Memory errors during mesh generation

I frequently receive an error similar to the following when I attempt to generate a mesh:

"Insufficient memory available. A Block of xxx,xxx KB could not be allocated"

Exactly what does this mean? How do I prevent this from happening?

Is the solution more hardware or is this an issue with my supermesh design, the number of elements, meshing algorithm, refinement, number of add-ins, etc. that can be solved by trying a different supermesh and other options?

Thanks.

Hi Dwaine,
it would be good to know what mesh generator you use. I would think that typically this is an indication of an issue with supermesh design, such as polygon and line nodes that are VERY close, but not identical. In this case the mesh generator tries to mesh the small space in between, and might end up with extraordinalily high numbers of elements - and thus lacking memory. Also small angles between supermesh element borders can cause such difficulties.

Cheers,
Peter

Sorry, I meant to include the technical details.

I am using FEFLOW 5.3 on a Windows XP machine with quad processors running at 2.66 GHz, 3.25 GB RAM.

I started using T-mesh. The model domain is regional in scope but I need a relatively fine mesh in the plume so I created a super-mesh with three polygons and four or five point add-ins. The intent was to specify 100,000 elements, then generate the mesh areally by specifying five times the suggested number of elements in the inner polygon, twice the suggested number in the intermediate polygon, and no additional elements in the outer polygon. I am sure that I do not have holes or overlaps in my polygons. The point add-ins are "close" to each other at the scale of the model as a whole, but not so close in terms of the real distance. I tried setting "medium", "low" and "no" refinement around the point add-ins. I don't recall if I tried adjusting the speed and quality control.

When that didn't work, I tried changing the element number distribution from 100,000 to 50,000 to 25,000, etc., and the multipliers for each polygon (2, 1.5, 1, for example). I also tried different supermesh designs and slightly modifying the geometry of the final supermesh and the polygons therein.

Finally, I moved to advancing front and finally got the mesh to generate. It's not pretty and no where near the size of the elements I want in the core area (even with globally refining the mesh, twice), and not refined at the extraction wells, but at least the mesh algorithm finished.

Let me also point out that I would get this error message more than once during the meshing using T-mesh. That is, certain super-elements would fail while others would work. This often resulted in a mesh being generated for certain super-elements and not others.

Dwaine,
If you send me the supermesh you have used, I'd try to find out what went wrong (support@dhi-wasy.de).

In general, T-Mesh uses a relatively slow algorithm, that also tends to create obtuse-angles triangles in some cases. At the time of its introduction in FEFLOW, however, it was probably one of the most modern and most flexible triangular meshing agorithms available. As in later versions FEFLOW includes Gridbuilder and the interface to Triangle, which are both way more sophisticated, we have decided to no longer support T-Mesh from version 6.0 on.

Peter

Sorry for my slow response.

The errors I experienced involved a test mesh for illustrative purposes only. Although I have seen these errors practically every time I try to generate a mesh and I want to understand why they occur to avoid these problems in the future, it is probably not useful to investigate this particular super mesh further. I will be preparing my real mesh in the coming weeks and if I see the same issues, I will let you know.

Thanks you Peter