You are on page 1of 14

Gang Wang

College of Fisheries,
Ocean University of China,
Qingdao 266003, China
Effects of Euler Angles of Vertical
e-mail: wg@stu.ouc.edu.cn

Liuyi Huang1
Cambered Otter Board on
Hydrodynamics Based on

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
College of Fisheries,
Ocean University of China,

Response Surface Methodology


Qingdao 266003, China
e-mail: huangly@ouc.edu.cn

Lei Wang
East China Sea Fisheries Research Institute, and Multi-Objective Genetic
Chinese Academy of Fishery Sciences,
Shanghai 200090, China
e-mail: 57085089@qq.com
Algorithm
Fenfang Zhao In this present work, effects of three Euler angles (angle of attack (AOA), angle of trim
College of Fisheries, (AOT), and angle of pitch (AOP)) of vertical cambered otter board on hydrodynamic char-
Ocean University of China, acteristics (drag coefficient (Cd), lift coefficient (Cl), center-of-pressure coefficients (Cp))
Qingdao 266003, China were studied based on numerical simulation combined with Kriging response surface meth-
e-mail: zhaoff@ouc.edu.cn odology (KRSM) and multi-objective genetic algorithm (MOGA). Wind tunnel experiments
were carried out to validate the accuracy of the response surface based on numerical simu-
lation. It was demonstrated that AOA had noticeable effects on Cd and Cl, while AOT and
Xinxin Wang AOP had fewer effects. The working posture of the otter board was recommended to lean
College of Fisheries,
inward (0 deg–6 deg) and forward (−10 deg–0 deg) to improve the lift-drag ratio without
Ocean University of China,
sacrificing Cl. The influences of AOT and AOP on positions of center-of-pressure points
Qingdao 266003, China
were less significant than that of AOA and decreasing with the increase of AOA. Besides,
e-mail: wxinxin@ouc.edu.cn
the response surface of hydrodynamic coefficients around the critical AOA was a decent
indicator of the occurrence of stall. Finally, three candidate cases were selected to
Rong Wan satisfy the high working efficiency by MOGA, which was consistent with the above recom-
National Engineering Research Center for Oceanic
mendations. This study provided a scientific reference of response surface experimental
Fisheries,
investigations methodology in the fishery engineering and the configuration of Euler
Shanghai Ocean University,
angles of otter board. [DOI: 10.1115/1.4048154]
Shanghai 201306, China;
College of Fisheries, Keywords: otter board, KRSM, MOGA, numerical simulation, hydrodynamics,
Ocean University of China, computational fluid dynamics, ocean space utilization
Qingdao 266003, China
e-mail: rwan@shou.edu.cn

1 Introduction hydrodynamic performance of otter board. Park et al. [5,6], Wang


et al. [7], and Shen et al. [8] studied effects of cambered ratio, dihe-
Otter board is an essential component to facilitate the expansion
dral, and sweepback angles and the existence of wing-end plates on
of trawl mouth in the horizontal direction, and its resistance
the hydrodynamics of otter board based on flume experiments or
accounts for 30% of the total resistance of the trawl system.
hydrogen bubbles tests, illustrating the most critical factors affect-
During the process of working, hydrodynamics of the otter board
ing the hydrodynamic performance are wing-end plate and dihedral
and its adaptability to marine environmental protection will directly
angle, followed by cambered ratio and sweepback angle. In the field
affect the efficiency of trawl catch [1]. Therefore, the improvement
of working inclinations, Lin et al. [9] analyzed that AOP had little
of hydrodynamic performances, reduce the resistance, and maintain
influence on the hydrodynamics and stability of the saucer-shaped
stability is the core issue in the design or optimization of the otter
otter board, which was rarely used in the fishery. Furthermore,
board.
the effects of Euler angles (AOA, AOT, and AOP) on the hydrody-
To better understand the hydrodynamic characteristics and carry
namics of vertical cambered trawl door with a low-AR was investi-
out the optimization of otter board, extensive studies were con-
gated. Still, conclusions might be one-sided as the interactions
ducted concentrating on the impacts of factors of structures and
among three Euler angles were not taken into consideration, and
working inclinations on hydrodynamics, aiming to optimize the
the optimal Euler angles combination to improve the working effi-
design and usage of otter board. In terms of structural investiga-
ciency of the otter board was not proposed either [10]. In summary,
tions, Patterson and Watts [2], Fukuda et al. [3], and Xu et al. [4]
previous studies lacked a comprehensive exploration of impacts of
conducted model experiments and numerical simulations on the
all Euler angles on the working efficiency of otter board, and that
influences of aspect ratio (AR) on the hydrodynamics of different
did not provide a recommended Euler angles configurations of
kinds of otter boards and founded AR has significant impacts on
otter board in the working state.
Response surface methodology (RSM), as an optimization
method of comprehensive experimental design and mathematical
1
Corresponding author. modeling, could predict the quantitative relationship among inde-
Contributed by the Ocean, Offshore, and Arctic Engineering Division of ASME for
publication in the JOURNAL OF OFFSHORE MECHANICS AND ARCTIC ENGINEERING.
pendent variables and induced results within the global scope of
Manuscript received January 31, 2020; final manuscript received June 11, 2020; design space by carrying out several tests of some representative
published online September 4, 2020. Assoc. Editor: Maurizio Collu. points. The optimal solution satisfying the objective function

Journal of Offshore Mechanics and Arctic Engineering APRIL 2021, Vol. 143 / 021401-1
Copyright © 2020 by ASME
could be obtained by the genetic algorithm (GA) generally. The direction, pointing rightward, and z is the vertical axis in the direc-
application of RSM and GA has been widely applied in the field tion of gravity. Meanwhile, the body reference frame, which is
of the chemical industry, biology, and civil engineering [11,12], attached to the otter board, is defined as Ɓ = {O; xb, yb, zb}, with
but it was rarely involved in the research area of fishery engineering. the origin O located at the center of the chord line at its trailing
RSM has apparent advantages over the orthogonal experiment in edge. xb follows the chord line, and yb is orthogonal to xb and on
terms of a large number of variables or response values to be the pseudo-symmetry plane, pointing outward. zb is perpendicular
assessed, as the complicated functional relationship [13]. Orthogo- to the xb-yb plane. As a demonstrated example, positive values of
nal design cannot obtain quantitative expressions or graphical links AOT, AOP, and AOA were also shown in Fig. 2, which means
among independent variables and attributive variables. Therefore, the model was configured to lean inward and back, respectively.

