mmcp-19-gerstner

Computational Model of Unsteady Hydromechanics of Large Amplitude Gerstner Waves
git clone https://git.igankevich.com/mmcp-19-gerstner.git
Log | Files | Refs

main.tex (12748B)


      1 \documentclass{webofc}
      2 
      3 \usepackage[varg]{txfonts}
      4 \usepackage{wrapfig}
      5 
      6 \begin{document}
      7 
      8 \title{Computational Model of Unsteady Hydromechanics\\of Large Amplitude Gerstner Waves}
      9 %\subtitle{Do you have a subtitle?\\ If so, write it here}
     10 
     11 \author{%
     12 \firstname{Alexander} \lastname{Degtyarev}\inst{1}%
     13 \and%
     14 \firstname{Ivan} \lastname{Gankevich}\inst{1}%
     15 \fnsep\thanks{\email{i.gankevich@spbu.ru}}%
     16 \and%
     17 \firstname{Nataliia} \lastname{Kulabukhova}\inst{1}%
     18 \and%
     19 \firstname{Vasily} \lastname{Khramushin}\inst{1,2}%
     20 }
     21 
     22 \institute{%
     23 Saint Petersburg State University, Universitetskaya
     24 Emb. 7-9, 199034 St.~Petersburg, Russia%
     25 \and%
     26 Alexey Krylov All-Russian Scientific Shipbuilder Society,  %with headquarters in 
     27 Saint Petersburg, % Scientific Society of Shipbuilders named after Alexey Krylov, 
     28 Russia%
     29 }
     30 
     31 \abstract{%
     32 The computational experiments in the ship fluid mechanics involve the
     33 non-stationary interaction of a ship hull with wave surfaces that include the
     34 formation of vortices, surfaces of jet discontinuities, and discontinuities in
     35 the fluid under the influence of negative pressure. These physical phenomena
     36 occur not only near the ship hull, but also at a distance where the waves break
     37 as a result of the interference of the sea waves with waves reflected from the
     38 hull.  In the study reported here we simulate the wave breaking and reflection
     39 near the ship hull. The problem reduces to determining the wave kinematics on
     40 the moving boundary of a ship hull and the free boundary of the computational
     41 domain.  We build a grid of large particles having the form of a parallelepiped
     42 and, in the wave equation instead of the velocity field we integrate streams of
     43 fluid represented by functions as smooth as the wave surface elevation field.
     44 We assume that within the boundaries of the computational domain the waves do
     45 not disperse, i.e.~their length and period stay the same. Under this
     46 assumption, we simulate trochoidal Gerstner waves of a particular period.
     47 This approach allows to simulate the wave breaking and reflection near the ship
     48 hull.  The goal of the research is to develop a new method of taking the wave
     49 reflection into account in the ship motion simulations as an alternative to the
     50 classic method which uses added masses.}
     51 
     52 \maketitle
     53 
     54 
     55 \section{Introduction}
     56 
     57 The ship and sea wave surface motions do not involve an abundance of geometric
     58 forms and physical phenomena. The formation of waves due to the ship motion and
     59 the interaction between the ocean and the atmosphere are governed only by the
     60 continuity condition for a heavy fluid and the law of conservation of the
     61 energy.
     62 
     63 A strict theoretical solution (and in fact the only solution) for
     64 large-amplitude wind waves at the surface of a heavy fluid was obtained in 1802
     65 by Franz Josef von Gerstner~\cite{gerstner1809}.  The generic trochoidal wave
     66 mathematical model has large dispersion~\cite{sommerfeld1945mechanik}, the
     67 speed of the wave propagation depends on their length and period. As a result
     68 \begin{itemize}\addtolength\itemsep{-1mm}
     69 	\item wave energy propagation speed becomes half the visible phase velocity
     70 		of the wave crests,
     71 	\item wave front constantly changes its phase, and
     72     \item waves are quantised into packets and the phenomena of wave
     73         transformation and propagation become nonstationary.
     74 \end{itemize}
     75 
     76 
     77 \section{Computational model of intense sea waves}
     78 
     79 \begin{wrapfigure}{l}{0.5\textwidth}
     80 	\centering
     81 	\includegraphics[width=0.5\textwidth]{graphics/01-gerstner.png}
     82 	\includegraphics[width=0.5\textwidth]{graphics/02-gerstner.png}
     83 	\caption{(Color online) Analytic solutions: progressive Gerstner wave (top),
     84 	a wave of critical height, as a standing wave (bottom).\label{fig-gerstner}}
     85 \end{wrapfigure}
     86 
     87 The Gerstner wave (Fig.~\ref{fig-gerstner}) is a cycloid with the radius
     88 $r_W=1.134\lambda_Wh_W/4\pi$  of the particle trajectory being fixed relative to
     89 the flat wave surface level \(z_W\), hence \(z\)-coordinates of the crest and
     90 trough are the same. Here \(\lambda_W\) is the wave length, \(h_W\) is the
     91 relative wave height defined on the interval \([0,1]\) with \(h_W=1\) being the
     92 maximum wave height for which the crest does not break
     93 (Fig.~\ref{fig-gerstner}). The vertical displacement of a fluid particle is
     94 given by
     95 
     96 \begin{equation*}
     97 	\zeta_Z = r_W \cos x_W \exp\left(-2\pi z_W / \lambda_W\right).
