I had a model fully constrained, but the solver returns a singular matrix error message.
The constraint was checked by right click the connection -> find underconstrained body.
Can anyone please help or advise?
Thanks in advance.
UPDATE - I think I see it now.
I had thought the study as it came was meshing the walls as sheet metal. I see now that it was solid, but meshing to a knife edge on the ends. So each section met line-on-line, and not necessarily with matching nodes. This will certainly put singularities in the solution.
I re-meshed with the curvature based mesher, with extreme max/min sizes (100/2 mm) until it meshed each cylinder and gave clean flat edges.
It might be easier if you attach your file or at least show a picture or two. There are some smart people on this forum but I don't think there are any mind readers. (I may be wrong about that last.)
Model attached to the original thread.
We need the part files, also.
I have zipped everything together and uploaded to the original thread.
The model doesn't contain any studies but looking at the assembly the problem part seems to be Instance 6 (1500 mm configuration). The "Manhole Boss" solid feature is having a hard time meshing with the sheet metal cylinder. Suppressing the "Cut-Extrude2" feature seems to give a smooth mesh so you may want to consider modeling that nozzle/boss as a surface instead of hollowing it out.
Hi Attiilio, thanks for your reply.
The Cut-Extrude2 feature needs to be there.
I have also tried to mesh the wall and the nozzle boss feature with surface element and leave everything else as solid. But I can't mesh the fillet even with a mesh control of 0.1mm element. Do you have any idea why?
The tangency of the fillet seems to not be merging with the face. That's not a solution I know but I can say from experience you are always better off extruding the nozzle into the surface. For instance, put the "Manhole Mounting Plane" at the end of the nozzle (400 mm?) and extrude to the cylindrical surface. After that it's trial and error mostly.
If the fillet definitely needs to be modeled you may want to consider making a detailed submodel of just that nozzle and juncture to get the stress profile in the fillet.
I have cracked the issue and all good at the moment.
What was the rigid connector for? There is already a global bonded connection in force.
I could see some mesh problems, and suspected parts weren't joining up as intended. I can't see just what's wrong with the existing study. The mix of solid and sheet features on the cut ring is one complication. Also the angle iron ring cannot be treated as a beam, since it is not being used as a beam (I think it was already solid).
So, I made a new study and got the following result in a couple minutes. To get this:
- Treat all as solid
- Curvature based mesh
- Activate "non-touching faces" for global bonded contact, as the sheet metal rings did not quite close (I tried a 2 mm gap).
This is a rare good application for shell elements, but getting that right adds complication, uncertainty, and human labor. My aging computer chews through the all-solid version in 34 seconds.
Hi Shawn, thank you for your reply.
First, the rigid connection was a old scenario. I didn't delete it as it doesn't make any difference to the simulation.
Second, I have tried your method in your second post, and it worked. Thank you.
1. I normally use standard mesher with the assistance of mesh control to mesh the model. I have investigated the mesh using the curvature-based mesher. I've found the distribution of the element on and around the fillet area is not symmetric or even close. I know the standard mesher can't produce a perfect symmetric mesh on a symmetric geometry, but the distribution is very close hence the result. So is the curvature-based mesher able to produce an accurate enough result on a non-symmetric geometry?
2. The ring is bonded with the bottom strake. What is the reason to turn on the "non-touching faces"?
I find the curvature based mesher to be fast and flexible. Symmetry of mesh is not critical if a solver is well conditioned. Mesh refinement at areas of high gradient is more important, and leaving in fillets is an easy way to get the refinement from the curvature driven mesh.
I turned on non-touching faces because in my first attempt the thin bands were not closed; the ends of the top piece moved opposite directions.
That is a pretty clear explanation.
I followed your suggested way and it worked.
Retrieving data ...