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
the optimal results cannot be more accurate compared with those The detailed relationship between Euler angles and gestures of
obtained from RSM combined with GA. Moreover, it is difficult the otter board is shown in Table 2.
for the analysis of orthogonal experimental results to consider the
interaction effects of independent values, while RSM is better at
2.3 Experimental Validation-Wind Tunnel Experiment. In
analyzing that. Besides, Kriging gives the best linear unbiased pre-
order to validate the accuracy of numerical simulation, the wind
diction of the intermediate values in a limited area, so the Kriging
tunnel experiment of the otter board was completed in the NH-2
method was used for modeling complex response surfaces [12].
low-speed wind tunnel located at Nanjing University of Aeronautics
Meanwhile, MOGA was also applied to search optimization
and Astronautics, China. This experiment was carried out in a small
points in RSM. The whole processed algorithm combing RSM
test section whose dimensions were 6 m (length) × 3 m (width) ×
and the MOGA is a hybrid variant of the popular non-dominated
2.5 m (height), and the cross-sectional area was 7.18 m2 (Fig. 3).
sorted genetic algorithm-II (NSGA-II) based on controlled elitism
The otter board model and the six-component mechanical tower-
concepts [14].
balance were jointed with a rob connection, which was used to
Computational fluid dynamics (CFD), which is a mature and
adjust the AOA of the model and located in the middle of the
effective numerical methodology, has been widely adopted in the
wind tunnel section.
investigations of hydrodynamics and flow characteristics in the
The AOA between the chord of the otter board and incoming
field of fishery engineering [4,15]. Besides, flume experiments,
flow was ranging from 0 deg to 70 deg. Keeping AOT and AOP
wind tunnel experiments, and sea trial tests could also be used to
remained at 0 deg, intervals of 2.5 deg were used for the AOA
carry out the research about the hydrodynamics of the otter board
ranging from 0 deg to 50 deg, followed by intervals of 5 deg
and validate the accuracy of CFD. In this paper, the impacts of
ranging from 50 deg to 70 deg. Wind speed 28 m/s was used in
three Euler angles (AOA, AOT, and AOP) of vertical cambered
this test when the Re = VL/ν reached 0.52 × 106 (in the range of self-
otter board on hydrodynamic characteristics (Cd, Cl, and Cp) were
modeling region of otter board, hydrodynamic coefficients would
studied based on CFD combined with KRSM and MOGA. Further-
not change significantly with Re). During the test, the six-
more, we have proposed an optimum combinatorial collocation of
component tower-balance remained static, and the rob connection
three Euler angles of the otter board in the working state.
could be rotated to achieve the AOA. The tower-balance could
obtain forces and moments through the conversion module. The
experimental schedule arrangement could produce 25 datasets at
2 Material and Methods each constant AOA, including hydrodynamics and stability param-
2.1 The Otter Board Model. The otter board investigated in eters listed above. Besides, all test data was carried out for bracket
this study was a two-main-panel vertical cambered otter board interference correction, accomplished by removing the effects of
(Fig. 1) used in West Africa, of which prototype (scale ratio: bare rob. However, according to the previous experience of wind
1/20) was used in the single trawler, with a power of 880 kW and tunnel test, this correction method could not altogether remove
working speed is 3 kn. The otter board model was made of the effects of bare rob, resulting in a little error between real
carbon structural steel Q235 and painted on the surface with anti- value and experimental value, especially when the AOA is small,
rust material, and its dimensions are as illustrated in Table 1. so the cases of which AOA was 0 deg–10 deg was discarded in
the following parts.
2.2 Euler Angles. To carry out the investigations of effects of
Euler angles of the otter board on the hydrodynamics, two frames of 2.4 Numerical Simulation of the Otter Board Model. Grid
reference should be defined to assist transforming data and chang- generation and boundary conditions should be discussed above
ing Euler angles (Fig. 2). The Earth or wind tunnel reference all. The outer computational domain was shown in Fig. 4, the
frame is defined as Ɛ = {E; x, y, z}, with the origin E on the same scale with the experimental wind tunnel. Due to the greatly
balance attachment point. x is the streamwise direction of the com- gradient change of various physical quantities around the otter
putational domain, pointing forward, y is the horizontal spanwise board, to ensure the accuracy of numerical simulation, an inner

Fig. 1 Otter board model and the angle of main panel and guide plate. Left: 1: main
panel; 2: guide plate; right: α: 30 deg; β: 25 deg; γ: 16 deg; δ: 6 deg, representing the
angle between chord of guide plate and horizontal direction.

021401-2 / Vol. 143, APRIL 2021 Transactions of the ASME


Table 1 Dimensions of otter board model

L (m) b (m) AR S (m2) e1 (m) e2 (m)

0.462 0.492 0.983 0.217 0.249 0.258

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
Fig. 3 Setup of otter board model in wind tunnel

Fig. 2 Two frames of reference (left view of otter board)

