Scientific Journals Zeszyty Naukowe

of the Maritime University of Szczecin Akademii Morskiej w Szczecinie

2020, 62 (134), 124–135
ISSN 1733-8670 (Printed) Received: 24.02.2019
ISSN 2392-0378 (Online) Accepted: 12.05.2020
DOI: 10.17402/427 Published: 29.06.2020

Experimental and numerical determination

of the hydrodynamic coefficients
of an autonomous underwater vehicle

Abdollah Sakaki1, Mohsen Sadeghian Kerdabadi2

Amirkabir University of Technology, Tehran, Iran
e-mail: 1 [email protected], 2 [email protected]

corresponding author

Key words: Reynolds averaged Navier–Stokes, computational fluid dynamics, linear coefficients, non-linear
coefficients, underwater vehicle
Abstract
Computational fluid dynamics (CFD) has progressed rapidly in the past fifty years and is now used in many
industrial fields, such as air, space, and marine engineering. CFD has an irreplaceable role in marine design and
scientific research, and its applications within this field continue to grow with the development of computers.
CFD is used to quickly and inexpensively simulate fluid behaviour using the Reynolds Averaged Navier–Stokes
(RANS) equations to calculate hydrodynamic coefficients, which are needed in manoeuvrability studies of
underwater vehicles (UWV). Here, these computations are performed for six geometrical shapes that represent
typical autonomous underwater vehicles (AUVs) currently in use. Resistance test simulations at up to 20o drift
angles were conducted for AUVs with different length-to-diameter ratios. The results were compared with the
experimental data and current quasi-experimental relationships, which suggested that the CFD predictions were
adequately precise and accurate. These predictions indicated that there was a non-linear relationship between
forces and moments and the lateral speed. Moreover, both linear and non-linear hydrodynamic coefficients were
calculated.

Introduction rotating arm experiments, and planar motion mech-

anism (PMM) tests (Kim et al., 2007; Obreja et
An analysis system based on CFD simulations al., 2010; Li & Duan, 2011; Fan et al., 2012; Pan,
has been developed to calculate the hydrodynam- Zhang & Zhou, 2012).
ic coefficients and hydrodynamic forces acting on UWVs can be classified as either manned or
UWVs. Numerical approaches have improved to unmanned systems. Manned systems can be catego-
a level of accuracy that allows them to be applied rized into two sub-classes: military submarines and
for practical marine vehicle resistance and indus- non-military submersibles, such as those operated to
trial propulsion computations. Recently, CFD has support underwater investigations and assessments.
also been used to determine hydrodynamic coeffi- Unmanned submersibles fall into several different
cients, which are required to evaluate the manoeu- sub-classes. The simplest and most easily described
vring characteristics of underwater marine vehi- are submersibles that are towed behind a ship, and
cles. Traditionally, the hydrodynamic coefficients the second type is called a remotely operated vehicle
of UWVs have been predicted using one of three (ROV), which is a tethered vehicle. The third type of
methods: potential flow methods, semi-empirical unmanned submersible is an unmanned untethered
methods, and experimental model tests. The last vehicle (UUV), which contains its own onboard
can be further divided into oblique towing tests, power but is controlled by a remote operator via

124 Scientific Journals of the Maritime University of Szczecin 62 (134)

 Experimental and numerical determination of the hydrodynamic coefficients of an autonomous underwater vehicle

