Application of Boundary Element

Anatoliy Khait1
Mem. ASME
School of Mechanical Engineering,
Faculty of Engineering,
Tel Aviv University,
Tel Aviv 6997801, Israel
e-mail: haitanatoliy@gmail.com
Lev Shemer
Mem. ASME
School of Mechanical Engineering,
Faculty of Engineering,
Tel Aviv University,
Tel Aviv 6997801, Israel
e-mail: shemer@eng.tau.ac.il
Method for Determination of the
Wavemaker Driving Signal
A method for the generation of steep nonlinear broad-banded wave trains having an arbi-
trary prescribed shape is developed. It is shown that the second-order contributions to the
velocity field are negligible in deep water, while the second-order bound components of the
surface elevation are significant. This fact allows improvement of an iterative method of
the wavemaker driving signal adjustment that increases the accuracy of excitation of
wave train with the prescribed free waves’ spectrum. The decomposition of the complex
amplitude spectrum of the surface elevation into free and bound components is based on
the approach adopted in the derivation of the Zakharov model. The iterative adjustment
of the driving signal is carried out using the numerical wave tank based on the boundary
element method. It is demonstrated that accurate wave train excitation is attained for dif-
ferent values of the wave steepness. The method allows decreasing the number of iterations
needed for the driving signal adjustment. The surface elevation values measured in the lab-
oratory wave tank agree closely with those obtained in the numerical simulations. The mea-
sured and the simulated frequency spectra are in agreement as well.
[DOI: 10.1115/1.4042942]

Introduction second-order in the wave steepness theory may be attributed to the

fact that it neglects higher-order contributions. A corrected transfer
Mechanically driven wavemakers are often applied to excite
function based on experimental results was offered in Ref. [5].
either deterministic or random wave sequences in laboratory
Houtani et al. [6,7] adopted the higher-order spectral method
conditions. The linear transfer function relating the oscillatory
(HOSM) to upgrade the results of Ref. [2] for directional wave exci-
wavemaker motion at a constant frequency to the resulting mono-
tation. The HOSM-WG (here WG stands for the wave generation)
chromatic surface elevation in the tank is given in Ref. [1]. The the-
method developed in Ref. [6,7] was shown to be superior to the con-
oretical analysis predicts that in addition to the designed sinusoidal
ventional wave generation when wave steepness increases, the fre-
wave, undesirable (the so-called evanescent) modes are excited in
quency spectrum broadens or the directional spreading becomes
the tank. These modes that result from inconsistency of the vertical
large. Particular attention was paid to the generation of the freak waves.
distribution of the horizontal component of the wavemaker motion
An alternative approach proposed in Refs. [3–5] is restricted to an
with that of the orbital wave velocity are indeed observed in exper-
accurate excitation of single-frequency waves. The approach is
iments; these waves decay fast with the distance from the wave-
based on the utilization of the force-feedback controller system
maker. Since the evanescent modes do not affect the wave field at
that employs the impedance matching procedure. In contrast to
large distances, they are often disregarded. Experiments, however,
Ref. [2], which applies theoretically calculated second-order
demonstrate that additional propagating modes, in particular at
driving signal correction, in Refs. [3–5] forces acting on the wave
the second harmonic of the dominant frequency, are observed.
board surface by force transducers are measured and serve as feed-
These undesirable propagating waves contaminate the designed
back to the computer-aided wavemaker system. An additional
wave field over much longer distances.
second-order correction was also implemented. It was found out that
Analytical second-order wavemaker theory for excitation of
wavemakers driven by force-controlled procedures, as opposed to
multiple harmonics in deep and intermediate-depth water was pro-
the displacement control, significantly reduce the content of the
posed by Schäffer [2]. He also suggested a second-order correction
second-order spurious waves. The method presented in Refs.
to the wavemaker motion in order to suppress the spurious propagat-
[3,4] also has an additional significant advantage due to its ability
ing waves. This theory was generally confirmed experimentally for
to simultaneously generate and absorb any unwanted waves, includ-
regular waves, wave groups, and irregular wave trains. Nevertheless,
ing the reflected ones, by a single mechanical device. On the other
certain deviations of the measured surface elevation from the theoret-
hand, it is limited to simple nearly monochromatic signals and
ical predictions based on the Schäffer theory [2] were observed,
requires high-precision force transducers to measure the wave
indicating that spurious waves are not completely removed. Spinne-
board torque, thus significantly complicating the experimental setup.
ken and Swan [3,4] and Aknin and Spinneken [5] re-examined
The need to generate accurately broad-banded wave trains
the theory of Schäffer for a wide range of excitation condi-
prompted the development of empirical methods for wavemaker
tions: shallow-to-intermediate, intermediate, and deepwater-wave
motion adjustment. To this end, an experimental technique to gen-
regimes, with wave steepness in the range of ε = 0.05–0.25 and for
erate the wave packet focusing at the prescribed location in the
either flap or piston wavemakers. Differences between the theoreti-
wave tank was applied in Ref. [8]. Adjustment of phases at the
cal predictions and experiments were found to exist for wave ampli-
wave board yielded satisfactory convergence of the wave energy
tudes as well as for phases of the spurious waves. These differences
at the specified focus point. In contrast, a numerical wave tank
persist for all experimental conditions. The limited accuracy of the
(NWT) based on the boundary element method (BEM) was used
in Ref. [9] to iteratively adjust the phases of the wave components
of the wave packet yielding the required wavemaker motion. The
1
Corresponding author. phase shifting was checked by a numerical wave probe and
Contributed by the Ocean, Offshore, and Arctic Engineering Division of ASME for
publication in the JOURNAL OF OFFSHORE MECHANICS AND ARCTIC ENGINEERING. Manu-
further verified experimentally.
script received August 31, 2018; final manuscript received February 3, 2019; published Shemer et al. [10,11] applied an experimentally determined
online March 20, 2019. Assoc. Editor: Felice Arena. wavemaker transfer function in order to determine amplitudes and

