You are on page 1of 11

173

2016,28(2):173-183
DOI: 10.1016/S1001-6058(16)60619-4

Hull form optimization of a cargo ship for reduced drag*

Fuxin HUANG, Chi YANG


Department of Physics and Astronomy, George Mason University, Fairfax, VA 22030, USA,
E-mail: fhuang@gmu.edu

(Received February 25, 2016, Revised April 19, 2016)

Abstract: Hydrodynamic optimization of the hull forms can be realized through the implementation and integration of computational
tools that consist of a hydrodynamic module, a hull surface representation and modification module, and an optimization module. In
the present paper, a new bulbous bow generation and modification technique has been developed and integrated into the hull surface
representation and modification module. A radial basis function based surrogate model is developed to approximate the objective
functions and reduce the computing cost. A multi-objective artificial bee colony optimization algorithm is implemented and integra-
ted into the optimization module. To illustrate the integrated hydrodynamic optimization tools, a cargo ship is optimized for reduced
drag. The optimal hull forms obtained are then validated computationally and experimentally. Validation results show that the prese-
nt tools can be used efficiently and effectively in the simulation based design of the hull forms for reduced drag.

Key words: simulation based design, ship hull form optimization, drag reduction, radial basis function, surrogate model,
Neumann-Michell theory

Introduction based on Euler/Reynolds Averaged Navier-Stokes


Hydrodynamic optimization is an important aspe- (RANS) equations. A low-fidelity CFD solver that
ct of ship design. Traditionally, hydrodynamic design accounts for the dominate physics can give a reasona-
optimization of ship hull forms is achieved using bly accurate flow prediction with much less compu-
empirical method based on the accumulated hull form ting time and it is thus good for the early stage design
database. With the rapid development of computer that might require a very large number of flow evalua-
technology, modeling methods and numerical techni- tions. A high-fidelity CFD solver that accounts for
ques, computational fluid dynamics (CFD) based ship more physics can give more accurate flow prediction
hull form design optimization has become more widely with much more computing time and it is thus good
used in ship design. for the validation of the final design. Both low-fidelity
Applications of CFD based hull form optimiza- and high-fidelity CFD tools have been integrated into
tion for reduced calm water resistance have been repo- the hydrodynamics module of the present simulation
rted in a significant number of studies. These studies based design tool.
attest to a rapidly growing interest in hydrodynamic In addition to the flow solver, the CFD based ship
optimization[1-23]. The CFD solvers used in these stu- hull form design optimization requires two other com-
dies can be classified into two categories: the low- ponents: ship hull form modification module and opti-
fidelity solver that is usually based on the potential mization module. In the hull form optimization proce-
flow theory with various approximations on the boun- ss, the low-fidelity flow solver is used to evaluate the
dary conditions, and the high-fidelity solver that is hydrodynamic performances of the ship hull forms that
are generated by the ship hull form modification tech-
niques according to the design variables updated by
* Biography: Fuxin HUANG, Male, Ph. D.,
the optimization algorithms.
Research Assistant Professor
Corresponding author: Chi YANG, E-mail: cyang@gmu.edu In the field of ship geometry modeling for hull
form optimization, a number of techniques have been
developed. These techniques can be divided into two
categories: conventional modeling techniques and pa-
174

