| Author | Message | ||
     Travis Anderson (tja) New member Username: tja Post Number: 1 Registered: 09-2005 |
I have been running a simulation with no problems under version 3. It pretty much consists of a 20 mm square stack of 8 different materials, with thicknesses from 0.1 to 10 mm. There is a small area for heat generation in the very center of the stack, with dimensions 1 um x 1 mm x 0.2 um. I have been attempting to run the code on version 5 for the past few days and it always gives me mesh generation errors, speficially a "too many neighbors" error. Initially I was trying to use the limited region command to avoid having a very small mesh area running vertically through the stack, but that gave me the same errors, so I resorted to the same code I had been running. Any suggestions as to what is causing the errors? Maybe scaling some of the x and z dimensions would help? | ||
| Robert G. Nelson (rgnelson) Moderator Username: rgnelson Post Number: 465 Registered: 06-2003 |
The version 4&5 mesh generators sometimes have this kind of trouble in making a rapid transition from small to large cell sizes. It is a statistical kind of thing, in that a linking decision at one stage creates a configuration that a later stage is unable to link. For this reason, it is very difficult to identify the offending link decision. With a dynamic range of sizes of 1:1e5, you are pushing the resolution limits of the mesh generator, and may be losing points to node merging. To test this, simply make your outer dimensions smaller or inner dimensions larger to see if the problem disappears. You can sometimes cure this simply by perturbing the conditions of the mesh generation, by: 1) changing NGRID (default=10 in professional 3d 2) changing GROW3 (cell growth rate, default=0.25), 3) by wrapping a very small feature in a somewhat larger region or figure to act as a mesh generation buffer, 4) by including thin neighboring layers in the LIMITED REGION specification, 5) by specifying explicit mesh densities in the vicinity of the small feature. If you send me the script, I will see if there is some other cause for the problem. |