You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245293139

Computation of Short-Crested Deepwater Waves

Article  in  Journal of Waterway Port Coastal and Ocean Engineering · May 2006


DOI: 10.1061/(ASCE)0733-950X(2006)132:3(157)

CITATIONS READS
15 60

5 authors, including:

Makoto Okamura Christian Kharif


Kyushu University Institut de Recherche sur les Phénomènes Hors Equilibre
45 PUBLICATIONS   285 CITATIONS    182 PUBLICATIONS   3,866 CITATIONS   

SEE PROFILE SEE PROFILE

A. J. Roberts
University of Adelaide
223 PUBLICATIONS   2,508 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Modelling shallow turbulent fluid dynamics and thin 3D fluid flows View project

Multiscale modelling of materials with fine structures View project

All content following this page was uploaded by Christian Kharif on 18 November 2014.

The user has requested enhancement of the downloaded file.


Computation of Short-Crested Deepwater Waves
M. Ioualalen1; M. Okamura2; S. Cornier3; C. Kharif4; and A. J. Roberts5

Abstract: Short-crested waves are three-dimensional waves that may be generated through a reflection of a two-dimensional Stokes wave
on a seawall. Thus, they are more likely to be observed near ports or any consequent marine structure. Two numerical methods are used
to compute three-dimensional surface gravity short-crested waves on deepwater. The first method is an asymptotic procedure and the
second computes a direct numerical solution. One of the main properties is the four-wave resonance. Such resonance introduces nonu-
niqueness with several solution branches connected through a turning point. We show that both computational methods are reliable for
nonresonant waves, but that the direct numerical method converges faster. For resonant waves, the direct method is more appropriate
because all solution branches can be obtained. The asymptotic method computes only one branch of solutions for any given parameter
values, and is uncertain around and past any turning point. Stability analysis of the branches shows that, although sporadic, an instability
associated with harmonic resonance is more likely to appear for one branch in the vicinity of the turning point. Consequently this could
amplify the unstable resonant mode. The nonuniqueness of the solution requires careful attention in every study on the impact of surface
waves on marine structures. It is shown here that the wave force exerted on a seawall may change drastically from one branch of solution
to another.
DOI: 10.1061/共ASCE兲0733-950X共2006兲132:3共157兲
CE Database subject headings: Sea walls; Resonance; Three-dimensional analysis; Waves forces; Water crest.

Introduction Ly = L / cos ␪ in the transverse direction y, where L is the incident


wavelength. The free surfaces displayed in Fig. 2 represent
Short-crested waves 共SCWs兲 are three-dimensional 共3D兲 water weakly 3D waves 共␪ = 80°兲 and fully developed 3D waves
waves first computed by Hsu et al. 共1979兲 up to the third order in 共␪ = 45°兲. SCWs admit two 2D limits: standing waves for ␪ = 0°
amplitude. The solutions were extended to 27th order for deep and progressive Stokes waves for ␪ = 90°.
water by Roberts 共1983兲 who computed their extreme forms and One main aspect of interest of SCWs is the appearance of
main properties. Later Roberts and Schwartz 共1983兲 and Okamura harmonic resonances. SCWs Eq. 共AII.7兲 may be described
共1996兲 computed fully numerical SCW solutions. SCWs can be through the synthetic form ␾tt + ␻2␾ = nonlinear forcing. The ve-
generated by the reflection of a two-dimensional 共2D兲 wave train locity potential ␾共x , y , z , t兲 may contain terms of the form
共crest of infinite length兲 arriving at a vertical seawall at some F j,k共x , y , t兲 = cos关jx − ␻t兴cos共ky兲, with j and k integers, that are
angle of incidence ␪ 共Fig. 1兲. This mode of propagation may solutions of the associated homogeneous equation. Physically,
generate a flow around vertical marine structures such as reflec- these modes propagate at the same velocity ␻ as the fundamental
tion from a wedge 共Marchant and Roberts 1990兲 representing an of the wave. The derived secularities and the associated phase
ideal ship’s bow. The reflection generates 3D waves of wave-
lengths Lx = L / sin ␪ in the direction of propagation x and

1
Research Scientist, Institut de Recherche pour le Développement,
Géosciences Azur, l’Observatoire Océanologique de Villefranche-Sur-
Mer, La Darse, B.P. 48, 06230 Villefranche-Sur-Mer F06235, France
共corresponding author兲. E-mail: mansour.ioualalen@geoazur.unice.fr
2
Associate Professor, Research Institute for Applied Mechanics,
Kyushu Univ., Kasuga, Fukuoka 816-8580, Japan.
3
M.D. Student, Institut de Recherche pour le Développement,
Géosciences Azur, l’Observatoire Océanologique de Villefranche-Sur-
Mer, La Darse, B.P. 48, 06230 Villefranche-Sur-Mer F-06235, France.
4
Professor, Institut de Recherche sur les Phénomènes Hors Equilibre,
49 rue F. Joliot-Curie, B.P. 146, F13384 Marseille Cedex 13, France.
5
Professor, Dept. of Mathematics and Computing, Univ. of Southern
Queensland, Toowoomba, Queensland 4350, Australia.
Note. Discussion open until October 1, 2006. Separate discussions
must be submitted for individual papers. To extend the closing date by
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos-
sible publication on August 31, 2004; approved on April 27, 2005. This
paper is part of the Journal of Waterway, Port, Coastal, and Ocean
Fig. 1. Short-crested wave formation. Reflection of plane wave onto
Engineering, Vol. 132, No. 3, May 1, 2006. ©ASCE, ISSN 0733-950X/
2006/3-157–165/$25.00. vertical seawall at angle ␪ of incidence.