a communications link. An autonomous underwater become increasingly used in the marine industry
vehicle (AUV) is an undersea system with its own during the design stage. Numerical methods are so
power that controls itself while performing a pre-de- precise and accurate that they are used in resistance
fined task. A further distinction between AUVs and and thrust calculations. In recent years, the CFD
UUVs is that AUVs require no communication method has also been used to calculate the hydrody-
during their mission, whereas UUVs require some namic coefficients for manoeuvrability predictions
level of communication to complete their mission (Ferziger, Peric & Leonard, 1997; Tyagi & Sen,
(Blidberg, 2001). 2006; Williams et al., 2006).
AUVs are automatic, intelligent vehicles that can Reynolds Averaged Navier–Stokes (RANS)
cover a given path and undertake a variety of activi- equations were first used to solve maritime prob-
ties, such as land surveying, identification, and mine lems more than 20 years ago, but they produced
detection, over a vast area without using cables. unsatisfactory results (Wilson, Paterson & Stern,
These ocean vehicles must be capable of rapid deci- 1998; Gentaz et al., 1999). With increased comput-
sion-making to cope with different conditions; there- ing capacities and recent progress in RANS models,
fore, their manoeuvrability and controllability are of great advances have been made in this field, and
great importance. In recent years, there have been CFD has become a crucial tool for various aspects of
many efforts to develop these vehicles. To design an UWV hydrodynamics for both research and design.
AUV, its manoeuvrability and controllability must One of the most recent and important applications
be accurately calculated using mathematical models. of CFD in the marine industry is its use to compute
The current mathematical models provide hydro- the hydrodynamic coefficients of marine vehicles
dynamic forces and moments, which are defined during captive model simulations. For example,
as hydrodynamic coefficients; therefore, accurately Sarkar developed a new computationally efficient
calculating these coefficients is essential for simulat- technique to simulate 2-D flow over axisymmetric
ing an AUV’s performance. AUVs using the PHOENICS CFD package (Sarkar,
Hydrodynamic coefficients are divided into three Sayer & Fraser, 1997). Fluent CFD code was used
types: linear damping coefficients, linear inertia to obtain the hydrodynamic coefficients of 3-D fins
coefficients, and non-linear damping coefficients. and an AUV, as well as to compute the linear and
Among these, temporal linear coefficients, in which nonlinear hydrodynamic coefficients of a SUBOFF
an AUV moves in a steady flow, affect the manoeu- submarine in an unrestricted fluid flow (Ray, Singh
vrability, while in unsteady flow, linear inertia coef- & Seshadri, 2009; Nazir, Su & Wang, 2010). The
ficients and non-linear coefficients should be taken transverse hydrodynamic coefficients of an AUV
into consideration. These coefficients are usually were computed using a commercial CFD package
determined by conducting system identification tests (Tyagi & Sen, 2006), and the hydrodynamic forc-
or experimental formulas, whose precision depends es and moments acting on an AUV due to control
on different conditions, such as the hull geometry surface deflection were investigated using ANSYS
and operating conditions. Most of these methods Fluent software (Dantas & de Barros, 2013). CFD
can only calculate linear coefficients, and there is was used to construct a platform for AUV hull shape
currently no experimental relationship for calculat- optimization, but only primitive particle swarm
ing the hydrodynamic coefficients of an AUV, and optimization (PSO) and multi-island genetic algo-
the application of ship or submarine-related rela- rithm (MIGA) methods were compared (Gao et al.,
tionships for an AUV produces many errors. Tem- 2016).
poral system identification is useful when adequate In this study, a new method is proposed for sim-
experimental data are available to design a suitable ulating hydrodynamic tests using a CFD software
mathematical model. Another problem is that during named CD-adapco, which discretizes Navier–Stokes
the initial design, manoeuvrability is not applicable. equations through the finite volume method with
Although PPM tests are one of the most common respect to the boundary and initial conditions. The
methods for calculating coefficients, they are expen- software was used to solve the final equation system,
sive and require long calculation times with many which allowed the hydrodynamic coefficients used
experimental errors. in the mathematical manoeuvre model to be calcu-
Several methods are used to obtain hydrody- lated. This computational model was highly suit-
namic coefficients, including theoretical approach- able, cost-effective, and flexible to use. Six models
es, semi-empirical formulas, captive model tests, with different length-to-diameter ratios and different
and CFD. As computers have developed, CFD has speeds were calculated and compared.

Zeszyty Naukowe Akademii Morskiej w Szczecinie 62 (134) 125

 Abdollah Sakaki, Mohsen Sadeghian Kerdabadi

Figure 1. Illustration of Myring AUV (Gao et al., 2016)

Description of the model where x is the axial distance to the nose tip; a, b, and
c are the lengths of the nose, middle and tail, respec-
A Myring-type body was used to parameter- tively; d is the middle hull diameter; n is the index of
ize the AUV hull shapes and was chosen due to its the nose shape; and θ is semi-angle of the tail. The
streamlined characteristics. The Myring AUV class section of the Myring AUV and Model dimensions
has already applied in aircraft fuselage and other are shown in Figure 1 and Table 1.
AUVs such as Maya and Remus, Pirajuba, and Gua-
nay II. A Myring-type AUV has three distinct parts: Table 1. Model Dimensions
the nose section, the middle body cylindrical section, Parameter Value Unit
and the tail section. The nose and the tail sections are 1 Hull maximum diameter (d) 0.234 m
defined by a semi-elliptical radius distribution along 2 Tail length (c) 0.279 m
the main axis, as given by the following equations 3 Nose length (a) 0.217 m
(Myring, 1976): 4 Middle body length (b) 1.246 m
1
5 Tail semi-angle (θ) 25 deg
1  xa 
2 n
Bow  r x   d1     6 Myring body parameter 2 –
2   a  
1  3d tan  Case study
Stern  r x   d   2   x  a  b  
2

2  2c c 
Experimental setup
 d tan 
  3  2  x  a  b 
3
(1) In this study, six AUVs were used to simu-
c c 
late hydrodynamic tests. The model with a 7.5

