Reg. Penetrations in a Contact Analysis Code

Hi,

 

I have developed a FORTRAN code to simulate the molding process of glass to make lenses. Hence, it involves a contact analysis between the glass and molds. The contact is modeled using the master and slave node concept and the mold is assumed to be rigid. For the application of this code, we are looking at highly precise dimensions to capture the curve of the mold on the glass after molding.

Now, I get very good results (curve shape) in a general sense (macro sense). But when I scalse down the position of nodes of the flexible body (glass) to microns, I see a lot of fluctuation/oscillation of the nodes and this happens due to penetrations of the glass into the rigid mold. These penetrations cause huge profile deviations as the accuracy we are looking here is of the order of microns.

I have tried changing the penalty parameter to a higher value, tried to offset penetration but always end up with the small penetrations. Can anyone suggest a way to get aroung this? How do I obtain absolutely no penetration? How do I obtain the exact curve of the mold surface on the glass?

 

Thanks, 

Shriram

 

 

 

 


Jayadeep U. B.'s picture

Use Lagrange Multiplier Method

Dear Shriram,

Penalty parameter method will lead to some amount of penetration always, as the contact forces are calculated as a product of the penalty parameter and penetration.  Increasing the penalty parameter should reduce penetration, but will lead to numerical difficulties also.  Hence, if you want extremely high accuracy, you can try using Lagrange Multiplier method for enforcing contact constraints, which exactly enforces the contact constraints, though the d.o.f as well as computational complexity are incresed.

Please refer the excellent book "Finite Element Procedures for Contact-Impact Problems" by Zhi-Hua Zhong.  My MTech thesis on "Finite Element Analysis of Dynamic Elasto-plastic Contact Problem with Friction" (available at: http://www.new.dli.ernet.in/cgi-bin/metainfo.cgi?&title1=Finite%20Elemen...) could also be of some help!

Regards,

Jayadeep U.B. 


Dear

Dear Jaydeep, 

 

 Thanks for the suggestions. I will look into it. However, I was unable to open your thesis link page. Is it possible for you to e-mail a pdf version of the same? (I hope it is not a problem)

 

My email address is: shriisall@gmail.com

 

Thanks once again, 

Shriram

 

 

 

 

 


Dear

Dear Jaydeep,

 

 

Can you also give me an idea of how complex is the implementation of the lagrange multiplier method as compared to the penalty method? Is it possible to modify the penalty method or the whole development itself is totally different?

 

Thanks, 

Shriram


Jayadeep U. B.'s picture

Contact implementation will have to re-written completely

Dear Shriram,

I feel that the code for contact implementation will have to rewritten completely, since the nodal contact forces also become primary unknowns in the Lagrange Multiplier method.

Regards,

Jayadeep


Hi Jayadeep,

Hi Jayadeep,

Can you give me your personal email id-?I am interested in discussing some fundamantal of Virtual work principle (which we had spoken a few days ago in another post)?

kajal