Topic: Documentation on ctrl file

Dear Borek,

How are you? It is nice to get back to you. Do you have some simple documentation on the use of ctrl file? It is difficult a bit for me to understand the control files. It seems you have update on it as well. Thanks,

Best Regards,
Xuejian

2 (edited by dr 02-03-2011 14:16:19)

Re: Documentation on ctrl file

Dear Xuejian

please find enclosed very very simple guide how to format control file (see attachment).

Daniel

Re: Documentation on ctrl file

Dear Daniel,

It is nice to see you here. I have a few short questions regarding the sample ctrl file in the unv2oofem section. I copied here,

group load_edge
elemprop bLoadNum 2
etype[11]

group support_nodes
nodeprop bc 2 1 1

group material_1
elemprop mat 1 crosssect 1
etype[44] PlaneStress2d

group material_2
elemprop mat 2 crosssect 1
etype[44] PlaneStress2d

"bLoadNum" is a keyword to specify the load? what does the etype[11], etype[44] mean? Could you give me an example on applying load on surface groups? Thanks,

Best Regards,
Xuejian

4

Re: Documentation on ctrl file

Dear Xuejian,

but your question is related to unv2oofem I have nothing common with.
Borek should navigate you to competent people.

D.

5

Re: Documentation on ctrl file

Dear Xuejian,
I started to prepare a manual describing the format of control file as well, but this will take some time.
But back to your question:
when "bloadnum" keyword is used in a record describing element properties, the values associated to this keyword determine the boundary loads to be applied. To select element edges or element surfaces on which the boundary load will be applied, the element type  (etype keyword) in this case determines the boundary element subjected to loading. For example etype[11] for linear edges, etype[41] for triangular surface, etype[42] for linear quad surfaces, etc. For example, linear brick element has eight boundary surface elements (linear quads), 12 boundary edge elements (linear edges).

sorry for the late reply,
Borek

Re: Documentation on ctrl file

Dear Borek,

Thanks a lot for your reply. Please let me know when you have the manual ready. Now I have another question. When I was using ConcreteDPM model, I got the message:

Plastic multiplier less than zero. Closest point projection doesnot converge. deltaLamda= -inf.

Could you give me some hints on what to do with it? My definition for the material is like this.

ConcreteDPM 1 d 0.0 e 4031.0 n 0.20 tAlpha 0.0 ft 0.30 fc 5.0 wf 2.0e-4 helem 2.0 yieldtol 1e-5 newtoniter 1000 (units: kips-in)

Thanks a lot.

Best Regards,
Xuejian

Re: Documentation on ctrl file

Dear Xuejian, I worked on the ConcreteDPM model with Peter Grassl.
The message means that the stress return algorithm does not converge,
but to find out why, I would need to trace the algorithm. Please could you
send me your input file that leads to this problem?
Best regards,
Milan Jirasek

8 (edited by Xuejian 25-03-2011 00:06:09)

Re: Documentation on ctrl file

Dear Professor,

Please see the input file I attached. The model is a wall with I-shape cross section. On the top nodes at Z=100, I applied the constant nodal loads in Z direction. In X direction, I apply nodal displacement from zero to 1 inch in two steps. Please review it. Thanks,

Best Regards,
Xuejian

Re: Documentation on ctrl file

Dear all,

This is just a follow-up problem. I try to do the same analysis with the updated model (con2dpm). It seems there are still some convergence problem saying "subincrementation required". I havenot looked into the model in detail, but any input on it is appreciated. Thanks,

Xuejian

Re: Documentation on ctrl file

A quick look into concretedpm2 at "subincrementation required" means that it is not in fact an error. It just notifies that it has to try a smaller strain increment internally to reach convergence. Of course, if you get a lot of subincrementations (all which halves the current strain increment) then you are unlikely to ever reach the full strain.
The only thing I could think of would be to use smaller time steps (and possible the linearized extrapolation, if you have non-zero Dirichlet boundary condition).

Please attach the inputfile (as small as possible).

Re: Documentation on ctrl file

Hi Mikael,

It seems the analysis froze after that. My input file may be too big for debug purposes. But anyway, I uploaded it. Thanks for the help,

Xuejian

Post's attachments

singlewall2_np1.in 495.95 kb, 3 downloads since 2013-02-08 

You don't have the permssions to download the attachments of this post.

Re: Documentation on ctrl file

Yes, this is indeed uncomfortably big for debugging.

You specifically use: "newtoniter 100000000" in your model. 100 million iterations? It'll take a while. Anything much over 10 seems overkill.
If it doesn't converge after that, you need to rethink the approach.
In this case, it gets stuck subdividing the strain increment infinitely. There should be a stop condition for this, where it just gives up (after say, 10 divisions, after which the increment is essentially zero already).
I will commit such changes. The reason why it doesn't converge is a whole other question though. ConcreteDPM2 doesn't seem to be the simplest of material models, so I can't really tell what you should do.