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006 / 157
locking generate the so-called harmonic resonances 共Roberts Formulation of Wave Propagation
1983兲. First consider marine structures: since the resonances ap-
pear for reflected waves they may have significant impact on the Consider an irrotational free surface flow on deep water of an
structures themselves. This should require attention for any engi- inviscid and incompressible fluid subject to gravity g. The frame
neering study on wave/structure interactions. Several studies have of reference R共O , x , y , z , t兲 is chosen such that the mean level of
already explored the forces exerted by SCWs on vertical seawalls the air–sea interface is the horizontal frame of coordinates
as design criterion for marine structures. Fenton 共1985兲 examined 共O , x , y兲. The vertical axis 共O , z兲 is oriented upward. The irrota-
the uniform SCW problem to third order and obtained the depth tional flow implies a velocity potential ␾共x , y , z , t兲 where t repre-
integrated wave forces and wave moment on the wall. Later sents time. The equation of mass conservation becomes the
Marchant and Roberts 共1987兲 via a computer-generated perturba- Laplace equation
tion expansion investigated the uniform wave to 35th order in
wave steepness and extended Fenton’s work to oblique incident ⵜ 2␾ = 0 for z 艋 ␩共x,y,t兲 共1兲
wave in shallow water. However, neither study computed fully where ␩共x , y , t兲 = free surface elevation. The solutions of
resonant SCW solutions. Second, consider operational physics: Laplace’s Eq. 共1兲 must satisfy a nonlinear dynamical condition 共2兲
the characteristics of a surface wave field provide information of constant free surface pressure derived from the Bernoulli equa-
about the transfer of momentum across the sea surface 共e.g., the tion, a nonlinear kinematic condition 共3兲 assuming that no par-
wind stress兲. If the resonances, due to intrinsic wave–wave inter- ticles of fluid crosses the surface, and a linear bottom condition
actions, affect the sea state on the same order of time scale as any 共4兲. On z = ␩共x , y , t兲, the dynamical condition is
atmospheric forcing, then the resonances may affect the momen-
tum transfer to the wave field. 1 pa
g␩ + ␾t + 兩ⵜ␾兩2 = −
Here we compare two methods to compute SCW fields, and 2 ␳
chose for that purpose two modes of propagation. The first con-
figuration describes a resonant propagation for which significant where pa = atmospheric pressure and ␳ = density.
harmonic resonance appears. The second mode exhibits a non- As is common, we include the supposed constant expression
resonant propagation for which no significant harmonic resonance pa / ␳ within the expression of the velocity potential ␾
appears. 1
g␩ + ␾t + 2 兩ⵜ␾兩2 = 0 on z = ␩共x,y,t兲 共2兲

␩ t + ␾ x␩ x + ␾ y ␩ y − ␾ z = 0 on z = ␩共x,y,t兲 共3兲

␾z = 0 for z → − ⬁ 共4兲
where Eq. 共2兲 is the dynamical; Eq. 共3兲 is the kinematic; and Eq.
共4兲 the bottom conditions.
In order to simplify the formulation, we write the system
of Eqs. 共1兲–共4兲 in nondimensional form where k−1 and
共gk兲−1/2 = length and time of reference, respectively, and k = wave
number. The new variables of the problem are: x̃ = kx, ỹ = ky,
z̃ = kz, t̃ = 共gk兲1/2t, ␾ ˜ 共x , y , z , t兲 = 关k2 / 共gk兲1/2兴␾共x , y , z , t兲, and
˜␩共x , y , t兲 = k␩共x , y , t兲. Henceforth omitting the tildes, the follow-
ing equations become:
ⵜ 2␾ = 0 for z 艋 ␩共x,y,t兲 共5兲

1
␩ + ␾t + 2 兩ⵜ␾兩2 = 0 on z = ␩共x,y,t兲 共6兲

␩ t + ␾ x␩ x + ␾ y ␩ y − ␾ z = 0 on z = ␩共x,y,t兲 共7兲

␾z = 0 for z → − ⬁ 共8兲
We look for 3D permanent doubly periodic solutions propagat-
ing at a phase velocity c = ␻ / k in the new frame of reference
R⬘共O⬘ , x1 , y 1 , z1 , t兲 moving with the wave, where ␻ = frequency;
x1 = m共x − ct兲; y 1 = ny; and z1 = z. m = sin ␪ and n = cos ␪ = wave
numbers in the x and y directions, respectively.
For simplicity, new coordinates x1, y 1, and z1 are written x, y,
and z, respectively. In this new frame of reference the following
equations become:
m2␾xx + n2␾yy + ␾zz = 0 for z 艋 ␩共x,y兲 共9兲

1
␩ − ␻␾x + 2 共m2␾2x + n2␾2y + ␾z2兲 = 0 on z = ␩共x,y兲 共10兲
Fig. 2. Short-crested wave fields for angles ␪ = 80° 共upper panel兲 and
␪ = 45° 共lower panel兲 ␩ t + m 2␾ x␩ x + n 2␾ y ␩ y − ␾ z = 0 on z = ␩共x,y兲 共11兲

