body," Journal of Ship Re- O’Dea, J., Hughes, M., Miller, R., Kring, D.,
search, Vol. 30, No. 4, 1986, pp. 256-274. Milewski, W., Hoffman, R., and Cary, C., 2007,
Stern, F., Yoo, S.Y., and Patel, V.C., "Interactive and “Computational Hydrodynamic Tools for High-
Large-Domain Solutions of Higher-Order Vis- Speed Sealift,” Transactions of The Society of
cous-Flow Equations," AIAA Journal, Vol. 26, Naval Architects and Marine Engineers, 114, pp.
No. 9, September 1988a, pp. 1052-1060, IIHR 55–81.
Reprint #753. Stern, F., Bhushan, S., Carrica, P., Yang, J., “Large
Stern, F., Kim, H.T., Patel, C., Chen, H.C., “A viscous- Scale Parallel Computing and Scalability Study
flow approach to the computation of propeller- for Surface Combatant Static Maneuver and
hull interaction”, Journal of Ship Research, vol. Straight Ahead Conditions using CFDShip-Iowa,”
32, no. 4, 1988b. 21st Parallel CFD Conference, May 18-22, Mof-
Stern, F., Hwang, W.S., and Jaw, S.Y., "Effects of fett Field, California, USA, 2009, pp. 52-56.
Waves on the Boundary Layer of a Surface- Stern, F., Sadat-Hosseini, H., Mousaviraad, M.,
Piercing Flat Plate: Experiment and Theory," Bhushan, S., “CHAPTER 4 - Evaluation of Sea-
Journal of Ship Research, Vol. 33, No. 1, 1989, keeping Predictions,” Gothenburg 2010 workshop
pp. 63-80. on CFD in Ship Hydrodynamics, 2010.
Stern, F., Choi, J.E., and Hwang, W.S., "Effects of Stern, F., Agdrup, K., Kim, S.Y., Hochbaum, A.C.,
Waves on the Wake of a Surface-Piercing Flat Rhee, K.P., Quadvlieg, F., Perdon, P., Hino, T.,
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Broglia, R., and Gorski, J., “Experience from Tahara, Y., Wilson, R Carrica, P., “Comparison of
SIMMAN 2008—The First Workshop on Verifi- Free-Surface Capturing and Tracking Approaches
cation and Validation of Ship Maneuvering Simu- in Application to Modern Container Ship and
lation Methods,” Journal of Ship Research, Vol. Prognosis for Extension to Self-Propulsion Simu-
55, No. 2, 2011a, pp. 135-147. lator”, Proc. of CFD Workshop Tokyo 2005, To-
Stern, F., Toxopeus, S., Visonneau, M., Guilmineau, kyo, Japan.
E., Lin, W.M., Grigoropoulos, G., “CFD, Poten- Takai, T., Kandasamy, M. and Stern, F., 2011, "
tial Flow, and System-Based Simulations of Verification and validation study of URANS sim-
Course Keeping in Calm Water and Seakeeping ulations for an axial waterjet propelled large high-
in Regular Waves for 5415M”, AVT-189 Special- speed ship," Journal of Marine Science and Tech-
ists' Meeting, Portsdown West, UK, 2011b. nology, 2011, Volume 16, Issue 4, pp 434-447.
Stern, F., Yoon, H., Yarbrough, D., Okay, M., Oztekin, Tallent, J. R., Yamashita, T., and Tsuchiya, Y.,
U., and Roszelle, B., “Hands-On Integrated CFD “Transformation Characteristics of Breaking
Educational Interface for Introductory Fluids Me- Waves”, Wave Water Kinematics, Vol. 178,
chanics: Invited Paper,” International Journal 1990, pp. 509-523.
Aerodynamics, 2012, in press. Temmerman, L., Leschziner, M. A., Hadziabdic, M.,
Strasser, C., Jasionowski, A. and Vassalos, D., Hanjalic, K., “A Hybrid Two-Layer URANS-LES
"Calculation of the Time-to-Flood of a Box- Approach for Large-Eddy Simulation at High
Shaped Barge using CFD", Proc. 10th Int’l Conf. Reynolds Numbers,” Int. J. Heat Fluid Flow, Vol.
Stability of Ships & Ocean Vehicles, St. Peters- 26, 2005, pp. 173–190.
burg, Russia, 2009, pp. 733–740. Thompson, J. F.; Warsi, Z. U. A., Mastin, C. W.
Suh, J., Yang, J., and Stern, F., “The effect of air-water “Numerical Grid Generation: Foundations and
interface on the vortex shedding from a vertical Applications,” North-Holland, Elsevier, 1985.
circular cylinder,” Journal of Fluids and Struc- Toutant, A., Chandesris, M., Jamet, D., Lebaigue, O.,
tures, Vol. 27, No. 1, 2011, pp. 1-22. “Jump conditions for filtered quantities at an un-
Sussman, M., “A parallelized, adaptive algorithm for der-resolved discontinuous interface Part 1: theo-
multiphase flows in general geometries,” Comp. retical development,” Int. J. Multiphase Flow, 35
Struct., Vol. 83, 2005, pp. 435-444. (12), 2009, pp. 1100-1118
Sussman, M., Dommermuth, D., “The numerical Toxopeus, S., 2009, “Deriving mathematical manoeu-
simulation of ship waves using Cartesian grid vring models for bare ship hulls using viscous
methods,” in: Proc. 23rd Symposium on Naval flow calculations,” J. Mar Sci Technol, Vol. 14,
Hydrodynamics, Val De Reuil, France, 2000. pp. 30–38.
Sussman, M., Puckett, E.G., “A Coupled Level Set and von Kerczek, C.H., Christoph, G., and Stern, F.,
Volume-of-Fluid Method for Computing 3D and "Further Developments of the Momentum Inte-
Axisymmetric Incompressible Two-Phase gral Method for Ship Boundary Layers," SAI Re-
Flows,” Journal of Computational Physics, Vol. port #8413046, May 1984, 38 pp.
162, Issue 2, 2000, pp. 301-337. Wackers, J., Koren, B., Raven,.H.C..van der Ploeg, A,
Sussman, M., Smereka, P. and Osher, S., “A level set Starke, A.R. Deng, G.B., Queutey, P., Visonneau,
approach for computing solutions to incompressi- M, Hino, T. Ohashi K., 2011 “Free-Surface Vis-
ble two-phase flow,” J. Comp. Phys., Vol. 114, cous Flow Solution Methods for Ship Hydrody-
1994, pp. 146-159. namics,” Archives of Computational Methods in
Tahara, Y., Stern, F., and Rosen, B., "An Interactive Engineering, Vol. 18(1), 2011, pp. 1-41.
Approach for Calculating Ship Boundary Layers Wackers, J., Deng G., Leroyer, A., Queutey, P.,
and Wakes for Nonzero Froude Number," Journal Visonneau, M., Adaptive grid refinement for hy-
of Computational Physics, Vol. 98, No. 1, Janu- drodynamic flows, (2012) Computers and Fluids,
ary 1992, pp. 33-53, IIHR Reprint #892. 55 ,pp. 85-100.