domain, 3 m (length) × 1.6 m (width) × 1.6 m (height), was setup


around the otter board, and more sophisticated grids were applied
to the inner domain compared with the outer domain. Unstructured
tetrahedral grids were adopted in this work. To capture gradient
change of quantities near surfaces of otter board, the density of
grids around the otter board surface was intensified by setting Fig. 4 The computational domain
eight layers inflations presented in Fig. 5, of which the first layer
thickness was 8 × 10−4 m, ensuring y+ ≥15. Mesh size of grids
was obtained through a grid independence test to insure stable The numerical simulation was carried out by ANSYS CFX 18.2,
and convergent calculations (Fig. 6), drag and lift force were which used element-based finite volume method [16] to solve the
selected as the target indices to inspect the test. Two indices got Reynolds-averaged Navier–Stokes equations. Considering the com-
stable after the gird number reached 5.69 × 106, the deviations of putational accuracy and load, an appropriate turbulence model
the drag force and lift force along the last three cases were less adapted to different hydrodynamics issues is significant in CFD.
than 1%. Therefore, considering the speed and accuracy of the Standard k–ɛ, RNG k–ɛ, and k–ɛ EARSM are ordinary high-
calculation, the grid numbers 5.69 × 106 was chosen. The total Reynolds number k–ɛ turbulence models in the CFX frame, so
number of nodes and elements were approximately 1.09 × 106 and the model selection process was carried out based on the above
5.69 × 106 in each case after meshing, respectively. three models. Scalable wall treatment, which is the improvement
The boundary conditions were illustrated in Fig. 4, inlet and of standard wall treatment and overcomes one of the significant
outlet were defined as the fluid’s entrance and exit of the computa- drawbacks of the standard wall function approach in that they can
tional domain in the x-direction, respectively. k and ϵ, which repre- be applied on arbitrarily fine meshes, was used for the near-wall
sent turbulent intensity of the entire domain, were initialized at the treatment. The adoption of high resolution in solving advection
inlet. In order to facilitate comparison with the wind tunnel test, the schemes and turbulence schemes ensured the calculation accuracy.
fluid media adopted in the numerical simulation was incompressible The convergence standard was 10−4, and the maximum iteration
air with T = 25 °C, ρ = 1.225 kg/m3, V = 28 m/s (keeping consistent step was 1000 in this study, i.e., the calculation stopped if the resid-
with experiments), ν = 15 × 10−6 m2/s and without heat transfer. ual of continuity, velocity, pressure, turbulent kinetic energy, and
Given that the fluid could discharge completely, the outlet boundary the energy-dissipation rate fell below 10−4.
was the pressure outlet with a relative pressure 0 Pa. In addition, all
surfaces of the otter board and other surfaces of the domain were
assumed to be the no-slip wall. 2.5 Data Statistics. The drag coefficient Cd and the lift coeffi-
cient Cl related to the hydrodynamics can be calculated using the
experimental or simulation value of drag force Fd and lift force
Fl, presented as follows:
Table 2 The correspondence between Euler angles and
gestures 2Fd
Cd = (1)
ρSV 2
AOT AOP
Euler AOA
angles + + − + − 2Fl
Cl = (2)
ρSV 2
Gestures Deflection Lean Lean Lean Lean
inward outward back forward where ρ is the fluid density, S is the plane area of the otter board, and
V is the incoming flow velocity. The moment coefficient and the

Journal of Offshore Mechanics and Arctic Engineering APRIL 2021, Vol. 143 / 021401-3
Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
Fig. 5 Grid generation in computational domain

initially obtained in wind tunnel coordinates, are straightforwardly


transformed to body coordinates using matrix R as follows:

Fb = RT (ψ, θ, φ)Fe
Mob = RT (ψ, θ, φ)Moe
⎛ ⎞ (8)
cθ cψ sφ sθ cψ − cφ sψ cφ sθ cψ + sφ sψ
⎜ cφ sθ sψ − sφ cψ ⎟
R(ψ, θ, φ) = ⎝ cθ sψ sφ sθ sψ + cφ cψ ⎠
−sθ sφ cθ cφ cθ
where ψ represents AOA, θ represents AOP, and φ represents AOT.
cx and sx denote cosine and sine, respectively, of the angle x.
Finally, Cm and Cp could be obtained through Eqs. (3)–(7) using
the converted forces and moments.

2.6 Response Surface Methodology and Multi-Objective


Genetic Algorithm. Built from the design of experiments
Fig. 6 Grid independence test
(DOE), RSM quickly provided the approximated values of the
output parameters throughout the design space without having to
perform a complete solution. Furthermore, MOGA was an efficient
center-of-pressure coefficient along the chord and span direction, method that could be used in goal-handling. It could be used to
associated with the stability performance of the otter board, are search for any candidate point, according to constrains. Simulta-
listed as neously, the response surface would be refined second through
2Mc MOGA study. The overall procedure consisted of three steps.
Cmc = (3) As presented in Fig. 7, central composite design (CCD), an
ρSLV 2
experimental design methodology widely used in establishing
response surfaces in the process of optimization studies [18], was
2Mb adopted to generate design points in the whole design space.
Cmb = (4)
ρSbV 2 Second, design points were used to construct response surfaces,
but the prediction accuracy might not be enough due to the
CN = Cd × sin AOA + Cl × cos AOA (5) limited quantities of design points. Afterward, the solution to this
problem was the refinement that we reconstructed the response
surface by increasing more orthogonal design points to the design
e1 − (Mc /N) e1 Cmc space until the relative diversities of results between numerical
C pc = = − (6)
L L CN simulation and response surface are less than 3%. The refinement
would stop if the relative deviations between numerical simulation
e2 − (Mb /N) e2 Cmb results and response surface results of the above six parameters
C pb = = − (7) were all less than 3%. Finally, optimization constraints were pre-
b b CN
defined before MOGA, and then the genetic algorithms optimizer
where Cmc and Cmb are the moment coefficients along the chord and was used to analyze the existing response surface, several candi-
span of otter board, respectively. Center-of-pressure coefficient of dates satisfied the contrasts would be picked up. To ensure the
the chord and the span (Cpl and Cpb) are equivalent to the positions quality of the response surface, the candidate points should be val-
of the hydrodynamics acting on the otter board in the chord and idated through numerical simulation. Moreover, the refinement
span directions [4]. routine mentioned above should be carried out again if not meet
The direction of Fd and Fl were related with the earth coordinates, the accuracy requirements.
i.e., it is relevant to the incoming flow direction, not the position of
otter board, so the calculation of Cd and Cl were acquired based on 2.6.1 Design of Experiments. CCD, as an effective means to
the earth coordinate frame directly. However, considering three gather more data with a few numbers of discrete design points,
Euler angles are changing when the otter board rotates about any was used in the design of experiments. This design method was
axis, all moment coefficients Cm and center-of-pressure coefficients composed of the following three parts: (1) a full factorial or frac-
Cp are usually given in body coordinates. Thus, Cd and Cl should be tional factorial design; (2) an additional design, often a star
transformed into the body coordinates to calculate normal force design in which experimental points are at a distance α from its
coefficient (CN), which is also based on body coordinates. Accord- center. More usually, α takes diverse values based on rotatabil-
ing to the research [17], the forces and moments concerning O, ity and orthogonality of the design and number of experiments;

