Page 472 - Practical Design Ships and Floating Structures
P. 472

447


        3 DISCRETIZATION AND SINGULARITY DISTRIBUTION
                                    -.
        To  solve the integral equation for  a(X,), the collocation method  is used. Field  points are chosen
        along the real boundary and sources are distributed outside the computational domain. A set of field
        points and the corresponding source points are chosen along the contours,  S, , S, ,and  S, , as shown
        in Fig. 2.

        In the DELTA method, the sources are distributed on the integration surface so that the source points
        never coincide with the field points and the integrals are nonsingular. In addition, a simple isolated
        sources rather than a distribution is used. The equivalent accuracy in the solution is then obtained.





                           SOURCE POINT

                                                         Dm

                          cp given




                              Figure 2: Model for numerical simulation
        The  singularities  are  distributed  above  the  field  points  on  the  free  surface  in  the
        normal  direction of the boundaries. Inside the body,  the isolated  singularities are placed along the
        normal direction from the field points in the body  surface. Numerical difficulties may occur in the
        vicinity of a sharp edge. One of the difficulties is that the singularity distribution may cross over the
        bisector of two body  surfaces or even the other side of the body surface since the desingularization
        distance is proportional to the locaI grid size.
        These  types  of  difficulties  can  be  avoided  by  careful  discretization and  desingularization. The
        desingularization distance near a sharp comer is modified so that the singularities are distributed on the
        bisector of the two body surfaces to avoid the cross over of the singularities beyond the centerline or
        the body surfaces.

        The nondimensional desingularization distance is set to be
                                        Ld = IdDm)=                             (3)
        Where  I,  reflects how far the integral equation is desingularized, Dm is the non-dimensional  local
        mesh size (usually the square root of the local mesh area in 3-D problems and the local mesh size in
        2-D problems). a is a parameter associated with the convergence of the solution as the mesh is refined.
        Cao,  Schultz & Beck  [2]  conducted numerical  tests  in  which  an  integral of  a  constant  source
        distribution over a square flat surface is evaluated at a point above the center of the square with  a
        distance given by Eqn. 3 They found that a = 0.5 and  I,  = 1 .O  are about the optimum values for the
        performance of the desingularization method.
        A linear system of  m x rnalgebraic equations is set to be
                               -
                                    -
                          Amxm Xm = Bm                            (4)
   467   468   469   470   471   472   473   474   475   476   477