Tahara, Y. and Stern, F., "Large-Domain Approach for Wallin, S., Johansson, A. V., “Modeling Streamline
Calculating Ship Boundary Layers and Wakes Curvature Effects in Explicit Algebraic Reynolds
and Wave Fields for Nonzero Froude Number," Stress Turbulence Models,” International Journal
Journal Computational Physics, Vol. 127, No. 2, of Heat and Fluid Flow, Vol. 23, 2002, pp. 721–
September 1996, pp. 398-411, IIHR Reprint 730.
#1210. Wang, Y., Simakhina, S., Sussman, M., “A hybrid
Tahara, Y., Katsuui, T., Himemno, Y. 2002 Computa- level set-volume constraint method for incom-
tion of ship Viscous flow at full scale Reynolds pressible two-phase flow,” Journal of Computa-
number, Journal of the Soc. Naval Architects of tional Physics, Vol., 231, 2012, pp. 6438–6471.
Japan, Vol. 92, 2002, pp. 89-101.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Wang, Z., Yang, J., Stern, F., “An improved particle waves,” Journal of Fluid Mechanics, 545, 2005,
correction procedure for the particle level set pp.291-328.
method,” Journal of Computational Physics, Vol. Weymouth, G., Wilson, R., and Stern, F., 2005,
228, Issue 16, 2009a, pp. 5819-5837. “RANS CFD predictions of pitch and heave ship
Wang, Z., Yang, J., Koo, B., and Stern, F, “A coupled motions in head seas”, Journal of Ship Research,
level set and volume-of-fluid method for sharp Vol. 49, pp. 80–97.
interface simulation of plunging breaking waves”, Wilson, R., Carrica, P. and Stern, F., “Simulation of
Inter. J. Multiphase Flow, Vol. 35, 2009b, pp. ship breaking bow waves and induced vortices
227-246. and scars”, Int. J. Num. Meth. Fluids, Vol. 54,
Wang, Z., Yang, J., Stern, F., “URANS Study of Air- 2007a, pp. 419-451.
Layer Drag Reduction in a High-Reynolds- Wilson, R.V., Nicholas, D.S., Mitchell, B., Karman,
Number Flat-Plate Turbulent Boundary Layer”, S.L., Betro, V.C., Hyams, D.G., Sreenivas, K.,
40th Fluid Dynamics Conference and Exhibit 28, Taylor, L.K., Briley, W.R., and Whitfield, D.L.,
June - 1 July 2010a, Chicago, Illinois. 2007b, “Simulation of a surface combatant with
Wang, Z., Yang, J., Stern, F., “Numerical simulations Dynamic Ship Maneuvers,” 9th International
of wave breakings around a wedge-shaped bow,” Conference on Numerical Ship Hydrodynamics,
28th Symposium on Naval Hydrodynamics, Pas- Ann Arbor, Michigan.
adena, California, September 12-17, 2010b. Wilson, R., Shao, J., and Stern, F., 2004, "Discussion:
Wang, Z., Suh, J., Yang, J., and Stern, F., “Sharp Criticisms of the" Correction Factor" Verification
Interface LES of Breaking Waves by an Interface Method," Journal of Fluids Engineering, 126, pp.
Piercing Body in Orthogonal Curvilinear Coordi- 704-706.
nates,” 50th AIAA Paper, January 2012a. Xing, T., Kandasamy, M., and Stern. F., “Unsteady
Wang, Z., Yang, J., Stern, F., “A new volume-of-fluid Free-Surface Wave-Induced Separation: Analysis
method with a constructed distance function on of Turbulent Structures Using Detached Eddy
general structured grids,” Journal of Computa- Simulation and Single-Phase Level Set,” Journal
tional Physics, Vol. 231, Issue 9, 2012b, pp. of Turbulence, Vol. 8, No. 44, 2007, pp. 1-35.
3703-3722. Xing, T., Carrica, P., and Stern, F., 2008, “Computa-
Wang, Z., Yang, J., Stern, F., “A simple and conserva- tional Towing Tank Procedures for Single Run
tive operator-splitting semi-Lagrangian volume- Curves of Resistance and Propulsion,” ASME J.
of-fluid advection scheme,” Journal of Computa- Fluids Eng., 130(2), pp. 1–14.
tional Physics, Vol. 231, Issue 15, 2012c, pp. Xing, T., Carrica, P. M., Stern, F., "Large-scale RANS
4981-4992. and DDES computations of KVLCC2 at drift an-
Wang, Z., Yang, J., Stern, F., “Parallel LES of gle 0 degree." Proceedings of Gothenburg 2010:
breaking wave over an immersed bump in or- A Workshop on CFD in Ship Hydrodynamics,
thogonal curvilinear coordinates,” ParCFD 2012, Gothenburg, Sweden, 2010.
22-26 May, 2012d, Atlanta, US. Xing, T. and Stern, F., “Factors of Safety for Richard-
Wang, Z., Suh, J., Yang, J., Stern, F., "Sharp Interface son Extrapolation,” ASME J. Fluids Eng, Vol.
Large-Eddy Simulation of Breaking Waves by an 132 (6), June 2010, pp.061403.
Interface Piercing Body in Orthogonal Curviline- Xing, T., Carrica, P., Stern, F., 2011, "Developing
ar Coordinates", Journal of Computational Phys- Streamlined Version of CFDShip-Iowa-4.5,"
ics (in preparation), 2012e. IIHR Technical Report No. 479.
Wang, Z.J., Parthasarathy, V. “A fully automated Xing, T., and Stern, F., 2011, "Closure to "Discussion
Chimera methodology for multiple moving prob- of 'Factors of Safety for Richardson Extrapola-
lems,” International Journal for Numerical Meth- tion'" (2011, ASME J. Fluids Eng., 133, p.
ods in Fluids, Vol. 33, 2000, pp. 919-938. 115501)," ASME Journal of Fluids Engineering,
Waniewski, T., "Air entrainment by bow waves", 133(11), p. 115502.
Doctoral dissertation, 1999, California Institute of Xing, T. Bhushan, S., and Stern, F. “Vortical and
Technology, Pasadena, California. Turbulent Structures for KVLCC2 at Drift Angle
Waniewski, T.A., Brennen, C.E. and Raichlen, F., 0, 12, and 30 Degrees,” in press, Ocean Engi-
“Bow wave dynamics,” J. Ship Res., Vol. 46, neering, 2012.
2002, pp. 1–15. Xiu, D., Karniadakis, G.E., “A Semi-Lagrangian High-
Watanabe, Y. and Saeki, H., “Velocity field after wave Order Method for Navier–Stokes Equations,”
breaking,” International Journal for Numerical Journal of Computational Physics, Vol. 172, Issue
Method and Fluids, 39, 2002, pp.607-637. 2, 2001, pp. 658-684.
Watanabe, Y., Saeki, H., Hosking, R., “Three-
dimensional vortex structures under breaking
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Yamazaki, R. 1”On the propulsion theory of ships on

