840

2012,24(6):840-847
DOI: 10.1016/S1001-6058(11)60311-9

NUMERICAL PREDICTION OF SUBMARINE HYDRODYNAMIC COEFFICIENTS USING CFD SIMULATION*

PAN Yu-cun
State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
Department of Naval Architecture, Naval University of Engineering, Wuhan 430033, China,
E-mail: [email protected]
ZHANG Huai-xin
State Key Laboratory of Ocean Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
ZHOU Qi-dou
Department of Naval Architecture, Naval University of Engineering, Wuhan 430033, China

(Received February 21, 2012, Revised September 3, 2012)

Abstract: The submarine Hydrodynamic coefficients are predicted by numerical simulations. Steady and unsteady Reynolds
Averaged Navier-Stokes (RANS) simulations are carried out to numerically simulate the oblique towing experiment and the Planar
Motion Mechanism (PMM) experiment performed on the SUBOFF submarine model. The dynamic mesh method is adopted to
simulate the maneuvering motions of pure heaving, pure swaying, pure pitching and pure yawing. The hydrodynamic forces and
moments acting on the maneuvering submarine are obtained. Consequently, by analyzing these results, the hydrodynamic
coefficients of the submarine maneuvering motions can be determined. The computational results are verified by comparison with
experimental data, which show that this method can be used to estimate the hydrodynamic derivatives of a fully appended submarine.
Key words: submarine maneuverability, hydrodynamic coefficients, Planar Motion Mechanism (PMM), dynamic mesh

Introduction 
During the submarine scheme design period, the
evaluation of maneuverability and stability is an important task. In practice, the six degree of freedom
maneuvering motion is decoupled into the horizontal
and the vertical motions, thus the problem can be simplified into a set of linear equations. Therefore, the
estimation of the hydrodynamic coefficients of these
motion equations is a key step to predict the motion of
the submarine.
Traditionally, the methods to predict the hydrodynamic derivatives of a submarine could be classified into three types: the semi-empirical method, the

* Project supported by the National Natural Science Foundation of China (Grant No. 11272213).
Biography: PAN Yu-cun (1980- ), Male, Ph. D. Candidate,
Lecturer
Corresponding author: ZHANG Huai-xin,
E-mail: [email protected]

potential flow method and the captive-model experiments including the oblique towing tests, the rotating
arm experiments and the Planar Motion Mechanism
(PMM) tests[1-4].
With the semi-empirical method, the complicated
submarine shape usually could not be taken into full
account. The potential theory could predict the inertial
hydrodynamic coefficients satisfactorily, but with the
viscous terms neglected. The PMM experiment may
be the most effective way, but it requires special facilities and equipment and it is both time-consuming
and costly, as not economical at the preliminary design stage.
An alternative method for determining the hydrodynamic derivatives is to use the Reynolds Averaged
Navier-Stokes (RANS) simulations to simulate the
captive-model tests numerically. The steady state CFD
was successfully applied to simulate the straight line
captive-model test for assessing the velocity based
coefficients of submerged vehicles. Tyagi and Sen[5]
investigated the transverse velocity based coefficients
of two typical Autonomous Underwater Vehicles

841

(AUV) using the RANS solver. Wu et al.[6] numerically simulated the steady straight line motion of the
SUBOFF model with and without attack angle, close
to an infinite level bottom. The motion-near-bottom
effects on the hydrodynamic force were investigated.
Hu et al.[7] used the CFX software to simulate the
maneuvering tests. The k  H model was adopted to
compute the positional hydrodynamic coefficients,
and the k  Z model was used to compute the rotational and other coupling hydrodynamic coefficients
for CR-02 AUV. The motion prediction based on
these calculated hydrodynamic coefficients enjoyed a
good agreement with the test at a lake.
From these static maneuvering motions mentioned above, only the velocity-based hydrodynamic
coefficients can be determined. In order to determine
the acceleration-based hydrodynamic coefficients, unsteady experiments such as the PMM tests should be
performed. An up-to-date application of the CFD to
the marine maneuverability can numerically simulate
the virtual PMM experiments to compute the unsteady
hydrodynamic forces and moments. Broglia et al.[8]
used a parallel CFD code to investigate the flow
around the KVLCC2 tanker during the pure swaying
maneuvering, with consideration of the free surface
effects. The motion of the vessel was simulated using
an overlapping mesh method with 8 blocks for a fixed
background and 20 fitted blocks moving with the hull.
The computed lateral force and the yaw moment agree
well with the experimental data with a relative error
less than 5.5% and 20%, respectively. Yang et al.[9]
simulated the flow around a naked KVLCC1 hull
undergoing the pure swaying motion in deep and shallow waters, with the effect of free surface ignored.
The aim of the present study is to explore the
possibility of developing a numerical method to evaluate the maneuvering characteristics of a submarine,
especially at the earlier stage of the design cycle. The
virtual towing tank and the PMM experiments are
conducted using the RANS solver to compute the
hydrodynamic forces and moments and the resultant
coefficients.
1. Numerical approach
1.1 Governing equations
Numerical simulations are performed with the
CFD software Ansys Fluent. The flow around the
vehicle is modeled using the incompressible, RANS
equations:

