Page 213 - Process Modelling and Simulation With Finite Element Methods
P. 213

200        Process Modelling and Simulation with Finite Element Methods

          Since the eigenvalues reported by this method are the decay rates if positive, we
          can conclude that all the eigenmodes are decaying, although one is neutral.  SO
          the  eigensystem  stability  analysis  shows  that  the  viscous  fingering  instability
          simulated  here  is  unconditionally  stable,  even  at  parameter  values  that  the
          analytic theory finds a long wave instability.
             We can visualize  the eigenmode solutions by tricking  FEMLAB 's  built-in
          postplot facility to see them as solutions.
          >>fem.sol.u=V(:,l)
          corresponds  to  the  eigenvector  with  the  smallest  magniture,  h=0.0000.  A
          standard  use  of  postplot to  give  concentration  contours  visualizes  the
          eigenfunction.
          postplot(fem,.  . .
              'geomnum', 1 ,. . .
              'context','local',. . .
              'contdata', ( 'c','cont','internal'} ,. . .
              'contlevels',20,.. .
              'contstyle','color', ...
              'contlabel','off ,...
              'contmaxmin','off', ...
              'contbar','on', ...
              'contmap','cool', ...
              'geom',  'on',. . .
              'geomcol','bginv',. . .
              'refine', 3, ...
              'contorder',2,.. .
              'phase',  0, ...
          'title',  'Time=200  lambda=0.0000 Contour: concentration of c  ', ...
              'renderer','zbuffer',. . .
              'solnum', 15, ...
              'axisvisible','on')
          Figures 5.12 and 5.13 demonstrate that the eigenvectors found this way represent
          discernable  patterns.  That  they  all  decay  implies  that  there  is  no  pattern
          formation due to instability. Note that the eigenmodes shown satisfy the pde and
          the  appropriate  homogeneous  boundary  conditions  as  well  - no  vertical  flux
          (flat) and uniform outflow.
             We are left in this subsection with the apparent disagreement between linear
          stability theory  [ 131 and linearization of the full solution.  Careful examination
          of  the  theory  and  the  simulations,  however,  suggests  that  the  simulations  in
          FEMLAB  are  too  restrictive  in  the  imposition  of  uniform  inlet  and  outlet
          boundary conditions.  Logically, if the inlet and outlet boundary conditions are
   208   209   210   211   212   213   214   215   216   217   218