Nose Mid-Body Tail L/D = 7.5

Figure 2. The model tested in the towing tank of Isfahan University of Technology

126 Scientific Journals of the Maritime University of Szczecin 62 (134)

 Experimental and numerical determination of the hydrodynamic coefficients of an autonomous underwater vehicle

length-to-diameter ratio was designed and manu- Table 2. Details of the six tested models
factured in the towing tank of Isfahan University LOA Moment centre LCB Ratio MC Ratio LCB
of Technology (Figures 2 and 3). Experiments were L/D
mm (nose) mm mm to LOA to LOA
also conducted in this towing tank (108×3×2.2 m), 7.5 1521 634 714 0.418 0.471
whose basin was equipped with a trolley that could 8.5 1724 736 815 0.427 0.473
provide a carriage speed up to 6 m/s with ±0.02 m/s 9.5 1927 838 915 0.435 0.475
accuracy. For force measurements, a 3-DOF dyna- 10.5 2130 940 1017 0.441 0.477
mometer was installed with 100 N load cells that
11.5 2333 1041 1118 0.446 0.479
were calibrated using calibration weights with 1%
12.5 2536 1143 1220 0.451 0.481
uncertainty. The experimental plan included per-
forming straight-ahead resistance runs at various
forward speeds (U = 0.5, 0.75, 1, 1.5, 2, 2.5, 3, and Governing equations
4 m/s).
Maneuvering equation

A body coordinate system o-xyz (Figure 5) was

defined to calculate the manoeuvring equations, such
that oz is the vertical axis and positive downward;
ox is the longitudinal axis and positive toward the
nose of the vehicle; and oy is the transverse axis and
positive toward the starboard side of a vehicle. If it
is assumed that the body is moving in the horizontal
plane o-xy, the origin o coincides with the centre of
mass, and the coordinate system coincides with the
principal axes of inertia. The motion equations may
be given as follows (Eq. (2)–(4)) for an AUV in the
Figure 3. Experimental setup
body coordinate system o-xyz that is moving relative
to the inertial coordinate system O-XYZ.
Other models were designed by Oceanic Con-
sulting Corporation (OCC) in Canada in 2005 and
were tested in 90-meter long and 12-meter wide
NTC-IOT towing tanks (Williams et al., 2006;
Azarsina, Williams & Issac, 2008; Azarsina & Wil-
liams, 2010). The length-to-diameter ratio of this
model was 8.5, and different models with the same
diameter and different lengths were rebuilt and test-
ed for improvement. A schematic view of the mod-
els is shown in Figure 4. Table 2 explains the six
tested models.
Figure 5. Coordinate system (O-XYZ is earth fixed coordi-
nate and o-xyz is the body-fixed coordinate system)
Nose Mid-Body Tail L/D = 8.5

L/D = 9.5
m u  rv   X (2)
Nose Mid-Body Tail

Nose Mid-Body Tail L/D = 10.5 m v  ru   Y (3)

Nose Mid-Body Tail L/D = 11.5 I z r  N (4)

Nose Mid-Body Tail L/D = 12.5 The first equation is the surge motion equation,
and the second and third equations are sway and yaw
Figure 4. Schematic of the five configurations tested in OCC motion equations, respectively. X and Y are external
(Williams et al., 2006) forces along the x and y-axis, respectively, and N is
the external moment about the z-axis; m is the mass

Zeszyty Naukowe Akademii Morskiej w Szczecinie 62 (134) 127

 Abdollah Sakaki, Mohsen Sadeghian Kerdabadi

of the body; Iz is the moment of inertia of the body and u', p' are the fluctuation velocity and pressure,
about the z-axis; u is body velocity along the x-di- respectively. The RANS equations are given as fol-
rection; and v is the velocity of the body along the lows for incompressible flow:
y-direction; u and v are the acceleration of the body
along the x and y directions, respectively; and r and  ui 
0 (8)
r are the angular velocity and acceleration around xi
the z-axis of the body.
 ui  
The nonlinear external forces X and Y and moment
t

x j

ui u j   uiu j 
N may be written as follows (Eq. (5)–(7)) according
to the Abkowitz model based on the quasi-steady-   u 
p  u j
state assumption, which states that the forces at any   g i    i   (9)
xi x j   x j xi 
instant depend on the motion parameters that define 
the instantaneous motion of the vessel. The body is
moving in a self-propulsion point, and the control where ρ is the fluid density; gi is the x, y, and z com-
surface is in the neutral condition. ponents of gravitational acceleration; μ is the fluid
dynamic viscosity; and  uiu j is the Reynolds stress
X  X u u  X u u  X uuu 2  X uuuu 3  X vv v 2  tensor components. The Reynolds stress tensor com-
 X rr r 2  X vr vr  X vvu v 2u  X rru r 2u  X uvruvr (5) ponents are estimated by turbulence models, which
are approximations of the physical phenomena
of turbulence. ρ is the density of the fluid, and the
Y  Yr r  Yv v  Yr r  Yv v  Yrrr r 3  Yvvv v 3  Yvrr vr 2 
Reynolds stress tensor is defined as:
 Yvvr v 2 r  Yvu vu  Yru ru  Yvuu vu 2  Yruuru 2
 u12 u1 u2 u1 u3 
(6)
 