021401-4 / Vol. 143, APRIL 2021 Transactions of the ASME


values in a limited space by taking the information from some
neighboring points instead of all the existed points [21]. Kriging
modeling is specialized in modeling multiple variables issues and
complex computer processes [22]. Compared with polynomial
regression modeling, Kriging modeling can use more flexible
experimental design as a pre-process and take correlation among
data into account. It could be used to fit sophisticated relationships
among data and form a high-quality response surface. The ordinary
Kriging model adopted in this paper are shown as

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
Y(x) = f (x) + Z(x) (9)
where Y(x) is the unknown target function representing the interest
response point, f (x) is the low-order polynomial function represent-
ing the global approximation of response surface, respectively. The
stochastic term Z(x) indicating the local behavior of model is shown
as follows:

n
Z(x) = λi r(xi , x) (10)
i=1

In addition, Z(x) obeys Gauss stochastic distribution with zero


mean and covariance:

Cov[Z(xi ), Z(xj )] = σ 2 R([r(xi , xj )]) (11)


here σ represents the variance of z(x), R is a symmetric correlation
2

matrix with ones along diagonal, of which dimension is (n0 × n0),


and r (xi, xj) is the Gaussian spatial correction function between
any two of the n0 sample point presented as follows:


M
j 2
r(x , x ) = exp −
i j
θk |xk − xk |
i
(12)
k=1

Fig. 7 Flowchart of the whole RSM and MOGA M is the number of design variables, θk is the unknown coeffi-
cient of correction. Furthermore, σ is the function of θk. xik and xjk
are the kth components of the sample points xi and xj, respectively.
(3) a central point [19]. In this study, range of AOA is selected from Predicted estimates, ŷ (x*), of the real values y(x*) at the untested
10 deg to 70 deg. Since excessive tilt would affect the stability of point x* is given by
the otter board and cause unexpected outcomes while working,
∧ ∧ ∧
upper and lower limits of AOT and AOP should be set maximally y = β + r T (x∗ )R−1 y − f β (13)
to 10 deg based on the production experience [20]. Fifteen cases are
required to finish the procedure of design of experiments, distribu-
ted in the Cartesian three-dimensional coordinate system (Fig. 8). r T (x∗ ) = [R(x∗ , x1 ), R(x∗ , x2 ), . . . , R(x∗ , xn0 )] (14)

2.6.2 Kriging Response Surface Methodology. Initially used in To solve Eq. (12), θk could be evaluated by solving the maximum
aid geologists analyzing mineral deposits, Kriging modeling is the likelihood estimation equations, maximizing the log-likelihood
approximation of the input/output relationships and provides the function shown in Eq. (15). In a word, the Kriging model could
linear unbiased estimation of the intermediate and unknown be determined once the θk is given

n0 ln (σ 2 ) + ln |R(θk )|
Maximize − (15)
2
subject to: 0 < θk < ∞.

2.6.3 Multi-Objective Genetic Algorithm. As mentioned previ-


ously, MOGA is characterized by a fast-non-dominated sorting
approach based on controlled elitism concepts. It supports all
types of input parameters and goal-handing issues. The Pareto
ranking scheme is done by a fast, non-dominated sorting method
that is an order of magnitude faster than traditional Pareto ranking
methods [14]. First, the initial population representing the best chro-
mosomes (parents) about the objective function is selected to run
MOGA. Then, MOGA is run and generates a new population
called offspring via cross-over and mutation. Specifically, cross-
over combines two parents to produce a new offspring [23]. The
principle of cross-over is that the higher the individual fitness is,
the higher the probability of reproduction. The kernel behind it is
that the new chromosome can be better than both of the parents if
it takes the best characteristics from each of the parents [24]. For
Fig. 8 Distribution of 15 cases through CCD continuous parameters in the cross-over process, the operator

Journal of Offshore Mechanics and Arctic Engineering APRIL 2021, Vol. 143 / 021401-5
linearly combines two-parent chromosome vectors to produce two
new offspring according to the following equations:

Offspring = α × Parent1 + (1 − α) × Parent2


(16)
Offspring = (1 − α) × Parent1 + α × Parent2
where α represents the cross-over probability that could be
described as the stability of the population.
The mutation alters one or more gene values in a chromosome