Journal of Offshore Mechanics and Arctic Engineering DECEMBER 2019, Vol. 141 / 061102-1
Copyright © 2019 by ASME
phases of the spectral components of the wavemaker driving signal
for generation of a propagating wave train with a prescribed shape.
The spatial version of the Zakharov equation was used as a theoret-
ical model for computation of spectral changes during the wave
train propagation along the experimental facility. Schmittner et al.
[12] utilized the phase–amplitude iteration scheme to experimen-
tally synthesize extreme deterministic wave sequences. They indi-
cate that this scheme is applicable even without a priorily
knowing the transfer functions of the wave generator. However,
their study demonstrates a noticeable discrepancy between the the-
oretical and the measured surface elevations. Buldakov et al. [13]
improved the iterative wave generation method by application of
the harmonics separation technique for linearization of the ampli-
tude spectrum. They showed that experimental adjustment of the
linear part of the spectrum only considerably improves the genera-
In laboratory facilities, waves are usually generated by the hori-
zontal motion of the wave board that prescribes the vertical distribu-
tion of the horizontal water velocity component. In linear
approximation, this condition allows one to determine the transfer
function c (1) [1]:
∂ψ (1) ∂S(1)
u(1) (x = 0, z = 0, t) ≡ − (x = 0, z = 0, t) = c(1) (z = 0, t) (4)
∂x ∂t
Here, S(z,t) is the horizontal wavemaker coordinate, u is the hor-
izontal velocity of the fluid, and x = 0 corresponds to the mean
wavemaker location. For weakly nonlinear waves, the horizontal
velocity can be expanded around the small parameter ε resulting
in the second-order boundary condition at the wavemaker. For
piston-type wavemaker, it reads as
   2 (1)  (1)

tion accuracy. The method requires four experimental runs (with ∂ψ ∂ψ (1)  ∂ψ (2)  ∂ ψ ∂ψ
different shifts in phases of the driving signal harmonics) at every = + − dt (5)
∂x ∂x x=0 ∂x x=0 ∂x2 ∂x
iteration to complete the separation of harmonics, thus resulting  
 
 

x=0
in a significant extension of the duration of experiments, especially u(1) u(2) u(2)
d
in large experimental facilities.
Existing theoretical models of nonlinear waves based on Schrö- Although Eq. (5) is valid for any vertical coordinate z, we further
dinger, Dysthe, Zakharov equations, etc. [14–18] describe evolution consider only the mean surface level z = 0. In Eq. (5), the linear
of free water gravity waves only, i.e., waves that satisfy the deep or velocity u (1) estimated according to Eq. (4) is supplemented with
intermediate-depth dispersion relation between the wave numbers k two additional second-order terms. The term u(2) d is caused by
and the angular frequencies ω: finite displacement of the wave board, while u (2) is associated
with the second-order bound waves η (2). In contrast to propagating
ω2 = gk tanh (kh) (1) waves η (1), bound waves are determined by free waves and do not
Due to nonlinearity, the free waves induce the corresponding satisfy the dispersion relation (1). Substitution of ψ (1) and ψ (2) into
higher-order bound waves that are totally defined by the carrier Eq. (5) yields the following:
free waves and cannot be controlled separately. ka
This paper is aimed at the improvement of the accuracy of the |û(1) |(ka, kh) = √
steep broad-banded wave trains generation in deepwater conditions. tanh (kh)
√
We develop a fully nonlinear method to adjust the driving signal of 3 cosh (2kh) tanh (kh)
the wavemaker to generate wave trains with the prescribed broad |û(2) |(ka, kh) = (ka)2 (6)
4 sinh4 (kh)
free waves’ spectrum. An additional benefit of this method is the
improvement of the convergence rate of the iteration process (ka)2
since it distinguishes harmonics according to their order of nonlin- d |(ka, kh) = 
|û(2)
2 tanh3 (kh)
earity. To improve the accuracy of the iteration procedure, measure-
ments of the surface elevation variation should be carried out in the The relative second-order contributions to the horizontal velocity
vicinity of the wavemaker. amplitudes at the free surface are plotted in Fig. 1 as a function
of the steepness ka and water depth kh. Deepwater conditions
Wave Generation Method exist when kh > π (tanh (kh) → 1), whereas the range π/10 < kh < π
(0.3 < tanh (kh) < 1) corresponds to the intermediate depth [1]. For
Nonlinear Wave Generation. Assumption of weak nonlinear- steep waves, the surface elevation amplitudes of the bound wave
ity within potential wave theory [1] allows perturbation expansion |η̂(2) | may become essential compared with |η̂(1) | even in deep
of the surface elevation η and potential ψ. Taking the wave steep- water. However, the amplitude of the second-order horizontal
ness ε = ka as a small parameter, k and a being the wave number
velocity |û(2) | vanishes with an increase of kh and in the limiting
and the surface elevation amplitude, the second-order expansion
case of deep water it is lower by few orders than the velocity ampli-
can be written as follows:
tude of the propagating wave |û(1) |. Therefore, it can be assumed


η(x, t) = η(1) (x, t) + η(2) (x, t) that the second-order velocity field can be neglected in the
(2) process of generation of deepwater waves, while the surface eleva-
ψ(x, z, t) = ψ (1) (x, z, t) + ψ (2) (x, z, t)
tion variations due to bound waves should be accounted for.
where x defines the propagation direction of unidirectional waves, z Figure 2 exhibits the velocity fields at different orders for the
is directed upward, and t is the time. The dimensionless amplitudes
 regular wave with ka = 0.3 and kh = π. It is seen that the
are defined as η̂ = ηA k and ψ̂ = ψ A k 3 /g, where ηA and ψA are the second-order velocities remain well below those of the first-order
corresponding dimensional values. For regular waves, the absolute even in the vicinity of the steep crest, while the shape of the
values of those amplitudes are given as follows [1]: surface elevation is noticeably asymmetric with respect to z = 0.
The difference in the behavior of η (2) and u (2) in Fig. 1 can be
|η̂(1) |(ka, kh) = ka related to the vertical velocity w of the surface elevation at a
fixed location x:
(ka)2 cosh (kh)[2 + cosh (2kh)]   
|η̂(2) |(ka, kh) = ∂η ∂ψ (1)  ∂ψ (2) 
4 sinh3 (kh) = − −
(3) ∂t x=0 ∂z x=0,z=0 ∂z x=0,z=0
ka
|ψ̂ (1) |(ka, kh)|z=0 = √  

