energies

Article

Numerical and Experimental Study of the Solo DuckWave Energy Converter

Jinming Wu 1,*, Yingxue Yao 2, Dongke Sun 1,*, Zhonghua Ni 1,* and Malin Göteman 3

1 School of Mechanical Engineering, Southeast University, Nanjing 211189, Jiangsu, China2 Shenzhen Graduate School, Harbin Institute of Technology, Shenzhen 518055, Guangdong, China;

yxyao@hit.edu.cn3 Department of Engineering Science, Uppsala University, 75121 Uppsala, Sweden;

malin.goteman@angstrom.uu.se* Correspondence: jinmingwu@seu.edu.cn (J.W.); dksun@seu.edu.cn (D.S.); nzh2003@seu.edu.cn (Z.N.)

Received: 20 March 2019; Accepted: 18 May 2019; Published: 21 May 2019�����������������

Abstract: The Edinburgh Duck is one of the highly-efficient wave energy converters (WECs).Compared to the spine-connected Duck configuration, the solo Duck will be able to use the pointabsorber effect to enhance its power capture performance. In this paper, a 3D computational fluiddynamic (CFD) model is developed to predict the hydrodynamic performance of the solo DuckWEC in regular waveswithin a wide range ofwave steepness until the Duck capsizes. A set ofexperiments was designed to validate the accuracy of the CFD model. Boundary element method(BEM) simulations are also performed for comparison. CFD results agree well with experimentalresults and the main difference comes from the friction in the mechanical transmission system.CFD results also agree well with BEM results and differences appear at large wave steepness as aresult of two hydrodynamic nonlinear factors: the nonlinear waveform and the vortex generationprocess. The influence of both two nonlinear factors iscombined to be quantitatively representedby the drag torque coefficient.The vortex generation process is found to cause a rapid drop ofthepressure force due to the vortexes taking away the kinetic energy from the fluid.

Keywords: solo Duck; computational fluid dynamic; wave steepness; experiment;hydrodynamic nonlinearity

1. Introduction

Growing demands of renewable and sustainable energy have been attracting public attentionto ocean energy, one of which is the wave energy. Wave energy is of great significance for its hugeresource, which has been estimated to be comparable with the world electricity consumption in 2012 [1],and high power density relative to wind and solar energy [2,3]. Wave energy converters (WECs) havebeen extensively studied since the 1970s as a result of the oil crisis, and several types of WECs havebeen proposed hitherto, such as attenuators, point absorbers, and terminators [4]. The EdinburghDuck, first proposed by Stephen Salter at the University of Edinburgh, is of the terminator type [5].

The Duck WEC was evolved from flapping devices [6]. It is sophistically designed and reacheda 92% hydrodynamic efficiency in regular wave tests [7]. Extensive experiments were performedduring 1975–1980 to study the performance of the Duck with axis compliant [8], torque limit [9]and slamming [10], etc. Evans proposed a two-dimension (2D) theoretical method based on theproportion of reflected and transmitted wave power, and showed that a 100% hydrodynamic efficiencyis possible for the asymmetrical Salter’s cam shape [11]. Mynett et al. proposed a 2D numericalstudy of the Duck based on linear theory of surface waves and floating bodies interaction, and thehydrodynamic efficiency and response are presented with the rotation axis rigidly or non-rigidly

Energies 2019, 12, 1941; doi:10.3390/en12101941 www.mdpi.com/journal/energies

Energies 2019, 12, 1941 2 of 19

supported, respectively [12]. Nebel used a complex-conjugate synthesizer in both theoretical andexperimental studies to predict the hydrodynamic performance of the Duck WEC, and reached a higherhydrodynamic efficiency over a broader bandwidth than previously achieved [13]. Above studies arefocused on the spine-connected Duck configurationas shown in Figure 1a. The solo Duck configurationis able to use the point absorber effect to further enhance its power capture performance [14], as shownin Figure 1b. Pizer proposed a three-dimensional linear wave-diffraction numerical model to predictthe hydrodynamic performance of a solo Duck, and found good agreement with both analytical andtheoretical results in the linear wave regime, but also showed a variety of unexpected singularities inthe damping matrix [15]. Cruz and Salter proposed a solo circular cylinder for desalination purposewith an offset pitching axis without the front beak to replace the original shape of the Duck to reducecost, and the influence of the axis position was evaluated to find the optimal configuration for thiskind of WEC [16,17]. Lucas et al. used a simplified one degree-of-freedom frequency-domain modelwith linear damping loads to optimize the performance of a solo circular cylinder, and showed thatfor a particular wave climate the optimal performance of the WEC can be acquired by relocating theballast [18].

Energies 2019, 12, x FOR PEER REVIEW 2 of 19

non-rigidly supported, respectively [12]. Nebel used a complex-conjugate synthesizer in both theoretical and experimental studies to predict the hydrodynamic performance of the Duck WEC, and reached a higher hydrodynamic efficiency over a broader bandwidth than previously achieved [13]. Above studies are focused on the spine-connected Duck configurationas shown in Figure 1a. The solo Duck configuration is able to use the point absorber effect to further enhance its power capture performance [14], as shown in Figure 1b. Pizer proposed a three-dimensional linear wave-diffraction numerical model to predict the hydrodynamic performance of a solo Duck, and found good agreement with both analytical and theoretical results in the linear wave regime, but also showed a variety of unexpected singularities in the damping matrix [15]. Cruz and Salter proposed a solo circular cylinder for desalination purpose with an offset pitching axis without the front beak to replace the original shape of the Duck to reduce cost, and the influence of the axis position was evaluated to find the optimal configuration for this kind of WEC [16,17]. Lucas et al. used a simplified one degree-of-freedom frequency-domain model with linear damping loads to optimize the performance of a solo circular cylinder, and showed that for a particular wave climate the optimal performance of the WEC can be acquired by relocating the ballast [18].

(a) (b)

Figure 1. Configuration of Ducks: (a) spine-connected Ducks; (b) solo Ducks.

Before any experimental investigation and deployment of a Duck WEC, its hydrodynamic performance analysis should be studied carefully. Accurate three-dimension (3D) theoretical models are hindered by the complex irregular shape of the Duck. Numerical boundary element methods (BEMs), such as the WAMIT software, are widely used in the design stages [7,19,20]. However, this method is based on the linear wave theory, hence will be limited within small wave steepness conditions. The computational fluid dynamic (CFD) method, which solves fully nonlinear Navier-Stokes equations covering the effect of viscous and turbulence, has been frequently used for hydrodynamic performance prediction in wave and floating body interaction process, and shows great potential in nonlinear effect prediction [21,22]. Wei et al. studied the viscous effect and slamming process on bottom hinged oscillating wave energy converters based on the CFD software ANSYS Fluent, and revealed that CFD resultsagree well with experiments in capturing local features of the flow as well as dynamics of the device [23,24]. Agamloh et al. utilized the CFD method based on the Comet software to perform coupled fluid structure interaction simulations, and the technique was expanded for an array of two buoys to determine the interference between them [25]. Lopez et al. implemented a 2DReynolds Average Navier-Stokes (RNAS) model to optimize the performance of OWC WECs, and found that numerical results agree well with physical experiments, and the model was utilized to find optimal damping for the WEC [26]. Penalba et al. identified the resources of nonlinearity of the WEC system and compared the approaches that model the nonlinearity [27]. Windt et al. tried to provide a best practice guideline for the application of CFD methods in wave energy conversion processes [28].

To the authors’ knowledge, little has been studied about the hydrodynamic modeling of the solo Duck WEC by the CFD method, which is required when predicting hydrodynamic performance in large wave steepness conditions, capsizing and other nonlinear situations. In this paper, a 3D CFD model is developed to predict the hydrodynamic performance of the solo Duck WEC in regular waveswithin a wide range of wave steepness until the Duck capsizes, and resultsare compared with both experimental and BEM results for validation and hydrodynamic nonlinear factors are

Figure 1. Configuration of Ducks: (a) spine-connected Ducks; (b) solo Ducks.

Before any experimental investigation and deployment of a Duck WEC, its hydrodynamicperformance analysis should be studied carefully. Accurate three-dimension (3D) theoretical modelsare hindered by the complex irregular shape of the Duck. Numerical boundary element methods(BEMs), such as the WAMIT software, are widely used in the design stages [7,19,20]. However, thismethod is based on the linear wave theory, hence will be limited within small wave steepness conditions.The computational fluid dynamic (CFD) method, which solves fully nonlinear Navier-Stokes equationscovering the effect of viscous and turbulence, has been frequently used for hydrodynamic performanceprediction in wave and floating body interaction process, and shows great potential in nonlineareffect prediction [21,22]. Wei et al. studied the viscous effect and slamming process on bottom hingedoscillating wave energy converters based on the CFD software ANSYS Fluent, and revealed that CFDresults agree well with experiments in capturing local features of the flow as well as dynamics of thedevice [23,24]. Agamloh et al. utilized the CFD method based on the Comet software to performcoupled fluid structure interaction simulations, and the technique was expanded for an array of twobuoys to determine the interference between them [25]. Lopez et al. implemented a 2DReynoldsAverage Navier-Stokes (RNAS) model to optimize the performance of OWC WECs, and found thatnumerical results agree well with physical experiments, and the model was utilized to find optimaldamping for the WEC [26]. Penalba et al. identified the resources of nonlinearity of the WEC systemand compared the approaches that model the nonlinearity [27]. Windt et al. tried to provide a bestpractice guideline for the application of CFD methods in wave energy conversion processes [28].

To the authors’ knowledge, little has been studied about the hydrodynamic modeling of the soloDuck WEC by the CFD method, which is required when predicting hydrodynamic performance inlarge wave steepness conditions, capsizing and other nonlinear situations. In this paper, a 3D CFDmodel is developed to predict the hydrodynamic performance of the solo Duck WEC in regular waveswithin a wide range of wave steepness until the Duck capsizes, and results are compared with both

Energies 2019, 12, 1941 3 of 19

experimental and BEM results for validation and hydrodynamic nonlinear factors are identified. InSection 2, a numerical wave tank (NWT) embedded with the Duck is established and the solutionmethod is specified. In Section 3, a set of experiments was designed to validate the CFD model. InSection 4, the BEM model is introduced. In Section 5, the convergence of the CFD model is firstlystudied, then the CFD model is validated by comparing the predicted hydrodynamic responses withboth experimental and BEM results, and finally vortex generation at the end of the Duck and thehydrodynamic force distribution along the longitudinal direction of the Duck are investigated.Forsimplicity, the scope of this paper is limited within regular waves, which lay the foundation of theWEC behavior in irregular waves.

2. The CFD Model

The CFD model is implemented based on the ANSYS Fluent 16.2 [29] software (Version 16.2,ANSYS: Canonsburg, PA, USA), which describes the fluid motion by the RAN Sequation.

2.1. The Geometry of the Duck

The geometry of the cross section of the solo Duck is shown in Figure 2. The cross section iscomposed of the paunch, stern and beak parts, and rotates about the axis O. The stern part is of circularshape with O as its center point, so that the pitch motion does not radiate waves afterward, R denotesthe radius of the stern part, G denotes the mass center, θ5 denotes the pitch angular displacement ofthe mass center with the counterclockwise direction as positive, and d denotes the depth of the rotationaxis. Oxyz is the Cartesian coordinate system with x in the incident wave propagating direction, y inthe longitudinal direction of the Duck, z in the opposite direction of the gravity acceleration, and z-xplane locating at the middle cross section of the Duck along the longitudinal direction.

Energies 2019, 12, x FOR PEER REVIEW 3 of 19

identified. In Section 2, a numerical wave tank (NWT) embedded with the Duck is established and the solution method is specified. In Section 3, a set of experiments was designed to validate the CFD model. In Section 4, the BEM model is introduced. In Section 5, the convergence of the CFD model is firstly studied, then the CFD model is validated by comparing the predicted hydrodynamic responses with both experimental and BEM results, and finally vortex generation at the end of the Duck and the hydrodynamic force distribution along the longitudinal direction of the Duck are investigated.For simplicity, the scope of this paper is limited within regular waves, which lay the foundation of the WEC behavior in irregular waves.

2. The CFD Model

The CFD model is implemented based on the ANSYS Fluent 16.2 [29] software(Version 16.2, ANSYS: Canonsburg, PA, USA), which describes the fluid motion by the RANSequation.

