Page 205 - Process Modelling and Simulation With Finite Element Methods
P. 205

192        Process Modelling and Sinzulation with Finite Element Methods
                               Ra        eigenvalue
                                    1705       0.024
                                    1706       0.01 6
                                    1707       0.009
                                    1708       0.002
                                    1709       -0.006
                                    1710       -0.014
                                    171 1      -0.022
                                    1712       -0.029
                                    1713       -0.037
             Table 5.2  Decay rates -hi (largest eigenvalue) with Ra near critical for aspect ratio 1:10.

         number leave no doubt that the critical mode oocurs near Rayleigh number 1708.
         The  triumph  of  eigenanalysis  coupled  with  FEM  is  to  find  numerically  the
          critical  Rayleigh  number  and  approximate  length  scale  associated  with  the
          critical  mode.  Although  the  linear  stability  theory  for  this  problem  is  not
          cumbersome,  for  many  situations  with  non-trivial  and  three-dimensional  base
          states,  that  cannot  always  be  claimed.  FEMLAB,  through  eigenanalysis,
         provides a consistency check on linear stability analysis of stationary states.  The
         numerical  technique  is  fair  more  robust,  however.   Eigenanalysis  can  be
         conducted  on any  solution, even of  transient problems.  Recall equation (5.4)
          shows  that  for  self-adjoint  operators,  the  eigenanalysis  predicts  the  time
          asymptotic dynamics of the linearized system.  For non-self adjoint operators, it
          has  been  demonstrated  that  pseudo-modes  that  are not  eigenmodes  can  grow
          rapidly before the time asymptotic  eigenmodes dominate. Trefethen et al.  [ 101
         identified  spiral pseudo-modes  as leading to transitions to turbulence in Couette
          and Poiseuille flows at much lower Reynolds Numbers than anticipated by linear
         theory. This was confirmed experimentally.  The extent to which eigenanalysis
          of transient flow problems identifies the fastest growing pseudomode in transient
          models for instantaneous states is a largely unexplored area, for non-self-adjoint
          operators.
             Here  we  have  shown  that  for  self-adjoint  operators,  the  FEM  model
          accurately reproduces the predictions of linear stability theory.


          5.3  Viscous Fingering Instabilities

          The Benard problem is a paradigm for instabilities of a stationary state,  Viscous
          fingering is an instability of a non-homogeneous state in motion - a less viscous
          fluid displacing a more viscous fluid.  Figure 5.7 shows the flow configuration
          for  miscible  viscous  fingering,  where  diffusion  tends  to  spread  out  viscous
          fingers  and  oppose  their  formation.  Nevertheless,  viscous fingering is a  long
   200   201   202   203   204   205   206   207   208   209   210