 

tanh (kh) w(1) w(2)

√  2 (1)
(7)
3 cosh (2kh) tanh (kh) ∂η ∂ψ(1) (1)
(1) ∂ ψ
|ψ̂ (2) |(ka, kh)|z=0 = (ka)2 + −η
8 sinh4 (kh) ∂x ∂x ∂z2 x=0,z=0
 

where kh is the dimensionless water depth. w(2)

st

061102-2 / Vol. 141, DECEMBER 2019 Transactions of the ASME

 Fig. 2 Normalized velocity fields of (a) the propagating and
(b) second-order components of a regular wave at kh = π and
ka = 0.3

Fig. 1 Relative contribution of (a) the second-order bound

waves to the surface elevation and (b) the second-order horizon-
tal velocity component, as a function of wave steepness and
water depth

Here, w (2) represents the second-order contribution to the verti-

cal velocity due to the instantaneous potential variation with z,
whereas w(2)st is caused by the motion of the free surface in the
Eulerian coordinate system. Substitution of (2) into (7) yields the
following:

|ŵ(1) |(ka, kh) = ka tanh (kh)
√
3 cosh (kh) tanh (kh)
|ŵ(2) |(ka, kh) = (ka)2
2 sinh3 (kh) (8)
(ka)2
st |(ka, kh) = √
|ŵ(2)
tanh (kh)
Figure 3 compares different contributions to the vertical velocity
of the free surface (8) for the regular wave with steepness ka = 0.3.
Both w (2) and wst(2) contribute significantly to the surface elevation
variation in an intermediate depth causing the appearance of the Fig. 3 Comparison of the contributions (8) to the vertical motion
bound waves η (2). In deep water, w (2) is negligibly small and of the free surface of regular wave with ka = 0.3
bound waves η (2) are stipulated by wst(2). This explains the differ-
ences in behavior of |η̂(2) | and |û(2) | in Fig. 1. Fig. 4. Within this approach, the surface elevation at xt is repre-
The unwanted spurious propagating waves [2] appear when sented in the spectral form containing free as well as bound compo-
both second-order terms u (2) and u(2) d in Eq. (5) are not taken into nents η(xt, t) = η (1) + η (2):
account in the wavemaker motion. Since u (2) is negligibly small  
in deep water, u(2) 
d is the only essential source of inaccuracies at
η(x, t) = Re aj e i[kj (x−xt )−ωj t]
(9)
the second-order in this case. As seen from Eq. (5), the expression
j
d incorporates only the first-order terms. The wavemaker
for u(2)
motion can therefore be adjusted iteratively so that it satisfies the where kj and ωj are the wave number and the angular frequency of
nonlinear boundary condition (5) in deep water. the jth component, respectively.
The wave generation method consists of three steps. First, the
free complex amplitude spectrum of the target wave train is deter-
Wave Generation in Deep Water. The suggested iterative mined and the surface elevation at the wavemaker location x = 0
generation method for a broad-banded wave train is most suitable is calculated. The linear transfer function c (1) (4) is then applied
for deepwater conditions. The iterative adjustment of the wave- to convert the temporal variation of the surface elevation into the
maker motion was performed with respect to the surface elevation required motion of the wave board. Next, the wave train is
variation with time recorded at the virtual probe location xt, see excited either experimentally or numerically employing a fully

Journal of Offshore Mechanics and Arctic Engineering DECEMBER 2019, Vol. 141 / 061102-3
 Fig. 4 Schematic of the NWT
nonlinear numerical wave tank using the given wavemaker motion,
and the full frequency spectrum of the surface elevation is deter-
mined at xt.
As shown earlier, for wave generation in deep water, the
second-order velocity field u (2) is vanishingly small and may be
neglected. Decomposition into free and bound waves according to
the order of nonlinearity based on the Zakharov theory [15–18] is
utilized here. The leading order free components that satisfy the dis-
persion relation (1) are separated from the contribution of higher-
order bound waves by an iterative procedure [10,11] based on the
analytic relations between the first-order free and the second-order
bound components of the surface elevation spectra (see the Appen-
dix). The spectrum of the propagating first-order waves is recalcu-
lated to the location of the wavemaker x = 0 by shifting the phases
of each harmonic according to (9).
At the next step, a modification of an iterative scheme given in
Ref. [12] is utilized for adjustment of the wavemaker motion
aimed at improvement of agreement of the linear part of the
surface elevation spectrum at x = 0 with the target one. For each
free harmonic ωj, the wavemaker amplitudes Swm(ωj) and phases
ϕwm(ωj) at iteration l + 1 are adjusted by correcting their values at
the previous iteration l by the difference between the target and
the measured values of the amplitude and the phase in the
complex surface elevation spectra:
 free
j,l+1 = S j,l + c {a j − a j,l }
Swm wm (1) tar
(10)
free
j,l+1 = ϕ j,l + {ϕj − ϕ j,l }
ϕwm wm tar
where c (1) is the linear transfer function and atar and ϕtar are ampli-
tudes and phases at the target free wave spectrum, respectively. The
iterations continue until convergence is attained.
Formulation of the Problem
We investigate here the excitation of a unidirectional surface
gravity wave train with a prescribed spectrum. A Gaussian-shaped
free-wave envelope of the surface elevation at the prescribed loca-
tion x = xf within the tank is considered as a generic shape of a wave
train with no restrictions on the spectral width:
   