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
from its initial state. It can result in an entirely new gene values
being added to the gene pool. With these new gene values, the
genetic algorithm might be able to arrive at a better solution than
was previously possible [25]. The mutation is an essential part of
the genetic search, as it helps to prevent the population from stag-
nating at any local optima. For continuous parameters, the polyno-
mial mutation operator (Eq. (17)) is applied to implement mutation:
Offspring = Parent + (upperbound − lowerbound) × δ (17)
where δ is the mutation rate reflecting the randomness of the
Fig. 9 Convergence chart with the increasing of refinement
algorithm.
points
Finally, the design points in the new population are updated, and
the whole optimization is validated for convergence. In this paper,
the initial samples of the population and the samples per iteration
0 deg to 10 deg were discarded. Raw data were transformed to Cd
were 100. The cross-over probability α was 0.98, and the mutation
(Eq. (1)), Cl (Eq. (2)). As shown in Fig. 10, the simulation results
probability δ was 0.01, respectively.
had a contemporary changing trend with the experimental results
in general. Cl was underestimated in the whole range of AOA
2.6.4 Refinement of Response Surface. In this research, refine- while Cd was not. It might be attributed to the oscillation of the
ment has two processes, including preliminary and secondary opti- otter board model in the direction perpendicular to the inflow. Cd
mizations. When the metamodeling is built first. Unless the increased while Cl increased first and then decreased with an
relative divergences between CFD and RSM results of refinements increasing AOA. The stall appeared at the critical AOA, which is
fall below 3%, new refinement points are added into the second for- 55 deg in the experiment and 47.5 deg or 50 deg in the numerical
mation of the response surface. In terms of the following optimiza- simulation, and the maximum lift coefficient was 2.09 (experiment)
tion, candidate points were used CFD simulations and comparison at the critical AOA. Furthermore, the maximum lift-drag ratio was
with the RSM values at the MOGA stage. Through this strict feed- 3.22 (experiment) with the AOA was 22.5 deg, and AOA under
back, the accuracy of the response surface would be guaranteed to working conditions of otter board (working AOA) was inferred
a great extent. ranging from 30 deg to 40 deg. In the range, the lift-drag ratio
During the generation of design points used in refinement, and the lift coefficient were pretty high, which could ensure that
random number generation and crowding distance separation per- the otter board could expand the expansion of trawl mouth while
centage (CDSP) were adopted to make sure the efficiency and valid- maintaining high working efficiency.
ity of the refinement process. CDSP has determined the minimum
allowable distance among new refinement points and existed
design points, implemented as a constraint in the search for refine-
ment points [26]. More specifically, this procedure uses the value
specified for this property to avoid areas close to failed design
points when generating the next refinement points. This property
specifies the minimum allowable distance between new refinement
points, providing a radius around failed design points that serves as
a constraint for refinement points. Finally, if two candidate refine-
ment points are closer together than the defined minimum distance,
only the first candidate is inserted as a new refinement point [27].
In general, the points number of the preliminary refinement would
have the dominant position of the response surface optimization pro-
cedure, as shown in Fig. 9. Despite some fluctuations in the early
period, it has shown a decreasing and stable gradually. Cd and Cl
need about 37 cases for the optimization (52 cases in the whole
process), while Cpb and Cpl would take approximately 67 cases
(82 cases in the whole process) to get converged, respectively. It
could be explained that the optimization of center-of-pressure coef-
ficients of the response surface depended on the determination of
moments and hydrodynamics at the same time so that it would be
more time-consuming and resourceful compared with that of hydro-
dynamic coefficients.

3 Results and Discussions


3.1 Validation Results and Turbulence Model Selection.
Figure 10 presented comparisons between numerical results (three
turbulence models) and experimental results. Due to the incomplete Fig. 10 Comparisons between numerical results (three turbu-
correction of bare rob in the small AOA, the results ranging from lence model) and experimental results

021401-6 / Vol. 143, APRIL 2021 Transactions of the ASME


Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020

Fig. 11 Variation of Cd at each constant AOA: (a) boxplot of whole AOA range and (b) variation of Cd in
the range of working AOA

Journal of Offshore Mechanics and Arctic Engineering APRIL 2021, Vol. 143 / 021401-7
Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020

Transactions of the ASME


Fig. 12 Variation of Cl at each constant AOA: (a) boxplot of whole AOA range and (b) variation of Cl
in the range of working AOA

021401-8 / Vol. 143, APRIL 2021


As mentioned earlier, choosing a reasonable turbulence model is AOP at each constant AOA, reflecting the overall change trend of
vital to carry out further investigations. From Fig. 10, Cd and Cl Cd under the influence of Euler angles. The height of each box rep-
obtained by k–ɛ EARSM model were much closer (below 10%) resents the range of hydrodynamics with AOT and AOP at each
to the experimental data compared with standard k–ɛ model and constant AOA. When AOA was in the scope of 10 deg–70 deg,
RNG k–ɛ model, so k–ɛ EARSM was adopted to carry out the Cd showed a fast then slowly increasing trend with the increase
further study. of AOA generally. AOA had prominent effects on Cd, while
AOT and AOP had fewer effects compared with AOA. In the
range of working AOA (Fig. 11(b)), with the two-dimensional
3.2 Analysis of Response Surface. To obtain the hydrody- changes of AOT and AOP, AOA had little significant effects on

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
namic performance of vertical cambered otter board under the the distribution of peaks and valleys of Cd in each response
working condition and stall condition, cases of which AOA surface, and small value of Cd obtained in the neighborhood of
ranging from 30 deg to 40 deg (working AOA) and 55 deg to which AOT = −10 deg–0 deg, AOP = 0 deg–10 deg, or AOT =
70 deg (critical AOA, stall condition) were chosen to analyze the 8 deg–10 deg, AOP = −10 deg–−2 deg, i.e., the appropriate config-
hydrodynamics response surface of AOA, AOT, and AOP. Two- uration of otter board body in the working condition to lean outward
dimensional contours were adopted to analyze visually (x-axis and back or lean inward and forward could reduce the resistance to a
and y-axis represent AOT and AOP, respectively, and color illus- certain extent. Furthermore, through quantitative analysis of the
trate response values of RSM). Results and discussions were illus- response surface, the posture, which is leaning outward and back,
trated as follows. was recommended for a broader range of available AOT and
AOP configurations, and the minimum of Cd was obtained in that
3.2.1 Drag, Lift Coefficients, and Lift-Drag Ratio. attitude. Moreover, Cd was large when AOT = −10 deg–0 deg,
Figure 11(a) was the boxplot of variation of Cd with AOT and AOP = −10 deg–−4 deg or AOT = 8 deg–10 deg, AOP = 8 deg–

Fig. 13 Variation of lift-drag ratio in the range of working AOA

Journal of Offshore Mechanics and Arctic Engineering APRIL 2021, Vol. 143 / 021401-9
10 deg, other hydrodynamic coefficients and actual working situa- which gesture was leaning outward and back was in the low-
tion not being considered, operators should avoid arranging board resistance state from the previous analysis. With the increase of
body to lean outward and forward or lean inward and back. AOA, the peak area of the lift coefficient of the response surface
Concerning Cl presented in Fig. 12(a), it was increasing first and had the trend to move toward the area where AOT was positive
then decreasing after the critical AOA within the AOA range of (lean inward) and AOP was negative (lean forward). Besides, the
0 deg–70 deg. AOA had prominent effects on Cl, while AOT and careful consideration of the lift-drag ratio of the otter board
AOP had fewer effects compared with AOA. Meanwhile, the prin- should be taken before drawing an instructive conclusion.
ciple of change of Cl with AOT and AOP was gradually consistent, According to the analysis results presented in the response
illustrating that change of AOA had no significant effects on distri- surface (Fig. 13), the global lift-drag ratio presented the decreasing

