Page 163 - Process Modelling and Simulation With Finite Element Methods
P. 163

150        Process Modelling and Simulation with Finite Element Methods

         into the FEMLAB GUI to carry on.  The definitions of u0 and v0 will now have
         the  desired  effects.  Annoyingly,  the  AddEdit  Couplings  dialogue window  no
         longer launches with this alteration.  But the computational model now solves.
             Advice  from  COMSOL  is  that  there  is  a  simpler  workaround  for  this
         apparent  bug in the coupling  variables,  which  allows you  to  solve this problem
         without  playing  around  with  m-files.  You  basically  define  a  weak  boundary
         mode  (as  opposed  to  a  weak  subdomain  mode).  The  make  sure  to  use  non-
         conservative  formulation  for  the  convection-diffusion  application  mode.  The
         reason for doing this is that the coupling variable or expression you define only
         needs to be accessed on a subdomain level, not on the boundary level.
         Pull down the Mesh menu and select Parameters.

             Mesh Mode
                     >>More
                     Number of Elements in Subdomain:  1 1000 2 10
                     Apply
                     OK
                 0
         As  is  typical  for  FEMLAB,  the  pairs  1  1000  2  10 put  1000 elements  in
         subdomain  1 and 10 in subdomain 2.
             This results in a 1013 element meshing.  Now for the Solver.  Pull down the
         Solver Menu and  select  Parameters.  Check the  Stationary Nonlinear  solver
         box, apply,  and  click on the  Solve button.  With no recycle  and  steady  state,
         FEMLAB finds the same solution for the reactor as before.  You can ramp up the
         recycle  ratio  gradually using  the Re-Start  button  on the Toolbar.  By f=10%, a
         modest variation in the position of the crossover point is noted.  The effect of the
         recycle  is  to  load  even  more  U  into  the  reactor,  as  well  as  to  speed  up  the
         throughput somewhat.
             The transient solution is a hard problem, but potentially more interesting as
         the “capacitor” takes a long time to charge in the buffer tank. Before undertaking
         the transient solution, however, we need to make some modest alterations to the
         PDE coefficients:
         Pull  down  the  Multiphysics menu,  select  surface  mode,  then  pull  down  the
         Subdomain menu and select View as Coefficients. Finally, select Parameters.

             Subdomain Mode (Surface Mode, View as Coefficients)
                     Select the dweak tab, domain 2
                     Replace US-time,  VS-time,  WS-time  with zeros in the
                     US,VS,WS coefficient boxes
                     APPlY
                     OK
   158   159   160   161   162   163   164   165   166   167   168