158 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006
Table 1. Resonance Angles ␪HR up to Order 27 of Truncation angles correspond to harmonic resonances of the linear wave
k j=1 j=2 j=3 j=4 j=5 共h = 0兲 that generate zero divisors in the asymptotic expansion.
For j and k → + ⬁, we reasonably conjecture the resonance angles
3 90.0000° — — — — are dense and, consequently, the radius of convergence of the
4 — 00.0000° — — — expansion should vanish at all angles. However, for most incident
5 90.0000° — — — — angles, the resonances appear at extremely high orders, and thus
6 — 52.2388° — — — they should be dissipated quickly since the high frequency modes
7 90.0000° — — — — are damped quickly with viscosity 共not included here兲. Thus, the
8 — 63.4349° — — — higher order resonances should be ignored for many practical
9 90.0000° — 00.0000° — — purposes. In an asymptotic expansion, the zero divisor problem
10 — 69.2952° — — — occurs with harmonic resonance. This problem cannot be avoided
11 90.0000° — 36.6992° — — near wave steepness h*, where h* is the wave steepness of
12 — 72.9761° — — — the nearest appearance of a noticeable resonance. Note that
13 90.0000° — 47.8696° — — resonances 共j = 1 , k ⬎ 2兲 are artificial. They appear for angle
14 — 75.5225° — — — ␪HR = 90° only and they can be removed through nonsecularity
15 90.0000° — 54.7356° — —
conditions 共Hsu et al. 1979兲.
Note that the triangular symmetry of SCWs 关␩共x , y兲 = ␩共y , x兲兴
16 — 77.3956° — 00.0000° —
excludes all mixed parity modes in the waves. Thus, any fre-
17 90.0000° — 59.5296° — —
quency correction occurs at odd orders only. In the following we
18 — 78.8342° — 28.0260° —
use either the wave steepness or the squared wave steepness
19 90.0000° — 63.1108° — — H = h2 to measure the amplitude of the SCWs.
20 — 79.9750° — 37.7612° — We analyze two configurations of SCWs. The first is a reso-
21 90.0000° — 65.9052° — — nant case 共␪ = 53°兲 which is close to the critical resonance angle
22 — 80.9026° — 44.2654° — ␪HR = 52.2388° that is due to the 共2,6兲 resonance. The second is a
23 90.0000° — 68.1546° — — nonresonant case 共␪ = 47.5°兲 which is far from all the angles of
24 — 81.6719° — 49.1066° — Table 1 共at least for orders lower than 27兲.
25 90.0000° — 70.0084° — 00.0000°
26 — 82.3206° — 52.9133° —
Solutions for Resonant Waves
27 90.0000° — 71.5651° — 22.6046°

Asymptotic Method
␾z = 0 for z → − ⬁ 共12兲 The asymptotic method was described by Roberts 共1983兲. Once
the first order linear solutions are chosen analytically, the
We look for doubly periodic solutions of permanent form like
␩ = 兺Nj,ka j,k cos共jx兲cos共ky兲 and ␾ = 兺Nj,kb j,k sin共jx兲cos共ky兲e关␣ j,kz兴.
The wave steepness is defined as half of the peak-to-trough height Table 2. Ratio between Two Successive Coefficients of Variables b1,1,
h = 1 / 2 关␩共0 , 0兲 − ␩共␲ , 0兲兴. bc2,6, and ␻ at Angle ␪ = 53° up to Order 25
The governing Eqs. 共9兲–共12兲 are solved using two methods: Order r br,1,1 / br+2,1,1 br,2,6 / br+2,2,6 ␻r−1 / ␻r+1
the first approach is the asymptotic method of Roberts 共1983兲; the 1 −21.66468000 — 5.563010216
second approach is the direct fully numerical method of Okamura
2 — — —
共1996兲.
3 0.1330347360 — 6.642493725
In the first approach 共Roberts 1983兲, the permament quantities
4 — — —
␩共x , y兲, ␾共x , y , z兲, and ␻ = cm = c sin ␪ are developed into succes-
sive powers of the small parameter h of the forms: 5 0.3587374090 — 1.434847593
⬁ ⬁ ⬁ 6 — 0.0327123106 —
␩共x , y兲 = 兺r=1 ␩rhr, ␾共x , y , z兲 = 兺r=1 ␾rhr, and ␻ = 兺r=0 ␻rhr. Roberts
共1983兲 looked for symmetrical solutions about the crest of the 7 0.0543535426 — 0.06791964918
wave 共x , y兲 = 共0 , 0兲. First order 共linear兲 solutions are ␻0 = 1, 8 — 0.0363684744 —
␩1 = cos共x兲cos共y兲, and ␾1 = sin共x兲cos共y兲ez. This initial choice im- 9 0.0371581800 — 0.04317144677
poses higher order solutions to be of the form: 10 — 0.0364479609 —
␩r = 兺rj,kar,j,k cos共jx兲cos共ky兲 and ␾r = 兺rj,kbr,j,k sin共jx兲cos共ky兲e␣ j,kz, 11 0.0364905782 — 0.03690255061
with ␣2j,k = 共jm兲2 + 共kn兲2. 12 — 0.0364600793 —
13 0.0364528857 — 0.03652715310
14 — 0.0364630595 —
Harmonic Resonance Appearance 15 0.0364524908 — 0.03650006652
16 — 0.0364653878 —
When merging Eqs. 共2兲 and 共3兲 for deep water, Roberts 共1983兲 17 0.0364545994 — 0.03650023788
showed that harmonic resonances appear when the right-hand 18 — 0.0364676677 —
side of the derived oscillator equation contains terms like 19 0.0364568569 — 0.03650243208
sin共jx兲cos共ky兲 which are solution of the homogeneous system. In 20 — 0.0364699475 —
practice, Roberts 共1983兲 found that a 共j , k兲 resonance appears for
21 0.0364591330 — 0.03650474921
critical angles satisfying cos2 ␪HR = 共j4 − j2兲 / 共k2 − j2兲.
22 — 0.0364722311 —
Table 1 gives angles ␪HR of incidence of the occurrence of
23 0.0364614092 — 0.03650708497
harmonic resonance in deep water up to order 27. These critical