2.1. The Geometry of the Duck

The geometry of the cross section of the solo Duck is shown in Figure2. The cross section is composed of the paunch, stern and beak parts, and rotates about the axis O. The stern part is of circular shape with O as its center point, so that the pitch motion does not radiate waves afterward, R denotes the radius of the stern part, G denotes the mass center, θ5 denotes the pitch angular displacement of the mass center with the counterclockwise direction as positive, and d denotes the depth of the rotation axis. Oxyz is the Cartesian coordinate system with x in the incident wave propagating direction, y in the longitudinal direction of the Duck, z in the opposite direction of the gravity acceleration, and z-x plane locating at the middle cross section of the Duck along the longitudinal direction.

Figure 2. Geometry of the cross section of the solo Duck.

2.2. The Geometry of the Numerical Wave Tank

Since the solo Duck is symmetrical relative to the z-x plane, only half of the Duck is modeled to reduce the computational cost. The geometry of the numerical wave tank is shown in Figure3. The length of the NWT is 4λ, where λ is wavelength of the incident wave, and is chosen as a result of the balance between results’ stability and computationalcost. The width and water depth of the NWT is the same as the experimental setup, and is 2.5 m and 4.5 m, respectively. Since the Duck WEC is of water surface piercing type, the hydrodynamic simulation of the Duck WEC is a two-phase flow problem. Therefore, another 1 m depth above thez= 0 plane is reserved for the air phase. The wave is generated at the inlet boundary, and absorbed in the wave absorbing zone, which lasts one wavelength as [28] suggests to avoid wave reflection. The Duck is placed 1.5λ away from the inlet boundary.

Figure 2. Geometry of the cross section of the solo Duck.

2.2. The Geometry of the Numerical Wave Tank

Since the solo Duck is symmetrical relative to the z-x plane, only half of the Duck is modeledto reduce the computational cost. The geometry of the numerical wave tank is shown in Figure 3.The length of the NWT is 4λ, where λ is wavelength of the incident wave, and is chosen as a result ofthe balance between results’ stability and computational cost. The width and water depth of the NWTis the same as the experimental setup, and is 2.5 m and 4.5 m, respectively. Since the Duck WEC isof water surface piercing type, the hydrodynamic simulation of the Duck WEC is a two-phase flowproblem. Therefore, another 1 m depth above the z = 0 plane is reserved for the air phase. The wave isgenerated at the inlet boundary, and absorbed in the wave absorbing zone, which lasts one wavelengthas [28] suggests to avoid wave reflection. The Duck is placed 1.5λ away from the inlet boundary.

Energies 2019, 12, 1941 4 of 19Energies 2019, 12, x FOR PEER REVIEW 4 of 19

Figure 3. Geometry of the numerical wave tank.

2.3. Boundary Conditions

To solve the RANS equation, boundary conditions should be specified to obtain the particular solution. Boundary conditions for all physical boundaries are listed in Table 1. Waves are generated at the Inlet boundary by using the velocity inlet boundary condition appended with a user defined functionspecifying the velocity and wave height time history as in [22]. At the Outlet and Top boundaries, pressure outlet boundary condition is chosen and the open channel flow option is activated to allow for specifying pressure distribution along the zdirection. At the Side, Bottom and Duck boundaries, wall boundary condition is utilized as in the experimental situation. Symmetric boundary condition is applied at the Symmetric boundary to set the velocity component inthe y direction to zero. In the wave absorbing zone, the kinetic energy of the fluid is dissipated by adding negative momentum source Sx and Sy into the momentum equations. The momentum sources are defined as:

ex

p e

x xS f ux x

−= −

−, (1)

ey

p e

x xS f vx x

−= −

−, (2)

where f is the dissipation rate, xeandxp are the position where the wave absorbing zone starts and ends, respectively, u and v are thefluid velocity in thex and y direction, respectively.

Table 1.Boundary conditions for all physical boundaries.

Physical Boundary Name Inlet Top, Outlet Symmetry Duck, Side,

Bottom Boundary condition Velocity inlet Pressure outlet Symmetry Wall

It should be mentioned that the sidewall may affect the captured power of the Duck WEC. For an axisymmetric floating oscillating water column device in a wave channel, Gomes et al. found that the sidewall will increase the captured power up to a maximum of 15% for regular waves due to the transverse sloshing mode [30]. However, for different types of WECs, the influence of the sidewall is expected to be different, since the radiated and diffracted waves, which generate transverse waves, behave differently. To quantitatively identify the influence of the sidewall in the case of the Duck WEC, captured power of the WEC should be calculated as a function of the width of the wave tank at different wave periods and wave heights, so that a database can be built for different situations of the wave tank. To keep the topic of this paper focused, the influence of the sidewall on the captured power will not be studied in detail in this paper, but is arranged as the work of the next phase.

Figure 3. Geometry of the numerical wave tank.

2.3. Boundary Conditions

To solve the RANS equation, boundary conditions should be specified to obtain the particularsolution. Boundary conditions for all physical boundaries are listed in Table 1. Waves are generatedat the Inlet boundary by using the velocity inlet boundary condition appended with a user definedfunctionspecifying the velocity and wave height time history as in [22]. At the Outlet and Topboundaries, pressure outlet boundary condition is chosen and the open channel flow option is activatedto allow for specifying pressure distribution along the z direction. At the Side, Bottom and Duckboundaries, wall boundary condition is utilized as in the experimental situation. Symmetric boundarycondition is applied at the Symmetric boundary to set the velocity component inthe y direction tozero. In the wave absorbing zone, the kinetic energy of the fluid is dissipated by adding negativemomentum source Sx and Sy into the momentum equations. The momentum sources are defined as:

Sx = − fx− xe

xp − xeu, (1)

Sy = − fx− xe

xp − xev, (2)

where f is the dissipation rate, xe and xp are the position where the wave absorbing zone starts andends, respectively, u and v are the fluid velocity in the x and y direction, respectively.

Table 1. Boundary conditions for all physical boundaries.

Physical Boundary Name Inlet Top, Outlet Symmetry Duck, Side, Bottom

Boundary condition Velocity inlet Pressure outlet Symmetry Wall

It should be mentioned that the sidewall may affect the captured power of the Duck WEC. For anaxisymmetric floating oscillating water column device in a wave channel, Gomes et al. found thatthe sidewall will increase the captured power up to a maximum of 15% for regular waves due to thetransverse sloshing mode [30]. However, for different types of WECs, the influence of the sidewall isexpected to be different, since the radiated and diffracted waves, which generate transverse waves,behave differently. To quantitatively identify the influence of the sidewall in the case of the DuckWEC, captured power of the WEC should be calculated as a function of the width of the wave tank atdifferent wave periods and wave heights, so that a database can be built for different situations of thewave tank. To keep the topic of this paper focused, the influence of the sidewall on the captured powerwill not be studied in detail in this paper, but is arranged as the work of the next phase.

Energies 2019, 12, 1941 5 of 19

2.4. The Mesh

The mesh near the Duck is shown in Figure 4. Hexahedral cells are used to discretize all fluidregions. Mesh resolution is increased near both the free water surface and the end of the Duck becauseof relative large variability of flow parameters in these regions. A sliding mesh method is employedto model the wave interaction with the Duck by specifying the angular velocity of the rotation zone,which is kept static relative to the Duck. The data exchange between the static zone and the rotationzone is realized by a pair of overlapping circular cylinder interfaces, at where the flow variables aresynchronized by interpolation. Compared to the dynamic mesh method, a sliding mesh method canavoid large element skew after large motion of the nodes so that the mesh quality can be guaranteed.The angular velocity of the rotation zone is the same as the Duck, hence can be determined from thesolution of equation of motion of the Duck as:

J..θ+ c

.θ+ Tg + Tw = 0, (3)

where J is the moment of inertia of the Duck, c is the linear damping coefficient of power take-off (PTO)system, Tg and Tw are the torques applied by the gravity and hydrodynamic force of the Duck relativeto O, respectively.

Energies 2019, 12, x FOR PEER REVIEW 5 of 19

2.4. The Mesh

The mesh near the Duck is shown in Figure4. Hexahedral cells are used to discretize all fluid regions. Mesh resolution is increased near both the free water surface and the end of the Duck because of relative large variability of flow parameters in these regions. A sliding mesh method is employed to model the wave interaction with the Duck by specifying the angular velocity of the rotation zone, which is kept static relative to the Duck. The data exchange between the static zone and the rotation zone is realized by a pair of overlapping circular cylinder interfaces, at where the flow variables are synchronized by interpolation. Compared to the dynamic mesh method, a sliding mesh method can avoid large element skew after large motion of the nodes so that the mesh quality can be guaranteed. The angular velocity of the rotation zone is the same as the Duck, hence can be determined from the solution of equation of motion of the Duck as:

0g wJθ cθ T T+ + + = , (3)

whereJ is the moment of inertia of the Duck, c is the linear damping coefficient of power take-off (PTO) system, Tg and Tw are the torques applied by the gravity and hydrodynamic force of the Duck relative to O, respectively.

Figure 4. The mesh near the Duck.

2.5. Solution Method

The RANS equation for the flow region is discretized by the finite-volume method (FVM). The volume of fluid (VOF) method is used to track the free surface position in this two phase (water and air) flow problem. In [23], 5 different built-in turbulence models of the ANSYS Fluent are cross compared to see their differences in modeling wave interaction with a flap-type WEC, and resultsshow that the difference can be neglected. In this paper, the renormalization group (RNG) k-ε modelis applied as the turbulence model. Boundary layer effects around walls are modeled by a standard wall function. Pressure implicit with splitting of operator (PISO) method is chosen to treat pressure-velocity coupling and pressure-based segregated solver only (PRESTO!) scheme is selected for pressure interpolation. The equation of motion of the Duck, i.e.,Equation (3), is solved by the Euler method. Small time step should be chosen to not only satisfy small Courant number to increase computational stability, but also reduce local and global error of this integration method.

3. Experimental Model

The experiments were performed in the wind tunnel and water flume (HTWF) laboratory at Harbin Institute of Technology. The wave tank is 50 m long, 5 m wide and 4.5 m water depth. Both regular and irregular waves can be generated. The wave period is between 0.5 s and 5 s, and the wave height ranges from2 cm to 40 cm. The wave height at the position of the Duck position was calibrated before the experiments. The dimension of the Duck cross section is 1:33 scaled compared to the full scale designed in [15].

Rotation zone

Interface pair

Figure 4. The mesh near the Duck.

2.5. Solution Method

The RANS equation for the flow region is discretized by the finite-volume method (FVM).The volume of fluid (VOF) method is used to track the free surface position in this two phase (waterand air) flow problem. In [23], 5 different built-in turbulence models of the ANSYS Fluent are crosscompared to see their differences in modeling wave interaction with a flap-type WEC, and resultsshowthat the difference can be neglected. In this paper, the renormalization group (RNG) k-εmodelis appliedas the turbulence model. Boundary layer effects around walls are modeled by a standard wall function.Pressure implicit with splitting of operator (PISO) method is chosen to treat pressure-velocity couplingand pressure-based segregated solver only (PRESTO!) scheme is selected for pressure interpolation.The equation of motion of the Duck, i.e.,Equation (3), is solved by the Euler method. Small time stepshould be chosen to not only satisfy small Courant number to increase computational stability, but alsoreduce local and global error of this integration method.

3. Experimental Model

The experiments were performed in the wind tunnel and water flume (HTWF) laboratory atHarbin Institute of Technology. The wave tank is 50 m long, 5 m wide and 4.5 m water depth.Both regular and irregular waves can be generated. The wave period is between 0.5 s and 5 s, and thewave height ranges from 2 cm to 40 cm. The wave height at the position of the Duck position was

Energies 2019, 12, 1941 6 of 19

calibrated before the experiments. The dimension of the Duck cross section is 1:33 scaled compared tothe full scale designed in [15].