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
butions of peaks and valleys in the response surface. Cl is maximum tendency with the increase of AOA within the range of working
when AOT = −10 deg–6 deg and AOP = −10 deg–0 deg, and Cl AOA. Lift-drag ratio was considerably large when AOT = 0 deg–
got the minimum value when AOT = −10 deg–−2 deg, AOP = 10 deg (lean inward), AOP = −10 deg–5 deg (lean forward or a
4 deg–10 deg, respectively. In other words, it is beneficial to little back) or AOT = −10 deg–0 deg (lean outward), AOP =
improve the lift coefficient of the otter board in the working condi- 0 deg–8 deg (lean back). However, based on the results of lift coef-
tion when the gesture is configured to lean outward wholly or ficients shown in the last paragraph, the lift coefficient of the otter
inward slightly and forward instead of leaning outward and back. board that is in the posture of leaning outward and back was
Nevertheless, its resistance was considerable when in the attitude small. As the consequence, it is recommended that this vertical cam-
of leaning outward and forward. However, the otter board of bered otter board should be configured to lean inward (0 deg–6 deg)

Fig. 14 Variation of Cpb in the range of working AOA

021401-10 / Vol. 143, APRIL 2021 Transactions of the ASME


and forward (−10 deg–0 deg) to improve the lift-drag ratio without determine the exact location of center-of-pressure on the otter
sacrificing lift coefficient. board before more conclusions could be drawn.
In the region of working AOA, the center-of-pressure coefficients
in the chord direction (Cpl) were slightly higher than 0.5, and as AOA
3.2.2 Center-of-Pressure Coefficients of the Span and Chord. increased, the overall Cpl was gradually reduced to 0.5 (Fig. 15). The
As can be noted from Fig. 14, the center-of-pressure coefficient in distribution of valley and peak of Cpl in all response surface contours
the span direction (Cpb) of each response surface changed slightly had little relationship with AOA. Cpl was large when AOT = 4 deg–
within the scope of working AOA, and the values were between 10 deg (lean inward), AOP = −10 deg–0 deg (lean forward) or AOT
0.48 and 0.57. The distribution of peaks and valleys remained = −10 deg–0 deg (lean outward), AOP = 0 deg–10 deg (lean back),

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
almost consistent with the increase of AOA. The change of AOT respectively. And Cpl got the small value when AOT = −10 deg–
had no significant effect on Cpb, that is to say, if other angles 0 deg (lean outward), AOP = −10 deg–−6 deg (lean forward) or
remained unchanged, there would be no significant effect on Cpb AOT = 2 deg–10 deg (lean inward), AOP = 6 deg–10 deg (lean
when the otter board was configured as an inward-leaning or back). In addition, Fig. 16 showed that the change regulation dia-
outward-leaning posture. In detail, Cpb decreased evenly from grams of the center-of-pressure point position with AOA were
0.57 to 0.48 with the change of AOP, indicating that Cpb fluctuated drawn by combining with the above data analysis of Cpb and Cpl.
slightly in the vicinity of 0.5, and the gesture is to lean back a little The red region represented the active region of center-of-pressure
to keep Cpb around 0.5. In summary, Cpb was kept stable around 0.5 varying with AOT and AOP at each constant AOA. Through this
with the change of AOA, AOT, and AOP. The center-of-pressure diagram, the effects of AOA, AOT, and AOP on the
coefficient in the chord direction (Cpl) should be considered to center-of-pressure were explicit that locations of point moved

Fig. 15 Variation of Cpl in the range of working AOA

Journal of Offshore Mechanics and Arctic Engineering APRIL 2021, Vol. 143 / 021401-11
away from the trailing edge of the otter board. Moreover, the area of separation on the suction side of the otter board with the increase of
the red region was gradually shrinking, illustrating the influences of AOA as the pressure increase is more significant around the critical
AOT and AOP on positions of the center-of-pressure point were AOA. If the separation occurs, the separated flow has a large area of
decreasing with the increase of AOA, and the impacts of AOT and vortex close to the leading edge of the otter board [4]. In addition,
AOP on Cpl was less than the impacts of AOT and AOP on Cpb, Takahashi et al. [28] investigated that boundary layer separation
which was intuitive in each diagram. occurs in the area with low shear stress.
Figure 17 showed that the results of changing of response sur-
3.2.3 Hydrodynamic Coefficients Analysis Around the Critical face contours of Cd and Cl with increasing AOA around the stall
Angle of Attack. According to the experimental results in Fig. 9, the point. It was shown that the distribution of the valley and peak of

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
critical AOA of the otter board was deduced to be 55 deg, which the response surface had an evident change with AOA for each
meant stall would happen at this point. There were lots of indicators hydrodynamic coefficient, indicating that the large amplitude per-
to characterize the occurrence of the stall. Cd or Cl increased rapidly, turbations of contours of response surface were related to the occur-
and then Cd increased slowly while Cl decreased around the critical rence of stall dramatically. The considerable separation of boundary
AOA and stall. Generally, the occurrence of the stall is often accom- layer might be accounted for the cause of this phenomenon, leading
panied by the separation of the boundary layer. There is a danger of to the large adverse-pressure-gradient in the streamwise direction of

Fig. 16 Change diagrams of the center-of-pressure point position with AOA (Color version
online).

