Discussion Forum

June 29, 2009 11:27am UTC in response to Wonseok Yoon

Re: Contact modeling and time dependent model(Structural mechanics module)

HI Yoon,
I had the same problem whit a contact simulation into a structural dynamic problem. I solved for this problem by defining the contact pairs among parts being in contact. and I disabled the augmented lagrangian solver (which works only for stationary problems). In other words you force comsol to solve the contact problem without calcutating the pressure field at contact interface. For your application, I think you have to:
1. define contact pair2.
2. assign the penalty normal pressure equal to E/h (where E is the estimated stiffness of your part and h is the mesh size). In this way you are modifing the default boundary constraint equation.
3. solve the problem with the time-dependent solver.

If any error, try to change the E value (more or less stiff).

Good luck

Pasquale

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

August 4, 2009 7:10pm UTC in response to Pasquale Franciosa

Re: Contact modeling and time dependent model(Structural mechanics module)

Hi Pasquale,

I was wondering if you could clarify part of the solution to this problem. If you use the time-dependent solver, there is no augmented lagrangian solver to disable. Any help would be great.

-Geoff

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

August 4, 2009 8:11pm UTC in response to zhaoyu chen

Re: Contact modeling and time dependent model(Structural mechanics module)

Hi, all.

I'm still having a convergence problem. The solver never find the converged solution.
I'm stuck, literally.

Sorry, guys.

Yoon

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

August 13, 2009 1:52pm UTC in response to zhaoyu chen

Re: Contact modeling and time dependent model(Structural mechanics module)

Here are some tips for contact modeling with time dependent solver.

1. Modify the contact pairs definition in order to use a penalty method instead of the default augmented lagrangian technique. To use a penalty contact method you just need to change the normal penalty
factor expression to E/h*n where E is the Young modulus of the material h
the element size and n a parameter that you will play with in order to improve the convergence and the accuracy of the model. Typically, the higher of n, the better. However, too high value of n leads to convergence problem. Need some trade-off.
On the other hand, the expression can be treated as a constant if your model is not linear isotropic material.
Instead of E/h*n, use constant and adjust the constant to get converged solution.
Also, I found the initial contact pressure affects the convergency too. Adjust the pressure if you need.
Sometimes, weak constraints on or off does as well if your model is highly nonlinear.
I don't know why.


2. You need to go to the solve menu > solver manager > solve for page and remove (deselect) the contact pressure variable from the list. With this method you are no longer computing for the contact pressure.


I hope this helps.

Yoon.

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 11, 2010 10:49am UTC in response to John Sidhom

Re: Contact modeling and time dependent model(Structural mechanics module)

Also.. if I'm modeling an articulating surface, such as a ball and joint, do these boundaries have to be defined as contact pairs and identity pairs? or just contact pairs?

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 11, 2010 4:06pm UTC in response to Marc Jouan

Re: Contact modeling and time dependent model(Structural mechanics module)

Hi, Louvet

Thank you for your model. I think this is really good.

Wonseok

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 11, 2010 8:13pm UTC in response to Wonseok Yoon

Re: Contact modeling and time dependent model(Structural mechanics module)

Hey,

The recommendation is sort of helping. The problem is converging to a solution, but I'm also workin in 3d. I seem to run the simulation and the answer isn't harmonic and the video isl ike a 1 second clip of a slight deformation. Do the setting in the solve parameters have to be BDF?

John

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 11, 2010 8:33pm UTC in response to John Sidhom

Re: Contact modeling and time dependent model(Structural mechanics module)

I'm sorry about the previous posts. Apparently, I only set the contact to be identity pairs and not contact pairs. When I set them to contact pairs, I am still getting an error..

Step Time Stepsize Res Jac Sol Order Tfail NLfail
0 0 out 100 12 100 0
Group #1: 50 6 50
Group #2: 50 6 50

Error:
Failed to find consistent initial values.
Last time step is not converged.


In the actual boundary conditions of the pair, I copied and pasted the same expression as your model. However, i replaced the E with the modulus, and the bottom portion with a mesh size of 1, and I ran the analysis. I also created a second group in the solve parameters.

John

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 12, 2010 10:04am UTC in response to Marc Jouan

Re: Contact modeling and time dependent model(Structural mechanics module)

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

March 15, 2010 4:40am UTC in response to Marc Jouan

Re: Contact modeling and time dependent model(Structural mechanics module)

In the boundary contact pn..

your model has the following:

E_smps/hmin_cp1_smps*min(1e-3*5^segiter,1)

where is E_smps and hmin_cp1_smps stored? When I enter this line into my 3d simulation, it can't find the values and can't run the simulation.

When I put my own values in, the simulation still will not converge.

John

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

November 23, 2010 10:43pm UTC in response to Wonseok Yoon

Re: Contact modeling and time dependent model(Structural mechanics module)

Hi, all,

I'm doing transient contact modeling. I tried to adjust the penalty factor and it works in some case. But it didn't work when finer mesh was used. I thought this may be due to smaller mesh size and I adjust the penalty factor to E/n (where n is a constant number), it still worked when coarse mesh was used and didn't work when finer mesh was used. Do you have any idea why this happened?

Any suggestion will be appreciated.

Hongqiang

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse

May 17, 2011 5:34pm UTC in response to

Re: Contact modeling and time dependent model(Structural mechanics module)

What do you mean by Quasi-static contact?

Reply  |  Reply with Quote  |  Send private message  |  Report Abuse