Page 181 - Computational Fluid Dynamics for Engineers
P. 181

168                        5.  Numerical  Methods  for  Model  Hyperbolic  Equations


         This  equation  gives  the  stability  requirement  for  the  solution  of  the  difference
         equation,  Eq.  (4.4.3a),  by  placing  a  constraint  on  the  size  of  the  time  step
         relative  to  the  size  of the  mesh  spacing.
            The von Neumann  or Fourier stability method  can also be applied to  examine
         the  stability  characteristics  of  hyperbolic  equations.  To  illustrate  this,  consider
         the  linear  convection  equation,  (5.1.1),




         and  replace  the  time  derivative  with  a  forward  difference,  Eq.  (4.3.8),  and  the
         spatial  derivative  with  a  central  difference.  Eq.  (4.3.7):  this  gives  the  simple
         Euler  explicit  scheme

                                                    ~
                                         +  c  l+ \  A  l l  =  0         (5.7.17)
                                  At             2Ax
         which  can  be  shown  to  be  unstable  by  applying  the  von  Neumann  stability
         analysis  (see Problem  5.7).  If  in the time  derivative, the term  uf  in Eq.  (4.7.17)
         is  replaced  by  an  average  value  between  grid  points  (i  -f  1)  and  (i  — 1),  so  that
         Eq.  (5.7.17)  becomes


                                  At           + C   2Ax      _ 0         ( 5 7 1 8 j
                                                                            - '
         the  scheme  becomes  stable  by  \cAt/Ax\  <  1.  This  differencing  in  the  time
         derivative  is  called  the  Lax  Method.
            To  apply  the  von  Neumann  stability  analysis  to  Eq.  (5.7.18),  again  assume
         an  error  in  the  form  given  by  Eq.  (5.7.7)  and  substitute  this  form  into  the
         difference  equation,  (5.7.18); the  amplification  factor  becomes

                             e aAt  =  cos(k mAx)  -  io  sin(fc m Az)    (5.7.19a)
         and  in terms  of phase  angle  0,  it  can  be  written  as

                                                    2
                                   2
                  G  =  |G| e^  =  ^cos (kmAx)  +  a 2  sin (fc m Ar)  e^  (5.7.19b)
         where
                                     t
                                 (j)  = a n  - 1  [—crtan(fc mZ\x)]
                                  aAt
         The  stability  requirement  \e \  <  1, when  applied  to  Eq.  (5.7.19)  yields

                                       a = cf x  < 1                       (5.7.20)
         where  a  =  cAt/Ax  is  called  the  Courant  number.  Eq.  (5.7.20)  is  known  as  the
         Courant-Friedrichs-Lewy  condition,  generally  referred  to  as the  CFL  condition.
            Figure  5.4  shows  a  polar  plot  of  Eq.  (5.7.19b)  for  several  Courant  numbers.
         For  a  =  1 all  frequency  components  are  propagated  without  attenuation  in  the
   176   177   178   179   180   181   182   183   184   185   186