021401-12 / Vol. 143, APRIL 2021 Transactions of the ASME


Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
Fig. 17 Variation of Cd and Cl with increasing AOA around the stall point: (a) Cd and (b) Cl

the turbulent boundary layer dominated the pressure distribution Besides, Cd and Cl were so mutually conflicting that no single
at the suction side of the guide plate around the critical AOA, and solution could fully satisfy the objectives. Three candidate points
the suction force cannot be provided to ensure lift coefficient con- were picked up for higher efficiency of working operations. Numer-
tinues increasing [29]. Accordingly, it is a decent indicator of the ical simulation was carried out to verify the availability of candidate
occurrence of the stall that contours of response surface around points selected by MOGA from the response surface. As the results
the critical AOA were quite different from that around other show in Table 3, the maximum relative errors for Cd and Cl were
AOA, but specific reasons for patterns with the change of AOA, below 2% between CFD and RSM results, which was acceptable
AOT, and AOP around the critical AOA should be studied in the in this MOGA process, so values predicted by the Kriging model
future study. were agreed well with real simulation results. Furthermore, postures
of cases selected by MOGA kept consistent with the recommended
3.3 Analysis of Multi-Objective Genetic Algorithm. In this postures analyzed by results of lift-drag ratio in Sec. 3.2.1, which
study, response surfaces of hydrodynamic coefficients of AOA, was leaning inward (0 deg–6 deg) and forward (−10 deg–0 deg).
AOT, and AOP were built based on the KRSM method. When optimization is updated, the best candidates (up to the
However, response surfaces in this study were corresponding to requested number) from the sample set could be obtained based
any two of the three parameters, such as the AOT-AOP response on the stated objectives. However, these results are not truly repre-
surface with a constant AOA, so it is difficult to manually search sentative of the solution set, as this approach obtains results by
for the expected value, even when the response surface is provided.
MOGA was then used to find the candidate point satisfying the
objective functions and limitations. The mathematical model of
the multi-objective problem could be developed as follows:
Objectives: Minimize Cd (AOA, AOT, and AOP), Maximize Cl
(AOA, AOT, AOP)
Subject to: 20 deg ≤ AOA ≤ 60 deg, −10 deg ≤ AOT, AOP ≤
10 deg

Table 3 Comparison of RSM and CFD results

Candidate 1 Candidate 2 Candidate 3

P1-AOA 35 deg 33.4 deg 35.8 deg


P2-AOT 2.9125 deg 0.4125 deg 7.9456 deg
P3-AOP −2.2457 deg −5.2086 deg −6.6901 deg
P4-Cd (RSM) 0.57823 0.54663 0.58872
P4-Cd (CFD) 0.58297 0.54757 0.59339
Predicted error 0.82% 0.17% 0.79%
P5-Cl (RSM) 1.5447 1.4968 1.5466
P5-Cl (CFD) 1.5589 1.5073 1.5671
Predicted error 0.92% 0.70% 1.30% Fig. 18 Two-dimensional tradeoff chart of MOGA for Cd and Cl
(Color version online).

Journal of Offshore Mechanics and Arctic Engineering APRIL 2021, Vol. 143 / 021401-13
ranking the solution by an aggregated weighted method. Schemati- References
cally, this represents only a section of the available Pareto fronts. [1] Sterling, D., 2000, The Physical Performance of Prawn Trawling Otter Boards
Figure 18 plotted the obtained Pareto-optimal front values for Cd and Low Opening Systems/by D. Sterling in Collaboration with the Australian
and Cl. Each point in Fig. 18 illustrated a specific optimal solution. Maritime Engineering CRC, The Gap, Qld, Brisbane.
[2] Patterson, R. N., and Watts, K. C., 1986, “The Otter Board as a Low-Aspect-Ratio
The best Pareto front is blue and the fronts gradually turn to red Wing at High Angles of Attack; An Experimental Study,” Fish. Res., 4(2),
(worst Pareto front). As presented, the Pareto-optimal points of pp. 111–130.
the MOGA approach were distributed at the left and upside of the [3] Fukuda, K., Hu, F. X., Tokai, T., and Matuda, K., 1999, “Effects of Aspect and
coordinate system, and the relationship between Cd and Cl is ordi- Camber Ratios on Hydrodynamic Characteristics of Biplane-Type Otter Board,”
Nippon Suisan Gakkaishi, 65(5), pp. 860–865.
narily antagonistic. [4] Xu, Q. C., Huang, L. Y., Zhao, F. F., Wang, X. X., Tang, Y. L., Liang, Z. L., Wan,

