8

8.1

Third order expansions
General

As something new we will introduce third order terms in the boundary element model. So from here let the perturbation expansions of  and be of the form  = "1 + "22 + "33 + O"4  = " 1 + "2 2 + "3 3 + O"4  8.1 8.2

These rede nitions of the expansions of  and do not contradict the former de nitions 2.44 and 2.44. Thus the results obtained in the previous parts of this work are still valid, but they may be improved by the introduction of third order terms. It is easily shown that the third order potential, 3, must satisfy the Laplaceequation, and thus the boundary integral equation 2.49 is valid for the third order potential. Analytically this is written as

@G , G @ 3 ds =

8. but for waves of intermediate nonlinearity the third order terms may make this model feasible. This means that the calculation time per time step will be increased by roughly 50 by introducing third order terms. For very weakly nonlinear waves it may not be worth the extra calculation time to obtain the third order potential.5.  x 8.0 In discrete form the boundary integral equations obtained for 3 will have the same coe cients as the equations for 1 and 2 as described in section 4.4 3 @n 3 0 @n0 . and this model will still be a lot faster than a full nonlinear BEM.2 Free surface and open boundary conditions +Z r23 = 0 .

68 . 8.1 General The most demanding task by introducing third order terms in the BEM. is the evaluation of new boundary conditions based on the dynamic 3.2.42.32 free surface boundary conditions modi ed by dissipative terms for a sponge layer.3 8.31 and kinematic 3. for instance given by 3.

@z2 + @x@z @x . 1 2 @z3 2 @x@z2 @x  3 @ 1 @ @ @  @ 3 + 6 @z3 @z . @z + @x @x +  x  2  @ @ @  @  @ 1 2 + 2 @z2 @t .8 .Taylor expanding the kinematic boundary condition from z = 0 we get @ . at z = 0 @  + g + 1 jrj2 +  x @t 2 2 . @z + @x @x +  x = O 4 . @  + @  @ +  x @t @z @x @x 2 2 @ 3 3 @ @ 2 @  + 1 2 @  @ . @z + @x @x +  x  3  @ 1 @ @  @  @ 3 + 6 @z3 @t . at z = 0 8.6 Similarly from the dynamic boundary condition we get @  + g + 1 jrj2 +  x @t 2   @ @  1 2 + @z  @t + g + 2 jrj +  x 1 @ 2 @  + g + 1 jrj2 +  x + 2 2 @z2  @t 2  3 1 @ @  1 2 3 + 6 @z3 @t + g + 2 jrj +  x = O 4 .7 + 8. @x @x = O 4 . @  + @  @ +  x @t @z @x @x   @ @  @  @ @ + @z @t . 2  @ 1 @ 1 @ 1 + 2 2 @t@z2 + 4 2 @z2 jrj2 + 2  x 2 @z 2   3 1 @ @  1 2 3 + + g + jrj +  x = O 4 . at z = 0 6 @z3 @t 2 69   + @ . @t . @  + 1 @ jrj2 +  x @  + @t@z 2 @z @z 3 2 . at z = 0 8.5   8.

@ 2 @ 1 . 1  x 2 @ 1 8. 1 @x@z @x + 2 1 @z31 8. @ 1 @ 2 . @ 21 3 3 @t @x @x @z @z 1 @t@z 2 @t@z @ . @ 1 @ 2 .jr j2 . @ 22 . @ 21 3 @t @n0 @x @x . @ 3 +  x = .11 . @ 1 @ 2 .Perturbation expanding  and in 8.1 1 1 1 2 2 @z @z @z 3 2 1 2 @ 1 .  x @ 2 . @ 3 +  x = .2 1 @t@z 2 1 @z 2 2 @ 3 . @ 1 @ 2 .6 and keeping only terms of third order in " we get @ 3 . @ 22 .9 + Similarly perturbation expanding 8. @ 2 @ 1 + @ 22 + @ 21 3 @t @z @x @x @x @x 1 @z2 2 @z2 2 @ 3 1 @ @ 1 1 2 .  x @ 1 .8 and keeping only terms of third order we get @ 3 + g +  x = .

@x @x 1 @x2.

@ 1 @ 2 .10 . @ 1 @ 2 3 3 @t @x . 1 2 @ @ 1 1 @ 8. 2 @x2 2 @ 1 . 1 @ @x @x @n0 2 1 @x2 @n0 @ 3 + g +  x = .

@x @n0 @n0 .

@ @ 1 . 1 @t 2 0 @n. @ @  2 .

 x 1 @ @n0 2 @n0 .  x @ 1 . @t @n0 1 @ @ 1 + @ 1 @ 21 . 1 @@x 1 @x @n0 @n0 @x2 2 .

