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

You are not logged in.

#1 2018-08-23 09:30:17

udc_Stdnt
Member
Registered: 2011-01-27
Posts: 29

Erreur numérique (floating point exception).

Hello,

I'm trying to run a vibro-acoustic simulation with a large mesh of the model (aprox. 1.5e6 nodes); when the stiffness matrix is calculated, the following error appear in the calculation of the center of gravity of an element:

  !--------------------------------------------------------------------------------------------------------------------------------------!
   ! <EXCEPTION> <DVP_2>                                                                                                                  !
   !                                                                                                                                      !
   ! Erreur numérique (floating point exception).                                                                                         !
   !                                                                                                                                      !
   ! --------------------------------------------                                                                                         !
   ! Contexte du message :                                                                                                                !
   !    Option         : RIGI_MECA                                                                                                        !
   !    Type d'élément : MEDKQU4                                                                                                          !
   !    Maillage       : malla_ac                                                                                                         !
   !    Maille         : M793346                                                                                                          !
   !    Type de maille : QUAD4                                                                                                            !
   !    Cette maille appartient aux groupes de mailles suivants :                                                                         !
   !    Group_Of_All_Faces                                                                                    !
   !    Position du centre de gravité de la maille :                                                                                      !
   !                                                                                                                                      !
   ! x=-179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326 !
   ! 889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204 !
   ! 144723168738177180919299881250404026184124858368.000000                                                                              !
   ! y=-179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326 !
   ! 889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204 !
   ! 144723168738177180919299881250404026184124858368.000000                                                                              !
   ! z=-179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326 !
   ! 889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204 !
   ! 144723168738177180919299881250404026184124858368.000000                                                                              !
   !                                                                                                                                      !
   !                                                                                                                                      !
   !                                                                                                                                      !
   ! Il y a probablement une erreur dans la programmation.                                                                                !
   ! Veuillez contacter votre assistance technique.                                                                                       !
   !--------------------------------------------------------------------------------------------------------------------------------------!

It seems that the calculation ends due the high value of the coordinates calculated (the mesh of the model is around 12000x4000x4000 mm): has anyone face this kind of issue before? Is the openBlas library related with this kind of problem? ( I read in older post about Floating Point exception in the newest version of Code Aster).

Thanks in advance and best regards,
Alex.

Offline

#2 2018-08-23 10:09:08

jeanpierreaubry
Guru
From: nantes (france)
Registered: 2009-03-12
Posts: 3,210

Re: Erreur numérique (floating point exception).

hello

the coordinates values of this element are properly astronomic
there is probably something wrong in the mesh
i suggest you open it in a viewer,
check its dimension and try to locate this element

jean pierre aubry

Offline

#3 2018-08-24 17:17:07

udc_Stdnt
Member
Registered: 2011-01-27
Posts: 29

Re: Erreur numérique (floating point exception).

Hi Jean Pierre,

thanks for the answer; looking for the element in the mesh, is a common quadrangular element which does not have any special geometry (see attached image). The coordinates of the node of the element are inside the mesh limits, so I think that the problem should appear from another reason.

Alex.


Attachments:
Element_793346.png, Size: 50.51 KiB, Downloads: 42

Offline

#4 2018-08-28 19:00:36

udc_Stdnt
Member
Registered: 2011-01-27
Posts: 29

Re: Erreur numérique (floating point exception).

Hello, I attach the .mess file of the execution.

During the execution, seems that the projection LIAISON_MAIL use PROJ_CHAM function, which shows several alarms trying to found the maillage_proj_loin_00000005.med file.
   
   !-------------------------------------------------------------------------------------!
   ! <A> <CALCULEL5_48>                                                                  !
   !                                                                                     !
   !  Vous utilisez la commande PROJ_CHAMP ou un mot clé nécessitant de "projeter"       !
   !  des noeuds sur des mailles (par exemple LIAISON_MAIL).                             !
   !  Il y a 2461 noeuds qui ont été projetés sur des mailles jugées distantes.          !
   !                                                                                     !
   !  Les noeuds sont jugés distants si :                                                !
   !   * la distance à la maille la plus proche est supérieure à DISTANCE_ALARME         !
   !     (si ce mot clé est utilisé).                                                    !
   !   * ou si la distance à la maille la plus proche est supérieure à 1/10ème           !
   !     de la taille de cette maille (si le mot clé DISTANCE_ALARME n'est pas utilisé). !
   !                                                                                     !
   !  Les 5 noeuds les plus éloignés ont été imprimés ci-dessus.                         !
   !                                                                                     !
   ! Risques et conseils :                                                               !
   !   * Un maillage constitué des noeuds distants a été imprimé au format MED.          !
   !     La visualisation de ce maillage pourra vous rassurer (ou non).                  !
   !     Le nom du fichier MED est : REPE_OUT/maillage_proj_loin_00000005.med            !
   !   * Le mot clé DISTANCE_MAX permet d'éviter que les noeuds trop distants ne soient  !
   !     projetés (ou "liés" quand on utilise le mot clé LIAISON_MAIL).                  !
   !   * Le mot clé DISTANCE_ALARME permet d'éviter cette alarme.                        !
   !                                                                                     !
   !                                                                                     !
   ! Ceci est une alarme. Si vous ne comprenez pas le sens de cette                      !
   ! alarme, vous pouvez obtenir des résultats inattendus !                              !
   !-------------------------------------------------------------------------------------!

