pull-out example (updating contact)

[abdelm4]

Hi,

This is a simple example of 48 mm length rebar, embedded in LDPM concrete mesh with the same thickness. The LDPM mesh is in contact with an elastic concrete mesh via master/slave constraint. There are ties between the rebar and LDPM mesh.
The rebar was pulled for 80 mm, exceeding the contact length.
As you can see, the rebar still experiencing forces/stresses even after being pulled for the full contact length.

I will attach the rest of modeling files in a reply.

Regards
Mohammed

[abdelm4]

Please find attached

[abdelm4]

For more explanation, the bar pulls out with constant force after friction even though it is totally outside the concrete. We worked more on that and tested Mars on CCI 'RPI supercomputer'. The openMP version gives exactly same results while the MPI gives different ones.
We also checked the global update option and confirmed that it does not create additional problems and in OMP the results are identical with and without global update.

[zhouxinwei]

We believe there has been a bit of confusion on our side. The BeamTetConstraintList with the RebarConcreteInteraction model for large deformation pull-out was working correctly at the end of January using the the RebarConcreteInteraction keyword in the list definition line. At the time, the previous class, that did not include large deformation and pull-out, was kept in the code but not accessible through input commands. Due to a misunderstanding, in a later version of Mars, the keyword RebarConcreteInteractionLD was assigned to the large deformation formulation and the keyword RebarConcreteInteraction to the older small deformation formulation. As such, the Run1.mrs input file that Mohammed attached to this forum started using the wrong formulation not valid for pull-out. If you run that input with the RebarConcreteInteractionLD keyword, it would work correctly. However, we don't want to keep both formulations available because the large deformation works all the times, and there is little extra cost, with respect to the previous formulations. We apologize for the inconvenience and we are fixing this immediately.

We rerun Run1.mrs by changing RebarConcreteInteraction to RebarConcreteInteractionLD and the updated load-displacement history is shown in Figure 1.

Again, we want to keep it simple. There will be only the large deformation formulation available and accessible using the RebarConcreteInteraction keyword.

The issue is fixed in version 2016.3.02.

[zhouxinwei]

The MPI issue is fixed in version 2016.5.00. Now the MPI simulation should give the same result as OpenMP.