w (ui )
=0
wxi
w ( U ui )
wu
wP
+
+ U u j i = U Fi 
wt
wx j
wxi

(1)

w
wx j

w ui

 U uicu cj
P

wx j

(2)

where ui is the time averaged velocity components

in Cartesian coordinates xi (i = 1, 2, 3) , U is the

fluid density, Fi is the body forces, P is the time

averaged pressure, P is the viscous coefficient, uic
is the fluctuating velocity components in Cartesian
coordinates, and  U uicu cj is the Reynolds stress
tensor.
The finite volume method is employed to discretize the governing equations with the second-order upwind scheme. The Semi-Implicit Method for the Pressure-Linked Equations (SIMPLE) is used for the pressure-velocity coupling. In order to allow the closure
of the time averaged Navier-Stokes equations, various
turbulence models were introduced to provide an estimation of the  U uicu cj . Here, the realizable k  H
model is chosen[10], as is used for applications in a
wide range of flows due to its robustness and economical merit, and the standard wall function is applied
for a better analysis of the turbulent viscous flow
around the wall.
1.2 Description of the model
The target studied in this paper is the SUBOFF
model[11], designed by the David Taylor Research
Center (DTRC). A series of captive-model experiments[12] were performed in the David Taylor Model
Basin on the towing carriage to measure the hydrodynamic force and moment acting on the model.
The entity model is a body of revolution, which
has a sail, no bow planes, two horizontal planes and
two vertical rudders, and a ring wing supported by
four struts in an X configuration. The overall
length of the SUBOFF model is 4.356 m, while the
length between the perpendiculars is 4.261 m, the
maximum diameter is 0.508 m.
The six degree of freedom motion of the submarine is normally described using two coordinate systems. The first is a right-handed, body-fixed coordinate system, with its origin at a point 2.013 m aft of
the forward perpendicular on the hull centerline. The
x -axis is positive pointing upstream. The y -axis is
positive pointing starboard and the z -axis is positive
pointing downward.
The second coordinate systen, an inertial reference frame, is used to define the translational and rotational motions of the body-fixed coordinate system in
the earth-fixed coordinates, as shown in Fig.1. In this
coordinate syetem, the position of the vehicles coordinate system is then expressed in [ , K , ] coordinates. The orientation of the body-fixed coordinate

842

system is described by Euler angles \ (yaw), T

(pitch), I (roll).

them. The grid is generated by such a hybrid mesh

strategy: in the inner region, a multi-block hexahedral
mesh is used to define the fluid immediately surrounding the vessel, which allows a detailed control of
the mesh parameters and the element quality. The
hexahedral mesh is also used in the outer region,
which is rather coarse, so the number of grids can be
reduced. The intermediate layer consists of unstructured tetrahedral grids, which can be conveniently remeshed in the case of the element deformation, see
Figs.3 and 4. The geometry modeling and the grid
generation are done by using the Gambit software.

Fig.1 Principal earth-fixed and body-fixed coordinate systems

Fig.4 Meshes of three sub-regions

Fig.2 Boundary condition for numerical simulation

1.3 Boundary conditions

The boundary conditions around the submarine
model are as follows: The inlet boundary is positioned
1.5 body-length upstream with an inflow velocity of
4 m/s (Reynolds number of 1.693 107 based on the
vehicle length), a pressure-outlet condition is defined
3 body-lengths downstream. Free slip wall boundary
conditions are applied to the 4 remaining walls 9 diameters away from the model and a no-slip boundary
condition is applied to the hull. Figure 2 shows the
boundary conditions for numerical simulation.

