Page 152 - Process Modelling and Simulation With Finite Element Methods
P. 152

Extended Multiphysics                 139

             pressure  at which separation of reacted  propylene and unreacted  cyclopropane
             will  occur.  The compressed gas is then condensed to bubble-point liquid and
             feed to a distillation column.  The column has 31 ideal stages with the feed onto
             the stage 16.  It operates with a total condenser and a molar reflux ratio of 8.4
             producing a distillate flow rate of 0.292 kg-molh.”
          Sound  familiar?  Such  scenarios populate  modules  on  “Process  Engineering
         Fundamentals.”   Why  is  it  extended  multiphysics?   Each  unit  operation
         constitutes  its  own  logical domain, connected to  the  others by  entry  and  exit
         points.  In the conceptual  design stage of  such a plant,  the unit  operations are
         treated  by  simplified  models  to  permit  facile exploration of  the  configuration
         space.  Process integration  by  means of  recycle  and  heat  exchanger networks
         adds greater  complexity  to  the  flowsheet,  and greater  scope for economies in
         operating and capital costs.  Eventually,  however,  the  process engineer has to
         give  detailed  designs  for  such  plant.   These  days  that  includes  process
         simulation,  typically  including  optimisation,  parametric  sensitivity  studies and
         transient analysis.  And even if the plant were designed a generation ago, process
          studies of this nature are common for retrofit  and optimisation.  In many cases,
         plant  were  over  designed  by  30-50%  (since  such  flexibility  is  a  common
          safeguard  in design), so now that  the plant is operational, efficiency  savings of
          30-50% should be achievable.  Thus has grown the burgeoning field of process
          systems optimisation.  And this is a regime for extended multiphysics.  If any of
         the  unit  operations  in  Figure  4.1  are  to  be  modelled  in  detail,  that  usually
         involves  a  spatial-temporal  PDE  where  the  simplified  model  used  in  design
         might have been a lumped parameter model.  For instance, suppose the reactor in
         Figure 4.1 is CSTR reactor  which is jacketed by a bath of its product liquid  (at
         500°C) before entering the heat exchanger proper.  Temporal fluctuations  in the
          reactor  temperature  propagate through the bath to the heat exchanger, requiring
         control  action, which  in  turn  leads to transients in  the  compressor operation.
          These  feed  into  the  distillation  column.  Presuming  the  separated  unreacted
          cyclopropane  is  recycled  back  to  the  feed  to  the  reactor,  the  temperature
          fluctuations  into  the  distillation  column  will  have  translated  into  composition
          fluctuations  in the recycle  stream, which will then effect the reactor  conversion,
          starting  the  whole  cycle  again.  The  plant  should  be  designed  to  dampen
         fluctuations  back  to  the  set  point,  rather  than  reinforce  them.   Extended
         multiphysics is in play at every level of process coupling.  In the linear flowsheet
          of Figure 4.1, it is possible to isolate the modelling of each unit operation, since
          the entry and exit points are the only overlaps.  It is still extended multiphysics if
         you  want  to  link  them  up  in  FEMLAB,  but  the  linkages are  simple.  But  if
         process  integration  enters  in,  then  the  linkages  may  be  more  thorough.  For
          instance, in distillation columns, differential heating and cooling of stages can be
          done to influence separation  efficiency  (with multiple  entry  and exit points for
          various  “fractions”).  These  streams  can  be  crossed  for  heat  integration and
   147   148   149   150   151   152   153   154   155   156   157