rametric modeling techniques[9]. Conventional mode- required for the CFD based hull form optimization, a
ling techniques build on a low level definition of the radial basis function (RBF) based surrogate model is
geometry, such as points. The points can be used to developed in this study to approximate the objective
define curves and curves can be used to define surfa- functions. Specifically, the simple (low-fidelity) CFD
ces. Therefore, these techniques offer a great flexibili- tool is used to evaluate the objective functions at
ty to the modification of the geometry and topology. sample points that are used to construct the RBF based
However, they require many parameters in order to surrogate model, and the MOABC algorithm is used
conduct a fair geometry modification. These parame- to search for the optimal solutions using the RBF
ters will serve as design variables in the optimization based surrogate model.
process. In order to modify the ship hull form in the The enhancement and integration of the compu-
optimization using conventional modeling techniques, tational tools discussed above have enabled the simu-
additional techniques, such as radial basis function lation-based design of hull forms in terms of hydrody-
(RBF) interpolation method[15] and modification fun- namic performance. The improved integrated compu-
ction method[8], can be implemented to reduce the tational tools are used for the design optimization of a
required parameters, i.e., design variables. Parametric cargo ship for reduced drag in the present study. The
modeling techniques, on the other hand, build on high- optimal hull forms obtained are then validated compu-
level entities. These entities are called form paramete- tationally and experimentally.
rs in the geometric modeling. The most prominent
advantage of parametric techniques is that small to
intermediate modifications can be produced very effi- 1. Problem statement
ciently. The parametric modeling of the hull form re- The aim for this study is to further develop a
quires few design variables. In the present study, a simulation based design tool and apply it to optimize
RBF interpolation method is combined with a parame- the hull form for reduced drag. In order to reduce the
tric modeling technique to modify the hull form loca- drag in the entire speed range, the present hull form
lly and globally in the optimization process. optimization problem can be formulated as follows:
Various optimization algorithms have been used
to minimize objective functions in the CFD based hull min{ fi }( X ) subject to X ∈ S ⊆ R d (1a)
form optimization. These algorithms can be divided
into two categories depending on whether the derivati- RTi − RTi , 0
ve of the objective function is required: the derivative fi = , i = 1, 2, 3 (1b)
based algorithms and the derivative free algorithms. RTi , 0
Popular optimization algorithms, such as steepest de-
scent, Newton, and conjugate gradient, are derivative where fi defines objective function, RTi and RTi , 0
based algorithms. With the help of gradient informa- denote the resistance obtained for the new hull form
tion, derivative based methods usually obtain the opti- and the initial hull form, respectively, at three given
mal solution efficiently. There are a number of appli- design speeds that cover the low, middle and high
cations on ship hull form optimization using derivati- speed, X represents the vector of design variables in
ve based optimization algorithms[6,23]. However, the d dimension defined in terms of the parameters asso-
optimal solution obtained by such algorithms is a local ciated with the hull form modification, and S is the
optimum, but not necessarily the global optimum. On feasible solution set resulting from removing the por-
the other hand, derivative free optimization algorithms tion of the space R d prohibited by the constraints. In
include the direct search methods, the evolutionary the present study, the constraint for the optimization is
based algorithms (such as genetic algorithm and diffe- taken as the displacement, i.e., the reduction of the
rential evolution algorithm), and the swarm intellige- displacement is less than one percent.
nce based algorithms (such as particle swarm optimi-
zation algorithms, ant colony optimization algorithms, Table 1 Main particulars of the Series 60 hull (CB = 0.6)
and bee colony optimization algorithms). Many of
these derivative free optimization algorithms have im- LWL Beam Draft Displacement
proved the capacity of finding the global optimum. 3.101 m 0.406 m 0.163 m 0.121 m3
Derivative free algorithms have also been widely used
in the ship hull form optimizations[12-21] due to their
merits. In this study, a single objective artificial bee For the purpose of illustration, the Series 60 hull
colony algorithm is extended to a multi-objective arti- model is taken as an initial hull and optimized for
ficial bee colony (MOABC) algorithm to ensure the reduced drag at three design speeds defined by Fr =
finding of the global optimums that meet several de- 0.22, 0.27 and 0.32 using the present tools. The main
sign needs at the same time. particulars of the Series 60 hull (CB = 0.6) model are
In order to further reduce the computing time listed in Table 1 and the 3-D view of the ship hull
175

model is shown in Fig.1. obtained in this step. Given a set of requirements and
certain criteria, an optimal solution that defines an
optimal hull form can be chosen from the Pareto-front
in the fourth step. Considering the fact that the opti-
Fig.1 3-D view of the original Series 60 hull model mal hull form is obtained using a surrogate model
based on a simple CFD tool, it is necessary to validate
the optimal hull form in this step by a RANS/Navier-
2. Methodology Stokes based advanced CFD tool or a model test.
The procedure for the hull form optimization The computational tools used in each step will be
using the present simulation based design tools is de- discussed in details in the following subsections.
scribed in Fig.2.