Fig.3 Surface grid of the model

1.4 Mesh definition

In order to simulate the motion of the model, the
fluid domain is split into three regions: an inner region,
an outer region, and an intermediate layer between

To numerically simulate the oscillatory motion

produced in the PMM tests, the User Defined Function (UDF) is used to control the motion of the vessel.
The outer domain remains fixed in space, while the
inner region containing the SUBOFF model moves or
rotates to simulate the motion induced in the PMM
experiment. It should be noted that the mesh in the
inner region remains locked in a position relative to
the motion of the vessel. Hence, the mesh of the intermediate layer is deformed to accommodate the motion of the inner region. The new node locations are
updated at each time step according to the calculation
of the UDF, while the overall mesh topology is maintained. Such a treatment guarantees a high quality of
the meshes around the vessel during the maneuvering
motion.

2. The grid independence

Before the CFD analysis, the sensitivity of the
solution to the resolution of the grid should be determined. Based on an initial grid, a series of successively refined grids were generated. The results from
the base grid and the refined grids were compared to
check the result variation with the grid refinement.
With the limited computational resource, in the
inner region the grid was refined with a ratio of 2
in three directions: the longitudinal, the transversal
and the normal directions, while in the intermediate

843

Table 1 Mesh sensitivity

Drift angle (o)

Mesh

Grid quantity

y+

X (N)

Y (N)

N (Nm)

Fine

14.36M

30

188.1370500

1.052097700

0.8278320

Medium

7.830M

43

191.9110500

0.821038280

0.2037840

Coarse

4.330M

60

189.3720700

1.908503300

0.4878590

Fine

14.36M

30

178.9098519

205.0880242

446.78590

Medium

7.830M

43

178.4934547

203.1697727

451.90772

Coarse

4.330M

60

181.2452483

215.5973220

451.44043

layer and the outer region, the refinement ratio was

less than 2 . The computational grids contained
from 4.33u106 cells for the coarse grid to 1.436u107
cells for the finest grid.
Besides the grid density, the y + value, i.e., the
thickness of the first cells adjacent the hull, is also
related with the accuracy of the numerical prediction.
In this study, the y + value varied from 30 for the
finest mesh to 60 for the coarsest. The mesh convergence test was carried out focusing on the forces and
the moments on the vehicle with a drift angle of 0o
and 3o. Table 1 summarizes the longitudinal forces
X , the lateral forces Y and the yaw moments N
computed in the tested mesh cases.
From the results in Table 1, it is seen that, on the
whole, the solutions do not change significantly from
the fine grid to the coarse grid, with only minor differences between results obtained by the fine and
medium grids, and a little bit larger differences
between the results obtained by the coarse and
medium grids. Therefore, the medium grid was chosen
for the maneuvering prediction.
Normally, the time step convergence investigation is a necessary step for an unsteady CFD simulation. In Turnocks paper[13], simulations with 50, 100
and 500 time steps per oscillation cycle were performed. For all three cases, the variations of the sway
force and the yaw moment were stabilized after less
than a quarter of an oscillation cycle.
Since transient simulations are required to solve
the multiple coefficient loops at each time step, further
time step refinement would be difficult due to the
hardware limitations. Indeed, the study of the temporal discretization convergence would significantly increase the overall cost. With a due review of the published data in literature[13-16], the scheme with 400
time steps per oscillation cycle was adopted for the
following transient computations.
3. Numerical simulation of maneuvering motion

3.1 Simulation of oblique towing tests

For the marine hydrodynamics, the CFD techni-

que has been developed mainly in the fields of resistance and propulsion. As is known, the oblique
towing test is a direct and explicit means of determining the static coefficients and is very similar to the
resistance test, except that the model has a fixed attitude during towing. Obviously, it is a logical and natural way to carry out the simulation of the steady oblique towing as an example of the application of the
CFD to the field of maneuverability.

Fig.5 Comparison of the computed and measured lateral force

Y c and yawing moment N c

The model was towed with a straight path at a

constant velocity, while for each run the model was
set at a prescribed pitch angle or heading angle. Thus
the heave velocity v or the sway velocity w of the
model change according to the following rule
v = U f sin E , w = U f sin D

(3)

844

Table 2 Comparison of hydrodynamic force/moment coefficients due to transverse velocity

Yvc

N vc

Z wc

M wc

Computed

0.028416

0.013739

0.01470

0.010090

55Experiment

0.027834

0.013648

0.01391

0.010324

CFD  Exp.
Exp.

