Page 125 - Process Modelling and Simulation With Finite Element Methods
P. 125

112        Process Modelling and Simulation with Finite Element Methods


          to “resolve” the noise, which is not particularly sensible.  Solution - turn the feature
          off!  Select  under  Solver  Parameters  the  Scaling  box,  and  in  the  pop-up
          dialogue box, select “none” rather than the default automatic.
             The second point about the pressure datum is discussed in detail in Chapter 5.
          The solutions here  are convergent, but  greater  accuracy and  faster  convergence
         result by specifying a pressure datum.
             To  implement  a  buoyant  force  that  varies  with  temperature,  edit  the
          appropriate subdomain setting for the IC NS mode to be:
                                     Fy = RaT                          (3.4)

          Actually,  the proper  dimensionless term is as below, but  given our scaling for
          temperature  set  in  the  constants,  they  are  equivalent.  If  you  wish  to  use
          temperatures  with  units,  then  edit  TO  and  TI  appropriately,  and  use  this
          substitute for (3.4).  Later, we will use temperatures with units, so I recommend
          its use:






          Add  Ra  to  the  constants list  and  set it  to Ra=l. Now  use  the  Restart  toolbar
         button,  which  uses  the  noise  velocity  field  as  an  initial  condition.  This  is
          actually a useful technique for introducing noise as an initial condition. Figures
          2.1 and 2.2 show the streamlines and isotherms at steady state.
             A quantity of central interest in thermal convection studies is the heat flux.
         The  natural  dimensionless  measure  of  heat  flux  is  the  ratio  of  the  total  time
          averaged rate of heat transport to the conductive rate of heat transport, termed the
          Nusselt number:



                                                                       (3.5)


         The overbar represents spatial average and the brackets time average.  FEMLAB
         permits  the  computation  of  the  separate terms  in  the numerator  as  subdomain
          integrations  under  the  Post menu.  Compute  the  dfluxT-cc  (conductive  heat
          flux) and cvfluxT-cc  (convective heat flux) integrated over the whole  domain.
          What is the corresponding Nusselt  number?  Does  this  surprise you  given the
          scale of  the streamfunction in Figure 2.1? Now try simulating Ra=50.  Do you
          get a converged solution?
   120   121   122   123   124   125   126   127   128   129   130