     98 \end{equation*}
     99 The horizontal displacement of the same fluid particle with respect to its
    100 initial position for progressive wave is given by an analogous equation, but
    101 with a shift by one fourth of the phase:
    102 \begin{equation*}
    103 	\zeta_X = -r_W \sin x_W \exp\left(-2\pi z_W / \lambda_W\right).
    104 \end{equation*}
    105 The critical wave height of the Gerstner waves (Fig.~\ref{fig-gerstner}) gives
    106 the correct ratio of the wave height to the wave length, but 60 degree slope
    107 limit for standing wave with steepness \(\approx{}1/4\) as well as 30 degree
    108 slope limit for progressive (traveling) wave with steepness \(\approx{}1/7\)
    109 are not correctly captured by the model.
    110 
    111 \begin{wrapfigure}{r}{0.5\textwidth}
    112 	\centering
    113 	\includegraphics[width=0.5\textwidth]{graphics/03-wind-wave.png}
    114 	\includegraphics[width=0.5\textwidth]{graphics/05-wind-wave.png}
    115 	\caption{(Color online) Simulation result: regular trochoidal waves with vertical
    116 	displacement of sea level and wind stress on the wave sea
    117 	surface. Propagating wind waves (top), extremely high wind waves
    118 	(bottom).\label{fig-trochoidal}}
    119 \end{wrapfigure}
    120 
    121 The main focus of state of the art mathematical and computational models that
    122 simulate wave groups is on stochastic properties rather than fluid
    123 mechanics~\cite{anastopoulos2016}.  Our model is a modified version of the
    124 Gerstner wave which includes wave groups. They are described as a dependency
    125 between fluid particle trajectory radius and instantaneous displacement of the
    126 particle with respect to the calm sea level (Fig.~\ref{fig-trochoidal}).
    127 
    128 We write the adjusted radius as \(R^A_W={}K^A{}r_W\left(\cos{}x_W-1\right)\), where
    129 \(K^A\in[1,\sqrt{2}]\) is the radius coefficient that makes the wave crests cnoidal
    130 and raises the mean sea level. We choose \(K^A\) to be slightly less than 1 to
    131 reduce the effect of gusty winds on the wave form and to prevent the formation
    132 of cycloidal loops in the wave crests, that appear for waves with overly large
    133 amplitude, which may occur as a result of the interference with waves
    134 heading from the opposite direction.
    135 
    136 The pressure on the windward slope of the wave is smaller, because the wind
    137 slides on the surface of the wave at a high speed makes the slope flatter,
    138 while on the leeward slope of the wave the wind speed drops significantly or
    139 even comes to nought and creates vorticity.
    140 
    141 The coefficient of wind stress \(K^W\) (the parameter that was shifted by one
    142 fourth of the phase) determines the asymmetry of steepness of windward and
    143 leeward slopes of the wave (Fig.~\ref{fig-trochoidal}):
    144 \(R^W_W=K^Wr_W\left(\sin{}x_W-1\right)\), where \(K^W\in[0,1]\). The
    145 coefficient is close to unity for fresh wind waves and close to nought for
    146 swell. For a two-dimensional sea surface, \(K^W\) is used in dot product
    147 between the wind and wave direction vectors:
    148 \begin{equation*}
    149 	\begin{aligned}
    150 		\zeta_Z &= r_W \cos x_W \exp\left(
    151 			2\pi \left[
    152 				-z_W + r_W K^A \left( \cos x_W - 1\right) + r_W K^W\sin x_W
    153 			\right] / \lambda_W
    154 		\right) \\
    155 		\zeta_X &= -r_W \sin x_W \exp\left(
    156 			2\pi \left[
    157 				-z_W + r_W K^A \left( \cos x_W - 1\right) + r_W K^W\sin x_W
    158 			\right] / \lambda_W
    159 		\right).
    160 	\end{aligned}
    161 \end{equation*}
    162 
    163 The energy conservation is defined by the Bernoulli's principle:
    164 \(\rho V^2/2+\rho g \zeta_W=\text{const}\),
    165 where the particle velocity \(V\) brings the largest contribution to balancing
    166 the pressure \(\rho{}g\zeta_W\) on the wave surface down to nought for breaking
    167 waves.
    168 
    169 \section{Trochoidal wave groups}
    170 
    171 \begin{wrapfigure}{l}{0.5\textwidth}
    172 	\centering
    173 	\includegraphics[width=0.5\textwidth]{graphics/06-group-wave.png}
    174 	\caption{(Color online) Trochoidal wave groups.