still water introduction”, Memoirs of the Faculty
of Engineering, Kyushu University, 27, 4, 1968.
Yang, J., Balaras, E., “An embedded-boundary
formulation for large-eddy simulation of turbulent
flows interacting with moving boundaries,” Jour-
nal of Computational Physics, Vol. 215, Issue 1,
2006, pp. 12-40.
Yang, J. and Stern, F., “Large-eddy simulation of
breaking waves using embedded-boundary/level-
set method,” In: 45th AIAA Aerospace Sciences
Meeting and Exhibit, Reno, Nevada, USA, 2007,
AIAA Paper 2007-1455.
Yang, J., Sakamoto, N., Wang, Z., Carrica, P.M. and
Stern, F., “Two phase level-set/immersed-
boundary Cartesian grid method for ship hydro-
dynamics”, Proc. 9th Inter. Conf. Numer. Ship
Hydrodynamics, Ann Arbor, Michigan, 2007.
Yang, J., Preidikman, S., and Balaras, E., “A strongly-
coupled, embedded-boundary method for fluid-
structure interactions of elastically mounted rigid
bodies”, J. Fluids Struct., Vol. 24, 2008a, pp. 167-
182.
Yang, J., Bhushan, S., Suh, J., Wang, Z., Koo, B.,
Sakamoto, N., Xing, T., Stern, F., Large-eddy
simulation of ship flows with wall-layer models
on Cartesian grids, Proc. 27th Symposium on Na-
val Hydrodynamics, 2008b, Seoul, Korea
Yang, J. and Stern, F., “Sharp Interface Immersed-
Boundary/Level-Set Method for Wave-Body In-
teraction,” Journal of Computational Physics,
Vol. 228, Issue 17, September 2009, pp. 6590-
6616.
Yang, J., Wang, Z., Yeon, S-M, Koo, B., and Stern, F.,
“High-fidelity curvilinear-grid two-phase flow
solvers for ship hydrodynamics,” Proceedings
29th Symposium Naval Hydrodynamics, Gothen-
burg, Sweden, 26-31 August 2012.
Yang, J. and Stern, F., “A simple and efficient direct
forcing immersed boundary framework for fluid–
structure interactions,” Journal of Computational
Physics, Vol. 231, Issue 15, 2012, pp. 5029-5061.
Yao, A. F. and Wu, C. H., “Incipient breaking on
unsteady waves on sheared currents,” Physics of
Fluids, 17, 2005, 082104-1–082104-10.
Zhang Z.-R., “Verification and validation for RANS
simulation of KCS container ship without/with
propeller”, Journal of Hydrodynamics, 22(5),
2010, pages 932-939.
Zlatko Zlatev, Z., Milanov., E., Chotukova, V.,
Sakamoto, N. and Stern, F., 2009, "Combined
model-scale EFD-CFD investigation of the ma-
neuvering characteristics of a high speed catama-
ran," 10th International Conference on Fast Sea
Transportation FAST 2009, Athens, Greece.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012
Table 1: Summary of instability studies available in the literature for canonical cases (shown by Grey shadow) and those performed using CFDShip-Iowa V4 for
geometries of ship hydrodynamics interest.
Scaling Parameters
Instability Geometry Vortex St=fU/L Comments
Length (Velocity) scales
Cylinders and backstep Separation bubble 0.07 – 0.09
Surface Piercing NACA 0024 Wave induced separation 0.0685±4.3%  Caused by the interaction of two opposite
Barehull Athena 0.088 – 0.148 vortices initiated by shear layer instability
Half wake width
Appended Athena Transom Vortex shedding 0.103±4.4%  Free-surface reduces both f and St
Karman-like
(Shear layer velocity)
Transom-Model 0.075  Responsible for transom wave unsteadiness
Wigley Hull at  = 45 and 60 Intersection of hull and tip vortices 0.08±4.2% for wetted transom flow
KVLCC2 at  = 30 Leeward bow vortices 0.0735  Averaged St ~ 0.088 for ship flows, higher
Barehull 5415 at =0 Port and Starboard Sonar dome NA end of the canonical flow range
5415 with BK at =20 Leeward sonar dome separation 0.132
Cylinder/airfoil junction with  Associated with two vortex system and show
Horshoe vortex separation 0.17-0.28
Obstacle Thickness Flatplate dual peak spectra
Horseshoe
(Freestream velocity) Rudder-Hull intersection 0.146±3.9%  St increases with obstacle angle of attack, and
Appended Athena
Strut-Hull intersection 0.053±2% sweep angle
Airfoils and Cylinders Boundary layer separation 0.0056±2%
Surface Piercing NACA 0024 Free-surface separation 0.00384±0.5%  Associated with boundary layer (BL)
BL separation at strut-hull intersec- separation
Appended Athena 0.0067±3%
Momentum thickness tion  St varies inversely with the non-dimensional
Shear-layer
(Shear layer velocity) Wigley Hull at  =60 BL separation at leeward keel 0.0003 adverse pressure gradient