The configuration of the experimental setup is shown in Figure 5a,b. A low speed disk-typepermanent magnet generator, which is often used in small wind turbines, is used as the PTO system,and it is connected to the rotation axis of the Duck by cascaded synchronizing wheels and gearsto increase speed. The speed-up system will drive the generator out of the non-linear speed zone,where the output voltage is not proportional to the input rotation speed, so that the linear dampingassumption in numerical simulations will be more credible. The load of the PTO system consists ofsymmetric three phase resistors, and the resistors are adjustable to find optimal damping coefficientfor power absorption. The torque transferred by the synchronizing belt is measured by a pair oftension meters, which are embedded into the synchronizing belt on both sides. The torque can bedetermined by the output voltage difference between the two meters firstly multiplied by a factor,which is obtained by the multiplication of the amplification coefficient of the mechanical and electricalsystems, and secondly multiplied by the radius of the synchronizing wheel. The synchronizing beltand tension meters are pre-tensioned to avoid force impact if the belt is suddenly tensioned from theslacked state. The pitch angular displacement of the Duck is measured by an encoder connected tothe rotation axis of the Duck on the other side also by a synchronizing belt. A Kistler9257B threecomponents dynamometer is used to measure the hydrodynamic forces in surge and heave directions.It is fixed on one side of the rotation axis supporting system assuming that the hydrodynamic forcesare distributed symmetrically relative to the middle section of the Duck. Output voltage signals of theall these transducers are imported to a data acquisition (DAQ) card, which transfers analog signals todigital signals, and then the output of DAQ card is transferred to a computer, where data is storedand analyzed.

Energies 2019, 12, x FOR PEER REVIEW 6 of 19

The configuration of the experimental setup is shown in Figures5a,b. A low speed disk-type permanent magnet generator, which is often used in small wind turbines, is used as the PTO system, and it is connected to the rotation axis of the Duck by cascaded synchronizing wheels and gears to increase speed. The speed-up system will drive the generator out of the non-linear speed zone, where the output voltage is not proportional to the input rotation speed, so that the linear damping assumption in numerical simulations will be more credible. The load of the PTO system consists of symmetric three phase resistors, and the resistors are adjustable to find optimal damping coefficient for power absorption. The torque transferred by the synchronizing belt is measured by a pair of tension meters, which are embedded into the synchronizing belt on both sides. The torque can be determined by the output voltage difference between the two meters firstly multiplied by a factor, which is obtained by the multiplication of the amplification coefficient of the mechanical and electrical systems, and secondly multiplied by the radius of the synchronizing wheel. The synchronizing belt and tension meters are pre-tensioned to avoid force impact if the belt is suddenly tensioned from the slacked state. The pitch angular displacement of the Duck is measured by an encoder connected to the rotation axis of the Duck on the other side also by a synchronizing belt. A Kistler9257B three components dynamometer is used to measure the hydrodynamic forces in surge and heave directions. It is fixed on one side of the rotation axis supporting system assuming that the hydrodynamic forces are distributed symmetrically relative to the middle section of the Duck. Output voltage signals of the all these transducers are imported to a data acquisition (DAQ) card, which transfers analog signals to digital signals, and then the output of DAQ card is transferred to a computer, where data is stored and analyzed.

(a)

(b)

Figure 5. Configuration of the experimental setup: (a) physical model; (b) simplified schematic diagram.

In the experiment, the radius of the stern part is 0.2 m; the depth of the rotation axis is 0.05 m; and the coordinate of the mass center is (−0.038 m, 0.046 m); and the mass of the Duck is 139.2kg.In numerical simulations, half of the Duck mass are specified since only half Duck is modeled.

4. BEM Model

For a more thorough comparison in this paper, BEM simulations are also performed to compare with both CFD and experimental results. WAMIT v7.0 [31] (Version 7.0, WAMIT Inc, Chestnut Hill, MA, USA) is employed to calculate the hydrodynamic response of the Duck in regular wave conditions.The pitch angular displacement amplitude and hydrodynamic force amplitude can be calculated from non-dimensional response amplitude operator (RAO) data by:

5 5 / 2θ θ HL= , (4)

31,3 1,3 / 2X X ρgHL= , (5)

Duck Shelf

Generator

Dynamometer

Tension meters

Encoder

Figure 5. Configuration of the experimental setup: (a) physical model; (b) simplified schematic diagram.

In the experiment, the radius of the stern part is 0.2 m; the depth of the rotation axis is 0.05 m;and the coordinate of the mass center is (−0.038 m, 0.046 m); and the mass of the Duck is 139.2kg.Innumerical simulations, half of the Duck mass are specified since only half Duck is modeled.

4. BEM Model

For a more thorough comparison in this paper, BEM simulations are also performed to comparewith both CFD and experimental results. WAMIT v7.0 [31] (Version 7.0, WAMIT Inc, Chestnut Hill,MA, USA) is employed to calculate the hydrodynamic response of the Duck in regular wave conditions.

Energies 2019, 12, 1941 7 of 19

The pitch angular displacement amplitude and hydrodynamic force amplitude can be calculated fromnon-dimensional response amplitude operator (RAO) data by:

θ5 = θ5HL/2, (4)

X1,3 = X1,3ρgHL3/2, (5)

where θ5 and θ5 are the dimensional and non-dimensional pitch angular displacement amplituderespectively, X1,3 and X1,3 are the dimensional and non-dimensional surge and heave hydrodynamicforce amplitude in the x and z direction respectively, ρ is the water density, g is the gravity acceleration,H is the wave height, and L is the characteristic length [31].

5. Results and Discussion

In this section, the convergence of the CFD model is first investigated. Then, the CFD modelis validated by comparing the predicted hydrodynamic responses with both experimental and BEMresults. Finally, the hydrodynamic nonlinearity identified by the CFD model and the hydrodynamicforce distribution along the longitudinal direction of the Duck are studied.

5.1. Convergence Study of the CFD Model

Before the CFD model is used to perform hydrodynamic simulations of the solo Duck WEC, itshould be both mesh and time step resolution independent. Figure 6a,b show the measured free surfaceelevation h at the Duck position with the Duck be removed from the NWT, and hydrodynamic responsesof the Duck at wave period T = 1.6 s and wave height H = 0.075 m. Here, τ is the non-dimensional flowtime defined by the flow time t divided by the wave period. In both subfigures, the curves behave in anapproximately sinusoidal way. The variation of the measured wave height only depends on horizontaland vertical, i.e., the x and z direction, mesh resolution. Therefore, convergence of the mesh resolutionin horizontal and vertical directions is investigated with measured wave height as the dependentvariable. On this basis, when the Duck is embedded into the NWT, the variation of the pitch angulardisplacement amplitude θ5 only depends on the longitudinal, i.e.,the y direction, mesh resolutionand the time step resolution. Hence convergence of the longitudinal mesh resolution and the timestep resolution will be studied with pitch angular displacement amplitude as the dependent variable.Since small vibrations of both H and θ5 are observed due to wave reflection, the two dependentvariables are averaged over three wave periods for comparison. In Figure 6b, the surge and heavehydrodynamic forces represent the forces applied by the water in the x and z direction, respectively.

Energies 2019, 12, x FOR PEER REVIEW 7 of 19

whereθ5 and θ5are the dimensional and non-dimensional pitch angular displacement amplitude respectively, X1,3 and X1,3are the dimensional and non-dimensional surge and heave hydrodynamic force amplitude in the x and z direction respectively, ρ is the water density, g is the gravity acceleration, H is the wave height, and Lis the characteristic length [31].

5. Results and Discussion

In this section, the convergence of the CFD model is first investigated. Then, the CFD model is validated by comparing the predicted hydrodynamic responses with both experimental and BEM results. Finally, the hydrodynamic nonlinearity identified by the CFD model and the hydrodynamic force distribution along the longitudinal direction of the Duck are studied.

5.1. Convergence Study of the CFD Model

Before the CFD model is used to perform hydrodynamic simulations of the solo Duck WEC, it should be both mesh and time step resolution independent. Figures6a,b show the measured free surface elevation h at the Duck position with the Duck be removed from the NWT, and hydrodynamic responses of the Duck at wave period T = 1.6 s and wave height H = 0.075 m. Here,τ is the non-dimensional flow time defined by the flow time t divided by the wave period. In both subfigures, the curves behave in an approximately sinusoidal way. The variation of the measured wave height only depends on horizontal and vertical, i.e., the x and z direction, mesh resolution. Therefore, convergence of the mesh resolution in horizontal and vertical directions is investigated with measured wave height as the dependent variable. On this basis, when the Duck is embedded into the NWT, the variation of the pitch angular displacement amplitude θ5 only depends on the longitudinal, i.e.,the y direction, mesh resolution and the time step resolution. Hence convergence of the longitudinal mesh resolution and the time step resolution will be studied with pitch angular displacement amplitude as the dependent variable. Since small vibrations of both H and θ5 are observed due to wave reflection, the two dependent variables are averaged over three wave periods for comparison. In Figure6b, the surge and heave hydrodynamic forces represent the forces applied by the water in thex and z direction, respectively.

(a) (b)

Figure 6. Measured free surface elevation at the Duck position with the Duck be removed from the NWT and hydrodynamic responses of the Duck at T = 1.6 s and H = 0.075 m: (a) measured free surface elevation and pitch angular displacement of the Duck; (b) surge and heave hydrodynamic forces applied on the Duck.

Figure7 shows the convergence study results at T = 1.6 s and H = 0.075 m. Figure7a shows the measured wave height as a function of the cell number per wavelength and per wave height. The wave height tends to be stable when the cell number per wavelength and per wave height exceeds 30 and 10, respectively. Figure7b shows the pitch angular displacement amplitude as a function of the cell number in the longitudinal direction and per wave period. Also, it reveals that when the cell number in the longitudinal direction and per wave period exceeds 32 and 360, respectively, the

4.75 5.25 5.75 6.25 6.75 7.25 7.75-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

0.06

0.08

h, m

τ

4.75 5.25 5.75 6.25 6.75 7.25 7.75-8

-6

-4

-2

0

2

4

6

8

θ 5, deg

θ5

η

5 5.5 6 6.5 7 7.5-150

-100

-50

0

50

100

150

τ

For

ce, N

Surge forceHeave force

Figure 6. Measured free surface elevation at the Duck position with the Duck be removed from theNWT and hydrodynamic responses of the Duck at T = 1.6 s and H = 0.075 m: (a) measured free surfaceelevation and pitch angular displacement of the Duck; (b) surge and heave hydrodynamic forcesapplied on the Duck.

Energies 2019, 12, 1941 8 of 19

Figure 7 shows the convergence study results at T = 1.6 s and H = 0.075 m. Figure 7a showsthe measured wave height as a function of the cell number per wavelength and per wave height.The wave height tends to be stable when the cell number per wavelength and per wave height exceeds30 and 10, respectively. Figure 7b shows the pitch angular displacement amplitude as a function ofthe cell number in the longitudinal direction and per wave period. Also, it reveals that when the cellnumber in the longitudinal direction and per wave period exceeds 32 and 360, respectively, the angulardisplacement amplitude tends to converge. Therefore, for a tradeoff betweenthe computational costandthe accuracy, for later simulations in this paper, the mesh resolution and time step are set according tothe critical cell number stated above.

Energies 2019, 12, x FOR PEER REVIEW 8 of 19

angular displacement amplitude tends to converge. Therefore, for a tradeoff betweenthe computational costand the accuracy, for later simulations in this paper, the mesh resolution and time step are set according to the critical cell number stated above.

(a) (b)

Figure 7. Convergence study results at T = 1.6 s, and H = 0.075 m: (a) measured wave height as a function of the cell number per wavelength and per wave height; (b) pitch angular displacement amplitude as a function of the cell number in the longitudinal direction and per wave period.

5.2. Comparison in Different Wave Heights

Table 2 shows the test cases for the comparison in different wave heights. Figure8 shows the comparison of the pitch angular displacement amplitude as a function of wave height among CFD, BEM and experimental results at T = 1.6 s. In Table 2, the difference of the wave height between the CFD and experimental model is due to the measured wave height at the position of the Duck being different from the set value at the wave maker. For example, if the set value of the wave height at the wave maker is 1.0 cm, the measured value of the wave height at the position of the Duck may be smaller than 1.0 cm. The reason behind this phenomenon is that the wave energy is dissipated during the propagation process. Since the dissipation rate of the wave energy may be different between the CFD and experimental model, the measured wave height at the position of the Duck may also be different, e.g.,0.9 cm for the CFD model and 0.8 cm for the experimental model. Therefore, the expectation of realizing an identical value of the wave height at the position of the Duck for both the CFD and experimental models cannot be easily achieved, but needs time-consuming tuning of the set value of the wave height at the wave maker for both models. Besides, since the wave height is served as the independent variable in Figure 8, whether or not the wave height in the CFD and experimental model is the same does not influence the variation tendency of the dependent variable. Therefore, the wave height between the CFD and experimental model needs not to be the same. The maximum of the wave height in the CFD model is larger than that of the experimental model due to capsizing of the Duck. Due to friction, the motion response of the Duck in the experimental model is slightly smaller than that in the CFD model. Therefore, the Duck in the experimental model is harder to capsize. Meanwhile, the maximum wave height in both models is the minimum wave height that induces the Duck to capsize. Therefore, the maximum of the wave height in the experimental model is larger than the CFD model.

