Page 273 - Process Modelling and Simulation With Finite Element Methods
P. 273

260        Process Modelling and Simulation with Finite Element Methods

          Optimization Method for ECT Inversion
          A  common  way  of  treating  inversion  problems  is  to  minimize  an  error  norm
          defined  on  the  discrepancy  between  the  predicted  solutions  of  the  forward
          problem and the measured boundary data.  How does it work?  First we need to
          package our forward ECT model as an m-file function.  The first step is to reset
          the  model  m-file  on  the  ECT  forward  problem,  insert  a  stationary  nonlinear
          (actually linear will suffice) solver step, and a boundary integration to round off
          the  computation.  Then  save  the  model  m-file  to  ect.m.  Copy  a  version  to
          ect2.m,  and  we  will begin  surgery  to  package it as  an  m-file  function.  Begin
          with the function call line
          function [ql,q2,q31 =ect2 (el,e2,e3)
          %  FEMLAB Model M-file  for ECT forward problem
          %  Generated 07-Nov-2002 09:53:56 by FEMLAB 2.3.0.145.
          flclear fern
          %  FEMLAB Version
          clear vrsn;
         vrsn.name='FEMLAB  2.3';
         vrsn .ma j or=O ;
         vrsn.build=145;
          fem.version=vrsn;

          %  Recorded command sequence
         ...
          %WZ: Edit the constants to pass the arguments el,e2, and e3
          %  Define constants
          fem.const={  ...
              'eO',    1, ...
                           .
              'el',    el,. .
              'e2',    e2,. . .
              'e3',    e3,. . .
              'e4',    0.05};
         ...
          %WZ: Manufacture the output data
          %  Integrate on subdomains
                                                     ...
         ql=postint(fem,'0.707107*phix+0.707107*phiy',
                  'cont',  'internal', ...
                  'contorder',2, ...
                  'edim',  1, . . .
                  'solnum', 1, . .
                             .
                  'phase', 0, ...
                  'geomnum',l, ...
                  'dl',    24,. . .
                  'intorder',4, ...
                  I context , I local )  ;
          %  Integrate on subdomains
         q2=postint(fem,  '0.707107*phix-0.707107*phiy', ...
                  'cont',  'internal', ...
                  'contorder',2, ...
                  'edim',  1,. . .
                              .
                  'solnum', 1,. .
                  'phase',  0, . .
                             .
                  'geomnum' ,1, . . .
   268   269   270   271   272   273   274   275   276   277   278