t 2
η(t) = η0 exp − cos (ω0 t) (11)
mT0
Here, η0 is the envelope amplitude, T0 and ω0 = 2π/T0 are the
carrier wave period and angular frequency, respectively, and the
parameter m can be adjusted to generate wave train with any
desired spectral width. The linear amplitude spectrum aj = a(ωj)
of the wave train also has a Gaussian shape. The phases of all
061102-4 / Vol. 141, DECEMBER 2019
spectral components are identical and equal to zero, thus at the pre-
scribed focusing location, the extreme wave crest appears with the
height equal to the sum of amplitudes of all free harmonics.
Since deepwater and intermediate-depth waves are dispersive,
phase velocities cj = ωj/kj differ for different harmonics, resulting
in wave train spreading away from the focusing location with the
corresponding reduction of the wave steepness [10,11].
The accuracy of the wave generation should be controlled as
close to the wavemaker as possible in order to avoid spectral
changes due to nonlinear effects. However, the surface elevation
at short distances from the wavemaker (usually not exceeding
about 3 mean water depths h) is affected by evanescent modes
[1]. Therefore, the wave generation accuracy is monitored here at
the target location xt = 2 m = 3.3 h (see Fig. 4).
Experimental Facility. The experiments were carried out in a
18 m long, 1.2 m wide wave tank filled to mean water depth of
h = 0.6 m. The tank is equipped with a paddle-type wavemaker
hinged at the bottom. The wave-energy absorbing sloping beach
that is 3 m long is located at the opposite end of the tank. Although
the reflection from the far end of the tank is significantly reduced by
the beach, the experiments are planned so that all measurements are
completed prior to the arrival of the wave train to the beach.
The experiments, including the wave generation and data record-
ing, are controlled by a single LABVIEW program ensuring precise
synchronization. The actual instantaneous paddle inclinations are
recorded as well. Resistance-type wave gauge can be moved
along the tank; in the present experiments, it is usually located at
xt = 2 m (see Fig. 4). The wave gauge was statically calibrated
prior to every experimental run using a fully automatic procedure.
The sampling interval was set at T0/128, where T0 is the carrier
wave period.
Parameters of the wave train (11) are as follows: T0 = 0.7 s,
m = 1.2, and carrier wave steepness ε0 = k0η0 = 0.25.
Numerical Wave Tank. The iterative adjustment of the wave-
maker motion is performed in the fully nonlinear two-dimensional
NWT based on transient BEM. This allows reduction of the exper-
imental effort for the determination of the wavemaker driving
signal.
The numerical domain Ω is bounded by the bottom Γb and lateral
ΓL boundaries, the wavemaker Γwm, and the free surface Γfs, as
shown in Fig. 4. Every boundary is defined by several nodes. The
nodes on Γb and ΓL are fixed and nodes on Γwm follow the
motion of the wavemaker, while the nodes on Γfs move freely deter-
mining the instantaneous shape of the free surface. Every boundary
contains end nodes that coincide geometrically with the end nodes
of the neighbor boundaries. These corner nodes are marked in Fig. 4
and treated in a special manner [19,20].
Transactions of the ASME
 Potential and irrotational fluid flow is governed by the Laplace
equation with appropriate boundary conditions: impenetrability
conditions for Γb and ΓL, the known linear time-dependent veloc-
ity profile for Γwm, and the kinematic and the dynamic boundary
conditions at the free surface Γfs [19–21]. Green’s identity is
applied to transform the Laplace equation into a boundary integral
equation:


∂ψ ∂ξ
α(rs )ψ(rs ) = (r)ξ(r, rs ) − ψ(r) (r, rs ) dΓ (12)
∂n ∂n
Γ Fig. 5 Assessment of effectiveness of the numerical wave
where r(x,z) is the radius vector in the coordinate system x–z
(Fig. 4), ξ(r, rs) = −1/(2π) ln(|r − rs|) is the fundamental solution
that represents the potential flow at point r due to a source
located at rs, α is a geometric coefficient, α = π for regular nodes
and α ≈ π/2 for corner nodes, and Γ = Γb ∪ ΓL ∪ Γwm ∪ Γ fs . For a
source located at the node p, the equation determines the value of
the potential ψp for a known normal gradient (∂ψ/∂n)p and vice
versa. Setting the position of the source at different grid nodes
leads to a system of linear equations.
Equation (12) is discretized by sliding cubic 4-node boundary
elements as shown schematically in Fig. 4. The central part of
every single boundary element is used for approximation of the
contour Γ, whereas the outer parts define the curvature of the
element. The full set of the boundary elements constitutes a contin-
uous closed curve Γ. Local curvilinear coordinate system s–n
(Fig. 4) was introduced to integrate Eq. (12) numerically by the
Gaussian quadrature method. A more detailed description of the
BEM can be found in Refs. [19–21].
The solution depends on the spatial resolution of the grid. The
convergence study disclosed that grid density at the free surface
of at least 40 nodes/m is required to obtain a grid-independent solu-
tion. Taking into account the width of the Gaussian spectrum and
the Nyquist condition for the highest frequencies, the grid density
at the free surface was taken to be 60 nodes/m. The number of
grid nodes for different boundaries was thus set as follows: Nfs =
360, Nb = 210, NL = 37, and Nwm = 37.
The radius vector of free surface nodes and their potentials were
expanded into a truncated second-order Taylor series; the time inte-
gration of the free surface boundary conditions was performed as
suggested in Ref. [21]. Quadratic five-node sliding element was
adopted to estimate the expansion terms. The value of the integra-
tion time step was taken as T0/1024 exceeding the Courant
number requirement for explicit schemes CFL ≤ 0.1. The Courant
number is defined as follows [20]:
 Δt
CFL = gh (13)
(Δr)min
where g is the acceleration due to gravity, Δt is the time step, and
(Δr)min is the minimal distance between adjacent nodes.
Free surface elevation is recorded at xt applying cubic interpola-
tion between neighboring nodes. The configuration of the domain
Ω is similar to the physical wave tank (h = 0.6 m) except for its
length since there is no need to simulate full length of the tank
for determination of the surface elevation at xt = 2 m. The
domain length was thus reduced to LNWT = 6 m. The numerical
absorbing beach is developed to decrease wave field contamination
by the reflected waves. The length of the beach was taken to be
Lb = 3 m, see Fig. 4.
Numerical Absorbing Beach. The absorbing beach is built by
modification of the dynamic free surface boundary condition:
∂ψ 1 p μ ∂2 ψ
= gz + |∇ψ|2 + bf (x) + 2 d 2 bf (x)
∂t 2 ρ ρ ∂s
 