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006 / 159
Table 3. Poles and Zeros of 关6 / 6兴 Padé Approximant for Frequency ␻ for Angle ␪ = 53°
Zeros Poles
Real part Imaginary part Real part Imaginary part
0.2718561935E + 02 0.0000000000E + 00 0.4525453091E + 01 0.0000000000E + 00
0.5183264613E + 00 0.0000000000E + 00 0.5198022723E + 00 0.0000000000E + 00
−0.4868999124E + 00 0.1376961768E + 00 −0.4870205522E + 00 0.1377589107E + 00
−0.4868999124E + 00 −0.1376961768E + 00 −0.4870205522E + 00 −0.1377589107E + 00
0.3643690422E − 01 0.2839575463E − 03 0.3643691540E − 01 0.2837929933E − 03
0.3643690422E − 01 −0.2839575463E − 03 0.3643691540E − 01 −0.2837929933E − 03
Note: Complex conjugate values are in boldface.

solutions are determined to higher and higher orders until the divisions on the coefficients bi⬎6,2,6 共for orders i ⬎ 6兲 contami-
asymptotic expansion is truncated. Then, the Shanks transform or nates rapidly the 共2,6兲 harmonic coefficient b2,6. Consequently,
Padé approximation 共Gilewicz 1978; Bender and Orzag 1981兲, past order 13, the radius of convergence for this coefficient is well
accelerates the convergence of the approximate solutions within determined. The prominence of the harmonic resonance 共2,6兲 is
the wave steepness h ⬍ h*, and converges past h*. However, it is confirmed by the behavior of the other variables in Table 2.
difficult to validate the solutions near h*, the resonance singular- In Table 3 we focus on the physical significance of real posi-
ity. The values of h* are close to the poles of the corresponding tive zeros and poles. Complex conjugate values also appear in the
Padé approximation. table 共in bold兲. However, the combination of the two values yields
Padé approximation of the ␻共H兲 = 共c0 + c1H + . . . + c M+NH M+N兲 two real values, that is, H = 0.03672 共h = 0.1916兲 and
truncated series constructs a rational function 关M / N兴␻ H = 0.03615 共h = 0.1901兲. Thus, we obtain two real values closely
共H兲 = 共p0 + p1H + . . . + p M H M 兲 / 共1 + q1H + . . . + qNHN兲 of M and N surrounding the value h = 0.1910.
order polynomials. The guiding principle is that the truncated Note the gaps between zeros-poles pairs are relatively small
series should be identical to the asymptotic expansion of the ra- and thus nearly cancel each other, suggesting that the harmonic
tional function up to order M + N. Thus, Padé approximation resonance is weak. Also, the real and positive pole H = 0.5198
places poles as well as zeros in and near the domain of interest. 共h = 0.721兲 may have a physical meaning. It could be related to
Consequently, the transformation of the expansion into a Padé the maximum load of the wave at this critical angle 共Roberts
fraction allows a continuous matching of the solutions before and 1983兲. The cone-like structure of extreme waves could lead to
past the poles incurred by harmonic resonance. The value of the singularities. This analysis allows a better understanding of the
closest pole estimates the radius of convergence of the asymptotic behavior of the truncated series solutions of SCWs when high
expansion, to within a certain range of precision due to truncation order harmonic resonances are voluntarily ignored.
itself. We now demonstrate the benefits of using the Shanks trans-
First, we compute 25th order solutions in order to illustrate the form to improve convergence of the expansion solution. Then, the
correspondence between the Padé approximation poles and the effects of higher order harmonic resonances are studied through
apparent radius of convergence of the expansions. the nature of the convergence, or otherwise, on the truncated
Table 2 illustrates what could be interpreted as a pseudoradius expansion.
of convergence of the expansion, for an increasing order, if no From the sixth order, the 共2,6兲 resonance contaminates all of
resonance appears past order 25. The ratio br,2,6 / br+2,2,6 of the the ␻ expansions 共Table 4兲. Without using the Shanks transform,
velocity potential harmonics 共2,6兲 converges rapidly to an ap- and from order 21, the frequency reaches an abnormal maximum
proximate value H = h2 = 0.03647 共h = 0.1910兲. At sixth order in that illustrates the divergence of the expansion past the radius of
the expansion the first small divisor directly amplifies the coeffi- convergence. For lower orders of truncation the maximum of fre-
cient b6,2,6 due to the proximity of the resonance angle ␪HR. Then, quency appears at higher wave steepness for which the poles are
the affected coefficient and the successive effects of the small larger 共not shown here兲. At low orders, we do not obtain a maxi-

Table 4. Asymptotic Method. Convergence of Frequency ␻ for Different Orders of Truncation at Angle ␪ = 53°.
h Order 5 Order 9 Order 13 Order 17 Order 21 Order 25 Order 29
共a兲 Without use of the Shanks transform
0.10 1.00180029 1.00180027 1.00180027 — — — —
0.15 1.00405827 1.00405799 1.00405793 1.00405790 1.00405790 1.00405789 1.00405789
0.20 1.00723365 1.00723173 1.00723036 1.00722872 1.00722675 1.00722439 1.00722155
0.25 1.01134063 1.01133179 1.01131527 1.01126705 1.01112569 1.01071131 1.00949697
0.30 1.01639749 1.01636552 1.01623491 1.01544390 1.01063485 0.981404522 0.803782573
共b兲 With use of the Shanks transform
0.10 1.00180030 1.00180027 1.00180027 — — — —
0.15 1.00405832 1.00405795 1.00405789 1.00405789 — — —
0.20 1.00723391 1.00723071 1.00723958 1.00723851 1.00723851 1.00723844 1.00723844
0.25 1.01134164 1.01128293 1.01133973 1.01134022 1.01134023 1.01134021 1.01134021
0.30 1.01640050 1.01643979 1.01638441 1.01639058 1.01639055 1.01639056 1.01639055

