Page 127 - Computational Fluid Dynamics for Engineers
P. 127

4.5  Finite-Difference  Methods  for Elliptic  Equations              113



         In  the  backward  sweep,  6j  is computed  from  the  following  recursion  formulas:

                       AJ6J  =  wj,                                       (4.4.34a)
                       Aj6j  =  Wj  -  CjSj+u  j  =  J  -  1, J  -  2 , . . . ,  0.  (4.4.34b)


         Example  4.4. Repeat  Example  4.1 using Keller's box method.  Compare your results  with
         those  obtained  with  the Crank-Nicolson  method.

         Solution.  The solution  of  Example  4.4 with  Keller's  box method  follows  the  procedure
         described  in subsection  4.4.3.  Essentially  after  generating the  finite-difference  grid,  initial
         profiles  at  x  =  0, we define  (si)j  to  (ss)j  in Eq.  (4.4.25a)  together  with  {r\)j  and  (r2)j.
         Then  we use  SOLV2  to  solve  the  linear  system.  Note  that  in  this  case  wall  and  edge
         temperatures  are  specified.
            A comparison  of numerical  results  obtained  with the Box method  together  with  those
         obtained  with  the  Crank-Nicolson  method  and the  analytical  method  is  given  in  Table
         E4.9.  A listing  of the computer  program  using  SOLV2  is given  in Appendix  A,  Example
         4.4.  As can be seen,  the predictions  of the box method  are in very  good  agreement  with
         the  other two.

         Table  E4.9.  Comparison  of  the  box
         method  results  with  CN  and  AS at  x  =
         0.30.  At  = 0.0100,  Ax  = 0.100

           t       CN        Box      AS
         .0000    .600      .6000     .6004
         .0100    .584      .5875     .5799
         .0200    .543      .5377     .5334
         .0300    .496      .4888     .4857
         .0400    .454      .4426     .4411
         .0500    .414      .4006     .4000
         .0600    .377      .3624     .3626
         .0700    .344      .3277     .3286
         .0800    .313      .2965     .2977
         .0900    .286      .2680     .2698
         .1000    .260      .2426     .2444






         4.5  Finite-Difference   Methods    for Elliptic  Equations


         To describe the numerical solution  of elliptic equations by  finite-difference  meth-
         ods,  consider  the  Poisson  equation  in  the  form
                                    2
                                            2
                                   d u     d u

         and  its  solution  for  a  rectangular  region  (Fig.  4.7)  defined  by  0  <  x  <  a,  0  <
         y  <  b, with  uniform  net  spacings  Ax  and  Ay  and  with
   122   123   124   125   126   127   128   129   130   131   132