Page 291 - Computational Fluid Dynamics for Engineers
P. 291

9.5  Differential  Equation  Methods                                  281



         (9.5.12)  are  known  on the  boundaries,  except  f^  (On  a  £ =  constant  boundary,
         the  same equations  for the  control  functions  result,  but  with  r^  as the  unknown
         quantity).  The  r m  term  must  be  solved  for  as part  of the  solution.  An  iterative
         procedure  is  set  up  in  order  to  evaluate  f^  as  the  solution  progresses.  This
         approach,  first  introduced  by  Sorenson  [5]  is  now  common,  with  modifications
         of the  basic  concept  introduced  in  various  codes.
            Once  the  form  of  the  control  functions  defined,  the  iterative  procedure  for
         the  solution  of the  system  (9.5.4)  can  proceed  as  follows:

         1.  A  starting  grid  is  generated  by  unidirectional  linear  interpolation  from  the
            boundary  points.  However,  since  the  final  grid  does  not,  in  general,  have
            uniform  spacing,  the  starting  grid  is  stretched  using  appropriate  algebraic
            formulas  according to  input  values  of the  cell spacing  at  the  inner  and  outer
            boundaries  (ASi  and  AS2).
         2.  Once  the  starting  grid  is  defined,  the  system  (9.5.4)  is  solved  numerically
            using Successive Line Over-Relaxation.  For the  first  iteration,  zero values  are
            assumed  for  p,  q,  r,  and  s.  All  the  fixed  derivatives  appearing  in  equations
            (9.5.12)  are  computed.
         3.  Given  the  initial  conditions  or  the  results  from  the  previous  iteration,  x^
            and  y vr],  at  j  =  1 and  j  =  j m a x  are  computed  using  special  one-sided  differ-
            ence formulas  suggested  by Sorenson  [5]. The  functions p, q, r,  and  s are  then
            evaluated  at  the  boundaries  using  equations  (9.5.7)  and  (9.5.12).  The  con-
            trol  functions  P  and  Q  are  then  evaluated  at  all  grid  points  using  relations
            (9.5.6).  For  numerical  stability,  these  control  functions  are  under-relaxed  to
            a  degree  specified  by  the  user.
         4.  Another  step  of  Successive  Line  Over  Relaxation  solution  is then  performed
            on  the  system  of  equations  (9.5.4).
         5.  Solutions  steps  3 and  4  are  repeated  until  convergence  is  attained.

         A typical  C-type  algebraic  grid  is shown  in Fig.  9.13a.  The  outer  boundaries  for
         this  type  of  grid  consist  of  a  semi-circle  and  a  rectangle.  The  r\ — 0  boundary
         moves  forward  from  the  rear  boundary  to  the  trailing  edge,  clockwise  around
         the  airfoil,  and  then  rearward  again. The  77  =  constant  family  of lines  form  open
         curves  resembling  the  letter  C.  The  £ =  constant  lines  join  the  inner  (airfoil)
         boundary  to  the  outer  boundary.  This  grid  was  used  as  the  starting  grid  for
         the  elliptically  generated  grid  shown  in  Fig.  9.13b.  The  grid  consists  of  70  x  30
         nodes,  with  a  spacing  of  0.3  times  the  airfoil  chord  c  at  the  outer  boundary,
         and  0.015c  at  the  airfoil  boundary.  Fifty  iterations  were  performed  to  obtain
         this  grid,  with  an  over-relaxation  factor  of  1.3  applied  on  the  main  equations,
         and  an  under-relaxation  factor  of  0.05  applied  to  the  control  functions.  The
         final  grid  obtained  is  relatively  smooth  and  exhibits  the  proper  characteristics.
         Figure  9.13c  provides  as  closer  view  of  the  grid  in  the  vicinity  of  the  airfoil,
         where  the  orthogonality  of the  boundary  cells  can  be  observed.
   286   287   288   289   290   291   292   293   294   295   296