N  N r r  N v v  N r r  N v v  N rrr r 3 
 ui u j   u2 u1 u22 u2 u3  (10)
 u  u  u  u  u 2 
 N vvv v 3  N vrr vr 2  N vvr v 2 r  N vu vu   3 1 3 2 3 

 N ru ru  N vuuvu 2  N ruuru 2 (7) Turbulence Model:

For the turbulence model, the k-ε model and
The coefficients on the right-hand side of the Reynolds stress transformation model were used.
equations are the hydrodynamic derivatives or The k-ε model is one of the most common turbulence
manoeuvring coefficients. Among the several avail- models used for aerodynamic and hydrodynamic
able methods for determining hydrodynamic coef- problems. After comparing the available models,
ficients, the PMM test is a single system that can this model was selected for the final simulations and
explicitly provide all damping and added mass coef- was defined with respect to two functions for param-
ficients required in the equations of motion. In this eters k and ε as:
paper, the OTT is simulated using CFD to obtain the
k    
linear parameters such as Yv and Nv and the nonlinear
t
 
  U k      t
k
 k   Pk   (11)
parameters such as Xvv, Yvvv, and Nvvv.   

Fluid flow modelling ρε

t
  ρU ε  
The unsteady viscous flow around a marine vehi-  μ   ε
cle is governed by the Navier–Stokes equations,    μ  t  ε   Cε1 Pk  Cε 2 ρε  (12)
 σ k   k
which can be applied to both laminar and turbulent
flow, but a very fine meshing is necessary to cap- and then:
ture all the turbulence effects in the turbulent flow
regime. RANS equations can also be used to model  2

Pk  t U U  U T  U 3t U  k
3
 
the turbulent flow. They are obtained based on statis-
tical tools known as Reynolds decomposition, where (13)
the flow parameters are decomposed into time-av-
eraged and fluctuation components, i.e. u  u  u, k2
t  C  (14)
p  p  p where u and p are the time-averaged, 

128 Scientific Journals of the Maritime University of Szczecin 62 (134)

 Experimental and numerical determination of the hydrodynamic coefficients of an autonomous underwater vehicle

Boundary conditions 1. Flow input (right boundary): Flow enters with

a steady, specific speed.
The appropriate boundary conditions on the flu- 2. Flow output (left boundary): Flow exits with
id domain boundaries and an AUV’s hull must be a steady pressure distribution.
used to create a well-posed system of equations. The 3. Symmetry (other boundaries): These boundaries
domain boundaries were split into patches as shown are defined by a symmetry condition to prevent
in Figure 6. The boundary conditions were chosen wall effects and fulfil the limited water condi-
to avoid backflow and lateral wall effects. There are tion.
two boundary conditions for the body surface: the 4. Wall (surface of the body): The surface of the
kinematic condition of no flow through the surface, body is defined as an impenetrable wall.
and a no-slip condition on the tangential velocity.
These are applied on the instantaneous wetted sur- Mesh generation
face of the AUV. For other boundaries, the symme-
try plane condition is a Neumann condition which There are different structured and unstructured
means that pressure, tangential velocities, and turbu- meshing strategies for solving different problems.
lence quantities have a zero gradient normal to the Simulations are conducted on unstructured trimmed
surface; however, for the normal velocity compo- meshes, while a trimmer meshing strategy is pro-
nent, a Dirichlet condition is applied. ficient for generating a high-quality mesh with the
lowest grid skewness. The overall view of the mesh
in the computational domain and around the AUV is
displayed in Figures 8 and 9, respectively. The mesh
distribution in the simulation range near the initial
model is shown in Figure 10.

Figure 6. Illustration of the computational domain

The finite volume method (FVM) is commonly

used to computationally solve RANS equations in
the computational domain by discretizing into finite
control volumes in which the discretized RANS
equations are solved. The domain dimensions are Figure 8. Computational domain
selected to be sufficiently large to prevent back-
flow at high drift angles. The distance from the inlet
boundary from the AUV nose tip is considered 1L,
the distance of outlet boundary from the AUV tail is
considered 2L, and the side boundaries are located at
0.5L (Figures 6 and 7).