Fig.3 The definition sketch of coordinate system

2.1 Drag evaluation


Consider the steady flow about a ship advancing
in calm water. Non-dimensional coordinates x ≡ ( x,
y, z ) ≡ X / L are defined in terms of a reference
length L , typically taken as the ship length. The z
axis is vertical and points upward, and the mean free
surface is taken as the plane z = 0 . The x axis is
chosen along the path of the ship and points toward
the ship bow. The definition sketch of the coordinate
Fig.2 Flowchart of the hull form optimization procedure system is shown in Fig.3. The Froude number Fr
and the Reynolds number Re are defined as
It can be observed from Fig.2 that the present
simulation based hull form design optimization proce- U
Fr = (2)
dure consists of four steps. After an initial hull form is gL
selected and the optimization objective functions and
constraints are defined, the first step is to select appro-
UL
priate ship hull surface modification techniques and Re = (3)
determine the design variables for the optimization ν
problem. In order to optimize hull form for reduced
drag, a steady ship flow problem needs to be solved where U is ship speed, g is gravitational accelera-
for all hull forms generated in the optimization proce- tion, and ν is kinematic viscosity of the water.
ss at given design speeds using a CFD tool so that The total drag of a ship hull is approximated by
objective functions can be evaluated. A simple CFD the summation of the wave drag and the friction drag.
tool based on linear potential flow theory can usually The wave drag is calculated by the Neumann-Michell
be used in the optimization process if the simple CFD (NM) based potential flow theory. A detailed descri-
tool is accurate enough in predicting the variations of ption of the formulation and the numerical solution
the flow solutions due to the change of the hull forms. procedure for the NM theory can be found in authorsʼ
In order to further reduce the computing cost associa- previous work[24-28]. The friction drag is determined
ted with the flow evaluations using a CFD tool, a by an empirical formula. The total drag coefficient is
surrogate model is developed to approximate the obje- defined as
ctive functions. Thus, the second step is to construct
the surrogate model in terms of the values of the obje- RT
ctive functions obtained using a simple CFD tool. CT = = CW + CF (4)
0.5ρU 2 S wet
Once the surrogate model construction is completed, it
can then be used by a multi-objective optimizer during
the search of the optimal hull forms in the third step. where ρ is the water density, U is the ship speed,
A set of optimal solutions called Pareto-front can be S wet is the wetted surface area, RT is the total drag
176

and CW is the wave drag coefficient evaluated by non- generated first from the initial hull using the geometry
modification tool developed in the present study. The
dimensionalizing the wave drag RW using a conven-
size of the bulbous bow is then optimized in terms of
tional way as the hydrodynamic performance, i.e., the minimization
RW of the total drag, in the present study. The top figure in
CW = (5) Fig.4 depicts a new initial hull with a bulbous bow
0.5ρU 2 S wet generated from the Series 60 hull model, and the
bottom figure in Fig.4 depicts the new initial hull and
and CF is the frictional drag coefficient evaluated a definition sketch for design variables (black dot) used
using ITTC 1957 Model-Ship Correlation Line as in the bulbous bow optimization. Specifically, the
follows shape of the bulbous bow is modified through three
design variables, i.e., the movable control point (black
0.075 dot) can be repositioned along lateral, longitudinal, and
CF = (6)
[lg( Re) − 2]2 vertical directions.

The simple CFD tool (computer code SSF) is


developed based on the NM theory and the ITTC 1957
Model-Ship Correlation Line. This simple CFD tool
has been used to evaluate the wave drag and the total
drag for various monohull forms and catamarans. Nu-
merical results predicted by this simple CFD tool are
in fairly good agreement with experimental measure- Fig.4 Fixed (a) and movable (b) control points (design varia-
ments[24-28,30,31]. Therefore, the simple CFD tool, SSF, bles) of the radial basis function interpolation method for
the modification of the ship hull with the initial bulbous
is employed to evaluate the total drag of the ship hull bow
form in constructing the surrogate model.
The shifting method is adopted to modify the
2.2 Design variable selection and ship hull surface entire ship hull body in the second step. In the shifting
modification method method, the longitudinal position of sections is shifted
In the present study, the Series 60 hull model is to modify the prismatic coefficient, the longitudinal
considered for the hull form optimization for reduced center of buoyancy, and the parallel mid-body of the
drag. Two hull modification steps are employed to initial hull. In the CFD-based hull form optimization,
generate new hull forms in the optimization process. the sectional area curve of the initial hull form is
The first one is to generate and modify a bulbous bow modified during the optimization process. The new
using radial basis function interpolation method, the hull form is obtained by moving the stations of the
second one is to modify the entire hull body using a initial hull form along the longitudinal direction. The
shifting method based on the modifications of the amount of the movement is determined by comparing
sectional area curve of the hull. the modified sectional area curve and the original one.
A radial basis function (RBF) based interpolation When the hull surface is defined by a discrete triangu-
method can be used to deform the shape of the geome- lation, the new hull surface can be obtained by moving
try. It has been used to generate candidate geometries the nodes of the triangular meshes according to the
in the shape optimization, such as ship hull form opti- movement defined at given stations.
mization[15,17-18]. A detailed description of the method The sectional area curve can be described in va-
can be found in authors’ previous work[15]. rious ways, e.g., spline polynomials. In the present
In order to modify the hull form, two types of study, the following formulation has been used to
control points (fixed and movable) are employed in the describe the sectional area curve:
RBF based interpolation method. These control points
can be defined either on or off the hull surface. The f n ( x) = f 0 ( x) + g ( x, a1a , a 2 a , a1 f , a 2 f ) (7a)
fixed control points are used to keep the hull surface
near them unchanged. The movable control points are   x − x1  
1/ 2

used as design variables to modify the ship hull surfa- g = a1a 0.5 1 − cos 2π  , x1 ≤ x ≤ a 2a
ce in the shape optimization process. Their positions   a 2 a − x1  
are determined by a given optimization algorithm to (7b)
minimize the objective functions. 1/ 2
In the present study, the surface modification   x − a 2a 
technique based on RBF interpolation method is used g = −a1a 0.5 1 − cos 2π  , a 2 a ≤ x ≤ x2
  a 2 a − x2  
in the first step for the bulbous bow generation and
modification. Specifically, an initial bulbous bow is (7c)
177