For all three models, the pitch angular displacement amplitude grows with wave height. CFD resultsagree well with BEM results at small wave heights, and presents almost linear relationship with wave height. However, at large wave heightswhen H≥ 0.12 m, difference emerges since the CFD curve shows a slower growing trend while the BEM one shows the same slope as at small wave heights. Actually, this can be explained by the hydrodynamic nonlinearityinduced at large wave heights. In [32], it is found that BEM models without the nonlinear term of drag force over-estimate the response of the WEC when compared to CFD models. For cases with high wave heights, the influence of the drag force will be more prominent since the magnitude of the drag force increases with the square of velocity. Compared to experimental results, the CFD curve

20 30 40 50 600.072

0.073

0.074

0.075

0.076

0.077

Cell number per wavelength

5 10 15 20

0.072

0.073

0.074

0.075

0.076

0.077

Cell number per wave height

H, m

Horizontal directionVertical direction

102.3 102.6 102.9

4.5

4.7

4.9

5.1

5.3

Cell number per wave period

0 10 20 30 40

4.5

4.7

4.9

5.1

5.3

Cell number in longitudinal direction

Θ5, d

eg

Time stepLongitudinal direction

Figure 7. Convergence study results at T = 1.6 s, and H = 0.075 m: (a) measured wave height asa function of the cell number per wavelength and per wave height; (b) pitch angular displacementamplitude as a function of the cell number in the longitudinal direction and per wave period.

5.2. Comparison in Different Wave Heights

Table 2 shows the test cases for the comparison in different wave heights. Figure 8 shows thecomparison of the pitch angular displacement amplitude as a function of wave height among CFD,BEM and experimental results at T = 1.6 s. In Table 2, the difference of the wave height between theCFD and experimental model is due to the measured wave height at the position of the Duck beingdifferent from the set value at the wave maker. For example, if the set value of the wave height at thewave maker is 1.0 cm, the measured value of the wave height at the position of the Duck may be smallerthan 1.0 cm. The reason behind this phenomenon is that the wave energy is dissipated during thepropagation process. Since the dissipation rate of the wave energy may be different between the CFDand experimental model, the measured wave height at the position of the Duck may also be different,e.g., 0.9 cm for the CFD model and 0.8 cm for the experimental model. Therefore, the expectationof realizing an identical value of the wave height at the position of the Duck for both the CFD andexperimental models cannot be easily achieved, but needs time-consuming tuning of the set value ofthe wave height at the wave maker for both models. Besides, since the wave height is served as theindependent variable in Figure 8, whether or not the wave height in the CFD and experimental modelis the same does not influence the variation tendency of the dependent variable. Therefore, the waveheight between the CFD and experimental model needs not to be the same. The maximum of the waveheight in the CFD model is larger than that of the experimental model due to capsizing of the Duck.Due to friction, the motion response of the Duck in the experimental model is slightly smaller than thatin the CFD model. Therefore, the Duck in the experimental model is harder to capsize. Meanwhile, themaximum wave height in both models is the minimum wave height that induces the Duck to capsize.Therefore, the maximum of the wave height in the experimental model is larger than the CFD model.

Energies 2019, 12, 1941 9 of 19

Table 2. Test cases for the comparison in different wave heights.

Tested Cases 1 2 3 4 5 6 7 8 9 10

Waveheight, cm

CFD model 1.79 3.70 5.61 7.51 9.41 11.30 13.20 15.10 17.00 18.91Experimental model 4.49 7.02 9.50 12.02 14.50 17.03 19.48 21.97 – –

Wave period, s 1.6

Energies 2019, 12, x FOR PEER REVIEW 9 of 19

showsthe same tendency as the experimental one, but predicts an overall larger value, especially at small wave heights. A possible explanation for thesephenomena is the friction loss in the mechanical transmission system. Actually, friction is very complex and its formation mechanism has still not been clearly revealed yet. Frictionalso varies nonlinearly with a lot of parameters, such as the velocity. For a Duck WEC, since the Duck oscillates periodically, the velocity of the Duck varies all the time, thus a fixed value of the friction cannot be obtained.Han et al. [33] tried to take friction into consideration by tuning the value of the friction inside the numerical model until the response of the numerical model equals that measured in the experimental model. However, for the comparison betweenthe numerical and the experimental model in this paper, the friction cannot be determined ahead. Anyway, although there are differences betweenthe numerical and experimental model, the difference is not significant overall, and experimental results can still quantitatively validate the numerical models to some extent.

Table 2.Test cases for the comparison in different wave heights.

Tested Cases 1 2 3 4 5 6 7 8 9 10

Wave height, cm

CFD model

1.79 3.70 5.61 7.51 9.41 11.30 13.20 15.10 17.00 18.91

Experimental model

4.49 7.02 9.50 12.02 14.50 17.03 19.48 21.97 – –

Wave period, s 1.6

Figure 8. Comparison of pitch angular displacement amplitude as a function of wave height among CFD, BEM and experimental results at T = 1.6 s.

Figure9 shows the comparison of the hydrodynamic force amplitude as a function of wave height at T = 1.6 s. Both surge and heave forces grow almost linearly with wave height. Similar to the variation tendency of pitch angular displacement amplitude with wave height, CFD results agree well BEM results at small wave heights, while diverges from BEM results at large wave heights with the heave force as an exception. Compared to experimental results, CFD results show only small difference in the whole wave height range. The main reason is that the friction in the mechanical transmission system does not influence the pressure distribution around the Duck. Therefore, the hydrodynamic forces are not influenced by the friction.

0 0.05 0.1 0.15 0.2 0.250

5

10

15

H, m

Θ5, d

eg

CFDBEMExp.

Figure 8. Comparison of pitch angular displacement amplitude as a function of wave height amongCFD, BEM and experimental results at T = 1.6 s.

For all three models, the pitch angular displacement amplitude grows with wave height.CFD results agree well with BEM results at small wave heights, and presents almost linear relationshipwith wave height. However, at large wave heightswhen H ≥ 0.12 m, difference emerges since the CFDcurve shows a slower growing trend while the BEM one shows the same slope as at small wave heights.Actually, this can be explained by the hydrodynamic nonlinearityinduced at large wave heights. In [32],it is found that BEM models without the nonlinear term of drag force over-estimate the response of theWEC when compared to CFD models. For cases with high wave heights, the influence of the drag forcewill be more prominent since the magnitude of the drag force increases with the square of velocity.Compared to experimental results, the CFD curve shows the same tendency as the experimental one,but predicts an overall larger value, especially at small wave heights. A possible explanation forthese phenomena is the friction loss in the mechanical transmission system. Actually, friction is verycomplex and its formation mechanism has still not been clearly revealed yet. Friction also variesnonlinearly with a lot of parameters, such as the velocity. For a Duck WEC, since the Duck oscillatesperiodically, the velocity of the Duck varies all the time, thus a fixed value of the friction cannot beobtained. Han et al. [33] tried to take friction into consideration by tuning the value of the frictioninside the numerical model until the response of the numerical model equals that measured in theexperimental model. However, for the comparison between the numerical and the experimental modelin this paper, the friction cannot be determined ahead. Anyway, although there are differences betweenthe numerical and experimental model, the difference is not significant overall, and experimentalresults can still quantitatively validate the numerical models to some extent.

Figure 9 shows the comparison of the hydrodynamic force amplitude as a function of wave heightat T = 1.6 s. Both surge and heave forces grow almost linearly with wave height. Similar to the variationtendency of pitch angular displacement amplitude with wave height, CFD results agree well BEMresults at small wave heights, while diverges from BEM results at large wave heights with the heaveforce as an exception. Compared to experimental results, CFD results show only small difference in thewhole wave height range. The main reason is that the friction in the mechanical transmission systemdoes not influence the pressure distribution around the Duck. Therefore, the hydrodynamic forces arenot influenced by the friction.

Energies 2019, 12, 1941 10 of 19Energies 2019, 12, x FOR PEER REVIEW 10 of 19

(a) (b)

Figure 9. Comparison of the hydrodynamic force amplitude as a function of wave height at T = 1.6 s: (a) surge force; (b) heave force.

5.3. Comparison in Different Wave Periods

In this section, the comparison is performed in a range of wave periods, from 0.6 s to 2.0 s. Table 3 shows the test cases for the comparison in different wave periods. Figure10 shows the comparison of pitch angular displacement amplitude as a function of wave period among CFD, BEM and experimental results at H = 0.075 m. For all three models, the angular displacement amplitude grows with wave period. CFD results agree well and almost coincides with BEM results when T≤ 1.4 s. When T> 1.4 s, CFD results are slightly smaller than BEM results. The difference is expected to be resulted from the viscosity that damps the response of the Duck in the CFD model. The angular displacement amplitude predicted by the CFD model is larger than experimental results a most wave periods except at T = 0.6 s and the discrepancy grows with wave period.

Table 3.Test cases for the comparison in different wave periods.

Tested Cases 1 2 3 4 5 6 7 8 Wave Period, s 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

Wave height, cm 7.5

Figure 10. Comparison of pitch angular displacement amplitude as a function of wave period between CFD, BEM and experimental results at H = 0.075 m.

Figure11 shows the comparison of the hydrodynamic force amplitude as a function of wave period at H = 0.075 m. In Figure11a, the surge force amplitude peaks at T = 1.4 s and decreases gradually on either side for all three models. The surge force amplitude of CFD resultsshow little difference from experimental resultswithin the wave period range from 0.8 s to 1.4 s, while is smaller than experimental results when T ≤ 0.8 s, and larger than them when T ≥ 1.4 s.

0 0.05 0.1 0.15 0.2 0.250

50

100

150

200

250

300

H, m

Sur

ge f

orce

am

plit

ude,

N

CFDBEMExp.

0 0.05 0.1 0.15 0.2 0.250

50

100

150

200

250

300

H, m

Hea

ve f

orce

am

plit

ude,

N

CFDBEMExp.

0.5 1 1.5 20

1

2

3

4

5

6

T, s

Θ5, d

eg

CFDBEMExp.

Figure 9. Comparison of the hydrodynamic force amplitude as a function of wave height at T = 1.6 s:(a) surge force; (b) heave force.

5.3. Comparison in Different Wave Periods

In this section, the comparison is performed in a range of wave periods, from 0.6 s to 2.0 s. Table 3shows the test cases for the comparison in different wave periods. Figure 10 shows the comparison ofpitch angular displacement amplitude as a function of wave period among CFD, BEM and experimentalresults at H = 0.075 m. For all three models, the angular displacement amplitude grows with waveperiod. CFD results agree well and almost coincides with BEM results when T ≤ 1.4 s. When T > 1.4 s,CFD results are slightly smaller than BEM results. The difference is expected to be resulted from theviscosity that damps the response of the Duck in the CFD model. The angular displacement amplitudepredicted by the CFD model is larger than experimental results a most wave periods except at T = 0.6 sand the discrepancy grows with wave period.

Table 3. Test cases for the comparison in different wave periods.

Tested Cases 1 2 3 4 5 6 7 8

Wave Period, s 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0Wave height, cm 7.5

Energies 2019, 12, x FOR PEER REVIEW 10 of 19

(a) (b)

Figure 9. Comparison of the hydrodynamic force amplitude as a function of wave height at T = 1.6 s: (a) surge force; (b) heave force.

5.3. Comparison in Different Wave Periods

In this section, the comparison is performed in a range of wave periods, from 0.6 s to 2.0 s. Table 3 shows the test cases for the comparison in different wave periods. Figure10 shows the comparison of pitch angular displacement amplitude as a function of wave period among CFD, BEM and experimental results at H = 0.075 m. For all three models, the angular displacement amplitude grows with wave period. CFD results agree well and almost coincides with BEM results when T≤ 1.4 s. When T> 1.4 s, CFD results are slightly smaller than BEM results. The difference is expected to be resulted from the viscosity that damps the response of the Duck in the CFD model. The angular displacement amplitude predicted by the CFD model is larger than experimental results a most wave periods except at T = 0.6 s and the discrepancy grows with wave period.