160 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006
Fig. 3. Frequency ␻ at order 21 of truncation as function of wave Fig. 5. Same as Fig. 4 for coefficient b1,1
steepness h, computed with asymptotic method: 共〫兲 solutions with
use of Shanks transform through ␧ algorithm procedure and 共쎲兲
without Shanks transform. Window panel is enlargement of figure
around pole.
converged solution is found 共if there is convergence兲; the conver-
gence criterion is that the relative error is less than 10−10 between
mum because the contamination by the small divisor is weak and two successive iterations. The second order approximation of Hsu
at fifth order the 共2,6兲 resonance has not appeared. Past the critical et al. 共1979兲 is the initial estimate for the Newton iteration. Then,
value of h = 0.1910, the asymptotic solutions do not converge. the initial estimate for another solution at a slightly different set
Below that value and for a given order r of truncation, the error of parameter values comes from the previous solution, provided
induced by neglecting higher orders is of order hr. the two wave steepnesses of the two successive computations are
The Shanks transform processed successfully a continuous close enough. Figs. 4 and 5 illustrate the behavior of the coeffi-
matching across the resonance since there is no decrease of the cients b2,6 and b1,1 as a function of the wave steepness parameter
frequency past h = 0.1910. Past this critical value and for a given computed at order 21 like the asymptotic method. Past a critical
order r of truncation, the error induced by neglecting higher or- wave steepness, close to the pole computed above for ␻, the
ders seems to be of order hr. However, this estimate relies on the solutions form three branches. Branch 共3兲 is equivalent to the
continuous matching being reasonable. This is not necessarily solution computed with the asymptotic method complemented
true because although the Shanks transform can generate con- with the Shanks transform. The two other branches, not accessible
verged answers, it may introduce a mathematical artefact as we with the asymptotic method, are linked through a turning point
see later. Fig. 3 illustrates the above discussion. In particular, it 共TP兲 whose value for ␻ is close to the value of the pole computed
shows how the Shanks transform places a pole into the truncated above for ␻. Although the behavior of the two coefficients is
series and thus exhibits singular behavior around the pole. comparable, they differ from each other in their alignment and
slope.
Fig. 6 illustrates the frequency ␻ as a function of the wave
Direct Numerical Method steepness. The frequency is influenced by the whole spectrum of
The numerical method, described by Okamura 共1996兲, solves the the wave mode coefficients bi,j. It also shows the three branches.
system of Eqs. 共9兲–共12兲 through an iterative Newton procedure. Table 5 illustrates the convergence of the branch solutions of
Similarly to the asymptotic approach, the order of truncation in the frequency ␻ as a function of the order of truncation and the
the wave’s harmonics is fixed and the iteration continued until a wave steepness. For Branch 共3兲, up to a wave steepness value
located around the pole, the convergence is similar to the
asymptotic method. Above that critical value, the convergence
deteriorates compared to the asymptotic approach. However, we
attribute more reliance to these slowly converged solutions than
to the converged solutions of the asymptotic method. The reason
is that convergence is attained with the asymptotic method mainly
through the use of the Shanks transform which is somewhat arti-
ficial and does not directly rely on the physical formulation of the
problem, whereas the direct numerical method solves the govern-
ing equations. Thus the precision of the asymptotic method, stated
above, cannot be validated, whereas the precision of the direct
method may be validated and is of order hr/3. The precision is
defined here as the error made in the solution at order r when
ignoring all higher orders. This value gives an idea of the impor-
tance of the high frequency mode and thus on the behavior of the
physical oscillation of the wave.
Branch 共2兲 appears from the sixth order of truncation due to
Fig. 4. Coefficient b2,6 as function of wave steepness h for direct the resonance 共2,6兲 and appears above a value of the wave steep-
numerical method ness close to the pole. Note that convergence is rapid at very low

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006 / 161
Solutions for Nonresonant Waves

