AHD

types of flows:

-- stationary/instationary
-- 2D/3D
   2D for instruction purposes and modeling, 3D for applications
-- compressible: experience is there, only for in-house codes
   experience with fluent-compressible is limited to a few 
   standard flow geometries
-- turbulent: models
   experience with k-epsilon and Reynolds stress flows (cyclone separator)
   Re stress gives some flow properties qualitatively, quantitatively
   results are not always very good
   other models have not been used/tested

-- separation

   for bluff bodies only, most of the time with a sharp corner in the boundary
   so that separation occurs always
   experience with square and round cylinders
   St and average forces OK, fluctuations wrong by (sometimes) orders of magnitude

-- complicated geometries:
   -- complicated physics
      inclusion of electric forces (laminar) with user subroutines
      the MHD module costs 7250 euro extra (eternal, no support included)
      
      inclusion of buoyancy but still shear driven
      buoyancy driven flows we prefer to do with in-house codes
      especially with horizontal heated walls

      inclusion of strong rotation
   
      water tunnel testing (complete circuit, with porous media or real grid)

   -- complicated grid
      grids up to 1000.000 points
      tests on 100.000 points
      grids and geometries generated in gambit (so not very difficult)
      gambit not capable of generating complicated geometries, for instance
      pipes bended in a 3D fashion

      most geomtries consist of a rectangular or cylindrical outline with
      a few objects included



-- geometry outline:
   -- gambit
   -- ProEngineer
      ProEngineer files cannot be generated with the standard (academic) license
      IGES can be generated but only a subset is supported by most packages, for
      not-too difficult geomtries it already went wrong
-- grids
   -- gambit
   all geometries and grids have been generated under GAMBIT, ICEM and other
   seem to have more acpabilities but we have no license

physics:
   -- multi-phase
      we have some experience with VOF (most in-house)
      but most experience is with particle-fluid, lagrangian
      (particles are tracked)
  -- heat transfer
     -- passive (shear driven)
     -- active (buoyancy induced)
  -- combustion
     Only in-house (DNS/LES), no experience with chamical reactions in 
     tabular form etc
  -- other physics
     -- MHD
              
-- parallel
   experience: speed up?
               ease of use?


-- specific geometries/flows/techniques



   -- flow around cylinders sufficiently high Re
   -- flow around sphere sufficiently high Re
   -- electrical field  working on particles
   -- finite size particles
    
-- user subroutines
   -- only simple cases: electrical field, inflow BC

-- typical geometries/flows that you do NOT with Fluent

   multi-physics: adding new equations, trying out FEMLAB
   single-user = 7000 euro
   academic    = 1000 euro?
   not parallel!!


   fluid-structure in fluent is very limited

   low Re number flows (under 5000 on bulk) are not well
   simulated because:
   -- turbulence models are not good at it
   -- the collocated discretisation in fluent is not energy-conserving
   -- all discretisations of the advective terms include a limiter
      (ALSO central differences!)

-- fluent vs other packages
   we have experience too with CFX 4.2 and CFX 5


   CFX 4.2

   CFX 4.2, the block-structured CFX
   CFX 4.2 had a good set of templates (including support!) for user subroutines, 
   which is mostly lacking in Fluent
   plotting etc I did not like (it used an AVS based plotting interface)
   it could do only block-structured grids
   it had problems with very fine grid cells, as are very common in the case
   of airfoils (like 1.e-5 times the cord length)

   2D/axi-symmetric objects were relatively easy to do, one had to use 1 grid cell
   in 1 direction (the z direction) although adapting existing geometries required
   care because the orientation of sub-geomtries had to be correct and no warning
   was issued in case it was not

   CFX 5

   CFX 5: un-structured
   I only tested the first two versions, which were so bad that we can
   call them beta versions 
   the  (unstructured) version was very inaccurate at the time, which was the main
   reason to switch to fluent
   moreover it could not do axi-symmetric or 2D objects, they had to
   be emulated using 3D geometries with homogeneous BC in the 3rd direction
   which led to problems on grid-refinement, especially for axi-symmetric problems
   (2D or axi-symmetric are often used for instruction purposes )
   the one thing I liked was the grid generator, it could build refined hex cell
   boundary layer grids near surfaces, also in 3D, and already had the
   possibilities of directing the refinement of the grid which have appeared only 
   recently in Fluent

   CFX has been bought by Ansys now