search

xcfem / xc

finite element analysis package for civil engineering structures.
http://www.xcengineering.xyz/html_files/software.html
GNU General Public License v3.0
262 stars 54 forks source link

Ill-conditioned problem? #58

Closed anaiortega closed 5 years ago

anaiortega commented 5 years ago

Hi, I'm trying to make a test on long-term prestress losses over a single section (see test_long_term_restress_loss.py in work-in-progress repository), but in the first stage (prestress transfer) it returns already a 2% error. It seems that the problem is ill-conditioned (the displacement in the element axis and rotation in its strong axis are free in one of its nodes). The message that pops up is:

XC::Node::checkReactionForce; the node: 2 has not constraints on DOF: 0 and, however, it has a reaction with value: 4385.68 on this degree of freedom, it seems that the solution method and/or tolerances are not well suited to the problem. (tol= 19.2473 XC::Node::checkReactionForce; the node: 2 has not constraints on DOF: 4 and, however, it has a reaction with value: 114.361 on this degree of freedom, it seems that the solution method and/or tolerances are not well suited to the problem. (tol= 19.2473

Does it ring any bells?

lcpt commented 5 years ago

Hi.

The prestressing steel material is not linear, so you need to use a nonlinear solver to reach equilibrium.

Regards. Luis.

anaiortega commented 5 years ago

Yeah, Newton-Raphson converges. The error is anyway 2%, I suppose it's acceptable. Thank you

lcpt commented 5 years ago

If you want to get an smaller error, you can reduce the tolerance.