Consider SCWs at angle ␪ = 47.5° where no harmonic resonance


is likely to appear until order 27 at least and, thus, only Branch
共3兲 is present. Table 6 shows that the good convergence of the
asymptotic method is obtained at a lower order of truncation than
for resonant waves. The lack of small divisors yields rapidly de-
caying coefficients in the expansions for the free surface eleva-
tion, the velocity potential, and the frequency; the contribution of
high order coefficients weakens quickly. The use of the Shanks
transform may provide two benefits: the first one is its ability to
process a continuous matching around the pole 共increase of the
radius of convergence兲 which is not needed here for nonresonant
waves; the second is to accelerate the convergence. Here, since
Fig. 6. Same as Fig. 4 for frequency ␻
the coefficients decrease quickly, the correspondence between the
truncated time series and a Padé approximant is obtained rapidly.
Thus no real benefit is obtained in using the Shanks transform.
orders. Thus this branch is related to relatively low frequency The direct method 共Table 6兲 is more efficient than the
modes and seems to be related directly to the mode 共2,6兲 that asymptotic one. For the asymptotic method, each coefficient of
oscillates quasifreely along this branch. one given harmonic is a truncated time series composed of a finite
Branch 共1兲 occurs similarly to Branch 共2兲 but convergence of number of subcoefficients of powers of the wave steepness. Thus,
this branch, to a comparable precision with the other branches, is for a specified desired accuracy, the coefficient needs more orders
difficult to attain. Figs. 4 and 5 show that this branch is very steep of calculation to verify the solution of the system of Eqs. 共9兲–共12兲
compared to the other branches. Consequently, past h = 0.20, the for the asymptotic method than for the direct method which pro-
Newton procedure diverges because efficient initial solutions on vides a single coefficient and thus verifies strictly the system of
this branch would require very close h values for consecutive equations.
computations, namely better than ⌬h = 10−5. This would require
many numerical runs to obtain one single solution. This is not the
primary objective of the present work. Contrary to the other two
branches, Branch 共1兲 solutions are less accurate. As a result, high Stability of Resonant Waves
frequency modes 共high order of truncation兲 play a crucial role on
that branch. We explore the linear stability of SCW solutions for the three
To conclude this section, the three branches behave distinctly. branches in order to better understand their mode of occurrence.
Branch 共3兲 is composed of all superharmonic modes with a de- We use the stability analysis developed by Ioualalen and Kharif
creasing influence of higher order modes. Branch 共2兲 is dedicated 共1993, 1994兲. Briefly, perturbations of the following form are
to the mode 共2,6兲 exclusively as a free mode or, at least, this added to the SCW solutions. Then, the full system of Eqs.
branch seems to ignore high order modes. Branch 共1兲, to the 共9兲–共12兲 is solved in a frame of reference propagating with the
contrary, is strongly affected by high order modes. wave and yields an eigenvalue problem

Table 5. Direct Numerical Method. Convergence of Branch 共3兲 of Frequency ␻ as Function of Wave Steepness at Angle ␪ = 53°.
h Order 5 Order 9 Order 13 Order 17 Order 21 Order 25 Order 29
共a兲 Branch 共3兲
0.10 1.00180027 1.00180027 — — — — —
0.15 1.00405804 1.00405789 1.00405789 — — — —
0.20 1.00723228 1.00758263 1.00718167 1.00718381 1.00717439 1.00717414 1.00717420
0.25 1.01133500 1.01461317 1.01287077 1.01274044 1.01181071 1.01176500 NCa
0.30 1.01637892 1.02184724 NCa 1.01649392 1.01292998 1.01343354 NCa
共b兲 Branch 共2兲
b
0.20 NEB 1.00723854 1.00723844 1.00723844 — — —
0.25 NEBb 1.01134023 1.01134021 1.01134021 — — —
0.30 NEBb 1.01639062 1.01639055 1.01639055 — — —
共c兲 Branch 共1兲
b c
0.189 NEB NE 1.00896032 1.00822033 1.00855784 1.00857384 1.00856056
0.192 NEBb 1.00683740 NCa 1.01019182 1.01093320 1.00930194 1.01054336
0.195 NEBb 1.00768976 NCa NCa 1.01299166 1.01377124 1.01217765
a
NC= nonconvergence.
b
NEB= nonexistent branch at this order.
c
NE= nonexistent branch for this wave steepness.

162 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006
Table 6. Convergence of Frequency ␻ as Function of Wave Steepness for Angle ␪ = 47.5°
h Order 5 Order 9 Order 13 Order 17 Order 21
共a兲 The asymptotic method without use of the Shanks transform
0.10 1.00145686 1.00145685 1.00145685 — —
0.20 1.00584012 1.00583928 1.00583927 1.00583927 —
0.30 1.01318774 1.01317760 1.01317710 1.01317707 1.01317707
共b兲 The asymptotic method with use of the Shanks transform
0.10 1.00145686 1.00145685 1.00145685 — —
0.20 1.00584016 1.00583928 1.00583927 1.00583927 —
0.30 1.01318830 1.01317744 1.01317710 1.01317707 1.01317707
共c兲 The direct numerical method
0.10 1.00145685 1.00145685 — — —
0.20 1.00583928 1.00583927 1.00583927 — —
0.30 1.01317690 1.01317707 1.01317707 — —

becomes unstable for h ⬍ 0.18935. It remains unstable down to


␾⬘ = e−i␴t 兺 兺 b j,kei共jx+ky兲e␥
j苸⌮ k苸⌮
j,kz
the TP. In the vicinity of the TP, the solution is unstable, then
共13兲 Branch 共1兲 remains unstable until h = 0.18907 where it restabi-
lizes. The h ranges for which Branches 共2兲 and 共1兲 are unstable
␩⬘ = e −i␴t
兺 兺 a j,ke
j苸⌮ k苸⌮
i共jx+ky兲
are approximately 20 times narrower than that for Branch 共3兲.
The, resonances associated with these branches are very sporadic.
with ␥ j,k = 关m2 j2 + n2 j2兴1/2. For Re共−i␴2,6兲 = 0 the 共2,6兲 mode is Thus, resonances of Branches 共2兲 and 共1兲 are unlikely to appear
neutrally stable. For Re共−i␴2,6兲 ⬍ 0, the mode is asymptotically although their growth rate 共0.168⫻ 10−2兲 is of the same order as
stable and for Re共−i␴2,6兲 ⬎ 0 the mode is unstable. that of Branch 共3兲. Thus, Branch 共3兲 is more likely to generate an
Following Ioualalen and Kharif 共1993兲, a superharmonic insta- observable resonance.
bility, associated to a 共2,6兲 harmonic resonance, corresponds to
the coalescence of the two eigenvalues ␴−2,6 and ␴2,6 at zero
frequency. This corresponds to a phase locking between the reso- Forces Exerted on Vertical Seawall
nant mode and the fundamental of the SCW in the moving frame
of reference. Indeed, the phase locking corresponds to a reso- Consider the nonuniqueness of the SCWs as may be generated by
nance phenomenon. reflection around marine structures. May such nonunique wave
Fig. 7 shows the stability of Branch 共3兲. The frequencies of the solutions generate significantly different wave forces encountered
resonant modes 共±2 , 6兲 coalesce around value of the pole at a in marine engineering? We try to answer the crucial question for
steepness value h = 0.194. One mode becomes unstable, whereas a typical case of a seawall resonant reflection. We chose a vertical
the other is asymptotically stable. Until that value, the two modes seawall immersed to a depth d 共nondimensonal兲 below the mean
are neutrally stable. The coalescence of the two eigenmodes gives zero level of the free surface.
rise to an instability which lasts until h = 0.215 with maximum of We compute at the crest 共x = 0兲 and at the trough 共x = ␲兲
instability of order 0.175⫻ 10−2 around h = 0.208. The resonance the nondimensional depth-integrated force per unit length
occurs in the range 关0.194,0.215兴 with a larger probability of oc- F共x , 0兲 = k2F⬘共x , 0兲 / ␳g, where we suppose that the seawall is
currence at h = 0.208.
Fig. 8 shows the stability of Branches 共2兲 and 共1兲. Branch 共2兲

