Professional Documents
Culture Documents
A Study On The Design Optimization of An AUV by Using Computational Fluid Dynamic Analysis
A Study On The Design Optimization of An AUV by Using Computational Fluid Dynamic Analysis
A Study on the Design Optimization of an AUV by Using Computational Fluid Dynamic Analysis
Taehwan Joung*, Karl Sammut*, Fangpo He*, and Seung-Keon Lee**
*School of Computer Science, Engineering and Mathematics, Faculty of Science & Engineering, Flinders University, Adelaide, Australia
**Department of Naval Architecture and Ocean Engineering, Pusan National University, Busan, Korea
Autonomous Underwater Vehicles (AUV's) provide an important In this paper, Computational Fluid Dynamics (CFD) tools are evaluated
means for collecting detailed scientific information from the ocean with the purpose of obtaining the hydrodynamic parameter (velocity,
depths. The hull resistance of an AUV is an important factor in pressure, etc.) estimates of an AUV with a ducted propeller. The design
determining the power requirements and range of the vehicle. This of an AUV is optimized using CFD analysis to minimise drag force.
paper describes a design method using Computational Fluid Dynamics
(CFD) for determining the hull resistance of an AUV under The methods reported in this paper for optimisation by CFD code are as
development. The CFD results reveal the distribution of hydrodynamic follows: (1) CFD results analysis and comparison with theoretical or
values (velocity, pressure, etc.) of the AUV with a ducted propeller. empirical equation for validation of reliability, (2) evaluation of an
The optimization of the AUV hull profile for reducing the total automatic element meshing method that generates a boundary layer
resistance is also discussed in this paper. This paper demonstrates that which allows for appendages such as fins and ducts, and produces a
shape optimization in conceptual design is possible by using a stable and robust analysis, and (3) searching and identifying optimum
commercial CFD package. The optimum designs to minimize the drag design variables to produce minimum resistance.
force of the AUV were carried out, for a given object function and
constraints.
INITIAL HULL DESIGN AND DRAG ESTIMATION OF
KEY WORDS: AUV; CFD (Computational Fluid Dynamics); THE AUV
Optimum design; Drag force; Drag Coefficient (CD)
Hull Design
INTRODUCTION At a conceptual design stage, the hull of an AUV can be divided into
distinct sections, namely the nose, middle section, tail, and propeller
An unmanned AUV (Autonomous Underwater Vehicle) is a versatile duct. The AUV hull has been designed based on the Myring hull profile
research tool for maritime archaeology, oceanographic and marine equations (Prestero, 2001) which is known to produce minimum drag
biology studies, defense applications, and oil and mineral exploration force for a given fineness ratio, that is, ratio of its length to its
and exploitation programs. Rapid progress in AUV development is maximum diameter (l/d). The curve shapes of the nose and tail sections
steadily increasing the reliability and endurance of such vehicles to are determined from equations (1) and (2), respectively.
operate in the harsh marine environment. Much work, however, still
1/ n
needs to be done in terms of optimizing the hull design to minimize
1 ⎡ ⎛ x−a⎞ ⎤
2
drag and increase propulsion efficiency. r ( x) = d ⎢1 − ⎜ ⎟ ⎥ (1)
2 ⎣⎢ ⎝ a ⎠ ⎦⎥
In previous studies, designers have employed empirical formulas or
used experimental derived data to estimate drag force of ships or
submerged bodies such as AUVs. However, the conventional empirical 1 ⎡ 3d tan(θ ) ⎤
r ( x) = d −⎢ 2 − ( x − ( a + b)) 2
formula is not able to accurately compute the drag of complex hull 2 ⎣ 2c c ⎥⎦ (2)
forms with appendages protruding from the hull. Although
experimental testing using a tow tank or cavitation tank can produce ⎡ d tan(θ ) ⎤
+⎢ 3 − 2 ⎥ ( x − ( a + b))3
very accurate predictions of drag, such testing requires considerable ⎣c c ⎦
time and effort, and expensive test facilities to obtain the vehicle’s
hydrodynamic characteristics. Consequently, a new drag estimation The designed shape of the AUV hull based on “Myring equation” and
method is needed for development of a specific AUV, which can be “NACA profile (NACA 6721)” is shown in Fig. 1. As the propeller
applied to a conceptual design. The new method should be efficient, blades rotate through the water, they generate high-pressure areas
696
behind each blade and low pressure areas in front of each blade. It is Estimation of the Drag Force
this pressure differential that provides the force to drive the vessel.