0.5L
L 2L

0.5L

Figure 7. Schematic of the computational domain

To solve the Navier–Stokes time-averaging equa-

tions, the boundary conditions of walls and the hull
wall were fulfilled, as shown in Figure 4, as: Figure 9. Mesh around the hull

Zeszyty Naukowe Akademii Morskiej w Szczecinie 62 (134) 129

 Abdollah Sakaki, Mohsen Sadeghian Kerdabadi

Table 3. Axial force for different grids

Number of grids Type of mesh Fx

1,456,325 Coarse –8.236
2,529,846 Medium –7.735
Figure 10. Mesh distribution in the simulation range near
4,077,078 Fine –7.542
the initial model

Prism layer refinement was applied to the hull to The convergence ratio is defined as follows:
improve the accuracy of the solution in the boundary  21
layer region. The turbulent flow inside the boundary R (15)
 32
layer was approximated by wall functions. A high
y+ wall treatment based on equilibrium turbulent where ε21 = s2 – s1 is the difference between the solu-
boundary layer theory was used as the wall function. tions of fine and medium grids, and ε32 = s3 – s2 is the
The mean value of y+ on the hull surface was around difference between the solutions of the medium and
30, indicating good refinement of the prism layer. coarse grids.
The distribution of y+ for the fine mesh on the hull is The possible convergence situations are:
shown in Figure 11. R > 1 ⇒ Grid divergence,
R < 0 ⇒ Oscillatory convergence,
0 < R < 1 ⇒ Monotonic grid convergence.
If grid convergence occurs, Richardson extrapo-
lation (also called h2 extrapolation) is used to esti-
mate the convergence rate. The fractional differ-
ence between solutions is defined as eij = (sj −si)/si;
hence, the order of the discretization is estimated as
follows:
loge32 / e21
p (16)
Figure 11. Distribution of y+ around the hull
log(r )
After that, the grid convergence index (GCI) is
Grid convergence defined as:
eij
Mesh sensitivity examination is the most straight- GCI ij  FS p (17)
forward and consistent technique for determining r 1
the order of discretization errors in numerical simu- In this equation, Fs is a safety factor that Roache
lations. In other words, numerical results can be con- (Roache, 1997) recommended for convergence
sidered precise and valid if their solution is indepen- studied with a minimum of three grids (Fs = 1.25).
dent of the grid. A mesh sensitivity study involves GCI indicates the difference between the calculated
implementing solutions on the CFD model, with and exact value and is also a measure of solution
sequentially refines grids with reduced mesh sizes changes upon additional grid refinement. A small
until the variables become independent of the mesh GCI value indicates that the solution is in the exact
size. Three different mesh sizes with a constant grid value range.
refinement factor (r = h2/h1 = h3/h2 = 1.65) were been The computed convergence ratio, order of discret-
chosen, in which hi is a characteristic dimension of ization, and GCI are shown in Table 4. The theoreti-
the model, for example, the AUV length that is used cal value for convergence is p = 2, and the difference
to measure the mesh discretization. To prevent errors is due to the orthogonal grid, problem nonlinearities,
arising from extrapolation, based on experience, it is turbulence modelling, etc.
recommended that r > 1.3. The corresponding solu-
Table 4. Estimated convergence ratio, the order of discreti-
tions for these cases are designated s1 through s3. zation, and GCI
The mesh study for simulations was examined
Fx
for the pure model with zero drift angle at U = 2 m/s,
and the corresponding forces and moment of each R 0.206
mesh were calculated. Mesh numbers, forces, and p 2.930
moments are shown in Table 3. GCIfine 0.009

130 Scientific Journals of the Maritime University of Szczecin 62 (134)

 Experimental and numerical determination of the hydrodynamic coefficients of an autonomous underwater vehicle

0.1

0.01

0.001
Residual

0.0001

1e₋05

1e₋06

1e₋07

50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 900 950 1000 1050
Iteration
Continuity X-momentum Y-momentum Z-momentum Tke Sdr

Figure 12. Residual of continuity, momentum, and turbulence parameters

 d  
1.5 3
Computational fluid dynamic simulations d 
C t  C f 1  1.5    7    (19)
 l  l  
The fluid flow around the model was simulat-
ed with and without a drift angle with respect to where Cf is the frictional resistance coefficient that
the fluid flow direction. For the case without a drift is calculated according to the ITTC 1957 friction
angle, the resistance can be obtained using resis- formula:
tance simulations. In the case with a drift angle 0.075
Cf  (20)
(the static drift angle), the damping coefficients log Re  22
dependent on the lateral velocity can be obtained.
All computations were performed with a SIM- where Re is the Reynolds number. The results
PLE algorithm for pressure-velocity coupling. The obtained for bare hull resistance are shown in Fig-
second-order upwind scheme was applied for the ure 13. A comparison of the experimental, numeri-
advection term in the momentum equation. The cal, and empirical results is presented in Figure 13,
most common method for checking the conver- which shows that the CFD solution provided a good
gence of the simulation results is to investigate the prediction of the experimental and empirical results
residual of each solved variable. In Figure 12, an for different velocities.
illustrated of residual is shown, which indicated
good convergence. 12
CFD
EXP
10 EMP
Resistance simulation
Resistance (N)