 
1/ 2 variables as well as the modification methods associa-
x − x3   ted with each design variable are defined in Table 2.
g = α1 f 0.5 1 − cos 2π  , x3 ≤ x ≤ α 2 f
  α 2 f − x3  
Table 2 Summary of the design variables
(7d)
No. Definition Modification Range
1/ 2
  x − α 2 f  1
Bulbous
RBF 1 ( x) [0.508, 0.52]
g = −α1 f 0.5 1 − cos 2π  , α 2 f ≤ x ≤ x4 bow length
  α 2 f − x4  
Bulbous
(7e) 2 RBF 1 ( y ) [0.00198, 0.00698]
bow width
Bulbous
g = 0 , elsewhere (7f) 3 RBF 1 ( z ) [−0.0316, − 0.0146]
bow height
Entrance
where f n ( x) denotes the new sectional area curve, 4 SAC α1 f [−0.012, 0.012]
angle
f o ( x) the initial sectional area curve, g ( x, a1a , a 2 a , Fore-body
5 SAC α 2 f [0.2, 0.35]
αα variation
1 f , 2 f ) the shape function, a1a , a 2a , α1 f and α 2 f

the parameters determined during the optimization, x1 6 Run angle SAC a1a [−0.015, 0.015]
and x4 are the end positions of the deformable after- After-body
7 SAC a 2a [−0.1, − 0.3]
body and fore-body, respectively, x2 and x3 are the variation

end positions of fixed middle body. The slopes of the


sectional area curve at the fore-body and after-body 2.3 Surrogate model construction
To further accelerate the optimization process, a
are defined by the parameters α1 f and a1a , respecti-
radial basis function based surrogate model is develo-
vely, and the locations of the fixed station at the fore- ped to approximate the objective functions during the
body and after-body are controlled by the parameter optimization process. The details of the RBF based
a 2a and α 2 f , respectively. By changing parameters surrogate model can be found in authors’ previous
work[19]. Four main steps are required to build a surro-
α1 f , α 2 f , a1a and a 2a , various sectional area curves
gate model in the ship hull form optimization for redu-
can be obtained, and the new hull forms can be produ- ced drag. Figure 6 shows the flowchart of these steps.
ced accordingly from the initial hull form by the shi-
fting method. A demonstration figure of the shifting
method is shown in Fig.5. The advantages of this
modification method are explained and illustrated by
Kim et al.[15].

Fig.6 Flowchart of the RBF surrogate model construction

The design of experiments (DOE) is a strategy for


selecting sample points in the design space that aim at
maximizing the amount of information acquired. In the
present study, the Latin hypercube sampling (LHS)
Fig.5 Shape function (a) and the sectional area curves (b) technique[32] is adopted to allocate 70 data points in
the design space. Once the sample points in the design
By combing two surface modification methods, space are generated, the candidate hull forms can be
there are seven design variables in total used in the produced using the surface modification tool. The total
present study. The definition and range of these design resistance can then be evaluated for these hull forms at
178