11 show the terms as they will look directly after the expansion.2.1 the third order expressions for the dynamic and kinematic free surface boundary conditions are quite 70 .Applying an implicit Adams method As we have seen in the previous section 8.9 and 8.10 and 8. + 8.2 Updating . 2 2 1 @ @ 1 @ 1 2 2 + 2 1 @t @x2 + 2  x 1 @x21 8.2.12 are rewritten to forms ready for implementation in a computer program.12 The equations 8. while the equations 8.

A1 = . We have chosen an explicit Adams-Bashforth method of higher order than before ka = 3 according to 4. and then use the kinematic boundk ary condition for nding new values of @ @t .complicated. So still letting indices m.25  with the coe cients A0 = 55=24.3. j " refer to time t = tm and space x = xj . as described in section 4. A2 = 37=24 and A3 = .59=24.9=24. Instead we will use an explicit AdamsBashforth method for the updating of k . we will not Taylor expand these equations in time. the new updating of will be  .3. and thus to avoid complicating things even more.

.

m+1.m.1. 55 @ 59 @ k k . 24 @t k.j  m.j = k.j + t 24 @t m.j   .

.

2.14 Choosing ka = 3. with Fk as right hand side. This procedure is often attributed to Adams-Moulton. The dynamic free surface boundary condition. we have the coe cients B0 = 9=24. 24 @tk m. 2.2. and this will also be done here. The method is often used as the corrector step in a predictor-corrector method.13 After some constructive conversations with Jesper Skourup. B1 = 19=24. Similarly to the explicit AdamsBashforth methods an implicit Adams-Moulton method is of the form ym+1 = ym + t B0fm+1 +    + Bka fm+1. B2 = .3.ka 8. 3g be written as .5=24 and B3 = 1=24. but we use the method for prediction of k only. can for k 2 f1. 3g 8. we chose to update  on the free surface by an implicit Adams method. even though the formula is due to Adams. k 2 f1. 37 @ 9 @ + 24 @tk .j  m.j  .

j +  xk. @ k @t m+1.15 .j = Fk.m+1.m+1.m+1.j + g k.j 8.

j = 1 + t B  x k.m.m+1.m+1.j + t B0 .m+1.j  we get "   .g k.j 0  .j + Fk. @  k in the equation for the implicit Adams-Moulton method Inserting @t m+1. 1 k.

.

.

j  8. @  @  @  +B1 @tk + B2 @tk + B3 @tk m.16 71 .1.j  m.2.j  m.

m+1.j  are known.When k. the dynamic surface boundary .j and k.m+1.

j  After having solved the boundary. conditions will provide @tk m+1. @  to be used in the following time steps.

@ . The method seems stable and reliable when the Courant number is below unity. @  + 1 2 @ 2 . stability can not be guaranteed.j at . @  + O4  = V . care should be taken to ensure that this does not happen.  @ 22 . and not improve the accuracy by much. +6 n r10 8. @  .e. see sections 3.13 from x = 0 we get  +  @ . say Cr = 0:8. at .integral equations the kinematic free sur@ k face boundary conditions supply @t m+1. but since this roughly would double the calculation time.13 and @ @z = 0. we have not implemented a corrector step.18 1 2 n. @ =.  @ 21 .3 and 3.@ 2 @x @x @x 2 @x @x  1 3 @ 3 . can in mathematical terms be described by a boundary condition on the form 3. Since the model may turn out to be unstable when the Courant number is greater than unity. 8.3 @x2 @x2 2 1 @x3     72 . The method described above has been implemented and is hereafter used in the third order model.3 Piston boundary condition A piston type wave-maker. @n 0 @x Taylor expanding 3. @@x 8.4.17 @x3 @x Perturbation expanding  and  and keeping only terms of third order we get 3 .f 0. 1 2 @ 31 = V . and since our method represents the predictor step only. also for the rst and second order terms. To improve stability a corrector step can be included in the model. i. It should be noted that the Adams-Bashforth-Moulton methods are normally used in predictor-corrector methods.

at .19 @ 3 = V +  @ 22 +  @ 21 n. @1 = .18 can be reduced to 8.1 = V and @ = 0 we get Using the Laplace-equation.3 1 2 @n0 @x2 @x2 2 2 . at .3 .20 73 .1 @z2 @t @z@t @z .1 @z       @ 31 = @ . @ 2 @ 1 = @ 2 @ 1 0 @x3 @x @z2 @z2 @x @z2 @n  @ 2 V = @ 2 . 1 @@z 2 r10 2 @z2 8.r10 Thus 8. @ @n0 n.  @ 21 . = Vn. @ 2 @1 = 0 = @z 2 n. @ 21 = .