 
adjustment of the phases. Further improvement of the convergence
term 1 term 2 (14)
 ∂ψ
p = −pd ρ gh
∂n
Journal of Offshore Mechanics and Arctic Engineering
absorption techniques
where μd, pd, and bε are parameters; and xorig and xend are coordi-
nates of both ends of the numerical absorbing beach (see Fig. 4).
Two approaches were considered: pressure action opposing the
free surface motion (term 1 in Eq. (14)) [22] and an artificial viscos-
ity (term 2) [23]. Beach function bf(x) defines the rate of increase of
both terms along the beach; bf (x ≤ xorig) = 0:

 
1 x − xorig
bf (x) = 1 + tanh bε (15)
2 xend − xorig
It was found that the absorption efficiency is mostly affected by
the dimensionless parameters μd and pd. Based on numerical simu-
lations performed to determine the optimal values for these param-
eters, the value of bε = 6 was selected. Numerical simulation results
for the 16 m long wave tank with no absorbing beach were taken as
a basis for estimation of the beach performance; the root-mean-
square (rms) deviation of the surface elevation profile from the
basis served as a measure for the waves damping effectiveness.
Figure 5 compares the beach effectiveness for different model
parameters. The most effective absorption occurs at pd = 2; this
value was adopted for subsequent simulations. The introduction
of artificial viscosity appeared to be less effective and thus was
not implemented. Note that the values of μd exceeding 2000 Pa · s
cause numerical instability.
Results and Discussion
Excitation of Steep Waves (εc = 0.21). Generation of initially
steep highly nonlinear waves is now investigated. It is expected
that strong nonlinear effects at the wavemaker may complicate
the generation process. High initial steepness is attained by linear
focusing of the Gaussian-shaped wave train (11) at xf = 4 m. The
surface elevation at x = 0 was calculated using the linear theory.
The wave steepness at x = 0 is defined as εc ≡ k0Ax=0 = 0.21, Ax=0
being the maximum of the free wave envelope at the wavemaker
location. The corresponding linear wavemaker motion was then
applied as the boundary condition for NWT.
The surface elevation spectrum estimated from the data recorded
by the virtual probe of NWT was decomposed into free and
second-order bound components as shown in Fig. 6(a). Although
bound components have peak amplitudes that are lower by an
order of magnitude than the free ones, their contribution to the
surface elevation is essential and cannot be ignored, see Fig. 6(b).
This figure also demonstrates that free modes exist in the frequency
range of the second-order bound waves, whereas wave amplitudes
in the target spectrum at those frequencies are negligibly small.
These waves are thus classified as spurious and should be elimi-
nated during an iterative adjustment of the driving signal.
The wavemaker motion adjustment procedure relies on Eq. (10).
The convergence of the free surface elevation to the target one
is attained in 3–4 iterations, as seen from Table 1. A very rapid
drop in the discrepancy during the first iteration results from fast
is mainly due to the modification of amplitudes. Subsequent itera-
tions do not affect the error noticeably. The remaining error may
be attributed to the limited accuracy of the NWT due to its finite
DECEMBER 2019, Vol. 141 / 061102-5
Fig. 6 Separation of the spectrum into free and bound
components
Table 1 Convergence of the adjustment procedure for steep
waves
 
2 0.5 
ηBEM − ηtarg ηtarg 
Iteration no.
1 26.3%
2 1.43%
3 1.05%
4 1.01%
spatial and temporal resolutions and the disregarded third- and
higher-order bound waves.
In Fig. 7(a), the simulated free amplitude spectra of surface ele-
adj
vation at the first and the final iterations (ainit
BEM and aBEM , respec-
tively) are plotted. The target spectrum at is also given for
comparison. The corresponding deviations of the initial and final
phases from the target are depicted in Fig. 7(b). The peak and the
low-frequency parts of the initial spectrum have lower amplitudes
than those of the target, whereas high-frequency modes are overval-
ued. Such an asymmetry signifies the importance of nonlinear
effects accompanying waves’ excitation. The phase shift between
the surface elevation and the wavemaker motion within the range
of meaningful modes is close to π/2 but not constant. A slight
increase is observed for higher frequencies, in agreement with
Ref. [5].
For most frequencies, the adjusted amplitude spectrum agrees
well with the target one, Fig. 7(a). A slight deviation can be identi-
fied at high frequencies where bound waves dominate. The phases
of the frequency harmonics of the adjusted wave train for all signif-
icant modes are also in agreement with the target, Fig. 7(b), except
for relatively high frequencies. The observed disagreement with the
target values of both amplitudes and phases can be attributed to the
finite spatial and temporal resolution of the NWT that imposes lim-
itations on the accuracy of simulation of short (high-frequency)
waves. It may also be caused by the disregarded third- and higher-
order bound waves.
For experimental verification of the simulated results, the wave-
maker motion computed by NWT served as a driving signal. In
experiments, however, a significant deviation of the actual motion
of the wavemaker paddle from the driving signal was observed.
The complex spectrum of the driving signal was therefore corrected
to improve agreement of the actual wavemaker motion with the pre-
dicted one by the NWT.
061102-6 / Vol. 141, DECEMBER 2019
Fig. 7 Amplitude and phase adjustment for free components of
the surface elevation