three given design speeds using the simple CFD tool problems. The details of the algorithm can be found in
SSF to construct three RBF based surrogate models. the references[33,34]. Compared with other population
After the RBF based surrogate model is constru- based optimization algorithms, one of the advantages
cted, the cross validation is performed to examine the of the ABC algorithm is that it requires only a few
accuracy of the model. The basic idea of the cross parameters: population size, i.e., the bee colony size
validation is to leave out one sample point, and then (CS), the maximum number of iterations (maxCycle),
predict it using the model constructed by the remaining and the maximum number of trials for the bees (limit).
sample points. The difference between the exact value In the present study, a multi-objective artificial
of the objective function at the given sample point bee colony (MOABC) algorithm is developed based
evaluated by the CFD tool and the approximate value on the single objective ABC algorithm proposed by
of the objective function at the given sample point Huang et al.[34] by introducing the dominated solution
predicted by the RBF based surrogate model constru- conception and an external archive for storing non-
cted with other sample points is calculated. If the dominated solutions. As a result, the present MOABC
difference is small enough, the model is valid. Other- algorithm only requires one more parameter in compa-
wise, increase the number of sampling points, and rison with the single-objective ABC algorithm. This
repeat the steps shown in Fig.6. In addition to the new parameter defines the external archive size (AS),
evaluation of three objective functions for each sample i.e., the maximum number of non-dominated solutions
hull form generated, the displacement of each sample during the iterations.
hull form is also calculated and recorded. Therefore, a Numerical experiments are conducted for the op-
total of four RBF based surrogate models are constru- timization of typical benchmark functions using the
cted to predict the resistance at three design speeds MOABC optimizer developed in this study. The resu-
and the displacement of the hull forms generated in lts show that CS = 100 , maxCycle = 100 , limit = 100
the optimization process. The cross validations for and AS = 100 are the appropriate choices and there-
these models are performed and the results are shown fore such settings are used in the present study.
in Fig.7. It can be observed from Fig.7 that the estima-
ted objective function values ( f E ) given by the surro-
gate model show a good agreement with the exact 3. Results and analysis
objective function values ( f C ) calculated by the CFD 3.1 Optimization results
tool directly or calculated in terms of the hull geome- The design optimization of the Series 60 hull
try directly. model for reduced drag at three given speeds is carried
out on a desktop PC (OpenSuse 12.1, CPU: Intel R
Xeon(R) X5260 @3.33G Hz, Memory: 24GB). After
the RBF based surrogate model is constructed, it takes
less than 10 s to obtain the optimal solutions using the
MOABC optimizer. The computing time for the entire
optimization that includes the construction of the RBF
based surrogate model is about 40 min on the PC with
above configurations.
The optimal solution set, i.e., Pareto-front, obtai-
ned by the MOABC optimizer for three objective fun-
ctions defined in Eq.(1b), is plotted in Fig.8 with
empty squares. It can be observed from Fig.8 that the
relationships between objective functions f1  f 2 ,
and f1  f3 show a strong trade-off in the Pareto-
front. This implies that the hull form with the mini-
mum resistance at lower speeds does not have mini-
mum resistance at higher speeds. The hull form with
the minimum resistance at higher speeds does not
Fig.7 Cross validations of the surrogate models about three have the minimum resistance at lower speeds either. It
objective functions (a-c) and the constraint (d)
can also be seen from Fig.8 that the relationship
2.4 Multi-objective artificial bee colony optimizer between objective functions f 2  f3 is almost linear.
The artificial bee colony (ABC) optimization al- Therefore, a special attention is required when desi-
gorithm is based on the intelligent foraging behavior gners select optimal designs from the solution of the
of the honey bee swarm and was proposed by Pareto-front. For example, the optimal solution deno-
Karaboga[33] to solve single objective optimization ted as Case 3 in Fig.8 represents an optimal hull form
179

that has a 2%, 8%, and 9% reduction of the drag at at given design speeds with respect to the original hull
three design speeds, respectively, with respect to the is listed in Table 3. It can be observed from Table 3
original hull. Two other hull forms, denoted as Case 1 that the maximum reduction of the displacement of
and Case 2 in Fig.8, are so chosen that the hull form in the optimal hull with respect to the original hull is
Case 1 has the largest drag reduction at the high speed 0.76%, which is consistent with the constraint that the
and the hull form in Case 2 has the largest drag redu- relative reduction of the displacement should be less
ction at the low speed. The value of the objective fun- than 1%.
ctions for the original Series 60 hull model should be
zero according to the definition given in Eq.(1b), Table 3 Relative changes of the displacement (V ) , wetted
which is denoted as Case 0 in Fig.8. surface area ( S wet ) , and the total resistance at
given design speeds ( RT1 , RT2 , RT3 ) of each optimal
hull with respect to the original hull
Case V/% S wet / % RT1 / % RT2 / % RT3 / %

1 −0.76 0.70 0.92 −14.21 −13.48


2 0.98 1.77 −2.07 3.95 −1.02
3 −0.02 1.02 −1.41 −8.84 −9.14

Fig.9 Comparisons of the total drag among the original hull


(Case 0) and three optimal hulls (Case 1-Case 3)

Fig.8 The Pareto-front in the objective function space

In order to compare the drag reduction, three


optimal hulls and the original hull, denoted as Cases
1-3 and Case 0 in Fig.8, respectively, are evaluated
using simple CFD tool SSF. The relative changes of
the displacement (V ) , wetted surface area ( S wet ) , and
Fig.10 Comparisons of the profiles between the original hull
the total resistance ( RT1 , RT2 , RT3 ) of each optimal hull (Case 0) and three optimal hulls (Case 1-Case 3)
180