Table 3.Test cases for the comparison in different wave periods.

Tested Cases 1 2 3 4 5 6 7 8 Wave Period, s 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

Wave height, cm 7.5

Figure 10. Comparison of pitch angular displacement amplitude as a function of wave period between CFD, BEM and experimental results at H = 0.075 m.

Figure11 shows the comparison of the hydrodynamic force amplitude as a function of wave period at H = 0.075 m. In Figure11a, the surge force amplitude peaks at T = 1.4 s and decreases gradually on either side for all three models. The surge force amplitude of CFD resultsshow little difference from experimental resultswithin the wave period range from 0.8 s to 1.4 s, while is smaller than experimental results when T ≤ 0.8 s, and larger than them when T ≥ 1.4 s.

0 0.05 0.1 0.15 0.2 0.250

50

100

150

200

250

300

H, m

Sur

ge f

orce

am

plit

ude,

N

CFDBEMExp.

0 0.05 0.1 0.15 0.2 0.250

50

100

150

200

250

300

H, m

Hea

ve f

orce

am

plit

ude,

N

CFDBEMExp.

0.5 1 1.5 20

1

2

3

4

5

6

T, s

Θ5, d

eg

CFDBEMExp.

Figure 10. Comparison of pitch angular displacement amplitude as a function of wave period betweenCFD, BEM and experimental results at H = 0.075 m.

Figure 11 shows the comparison of the hydrodynamic force amplitude as a function of wave periodat H = 0.075 m. In Figure 11a, the surge force amplitude peaks at T = 1.4 s and decreases gradually oneither side for all three models. The surge force amplitude of CFD resultsshow little difference fromexperimental resultswithin the wave period range from 0.8 s to 1.4 s, while is smaller than experimentalresults when T ≤ 0.8 s, and larger than them when T ≥ 1.4 s. Compared to BEM results, the force

Energies 2019, 12, 1941 11 of 19

predicted by the CFD model is slightly larger in the whole wave period region. In Figure 11b, theheave force amplitude grows with wave period for all three models. In fact, these amplitudes equalthe surge and heave excitation force since the WEC is static in both degrees of freedom. The shapeof the curves is characterized by either a bell-like shape with the peak value appears at the waveperiod of 1.4 s, or by a nearly monotonically varying trend. The data that forms the shape can beapproximately obtained by solving the Laplace equation, with the boundary condition defined bythe geometry of the WEC. Therefore, the shape of the curves, no matter it is the bell-like form for thesurge mode or the nearly straight form for the heave mode is influenced mostly by the geometry of theWEC. The difference between CFD and experimental results is smaller than that between BEM andexperimental results. One possible reason for the discrepancy between CFD and experimental resultsmay be the installation error of the mechanical equipments, which may cause the measured surge andheave excitation forces be coupled with the excitation force of other degrees of freedom. The complexcoupling effect may cause the resultant measured excitation force from the experiment be larger thanthat of the numerical models. From above comparison results, it is revealed that the CFD model canbetter predict the hydrodynamic response of the Duck than the BEM model for this experimental setup.

Energies 2019, 12, x FOR PEER REVIEW 11 of 19

Compared to BEM results, the force predicted by the CFD model is slightly larger in the whole wave period region. In Figure11b, the heave force amplitude grows with wave period for all three models. In fact, these amplitudes equal the surge and heave excitation force since the WEC is static in both degrees of freedom. The shape of the curves is characterized by either a bell-like shape with the peak value appears at the wave period of 1.4s, or by a nearly monotonically varying trend. The data that forms the shape can be approximately obtained by solving the Laplace equation, with the boundary condition defined by the geometry of the WEC. Therefore, the shape of the curves, no matter it is the bell-like form for the surge mode or the nearly straight form for the heave mode is influenced mostly by the geometry of the WEC. The difference between CFD and experimental results is smaller than that between BEM and experimental results. One possible reason for the discrepancy between CFD and experimental results may be the installation error of the mechanical equipments, which may cause the measured surge and heave excitation forces be coupled with the excitation force of other degrees of freedom. The complex coupling effect may cause the resultant measured excitation force from the experiment be larger than that of the numerical models. From above comparison results, it is revealed that the CFD model can better predict the hydrodynamic response of the Duck than the BEM model for this experimental setup.

(a) (b)

Figure 11. Comparison of the hydrodynamic force amplitude as a function of wave period at H = 0.075 m: (a) surge force; (b) heave force.

5.4. Influence of Wave Steepness

From the comparison performed in Sections 5.2 and 5.3, we find that the difference of the predicted hydrodynamic responses between the CFD and BEM models appears at different wave heights and wave periods.In order to quantitatively describe the difference of predicted results between the CFD and BEM model, the concept of relative difference is introduced as:

CFD BEM

CFD

φ φν

φ−

= , (6)

whereφCFD denotes results from the CFD model and φBEM for the BEM model. Relative difference is defined to measure to what extent the two results differ from each other. Figure12 shows the relative difference of pitch angular displacement amplitude as a function ofwave height and wave period. It can be seen that the relative difference between the CFD and BEM model generally increases with wave height, but decreases with wave period.

Since the BEM model is linearised from the CFD model, we can attribute the above difference to the hydrodynamic nonlinearfactors that are taken into consideration in the CFD model, including the nonlinearity of the wave, i.e., the nonlinear waveform, and the nonlinearity of the Duck motion, i.e., the vortex generation process. High nonlinearity of the wave means relatively large wave heights, resulting in relatively large motion response of the Duck, i.e. high nonlinearity of the Duck motion. Therefore, the nonlinearity of the Duck motion is positively correlated to the nonlinearity of the wave.Therefore, the nonlinearity of the wave can be seen as an index to represent the total

0.5 1 1.5 220

30

40

50

60

70

80

90

T, s

Sur

ge f

orce

am

plit

ude,

N

CFDBEMExp.

0.5 1 1.5 20

30

60

90

120

150

T, s

Hea

ve f

orce

am

plit

ude,

N

CFDBEMExp.

Figure 11. Comparison of the hydrodynamic force amplitude as a function of wave period atH = 0.075 m: (a) surge force; (b) heave force.

5.4. Influence of Wave Steepness

From the comparison performed in Sections 5.2 and 5.3, we find that the difference of the predictedhydrodynamic responses between the CFD and BEM models appears at different wave heights andwave periods.In order to quantitatively describe the difference of predicted results between the CFDand BEM model, the concept of relative difference is introduced as:

ν =ϕCFD −ϕBEM

ϕCFD, (6)

where ϕCFD denotes results from the CFD model and ϕBEM for the BEM model. Relative difference isdefined to measure to what extent the two results differ from each other. Figure 12 shows the relativedifference of pitch angular displacement amplitude as a function ofwave height and wave period.It can be seen that the relative difference between the CFD and BEM model generally increases withwave height, but decreases with wave period.

Since the BEM model is linearised from the CFD model, we can attribute the above difference tothe hydrodynamic nonlinear factors that are taken into consideration in the CFD model, including thenonlinearity of the wave, i.e., the nonlinear waveform, and the nonlinearity of the Duck motion, i.e.,the vortex generation process. High nonlinearity of the wave means relatively large wave heights,resulting in relatively large motion response of the Duck, i.e., high nonlinearity of the Duck motion.Therefore, the nonlinearity of the Duck motion is positively correlated to the nonlinearity of the wave.

Energies 2019, 12, 1941 12 of 19

Therefore, the nonlinearity of the wave can be seen as an index to represent the total nonlinearity.Normally, the nonlinearity of the wave can be measured by the wave steepness defined as:

k =Hλ

. (7)

The wave steepness is also shown in Figure 12 as a set of contour lines. It can be clearly seenthat the difference betweenCFD and BEM results grows with wave steepness. Figure 13 shows thenon-dimensional free surface elevation, η = 2h/H, where h denotes the free surface elevation, and Hdenotes the wave height, as a function of the non-dimensional flow time for different wave steepnessat T = 1.6 s and H = 0.16 m, respectively. When looking into Figure 13, it can be found that: withthe increase of wave steepness, the wave crest becomes shaper and the wave trough becomes flatter.This process is similar to the transient process from the Airy wave to the second order Stokes wave,indicating that the nonlinearity of the wave increases with wave steepness.With the increase ofwavenonlinearity, the drift torque drives the Duck to capsize when the wave condition exceed a critical value.In this situation, since the Duck departures from its original equilibrium position, the comparisonof the pitch angular displacement amplitude of the Duck between the CFD and BEM model will bemeaningless. Therefore, a dashed line is shown in Figure 12 to distinguish the useful data in the leftpart from the useless date in the right part.

Energies 2019, 12, x FOR PEER REVIEW 12 of 19

nonlinearity.Normally, the nonlinearity of the wave can be measured by the wave steepness defined as:

Hkλ

= . (7)

The wave steepness is also shown in Figure 12 as a set of contour lines. It can be clearly seen that the difference betweenCFD and BEM results grows with wave steepness. Figure13 shows the non-dimensional free surface elevation, η = 2h/H, where h denotes the free surface elevation, and H denotes the wave height, as a function of the non-dimensional flow time for different wave steepness at T = 1.6 s and H = 0.16 m, respectively. When looking into Figure 13, it can be found that: with the increase of wave steepness, the wave crest becomes shaper and the wave trough becomes flatter. This process is similar to the transient process from the Airy wave to the second order Stokes wave, indicating that the nonlinearity of the wave increases with wave steepness.With the increase ofwave nonlinearity, the drift torque drives the Duck to capsize when the wave condition exceed a critical value. In this situation, since the Duck departures from its original equilibrium position, the comparison of the pitch angular displacement amplitude of the Duck between the CFD and BEM model will be meaningless. Therefore, a dashed line is shown in Figure 12 to distinguish the useful data in the left part from the useless date in the right part.

Figure 12. Relative difference of the pitch angular displacement amplitude as a function ofwave height and wave period. Contour lines of the wave steepness are shown by solid lines, while the critical capsizing condition for the Duck is shown by the dashed line.

(a) (b)

Figure 13. Non-dimensional free surface elevation as a function of the non-dimensional flow time at different wave steepness: (a) at T = 1.6 s; (b) at H = 0.16 m.

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

1

1.2

1.4

1.6

1.8

2

H, m

T, s

0.01

0.01

0.01

0.03

0.03

0.03

0.03

0.05

0.05

0.05

0.07

0.07

0.09

0.09

0.11

Critica

l cap

sizing

cond

ition

-0.2

0

0.2

0.4

0.6

0.8

6.4 6.6 6.8 7 7.2-1.5

-1

-0.5

0

0.5

1

1.5

τ

η

TheoryH =0.04 m, k =0.010H =0.12 m, k =0.030H =0.20 m, k =0.050

6.4 6.6 6.8 7 7.2-1.5

-1

-0.5

0

0.5

1

1.5

τ

η

TheoryT =2.0 s, k =0.026T =1.6 s, k =0.040T =1.0 s, k =0.103

Figure 12. Relative difference of the pitch angular displacement amplitude as a function ofwave heightand wave period. Contour lines of the wave steepness are shown by solid lines, while the criticalcapsizing condition for the Duck is shown by the dashed line.

The paragraphs above attributed the difference between the CFD and BEM model to hydrodynamicnonlinear factors, and the reason can be explained as follows. The BEM model is based on the LaplaceEquation, which is originated from the Navier-Stokes Equation with the assumptions of inviscid fluid,incompressible and irrotational flow. To solve the Laplace Equation, the linear wave theory, whichassumes that both the wave steepness and motion response of the WEC are small, is employed. On theother hand, the CFD model used in this work is based on the RANS Equation, which is also originatedfrom the Navier-Stokes Equation but with the turbulence effect be approximated by empirical models.Although both two models adopt approximations, the CFD model is closer to real flow conditions thanthe BEM model since the fluid can be viscid, and the flow can be compressible and rotational. With theincrease of wave steepness, large wave steepness may violate basic assumptions of the linear wavetheory, resulting in the BEM model loss the ability to accurately predict the hydrodynamic response ofthe WEC. Therefore, the CFD model is expected to be more accurate than the BEM model in this case.