2.09%

0.67%

5.68%

2.27%

where U f is the uniform inflow velocity and D is

the attack angle, E is the drift angle.
The computed hydrodynamic force and moment
are non-dimensionalized as follows
Yc=

Y
N
, Nc =
1
1
UU f2 L2
UU f2 L3
2
2

(4)

where L is the length of the model. As shown in

Fig.5, the variation trends of the transverse force and
the moment with the drift angles are well predicted.
The agreement between the numerical results and the
experimental data[12] is good. Similarly, the simulated
and the measured normal force Z c and the pitching
moment M c are also in good agreement.
The values of the static derivatives are determined from the slopes at the origin of the curves of the
force and moment coefficients versus the normal or
the lateral velocity. The results of the calculation and
the experiment are compared in Table 2. The correlations between the model tests and the CFD simulations are very good, with less than 5% of discrepancy.
Based on this comparison, it can be preliminarily concluded that the CFD calculation is adequate for the investigation of the influences of the oblique velocity on
the steady motion of the submarine model.
3.2 Numerical simulation of planar motion mechanism
tests
The PMM generates two kinds of motions: the
translation and the rotation, imposed on the vehicle as
it travels down the tank at a constant forward velocity.
The sinusoidal motion can be designed in such a
manner as to produce the desired conditions of hydrodynamically pure heaving, pure swaying, pure
yawing and pure pitching. Thus, the rotary-based
and acceleration-based coefficients can be explicitly
determined.
The system is designed for obtaining the hydrodynamic characteristics of deeply submerged bodies
in either the vertical or horizontal planes of motion.
For simplicity, the mode of operation applied to the
submarine in the vertical plane is discussed here.

Fig.6 Trajectory of model during pure heaving test

3.2.1 Pure heaving

During the pure heaving motion, the models
Center of Gravity (CG) moves in such a sinusoidal
path that the pitch angle T remains zero, as shown in
Fig.6. The variation of the vertical displacement, the
velocity and the acceleration are given by the following equations.

] = a sin Z t , T = T = 0 , w = aZ cos Z t ,
w = aZ 2 sin Z t

(5)

where T and T are the angle and the angular velocity in the direction of rotating around the y axis,
w and w are the vertical velocity and the acceleration, a is the amplitude, Z is the circular frequency of the heaving motion.

Fig.7 Force and moment acting on the hull during pure heaving
test ( f = 0.2 Hz)

845

Fig.8 The oscillating wake pattern behind SUBOFF in pure heaving motion

The normal force and the pitch moment acting on

the hull are monitored as the model oscillates with
different frequencies (0.1 Hz, 0.2 Hz and 0.3 Hz). The
unsteady RANS simulations are performed for 8
cycles, with the first 7 cycles allowing the system to
settle down before the monitoring. Figure 7 shows the
time history of the variation of the force and the
moment acting on the hull with the oscillating frequency of 0.2 Hz. From these results, the translatory velocity-based and the acceleration-based coefficients Z wc ,
M wc , Z wc , M wc can be determined using the Fourier
expansion, as shown in Table 3.

velocity field around the vehicle is given on Fig.8.

Fig.9 Trajectory of model during pure pitching test

Table 3 The hydrodynamic coefficients of the SUBOFF

model
Computed

Experimental

CFD  exp.
exp.

Z wc

0.0181100

0.014529

24.65%

M wc

0.0006250

0.000561

11.41%

Z wc

0.0157000

0.013910

12.87%

M wc

0.0080190

0.010324

22.33%

YVc

0.0189900

0.016186

17.32%

NVc

0.0005665

0.000396

43.56%

Yvc

0.0303700

0.027834

9.110%

N vc

0.0131200

0.013648

3.870%

Z qc

0.0006292

0.000633

0.600%

M qc

0.0009468

0.000860

10.09%

Z qc

0.0077810

0.007545

3.130%

M qc

0.0034680

0.003702

6.320%

Yrc

0.0003602

0.000398

9.500%

N rc

0.0010130

0.000897

12.93%

Yrc

0.0046610

0.005251

11.24%

N rc

0.0042710

0.004444

3.890%

As the SUBOFF model oscillates vertically, the

flow pattern around the vehicle varies with time. As
an example, an instantaneous representation of the

Fig.10 Force and moment acting on the hull during pure pitching test ( f = 0.2 Hz)

3.2.2 Pure pitching

The pure pitching motion is one in which the
model CG moves in a sinusoidal path while the model
axis remains tangent to the path, that is, the angle of
attack D remains zero, as shown in Fig.9. In this
case, the pitch angle traces (T , T, T) are of primary
interest.

