Page 82 - Computational Fluid Dynamics for Engineers
P. 82

2.6  Classification  of  Conservation  Equations                       67



         disturbances  in  a boundary  layer  on  a  flat  plate  by  means  of  a vibrating  ribbon
         held  parallel  to  the  plate  and  normal  to  the  freestream.  The  measured  neutral
         stability  curve,  critical  Reynolds  number  and  amplification  rates,  discussed  in
         Chapter  8,  were  found  to  agree  well  with  the  predictions  of  the  linear  stability
         theory.



         2.6  Classification  of  Conservation    Equations

         To understand  some  of the  basic  ideas  underlying  the  classification  of  conserva-
         tion  equations,  we  examine  the  general  linear  second-order  partial-differential
         equation  in  two  independent  variables:

                       2
                               2
                                       2
                      d u   _  d u    d u     Bu    du    „    7 /  ,      /rt  ^  ^
               LU         +  b    +      +  d   +     +  /U  HX V)        (2 6 1}
                  "  °&*    ^dy     °W      d~x   %        "   >           " -
         Rotation  of the  xy-plane  into  the  x^-plane  by  the  coordinate  transformation,
                         x  =  x cos 6 +  y sin #,  y  —  — x sin 6 +  y cos 6  (2.6.2)
         allows  Eq.  (2.6.1)  to  be  transformed  to
                                        2
                                2
                         -    _d u    _d u    -du   _du    p                / n / l o .

        This  removal  of the  mixed  derivative term  by  choosing  6 so that  b =  0 is  similar
        to  the  transformation  applied  to  the  general  quadratic  equation;  the  new  coor-
        dinate  axes  are  aligned  with  the  principal  axes  of the  conic  section.  Indeed,  we
         adopt  the  terminology  for  the  conies  to  classify  the  differential  equations.
           Thus  if  ac  >  0  (i.e.,  a  and  c  have  the  same  sign),  we  call  Eq.  (2.6.3)  an
         elliptic  equation;  if  ac  <  0,  we call  Eq.  (2.6.3)  a  hyperbolic  equation;  and  finally
         if  ac  — 0  (but  both  do  not  vanish),  we  call  Eq.  (2.6.3)  a  parabolic  equation.  As
                                     2         2
        with  the  conies,  we note that  b  — 4ac  — b  — 4ac  is an  invariant  of the  rotation.
                                                                      2
        Thus  we say that  Eq.  (2.6.1)  is elliptic,  hyperbolic,  or  parabolic  if  b  — 4ac  <  0,
         >  0,  or  =  0,  respectively.
           In the elliptic case we can stretch  or compress the coordinates  by  introducing

                                 x=-=,         V=^f                        (2.6.4)

         and  we get  finally

                                2
                                       2
                                d u   d u   =du    =du
                          Lu                d
                             =  7^2 2  +  ~^2 + ^E  +  ^  +  fu  =  h      (2.6.5)
                                <9I   9 ^    OS
         The  leading  term  here  is the  well-known  Laplacian  in  the  xp-plane.  So  we  see
         that,  by  a  simple  rotation  and  stretching  of the  coordinate  system,  any  elliptic
         equation  in  the  plane  can  be  reduced  to  the  Laplacian,  plus  lower-order  terms.
   77   78   79   80   81   82   83   84   85   86   87