max
Fig. 8 Adjustment of the numerically simulated and the actual
spectra of the wavemaker motion
The required correction of the driving signal is demonstrated in
Fig. 8. Here, Sinit
BEM represents the amplitude spectrum of the actual
paddle motion resulting from the initially computed driving
signal, whereas Sadj adj
BEM and ϕBEM correspond to the amplitudes and
phases of the paddle motion, respectively, at NWT after the final
adjustment of the required driving signal. The amplitudes and
phases of the driving signal that causes the required paddle motion
drv and ϕdrv . This correc-
corr
in experiments are depicted in Fig. 8 as Scorr
tion of the driving signal results in the root-mean-square deviation of
the actual motion from the designed one of less than 1%.
It is seen from Fig. 8(a) that the corrected amplitude spectrum
of the driving signal Scorr
drv is shifted to higher frequencies as com-
pared with the spectrum of the wavemaker motion. Thus, the trans-
fer function correction that relates the driving signal to the actual
paddle motion is frequency-dependent. The phase difference
adj
drv − ϕBEM is, however, almost constant within a wide frequency
ϕcorr
range remaining close to −π/2. It should be stressed that the devia-
tion of the driving signal from the actual paddle motion is a feature
of the particular wavemaker and, therefore, the importance of the
suggested driving signal correction may be wavemaker-dependent.
Upon completion of the driving signal correction procedure, the
surface elevation at the position of xt was measured, and the appro-
priate spectrum was calculated. The free waves were separated from
the bound components as described previously, and the results are
Transactions of the ASME
Fig. 9 The measured spectrum of free waves after correction of
the wavemaker motion
plotted in Fig. 9. Although a small discrepancy for the high-
frequency part remains, very good agreement is obtained at lower
frequencies. These results confirm the applicability of the suggested
method and NWT based on BEM for derivation of the driving
signal for generation of unidirectional broad-banded gravity waves.
Comparison of the simulated and the measured waveforms with
the target is carried out in Fig. 10. The plotted experimental and
numerical data contain free as well as bound components. The
target linear waveform containing only free waves was therefore
augmented by corresponding bound components computed using
expressions given in Refs. [16–18]. The experimental and numeri-
cal waveforms agree well with the resulting modified target surface
elevation variation with time.
Excitation of Waves With Lower Steepness (εc = 0.12). The
effect of the initial wave steepness on the accuracy of the generation
is now assessed. To this end, the focusing location of the Gaussian
wave train (11) was moved farther away from the wavemaker to
xf = 8.5 m, while retaining all other governing parameters including
the envelope amplitude η0. The linear dispersion causes spreading
of the wave train at the wavemaker with the corresponding reduc-
tion of the height of the steepest crest. The location of the wave
gauge and of the virtual probe remained unchanged.
As in the case of steep waves excitation for xf = 4 m, the surface
elevation at the wavemaker was transformed into the wavemaker
motion by application of the linear transfer function. The obtained
signal served at the initial iteration for the wavemaker boundary
condition in the NWT. The amplitude spectrum of the surface ele-
vation derived from the virtual probe output and decomposed into
free and bound modes is plotted in Fig. 11(a), while the wave
forms are given in Fig. 11(b). The amplitudes of the bound waves
Fig. 10 Comparison of experimental, numerical, and target wave forms
Journal of Offshore Mechanics and Arctic Engineering
Fig. 11 Separation of the surface elevation spectrum into free
and bound waves
Table 2 Convergence of the adjustment procedure for
medium-steep waves

 2 0.5  
ηtarg 
ηBEM − ηtarg
Iteration no.
max
1 40.5%
2 1.12%
3 0.67%
4 0.60%
5 0.59%
in Fig. 11(a) are notably lower than those in Fig. 6(a), as expected
for less steep waves. There is no overlapping of free and bound
waves in the spectral domain, thus enabling improved separation
of harmonics.
The convergence of the iteration procedure required for the deter-
mination of the driving signal in the NWT is illustrated in Table 2.
The accuracy is now nearly twice better than that in Table 1, but
four iterations are still needed to arrive at the optimal adjusted
wavemaker motion. The accuracy of the waves’ generation thus
strongly depends on the initial wave steepness, whereas the conver-
gence rate of the iterative method remains almost independent of ε.
Figure 12 demonstrates the variation of the free wave spectrum
during the iterative adjustment of the paddle motion. While the
peak and the low-frequency parts of the initial amplitude spectrum
are slightly underestimated, the high-frequency region is very close
DECEMBER 2019, Vol. 141 / 061102-7

