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

You are not logged in.

#1 2019-11-22 10:35:51

Registered: 2017-12-31
Posts: 59

Mohr-Coulomb convergence issue on a D_PLAN model

Hello everybody,

I am experiencing big throubles with Mohr-Coulomb applied on a D_PLAN model.

I have two major problems not still solved also after reading several topics:

1) This simple D_PLAN model (annexed) doesn't converge. I tried to increase precision (PREC=16), I changed meshing, I tried to reduce timestep and to change some parameters in Newton alghoritm, but the result is always the same: in a total defined analysis time of 10, it stops at 5 for convergence issue.

2) In a second model I introduced some Poutre elements to simulate retaining wall, its foundation and an anchoring cable. It works until the simulation is elastic ( Resn_01 in the comm file), but when Mohr_Coulomb starts (Resn_02) it stops at the first timestep for problems with poutre elements.

   ! <EXCEPTION> <CALCUL_37>                                                                                   !
   !                                                                                                           !
   ! Erreur utilisateur :                                                                                      !
   !   -> Le TYPE_ELEMENT MECA_POU_D_E  ne sait pas encore calculer l'option:  RIGI_MECA_ELAS.                 !
   !                                                                                                           !
   !   -> Risques & Conseils :                                                                                 !
   !    * Si vous utilisez une commande de "calcul" (THER_LINEAIRE, STAT_NON_LINE, ...), il n'y a pas          !
   !      moyen de contourner ce problème. Il faut changer de modélisation ou émettre une demande d'évolution. !
   !                                                                                                           !
   !    * Si c'est un calcul de post-traitement, vous pouvez sans doute "éviter" le problème                   !
   !      en ne faisant le post-traitement que sur les mailles qui savent le faire.                            !
   !      Pour cela, il faut sans doute utiliser un mot clé de type "GROUP_MA".                                !
   !      S'il n'y en a pas, il faut faire une demande d'évolution.                                            !

Last edited by ingladisa (2019-11-22 10:43:20)

Soil_Example.zip, Size: 420.65 KiB, Downloads: 39


#2 2019-11-25 11:31:56

Registered: 2017-09-08
Posts: 3

Re: Mohr-Coulomb convergence issue on a D_PLAN model


as for the first model's issue, find attached an attempt to solve it.
I saw a main problem in BCs application: a std. "geotechnical" condition for boundary displacement, especially with gravity-driven problems such as slope stability, should always let free vertical displacement except from base nodes/elements.
Plus always consider the constitutive model limitations: in case of Mohr-Coulomb, you do not have the ability to follow post-peak deformation as the numerical model would become unstable after reaching ultimate state.
In fact the attached model converges only when allowing higher values of drained cohesion (namely around 5 kPa). Below this threshold the model will fail (as the slope would).
Since post peak isn't so relevant in slope stability problems, I see no issue in the overall model set-up. For more complex interaction problems (such as retaining structures) please consider more suitable constitutive models to grasp more realistic deformations. For sand-like I suggest CJS, although you have to deal with harder parametrization. You can also try with more Plaxis-like models such as HS Small via MFront (you can find models' formulation at e.g. soilmodels.com website).

Hope this helps

Soil_Example_mod.zip, Size: 689.64 KiB, Downloads: 37


#3 2019-12-10 18:24:00

Registered: 2013-09-26
Posts: 107

Re: Mohr-Coulomb convergence issue on a D_PLAN model

We recently worked on a MFront implementation of the MohrCoulomb behaviour.
If you are interested by giving it a try, you may contact me via the MFront website (can't post link on the forum, sorry).
Thomas Helfer