FE^2 solution

[zhiran]

Dear FEAP users:

I am new in feap.  I am learning to compute a 10x10x20 block using FE^2, but I am running into some problems: 
1. When I'm using FE^2, when my solution command contains the 'tangent,,1' or 'time', the program will get stuck in that command, and subsequent commands will not be executed.  At the same time, the output is displayed only to the 'Command tangent' or 'Command time'  By the way, I'm using version 8.5.2.  I used the command 'mpirun -n 2 ./feap' to execute feap.  The input files are 'Imacro' and 'Imicro'.
2. Another problem is that the nodal forces I applied on some points didn't seem to work. 
The attached file contains the input file I used and the generated output file.  I'd appreciate it if you could give me some advice.

Sincerely,
Zhiran

[Prof. S. Govindjee]

You have a couple of issues:

1) you have set the RVE file to be Imacro; it should be Imicro
2) your Imicro file's first line should just be the node,elements,mats,etc. line (not feap line)
3) in your solve_mpi file the command mpi should be replaced by fe2 every where that it appears
[4) I also removed the 'its' since I forgot what it is used for and was too lazy to look it up in the manual.]

With these changes the file ran.

As a general remark, I suggest first getting things to work on a much much smaller problem!

[zhiran]

Thank you for your prompt reply.  I should not have set the RVE file incorrectly.  My goal is to use a 1x1x1 RVE to calculate some features of a 10x10x20 block.  This RVE contains some inclusions and splits many meshes.  I have modified the files according to your 2-4 suggestions, but the result doesn't seem to have changed. 

[Prof. S. Govindjee]

You must have missed something.  I'll post the files in just a bit.

[Prof. S. Govindjee]

Here are the files I ran.  I deleted the extra junk in your files.  Run a diff to see what I changed.

Also note, I removed the ELINK at the end of Imicro since you fixed all the boundary nodes.

[Prof. S. Govindjee]

Additional problems with your RVE file (Imicro):

1. some of your elements have incorrect node ordering, leading to negative jacobians (or FEAP's check command is failing)

2. the layout of the node numbering is not ideal and this will lead to very poor structure of the tangent matrix.  it will be useful to renumber your nodes to minimize the bandwidth of the tangent matrix for the RVE mesh (Imicro).  another option is to use a solver that will do this automatically for you, say, pardisio -- which can be compiled into FEAP.

[zhiran]

Thank you for your help.  I followed your advice and modified the input file.  I could see that the force was applied in the generated plot, but I did not get the contour plot.  Later, I tried to use a simple RVE model to calculate this 10x10x20 block, but the result of my calculation was still stuck in the ‘time’ command in the loop, and no contour map of displacement or stress was obtained.  If I comment out the loop, I also don't get 'plot cont', and I get 'all zero values'.

[Prof. S. Govindjee]

Try this file as your RVE file

[zhiran]

I tried this RVE file, but it still doesn't work.

[Prof. S. Govindjee]

There must be something wrong with your installation.  I just ran the files with version 8.5 and it ran fine.

Can you run 'ls' on the fe2 folder?  also can you post the output of 'ls program/mpi*'?

[Prof. S. Govindjee]

Just to be sure, make sure your Imacro files is now calling Imicro2 as the rve file.