Fig. 12 Amplitude and phase adjustment for free components
utilizing NWT
Fig. 13 Experimental spectrum of free waves after correction of
the wavemaker motion
to the target right from the beginning. This is the main difference
between excitation by the wavemaker of waves with different steep-
ness (cf., Figs. 7 and 12).
The initial phase difference in Fig. 12(b) is close to π/2 but varies
slightly, similar to Fig. 7(b). The phases of the adjusted wave train
are almost identical to those of the target within the whole range of
significant modes contrary to Fig. 7(b). This improvement can be
Fig. 14 Comparison of experimental, numerical, and target wave forms
061102-8 / Vol. 141, DECEMBER 2019
Fig. 15 Adjustment of (a) amplitudes and (b) phases in the
spectra of the wavemaker motion needed for excitation of wave
trains with different steepnesses
attributed to the reduced contribution of spurious waves, especially
at high frequencies.
The correction of the driving signal is still required for this
lower steepness case. The procedure applied is identical to that
described in relation to Fig. 8. The measured spectrum of free com-
ponents and the corresponding temporal variation of the surface
elevation comprising all free and bound modes are compared with
the target in Figs. 13 and 14, respectively. The convergence of
the measured records with the target improved noticeably as com-
pared with the case of steeper waves at the wavemaker (cf., Figs.
9 and 10).
Figure 15 compares the correction of the wavemaker motion
obtained as a result of the iterative adjustment in the NWT for
wave trains of different steepness. The corrections to both ampli-
tudes and phases of the wavemaker motion are almost independent
of εc in the vicinity of the peak frequency fp, cf., Figs. 15(a) and
15(b). However, a more significant correction is required for a
steeper wave train (εc = 0.21) at frequencies closer to the free
wave spectrum band denoted by vertical dashed lines in Fig. 15.
This observation agrees with the theory of Schäffer [2], where the
impact of steepness of the dominant free waves on the amplitudes
of the spurious propagating waves is reported.
Conclusions
Theoretical analysis of regular steep waves demonstrated that
the second-order contributions to the instantaneous velocity field
are negligibly small in deep water. This simplifies the iterative
Transactions of the ASME
adjustment of the wavemaker driving signal for generation of unidi- η, ψ = surface elevation and potential
rectional broad-banded surface gravity waves with the prescribed ω, k = angular frequency and wave number
nonlinear spectrum of surface elevation. Expressions obtained in η̂, ψ̂, û, ŵ = dimensionless amplitudes of surface elevation,
the course of derivation of the Zakharov model [15–18] are potential, horizontal, and vertical velocities
applied for decomposition of the surface elevation spectrum into
components according to their order of nonlinearity [10,11]. The
developed method allows considerable improvement of the numeri-
cal effectiveness of the computational procedure, including a higher Appendix: Analytic Relation Between First-Order Free
convergence rate of the iteration process and the wave generation and Second-Order Bound Components of the Surface
accuracy. The importance of verification of the wave generation Elevation Spectra According to Zakharov Theory
accuracy in the vicinity of the wavemaker should be stressed. Surface elevation due to each spectral component of the first-
The fully nonlinear numerical wave tank based on the boundary order free waves may be presented as follows:
element method served for the iterative adjustment of the wave-  
maker motion. The results of the numerical simulations agree (1) (1)
η(1) = Re a(1) ei(k x−ω t) (A1)
well with the experimental data. The duration of the simulations
was reduced by shortening the numerical wave tank and by imple- where a (1) is its complex amplitude. The Zakharov theory [15–18]
mentation of the numerical beach that effectively prevents wave
n , n = 1 … N,
relates the complex spectrum of the free harmonics a(1)
reflection from the far end. The artificial pressure incorporated to the complex amplitudes of the corresponding second-order bound
into the dynamic free surface boundary condition was found to be components. For every pair of first-order harmonics ωm(1) and ω(1) n
significantly more effective for damping the reflected waves in with complex amplitudes am(1) and a(1)n , three second-order bound
the numerical procedure as compared with the introduction of a waves appear with the following frequencies ωj , wave numbers
(2)
viscous dissipative term. It should be stressed that the application k(2) (2)
j , and complex amplitudes aj :
of the numerical wave tank for adjustment of the driving signal is
particularly important in large-scale experiments. 1 = ωm + ωn ; k1 = km + kn
• ω(2) (1) (1) (2) (1) (1)

The accuracy of the generation process was found to be strongly 

dependent on the wave steepness at the wavemaker. Considerable 2gω p,1 V1 (ω p,1 , ω(1)
m , ωn , k1 , km , kn ) (1) (1)
(1) (2) (1) (1)
deviation of the free wave spectra from the target was observed a1 = −π
(2)
am an
ω(1)
m ωn
(1)
ω p,1 − ω(1)
m − ωn
(1)
for steep waves with the characteristic steepness exceeding εc ≈
0.2. Reduction of wave steepness at the wavemaker to εc ≈ 0.12 (A2)
substantially reduces the contribution of unwanted waves. The iter-
ative adjustment of the driving signal applied in the present study
demonstrated its effectiveness in the generation of the waves that • ω(2)
2 = −ωm + ωn ; k2 = −km + kn
(1) (1) (2) (1) (1)

are steep in the vicinity of the wavemaker. 

2gω p,2 V2 (ω p,2 , ω(1)
m , ωn , k2 , km , kn ) (1) * (1)
(1) (2) (1) (1)
a2 = −π
(2)
(am ) an
Funding Data ωm ωn
(1) (1)
ω p,2 + ω(1)
m − ωn
(1)

• Israel Ministry of Science, Technology and Space (Grant No. (A3)

3-12473).
• ω(2)
3 = −ωm − ωn ; k3 = −km − kn
(1) (1) (2) (1) (1)

Nomenclature 
h = mean water depth 2gω p,3 V3 (ω p,3 , ω(1)
m , ωn , k3 , km , kn ) (1) * (1) *
(1) (2) (1) (1)
a3 = −π
(2)
(am ) (an )
xf = focusing location of the wave train relative to the ωm ωn
(1) (1)
ω p,3 + ω(1)
m + ωn
(1)

wavemaker
(A4)
xt = location of the wavegauge in the experimental and
numerical wave tanks
a, ϕ = amplitudes and phases of the surface elevation
Here * denotes complex conjugate. In (A2)–(A4), the angular fre-

Sdrv, ϕdrv = amplitudes and phases of the driving signal
Swm, ϕwm = amplitudes and phases of the wavemaker motion quency ω p,j = gkj(2) tanh (kj(2) h) corresponds to that of a free wave
εc = wave steepness calculated as a product of the with the wave number k(2) j . The radian frequencies ωp,j do not
amplitude of the wave train envelope and the carrier appear in the bound waves’ spectrum and are used to calculate
wave number the coefficients:

     

1 g 1 ω0 ω1 ω2 2 ω2 ω0 ω1 2
V1 (ω0 , ω1 , ω2 , k0 , k1 , k2 ) = + k1 k2 − − k0 k1 (A5)
4π 2 2 ω1 ω2 g ω0 ω1 g