The resistance test was simulated for a bare hull 6

at U = 0.5–4 m/s with an increment of 0.5 m/s to
investigate the axial force on an AUV. As noted, the 4

straight-ahead resistance experiment was performed 2

at U = 0.5, 0.75, 1, 1.5, 2, 2.5, 3, and 4 m/s, and the
resistance of the bare hull was estimated based on 0
0 0.5 1 1.5 2 2.5
the empirical equation. In this method, the resistance Velocity (m/s)
was predicted using: Figure 13. Comparison of computed, experimental, and
empirical resistance vs. velocity
1
RBarehull  AV 2C t (18)
2 Results
where A is the wetted surface area, V is the velocity,
ρ is the water density, and Ct is the total drag coeffi- Simulations were conducted on six symmetric
cient of the hull that is calculated as follows: bare hulls. The initial bare hull is shown in Figures

Zeszyty Naukowe Akademii Morskiej w Szczecinie 62 (134) 131

 Abdollah Sakaki, Mohsen Sadeghian Kerdabadi

1 and 4, with a length-to-diameter ratio of 7.5. angles of attack were simulated for four models
Other models with the same diameter and 8.5, 9.5, (L/D = 8.5, 9.5, 10.5, and 12.5) in 2-degree intervals
10.5, 11.5, and 12.5 length-to-diameter ratios were with a speed of 2 m/s. The results were compared
designed for conditions in which more space for with the experimental results in terms of longitudi-
equipment or increased battery capacity was needed; nal force, lateral force, and yawing moment (Fig-
therefore, a set of simulations was conducted using ures 15, 17, and 19) (Azarsina & Williams, 2010).
different speeds for these models. At a permanent angle of attack, upon increasing
the length-to-diameter ratio, the longitudinal force
Resistance tests increased because of an increase in the AUV hull
surface, which increased the total resistance. The
Resistance simulations were conducted for longitudinal force graph was symmetric relative to
6 hulls at speeds of 1, 2, 3, and 4 m/s. All simulations the zero-degree angle in terms of drift angle due to
were conducted in the absence of an angle of attack. the symmetry of the hull relative to the longitudinal
An example of the speed distribution around the hull axis of the vehicle. In other words, the longitudinal
model is shown in Figure 14. force of the vehicle relative to the drift angle was
The longitudinal force in each simulation was an even function, and the drift angle was equivalent
calculated, and the calculation results for each mod- to the dimensionless lateral speed. The lateral force
el are shown in Table 5 and were also compared graph in terms of the drift angle was an odd func-
with the experimental results (Williams et al., 2006). tion, i.e., the rotation direction of the vehicle affect-
Due to the hull symmetry in the tests with zero ed the output force. At this stage, it was repeatedly
angle of attack, the lateral force should also be zero. observed that the lateral force increased at a per-
To ensure the accuracy of results, the force value in manent drift angle and increased length-to-diame-
the lateral direction was also calculated. A value of ter ratio. The yawing moment coefficient graph as
approximately 10–4 N was obtained, which indicated a function of the angle of attack also produced an
the accuracy of results. odd function (Figure 19).
The velocity distribution around the hull at a drift
Yaw static test angle of 0° and a velocity of 2 m/s is shown in Figure
16, while Figure 18 shows the velocity distribution
To calculate the hydrodynamic coefficients, around the hull at a drift angle of 20° and a velocity
vehicle movements at –20 to +20 degrees of of 2 m/s.

Figure 14. Velocity distribution around hull at 1 m/s velocity

Table 5. Resistance results with experimental and CFD

Model resistance Model resistance Model resistance
LDR = 9.5 N LDR = 8.5 N LDR = 7.5 N V
m/s
Error % CFD EXP Error % CFD EXP Error % CFD EXP
5.43 10.86 10.3 4.18 9.81 9.4 4.36 7.89 7.56 2
Model resistance Model resistance Model resistance
LDR = 12.5 N LDR = 11.5 N LDR = 10.5 N
5.17 12.3 11.6 3.84 11.63 11.2 3.49 10.97 10.6 2

132 Scientific Journals of the Maritime University of Szczecin 62 (134)

 Experimental and numerical determination of the hydrodynamic coefficients of an autonomous underwater vehicle

16.5

15.5

EXP(LDR=12.5)
14.5
EXP(LDR=10.5)
Axial Force (N)