KVLCC2 at  = 30 BL separation at leeward bow 0.00101  St ~ 0.001 – 0.003 for ship flows are lower
Barehull 5415 at =0 Sonar dome separation NA than those for canonical case
5415 with BK at =20 Leeward Sonar dome separation 0.0016 – 0.0059
Cylinder, Backstep, Square Rib Separation bubble 0.073 – 0.12  Exhibits a periodic enlargement and shrink-
Reattachment length
Flapping Surface Piercing NACA 0024 Free-surface separation bubble 0.28 age of recirculation region
(Freestream velocity)
Barehull Athena Wake growth and decay 0.144  St has wide range
Delta Wing Tip vortex 0.75-1.35
 Vortices for static drift case show helical
Wigley Hull at  =60 Leeward keel vortices NA
streamline pattern
Distance along Leeward Fore-body Side 1.25-1.35
 Limited studies show good correlation with
Helical vortex core KVLCC2 at  = 30 Leeward Fore-body Bilge 1.35-1.45
delta wing tip vortex scaling
(Freestream velocity) Leeward Aft-body bilge 1.8-2.25
 Confirm the identity of these instabilities by
Windward BK vortex 1.2 – 1.4
5415 with BK at =20 comparing with slender fuselage vortices
Fore-body keel vortex 0.95 – 1.15
Free-surface Bare hull Athena  Occurs for dry transom flow
NA Rooster tail wave breaking NA
instability Transom-model  Shoulder waves traveling towards centerline
Vortex induced Ship length Wetted Transom appended Associated with transom vortex  Unsteady pitch and heave motions due to
2.19
motions (Freestream velocity) Athena shedding Karman-like transom vortex shedding
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012
Table 2 Summary of calm water resistance, sinkage and trim CFD V&V studies
Reference Geometry Fr Grid Resistance Sinkage Trim
|ECT|%D USN%S UD%D Es|%D USN%S UD%D |Et|%D USN%S UD%D
0.28 4.3 0.64 7.4 4.71 10.4 4.7
Carrica et al. (2007) 5415 3M
0.41 1.5 0.61 1.5 2.93 1.11 0.87
0.28 3.7 0.64 9.5 4.71 2.2 4.7
Xing et al. (2011) 5415 1.3M
0.41 4.5 0.61 4.5 2.93 19.3 0.87
Bare hull Athena 0.2 - 1.0 1M 2.1 2.53 1.5 7.7 1.6 29.3 9.6 15.3 8.1
Xing et al. (2008)
Propelled Appended Athena 0.2 - 0.84 2.2M 4.5 8.1 5
0 - 0.6, Roll
3.3M 2.6 0.42 3.62 2.3 14.15 2.4
Sadat-Hosseini et al. f = 10°
ONR Tumblehome
(2010) 0 - 0.6, Roll
3.3M 2.5 2.47 10.03
f = 20°
Stern et al. (2007) HSSL-Delft catamaran 0.2 - 0.65 - 8 23 17
Castiglione et al. (2011);
Delft catamaran 0.18 - 0.75 5.4M 5.2 18.5 16.1
Zlatev et al. (2009)
DTMB 5594 0.511 1.8M 0.78 0.8 14.3
Kandasamy et al. (2010)
5594, Water Jet self-propelled 0.511 1.8M 4.6 9 13.7
JHSS, Bare hull 0.34 29M 2.2 3.6 5.8 11.6 13.7
Takai et al. (2011) JHSS, Water Jet self-
0.34 13M 0.2 1.1 1.2 10.27 27.4
propelled
0.0 (double
Ismail et al. (2010) KVLCC2, FXst 4.6M 3.45 3.79 0.7
tanker)
Athena bare hull with skeg at 0.48 4.8 1.6 1.5 9.5 15.5 29.3 6.8 2.6 8.1
full-scale 0.8 3.14 32.3 5.3
Bhushan et al. (2009) 4.8M
Self-propelled fully appended 0.432, 0.575,
23 4.9
Athena at full-scale 0.839
Full-scale appended Athena
Bhushan et al. (2012c) 0.25 43M 4.22 5.28 1.5
(DES), FXσt
Bhushan et al. (2012b) T-Craft (SES/ACV) 0.2, 0.4, 0.6 14.7M 7.8 1.5 9.8 8.3 14.7 6.3
Fully appended ONR
Mousaviraad et al. (2012) 0.2, 0.3, 0.4 10.2M 4.5 3.1 3.3
Tumblehome in head winds
KVLCC2, FXst 0.1423 1.7
KVLCC2, FXst 0.26 0.8
5415, FXst 0.138 - 0.41 2.5
Gothenburg 2010 KVLCC2, FRzq < 0.2 2.1 33.3 7.5
(Stern et al., 2010) < 0.2 55.6 30.5
KCS, FRzq 1.64
> 0.2 7.5 3.62
< 0.2 31.4 164.6
5415, FRzq 3
> 0.2 8.8 23
< 0.2 34.7 54.7
Average 3.3 2.5 1.5
> 0.2 9.7 6.9 12.3 11.0 6.7 4.6
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012
Table 3 Summary of seakeeping in regular head waves CFD V&V studies
UT (%S1) UG (%S1) USN (%D) Reported UV (%D) E%D for amp, E%2π for phase
(0th amp, 1st amp, (0th amp, 1st amp, (0th amp, 1st amp, (0th amp, 1st amp, 1st
Grid V&V (0th amp, 1st amp, 1st phase)
Study Geom. Code λ/L (ak) Fr 1st phase) 1st phase) 1st phase) phase)
(M) Method
1st 2nd
CT z θ CT z θ CT z θ CT z θ CT z θ
order order
- - - - - - - - - - - - - - -
21.4 6.0 0.7 2.8 0.7 1.3 21.6 6.0 1.5 21.8 6.5 2.9 0.8 6.6 2.3
Weymouth et 0.11- CFDShip- 1.25 Correction 4.4 0.8
Wigley 0.3 - - - - - - - - - - - - - - -
al. (2005) 0.29 Iowa V3 (0.018) Factor
21.4 6.0 0.7 2.8 0.7 1.3 21.6 6.0 1.5 21.8 6.5 2.9 0.8 6.6 2.3
9.4 1.6 9.7 10.4 3.2 2.6
2.0 - - 2.0 - - 2.0 - - - 21.3 3.3