Energies 2019, 12, 1941 13 of 19

Energies 2019, 12, x FOR PEER REVIEW 12 of 19

nonlinearity.Normally, the nonlinearity of the wave can be measured by the wave steepness defined as:

Hkλ

= . (7)

The wave steepness is also shown in Figure 12 as a set of contour lines. It can be clearly seen that the difference betweenCFD and BEM results grows with wave steepness. Figure13 shows the non-dimensional free surface elevation, η = 2h/H, where h denotes the free surface elevation, and H denotes the wave height, as a function of the non-dimensional flow time for different wave steepness at T = 1.6 s and H = 0.16 m, respectively. When looking into Figure 13, it can be found that: with the increase of wave steepness, the wave crest becomes shaper and the wave trough becomes flatter. This process is similar to the transient process from the Airy wave to the second order Stokes wave, indicating that the nonlinearity of the wave increases with wave steepness.With the increase ofwave nonlinearity, the drift torque drives the Duck to capsize when the wave condition exceed a critical value. In this situation, since the Duck departures from its original equilibrium position, the comparison of the pitch angular displacement amplitude of the Duck between the CFD and BEM model will be meaningless. Therefore, a dashed line is shown in Figure 12 to distinguish the useful data in the left part from the useless date in the right part.

Figure 12. Relative difference of the pitch angular displacement amplitude as a function ofwave height and wave period. Contour lines of the wave steepness are shown by solid lines, while the critical capsizing condition for the Duck is shown by the dashed line.

(a) (b)

Figure 13. Non-dimensional free surface elevation as a function of the non-dimensional flow time at different wave steepness: (a) at T = 1.6 s; (b) at H = 0.16 m.

0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2

1

1.2

1.4

1.6

1.8

2

H, m

T, s

0.01

0.01

0.01

0.03

0.03

0.03

0.03

0.05

0.05

0.05

0.07

0.07

0.09

0.09

0.11

Critica

l cap

sizing

cond

ition

-0.2

0

0.2

0.4

0.6

0.8

6.4 6.6 6.8 7 7.2-1.5

-1

-0.5

0

0.5

1

1.5

τ

η

TheoryH =0.04 m, k =0.010H =0.12 m, k =0.030H =0.20 m, k =0.050

6.4 6.6 6.8 7 7.2-1.5

-1

-0.5

0

0.5

1

1.5

τ

η

TheoryT =2.0 s, k =0.026T =1.6 s, k =0.040T =1.0 s, k =0.103

Figure 13. Non-dimensional free surface elevation as a function of the non-dimensional flow time atdifferent wave steepness: (a) at T = 1.6 s; (b) at H = 0.16 m.

5.5. Vortex Generation Processat the End of the Duck

In the above section, we inferred that one of the hydrodynamic nonlinear factors is from theDuck motion, which is mainly embodied by the vortex generation phenomenon around the Duck,especially at the end the Duck. In this section, we try to study the vortex generation process around theWEC using the validated CFD model. For the solo Duck WEC, the vortex generation process is a newchallenge since it does not appear for the spine-connected Duck since there are almost no gaps betweenthe Ducks. The pressure difference between the blocked flow in the front area of the Duck and freeflow at the end causes vortexes to be generated, which is also very common in aerodynamic designof airplane wings [34] and wind turbine blades [35], where blade tip losses are caused by pressuredifferences. The flow passing the end of a Duck is similar to a steady flow passing an obstacle withleading and trailing rectangular corners, as shown in Figure 14. In the upstream, the blockage of theobstacle causes a vertical velocity component of the flow in the front area of the obstacle. At the leadingcorner, the sharply deceleration of the vertical velocity in the vertical wall boundary layer causes anadvert pressure gradient leading to flow separation. Under the compression of the main stream, theseparation zone is restricted to a limited zone, and reattachment of the boundary layer happens whenthe horizontal wall is suitably long. At the trailing corner, the same phenomenon appears since thehorizontal boundary layer reaches another corner, and this will cause another flow separation. In boththe upstream and downstream positions, standing vortexes are generated and denoted as P and Q,respectively. Because a rectangular angle is the infinite approximation of a sufficiently large curvature,the leading and trailing corners act as the fixed flow separation points for the geometry. These vortexgeneration phenomena are extensively studied in [36–39]. The direction of P and Q is perpendicular toplane of the main flow with the polarity be determined so that the peripheral velocity of the vortexfollows the main flow.

Energies 2019, 12, x FOR PEER REVIEW 13 of 19

The paragraphs above attributed the difference between the CFD and BEM model to hydrodynamic nonlinear factors, and the reason can be explained as follows. The BEM model is based on the Laplace Equation, which is originated from the Navier-Stokes Equation with the assumptions of inviscid fluid, incompressible and irrotational flow. To solve the Laplace Equation, the linear wave theory, which assumes that both the wave steepness and motion response of the WEC are small, is employed. On the other hand, the CFD model used in this work is based on the RANS Equation, which is also originated from the Navier-Stokes Equation but with the turbulence effect be approximated by empirical models. Although both two models adopt approximations, the CFD model is closer to real flow conditions than the BEM model since the fluid can be viscid, and the flow can be compressible and rotational. With the increase of wave steepness, large wave steepness may violate basic assumptions of the linear wave theory, resulting in the BEM model loss the ability to accurately predict the hydrodynamic response of the WEC. Therefore, the CFD model is expected to be more accurate than the BEM model in this case.

5.5. Vortex Generation Processat the End of the Duck

In the above section, we inferred that one of the hydrodynamic nonlinear factors is from the Duck motion, which is mainly embodied by the vortex generation phenomenon around the Duck, especially at the end the Duck. In this section, we try to study the vortex generation process around the WEC using the validated CFD model. For the solo Duck WEC, the vortex generation process is a new challenge since it does not appear for the spine-connected Duck since there are almost no gaps between the Ducks. The pressure difference between the blocked flow in the front area of the Duck and free flow at the end causes vortexes to be generated, which is also very common in aerodynamic design of airplane wings [34] and wind turbine blades [35], where blade tip losses are caused by pressure differences. The flow passing the end of a Duck is similar to a steady flow passing an obstacle with leading and trailing rectangular corners, as shown in Figure14. In the upstream, the blockage of the obstacle causes a vertical velocity component of the flow in the front area of the obstacle. At the leading corner, the sharply deceleration of the vertical velocity in the vertical wall boundary layer causes an advert pressure gradient leading to flow separation. Under the compression of the main stream, the separation zone is restricted to a limited zone, and reattachment of the boundary layer happens when the horizontal wall is suitably long. At the trailing corner, the same phenomenon appears since the horizontal boundary layer reaches another corner, and this will cause another flow separation. In both the upstream and downstream positions, standing vortexes are generated and denoted as P and Q, respectively. Because a rectangular angle is the infinite approximation of a sufficiently large curvature, the leading and trailing corners act as the fixed flow separation points for the geometry. These vortex generation phenomenaare extensively studied in [36–39]. The direction of P and Q is perpendicular to plane of the main flow with the polarity be determined so that the peripheral velocity of the vortex follows the main flow.

Figure 14. A steady flow passing an obstacle with leading and trailing rectangular corners, and standing vortexes,P and Q, are generated as a result of the boundary layer separation.

Figure15 shows the fluid velocity and vorticity distribution around the end of the Duck at T = 1.6 s and H = 0.12 m. In the left column, the subfigures show the approximately undisturbed fluid velocity vector distribution in the plane y= 2.4 m. The hollow circle represents the rotation axis of the pitch motion. Within a wave period, as shown in the linear wave theory, the velocity direction of water particles at a fixed point is of circular type, as shown in Figure 15i and the time sequence follows the clockwise principle as A→B→C→D→A. In the right column, vorticitydistributionof the main vortexes are shown in either the free surface, i.e.,z = 0 plane, or the perpendicular surface,

P

Q

Figure 14. A steady flow passing an obstacle with leading and trailing rectangular corners, and standingvortexes, P and Q, are generated as a result of the boundary layer separation.

Figure 15 shows the fluid velocity and vorticity distribution around the end of the Duck at T = 1.6 sand H = 0.12 m. In the left column, the subfigures show the approximately undisturbed fluid velocityvector distribution in the plane y = 2.4 m. The hollow circle represents the rotation axis of the pitchmotion. Within a wave period, as shown in the linear wave theory, the velocity direction of water

Energies 2019, 12, 1941 14 of 19

particles at a fixed point is of circular type, as shown in Figure 15i and the time sequence followsthe clockwise principle as A→B→C→D→A. In the right column, vorticitydistributionof the mainvortexes are shown in either the free surface, i.e., z = 0 plane, or the perpendicular surface, i.e., x= 0plane. Since the peripheral velocity of the Duck is small relative to the fluid velocity, the Duck can beconsidered as standing still in the flow field. At time instance A, the main flow is in the x-y plane, i.e.,z = 0 plane. As illustrated in Figure 14, the main vortexes will be generated in the z direction. The Ptype vortex is generated at the leading cornerat the end, while the Q type vortex is at the trailing corner,and the direction of these vortexes are both positive so that the peripheral velocity of the vorticesfollows the main flow, and this is confirmed in Figure 15b.

At time instance B, the main flow is in the y-z plane, i.e., x = 0. Similar to the above principle, theP and Q vortexes are generated and both in the positive x direction as confirmed in Figure 15d. At timeinstance C, the situation is rightly opposite to time instance A. The location of P and Q vortexes arereversed, and both are in the negative z direction as shown in Figure 15f. At time instance D, bothlocation and direction of P and Q vortexes are reversed from time instance B as confirmed in Figure 15h.

From above discussions, we find that: at different time instances within a wave period, themagnitude and direction of the generated vortexes are different, and vary periodically. Vortexes aregenerated due to the relative motion between the WEC and the fluid, thus part of the kinetic energy ofthe WEC will be devoted to feed the vortex generation process, resulting in the captured power of theWEC be reduced.

The vortex generation process causes the static pressure in the downstream side of the obstaclebe smaller than that of the upstream side, resulting in a force resisting the relative motion of theobstacle, e.g., a resisting torque for the Duck in this paper. This resisting force is called the dragforce, whose magnitude is proportional to the square of the velocity and direction is opposite to thevelocity. Therefore, the nonlinearity of the Duck motion, which is embodied by the vortex generationphenomenon, can be quantitatively represented by the drag torque. Besides, since the analysis inSection 5.4 revealed that the nonlinearity of the wave is positively correlated to the nonlinearity ofthe Duck motion, we combine the two nonlinear factors together so that the total nonlinearity isquantitatively represented by a single drag torque. As in [40], the drag torque of the Duck in the pitchdegree-of-freedom can be defined as:

Td = −Cd.θ∣∣∣∣ .θ∣∣∣∣, (8)

where Cd is the drag torque coefficient. Then, according to the Newton’s Second Law, we can obtainthe equation of motion of the Duck as:

J..θ = Te + Tr + Th + Td, (9)

where Te is the excitation torque, Tr is the radiation torque, and Th is the hydrostatic torque. By solvingEquation (9), we can obtain the motion response of the Duck in the time domain considering the dragtorque. However, the drag torque coefficient in Equation (8) for a given wave condition cannot beobtained directly. In this paper, this drag torque coefficient is obtained in an indirect way as in [33], bytuning its value until the motion response of the Duck obtained from Equation (9) equals that of theCFD model.

Figure 16 shows the drag torque coefficient as a function of wave height at T = 1.6 s andc = 12 N·m·s/rad. At small wave heights, the drag torque coefficient is almost zero. However, at largewave heights, the drag torque coefficient increases significantly with wave height. This is due to thenonlinear behavior of the wave and the vortex generation phenomenon is more prominent at largewave heights. The observed variation tendency in Figure 16 indicates that the drag torque coefficienteffectively reflected the influence of the nonlinearity of the wave and the Duck motion. This givesthe opportunity to build a database of the drag torque coefficient as a function of wave height, waveperiod, etc, so that the hydrodynamic nonlinearity can be modeled with reasonable accuracy withoutthe time-consuming CFD model.

Energies 2019, 12, 1941 15 of 19

Energies 2019, 12, x FOR PEER REVIEW 14 of 19

i.e.,x= 0 plane. Since the peripheral velocity of the Duck is small relative to the fluid velocity, the Duck can be considered as standing still in the flow field. At time instance A, the main flow is in the x-y plane, i.e., z = 0 plane. As illustrated in Figure 14, the main vortexes will be generated in the z direction. The P type vortex is generated at the leading cornerat the end, while the Q type vortex is at the trailing corner, and the direction of these vortexes are both positive so that the peripheral velocity of the vortices follows the main flow, and this is confirmed in Figure 15b.

