flash3780
Mechanical
- Dec 11, 2009
- 829
I'm having a bit-o-trouble with one of my 3D models and I was wondering if anyone on here might have some insight. I have a pretty large thermal model with quite a few types of boundary conditions. One of the boundaries is a volumetric heat generation (in some heater cartridges). If I apply 25% or so of the full load, the model converges fine. I've found that if I apply 50% or so of the required load, the model converges with some difficulties, small equation solver pivot terms, bisections, etc. However, the model errors out if I apply the full load (becomes unstable, negative absolute temperatures as DOFs fluctuate wildly, etc.).
I've got thermal contacts (augmented lagrange) throughout the model handling gap convection and gap radiation, and convection in cavities within the model are handled with SURF152 elements tied to a space node. If anything were to cause these sorts of convergence problems, I'd expect it to be the contacts. Though, I've used the exact same contact setup on a very similar model (with a few extra pieces in it).
The results look pretty as expected for the cases that converged (no crazy temperatures or anything). I'm wondering if anyone has run into this sort of thing before and might have some tips. Not sure what's making my model blow up with what seems like a relatively benign change. I'm thinking that perhaps I should be using a different solver. I'm currently using the sparse-matrix direct solver with nropt,full.
I suppose that I could try switching to an iterative solver or changing to nropt,modi... but beyond that, I'm stuck. Any suggestions? I've attached the heat convergence, the max dof increment, and the final temperature profile (or at least the part that can be posted on the internet) below.
Heat Convergence
MAX DOF Increment
Temperatures
Thanks for the help.
I've got thermal contacts (augmented lagrange) throughout the model handling gap convection and gap radiation, and convection in cavities within the model are handled with SURF152 elements tied to a space node. If anything were to cause these sorts of convergence problems, I'd expect it to be the contacts. Though, I've used the exact same contact setup on a very similar model (with a few extra pieces in it).
The results look pretty as expected for the cases that converged (no crazy temperatures or anything). I'm wondering if anyone has run into this sort of thing before and might have some tips. Not sure what's making my model blow up with what seems like a relatively benign change. I'm thinking that perhaps I should be using a different solver. I'm currently using the sparse-matrix direct solver with nropt,full.
I suppose that I could try switching to an iterative solver or changing to nropt,modi... but beyond that, I'm stuck. Any suggestions? I've attached the heat convergence, the max dof increment, and the final temperature profile (or at least the part that can be posted on the internet) below.
Heat Convergence
MAX DOF Increment
Temperatures
Thanks for the help.