Negative Jacobian Error After Using Mesh Extruder

377 views
Skip to first unread message

William Harris

unread,
Dec 21, 2015, 7:22:25 PM12/21/15
to moose-users
I created a 2D mesh in Gmsh and read it as the input for my MOOSE simulation using the native .msh file format.


My simulation ran fine over this mesh. I then used the Mesh Extruder mesh modifier to create a 3D mesh by extruding the above surface.


When I repeated the simulation, I got an error about a negative Jacobian:

ERROR: negative Jacobian -0.00140208 at point (x,y,z)=(-0.83774, 0.429148, 4.88453) in element 14853

I am doing a steady-state diffusion simulation over the volume, with constant temperature boundary conditions on all surfaces. I have attached my input file.


I then tried to extrude the surface in Gmsh before reading it into MOOSE. However, although the mesh is 3D (which I can view in Gmsh, as shown below)...


...MOOSE only reads it as a 2D surface (like the picture at the top). I can't seem to get the simulation to run in three dimensions, despite trying two different ways. Are there any suggestions as to how to debug thus?


Thank you,

Wil





ThermalHydraulics.i

Yaqi Wang

unread,
Dec 21, 2015, 7:49:13 PM12/21/15
to moose...@googlegroups.com
Can you do --mesh-only on you 2d mesh and generate an exodus mesh file. And then let your 3D input with the extruder using that exodus file. Negative jacobian means the local node numbering on element is not following the convention numbering. Libmesh could do some renumbering but not for the extruded mesh.

Sent from my iPhone
--
You received this message because you are subscribed to the Google Groups "moose-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to moose-users...@googlegroups.com.
Visit this group at https://groups.google.com/group/moose-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/22a7cfa4-f790-4170-a929-fd90953615cc%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
<ThermalHydraulics.i>

William Harris

unread,
Dec 21, 2015, 8:00:36 PM12/21/15
to moose-users
Yaqi,

I am not familiar with --mesh-only. Where do I put it into my input file?

Thank you,
Wil

Yaqi Wang

unread,
Dec 21, 2015, 8:05:38 PM12/21/15
to moose...@googlegroups.com
It is a command line parameter. If you run yourapp-opt without argument you will see it.

Sent from my iPhone

William Harris

unread,
Dec 21, 2015, 8:24:38 PM12/21/15
to moose...@googlegroups.com
Yaqi,

That did not seem to get rid of the error. I wrote the 2D mesh to an exodus file, read that into another simulation as the initial 2D mesh, and extruded it. However, the negative Jacobian error was still there when I performed the 3D simulation.

Wil

--
You received this message because you are subscribed to a topic in the Google Groups "moose-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/moose-users/REtAQ0b7xXk/unsubscribe.
To unsubscribe from this group and all its topics, send an email to moose-users...@googlegroups.com.

Visit this group at https://groups.google.com/group/moose-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/FA29AC78-A238-4974-B224-6309844186C0%40gmail.com.

Yaqi Wang

unread,
Dec 21, 2015, 8:37:59 PM12/21/15
to moose...@googlegroups.com
 Oh wait, there is a bug not fixed yet see moose issue #3554. The fix has been done in libmesh but has not been pushed into moose libmesh submodule. But I doubt that is causing the error here. Can you use the exodus output of your 2d calculation really quick? Just want make sure. You should go to you libmesh fold and use git to update it manually. 

Sent from my iPhone

William Harris

unread,
Dec 21, 2015, 8:48:52 PM12/21/15
to moose...@googlegroups.com
Yaqi,

Yes, the simulation works fine with the 2D exodus output mesh.

Wil

To view this discussion on the web visit https://groups.google.com/d/msgid/moose-users/F0D22738-B391-4664-8A20-7CE4A7D5CF6D%40gmail.com.

Yaqi Wang

unread,
Dec 21, 2015, 9:06:01 PM12/21/15
to moose...@googlegroups.com
Great! Keep in mind that that bug could affect you calculations. You do want to update and compile your libmesh manually or wait for next moose libmesh update.

Sent from my iPhone
Reply all
Reply to author
Forward
0 new messages