13.5 EXP(LDR=9.5)
EXP=LDR9.5)
EXP(LDR=8.5)
12.5
CFD(LDR=12.5)
11.5 CFD(LDR=9.5)

10.5
CFD(LDR=10.5)
CFD(LDR=8.5)
9.5

8.5
-22 -20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 22
Drift Angle (deg)

Figure 15. Axial force versus drift angle (V = 2 m/s)

Figure 16. Velocity distribution around the hull (drift angle = 0°, V = 2 m/s)

2.5

1.5
EXP(LDR=8.5)
Lateral force Coefficient

1
EXP(LDR=9.5)
0.5 EXP(LDR=10.5)
EXP(LDR=12.5)
0
-25 -20 -15 -10 -5 0 5 10 15 20 25 CFD(LDR=8.5)
-0.5
CFD(LDR=9.5)
-1 CFD(LDR=10.5)
CFD(LDR=12.5)
-1.5

-2

-2.5
Drift Angle(deg)

Figure 17. Lateral force coefficient versus drift angle (V = 2 m/s)

Zeszyty Naukowe Akademii Morskiej w Szczecinie 62 (134) 133

 Abdollah Sakaki, Mohsen Sadeghian Kerdabadi

Figure 18. Velocity distribution around the hull (drift angle = 20°, V = 2m/s)

0.3

0.2

EXP(LDR=8.5)
Yaw Moment Coefficient

0.1 EXP(LDR=9.5)
EXP(LDR=10.5)
EXP(LDR=12.5)
0
-25 -20 -15 -10 -5 0 5 10 15 20 25 CFD(LDR=8.5)
CFD(LDR=9.5)
-0.1
CFD(LDR=10.5)
CFD(LDR=12.5)
-0.2

-0.3
Yaw Angle(deg)

Figure 19. Yaw moment coefficient versus drift angle (V = 2 m/s)

Hydrodynamic coefficient calculations   v3

N   Nv  v  Nvvv (23)

With respect to the axial force results, the lateral where v' = sin α. Unlike many methods, this does
force coefficient, lateral moment coefficients at drift not have the problem of an acute angle of attack
angles from –20° to +20°, the dimensionless lateral approximation; therefore, it is acceptable for v' to
force, or the lateral force can be calculated using the be a greater value. Table 6 shows the hydrodynam-
drag and lift coefficients: ic coefficients of AUVs with different L/D. Using
curve processing for the data, Y'v, Y'vv, N'v and N'vv
Y   CD sin   CL cos (21) coefficients can be calculated (Table 6).

where CL and CD are the lift and drag coefficients, Table 6. Hydrodynamic coefficients
respectively, and α is the angle of attack or drift L/D Y'v Y'vvv N'v N'vvv
angle. The dimensionless moment is equal to the yaw
8.5 0.0127 0.000006 0.044 0.00003
moment. To calculate the hydrodynamic coefficients
for the calculated force and moment, a third-order 9.5 0.0142 0.00006 0.0539 0.00002
model should be defined as follows: 10.5 0.0157 0.00007 0.0563 0.00004
12.5 0.0185 0.000018 0.0679 0.0005
  v3
Y   Yv  v  Yvvv (22)

134 Scientific Journals of the Maritime University of Szczecin 62 (134)

 Experimental and numerical determination of the hydrodynamic coefficients of an autonomous underwater vehicle