However, losses occur at the tip of each blade as water escapes from The axial drag force acting on a body moving at a constant velocity in a
the high pressure side of the blade to the low pressure side, resulting in fluid medium is approximated by the expression (5). The drag force of
reduced efficiency in terms of pushing the vessel forward. To obtain the an AUV, therefore, can be estimated by using the following formula
most thrust, a propeller must move as much water as possible in a given (Michael, 2003).
time. A nozzle will reduce these propeller losses, especially when a
high thrust is needed at a low vehicle speed. A Rice speed nozzle 1
profile has been employed for our AUV design concept, since its Rf = ρC f A f u u = X u u u u (5)
coefficient of drag is over 17 times less than that of a conventional Kort 2
nozzle. A section of a designed nozzle has been developed based on the
‘NACA profile (NACA 6721)’ as showed in Table 1. where, Cf is the coefficient of friction drag obtained from equation (3)
and (4), ρ is the density of the fluid, Af is a submerged area of a hull
and u is the fluid velocity or the advanced speed of the AUV. This
estimation for the drag force can provide useful information for the
powering requirements at the early stage of design. However, the
estimation of the drag obtained only by considering form factor (1+k)
could be uncertain, if there is complex shape equipment attached on the
AUV such as an antenna, DVL (Doppler Velocity Log), ADCP
(Acoustic Doppler Current Profilers), camera or other protruding
equipment.
Fig. 1 A conceptual design of the AUV Governing Equation of the CFD Analysis
The fluid flow around the AUV has been modelled using the
Table 1. Section of the Nozzle - NACA profile (NACA 6721)
commercial CFD code ANSYS CFX 11.0. For these calculations, the
governing equations are the Navier-Stokes equations and continuity
Thickness 21 % of wing chord (Length of chord = 1) equation under the assumption of incompressible fluid. The Navier-
Position 0.7 Position of max camber (0 - 1) Stokes equations considered in ANSYS CFX 11.0 is the isothermal
Reynolds Averaged Navier-Stokes (RANS) shown in equations (6) and
Camber 0.06 % of wing chord (Length of chord = 1) (7) (Wilcox, 1998; Seo et al, 2005)
697
program in ANSYS-WORKBENCH. and allows solutions to be quickly generated by control of the iteration
times. What is required is a tool that is not only easy to use, but also
Mesh Generation able to provide a satisfactory reliability analysis result with the desired
degree of convergence.
‘Tetrahedral’ and ‘Pyramid’ elements are normally employed for
generating nodes and elements in the fluid domain. These elements are
suitable for representation of a complex geometry such as a nozzle.
However, such elements are not suitable to resolve the boundary layer
adjacent to the solid body (Nishi, 2007). Therefore, ‘Prism’ elements
are selected for generating meshes around the body surface as these are
the most appropriate for a boundary layer mesh (ANSYS-CFX, 2007).
Fig. 2 shows the various meshed sections which are merged, and
embodied for the CFD analysis by the “ANSYS-CFX-MESH” mesh
generator. The size of the fluid domain around the tank is big enough so
as to not cause any error due to blocking effects if the walls of the tank
significantly restrict flow around the hull. The water tank size should be
decided after thorough preliminary analysis, and experimental or
theoretical prediction so as not to make the fluid domain too small thus
adversely affecting the CFD analysis, nor unduly large which would
unnecessarily increase the computation time.
In order to determine the velocity of the AUV, the inlet velocity from
the front of the water tank was set to be equivalent to the advanced
AUV velocity. The constraint of the outlet (opposite side of the inlet) is
that there is no relative pressure, and the ‘Free Surface Condition: no
slip condition’ was allocated to the remaining sides of the water tank.
Given that both the AUV geometry and the boundary conditions are
symmetrical about the centre plane, the symmetric condition was used
for modelling the AUV instead of the total model as shown Fig. 3.
Turbulence Model
The degree of the nozzle (Ψ) defined as the angle of attack of the
nozzle profile was considered as one design variable satisfying the
given constraints (0° ≤ Ψ ≤ 20°) in order to minimize the overall drag
resistance of the AUV body including the nozzle. The above
summarized specifications for the “pre-processing” stage including
mesh generation are showed in Table 2.
698
Table 2. Principal conditions employed in the numerical computation
Water tank size 7,000×3,000×3,000 mm3
132,308 ∼ 158,611
Total no. of elements (nodes)
(30,685 ∼ 45,303)
If the results are satisfied at a certain value, a user can stop the run and
move to “post-processing” to see the results. Fig. 4 shows the “solve-
process” being monitored while ANSYS-CFX solver is running. Note
that the value of the X-direction drag force of the AUV hull converged
after nearly 30 iterations. The user, therefore, can decide the number of (a) Default (Momentum etc.)
iteration for convergence and determine when to stop in order to save
CPU time.