In order to examine the performance of the plotted in Fig.11, respectively. The comparison of the
chosen optimal hull forms at off design speeds, the sectional area curves (SAC) from the original hull and
total drag of the optimal hull forms are evaluated by three optimal hulls are plotted in Fig.12. As shown in
the simple CFD tool SSF for a range of speeds. The Figs.10-11, a large modification of the entire hull
comparisons of the total drag between the original hull form can be observed in all three optimal hulls. This is
and the optimal hulls are plotted in Fig.9. It can be consistent with the fact that both local and global hull
observed from Fig.9 that the hull form in Case 3 surface modifications are allowed in this study. It can
results in a reduction of the drag in the entire speed also be observed that the fairness and the practical
range, while the hull forms in Case 1 and Case 2 manufacturability are well preserved for the optimal
provide maximum reduction of the drag at high speeds hull forms obtained, which demonstrates that the pre-
and low speeds, respectively. sent hull form modification technique is very effecti-
ve.

Fig.12 Comparisons of the sectional area curves from the ori-


ginal hull (Case 0) and three optimal hulls (Case 1-Case
3)

It should be noted that the shape of the optimal


hull forms, i.e., the size of the bulbous bow, the longi-
tudinal distribution of the displacement, depends on
the definition of the objective function, i.e., the design
speeds. The hull form with a fat bulbous bow shows
the better drag reduction at the high speed (Case 1),
but worse performance at the low speed, vice versa
(Case 2). Narrow sectional shapes of the after-body
are obtained in Case 1, while a wider sectional shape
is obtained in Case 2. Finally, a combination of the
modification of the Case 1 and Case 2 can be obse-
rved in Case 3, which results in a moderate reduction
of the drag in the entire design speed range.

3.2 Validation of the optimal hulls using a high-


fidelity CFD tool
The present simulation-based design of the opti-
mal hull forms is based on the simple (low-fidelity)
CFD tool for the evaluation of the objective functions
that define the hydrodynamic performance of the hulls.
Therefore, it is necessary to validate the hydrodyna-
mic performance of the optimal hull forms obtained
using a high-fidelity CFD tool. In the present study, a
Fig.11 Comparisons of the body plans between the original hull
(Case 0) and three optimal hulls (Case 1-Case 3) high-fidelity open source CFD solver, called
OpenFOAM (www.openfoam.com), is used to evalua-
The comparisons of the profiles between the ori- te the total drag of the original hull and the optimal
ginal hull (Case 0) and three optimal hulls (Case 1- hull obtained from Case 1 that has the largest drag
Case 3) are plotted in Fig.10, respectively. The com- reduction at high speeds. OpenFOAM has been vali-
parisons of the body plans between the original hull dated in authors’ work for the prediction of the total
(Case 0) and three optimal hulls (Case 1-Case 3) are drag of a ship advancing in calm water at a constant
181

speed[35]. rated by the original hull form and the optimal hull
According to the prediction results from the form obtained in Case 1 at the low speed ( Fr = 0.22)
OpenFOAM, the relative drag of the optimal hull form are comparable, which implies that there is very small
obtained in Case 1 with respect to the original hull is change in the wave resistance between these two hulls.
reduced by 8.07% at Fr = 0.27 and by 14.5% at So the increase of the total resistance at the low speed
Fr = 0.32 and increased by 3.8% at Fr = 0.22 The is mainly from the friction resistance due to the in-
trend predicted by the high fidelity solver is consistent crease of the wetted surface area. The results of wave
with the results obtained from the simple CFD tool pattern comparison are consistent with these of wave
based optimization tools. resistance predictions.

3.3 Validation of the optimal hulls by model tests


In order to validate the optimal hull forms obtai-
ned, a series of model tests were conducted at Wuhan
University of Technology. The initial hull model and
the optimal hull model obtained for the best perfor-
mance at the high speed (Case 1) were built and tested
in the towing tank that is 102 m long, 10.5 m wide,
and 2 m deep for a series of speeds ranging from
Fr = 0.18 to 0.34. Figure 14 shows the original hull
model and the optimal hull model used in the model
tests.

Fig.14 The optimal hull model (a) and the original hull model
(b) used in the model tests

Fig.15 Comparison of the total resistance obtained from the


experimental measurements for the original hull model
and the optimized hull model from Case 1