- - - - - - - - - - 2.8 2.7
Carrica et al. DTMB 0.38- CFDShip- 1.5 Correction 1.8 12.3
0.28 - - - - - - - - - - - 0.3 1.2
(2007) 5512 2.96 Iowa V4 (0.025) Factor
2.0 - - 2.0 - - 2.0 - - - 8.1 2.4
2.0 2.0 2.0 5.3 7.0
- - - - - - - - - 5.6 -5.1 143.8
1.5 10.3 2.6 1.5 10.3 2.6 10.5 10.3 2.6 82.3 82.3 -10.1
Simonsen et al. 1.8- CFDShip- 1.15 Correction 15.3 55.5
KCS 0.26 - - - - - - - - - - 1.1 1.1 -2.6
(2010) 3.8 Iowa V4 (0.025) Factor
1.5 10.3 2.6 1.5 10.3 2.6 10.5 10.3 2.6 29.7 26.1 43.7
4.8 4.8 7.8 33.2 35.4
- - - - - - - - - - - - - - -
- 1.5 0.5 - 3.3 1.7 - 3.6 1.7 - 4.4 3.0 - 9.4 0.1
Castiglione et DELFT 0.7- CFDShip- 1.806 Correction 4.8 -
0.75 - - - - - - - - - - - - -
al. (2011) Cat. 5.4 Iowa V4 (0.025) Factor
- 1.5 0.5 - 3.3 1.7 - 3.6 1.7 - 4.4 3.0 - 9.4 0.1
1.0 2.5 2.7 3.7 4.8 4.8
- - - - - - - - - - - - - - -
0.839- - 5.7 2.2 - 4.9 4.9 - 9.1 5.6 - 9.2 5.8 - 3.8 2.2
Mousaviraad et DTMB 2.8- CFDShip- Factor of 3.4 -
2.977 0.34 - 2.0 2.8 - 2.8 4.8 - 3.4 5.6 - 9.1 14.1 - 2.3 5.3
al. (2010a) 5512 22.1 Iowa V4 Safety
(0.025) - 3.8 2.5 - 3.9 4.9 - 6.2 5.6 - 9.1 9.9 - 3.0 3.8
3.2 4.4 5.9 9.5 3.4 3.4
- 6.5 12.5
CFDShip- - 4.3 2.5
Carrica et al. DTMB 1.5 4.5 9.5
71 Iowa V4 0.41 - - - - - - 6.2 5.0
(2010) 5512 (0.025)
(DES) - 5.7 6.7
6.2 7.0
- - - 1.9 2.0 8.9 1.9 2.0 8.9 8.7 9.6 22.3 18.5 54.0 54.0
Gothenburg 21.4 4.4 1.1 2.2 4.8 2.6 11.6 6.5 2.9 21.6 11.0 5.3 47.7 10.0 10.0
KVLCC2; Avg.= 14.6 44.0
2010 (Stern et Various - 2.0 2.8 - 2.8 4.8 - 3.4 5.6 - 5.7 7.2 20.3 17.3 17.3
KCS 2.6
al., 2010) 21.4 3.2 2.0 2.0 3.2 5.4 6.7 4.0 5.8 15.1 8.8 11.6 28.8 27.1 27.1
8.9 3.5 5.5 11.8 27.7 29.3
- - - 2.0 2.0 8.9 2.0 2.0 8.9 5.3 9.6 22.3 12.0 19.2 53.4
21.4 4.4 1.1 2.2 4.8 2.6 11.6 7.1 2.9 17.9 8.3 3.9 43.6 17.0 1.4
8.2 31.7
Average (Average fine grid=15M) - 2.0 2.8 - 2.8 4.8 - 3.4 5.6 - 7.4 10.7 10.7 5.4 5.2
21.4 3.2 2.0 2.1 3.2 5.4 6.8 4.2 5.8 11.6 8.4 12.3 22.1 13.9 20.0
8.9 3.6 5.6 10.8 18.7 20.0
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012
Table 4 Summary of forced motions maneuvering CFD V&V studies
Linear
Forces & Moments Nonlinear
Derivative
Fine Grid Derivative
Study Geom. Code Fr Conditions X Y N Yʹ Nʹ
(M)
USN UV E USN UV E USN UV E E E E
%D %D %D %D %D %D %D %D %D %D %D %D
CFDShip-
Ismail 0.0 (double
KVLCC2 4.6 Iowa Static Drift, β=12° 2.5 2.6 13.2 0.1 0.7 4.6 2.2 2.5 2.8
et al. (2010) tanker)
V4
HTC, EFD: 0.13- Static Drift, β up to 30° 20.6 8.9
Toxopeus
MARIN LNG, 3.8 PARNASSOS 0.18;
(2009) Pure Yaw, γ up to 0.6 21.9 30.8
KVLCC2M CFD: 0.0
Phillips ANSYS EFD: 0.14;
KVLCC2 2.1 Static Rudder, δ=10° 17.5 18.8 6.3 1.4 2.9 11.9 0.5 2.5 3.7
et al. (2009) CFX CFD: 0.0
CFDShip-
Bhushan
DTMB 5415 250 Iowa 0.28 Static Drift, β=20° 0.4 7.9 4.6 0.6 4.2 3.7 1.4
et al. (2011)
V4 (DES)
0.0 (double
KVLCC2M Static Drift, β up to 12° 2.6 4.6 9.2
Jacquin tanker)
ICARE
et al. (2006) 0.132,
HTC Static Drift, β up to 30° 10.5 7.3 14.3
0.238
Static Drift, β=10° 8.4 9.7 9.4 5.4 3.1 4.9 4.4
CFDShip-
Sakamoto Pure Yaw, γ=0.3 7.6 10.0 17.1 34.5 37.4 34.0 10.1 10.9 8.8 35.2 7.7
DTMB 5415 20 Iowa 0.28
et al. (2012) Pure sway, βmax up to 10° 8.0 9.2 4.5 4.8 4.9
V4
Yaw & Drift 21.6 10.0 7.9 15.4 91.3
Wilson
DTMB 5415 7.1 Tenasi 0.28 Pure sway 31.0 5.0 24.0
et al. (2007)
Fully appended Static Drift, β up to 11.7° 0.8 2.9 7.7
CFDShip- 0.2,
Mousaviraad ONR Pure sway, βmax=10° 7.4 2.7 14.6
10.2 Iowa 0.3,
et al. (2012) Tumblehome Pure Yaw, ψmax up to 17.7° 5.6 5.9 3.2
V4.5 0.4
in head winds Yaw & Drift; β=10°, ψmax=8.9° 4.2 2.3 4.1
KVLCC1, Static Rudder, δ=10° 27.0 55.0 29.0 14.5 59.0
SIMMAN 2008; Various
KVLCC2, Static Drift, β up to 12° 18.1 7.0 10.1 14.5 59.0
Stern Avg.=5.1 URANS Various
KCS, Pure Sway 21.3 15.4 15.4 8.0 8.0 49.0
et al. (2011) solvers
5415 Pure Yaw 32.8 25.7 25.7 23.0 23.0 32.0
Static Rudder 17.5 18.8 16.6 1.4 2.9 33.4 0.5 2.5 16.4 14.5 59.0
Static Drift 3.8 6.7 8.4 0.4 2.5 5.1 2.2 2.5 6.9 13.3 24.1
Pure Sway 16.9 8.1 14.6 6.4 6.4 49.0
Average Pure Yaw 7.6 10.0 18.5 34.5 37.4 21.9 10.1 10.9 12.5 26.7 20.5 32.0
Yaw & Drift 12.9 6.1 6.0 15.4 91.3
9.6 11.8 14.7 12.1 14.3 14.9 4.3 5.3 11.3 15.3 40.3 40.5
AVERAGE
|E|=13.6 27.8 40.5
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012
Table 5 Summary of CFD simulations for maneuvering and system identification