    175 	\label{fig-group}}
    176 \end{wrapfigure}
    177 
    178 In our modified model we simulate two wave surfaces simultaneously: one for
    179 regular waves with normal length and one for waves with nine times higher
    180 length, that propagate under the same laws but with half speed
    181 (Fig.~\ref{fig-group}). The product of these surfaces allows to simulate wave
    182 groups.
    183 
    184 On the first entry the profile of the long wave is given by a specific
    185 smoothing function the form of which is close to the phase wave profile. This
    186 function defines continuous changes of the wave front phase which is needed to
    187 simulate waves produced by the ship.
    188 
    189 \begin{wrapfigure}{r}{0.6\textwidth}
    190 	\centering
    191 	\includegraphics[width=0.6\textwidth]{graphics/waves-01.png}
    192 	\caption{(Color online) Large-amplitude trochoidal waves.
    193 	\label{fig-waves-1}}
    194 \end{wrapfigure}
    195 
    196 There is also a simpler approach to simulate wave groups: a superposition of
    197 regular waves with slightly different periods propagating in opposite
    198 directions. The interference of waves of comparable lengths produces beats, in
    199 which the ninth wave has double height and is standing wave. This approach
    200 generally gives satisfactory wave surfaces, but does not work for waves
    201 produced by the ship, because these have intricate wave fronts.
    202 
    203 
    204 \section{Direct numerical simulation of sea waves}
    205 
    206 We use an explicit numerical scheme to simulate a sea wave surface that
    207 satisfies the continuity equation; we call it direct numerical simulation
    208 (Fig.~\ref{fig-waves-1}). We use the following definitions for three sea wave
    209 systems, that are used in the scheme.
    210 %\vspace{-0.5\baselineskip}
    211 \begin{itemize}	\addtolength\itemsep{-1mm}
    212     \item Fresh wind waves have a period of 6--8 seconds near the shore and up
    213         to 10--12 seconds in the open ocean. The height of the wave is close to
    214         critical, that typically corresponds to 6 on the Beaufort scale with
    215         wave crests larger than 5--6 metres.
    216 
    217     \item Fresh swell waves skew from mean wind direction by \(\approx{}30\)
    218         degrees. When the storm in northern hemisphere increases, wind
    219         direction goes counterclockwise and vice versa, i.e. the swell is
    220         always present in the ocean. Swell waves are comparable to wind waves:
    221         their height is two times smaller than the critical wave height, and
    222         their length is 1.5--2 times larger.
    223 
    224     \item Old swell waves are long waves that come from higher latitudes. Their
    225         height is two times smaller than than of the wind waves and fresh
    226         swell, their length is two times larger, and their direction is close
    227         to meridional (i.e.~south in northern latitudes and vice
    228         versa\footnote{The wind blows into the compass rose, the waves
    229         propagate in the direction of the rose.}).
    230 
    231 \end{itemize}
    232 
    233 \begin{wrapfigure}{l}{0.55\textwidth}
    234 	\centering
    235 	\includegraphics[width=0.55\textwidth]{graphics/07-exp-1.png}
    236 	\caption{(Color online) In the course of the simulation we visualise all three
    237 		wave systems and create a view of ship hull dynamics
    238 		and sea wave profiles in a different convenient scale.
    239 	\label{fig-waves-2}}
    240 \end{wrapfigure}
    241 
    242 These waves may add up in unfavourable way to a wave with the height of 13--15
    243 metres, however, in real world mean wave height will be 8--10 metres. Wave
    244 groups have the ninth wave with double height, breaking crest and wave slope larger
    245 than 45 degrees.
    246 
    247 Oceanographers use well-estab\-lished solutions~\cite{poplavskii1997} for regular
    248 progressive waves of arbitrary shape. Using trochoidal waves as a source, we
    249 fix wave periods and speeds in time to satisfy continuity equation and
    250 energy conservation law. We simulate all three wave systems (described above)
    251 simultaneously and indepedently (Fig.~\ref{fig-waves-2}) and add individual
    252 wave surfaces together to produce the resulting wave surface.
    253 
    254 \vspace{-0.5\baselineskip}
    255 
    256 \section{Conclusion}
    257 
    258 We use explicit numerical schemes to simulate a modified version of the
    259 Gerstner waves.  We simulate particle drift in the upper fluid layers by
    260 changing the curvature of the trajectory depending on the instantaneous change
    261 of the wave surface elevation.  Our model is nonstationary, hence ship motions
    262 can also be nonstationary.  The computational power of a desktop computer is
    263 enough for performing such simulations in real-time, and these types of
    264 simulations can even be performed on the board of the ship to chose efficiently
    265 the optimal mode of ship operation.
    266 
    267 %\begin{acknowledgement}
    268 
    269 \vspace{-0.5\baselineskip}
    270 
    271 \subsection*{Acknowledgement}
    272 
    273 \vspace{-0.5\baselineskip}
    274 
    275 Research work is supported by Saint Petersburg State University (grant
    276 no.~26520170 and~39417213).
    277 %\end{acknowledgement}
    278 
    279 \vspace{-0.5\baselineskip}
    280 
    281 \bibliography{refs.bib}
    282 
    283 \end{document}