T = T 0 sin Z t , q = T = T 0 Z cos Z t ,
q = T = T 0 Z 2 sin Z t , w = w = 0

(6)

where T0 is the amplitude, q and q are the angular velocity and the acceleration in the direction of rotating around Y axis. The normal force and the pitch
moment acting on the hull are monitored as the model
oscillates with different frequencies (0.1 Hz, 0.2 Hz
and 0.3 Hz). The time history variation of the force

846

Fig.11 The oscillating wake pattern calculated behind SUBOFF in pure pitching motion

and the moment acting on the hull with the oscillating

frequency of 0.2 Hz is shown in Fig.10. From these
results, the rotary rate-based and the accelerationbased coefficients Z qc , M qc , Z qc , M qc can be determined using the Fourier expansion, as shown in
Table 3.
The flow pattern around the vehicle is shown in
Fig.11. In this case, the characteristics of the wake
pattern are different from those shown in Fig.8. In
studying the velocity field, it can be seen that the rotary motion imposed on the SUBOFF model has a distinct impact on the flow asymmetry and loading.
3.2.3 Results
The motion of the pure swaying is similar to that
of the pure heaving, and the motion of the pure
yawing is similar to that of the pure pitching. From the
results of the above four virtual PMM tests, 16 hydrodynamic coefficients can be determined, as shown in
Table 3.
The resulting predictions of the hydrodynamic
coefficients show a good correlation with the experimental data. Most of the discrepancies between the
predicted hydrodynamic coefficients and the measured
values are in the range 0.5%-15% except for a few
cases, which are in an acceptable level of accuracy for
the preliminary design. The possible source of error
lies in the insufficient mesh resolution and the inadequacy of the turbulence model[16].
In general, there are several characteristics worth
mentioning. First, the coefficients Z wc , M wc , Yvc ,
N vc determined from the oblique towing simulations
are in better agreement with the experimental data
than those from the PMM simulations. Second, the
precision of the acceleration-based
leration-based coefficients is
lower than that of the velocity-based coeff
oeffiicien
cients.
Among the 16 coefficients, the coefficient NVc has
the largest discrepancy, but its value still remains in
the same order of magnitude with the experimental
ones.
3.3 Discussions
The overall results indicate that the RANS
method can predict the hydrodynamic coefficients in
the same level of accuracy as the model test based

method. However, the present method might be improved.

In the submarine design, the computational cost
may be the main concern that prevents the application
of the CFD method. In this study, the unsteady simulations are carried out on a desktop PC using an Intel
Core Processor i5 2500 with 16 GB of RAM. A PMM
simulation takes approximately 100 h to complete.
Thus the total runtime for a set of linear coefficients
would be 16 d on a single machine.
Another important issue is that in the preliminary
design stages for the submarine maneuverability, a
great number of iterations is required, with considerations of various combinations of sizes, locations and
configurations of the control surfaces. During such a
design process, the changes in geometry in the iteration will make the problem very complex in using the
CFD methods and a vast amount of time will be consumed to generate the CFD meshes.
Therefore, a compromise between the semi-empirical method and the RANS method would be more
attractive. That is, at the first estimate loop, the semiempirical method is used to determine a small number
of design alternatives. And they will be further assessed with the RANS method during the next optimization loop. The accuracy of these predictions is balanced with the calculation speed.