Turning Circle E%D Zigzag E%D

1st 2nd
Type Author Code Method Propulsion Rudder Grid Geometry Cases DOF A TR TD D Ave.
overshoot overshoot

Finite Volume,
424K 6.2 11.2 7.2 8.20
RANSE code Spalart and All- 3DOF
Muscari et al. Body force Actual
Maneuver developed at Maras, no free KVLCC2 turning (double
(2008a,b) propeller rudder
INSAEN surface,dynamic 3.4M model) 8.4 5.3 0.8 4.83
overset Grid
Finite difference, k-
Bhushan et al. w, level set, Body force Actual
Maneuver CFDShip-Iowa 7M 5415M zigzag 6DOF No data
(2009) dynamic overset propeller rudder
grid
Finite difference,
5.6M zigzag 30.6 36 33.30
Carrica and Stern DES, level set, Actual Actual
Maneuver CFDShip-Iowa KVLCC1 6DOF
(2008) dynamic overset propeller rudder
14.9M turning 8 24 0.65 10.88
grid
Finite difference, k- zigzag and
7M 2.6 9.5 5.18 5.69 6.97 6.69 6.11
Carrica et al. w, level set, Body force Actual turning
Maneuver CFDShip-Iowa 5415M 6DOF
(2008a,2012a) dynamic overset propeller rudder
12M turning in waves No data
grid

5415M 6DOF No data

ICARE
Drouet et al. Finite volume, k-w, Body force
Maneuver developed by Side force - turning 3DOF 0.3 0.30
(2008) VOF, dynamic grid propeller Humburg test
ECN, France
case
6DOF 5 5.00

ICARE Finite volume, k-w,

Ferrant et al. Body force
Maneuver developed by VOF, dynamic Side force - 5415M Turning in waves 6DOF No data
(2008) propeller
ECN, France overset grid
Finite difference, k-
Araki et al. w, level set, Body force Actual zigzag and
Maneuver and SI CFDShip-Iowa 12.1M OT 6DOF 1.23 2.29 1.34 1.54 2.23 1.73
(2012a) dynamic overset propeller rudder turning
grid
Finite difference, k-
Araki et al. w, level set, Body force Actual zigzag and
Maneuver and SI CFDShip-Iowa 12.1M OT 6DOF Not reported
(2012b) dynamic overset propeller rudder turning in wave
grid
ICARE Finite volume, k-w,
Jacquin et al. Body force
Maneuver developed by VOF, dynamic Side force - S60 turning 6DOF No data
(2006) propeller
ECN, France overset grid
Finite Volume,
Durante et al. 6.2M 7.35 3.15 5.24 4.06 4.95
RANSE code Spalart and All-
(2010) Body force Actual
Maneuver developed at Maras, no free Tanker turning 6DOF
Dubbioso et al. propeller rudder
INSAEN surface, dynamic 0.77M 10.6 9.42 4.96 12.26 9.31
(2012)
overset Grid
turning and
Finite difference, k-
zigzag in calm
Mousaviraad et al. w, level set, Body force Actual
Maneuver CFDShip-Iowa 8.6-9.3M SES/ACV water and waves 6DOF No data
(2012) dynamic overset propeller rudder
for deep and
grid
shallow
Total 6.34 9.27 4.10 4.80 13.04 14.97 8.46
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Table 6 Summary of CFD simulations for course keeping and stability