Fig. 7. Stability results for Branch 共3兲 as function of wave steepness


h. Eigenvalue frequency 关−R共␴±2,6兲兴 is represented by 共〫兲 sign and
growth rate 关−I共␴±2,6兲兴 is represented by 共⽧兲 sign. Fig. 8. Same as Fig. 7 for Branches 共2兲 and 共1兲

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006 / 163
aligned along the x axis 共y = 0兲 and F⬘共x , 0兲 is the effective force
共Fenton 1985; Marchant and Roberts 1987兲


␩共x,0兲
F共x,0兲 = p共x,0,z兲dz 共14兲
−d

where p = associated nondimensional pressure exerted on the sea-


wall. With the Bernoulli expression of the given pressure being
1
p = −z + ␻␾x − 2 共m2␾2x + n2␾2y + ␾z2兲, the force is

冕 冋 册
␩共x,0兲
1 1 1
F共x,0兲 = d2 − ␩2 + ␻␾x − 共m2␾2x + n2␾2y + ␾z2兲 dz
2 2 −d 2
共15兲
The component 关−共1 / 2兲d 兴 = hydrostatic force referenced to
2

the flat surface and the residual Fw共x , 0兲 = effective force due to
the waves. The integral along the vertical axis is processed by a
simple trapezoidal rule. Dimensional values of the wave steepness
and depth are of the form hd = Lh / 2␲ and dd = Ld / 2␲ respectively,
where L = wavelength of the dimensional incident wave chosen
here to be 100 m. As a comparison and validation we placed at
depth d = 0.8 共dd ⬇ 12.73 m兲 as in Fig. 9 of Marchant and Roberts
共1987兲 for angle of resonance ␪ = 53°. In their Fig. 9 they consid-
ered a wave steepness of h = 0.16 共hd ⬇ 2.54 m兲. Marchant and
Roberts used the asymptotic method extended to finite depth. As a
result they obtained only the Branch 共3兲 solution. First our Fig. 9 Fig. 9. Variation of depth integrated force per unit length F / d2 with
shows the same result for Branch 共3兲 at h = 0.16 if we place at wave steepness h for ␪ = 45°
␪ = 53° in their Fig. 9. The onshore force is on the crest, the
offshore force is on the trough, and the numerical values are
identical. The new contribution of our figure is the calculation of the angle ␪ = 47.5° that is far from the critical angles of resonance
the wave force exerted on the seawall due to the two other branch occuring up to reasonable order 27 of truncation.
solutions to which Marchant and Roberts 共1987兲 have no access. The asymptotic method gave SCW solutions in the form of a
The main crucial feature is that, past the TP, the three branches of truncated series. For the resonant case, we recovered SCW solu-
solutions give wave forces very different from each other 共at the tion properties reported by Roberts 共1983兲. In particular, Padé
same order as their absolute value兲. Moreover Branch 共3兲 does not approximation has zeros and poles of the series that identify reso-
represent the maximum estimated wave force. The maximum nance. The solutions converge rapidly 共relative to the order r of
forces occur on Branch 共2兲 for the onshore force and on Branch truncation兲 for values of the wave steepness weaker than the es-
共1兲 for the offshore force, respectively. As a conclusion, in the timated pole h* associated with the resonance 共2,6兲. For larger
case of a harmonic resonance, Branch 共3兲 共or the asymptotic wave steepness, the Shanks transform 共␧ algorithm numerical
method solution兲 is not an interesting feature for marine structure procedure兲, derived from the Padé approximation, extends the
engineering 共in our parameter case study兲. A complete full para- radius of convergence past h* to obtain a converged solution at
metrical study should be undertaken but it is not the aim of the approximately the same precision as that below h*. However, the
present work. solutions are approximate ones only and the convergence ob-
tained with the Shanks transform is not robust because the trans-
form needs to take into account the physical system of the wave
Conclusion itself. The asymptotic method complemented with the Shanks
transform introduces a pole and thus a vertical asymptote. Since
SCWs in deep water have been computed with two different the system is physical, the solution cannot tend to infinity around
methods: asymptotic and direct fully numerical. The computa- the asymptote. This suggests the presence of TP and multiple
tions have been processed for two distinct wave configurations: branches of the solution that are not, however, reachable with the
resonant and nonresonant. The resonances derive from SCW in- method. Only Branch 共3兲 can be obtained with the method.
trinsic interactions without any outside forcing. This is due to, in The fully numerical method resolves all three branches. The
a linear description, some of the 共i , j兲 modes of the SCW har- first branch, Branch 共3兲, corresponds to the one computed with
monic structure propagating at the same phase speed of the SCW the asymptotic method and covers the whole range of the autho-
fundamental mode, yielding secularities. We stress that any engi- rized wave steepness, that is, from a flat surface to near breaking
neering study on the effects of surface waves on marine structures of the wave. Its convergence is linked to the order of numerical
should take carefully into account the resonance phenomena. In truncation. It is affected by all harmonics of the wave but with a
the present work the 共2,6兲 resonance has been explored because it decreasing influence of higher modes. The two other branches
is the lowest order resonance and thus it is likely to be the stron- appear only past h*. Branch 共2兲 converges rapidly as a function of
gest resonance in its characteristics. For the resonant wave con- the order of truncation immediately after the occurrence of the
figuration, we chose an angle ␪ = 53° close to the critical 共2,6兲 mode. Thus, this branch seems to correspond to the 共2,6兲
resonance angle ␪HR = 52.2388° of the 共2,6兲 harmonic resonance free mode oscillation or at least to a relatively low frequency
in deep water. For the nonresonant wave configuration, we chose evolution of the SCW. Branch 共1兲 converges slowly with increas-

