homogeneous stress for a tensile test

When you block only axial displacement, the model is underconstrained (rigid body motions occur). You have to constrain all 6 RBM’s with boundary conditions. In case of tensile test this can be done in one of two ways: either fix the end of the sample and read stresses away from this constraint or utilize symmetry of the model.

Ok,

interesting, then it means that i need anyway to fix completely one side, but i don't get the point when you say that i can use the symmetry of my model.

what are you refering to ?

thanks for your answer

If your model’s geometry is symmetric then you can draw only part of it (one-eighth should be possible in this case) and apply symmetry boundary conditions to the regions of cut. These will block displacements in normal directions hence eliminating rigid body motions. If the sample is cylindrical then axisymmetric model is yet another option.

What kind of mesh do you have? Is it a shell mesh representing a long test sample with a rectangular section? Or, is it a solid mesh of a circular rod?

Ok, i give you the dimensions of the solid i m stuying (rectangular section, width=20 mm, height=40 mm and length=100 mm). I'm a beginner in Patran so considering the profile of the part I have chosen to work with an hexahedral mesh (solid mesh i suppose). What is a shell mesh ?, what is the advantage to work with this type of meshing ?

Thanks

Here's some suggestions:

[ol 1]
[li]Make sure you have a relatively fine mesh if you want to look at stresses. Based on your dimensions, you may want to use an element edge length of 2 mm or even 1 mm.[/li]
[li]By default, Patran will plot averaged results for fringe plots (which will miss absolute peak stresses). You can't turn averaging off from a "Create->Quickplot" you need to use "Create->Fringe" in the Results Form in Patran. I believe the actual setting is something like, "Domain: None" in one of the fringe plot settings.[/li]
[li]When you removed the Tx constraint from one end of the mesh, the model was free to move in the Y-direction which resulted in the negative pivot ratio error in the f06 output file. This is because you can't have rigid body motion of your model in a linear statics solution. You need to constrain at least one node in the model from y-displacement.[/li]
[/ol]

Question: When you successfully ran your model with the full constraints on one end, what didn't you like about the results? If I was going to simulate a tension test, I would constrain all degrees of freedom at one end and pull on the other end.