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/143/2/021401/6563757/omae_143_2_021401.pdf by NTNU Universitets Biblioteket, Liu Yi Huang on 28 September 2020
R., Sun, P., Liu, C. D., Cheng, H., and Zhao, Y. P., 2017, “Effects of Aspect Ratio
on the Hydrodynamic Performance of Full-Scale Rectangular Otter Board:
Numerical Simulation Study,” Ocean Eng., 142, pp. 338–347.
4 Conclusion [5] Park, C. D., Matuda, K., and Hu, F. X., 1996, “Effects of Dihedral and Sweepback
Angles on Lift and Drag of the Cambered Otter Board,” Nippon Suisan
In this research, the effects of Euler angles of a vertical cambered Gakkaishi, 62(6), pp. 920–927.
otter board on hydrodynamic performance were studied based on [6] Park, C. D., Matuda, K., and Tokai, T., 1994, “Flow Visualization Around
numerical simulation combined with KRSM and MOGA. The accu- Cambered Plates Using Hydrogen Bubbles,” Nippon Suisan Gakkaishi, 60(4),
pp. 485–491.
racy of KRSM and MOGA based on numerical simulation was suf- [7] Wang, M., Wang, J., Zhang, X., Yu, Y., and Xu, B., 2004, “Hydrodynamic
ficient for the validation of wind tunnel experiments. Based on this Characteristics of Vertical V Type Otter Board,” J. Fish. China, 28(3), pp. 311–315.
research, AOA had noticeable effects on Cd and Cl, while AOT and [8] Shen, X., Hu, F., Kumazawa, T., Shiode, D., and Tokai, T., 2015, “Hydrodynamic
AOP had fewer effects in contrast to AOA, and the influences of Characteristics of a Hyper-Lift Otter Board With Wing-End Plates,” Fish. Sci.,
81(3), pp. 433–442.
AOT and AOP on positions of the center-of-pressure point were [9] Lin, J., Sato, O., Nashimoto, K., and Yamamoto, K., 1989, “Efficiency and
less significant than that of AOA and decreasing with the increase Stability of Saucer-Shaped Otter Boards With Changing Attack Angle and Heel
of AOA. Second, the working posture of the otter board was recom- Angle,” Nippon Suisan Gakkaishi, 55(2), pp. 295–300.
mended to lean inward (0 deg–6 deg) and forward (−10 deg–0 deg) [10] Liu, J., Huang, H., Chen, S., Li, L., Wu, Y., Xu, G., and Rao, X., 2013,
“Hydrodynamic Characteristics of Low Aspect Ratio Vertical Cambered Otter
to improve the lift-drag ratio without sacrificing Cl. Third, response Board,” J. Fish. China, 37(11), pp. 1742–1749.
surface of hydrodynamic coefficients around the critical AOA was a [11] Bucher, C. G., and Bourgund, U., 1990, “A Fast and Efficient Response Surface
decent indicator of the occurrence of stall. At last, the optimal Approach for Structural Reliability Problems,” Struct. Saf., 7(1), pp. 57–66.
results in the processes of KRSM and MOGA were beneficial to [12] Wang, G. R., Chu, F., Tao, S. y., Jiang, L., and Zhu, H., 2015, “Optimization
Design for Throttle Valve of Managed Pressure Drilling Based on CFD
improve the efficiency of the otter board. In other words, the opti- Erosion Simulation and Response Surface Methodology,” Wear, 338–339,
mization method was a useful analytical tool for fishing gear pp. 114–121.
design and optimization. [13] Wang, H., Geng, H., Guo, J., and Li, C., 2014, “The Distinction of Orthogonal
Design and Response Surface Methodology Used to Distillation System,”
J. Heibei Univ. Technol., 43(1), pp. 2–6.
[14] Deb, K., Pratap, A., Agarwal, S., and Meyarivan, T., 2002, “A Fast and Elitist
Multiobjective Genetic Algorithm: NSGA-II,” IEEE Trans. Evol. Comput.,
Acknowledgment 6(2), pp. 182–197.
[15] Wang, G., Wan, R., Wang, X. X., Zhao, F. F., Lan, X. Z., Cheng, H., Tang, W. Y.,
The present work was supported by National Natural Science and Guan, Q. L., 2018, “Study on the Influence of Cut-Opening Ratio,
Foundation of China (Project Nos. 31072246, 31802349, and Cut-Opening Shape, and Cut-Opening Number on the Flow Field of a Cubic
41606110) and Fund CAFS (No. 2019HY-XKQ03). Artificial Reef,” Ocean Eng., 162, pp. 341–352.
[16] Filippini, G., Maliska, C. R., and Vaz, M., 2014, “A Physical Perspective of the
Element-Based Finite Volume Method and FEM-Galerkin Methods Within the
Framework of the Space of Finite Elements,” Int. J. Numer. Methods Eng.,
98(1), pp. 24–43.
Conflict of Interest [17] Mellibovsky, F., Prat, J., Notti, E., and Sala, A., 2015, “Testing Otter Board
There are no conflicts of interest. Hydrodynamic Performances in Wind Tunnel Facilities,” Ocean Eng., 104,
pp. 52–62.
[18] Selukar, R. S., Shah, A. K., and Mishra, S. N., 1995, “Efficiency Comparisons of
Central Composite Designs,” J. Stat. Comput. Simul., 52(2), pp. 177–183.
[19] Lundstedt, T., Seifert, E., Abramo, L., Thelin, B., Nyström, Å, Pettersen, J., and
Nomenclature Bergman, R., 1998, “Experimental Design and Optimization,” Chemom. Intell.
Lab. Syst., 42(1–2), pp. 3–40.
b = span length [20] Cui, J., 1987, Fishing Gear and Fishing Science, China Agriculture Press,
k = turbulent kinetic energy Beijing.
L = chord length [21] Gano, S. E., Renaud, J. E., Martin, J. D., and Simpson, T. W., 2006, “Update
Strategies for Kriging Models Used in Variable Fidelity Optimization,” Struct.
N = normal force loading on the otter board Multidiscip. Optim., 32(4), pp. 287–298.
S = otter board projected area [22] Sacks, J., Welch, W. J., Mitchell, T. J., and Wynn, H. P., 1989, “Design and
V = incoming velocity Analysis of Computer Experiments,” Stat. Sci., 4(4), pp. 409–423.
e1 = distance between center of gravity and leading edge in the [23] Murata, T., and Ishibuchi, H., 1995, “MOGA: Multi-Objective Genetic
Algorithms,” Proceedings of the IEEE Conference on Evolutionary
chord direction Computation, pp. 289–294.
e2 = distance between center of gravity and leading edge in the [24] Pawar, S. N., and Bichkar, R. S., 2015, “Genetic Algorithm With Variable Length
span direction Chromosomes for Network Intrusion Detection,” Int. J. Autom. Comput., 12(3),
Cd = drag coefficient pp. 337–342.
[25] García-Martínez, C., Rodriguez, F. J., and Lozano, M., 2018, “Genetic
Cl = lift coefficient Algorithms,” Handbook of Heuristics, Springer, Cham.
Cp = center-of-pressure coefficient [26] Zhang, J., and Xing, L., 2017, “A Survey of Multi Objective Evolutionary
Cpl = center-of-pressure coefficient of the chord Algorithms,” IEEE International Conference on Computational Science and
Cpb = center-of-pressure coefficient of the span Engineering (CSE) and IEEE International Conference on Embedded and
Ubiquitous Computing (EUC), Guangzhou, China, July 21–24, pp. 93–100.
CN = normal force coefficient [27] Raquel, C. R., and Naval, P. C., 2005, “An Effective Use of Crowding Distance in
Fd = resistance parallel to the incoming flow Multiobjective Particle Swarm Optimization,” GECCO 2005—Genetic and
Fl = lift force perpendicular to the incoming flow Evolutionary Computation Conference, New York, NY, pp. 257–264.
Mc = the moment along the chord [28] Takahashi, Y., Fujimori, Y., Hu, F., Shen, X., and Kimura, N., 2015, “Design of
Trawl Otter Boards Using Computational Fluid Dynamics,” Fish. Res., 161,
Re = Reynolds number pp. 400–407.
ɛ = turbulent dissipation rate [29] Cebeci, T., and Cousteix, J., 2005, Modeling and Computation of
ν = turbulent kinematic viscosity Boundary-Layer Flows, Springer, New York.

021401-14 / Vol. 143, APRIL 2021 Transactions of the ASME

You might also like