Type Author Code Method Propulsion Rudder Grid Geometry Cases DOF Comparison with EFD
Finite
difference, k-w, broaching in
Course Keeping Sadat-Hosseini Body force Actual
CFDShip-Iowa level set, 3.7M OT following regular 6DOF Time history of motions
and Stability et al. (2011) propeller rudder
dynamic overset waves
grid
Finite
difference, k-w,
Course Keeping Mousaviraad Body force Actual hurricane
CFDShip-Iowa level set, 3.7M OT 6DOF no EFD data
and Stability et al. (2008) propeller rudder CAMILLE
dynamic overset
grid
Finite
Carrica et al.
difference, k-w,
Course Keeping (2008b) and Body force Actual broaching in
CFDShip-Iowa level set, 3.7M OT 6DOF no EFD data
and Stability Huang et al. propeller rudder irregular waves
dynamic overset
(2008)
grid
Finite
difference, k-w, broaching in
Course Keeping Carrica et al. Actual Actual
CFDShip-Iowa level set, 21M OT following regular 6DOF Time history of roll, pitch and yaw
and Stability (2012b) propeller rudder
dynamic overset waves
grid
Finitie volume,
Strasser et al. compressible, k- damaged flooding in calm Time history of motions and water
Stability - No propeller No rudder - 6DOF
(2009) e,VOF, grid re- barge water height
meshing
Finite
roll decay and
difference, k-w,
Sadat-Hosseini 6.3- damaged motions with Time history of motions and water
Stability CFDShip-Iowa level set, No propeller No rudder 6DOF
et al.(2012) 20M SSRC flooding in calm height
dynamic overset
water and waves
grid
Finite
difference, k-w, roll decay with
Body force Actual 6.3-
CFDShip-Iowa level set, no fins or
propeller rudder 18.6M
dynamic overset passive fins
grid
Stern et al. Time history of 6DOF motions, BK, fin
Course Keeping ISIS-CFD 5415M 6DOF
(2011) and rudder forces
developed by Finite volume, k- roll decay,
Fluid w, VOF, Body force Fixed forced roll and
5.9M
Mechanics unstructured propeller rudder motions in
Laboratory, grid waves
France
Finite volume,
OVER- Spalart and All-
Dreyer and Body force submarine Time history of pitch and depth for
Course keeping REL_TCURS, Maras, VOF, Actual fins 6.4M overtaking 6DOF
Boger (2010) propeller and ship submarines
ARL dynamic overset
grid
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012
Table 7 Experimental studies of bow wave breaking
Waniewski et al. (2002) Karion et al. (2004)
Experiment Geometry Deflecting plate in small & large flumes Bow wedge (Model No. 5605) in deep water towing basin
Setup Wedge in a towing tank Bow entrance angle θ=20˚ & 40˚
Half wedge angle θ=13˚ & 26˚ Flare angle φ=20˚ & 0˚
Dihedral angle φ=0˚ (for small flume, φ=0˚ to 15˚)
Facility Small flume: 40 m long, 1.09 m wide Deep water towing basin: 575 m long, 15.5 m wide, and 6.3 to 6.5 m deep.
Large flume:2.65 m long, 0.459 m wide
Towing tank: 126m long, 7.5 m wide, and 3.7 m deep
Approaches Contact line measured using free surface probes in the Quantitative visualization (QViz) technique for the wave free surface
flumes High-speed video camera for spray of droplets
Bow wave profile measured using videos in towing tank

Test conditions Small flume: Re=1.1×104 to 1.6×104; Fr=2.57 to 6.46 Re= 4.1×105 to 7.0×106; Fr= 0.2 to 1.4
Large flume: Re=1.4×105 to 2.0×105; Fr=2.57 to 3.29 We= 11 to 2800
Towing Tank: Re=0.43×105 to 7.0×105; Fr= 1.27 to 7.77

Experimental Results & 1. Bow wave profile: mainly wave elevation 1.Contour plots of free surface elevation
observations 2. Plunging jet shape: jet thickness & impingement angle 2. Surface fluctuations
3. Scaling analysis 3.Surface roughness measurement
4.Free surface disturbance 4.Extent of breaking
5. Air entrainment (separate paper) 5. Spray droplets formation, number, size

Summary 1. The bow wave flow is highly nonlinear, there appears to 1. Wave breaking occurs when both Froude and Reynolds numbers exceed
be no satisfactory analytical solution. a critical level.
2. The wave is weakly dependent on dihedral angle and 2. A critical Webber number must be exceeded for spray generation to
depends on the bow half-angle. occur.
3. Surface disturbances were observed, which are likely 3. Scaling the maximum wave heights by the quantity ( Fr1.5D), collapses
gravity waves on the surface of the bow wave. They seem to the data fairly well.
be responsible for the breakup of the jet into strings of
droplets and for the periodic nature of the bubble clouds
produced by the wave breaking process.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 1. Flow chart demonstrating the components of the ship hydrodynamics computational methods.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

(a)

(b)

Fig. 2. Examples for sea modeling: a) exact potential solution for a linear wave component and generated random seas inside the
computational domain, b) snapshots of ship in three sisters rogue waves simulations.