Verification
In order to ensure the verification of the CFD analysis, the bare hull of
the AUV was employed, not considering nozzle part. The drag force
predictions from the CFD results and the ‘ITTC 1957 correlation line’
have a high degree of correspondence as shown in Fig. 5. The results
show that the form factor predicted by equation (2) is useful for the
estimation. There are drag differences between ‘ANSYS-CFX’ and
‘ITTC 1957 correlation line’, because ANSYS-CFX considers the total
drag while the ‘ITTC 1957 correlation line’ only considers the skin
friction drag. The pressure and skin friction distribution along the AUV
are shown in Fig. 6.
The CFD results can thus be validated by the ‘ITTC 1957 correlation
line’, and demonstrated to be reliable and useful for further research
such as optimizing the nozzle shape.
699
Fig. 5 Comparisons of drag predictions for the AUV (w/o duct)
Fig. 8 Pressure distribution around the AUV at 3 m/s
A design optimization was carried out to find the optimum value of the
nozzle angle, since it has a high sensitivity value with respect to the
total resistance and drag, as well as AUV manoeuvrability. The object
function for optimum design is the X-direction (+ is the advanced
direction) drag force, and the constraint is the nozzle angle (0°~20°).
For the purpose of finding the optimum value of the object function, the
three tools, that is, 3D-CAD program, mesh generation program and
ANSYS-DesignXplorer (DX), are linked together. ANSYS-
DesignXplorer is the optimization program, which sets up the relations
between input parameter from the 3D CAD program (DesignModeler
(DM), AUTOCAD etc.) and output parameters. The typical application
workflows of ANSYS-DesignXplorer comprise two different methods
as shown in Fig. 10. The ANSYS-DesignXplorer provides three
Fig. 7 Pressure contour around the AUV hull at 3 m/s optimization methods; Design of Experiments (DOE), Variational
700
Technology (VT) and Design for Six Sigma (DFSS) (ANSYS-CFX, (Ψ) by DOE has a high correspondence with the Direct Searching
2007). From the three methods, Design of Experiments (DOE) was Method, therefore, the results of the DOE method can be accepted to be
employed for the optimization due to its simplicity and reliability. That reliable.
is, DOE creates a matrix and builds an approximation model, and
optimization is performed by sampling against this approximation From the results of the CFD analysis, shown in Fig. 13(a), it can be
method. seen that if the nozzle angle is lower than the optimum angle (-11°),
then the value of the object function (drag force) becomes much higher,
even though mass flow of the water is a bit larger, when compared with
the drag force at the optimum angle. On the other hand, if the nozzle
angle is higher than optimum angle (-11°), not only does the value of
the object function (drag force) become much higher, but also vortices
can occur behind of the nozzle which reduce the nozzle efficiency, as
shown in Fig. 13(c).
The DOE method by CCD sampling was verified by using the Direct
Searching Method to ensure its reliability.
As shown in Fig. 11, the optimum value of the design variable (Ψ) was
obtained as -11.11° from the CFD analysis by DOE, and the values of
the object function (drag force) were 3.00 N (@ 1 m/s), 11.12 N(@ 2
m/s), and 23.98 N (@ 3 m/s). The result of the Direct Searching Fig. 12 X-directional drag force acting on the AUV (Direct Searching
Method, which used for verifying the DOE-CCD result, showed that Method)
the optimum value of the design variable (Ψ) was obtained as -11°, and
the values of object function were 2.98 N (@ 1 m/s), 11.05 N (@ 2
m/s), and 23.76 N (@ 3 m/s). The optimum value of the design variable
701
(3) Two possible optimum design methods, ‘Design of Experiment
(DOE)’ and ‘Direct Searching Method (What-if)’, have been
researched and employed for finding the optimum design value (nozzle
angle) with the minimum value of the object function (drag force of the
AUV).
Future Work
The velocity and pressure distribution around the AUV and drag
estimation, which are difficult to obtain from an experimental tow tank
(a) 0° of the Nozzle (Initial value)
test, were obtained by CFD analysis. Similarly, the optimum design
value of the nozzle angle was also obtained using CFD analysis. The
effect of the propulsion force have not however been considered in this
paper.
Relative rotation efficiency (ηR), thrust deduction fraction (t), and wake
factor (w) of the nozzle should be carried out for more reliable
optimization of the shape. A CFD analysis including a rotating
propeller is now being studied, and will be verified by an experimental
test in a tow tank.
ACKNOWLEDGEMENTS
(b) -11° of the Nozzle (Optimum value)
The authors would like thank to Professor Yoshiki Nishi of Yokohama
National University, Professor Wataru Koterayama, and Professor
Masahiko Nakamura of Kyushu University in Japan for their practical,
technical support, and productive international cooperation.
REFERENCES
702