Page 259 - Computational Fluid Dynamics for Engineers
P. 259

8.2  Solution  of the  Orr-Sommerfeld  Equation                       249



            We  note  from  the  difference  equations  defined  by  Eqs.  (8.2.3)  and  (8.2.5)
         that  the  solution  of  Eq.  (4.4.29)  depends  upon  the  three  parameters  a,  u  and
         R  and  we  can  denote  this  dependence  by  writing

                                       6  =  ^(a,u;,R)                     (8.2.10)

         Recalling  that  a  is  complex,  and  that  u  is  real  in  spatial-amplification  theory,
         the  above  relation  implies  that  the  solution  of  Eq.  (4.4.29)  depends  upon  four
         scalars. With  any  two  of these  scalars  fixed,  the  remaining  scalars  can  be  com-
         puted  in  such  a  way  that  the  missing  boundary  condition  0'(O)  =  0  is  satisfied.
         In  our  finite-difference  notation,  this  corresponds  to  the  condition

                                      f 0(a,uj,R)  =  0                   (8.2.11)


         8.2.2  Eigenvalue  Procedure
         To  discuss the  solution  procedure  for  Eq.  (8.2.11),  let  us  consider  the  computa-
         tion  of  a r ,  u  and  R  for  a  specified  value  of  c^.  In  this  case  since  UJ  is  real,  the
         solution  of  Eq.  (8.2.11)  can  be  obtained  by  fixing  one  parameter  and  solving
         for  the  remaining  two.  The  choice  of the  fixed  parameter  depends  on  the  slope
         of  da r/dR  so that,  for  example,  it  is more  convenient  to  fix  R  and  solve  for  a r
         and  uj away  from  the  critical  Reynolds  number  where  da r/dR  is  small.  In  the
         region  of  critical  Reynolds  number,  da r/dR  is  large  and  increases  as  R  —*  R cr
         and  it  is necessary  to  specify  a r  and  solve  for  u  and  R.  To  explain  these  points
         further,  consider  the  eigenvalue  problem  corresponding  to  small  da r/dR.  Since
         /o  is  complex,  OLI is  given  and  R  is  fixed,  Eq.  (8.2.11)  represents  two  equa-
        tions  with  two  unknowns  (a r,u)  and  the  equations  can  be  solved  by  Newton's
                                    v
        method.  Specifically,  if  (a u  ,uo )  are the  i/-th  iterates,  the  (z/-f- l)-th  iterates  are
        determined  by  using
                                                 v
                                            u
                                     <  + 1  =a r+6a r                   (8.2.12a)
                                       u+1   /     v
                                     u>   =(J  +  8u                     (8.2.12b)
                                           v
                                                   V
         in  Eq.  (8.2.11),  expanding  /o  about  a T  and  UJ  and  retaining  only  linear  terms
         in the  expansion.  This  gives the  linear  system  by taking ^  + 1  =  0 and  f" +  =  0
                                                            /



                              +        f a +         & / ,               <8 2 i3b)
                            ^ (^)" f (^)"                =°                - -
         Here,  for  convenience,  we  have  dropped  the  subscript  0  on  /  and  used  r  and  i
        to  denote  the  real  and  imaginary  parts  of  /  at  the  wall.  The  solution  of  Eqs.
         (8.2.13)  is,                                                     (8214a)
                           ^ = iHi)"^(i)l                                   -
   254   255   256   257   258   259   260   261   262   263   264