Welcome to the forums. Please post in English or French.

You are not logged in.

#1 Re: Code_Aster usage » [SOLVED]Extract Large Sparse Matrix and then use them in Python » 2013-03-13 10:37:22

Hi,
There is no built-in feature to extract a matrix in *sparse* format from Aster to Python.
The feature for dense matrices in implemented in catapy/entete/co_matr_asse.capy.

Otherwise you can rely on Aster's commands such as COMB_MATR_ASSE (to combine assembled matrices), RESOUD (to solve linear systems), and so on....

AA

#2 Re: Code_Aster development » Proposal for 2nd Code_Aster and other FE codes Hackathon in March 2013 » 2013-02-22 14:05:37

Dear Nicolas,

Thank you for proposing this new Code_Aster Hackathon.
Considering the Code_Aster team, I think March 20th would be the preferred date, since we will participate both the Pronet and Users' Day meetings.

Best regards,
N.T.

#3 Re: Code_Aster development » [SOLVED] Implementing a partitioned integration sheme » 2013-02-15 13:15:34

Dear Richard,

You can check the zzzz215c test-case : it gives an example of the reading of pressure field from Code_Saturne, the projection on the solid mechanics mesh and the solution of the solid mechanics problem.

AA

#6 Re: Code_Aster usage » [SOLVED] MODE_ITER_SIMULT with QZ fails with complex MATR_A and MATR_C » 2013-02-05 09:54:21

Dear Lagrange,

There is no problem in Aster with the von Mises stress for shells. You must keep in mind these elements are integrated by layers. So, in order to get a meaningfull post-processing view, you have first to extract the desired field on the desired layers. This is the role of the POST_CHAMP operator.
Then you can get the attached view. Notice that the component VMIS of the SIEQ_ELNO field is displayed ; this field can be discontinuous but, given the simplicity of the problem you are solving, we get a smooth stress field.

Also note that the *_NOEU fields are always smooth since they come from an average process.

Also note the view was produce with the v11 of Aster in which great improvements have been done in the post-processing features.

AA

#7 Re: Code_Aster usage » [solved] Why does the solve time increase if the matrix has fewer eqs? » 2013-01-27 21:13:40

Anna, you can easily specify the number of threads when using multi-threaded BLAS.
For instance, with MKL, just write at the beginning of your command file :

import os
os.environ['MKL_NUM_THREADS']=4

AA

#8 Re: Code_Aster usage » *SOLVED* Thermal flux through a boundary surface » 2013-01-27 21:09:14

You can blame it on me! I had a problem during the commit ; this should be fixed next week. Sorry for the delay.

AA

#9 Re: Code_Aster usage » Iter_glob_maxi » 2013-01-23 16:55:43

Convergence is never guaranted. A reasonnable upper limit is, say, 50. You should activate the automatic time stepping feature thanks to DEFI_LIST_INST.

#10 Re: Code_Aster usage » *SOLVED* Thermal flux through a boundary surface » 2013-01-21 11:38:57

Hi Olivier.

There will be a correction in the next development release, available next thursday.
FYI, I join the command file that greatly simplifies the post-processing.

AA

#11 Re: Code_Aster usage » [SOLVED] Bug CREA_TABLE / RESU / NOEUD avec liste de noeud » 2013-01-17 11:11:45

Jean-Pierre, thank you for confirming you do not encounter problems with the test. FYI, the Python that is used is the one specified in you config.txt file.

Humbert, I am afraid you have a problem on your system. I am using Ubuntu 12.04 with Python 2.7.

AA

#12 Re: Code_Aster usage » [SOLVED] Bug CREA_TABLE / RESU / NOEUD avec liste de noeud » 2013-01-16 13:13:09

Dear Humbert,

I tried the following code on the foma01a test case. It runs fine on my installation.
Can you give it a try.