4. Conclusions
A method for the unsteady RANS simulation for
the submarine maneuverability is proposed. The
method can successfully be used to calculate the flow
around a submarine model, and the force and the
moment during the steady oblique towing and dynamic PMM motion. The predictions of the static, rotary,
acceleration coefficients of the submarine model
enjoy an acceptable level of accuracy. The CFD
method is shown to be able to provide a good estimate
of the maneuvering coefficients for the fully appended
submarine model. However, more studies ae required
for more advanced turbulence models, finer grid resolutions and additional verifications and validations
before such simulations can be applied with a high
degree of confidence.
The main drawbacks of using the CFD method in

847

the submarine design are that it is time-consuming to

obtain the flow solution and to carry out the mesh
generation. Advances in the parallel computing and
the processor speed can reduce the total simulation
time. In view of the iterative design changes, the automatic grid generation should be a desirable technique.

References
[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

KIM Y. G., KIM S. Y. and KIM H. T. et al. Prediction

of the maneuverability of a large container ship with
twin propellers and twin rudders[J]. Journal of Marine
Science and Technology, 2007, 12(3): 130-138.
LI Gang, DUAN Wen-yang. Experimental study on the
hydrodynamic property of a complex submersible[J].
Journal of Ship Mechanics, 2011, 15(1): 58-65(in
Chinese).
OBREJA D., NABERGOJ R. and CRUDU L. et al. Identification of hydrodynamic coefficients for manoeuvring simulation model of a fishing vessel[J]. Ocean
Engineering, 2010, 37(8): 678-687.
FAN Shi-bo, LIAN Lian and REN Ping et al. Oblique
towing test and maneuver simulation at low speed and
large drift angle for deep sea open-framed Remotely
operated vehicle[J]. Journal of Hydrodynamics, 2012,
24(2): 280-286.
TYAGI A., SEN D. Calculation of transverse hydrodynamic coefficients using computational fluid dynamic
approach[J]. Ocean Engineering, 2006, 33(5-6): 798809.
WU Ban-shan, XING Fu and KUANG Xiao-feng et al.
Investigation of hydrodynamic characteristics of submarine moving close to the sea bottom with CFD
methods[J]. Journal of Ship Mechanics, 2005, 9(3):
19-28.
HU Zhi-qiang, LIN Yang and GU Hai-tao. On Numerical computation of viscous hydrodynamics of unmanned underwater vehicle[J]. Robot, 2007, 29(2): 145150(in Chinese).
BROGLIA R., MASCIO A. D. and AMATI G. A. parallel unsteady RANS code for the numerical simulations
of free surface flows[C]. 2nd international Conference on Marine Research and Transportation. Naples,
Italy, 2007.

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

YANG Yong, ZOU Zao-jian and ZHANG Chen-xi.

Calculation of hydrodynamic forces on a KVLCC hull
in sway motion in deep and shallow water[J]. Chinese
Journal of Hydrodynamics, 2011, 26(1): 85-93(in
Chinese).
PHILLIPS A. B., TURNOCK S. R. and FURLONG M.
Influence of turbulence closure models on the vortical
flow field around a submarine body undergoing steady
drift[J]. Journal of Marine Science and Technology,
2010, 15(3): 201-217.
GROVES N., HUANG T. T. and CHANG M. S. Geometric characteristics of DARPA SUBOFF models
(DTRC Models Nos. 5470 and 5471)[R]. DTRC/SHD
1298-01, 1989, 1-75.
RODDY R. F. Investigation of the stability and control
characteristics of several configurations of the DARPA
SUBOFF model (DTRC model 5470) from captivemodel experiments[R]. DTRC/SHD 1298-08, 1990, 1108.
TURNOCK S. R., PHILLIPS A. B. and FURLONG M.
et al. URANS simulations of static drift and dynamic
manoeuvres of the KVLCC2 tanker[C]. Proceeding of
SIMMAN International Manoeuvring Workshop.
Copenhagen, Demark, 2008.
PHILLIPS A. B., TURNOCK S. R. and FURLONG M.
The use of computational fluid dynamics to aid costeffective hydrodynamic design of autonomous underwater vehicles[J]. Journal of Engineering for the
Maritime Environment, 2010, 1(1): 1-16.
PHILLIPS A. B., FURLONG M. and TURNOCK S. R.
Virtual planar motion mechanism tests of the autonomous underwater vehicle autosub[C]. STG-Conference/Lectureday CFD in Ship Design. Hamburg,
Germany, 2007.
STERN F., AGDRUP K. and KIM S. Y. et al. Experience from SIMMAN 2008-the first workshop on verification and validation of ship maneuvering simulation
methods[J]. Journal of Ship Research, 2011, 55(2):
135-147.