164 / JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006
ing order of truncation. This suggests the strong effects of high for his help and Mrs. Annie Tranchard for her technical support,
order harmonics on this branch. Consequently, the branch is very both from IFREMER. Finally the writers are grateful to the edi-
sensitive to high frequency modes of oscillation of the SCW field. tors and referees who contributed to a substantial improvment of
As a result, the distinct branches of the resonant SCW solutions the first draft of the manuscript, in particular through their sug-
seem to be sensitive to different time scales of wave evolution. gestion to compute one engineering application.
For the nonresonant case, only Branch 共3兲 exists, and solutions
from the direct method converge faster than those obtained from
the asymptotic method. References
The stability of the solution for the resonant configuration has
been explored in order to better understand the behavior of the Bender, C. M., and Orszag, S. A. 共1981兲. Advanced mathematical meth-
SCW solution branches. Branch 共3兲 is unstable in a range of the ods for scientists and engineers, McGraw-Hill, New York.
wave steepness of order ⌬h = 0.021 around h = 0.2045. This range Fenton, J. D. 共1985兲. “Wave forces on vertical walls.” J. Waterw., Port,
is small. Branches 共2兲 and 共1兲 and the TP are unstable in a range Coastal, Ocean Div., Am. Soc. Civ. Eng., 111共4兲, 693–718.
of the wave steepness of order ⌬h = 0.001 around h = 0.1885 with Gilewicz, J. 共1978兲. “Approximants de Padé.” Lecture notes in mathemat-
a comparable growth rate. This latter range is at least 1 order of ics, Vol. 667, Springer, New York.
magnitude narrower than that of Branch 共3兲. As a consequence, Hsu, J. R. C., Tsuchiya, Y., and Silvester, R. 共1979兲. “Third-order ap-
Branch 共3兲 is more likely to amplify the superharmonic mode proximation to short-crested waves.” J. Fluid Mech., 90, 179–196.
Ioualalen, M., and Kharif, C. 共1993兲. “Stability of three-dimensional pro-
共2,6兲 via resonance.
gressive gravity waves on deep water to superharmonic disturbances.”
Finally we computed, for one case, the wave force per unit Eur. J. Mech. B/Fluids, 12共3兲, 401–414.
length exerted by resonant SCWs on a vertical seawall. The result Ioualalen, M., and Kharif, C. 共1994兲. “On the subharmonic instabilities of
indicates a drastic change of the force from one branch of solu- steady three-dimensional deep water waves.” J. Fluid Mech., 262,
tions to another. Moreover, Branch 共3兲 which is usually computed 265–291.
共Fenton 1985; Marchant and Roberts 1987兲 is not associated with Marchant, T. R., and Roberts, A. J. 共1987兲. “Properties of short-crested
the maximum of the force intensity. Instead, waves on Branches waves in water of uniform depth.” J. Aust. Math. Soc. Ser. B, Appl.
共2兲 and 共1兲 exert maximum onshore and offshore forces, respec- Math., 29, 103–125.
tively. Marchant, T. R., and Roberts, A. J. 共1990兲. “Reflection of nonlinear deep-
water waves incident onto a wedge of arbitrary angle.” J. Aust. Math.
Soc. Ser. B, Appl. Math., 32, 61–96.
Okamura, M. 共1996兲. “Notes on short-crested waves in deep water.”
Acknowledgments J. Phys. Soc. Jpn., 65, 2841–2845.
Roberts, A. J. 共1983兲. “Highly nonlinear short-crested water waves.”
The writers thank the SST service at the French Ambassy at J. Fluid Mech., 135, 301–321.
Tokyo for their valuable and constant support, in particular M.M. Roberts, A. J., and Schwartz, L. W. 共1983兲. “The calculation of nonlinear
Robert Farhi and Michel Israel. Thanks also go to M. Ivan Conesa short-crested gravity waves.” Phys. Fluids, 26共9兲, 2388–2392.

JOURNAL OF WATERWAY, PORT, COASTAL, AND OCEAN ENGINEERING © ASCE / MAY/JUNE 2006 / 165

View publication stats

You might also like