Page 83 - Computational Fluid Dynamics for Engineers
P. 83

68                                                 2.  Conservation  Equations



            In  the  hyperbolic  case,  with  a  >  0,  c  <  0  and  using  Eq.  (2.6.4),  we  obtain
         from  Eq.  (2.6.3)

                                 2
                                        2
                                d ?/  <9 ?/  =9w   = 9w
                                            d
                                                   g
                          Lu  =  ^ ^ 2  "  -£=2 + 7 ^  + —  +fu  =  h      (2.6.6)
                                        2
                                <9x   dy      dx     dy
         Here  the  leading  terms  form  the  wave  operator  and
                                   2
                                             2
                                  d u/dx 2  -  d u/dy 2  =  0
         is the  wave equation,  so that  any  hyperbolic  equation  in the  plane  is  analogous
         to  the  wave  equation,  with  additional  lower-order  terms.
            In  the  parabolic  case,  with  a  >  0,  c =  0,  e  <  0 and  using
                                   S = - 4 ,    £ = R                      (2-6.7)

        rather  than  Eq.  (2.6.4),  we  rewrite  Eq.  (2.6.3)  as

                                    2
                                   d u    du   =du
                              L l /            d
                                 =  T^2 2  ""  7SE + ^H  +  fu  =  h       (2.6.8)
                                   dx     dy     dx
        The  leading  terms  here  form  the  heat  or  diffusion  operator,  that  is
                                    2
                                   d u/dx 2  -  du/dy  =  0
        is the  heat-conduction  equation  with  u  representing  a  nondimensional  temper-
        ature.
           One  of the  basic  differences  between  the  various  types  of  partial-differential
        equations  is  in  their  "domains  of  dependence"  and  their  "regions  (or  domains)
        of  influence."  Suppose  each  Eq.  (2.6.1)  is  to  be  solved  in  some  region  R  of
        x7/-space  with  boundary  B.  Then  the  domain  of  dependence  of  a  point  P  in
        R  consists  of  all  those  points  on  B  which  are  required  in  order  to  uniquely
        determine  the  solution  at  the  point  in  question  (see  Fig.  2.3).  Conversely,  the
         "region  of  influence"  of  a  point  P  consists  of  all  those  points  in  R  at  which  the
        solution  is  altered  when  a  change  in  the  solution  at  the  point  P  occurs  (see
        Fig.  2.4).  In  general,  elliptic  equations  have  the  property  that  the  domain  of
        dependence  of  any  point  is  a  curve  or  surface  completely  enclosing  the  point,
        as  in  Fig.  2.3a.  In  parabolic  and  hyperbolic  equations  this  is  not  the  case.  The
        extent  of the  domain  of dependence  of  a  point  is determined  by the  intersection
        of  the  so-called  characteristic  curves  through  that  point  with  the  "physical"
        boundary  B.  The  characteristics  also  form  part  of the  boundary  of the  domains
        of  influence,  see  Fig.  2.4.  The  total  number  of  characteristics  is  equal  to  the
        number  of dependent  variables.  For  a parabolic  system  they  all  coincide  with  a
        line normal  to  the  timelike  direction  (Fig.  2.4b),  whereas  they  are  distinct  for  a
        hyperbolic  system.  An  elliptic  equation  has  no  real  characteristics.  For  further
        details,  see  Hirsch  [11].
   78   79   80   81   82   83   84   85   86   87   88