The comparison of the total resistance obtained


from the experimental measurements for two models
Fig.13 Comparison of wave patterns between the original hull is plotted in Fig.15. It can be observed from Fig.15
and the optimal hull obtained from Case 1 at three
speeds
that the optimal hull model obtained from Case 1 has
better performance at high speeds but worse performa-
The comparison of the wave patterns between the nce at low speeds, which is consistent with the results
original hull and the optimal hull obtained in Case 1 at from the initial numerical predictions and the late vali-
three speeds is shown in Fig.13. It can be seen from dations with the high-fidelity CFD tool. Specifically,
Fig.13 that the waves generated by the optimal hull the total resistance of the optimal hull obtained from
form at high speeds ( Fr = 0.27 and 0.32) are smaller Case 1 is increased by 5.93% at Fr = 0.22 with respe-
than these by the original hull form. The waves gene- ct to the original hull, while the total resistance of the
182

optimal hull obtained from Case 1 is reduced by 6.42% mic optimization of a trimaran[J]. Ship Technology Re-
and 13.10% at speeds Fr = 0.27 and 0.32 with respe- search, 2002, 49(2): 70-92.
[5] PERI D., CAMPANA E. Multidisciplinary design optimi-
ct to the original hull.
zation of a naval surface combatant[J]. Journal of Ship
Research, 2003, 47(1): 1-12.
[6] CHEN P., HUANG C. H. An inverse hull design problem
4. Conclusions in optimizing the desired wake of ship[J]. Journal of ship
A simulation based design tool is further develo- Research, 2002, 46(2): 138-147.
ped and enhanced in this study and applied to the [7] TAHARA Y., STERN F. and HIMENO Y. Computational
fluid dynamics–Based optimization of a surface combata-
design optimization of a cargo ship for reduced drag.
nt[J]. Journal of ship Research, 2004, 48(4): 273-287.
This design process employs two simple ship hull [8] SAHA G. K., SUZUKI K. and KAI H. Hydrodynamic
surface modification methods to modify the hull form optimization of ship hull forms in shallow water[J]. Jour-
locally and globally during the optimization process. nal of Marine Science and Technology, 2004, 9(2): 51-
The NM-theory based simple CFD tool is used for the 62.
drag (objective function) evaluation and for the const- [9] HARRIES S., ABT C. and HOCHKIRCH K. Modeling
meets simulation-process integration to improve design[C].
ruction of a RBF based surrogate model to approxi-
Honorary Colloquium for Prof. Hagen, Prof. Schluter
mate the objective functions. A newly developed and Prof. Thiel. Duisburg, Germany: University of
MOABC optimizer is used together with RBF based Duisburg-Essen, 2004 .
surrogate model to determine the optimal designs very [10] TAHARA Y., TOHYAMA S. and KATSUI T. CFD based
efficiently. A set of optimal designs, i.e., optimal hulls, multi-objective optimization method for ship design[J].
are obtained and analyzed. For the purpose of valida- International Journal for numerical methods in fluids,
2006, 52(5): 499-527.
tion, a high-fidelity solver is used to evaluate the
[11] CAMPANA E. F., PERI D. and TAHARA Y. et al. Shape
performance of the optimal hull forms obtained. The optimization in ship hydrodynamics using computational
model tests are conducted to validate the performance fluid dynamics[J]. Computer Methods in Applied Me-
of the optimal hull forms as well. chanics and Engineering, 2006, 196(1): 634-651.
The comparison of the numerical results and [12] ZALEK S. F., PARSONS M. G. and BECK R. F. Evolu-
experimental measurements show that the present op- tionary multicriterion optimization for propulsion and sea-
keeping[C]. 26th ONR Symposium on Naval Hydrody-
timization tools can produce superior hull forms with
namics. Rome Italy, 2006.
reduced drag. This study also demonstrates that the [13] KIM H. Y., YANG C. and LÖHNER R. et al. A practical
present multi-objective optimization tools can be used hydrodynamic optimization tool for the design of a mono-
in the simulation based design of hull forms for redu- hull ship[C]. The 18th International Offshore and Polar
ced drag in the preliminary and early stage design due Engineering Conference. Vancouver, Canada, 2008.
to its simplicity and efficiency. [14] KIM H., YANG C. and KIM H. Y. et al. Hydrodynamic
optimization of a modern container ship using variable
fidelity models[C]. The Nineteenth International Off-
shore and Polar Engineering Conference. Osaka, Japan,
Acknowledgements 2009.
This work was sponsored by the Office of Naval [15] KIM Hyunyu1 Y., YANG Chi. A new surface modifica-
Research (ONR). Ms. Kelly Cooper is the technical tion approach for CFD-based hull form optimization[J].
monitor. The authors would also like to thank Dr. Journal of Hydrodynamics, 2010, 22(5Suppl.): 520-525.
[16] KIM H. Y., YANG C. Hydrodynamic optimization of
Baiwei Feng and Dr. Haichao Chang for conducting
multihull ships[C]. Proceedings of the FAST2011.
the model tests sponsored by the National Natural Honolulu, Hawaii, USA, 2011.
Science Foundation of China (Grant Nos. 51279147, [17] KIM H. Y., YANG C. Design optimization of bulbous
51179143 and 51479150) and providing the experime- bow and stern end bulb for reduced drag[C]. The Twenty-
ntal data for the validation. third International Offshore and Polar Engineering
Conference. Anchorage, Alaska, USA, 2013.
[18] HUANG F., KIM H. Y. and YANG C. A new method of
ship bulbous bow generation and modification[C]. The
References Twenty-fourth International Offshore and Polar Engi-
neering Conference. Busan Korea, 2014.
[1] PERCIVAL S., HENDRIX D. and NOBLESSE F. Hydro- [19] HUANG F., WANG L. and YANG C. Hull form optimi-
dynamic optimization of ship hull forms[J]. Applied zation for reduced drag and improved seakeeping using a
Ocean Research, 2001, 23(6): 337-355. surrogate based method[C]. The Twenty-fifth Internatio-
[2] HINO T. Shape optimization of practical ship hull forms nal Offshore and Polar Engineering Conference. Kona,
using navier-stokes analysis[C]. Proceedings of the 7th Big Island, Hawaii, USA, 2015.
International Conference on Numerical Ship Hydrody- [20] HUANG F., WANG L. and YANG C. et al. Hull form
namics. Nantes, France, 1999. optimization of a TriSWACH for reduced drag[C]. Pro-
[3] PERI D., ROSSETTI M. and CAMPANA E. Design opti- ceedings of the FAST2015. Washington DC, USA, 2015.
mization of ship hulls via CFD techniques[J]. Journal of [21] WANG L., HUANG F. and YANG C. et al. Hydrodyna-
Ship Research, 2001, 45(2): 140-149. mic optimization of a wedge hull[C]. Proceedings of the
[4] YANG C., SOTO O. and LOHNER R. et al. Hydrodyna- FAST2015. Washington DC, USA, 2015.
183