A

B

C

D

Figure 15. Fluid velocity and vorticity distribution at the end of the Duck at T = 1.6 s and H = 0.12 m. The subfigures in the left column show the approximately undisturbed fluid velocity vector distribution in the plane y = 2.4 m, with subfigures (a,c,e,g) for time instances A, B, C and D, respectively. The subfigures in the right column show the vorticity of the main vortex in either the free surface, i.e.,z = 0 plane, or the perpendicular surface, i.e., x = 0 plane, with subfigures (b,d,f,h)

z-vorticity

x-vorticity

z-vorticity

x-vorticity

(a) (b)

(c) (d)

(h)

(i)

0 0.4

Velocity magnitude

0.04 0.12 0.16 0.24 0.32 0.36

-10 -8 -6 -4 -2 0 2 4 6 8 10

-18.0 -14.4 -10.8 -7.2 -3.6 0 3.6 7.2 10.8 14.4 18.0

-10 -8 -6 -4 -2 0 2 4 6 8 10

-18.0 -14.4 -10.8 -7.2 -3.6 0 3.6 7.2 10.8 14.4 18.0

0.08 0.28 0.20

P+ Q+

P+

Q+

P-

Q-

P-

Q-

z = 0 plane

x = 0 plane

z = 0 plane

x = 0 plane

(e) (f)

x z

x z

x z

x z

y

y

z

y

y

z

(g)

Figure 15. Fluid velocity and vorticity distribution at the end of the Duck at T = 1.6 s and H =

0.12 m. The subfigures in the left column show the approximately undisturbed fluid velocity vectordistribution in the plane y = 2.4 m, with subfigures (a,c,e,g) for time instances A, B, C and D, respectively.The subfigures in the right column show the vorticity of the main vortex in either the free surface, i.e.,z = 0 plane, or the perpendicular surface, i.e., x = 0 plane, with subfigures (b,d,f,h) for time instancesA, B, C and D, respectively. Subfigure (i) shows the velocity direction of water particles at differenttime instances.

Energies 2019, 12, 1941 16 of 19

Energies 2019, 12, x FOR PEER REVIEW 15 of 19

for time instances A, B, C and D, respectively. Subfigure (i) shows the velocity direction of water particles at different time instances.

At time instance B, the main flow is in the y-z plane, i.e., x = 0. Similar to the above principle, the P and Q vortexes are generated and both in the positive x direction as confirmed in Figure 15d. At time instance C, the situation is rightly opposite to time instance A. The location of P and Q vortexes are reversed, and both are in the negative z direction as shown in Figure 15f. At time instance D, both location and direction of P and Q vortexes are reversed from time instance B as confirmed in Figure 15h.

From above discussions, we find that: at different time instances within a wave period, the magnitude and direction of the generated vortexes are different, and vary periodically. Vortexes are generated due to the relative motion between the WEC and the fluid, thus part of the kinetic energy of the WEC will be devoted to feed the vortex generation process, resulting in the captured power of the WEC be reduced.

The vortex generation process causes the static pressure in the downstream side of the obstacle be smaller than that of the upstream side, resulting in a force resisting the relative motion of the obstacle, e.g., a resisting torque for the Duck in this paper. This resisting force is called the drag force, whose magnitude is proportional to the square of the velocity and direction is opposite to the velocity. Therefore, the nonlinearity of the Duck motion, which is embodied by the vortex generation phenomenon, can be quantitatively represented by the drag torque. Besides, since the analysis in Section 5.4 revealed that the nonlinearity of the wave is positively correlated to the nonlinearity of the Duck motion, we combine the two nonlinear factors together so that the total nonlinearity is quantitatively represented by a single drag torque. As in [40], the drag torque of the Duck in the pitch degree-of-freedom can be defined as:

d dT C θ θ= − , (8)

where Cd is the drag torque coefficient. Then, according to the Newton’s Second Law, we can obtain the equation of motion of the Duck as:

e r h dJθ T T T T= + + + , (9)

where Te is the excitation torque, Tr is the radiation torque, and Th is the hydrostatic torque. By solving Equation (9), we can obtain the motion response of the Duck in the time domain considering the drag torque. However, the drag torque coefficient in Equation (8) for a given wave condition cannot be obtained directly. In this paper, this drag torque coefficient is obtained in an indirect way as in [33], by tuning its value until the motion response of the Duck obtained from Equation (9) equals that of the CFD model.

Figure 16. The drag torque coefficient as a function of wave height at T = 1.6s and c = 12N·m·s/rad.

0 0.05 0.1 0.15 0.20

5

10

15

20

25

H, m

Cd,

N.m

.s2

Figure 16. The drag torque coefficient as a function of wave height at T = 1.6 s and c = 12 N·m·s/rad.

5.6. Hydrodynamic Force Distribution Along the Longitudinal Direction of the Duck

In order for the solo Duck WEC to survive in harsh offshore climates, the hydrodynamic forcedistribution along the longitudinal direction of the Duck should be carefully considered in designingthe intensity and stiffness of the WEC structure. Figure 17 shows the comparison of surge and heaveforces per unit square meter along the longitudinal direction of the Duck among 3D CFD, pseudo 3DCFD and BEM results at t = 8.544 s when T = 1.6 s and H = 0.08 m.

Energies 2019, 12, x FOR PEER REVIEW 16 of 19

Figure 16 shows the drag torque coefficient as a function of wave height at T = 1.6 s and c = 12 N·m·s/rad. At small wave heights, the drag torque coefficient is almost zero. However, at large wave heights, the drag torque coefficient increases significantly with wave height. This is due to the nonlinear behavior of the wave and the vortex generation phenomenon is more prominent at large wave heights. The observed variation tendency in Figure 16 indicates that the drag torque coefficient effectively reflected the influence of the nonlinearity of the wave and the Duck motion. This gives the opportunity to build a database of the drag torque coefficient as a function of wave height, wave period, etc, so that the hydrodynamic nonlinearity can be modeled with reasonable accuracy without the time-consuming CFD model.

5.6. Hydrodynamic Force Distribution Along the Longitudinal Direction of the Duck

In order for the solo Duck WEC to survive in harsh offshore climates, the hydrodynamic force distribution along the longitudinal direction of the Duck should be carefully considered in designing the intensity and stiffness of the WEC structure. Figure17 shows the comparison of surge and heave forces per unit square meter along the longitudinal direction of the Duck among3D CFD, pseudo 3D CFD and BEM results at t = 8.544 s when T = 1.6 s and H = 0.08 m.

(a) (b)

Figure 17. Comparison of hydrodynamic forces per square meter along longitudinal direction of the Duck among 3D CFD, pseudo 3D CFD and BEM results at t = 8.544 s when T = 1.6 s and H = 0.075 m: (a) surge force; (b) heave force.

The pseudo 3D CFD simulation is intended for simulating the hydrodynamic performance of the spine-connected Duck, and is implemented by applying the symmetric boundary at both ends of the Duck. Correspondingly, 3D CFD and BEM simulations are performed for the solo Duck. The hydrodynamic forces keep almost constant in the longitudinal direction in pseudo 3D CFD simulations, while vary continuously in 3D CFD and the BEM simulations, as a result of the diffracted and radiated wave in the y direction, which cannot be developed in pseudo 3D CFD simulations. This indicates that the structure design of the solo Duck should be different from that of the spine-connected Duck when considering the longitudinal force distribution. In both 3D CFD and BEM simulations, when approaching the end of Duck, the surge force decreases, while the heave force gradually increases. One interesting finding in both subfigures is the rapid jump of the forces at the end of the Duck in 3D CFD simulations, and is quite remarkable for the heave force. Figure 18 shows the comparison of the dynamic pressure contour between 3D CFD and pseudo 3D CFD simulations at t = 8.544 s, when T = 1.6 s and H = 0.075 m. It can be seen that the position where dynamic pressure dramatically varies, which is not observed in pseudo 3D CFD simulations, coincides with where vortexes are generated. Actually, the vortex generation process is due to pressure differences around the end of the Duck. The pressure near the end of the Duck quickly drops down since the vortexes take away the kinetic energy from the fluid, resulting in the rapid jump of the pressure force. The comparison shows that the CFD model can subtly capture the local feature of the vortex generation process and its influence on the hydrodynamic forces. Therefore,

0 0.2 0.4 0.6 0.8 170

80

90

100

110

120

z, m

Sur

ge f

orce

per

suq

are

met

er, N

/m2

3D CFDPseudo 3D CFDBEM

0 0.2 0.4 0.6 0.8 190

100

110

120

130

140

150

z, m

Hea

ve f

orce

per

suq

are

met

er, N

/m2

3D CFDPseudo 3D CFDBEM

Figure 17. Comparison of hydrodynamic forces per square meter along longitudinal direction of theDuck among 3D CFD, pseudo 3D CFD and BEM results at t = 8.544 s when T = 1.6 s and H = 0.075 m:(a) surge force; (b) heave force.

The pseudo 3D CFD simulation is intended for simulating the hydrodynamic performanceof the spine-connected Duck, and is implemented by applying the symmetric boundary at bothends of the Duck. Correspondingly, 3D CFD and BEM simulations are performed for the soloDuck. The hydrodynamic forces keep almost constant in the longitudinal direction in pseudo 3DCFD simulations, while vary continuously in 3D CFD and the BEM simulations, as a result of thediffracted and radiated wave in the y direction, which cannot be developed in pseudo 3D CFDsimulations. This indicates that the structure design of the solo Duck should be different from that ofthe spine-connected Duck when considering the longitudinal force distribution. In both 3D CFD andBEM simulations, when approaching the end of Duck, the surge force decreases, while the heave forcegradually increases. One interesting finding in both subfigures is the rapid jump of the forces at theend of the Duck in 3D CFD simulations, and is quite remarkable for the heave force. Figure 18 showsthe comparison of the dynamic pressure contour between 3D CFD and pseudo 3D CFD simulationsat t = 8.544 s, when T = 1.6 s and H = 0.075 m. It can be seen that the position where dynamicpressure dramatically varies, which is not observed in pseudo 3D CFD simulations, coincides withwhere vortexes are generated. Actually, the vortex generation process is due to pressure differences

Energies 2019, 12, 1941 17 of 19

around the end of the Duck. The pressure near the end of the Duck quickly drops down since thevortexes take away the kinetic energy from the fluid, resulting in the rapid jump of the pressure force.The comparison shows that the CFD model can subtly capture the local feature of the vortex generationprocess and its influence on the hydrodynamic forces. Therefore, the CFD model is recommended forthis kind of calculation. From the above discussion, it is found that the hydrodynamic forces of thesolo Duck are different from that of its spine-connected counterpart. The resultant surge force of thesolo Duck is smaller than the spine-connected Duck, which means that the modulus of the Duck crosssection in the x direction can be reduced, and furthermore, the mooring force in this direction can alsobe reduced. Meanwhile, the resultant heave force of the solo Duck is also smaller, which means thatthe modulus of the Duck section in the z direction can be reduced, and likewise for the mooring force.Overall, the smaller dynamic surge and heave forces will reduce the fatigue load on the Duck WEChull and the mooring system, which will result in the fatigue life of the Duck to be extended.

Energies 2019, 12, x FOR PEER REVIEW 17 of 19

the CFD model is recommended for this kind of calculation. From the above discussion, it is found that the hydrodynamic forces of the solo Duck are different from that of its spine-connected counterpart. The resultant surge force of the solo Duck is smaller than the spine-connected Duck, which means that the modulus of the Duck cross section in the x direction can be reduced, and furthermore, the mooring force in this direction can also be reduced. Meanwhile, the resultant heave force of the solo Duck is also smaller, which means that the modulus of the Duck section in the z direction can be reduced, and likewise for the mooring force. Overall, the smaller dynamic surge and heave forces will reduce the fatigue load on the Duck WEC hull and the mooring system, which will result in the fatigue life of the Duck to be extended.

(a)

(b)

Figure 18. Comparison of dynamic pressure contour between 3D CFD and pseudo 3D CFD simulations at t = 8.544 s, when T = 1.6 s and H = 0.075 m: (a) 3D CFD; (b) pseudo 3D CFD.

