Convergence problem

[ananthaces]

Hello, I am running a simple curved sandwhich panel with large displacements on. the boundary conditions are : it is pinned on one end and allowed to translate on the other.I have applid pressure to the inner surface of the beam. It is a non-linear analysis and i am not able to get it converged. Could anyone give me some idea please. Thanks for yout time.

 

[ThomasH]

There are a lot of remedies for convergence problems depending on what exactly is nonlinear in the analyses.

First try a linear analysis, second increase the number of loadsteps and print everything. Then you can follow the solution and perhaps find the problem.

However, a thought, curved panel pinned at one end and free at the other. Isn't that a mechanism, the first end can rotate and the second does not stop global rotation.

Run a frequency analysis to find any free-body motion.

Good Luck

Thomas

 

[ananthaces]

hi..thnx for ur reply. actually there is no free body motionm as one end is allowed to translate in the x-direction alone and constrianed in y and Z while the other end is pinned. its non-linear due to large deflection. Is there any other setting which could resolve this issue?

 

[GBor]

Can you tell us a little more about your convergence parameters? What timestep are you using and what is the anticipated total deflection of you plate? What material type are you using? Thomas is correct in saying there are many things that can cause convergence problems.

Garland E. Borowski, PE

 

[ThomasH]

Like I said before, try using a smaller load increment to trace the problem. Do you get any results at all?

Also, does a nonlinear analysis without "large deflection work"? In other words, is that the problem?

Also, which Nastran do you use (MSC NX, NEi etc)?

Regards

Thomas

 

[ananthaces]

Hello, I am using MSC Nastran. When in run the analysis without large displacement on it runs fine. I am using PCOMP as it is a honeycomb curved panel. I used 20 increments, and 40 step.played around it a bit. I also set the k6rot value to 1e5.it runs for few steps and then dsnt converge. So,how do i get it to converge now? Thanks for your help.

 

[ThomasH]

First I would increase the number of increments to say 200.
Then I would set the k6rot value to default.

Also, do you get any results, you say a few steps. The error messages in the .f06 file can indicate what goes wrong.

Unfortunately all I can do is some "blind" guessing.

Good Luck

Thomas

 

[ananthaces]

hi there thomas..i do get some results..like it solves till 0.6 and then it dsnt converge at all. i wonder why it happens. and also one more question. how to handle crumbling of the elements when runnin a modal analysis? any idea? Thanks.

 

[ThomasH]

If it solves up to a limit, in your case 0.6, this can be due to a number of reasons. If for example you work with plasticity in can be "a simple" structural faliure. The should be some info in the f06 - file.

As for the "crumbling" of elements I don't know what you mean so no ideas right now, sorry.

Regards

Thomas

 

[ananthaces]

ok, here is my problem.the curved beam with large displacements on converges if i apply on an internal pressure but whn i apply an additional compressive load on the ends of the beam it doesnt converge.any idea as to why? thanks.

 

[ThomasH]

A compressive load means a reduced stiffness in some respects. You can compare this to, say, a prestressed wire.

So your system without compressive load converges but not with the compressive load.

A few ideas:

First try with only the compressive load. You might have some kind of buckling instability present.

Another method is to first create a loadcase with internal pressure only and then a second loadcase where you add the compression.

Since you don't mention it I guess this is not a case of plastic faliure.

You can also "play" with the convergence criteria. If you set them "wrong" you might be able to get a incorrect solution that can indicate what the problem is.

Good Luck

Thomas