CalculiX generate the substructure stiffness matrix 1

[taotaomy]

Hello everyone
recently I'm trying to use Calculix to generate the substructure stiffness matrix. But I don't know how to control the output directory of the mtx file.

Then I read the source code file ccx_2.10.c and notice there is a parameter about the output file. just post the code as follow
for(i=1;i<argc;i++){
if(strcmp1(argv,"-o")==0) {
strcpy(output,argv[i+1]);break;}
}
but when I try to use the -o parameter with ccx.exe, it doesn't show any different with or without using output parameter.

Another problem I encountered is about element stiffness matrix output problem, since there isn't an example for users.

Thanks for any help!

 

[whitwas]

Can you not put a path in the FILE NAME parameter of *SUBSTRUCTURE MATRIX OUTPUT?

You'll probably have more luck asking on the CalculiX forum where the programmer sometimes answers questions:

https://groups.yahoo.com/neo/groups/CALCULIX/conversations/messages

 

[taotaomy]

Hello, whitwas!
I already tried, it works. And I also write a bat file to cd to the inp directory, and the output file would be created in the same path.

Another problem is, why the stiffness matrix generated by CalculiX is different from the one generated by Abaqus with the same inp file?
Can anyone help me to figure out which one is correct.

Thanks!

 

[whitwas]

There's no reason to expect different programs to produce the same stiffness matrix. They might order the DOFs differently, use different element formulations or even scale some values differently. In this case, since you have 2-node beams, CCX expands them to 8 node hex elements with no rotational DOFs. It might also insert constraint equations to join them if they're not in a straight line. Abaqus probably does it in the more traditional way with rotational DOFs that would look completely different.

To test if it's correct, it's probably easiest to solve it and see if you get the correct displacements - not forgetting to check for mesh convergence.

 

[taotaomy]

I didn't expect totally exact the same results, but the two results are totally different things.
I recover these two output mtx files. The data structure are different.
Calculix result

Abaqus result

I choose to believe abaqus result, and there may be some software bugs in ccx, or I used ccx with a wrong way.
The stiffness matrix is attached as below.

 

[taotaomy]

I also tested the S4R element to export stiffness matrix, but seem like calculix gives the wrong answer.

 

[whitwas]

Yea that is suspicious. In CCX's matrix, the displacement DOFs in each row don't sum to zero which seems to indicate there's some constraint to ground when there shouldn't be. For Abaqus's matrix they do sum to zero.

There have been bugs with beams and shells in CCX related to the way it converts them to solids internally. If anything was going to work right, it would be solid elements.

 

[taotaomy]

I read the ccx USER’S MANUAL, and the beam or shell elements are expanded into solid elements. For example, "S4 and S4R four-node shell elements are expanded into three dimensional C3D8I and C3D8R elements, respectively. "
What I need, is the original shell element and exporting the stiffness matrix for the another application to use.
So if there isn't a recovering progress, the directly exported stiffness matrix would not suitable for my using case.

Does anyone has a solution for my trouble, or substitute scheme?

Thanks!

 

[whitwas]

Do you want the complete stiffness matrix or just a few nodes to use it as a superelement? For the complete matrix, my software Mecway (see Google) can do that using traditional 5 DOF shell elements. It can also import a mesh in .inp format. The matrix file format is a row,column,value triple on each line. It's not sorted by node number so it also outputs a mapping between the matrix row numbers and node numbers.

 

[taotaomy]

Thank you