I have done a previous modal simulation with the same mesh, and I don't have this kind of problem, so I suspect that during the execution of this command the mesh does not do a good calculation of the Gravity Center.

Thanks in advance.
Alex.


Attachments:
Acoustic.mess, Size: 798.3 KiB, Downloads: 18

Offline

#5 2018-09-06 12:25:03

biscarri
Member
Registered: 2012-01-25
Posts: 58

Re: Erreur numérique (floating point exception).

It is a very strange situation, to obtain infinite values as the coordinates of the center of gravity corresponding to a regular shaped QUAD element, which suggests there is a divided by zero issue for this calculation.

As the error message says, it may be a programming error...

       ! Il y a probablement une erreur dans la programmation.                                                                                !
       ! Veuillez contacter votre assistance technique.             

Does anybody know a reported bug of this kind?

Lluis

Offline

#6 2018-09-08 12:07:23

udc_Stdnt
Member
Registered: 2011-01-27
Posts: 29

Re: Erreur numérique (floating point exception).

Thanks Biscarri for your comment; I add more information abut the same element. The calculation of the CDG is done in Salome when you select the element, and the values obtained there seem in the correct units range of the mesh (see attached).

Waiting for more ideas to solve this issue...

Best regards,
Alex.


Attachments:
Element_793346_Salome.png, Size: 158.89 KiB, Downloads: 13

Offline

#7 2018-09-08 22:02:45

Volker
Member
From: Chemnitz
Registered: 2016-05-23
Posts: 64

Re: Erreur numérique (floating point exception).

How did you produce your mesh?
And did you try to delete that suspicious element for testing?
Kind regards Volker

Offline

#8 2018-09-14 07:51:56

udc_Stdnt
Member
Registered: 2011-01-27
Posts: 29

Re: Erreur numérique (floating point exception).

Hi Volker, thanks for idea.

The mesh is imported in med format; I have launched a new simulation deleting this elements, and the same problem is produced in another element, so it seems to be recursive.

I have tried to print the Center of Gravity of the elements using POST_ELEM command, obtaining the same error.I 'm not shure that the mesh is the root of the problem...

Best regards,
Alex.

Offline

#9 2018-09-14 08:49:20

jlf
Member
Registered: 2007-11-22
Posts: 194

Re: Erreur numérique (floating point exception).

Hello,

You have the message "<A> <CALCULEL5_48>" and the name, coordinates values of 5 nodes.
Could you gives us these names and values.

Did you visualize the file "REPE_OUT/maillage_proj_loin_00000005.med" to check the projection.
To obtain the med file you must add a 'repe' as 'R' (like result), in  your "astk" study.

JLF

Offline

#10 2018-09-14 11:36:33

biscarri
Member
Registered: 2012-01-25
Posts: 58

Re: Erreur numérique (floating point exception).

Hi jlf, thanks for joining the thread.

This vibroacoustic model consists of a closed vehicle structure, basically meshed using shell elements, and the inner air cavity meshed with 3D elements. Structure and fluid models are coupled through FLUI_STRU interface.

The cavity skin mesh (interface mesh) is decomposed in several groups, and so are the inner surfaces of the structure in an analogous way. Structure and cavity groups are paired to define the link between both meshes using the command LIAISON_MAIL. 

Since cavity skin and structure are not in contact, those 'long projection' warnings are expected. Projection lengths are compatible with the actual geometry of the model.

Before using the real mesh (>1.5Mnodes) we have tested the vibroacoustic analysis process using a simple structure mesh (prism) with a contracted inner cavity mesh, separated from it. Everything was OK, and we obtained coherent results.

The problem arises when using the 'big' mesh. But the error (element center of gravity overflow) apparently does not have relation with LIAISON_MAIL node projection, because it appears when executing ASSEMBLAGE/MATR_ASSE/OPTION=RIGI_MECA command. As it is explained in this thread, the element producing this error is a regular one, without any special distortion.

To be noticed that the same big structure mesh has been used to perform a modal analysis (CALC_MODES) satisfactorily.

Thanks for your help,
Lluis

Last edited by biscarri (2018-09-14 11:38:35)

Offline

#11 2018-09-14 13:17:39

Volker
Member
From: Chemnitz
Registered: 2016-05-23
Posts: 64

Re: Erreur numérique (floating point exception).

Hi udc_Stdnt,
I had a short look on the .mess file.
The names of the groups are similar to meshes produced with hypermesh.
In this case lots of problems are possible (at this point normally I say good night ;-)
Check that M793346 is connected to the neighbour elements.
Can you place at each node of M793346 a constrain in DX, DY, DZ, DRX, DRY, DRZ = 0, just for testing?

Greetings Volker

Offline

#12 2018-09-16 17:07:19

biscarri
Member
Registered: 2012-01-25
Posts: 58

Re: Erreur numérique (floating point exception).

Hi Volker,

The same structure mesh (including element M793346) has been used to perform a modal analysis (CALC_MODES) satisfactorily, so that we are quite sure there is not any kind of modeling problem with this particular shell element.

Thank's for your comments!

Offline

#13 2018-09-17 08:43:29

Volker
Member
From: Chemnitz
Registered: 2016-05-23
Posts: 64

Re: Erreur numérique (floating point exception).

Hi again,
I have done modal analysis without any kinematic constraints (Im not sure that I have done it with CA but with other codes).
Greetings Volker

Last edited by Volker (2018-09-17 08:44:27)

Offline

Board footer