[22] HAN S., LEE Y. S. and CHOI Y. B. Hydrodynamic hull [29] ZHANG C., HE J. and ZHU Y. et al. Stationary phase and
form optimization using parametric models[J]. Journal of numerical evaluation of far-field and near-field ship
Marine Science and Technology, 2012, 17(1): 1-17. waves[J]. European Journal of Mechanics-B/Fluids,
[23] PARK J.-H., CHOI J. E. and CHUN H. H. Hull-form opti- 2015, 52: 28-37.
mization of KSUEZMAX to enhance resistance performa- [30] HUANG F., LI X. and NOBLESSE F. et al. Illustrative
nce[J]. International Journal of Naval Architecture and applications of the Neumann–Michell theory of ship
Ocean Engineering, 2015, 7(1): 100-114. waves[C]. 28th Workshop on Water Waves and Floa-
[24] YANG C., KIM H. Y. and NOBLESSE F. A practical ting Bodies. L’Isle sur la Sorgue, France, 2013.
method for evaluating steady flow about a ship[C]. Pro- [31] YANG Chi, HUANG Fuxin and KIM Hyunyul. Hydrody-
ceedings of the FAST2007. Shanghai, China, 2007, 118- namic optimization of a TriSWACH[J]. Journal of Hy-
126. drodynamics, 2015, 26(6): 856-864.
[25] NOBLESSE F., HUANG F. and YANG C. The [32] MCKAY M., CONOVER W. and BECKMAN R. A com-
Neumann–Michell theory of ship waves[J]. Journal of parison of three methods for selecting values of input
Engineering Mathematics, 2013, 79(1): 51-71. variables in the analysis of output from a computer code[J].
[26] YANG Chi, HUANG Fuxin and NOBLESSE Francis. Technometrics, 1979, 21(2): 239-245.
Practical evaluation of the drag of a ship for design and [33] KARABOGA D. An idea based on honey bee swarm for
optimization[J]. Journal of Hydrodynamics, 2013 25(5): numerical optimization[R]. Vol. 200. Technical report tr06,
645-654. Kayseri, Turkey: Erciyes university, 2005.
[27] HUANG F., YANG C. and NOBLESSE F. Numerical [34] HUANG F., WANG L. and YANG C. A new improved
implementation and validation of the Neumann–Michell artificial bee colony algorithm for ship hull form optimiza-
theory of ship waves[J]. European Journal of Mechani- tion[J]. Engineering Optimization, 2016, 48(4): 672-686.
cs-B/Fluids, 2013, 42: 47-68. [35] YANG Chi, HUANG Fuxin and WANG Lijue et al.
[28] NOBLESSE F., HUANG F. and YANG C. Evaluation of Numerical simulations of highly nonlinear steady and
ship waves at the free surface and removal of short unsteady free surface flows[J]. Journal of Hydrodynami-
waves[J]. European Journal of Mechanics-B/Fluids, cs, 2011, 23(6): 683-696.
2013, 38: 22-37.

You might also like