Page 154 - Basic Structured Grid Generation
P. 154

Differential models for grid generation  143

                        grid-cell area. We start with a single boundary curve on which η is taken to be zero.
                        The two first-order partial differential equations for the cartesian co-ordinates x, y as
                        functions of ξ, η are:

                                              g 12 = g 1 · g 2 = x ξ x η + y ξ y η = 0,
                                                |g 1 × g 2 |= x ξ y η − x η y ξ = V,      (5.111)
                        where the first equation imposes orthogonality, and the second implies, by eqn (1.43),
                        that V is a measure of cell-area (assuming that δξδη is the same for each grid cell).
                        In general we can let V be a function of ξ, η, so that, for example, we could increase
                        grid-density near the boundary η = 0byensuringthat V is small there.
                          It follows from eqns (5.111) that

                                                       V            V
                                                x η =−   y ξ ,  y η =  x ξ ,              (5.112)
                                                      g 11         g 11
                                       2
                                              2
                        where g 11 = (x ξ ) + (y ξ ) .
                          The subdirectory Book/hyper.gds on the accompanying disk contains a program for
                        solving these equations numerically using a marching procedure. The cell-area V is
                        prescribed in the form
                                                         √     −λ(1−η)
                                                   V = K g 11 e      ,                    (5.113)
                        where K and λ are constants, and the equations are discretized with first-order accuracy
                        in the η-direction and second-order accuracy in the ξ-direction as

                                       (x i,j+1 − x i,j )  V      y i+1,j − y i−1,j
                                                    =−                           ,
                                             η            g 11        2 ξ
                                                              i,j

                                       (y i,j+1 − y i,j )  V    x i+1,j − x i−1,j
                                                    =                          ,          (5.114)
                                             η          g 11        2 ξ
                                                            i,j
                        where
                                                               2                  2
                                                x i+1,j − x i−1,j  y i+1,j − y i−1,j
                                      (g 11 ) i,j =             +                  .
                                                    2 ξ                 2 ξ
                          These equations determine explicitly the values x i,j+1 and y i,j+1 in terms of val-
                        ues x i−1,j ,x i,j ,x i+1,j ,y i−1,j ,y i,j ,and y i+1,j . Hence the method itself may be called
                        explicit. Only data on values of x and y on the initial boundary η = 0 are required
                        to start the process. The numerical solution starts from this boundary and marches
                        outward in the direction of increasing η, eventually constructing new co-ordinate lines
                        ξ = 0, 1and η = 1.



                           5.11 Solving the hosted equations

                        5.11.1 An example


                        In this section we illustrate the task of solving a partial differential equation, given
                        that the solution domain has been discretized by the generation of a suitable grid. We
   149   150   151   152   153   154   155   156   157   158   159