Unexpected lateral displacement under gravity loading

[rafetsisman]

Hi,

I have modelled a 2D soil deposit with "quad" elements by using "pressure independent multiyield" material. Lateral sides and bottom side of the model are fixed in X direction and Y direction, respectively. The soil domain is modelled to be symmetrical. However, gravity analysis results yield non-zero lateral displacement at the center of the domain.

I use "update material stage" command to change material status from elastic to plastic. For elastic stage of the gravity analysis we get nearly 0 (1e-20) lateral displacements, however, for plastic stage lateral displacements are at level of 1e-3. Is there a problem with that material?

Thanks,

Rafet

[Jhno]

I use the PIMY material routinely and there is no such problem with it.
I don't understand thought, you say your artifial boundary are fixed in x direction? There should not be much lateral displacement then.
Maybe start with a very small soil domain and check if this still happens.
I run with over 500 m wide model + structure and I obtain perfect symetry so the problem is in the way you defined the domain.

One thing you should do is plot your domain using GiD software. This way you can visualize the displacement and perhaps see where something goes wrong.

Good luck.

[selimgunay]

For the plastic case, it could be due to numerical errors. What is your convergence tolerance?

[rafetsisman]

Jhno wrote:
> I use the PIMY material routinely and there is no such problem with it.
> I don't understand thought, you say your artifial boundary are fixed in x
> direction? There should not be much lateral displacement then.
> Maybe start with a very small soil domain and check if this still happens.
> I run with over 500 m wide model + structure and I obtain perfect symetry
> so the problem is in the way you defined the domain.
>
> One thing you should do is plot your domain using GiD software. This way
> you can visualize the displacement and perhaps see where something goes
> wrong.
>
> Good luck.

The model I defined is also a small scale one, width of 10 m and height of 10 m. Actually I get smaller displacement values for model with smaller dimensions, but not at the level of 1e-20 etc. I checked the boundary conditions, node locations, element orientations for many times but still have this problem.

Thanks

[rafetsisman]

selimgunay wrote:
> For the plastic case, it could be due to numerical errors. What is your
> convergence tolerance?

The convergence tolerance is set to 1e-3, could this cause that problem? By the way, despite the larger tolerance, convergence of Krylov Newton algorithm occurs at nearly 300th iteration. Is it normal?

Thanks

[Jhno]

No, i don't think it is if you have a simple model.
Did you check example on the OS wiki? They are very well prepared.
http://opensees.berkeley.edu/wiki/index ... _Analysis)

[selimgunay]

1e-3 is a large tolerance, lateral displacement is likely due tho that. Which norm are you using? Please try 1e-8 first and then 1e-6 if that does not work.

Krylow Newton is supposed to accelerate convergence, 300 is not a very large number for iterations, depending on the specifics of your model, it is normal.

[rafetsisman]

Jhno wrote:
> No, i don't think it is if you have a simple model.
> Did you check example on the OS wiki? They are very well prepared.
>
> http://opensees.berkeley.edu/wiki/index ... _Analysis)

Yes, I built the model by using these examples as a basis.

[rafetsisman]

selimgunay wrote:
> 1e-3 is a large tolerance, lateral displacement is likely due tho that.
> Which norm are you using? Please try 1e-8 first and then 1e-6 if that does
> not work.
>
> Krylow Newton is supposed to accelerate convergence, 300 is not a very
> large number for iterations, depending on the specifics of your model, it
> is normal.

I tried 1e-6 tolerance by setting the maximum number of iterations to 2000, and it converged at iteration 1587, is it also normal? By the way, I am using normalized displacement increment test.

[selimgunay]

It is normal as long as the unexpected lateral displacement disappears. Is it smaller in the analysis with tolerance of 1e-6?

[rafetsisman]

selimgunay wrote:
> It is normal as long as the unexpected lateral displacement disappears. Is
> it smaller in the analysis with tolerance of 1e-6?

Yes, it is. The lateral displacements are at level of 1e-13 now (for the tolerance of 1e-6). Thank you for your kind answers.