Contact, seperation, maximum activation distance etc

[autodesk428JLG]

Hello sirs!

Im running some linear and nonlinear buckling analysis with contact between some profiles and yesterday i ran some and i didnt find convergence. However, i think i misunderstood something.

So, im wondering, if you have plates/parts that are side to side with each other, welded on top and have the possibilitiy so slide against each other on the surfaces below the weld, how do you set up the analysis correct? See picture below:

At first, i modeled the midsurfaces as they are IRL, hence, i had a seperation in my model of the midsurface planes but then the distance between the plates are so far away that the midplane nodes never connect and hence, i didnt get convergence.

So, should the assembly be set up to be that the midsurfaces are really close together, correct? Like 0.2 mm or something.  Then i got convergence. So my first tries, in my picture below, i first modeled the midsurfaces with a distance between them with 17 mm. Then i used the formula in tip #18 for the max activation distance. And that didnt converge.

Then, i set it up with a distance of zero in the assembly and run the analysis, maybe i could have used a real small distance in order to be able to select the surfaces easier. Then things works, see results in the end. So my question is pretty much if this is correct analysed of the situation i have?

Thanks in advance, im using Inventor nastran 2023,

Daniel

[John_Holtz]

Hi @autodesk428JLG 

The first comment is buckling does not handle separation contact. Buckling is a form of modal analysis. A modal analysis assumes the model vibrated the same in both directions, and that would not happen if the parts could separate in one direction. The modal analysis converts the separation contact to bonded contact. You should either change the contact type from separation to bonded (so that anyone looking at the model does not think that the separation contact works as separation), or remove the contact so that the parts are free.

If you were doing a linear analysis, then separation contact is handled. (But friction is not handled in a linear analysis, so set it to blank or 0.) In your situation where the midplane of the plates cause the shells to be separated by some distance (equal to the sum of half the thicknesses), you can use the Penetration Surface Offset to account for the material that exists in reality but not in the model. From the Autodesk Knowledge Network, see Difference between Maximum Activation Distance and Penetration Surface Offset in Nastran contact def....

A nonlinear analysis supports separation contact and friction.

The article about the Penetration Surface Offset also indicates another important setup for shells: the normal direction of the two shells should point toward each other for symmetry contact. If the contact type is unsymmetric, then only the primary shell normal direction needs to point toward the secondary shell.

Let us know if you have any other questions.

John

[autodesk428JLG]

Hi John!

Thanks for your reply, i really appreciate it! : )

So, in my case, using non-linear buckling i have basically used "bonded contact" where i defined separation, and no need to define a penetration distance in that case either? Would it in your professional opinion, be a decent approximation of my case in reality? I mean, thats all i got i guess, if i use the surfaces as free it would be to conservative, right? I feel when it buckle those surfaces will most likely follow eachother like a bonded connection?

Best regards,

Daniel

[John_Holtz]

Hi Daniel,

I do not understand what you are trying to analyze and what you have modeled, so it is hard to give an opinion about the engineering. (Even if I did understand everything, you need an engineer competent in your field to provide good suggestion, and that is not me. 🙂) What I would suggest is running the analysis with bonded, another analysis using sliding contact (maybe fair for the condition of sliding plates), and maybe another using free contact. If the results of bonded and sliding are similar, then one of those results may be the answer. (Not sure if free is a good approximation, but if the results are similar, this indicates it really doesn't matter for the design.) 

John

[autodesk428JLG]

Hi John!

I see, but i was referring to the buckling analyses, you said that when it comes to linear and nonlinear buckling, there is no option to use contacts like separation right? Only bonded. or did i misunderstood 😞

[John_Holtz · Accepted solution]

When I wrote sliding contact, I should have written "sliding/no separation". That contact type is different than separation.

This is the basic list of contact types and the 22 possible analysis types that they are applicable for:

• bonded. All 22 analysis types.
• offset bonded. All 22 analysis types
• separation. linear static, nonlinear static, nonlinear transient response, maybe linear direct transient response, explicit dynamics. Other analysis types either ignore it, or treat it as bonded (such as modal). The reason it is supported by only a few analysis types is because the results are different "in tension versus compression", so the analysis needs to be able to handle it. Modal analysis cannot handle it (and therefore, frequency response, random response, response spectrum) because those analysis types assume equal vibration "in tension and compression".
• sliding/no separation. I am not sure about thermal and explicit dynamics, but all other stress analysis types should handle this. Sliding/no separation behavior is equal "in tension versus compression" because the parts do not separate or pass through each other. When I referred to sliding contact in my previous reply, this is the type of contact that I was thinking of. Note that sliding was created for flat-to-flat geometry. It does not work very well with curved geometry. (Curved geometry tends to act more like bonded and freely "sliding".)
• separation/no sliding. Probably the same as separation.
• shrink fit/sliding. nonlinear static, maybe nonlinear transient response.
• shrink fit/no sliding. nonlinear static, maybe nonlinear transient response.

(One clarification: the Nastran solver supports additional analysis types, such as DDAM. Twenty two (22) analysis types is the number supported by the Inventor interface.)

John