Volume seems to mesh fine, despite error message

I have recently modified a simulation on which I have been working for a few weeks and am getting good quality surface meshes as judged by the "A" and "E Ratio” scalar results, with very few narrow triangular elements. Most have an E Ratio of > 0.5. When I mesh the volume, it seems to mesh fine, per the "Vol" and "Vol Ratio" results. The entire volume is filled with a mesh. I have no triangulator or tetulator problems, and I have all the diagnostics turned on to show any problems in the log. The log reports none.

Feature-wise, the only difference between this simulation geometry and the previous simulation geometry (on which the solver ran fine) is the addition of an embedded face, which I turned into a two sided face that is a fixed-velocity inlet face. Both faces of the two sided face have the same BCs, the same air velocity, and the same flow direction per the X, Y, Z coordinate entries in the respective property panes. The geometry itself was made from a planar surface imported from Rhino as a STEP file that I revolved in Caedium to generate a solid. Then I imported the face to be embedded, Boolean subtracted it from the solid, and the geometry was all done. Most of the Rhino geometry elements were simply copied from those that worked well with the previous simulation.

However, after I attach vector arrows to the main seed face (there are four) and try to run the simulation, I immediately get the "FOAM FATAL ERROR: illegal cell label -1 in neighbor addressing for face...." message. This always happens at the very start, without the solver even appearing to start running. I also tried it with vector arrows attached to all four seed faces at the same time, and got the same error message.

I have read and re-read the Symscape page "How do I fix mesh and solver failures?" several times.
I have shaded all model faces. I have scrutinized the meshes and they seem okay, but I am not entirely sure I know how to evaluate the quality of volume meshes.
I initially set all faces to be walls, then went back and changed certain ones as needed to inlets, inlets/outlets, and outlets, so everything has some sort of BC set. After I could not get the solver to run, I went back and did this again.

I have the residual monitor up and ready to go. Sometimes - not always - it shows a residual trace going for a fraction of an iteration horizontally and then stopping.

Thanks very much in advance for your thoughts.

No votes yet

The mesh is corrupt, check your volume elements

That error message means that the mesh is corrupt, most likely the volume mesh failed at some point. If you are sure that the surface meshes had not errors - often the cause of a volume mesh failure - then there is likely to be a region within the volume (maybe between faces in close proximity) that is causing a problem.

I'll assume that your double-sided face was subtracted from the volume with the Topology Only option set to True and that the face doesn't touch or intersect any other faces - if not this process will not work.

Changes to the Physics or trying to visualize results will not correct a meshing failure. You need to focus on the location of poor/corrupt volume elements as described in the section "Volume Mesh Quality" in "How do I fix mesh and solver failures?". Specifically:

To view the volume mesh ... use the Scalar Fields->Vol Ratio in the Results Tool Palette, and set the volume view properties: Style = Flat and Transparent = Off. Also make sure your faces are transparent, otherwise you won't see the interior volume.

An ideal element will have a Vol Ratio = 1, and so focus on the low valued elements with the threshold sliders to see if there is a particular region causing problems. If you identify problems try using the Accuracy tool to locally modify the mesh size on nearby edges/faces.

Updated FAQ with screenshots

I just updated the FAQ on diagnosing mesh and solver failures with screenshots, so take another look to see if it helps clarify what to look for and correct.

Problem solved.

Thanks for your suggestion about the 2-sided face. That was the problem. As soon as I fixed that, the simulation ran and converged the first time. I used the relaxation factors from my previous simulation and woke up this morning to a nicely converged simulation. Thanks again.

Trending Now