AA

def function(x):
    z = []
    for s in x:
        z.append(s)
    
    depl_tab = CREA_TABLE(RESU=_F(RESULTAT   = RESU,
                                  NOM_CHAM   = 'DEPL',
                                  NOM_CMP    = ('DX','DY','DZ',),
                                  NOEUD      =  z,
                                  ),
        )


DEBUT(PAR_LOT='NON',)

MA=DEFI_MATERIAU(ELAS=_F(E=200000.0,
                         NU=0.3,),);

MAIL=LIRE_MAILLAGE(FORMAT='MED',);

MODE=AFFE_MODELE(MAILLAGE=MAIL,
                 AFFE=_F(TOUT='OUI',
                         PHENOMENE='MECANIQUE',
                         MODELISATION='C_PLAN',),);

MAIL=MODI_MAILLAGE(reuse =MAIL,
                   MAILLAGE=MAIL,
                   ORIE_PEAU_2D=_F(GROUP_MA='haut',),);

MATE=AFFE_MATERIAU(MAILLAGE=MAIL,
                   AFFE=_F(TOUT='OUI',
                           MATER=MA,),);

CHAR=AFFE_CHAR_MECA(MODELE=MODE,
                    DDL_IMPO=(_F(GROUP_MA='bas',
                                 DY=0.0,),
                              _F(GROUP_MA='gauche',
                                 DX=0.0,),),
                    PRES_REP=_F(GROUP_MA='haut',
                                PRES=-100.0,),);

RESU=MECA_STATIQUE(MODELE=MODE,
                   CHAM_MATER=MATE,
                   EXCIT=_F(CHARGE=CHAR,),);


xx=('N1','N2')
function(xx)

FIN();

#13 Re: Code_Aster usage » [SOLVED] Bug CREA_TABLE / RESU / NOEUD avec liste de noeud » 2013-01-16 08:42:34

Dear Humbert,

I tried to reproduce your problem without success. I used the stable and unstable versions.

I am afraid something is wrong in your installation. Perhaps try to rebuild a fresh version.

AA

#14 Re: Code_Aster usage » [SOLVED]MACR_RECAL: fitting parameters of charge » 2013-01-11 15:27:17

Hi,

We are considering your problem.
You should have some news soon.

AA

#15 Re: Code_Aster usage » AFFE_CARA_ELEM with 100,000 mass elements takes hours to complete » 2013-01-09 13:45:04

Dear Tom,

Indeed Code_Aster's supervisor is not designed to deal with "element by element" assignment, rather "group of element by group of element" sharing the same properties.
You could send us your files so that we manage a profiling of the code. You can use codeaster at edf dot fr.

AA

#17 Re: Code_Aster usage » RBE3 Tested [SOLVED] » 2012-12-17 14:55:00

Thank you very much for your positive feedback.

AA

#18 Re: Code_Aster usage » error in doc regarding NEWMARK schema (R5.02.05) » 2012-12-11 09:53:01

Dear Johannes,

Thanks for the feedback ; we'll answer soon your question.

AA

#19 Re: Code_Aster usage » [SOLVED]11.2.24 much slower than 10.7? » 2012-11-27 15:09:14

Dear Jean-Pierre,

Thanks for the bug report ; we have changed a little bit the process of warnings emission. In the next release (this week), you should see a great improvement.

AA

#20 Re: Code_Aster development » [SOLVED] MPI communitaction » 2012-11-23 13:09:19

Hi,

The MPI_Init is done in the C layer of Aster : $ASTER_ROOT/NEW11/bibc/supervis/python.c.

int
_MAIN_(argc, argv)
    int argc;
    char **argv;
{
    int ierr;
#ifdef _USE_MPI
    int rc;
    rc = MPI_Init(&argc,&argv);
    if (rc != MPI_SUCCESS) {
         fprintf(stderr, "MPI Initialization failed: error code %d\n",rc);
         abort();
    }

AA

#21 Re: Code_Aster usage » [SOLVED] Possible d'imprimer un vecteur assemblé obtenu ? » 2012-10-17 20:52:53

Dear Li,

What you want is :

IMPR_RESU(RESU=_F(CHAM_GD=myVec))

By default, it will be printed in the .resu file (logic unit 8).

AA

#22 Re: Code_Aster usage » first trial of REPERE_LOCAL: disappointing! » 2012-10-17 10:39:04

Dear Jean-Pierre,

Can you provide me with your mesh and CARA_ELEM so that I can try to produce the same output I did with Gmsh?

AA

#23 Re: Code_Aster usage » [SOLVED]annoying feature with AFFE_CARA_ELEM, POUTRE in 11.2.16 » 2012-10-09 16:40:28

I do agree : in the of a very thin square section, the effect of shear should be low and allowing the use of Euler beams.

You can check the AFFE_CARA_ELEM doc to get information of the way we now compute the effect of shear for Timoshenko beams.

