10/11/12

I was very excited to get the new data and begin analysing it. First, I wanted to check out the mesh of the octant simulation. This is what I found:

From this 3d plot, we see that the mesh is squarish in the y-z plane, but roundish on the other 2 faces of the cube. Depending on which face I chose to make plots of, the mesh will be slightly different (and the way the fluid flows will be slightly different as well):

A pseudo-color plot of the density can be found here:

As you can see from this movie, the ambient is very poorly resolved. I wonder about the consequences of this. The fluid variables for each cell are calculated by considering gradients along cell-cell interfaces. If the cell is course, as in the ambient, then one value of density is assigned to that course cell, whereas if that one cell was split into many little cells, there may be more structure than is being seen at a lower resolution. Now, the structure may be missing here, but is there a sufficient loss of information? The collapse seems to be (largely) spherical, so the loss of information would be in the radial direction. Is this right? Is it like the density is being "averaged" so-to-speak over a larger volume = volume of coarse cell..?

I tried to get more refinement in the ambient, so I removed the de-refinement object from my problem module (last post). I expected refinement based on a) jeans in the ambient, and b) density gradient in the sphere. The grids were initialized as expected. However, later in the simulation, BH killed the simulation, reporting a seg. fault. A look at the last frame revealed massive refinement throughout the grid:

The difference in the refinement calls between this and yesterday's post, is in this most current set-up, I had astrobear refine on Jeans length.

10/10's directory is /alfalfadata/erica/DEBUG/Matched_From_Bluehive 10/11's is /grassdata/erica/from_bluehive/MatchedMeshMadness

I'd like to figure out what is the refinement being triggered from?

Attachments (5)

Comments

No comments.