6. Conclusions

This paper developed a CFD model to predict the hydrodynamic performance of a solo Duck WEC in regular waves within a wide range of wave steepness until the Duck capsizes. The CFD model, in which the sliding mesh technology is employed, is validated by comparing the predicted hydrodynamic responses, i.e., the pitch angular displacement amplitude, the surge and heave hydrodynamic forces, with that predicted by the experimental and BEM models and some nonlinear phenomena are studied.

CFD results agree well with experimental results with the main difference coming from the friction brought by the mechanical transmission system. CFD results also agree well with BEM results with the difference resulting from hydrodynamic nonlinearity, which is identified as two factors that is taken into consideration in the CFD model: the nonlinearity of the wave, i.e., the nonlinear waveform, and the nonlinearity of the Duck motion, i.e., the vortex generation process. The influence of both two nonlinear factors is combined to be quantitatively represented by the drag torque. This provides the opportunity to build a database of the drag torque coefficient as a function of wave height, wave period, etc., so that these nonlinear factors can be modeled with reasonable accuracy without the time-consuming CFD model. The vortex generation phenomenon is found to cause a rapid jump of the pressure force due to the vortexes taking away the kinetic energy from the fluid. The hydrodynamic surge and heave forces and required mooring forces of the solo Duck is smaller than that of its spine-connected counterpart owing to the three-dimensional characteristic of the structure.

Author Contributions: J.W. performed the numerical simulation and prepared the manuscript. Y.Y., D.S., Z.N. and M.G. supervised the wave power project.

Funding: The research was supported by Shenzhen Science and Technology Innovation Committee Grant 20160221 and the Swedish Research Council (VR) Grant 2015-04657.

Conflicts of Interest:The authors declare no conflict of interest.

References

1. Gunn, K.; Stock-Williams, C. Quantifying the global wave power resource. Renew. Energy2012, 44, 296–304. 2. International Energy Agency (IEA). Key World Energy Statistics 2016; IEA: Paris, France, 2016; pp. 35–36.

Figure 18. Comparison of dynamic pressure contour between 3D CFD and pseudo 3D CFD simulationsat t = 8.544 s, when T = 1.6 s and H = 0.075 m: (a) 3D CFD; (b) pseudo 3D CFD.

6. Conclusions

This paper developed a CFD model to predict the hydrodynamic performance of a solo Duck WECin regular waves within a wide range of wave steepness until the Duck capsizes. The CFD model, inwhich the sliding mesh technology is employed, is validated by comparing the predicted hydrodynamicresponses, i.e., the pitch angular displacement amplitude, the surge and heave hydrodynamic forces,with that predicted by the experimental and BEM models and some nonlinear phenomena are studied.

CFD results agree well with experimental results with the main difference coming from the frictionbrought by the mechanical transmission system. CFD results also agree well with BEM results with thedifference resulting from hydrodynamic nonlinearity, which is identified as two factors that is takeninto consideration in the CFD model: the nonlinearity of the wave, i.e., the nonlinear waveform, and thenonlinearity of the Duck motion, i.e., the vortex generation process. The influence of both two nonlinearfactors is combined to be quantitatively represented by the drag torque. This provides the opportunityto build a database of the drag torque coefficient as a function of wave height, wave period, etc., so thatthese nonlinear factors can be modeled with reasonable accuracy without the time-consuming CFDmodel. The vortex generation phenomenon is found to cause a rapid jump of the pressure force due tothe vortexes taking away the kinetic energy from the fluid. The hydrodynamic surge and heave forcesand required mooring forces of the solo Duck is smaller than that of its spine-connected counterpartowing to the three-dimensional characteristic of the structure.

Author Contributions: J.W. performed the numerical simulation and prepared the manuscript. Y.Y., D.S., Z.N.and M.G. supervised the wave power project.

Funding: The research was supported by Shenzhen Science and Technology Innovation Committee Grant 20160221and the Swedish Research Council (VR) Grant 2015-04657.

Conflicts of Interest: The authors declare no conflict of interest.

Energies 2019, 12, 1941 18 of 19

References

1. Gunn, K.; Stock-Williams, C. Quantifying the global wave power resource. Renew. Energy 2012, 44, 296–304.[CrossRef]

2. International Energy Agency (IEA). Key World Energy Statistics 2016; IEA: Paris, France, 2016; pp. 35–36.3. Falnes, J. A review of wave energy extraction. Mar. Struct. 2007, 20, 185–201. [CrossRef]4. Hong, Y.; Waters, R.; Bostrom, C.; Eriksson, M.; Engstrom, J.; Leijon, M. Review on electrical control strategies

for wave energy converting systems. Renew. Sust. Energy Rev. 2014, 31, 329–342. [CrossRef]5. Salter, S.H. Wave power. Nature 1974, 249, 720–724. [CrossRef]6. Salter, S.H. Characteristics of a rocking wave power device. Nature 1975, 254, 504–506.7. Cruz, J. Ocean Wave Energy Current Status and Future Perspectives; Springer: Berlin, Germany, 2008.8. Jeffrey, D.C.; Richmond, D.J.E.; Salter, S.H.; Taylor, J.R.M. Study of Mechanisms for Extracting Power from Sea

Waves; Second Year Report on Edinburgh Wave Power Project; University of Edinburgh: Edinburgh, UK,September 1976.

9. Salter, S.H. The development of the Duck concept. In Proceedings of the Wave Energy Conference, London,UK, 22–23 November 1978.

10. Jeffrey, D.C.; Keller, G.J.; Mollison, D.; Richmond, D.J.E.; Salter, S.H.; Taylor, J.R.M.; Young, I.A. Fourth YearReport on Edinburgh Wave Power Project; University of Edinburgh: Edinburgh, UK, 1978.

11. Evans, D.V. A theory for wave-power absorption by oscillating bodies. J. Fluid Mech. 1976, 77, 1–25.[CrossRef]

12. Mynett, A.E.; Serman, D.D.; Mei, C.C. Characteristics of Salter’s cam for extracting energy from ocean waves.Appl. Ocean Res. 1979, 1, 13–20. [CrossRef]

13. Nebel, P. Maximising the efficiency of wave energy plant using complex-conjugate control. J. Syst. ControlEng. 1992, 206, 225–306.

14. Wu, J.; Yao, Y.; Li, W.; Zhou, L.; Göteman, M. Optimizing the Performance of Solo Duck Wave EnergyConverter in Tide. Energies 2017, 10, 289.

15. Pizer, D. Numerical Modeling of Wave Energy Absorbers; Technical Report; University of Edinburgh: Edinburgh,UK, 1994.

16. Cruz, J.; Salter, S.H. Numerical and experimental modeling of a modified version of the Edinburgh Duckwave energy device. Proc. Inst. Mech. Eng. Part M J. Eng. Marit. Environ. 2006, 220, 129–147.

17. Cruz, J.; Salter, S. Update on the design of an offshore wave powered desalination device. In Proceedings ofthe OWEMES 2006-Civitavecchia, Rome, Italy, 20–22 April 2006.

18. Lucas, J.; Salter, S.H.; Cruz, J.; Taylor, J.R.M.; Bryden, I. Performance optimization of a modified Duck throughoptimal mass distribution. In Proceedings of the 8th European Wave and Tidal Energy Conference, Uppsala,Sweden, 7–10 September 2009.

19. Pastor, J.; Liu, Y. Power Absorption Modeling and Optimization of a Point Absorbing Wave Energy ConverterUsing Numerical Method. J. Energy Res. Technol. 2014, 136, 021207. [CrossRef]

20. Pastor, J.; Liu, Y. Frequency and time domain modeling and power output for a heaving point absorber waveenergy converter. Int. J. Energy Environ Eng. 2014, 5, 101. [CrossRef]

21. Li, Y.; Lin, M. Regular and irregular wave impacts on floating body. Ocean Eng. 2012, 42, 93–101. [CrossRef]22. Hadzic, I.; Hennig, J.; Peric, M.; Xing-Kaeding, Y. Computation of flow-induced motion of floating bodies.

Appl. Math. Model. 2005, 29, 1196–1210. [CrossRef]23. Wei, Y.; Rafiee, A.; Henry, A.; Dias, F. Wave interaction with an oscillating wave surge converter, Part I:

Viscous effects. Ocean Eng. 2015, 104, 185–203. [CrossRef]24. Wei, Y.; Abadie, T.; Henry, A.; Dias, F. Wave interaction with an oscillating wave surge converter, Part II:

Slamming. Ocean Eng. 2016, 113, 319–334. [CrossRef]25. Agamloh, E.B.; Wallace, A.K.; Jouanne, A.V. Application of fluid-structure interaction simulation of an ocean

wave energy extraction device. Renew. Energy 2008, 33, 748–757. [CrossRef]26. Lopez, I.; Pereiras, B.; Castro, F.; Iglesias, G. Optimisation of turbine-induced damping for an OWC wave

energy converter using a RANS-VOF numerical model. Appl. Energy 2014, 127, 105–114. [CrossRef]27. Penalba, M.; Giorgi, G.; Ringwood, J.V. Mathematical modelling of wave energy converters: A review of

nonlinear approaches. Renew. Sustain. Energy Rev. 2017, 78, 1188–1207. [CrossRef]

Energies 2019, 12, 1941 19 of 19

28. Windt, C.; Davidson, J.; Ringwood, J.V. High-fidelity numerical modelling of ocean wave energy systems: Areview of computational fluid dynamics-based numerical wave tanks. Renew. Sustain. Energy Rev. 2018, 93,610–630. [CrossRef]

29. ANSYS Inc. ANSYS Fluent 16.2 Manual; ANSYS Inc.: Canonsburg, PA, USA, 2016.30. Gomes, R.P.F.; Henriques, J.C.C.; Gato, L.M.C.; Falcão, A.F.O. Wave power extraction of a heaving floating

oscillating water column in a wave channel. Renew. Energy 2016, 99, 1262–1275. [CrossRef]31. WAMIT Inc. WAMIT User Manual Version 7; WAMIT Inc.: Chestnut Hill, MA, USA, 2013.32. Bhinder, M.A.; Babarit, A.; Gentaz, L.; Ferrant, P. Potential time domain model with viscous correction

and CFD analysis of a generic surging floating wave energy converter. Int. J. Mar. Energy 2015, 10, 70–96.[CrossRef]

33. Han, J.; Maeda, T.; Kinoshita, T.; Kitazawa, D. Towing test and motion analysis of motion controlledship—Based on an application of skyhook theory. In Proceedings of the 12th International Conference on theStability of Ships and Ocean Vehicles (STAB2015), Glasgow, Scotland, 14–19 June 2015.

34. Devenport, W.J.; Rife, M.C.; Liapis, S.; Folin, G.J. The structure and development of a wing-tip vortex. J.Fluid Mech. 1995, 312, 67–106. [CrossRef]

35. Wood, D.H.; Okulov, V.L.; Bhattacharjee, D. Direct calculation of wind turbine tip loss. Renew. Energy 2016,95, 269–276. [CrossRef]

36. Yu, D.; Kareem, A. Parametric study of flow around rectangular prisms using LES. J. Wind Eng. Ind. Aerodyn.1998, 77–78, 653–662. [CrossRef]

37. Arslan, T.; Malavasi, S.; Pettersen, B.; Andersson, H.I. Turbulent flow around a semi-submerged rectangularcylinder. J. Offshore Mech. Arct. Eng. 2013, 135, 1–11. [CrossRef]

38. Sherry, M.J.; Jacono, D.L.; Sheridan, J.; Mathis, R.; Marusic, I. Flow separation characterization of a forwardfacing step immersed in a turbulent boundary layer. In Proceedings of the Sixth International Symposiumon Turbulence and Shear Flow Phenomena, Seoul, Korea, 22–24 June 2009; pp. 1325–1330.

39. Shimada, K.; Ishihara, T. Application of a modified k-ε model to the prediction of aerodynamic characteristicsof rectangular cross-section cylinders. J. Fluid Struct. 2002, 16, 465–485. [CrossRef]

40. Sjökvist, L.; Wu, J.; Ransley, E.; Engström, J.; Eriksson, M.; Göteman, M. Numerical models for the motion andforces of point-absorbing wave energy converters in extreme waves. Ocean Eng. 2017, 145, 1–14. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open accessarticle distributed under the terms and conditions of the Creative Commons Attribution(CC BY) license (http://creativecommons.org/licenses/by/4.0/).