Conclusions 5. Fan, S.-B., Lian, L., Ren, P. & Huang, G.-L. (2012) Oblique
towing test and maneuver simulation at low speed and large
drift angle for deep sea open-framed remotely operated ve-
Maneuverability is an important hydrodynamic
hicle. Journal of Hydrodynamics, Ser. B 24 (2), pp. 280–286.
quality of a marine vehicle and should be predicted 6. Ferziger, J.H., Peric, M. & Leonard, A. (1997) Computa-
during the various design stages. There are various tional methods for fluid dynamics. AIP.
models to predict the manoeuvring properties of 7. Gao, T., Wang, Y., Pang, Y. & Cao, J. (2016) Hull shape
a marine vehicle, the most popular of which is the optimization for autonomous underwater vehicles using
CFD. Engineering Applications of Computational Fluid
Abkowitz model, in which the external forces and Mechanics 10 (1), pp. 599–607.
moments are defined using hydrodynamic deriv- 8. Gentaz, L., Guillerm, P.E., Alessandrini, B. & Delhom-
atives or coefficients. These hydrodynamic coef- meau, G. (1999) Three-dimensional free surface viscous
ficients should be found in advance to predict the flow around a ship in forced motion. 7th International Con-
ference on Numerical Ship Hydrodynamics.
manoeuvring properties of a marine vehicle.
9. Kim, Y.-G., Kim, S.-Y., Kim, H.-T., Lee, S.-W. & Yu, B.-S.
The computations were validated with exper- (2007) Prediction of the maneuverability of a large container
imental tests performed in the towing tank of the ship with twin propellers and twin rudders. Journal of Ma-
Isfahan University of Technology. The comparisons rine Science and Technology 12 (3), pp. 130–138.
of the measured resistances and the computational 10. Li, G. & Duan, W.-Y. (2011) Experimental study on the hy-
drodynamic property of a complex submersible. Journal of
results indicated that the CFD computations were Ship Mechanics 15 (1–2), pp. 58–65.
reliable. OTT simulations were performed over 11. Myring, D.F. (1976) A theoretical study of body drag in
a wide range of drift angles to compute the trans- sub-critical axisymmetric flow. Aeronautical Quarterly 27
verse velocity-dependent coefficients. All linear and (3), pp. 186–194.
12. Nazir, Z., Su, Y.-M. & Wang, Z.-L. (2010) A CFD based
nonlinear coefficients were obtained using the time
investigation of the unsteady hydrodynamic coefficients of
mean of hydrodynamic forces, and the moments 3-D fins in viscous flow. Journal of Marine Science and Ap-
were calculated using CFD simulations. The coef- plication 9 (3), pp. 250–255.
ficients were obtained using suitable curve fittings. 13. Obreja, D., Nabergoj, R., Crudu, L. & Păcuraru-Popoiu,
In the first section of this study, a new method S. (2010) Identification of hydrodynamic coefficients for
manoeuvring simulation model of a fishing vessel. Ocean
was proposed for hydrodynamic simulations using Engineering 37 (8), pp. 678–687.
CFD with the aid of CD-adapco software. Hydro- 14. Pan, Y.-C., Zhang, H.-X. & Zhou, Q.-D. (2012) Numerical
dynamic forces and coefficients were calculated for prediction of submarine hydrodynamic coefficients using
different AUV models at different speeds. In the ini- CFD simulation. Journal of Hydrodynamics, Ser. B 24 (6),
pp. 840–847.
tial design stages, this method can be used to predict
15. Ray, A., Singh, S. & Seshadri, V. (2009) Evaluation of lin-
the manoeuvrability of an AUV and to design con- ear and nonlinear hydrodynamic coefficients of underwater
trol systems for these AUVs. Another characteristic vehicles using CFD. ASME 2009 28th International Confer-
of this method was its ability to calculate non-lin- ence on Ocean, Offshore and Arctic Engineering, American
ear hydrodynamic coefficients, which are essential Society of Mechanical Engineers.
16. Roache, P.J. (1997) Quantification of uncertainty in compu-
for the application of non-linear manoeuvre models. tational fluid dynamics. Annual Review of Fluid Mechanics
Finally, the non-linear damping coefficients were 29 (1), pp. 123–160.
calculated in the horizontal plane. 17. Sarkar, T., Sayer, P. & Fraser, S. (1997) A study of auton-
omous underwater vehicle hull forms using computational
fluid dynamics. International Journal for Numerical Meth-
References ods in Fluids 25 (11), pp. 1301–1313.

1. Azarsina, F. & Williams, C.D. (2010) Manoeuvring sim- 18. Tyagi, A. & Sen, D. (2006). Calculation of transverse hy-
ulation of the MUN Explorer AUV based on the empirical drodynamic coefficients using computational fluid dynamic
hydrodynamics of axi-symmetric bare hulls. Applied Ocean approach. Ocean Engineering 33 (5), pp. 798–809.
Research 32 (4), pp. 443–453. 19. Williams, C., Curtis, T., Doucet, J., Issac, M. & Azarsi-
2. Azarsina, F., Williams, C. & Issac, M. (2008) Modelling na, F. (2006) Effects of hull length on the hydrodynamic
the hydrodynamic sway force exerted on the bare-hull of loads on a slender underwater vehicle during manoeuvres.
an axi-symmetric underwater vehicle in lateral acceleration OCEANS 2006, IEEE.
manoeuvres. OCEANS 2008, IEEE. 20. Wilson, R., Paterson, E. & Stern, F. (1998) Unsteady
3. Blidberg, D.R. (2001) The development of autonomous RANS CFD method for naval combatant in waves. In Pro-
underwater vehicles (AUV); a brief summary. ICRA, IEEE. ceedings of the 22nd Symposium on Naval Hydrodynamics,
4. Dantas, J.L.D. & de Barros, E.A. (2013) Numerical anal- Washington, D.C., pp. 532–549.
ysis of control surface effects on AUV manoeuvrability. Ap-
plied Ocean Research 42, pp. 168–181.

Zeszyty Naukowe Akademii Morskiej w Szczecinie 62 (134) 135