      

1 g ω2 ω0 ω1 2 ω1 ω0 ω2 2 ω0 ω1 ω2 2
V2 (ω0 , ω1 , ω2 , k0 , k1 , k2 ) = + k0 k1 − − k0 k2 − − k1 k2 (A6)
4π 2 ω0 ω1 g ω0 ω2 g ω1 ω2 g

    
1 g ω2 ω0 ω1 2 1 ω0 ω1 ω2 2
V3 (ω0 , ω1 , ω2 , k0 , k1 , k2 ) = + k0 k1 + + k1 k2 (A7)
4π 2 ω0 ω1 g 2 ω1 ω2 g

Journal of Offshore Mechanics and Arctic Engineering DECEMBER 2019, Vol. 141 / 061102-9
References
[1] Dean, R. G., and Dalrymple, R. A., 1991, Water Wave Mechanics for Engineers
and Scientists, World Scientific, Singapore.
[2] Schäffer, H. A., 1996, “Second-Order Wavemaker Theory for Irregular Waves,”
Ocean Eng., 23, pp. 47–88.
[3] Spinneken, J., and Swan, C., 2009, “Second-Order Wave Maker Theory Using
Force-Feedback Control. Part I: A New Theory for Regular Wave Generation,”
Ocean Eng., 36(8), pp. 539–548.
[4] Spinneken, J., and Swan, C., 2009, “Second-Order Wave Maker Theory Using
Force-Feedback Control. Part II: An Experimental Verification of Regular
Wave Generation,” Ocean Eng., 36(8), pp. 549–555.
[5] Aknin, D., and Spinneken, J., 2017, “A Laboratory Investigation Concerning the
Superharmonic Free Wave Suppression in Shallow and Intermediate Water
Conditions,” Coastal Eng., 120, pp. 112–132.
[6] Houtani, H., Waseda, T., Fujimoto, W., Kiyomatsu, K., and Tanizawa, K., 2015,
“Freak Wave Generation in a Wave Basin With HOSM-WG Method,” ASME
Paper No. OMAE2015-42284.
[7] Houtani, H., Waseda, T., Fujimoto, W., Kiyomatsu, K., and Tanizawa, K., 2018,
“Generation of a Spatially Periodic Directional Wave Field in a Rectangular Wave
Basin Based on Higher-Order Spectral Simulation,” Ocean Eng., 169, pp. 428–441.
[8] Chaplin, J. R., 1996, “On Frequency-Focusing Unidirectional Waves,”
Int. J. Offshore Polar Eng., 6(2), pp. 131–137.
[9] Lugni, C., 2000, “An Investigation on the Interaction Between Free-Surface
Waves and Floating Structure,” Ph.D. thesis, University of Rome, Italy (in Italian).
[10] Shemer, L., Goulitski, K., and Kit, E., 2006, “Steep Waves in Tanks: Experiments
and Simulations,” ASME Paper No. OMAE2006–92547.
[11] Shemer, L., Goulitski, K., and Kit, E., 2007, “Evolution of Wide-Spectrum
Unidirectional Wave Groups in a Tank: An Experimental and Numerical
Study,” Eur. J. Mech. B Fluids, 26(2), pp. 193–219.
061102-10 / Vol. 141, DECEMBER 2019
[12] Schmittner, C., Kosleck, S., and Hennig, J., 2009, “A Phase-Amplitude Iteration
Scheme for the Optimization of Deterministic Wave Sequences,” ASME Paper
No. OMAE2009-80131.
[13] Buldakov, E., Stagonas, D., and Simons, R., 2017, “Extreme Wave Groups in a
Wave Flume: Controlled Generation and Breaking Onset,” Coastal Eng., 128,
pp. 75–83.
[14] Mei, C. C., 1989, The Applied Dynamics of Ocean Surface Waves, World
Scientific, Singapore.
[15] Zakharov, V. E., 1968, “Stability of Periodic Waves of Finite Amplitude on the
Surface of a Deep Fluid,” J. Appl. Mech. Tech. Phys., 9, pp. 190–194.
[16] Stiassnie, M., and Shemer, L., 1984, “On Modification of Zakharov Equation for
Surface Gravity Waves,” J. Fluid Mech., 143, pp. 47–67.
[17] Stiassnie, M., and Shemer, L., 1987, “Energy Computations for Evolution of
Class I and II Instabilities of Stokes Waves,” J. Fluid Mech., 174, pp. 299–312.
[18] Krasitskii, V. P., 1994, “On the Reduced Equations in the Hamiltonian Theory of
Weakly Nonlinear Surface Waves,” J. Fluid Mech., 272, pp. 1–20.
[19] Grilli, S. T., and Svendsen, I. A., 1990, “Corner Problems and Global Accuracy in
the Boundary Element Solution of Nonlinear Wave Flows,” Eng. Anal. Bound.
Elem., 7(4), pp. 178–195.
[20] Grilli, S. T., and Subramanya, R., 1996, “Numerical Modeling of Wave
Breaking Induced by Fixed or Moving Boundaries,” Comput. Mech., 17,
pp. 374–391.
[21] Grilli, S. T., Skourup, J., and Svendsen, I. A., 1989, “An Efficient Boundary
Element Method for Nonlinear Water Waves,” Eng. Anal. Bound. Elem., 6(2),
pp. 97–107.
[22] Grilli, S. T., and Horrillo, J., 1997, “Numerical Generation and Absorption of
Fully Nonlinear Periodic Waves,” J. Eng. Mech., pp. 1060–1069.
[23] Tian, Z., Perlin, M., and Choi, W., 2010, “Energy Dissipation in Two-
Dimensional Unsteady Plunging Breakers and an Eddy Viscosity Model,”
J. Fluid Mech., 655, pp. 217–257.
Transactions of the ASME