Fig. 3. Waterjet flow modeling for JHSS at Fr=0.34 (top) and Delft catamaran at Fr=0.53 (bottom).
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

(a)

(b)

(c)
Fig. 4. CFDShip-Iowa V4 predictions for (a) streamwise velocity profile and cross-plane velocity streamline,
(b) turbulent kninetic energy and (b) shear stress ̅̅̅̅̅ at nominal wake plane x/L=0.935 using isotropic (BKW) and anisotropic
(ARS) models for straight ahead 5415 simulations on 50M grid at Re = 5.13106, Fr = 0.28 are compared with
experimental data (EFD).
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

(a)

(b) 50M grid-DES (c) 9M grid, URANS (d) 9M grid-DES

Fig. 5. (a) Isosurfaces of Q=300 for instantaneous solution DES solution on 9M grid. Inset figures are obtained using averaged
solution. Three different types (A, B and C) of juncture vortices are marked and associated dominant frequency modes are shown.
Contours are of the absolute pressure with levels from -0.5 to 0.1 at an interval of 0.02. Vortical structures at the transom corner
obtained using (b) DES on 50M grid and (c) URANS on 9M grid. (d) Instantaneous flow separation at Y = 0.01 plane is shown for
DES on 9M grid.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012
Fore-body keel
vortex

Sonar dome
vortex

(a)

Sonar dome Fore-body

vortex keel vortex

(b) (c)

(d) (e)
Fig. 6. (a) Vortical structures predicted by CFDShip-Iowa V4 using DES model on 300M grid for straight ahead 5415 at Re =
5.13106, Fr = 0.28. The flow does not show small scale turbulent structures, but resolves vortical structures and their interaction with
the boundary layer very well. Contours of the streamwise vorticity at (b) x/L = 0.2, (c) x/L = 0.6 and (d) x/L = 0.8 are compared with
experimental data. (e) Streamwise velocity profile and cross-plane velocity streamline at nominal wake plane is compared with
experimental data.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 7. Vortex system of KVLCC2 (isosurface of Q 200 colored by helicity) at 0 30 : (a) bow view and (b)
bottom view.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

(a)

(b)
Fig. 8. (a) Large scale vortical structures and instabilities are identified for 5415 at  = 20 static drift using CFDShip-Iowa V4 DES
simulations. The inset on right topmost corner shows the small scale structures predicted on 250M grid. (b) Initial comparison of
streamwise velocity contour and cross flow vectors and wake at X=0.935 shows very good agreement with the ongoing experiments.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

(a)

Re = 2.2107 Re = 2.0109
(b) (c)
Fig. 9. (a) Fully appended and barehull Athena resistance predictions for model and full scale using near wall turbulence model
(BKW) and wall-functions (WF) are compared with experimental data and ITTC line. Local flow field for model and full scale
barehull Athena: (b) boundary layer profiles colored by streamwise velocity; and (c) transom free surface wave elevation contour for
Fr=0.48. As expected, the full-scale boundary layer is thinner than in model-scale, and the free surface elevation pattern is not
significantly affected by the Reynolds number.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 10. Examples of air flow modeling for ONR Tumblehome in static conditions

Fig. 11. Examples of CFDShip-Iowa ship exhaust plumes simulations for ONR Tumblehome during a broaching event.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 12. Instantaneous air/water interface colored by elevation for DTMB 5512 at Fr = 0.41

Fig. 13. Instantaneous vortical structures colored by streamwise vorticity for DTMB 5512 at Fr = 0.41 for air phase, Q = 500.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 14. URANS free-surface wave elevations for calm water resistance, sinkage and trim computations of T-Craft (ACV/SES) at Fr =
0.2 – 0.6 (Bhushan et al., 2012b).

Fig. 15. Resistance and propulsion curves for Athena obtained using the single-run procedure (Xing et al., 2008).
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 16. Free surface and hull pressure distributions at an instance during regular head wave simulation for side by side ship-ship
interactions for Hope and Bobo (Mousaviraad et al., 2011).

Fig. 17. Captive pure yaw maneuvering for ONR Tumblehome including wind effects (Mousaviraad et al., 2012b).

Fig. 18. Parametric rolling studies for ONR Tumblehome (Sadat-Hosseini et al., 2010).
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 19. Free surface colored with the absolute velocity and Isocontours of total vorticity during turning circle maneuvering in calm
water for 5415M with body force propeller model (Carrica et al., 2012b)

(a) (b)
Fig. 20. Free running simulations for T-Craft (SES/ACV): (a) turning circle in calm water with water jet propulsion; (b) straight ahead
free running with air-fan propulsion (Mousaviraad et al., 2012a)

(a) (b) (c)

Fig. 21. Improving system based predictions using CFD outputs along with system identification technique: (a) The grid topology for
CFD model of free running ONR Tumblehome with body force propeller model; (b) Improved system based prediction for zigzag
simulation in calm water; (c) Improved system based prediction for zigzag in following waves (Araki et al., 2012a,b).
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

(a) (b)
Fig. 22. Broaching simulation of free running ONR Tumblehome in following waves using: (a) body force propeller model (Sadat-
Hosseini et al., 2011); (b) actual propeller (Carrica et al., 2012b)

Fig. 23. The grid topology and the instant view of CFD solution during course keeping in beam waves for 5415M with active fins and
body force propeller model (Sadat-Hosseini et al., 2012a).

Fig. 24. CFD simulation of course keeping in irregular beam waves with JONSWAP spectrum for 5415M with active fins and body
force propeller model (Sadat-Hosseini et al., 2012b).
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 25. CFD solution of flooding free SSRC cruiser with two-room compartment in beam waves at six instants (Sadat-Hosseini et al.,
2012c).

Fig. 26. Two-phase flow past a surface-piercing cylinder. Left: Instantaneous air-water interface; Right: instantaneous vortical
structures colored by pressure, .
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 27. Wave profiles for both experiment (Waniewski et. al, 2002) and simulation.

Fig. 28. Close-up view of the bow sheet breakup Top: Experiment (Karion et al., 2004); Bottom: Simulations.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 29. Wave breaking process behind a submerged bump. Left: Experiment; right: Simulation.

Fig. 30. Top view of the wave breaking behind a submerged bump, grid 768×256×64.
 29th Symposium on Naval Hydrodynamics
Gothenburg, Sweden, 26-31 August 2012

Fig. 31. Close-up view of the wave breaking region for the bump flow, grid 2.2 billion.