AA

#24 Re: Code_Aster usage » [SOLVED]annoying feature with AFFE_CARA_ELEM, POUTRE in 11.2.16 » 2012-10-09 08:24:54

Hi Jean-Pierre,

This is due to the Discontinuity of the values of the areas reduced for the shears in Timoshenko beams.
Here is the text of the commit :

TITLE
    MN17 Discontinuity of the values of the areas reduced for the shears
FONCTIONNALITE
   Objectives:
   ===========
   This file proposes to determine the coefficients of correction of shears for
   modelizations beams with circular and rectangular section of Timoshenko. 
   Currently, the values used are those of the literature. 
   
   For the circular section in Code_Aster one uses a formula given in
   AFFE_CARA_ELEM which makes it possible to evaluate the coefficients. This function is discontinuous in
   0.9, moreover it have no reference there. The beams which require these coefficients of
   correction are those of Timoshenko, one thus uses the method of Timoshenko for
   compute. This method is that which is used by MACR_CARA_POUTRE.
   One makes a study distributed while varying the geometrical characteristics of
   several sections in two dimensions in order to compute the new coefficients, which
   will be interpolated in AFFE_CARA_ELEM. They will be then in coherence with the beams of
   Timoshenko.
   
   Even step for the rectangular section. For these sections the only distinction was
   full or hollow section. By formula of interpolation, just 2 values.
   
   Realization:
   =============
   Circular section:
   --------------------
   - Installation of a script python to generate several meshes 2D representing of
   sections of tubes with various thickness.
   Realization of 11 meshes while varying the report of the radii interior and external
   from 0 to 0.9 (0.0, 0.05,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8 and 0.9). It is specified that
   theoretical value for a ratio of 1 is known (AY=AZ=2).
   - Installation of a study distributed with Code_Aster to determine the values of
   AY=AZ for each generated mesh.
   - Execution of several studies with several smoothnesses of mesh by observing them
   results in order to determine the optimal smoothness to use (when results
   vary according to refinement).
   
   Rectangular section:
   -----------------------
   - Installation of a script python to generate several meshes 2D representing of
   sections of tubes with various thickness.
   Realization of 122 meshes while varying the reports alpha and beta from 0 to 0.95
   (0.0, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.95). One stops with a ratio of 0.95. 
   values ranging between 0.95 and 1 will thus not be envisaged. that corresponds to one
   section which is not any more very physical.
   - Installation of a study distributed with Code_Aster to determine the values of AY
   and AZ for each generated mesh.
   - Execution of several studies with several smoothnesses of mesh by observing them
   results in order to determine the optimal smoothness to use.
   
   Development:
   ---------------
   The results of the two studies are stored in a new routine "coeces.f" . Those of
   the circular part are indicated in the form of a function and those of the part
   rectangular, in the form of a three-dimensions function. This routine is systematically called by
   "aceapo.f" , therefore to each use of AFFE_CARA_ELEM turns over systematically
   function and three-dimensions function. Those are then used in "affgen.f"  where a call to
   "fointe.f"  is carried out to interpolate the values of the coefficients of correction of
   shears according to the geometrical parameters of the sections.
   Checks are carried out in “affgen.f†:
   - circular part: 0<alpha<1,
   - rectangular part: 0<alpha and 0<beta. Moreover, one adds a fatal error if one
   user informs SECTION='RECTANGLE' with a alpha>0.95 or a beta>0.95: 
   	MODELISA10_15
   	“The geometry of the section used is not envisaged by the operand SECTION =
   “RECTANGLE† of AFFE_CARA_ELEM.
   	One of edges is too fine. 
   	Use the operand SECTION = “GENERALE†.†
   U4.42.01 documentation is impacted consequently.

#25 Re: Code_Aster usage » [solved] slit annular plate test case: singular matrix... » 2012-09-18 08:14:06

Pierre,

When speaking about boundary conditions, I was thinking about the fact that Calculix really eliminates the rows and columns of the imposed dof while Aster puts 1 and 0.

Considering the NPREC mechanism, it is as old as Aster (the legacy LDLT algorithm had it).
I think that Spooles has the same kind of mechanism, it is simply not "tuned" as the one of Aster. In Aster, we try to guess the number of lost digits and when it is larger that 8 (a double has lost the half of its significant digits), we stop the solution.

AA