Dhondtguido / CalculiX

This repository contains the source files of CalculiX, a three-dimensional Finite Element Program (www.calculix.de).
GNU General Public License v2.0
72 stars 16 forks source link

Error executing the attached files #81

Open anoopngmys opened 4 months ago

anoopngmys commented 4 months ago

Hello,

I have 2 input files e_5905.inp and e_5906.inp. I am trying to run these files using Calculix linux executable. I have noticed that when I run the e_5905.inp file, it generates .sti and .dof files successfully. However, when I run the e_5906.inp file, it doesn't produce these files. I'm curious to understand why this is happening.

Attached are the input files in .txt format, as .inp format couldn't be attached. Thank you for your understanding

e_5905.txt e_5906.txt

victorkemp commented 4 months ago

Works for me on windows with CCX 2.21. When you have the failure, does it exit properly with "Job finished" message and produce the .frd and other files? If not, it might have crashed for some reason.

e_5906.dof:

855.1
855.2
855.3
6213.1
6213.2
6213.3
7485.1
7485.2
7485.3
7487.1
7487.2
7487.3
anoopngmys commented 4 months ago

e_5906.inp exit without "Job finished" message with below files only

image

Is it possible to run 2 input files using linux executable and check?

JuanP74 commented 4 months ago

what is surprising in fact is that e_5905.inp works, since you do not have any boundary conditions in a static step, the other one behaves as expected: matrix is found to be singular as cast by spooles, for example. If you set up both problems properly both will work. Nevertheless the behaviour under wrong input is surprising.

anoopngmys commented 3 months ago

Hey @JuanP74, but the input file provided works fine with windows executable and not with linux executable. even the same result should be expected from windows executable right?

victorkemp commented 3 months ago

Where there can be rigid body motion, which these files have, it sometimes solves, sometimes doesn't. So it could depend on the node coordinates, solver, OS, or the phase of the moon. Pardiso seems to be more robust to that sort of thing than Spooles, but you never know.

Although you'd expect CPS elements to not need E3 and they often work fine with E3=0, they turn out to require non-zero E3 to work reliably.

anoopngmys commented 2 months ago

@victorkemp, considering that the CalculiX Linux executable on the website includes the Spooles solver, is it possible to obtain an executable linked with other solvers like Pardiso ?