Robert - I had a good meeting with Yannis this afternoon. He is concerned that the strain condition in the specimen after each time step as currently coded is not stable and suggested that a few iterations be run at each displacement interval to ensure that the element moduli converge before moving onto the next time step or displacement interval. To test his idea I ran a few iterations at a constant displacement and found that the forces do fluctuate quite a bit but appear to converge (somewhat) after about 10 iterations. I presume the moduli of the elements would also stabilize.



On Tue, Jan 18, 2011 at 10:48 AM, Robert Cimrman <cimr...@ntc.zcu.cz> wrote:
On Tue, 18 Jan 2011, Andre Smit wrote:
Robert

I've revised the code loops as shown in [1]. Seems to work. Yes, I still
find myself thinking in FORTRAN :(

You will get used to it :)

The code works for me (and is much shorter and readable now).

As a matter of fact, the functions in sfepy.mechanics.matcoefs could be easily vectorized as well, and subsequently your glame() - it's another EasyToFix issue topic (new issue 150).


I'm looking at the hyperelastic option now. I've also asked Yannis
Tassoulas to consider joining as a co-author. He may have some ideas
regarding plasticity options.

OK. Plasticity would come handy - I still have not got to implementing it (first, I would have to study it :]).


r.

[1] http://paste.pocoo.org/show/322755/


On Tue, Jan 18, 2011 at 8:14 AM, Andre Smit <freev...@gmail.com>
wrote:
     Thanks Robert!

     I'm working on the revisions and updates and will get back.
     The abstract is due at the end of January but I hope to get
     the bulk of the paper done by then as well, which is due in
     Feb sometime i.e. if abstract is accepted.


On Mon, Jan 17, 2011 at 7:13 AM, Robert Cimrman
<cimr...@ntc.zcu.cz> wrote:
     At [1] I have posted the code with "correct" avgqp()
     function. Yours was ok, as long as all the quadrature
     point weights were the same.

     The new implementation defines an auxiliary material
     with the quadrature point data, and calls the
     de_volume_average_mat term. Note that the order of the
     quadrature has to be the same as the one used for
     evaluating the strain.

     Then it seems to me, that you could avoid all those
     loops in
     functions Ft(), Fc(), Ec(), by using numpy vectorized
     operations, i.e. numpy.exp() instead of math.exp().
     Also the failure loops (around line 193) could be IMHO
     vectorized using numpy.where() and fancy indexing.
     Try it as a NumPy excercise :) - you get big speed gain
     by this, and also the code would be shorter and more
     readable. I can give you a hand if you get stuck, of
     course.

     cheers,
     r.

     [1] http://paste.pocoo.org/show/322183/


On Mon, 17 Jan 2011, Robert Cimrman wrote:

     Hi Andre,

     (sorry for the delayed response - I'm
     accommodating in Fribourg again, and have not yet
     wifi at home...)

     On Fri, 14 Jan 2011, Andre Smit wrote:

           Robert

           I think I have something to share.
           How would you feel about co-authoring
           a paper to this conference - mainly
           to further expose SfePy but also to
           check that I'm not applying it
           incorrectly?


     That would be great, thanks for proposing it!

           Using the strain rate dependency
           example you provided I recoded an
           example looking at the failure of a
           cylindrical specimen under
           compression. The cylinder is made up
           of elements that I treat
           individually or discretely, checking
           the strain rates in each and using a
           relationship established in the lab,
           I calculate the compressive
           strengths in the elements based on
           these strain rates. If the stress in
           an element exceeds its strength then
           the element fails and I assign a low
           stiffness to that element. The code
           I'm using is available here:
           http://paste.pocoo.org/show/320646/


     I will look at the code and try it ASAP. What is
     the deadline, anyway?
     A few notes straight from my head follow.

     What happens with the linear elastic assumption
     when you reduce stiffness in some elements? The
     code has some hyperelastic terms, in case they
     would be needed.

           Would you mind looking and commenting
           on this code. I'm a bit skeptical
           about the avgqp function which
           averages the strains at the quad
           points.
           Note that I use a normal distribution
           to vary the moduli of the elements
           initially.  


     To get averaged element values of something
     defined in quadrature points, one can use either
     directly de_volume_average_mat term (a fake
     material definition would then be needed, I can
     show you), or use directly similar code to the
     one in the body of that term.
      
           I've attached the mesh: cylgeo.vtk.

           What I aim to do with the paper is
           demonstrate the influence of specimen
           geometry on strength. For simple
           compression and tension tests (on
           cylinders) this is trivial but with
           indirect tensile tests, for example,
           there is both compression and tension
           failure, and I hope to show using
           FEM that the strengths of materials
           determined using these tests are
           significantly influenced by the
           geometry of the specimen.     


     Interesting!

           The force-displacement curves
           currently generated by the code look
           promising. I'm not sure the analysis
           can be refined by using smaller
           elements?


     The mesh is not very fine, yes. Another way to
     refine the analysis might be to use another
     material relation (hyperelastic?).

     r.

           On Fri, Jan 14, 2011 at 3:19 AM,
           Robert Cimrman <cimr...@ntc.zcu.cz>
           wrote:
                I see :) Good luck then!

                r.


           On Thu, 13 Jan 2011, Andre Smit
           wrote:

                No plans currently - still
           working on an analysis.
                Maybe if I can get it
                done before the end of the month
           :)

                On Thu, Jan 13, 2011 at 2:43 AM,
           Robert Cimrman
                <cimr...@ntc.zcu.cz>
                wrote:
                     Hi Andre,

                     are you planning to go
           there?

                     r.


                On Mon, 10 Jan 2011, Andre Smit
           wrote:

                     FYI:

                   
              
             http://congress.cimne.com/CFRAC2011/frontal/default.asp


                     --
                     Andre


           --
           You received this message because you
           are subscribed to the Google
           Groups "sfepy-devel" group.
           To post to this group, send email to
           sfepy...@googlegroups.com.
           To unsubscribe from this group, send
           email to
           sfepy-devel...@googlegroups.com.
           For more options, visit this group at
           http://groups.google.com/group/sfepy-devel?hl=en.




           --
           Andre

           --
           You received this message because you
           are subscribed to the Google Groups
           "sfepy-devel" group.
           To post to this group, send email to
           sfepy...@googlegroups.com.
           To unsubscribe from this group, send
           email to
           sfepy-devel...@googlegroups.com.
           For more options, visit this group at
           http://groups.google.com/group/sfepy-devel?hl=en.



     --
     You received this message because you are
     subscribed to the Google Groups "sfepy-devel"
     group.
     To post to this group, send email to
     sfepy...@googlegroups.com.
     To unsubscribe from this group, send email to
     sfepy-devel...@googlegroups.com.
     For more options, visit this group at
     http://groups.google.com/group/sfepy-devel?hl=en.



--
You received this message because you are subscribed to the
Google Groups "sfepy-devel" group.
To post to this group, send email to
sfepy...@googlegroups.com.
To unsubscribe from this group, send email to
sfepy-devel...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/sfepy-devel?hl=en.




--
Andre




--
Andre

--
You received this message because you are subscribed to the Google Groups
"sfepy-devel" group.
To post to this group, send email to sfepy...@googlegroups.com.
To unsubscribe from this group, send email to
sfepy-devel...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/sfepy-devel?hl=en.



--
You received this message because you are subscribed to the Google Groups "sfepy-devel" group.
To post to this group, send email to sfepy...@googlegroups.com.
To unsubscribe from this group, send email to sfepy-devel...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/sfepy-devel?hl=en.




--
Andre