Update to a long thin walled cylinder, fixed both ends L=1, r=0.5, t=0.012, E=200E9,v=.3, external pressure 6.76E6
Values across abaqus, Elmer, nastran are very close
Folder buck4 in
https://github.com/mrkearden/Thin-Wall- ... r-Buckling
Thin Wall Cylinder Buckling
-
- Site Admin
- Posts: 4841
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: Thin Wall Cylinder Buckling
Hi Kevin,
Great that mesh convergence is achieved. That shows that codes are consistent!
To avoid confusion between higher order nodal or p-elements:
- Only either can be active at the same time
- Higher order nodal elements are created by the meshes and can have better geometry description
- p-elements always start from linear basis and use hierarchy of basis functions
- the "increase element order" was included inside ElmerSolver to be able to run larger cases where also "Mesh Levels" is used
- Both the internal increased element order and p-elements can follow some selected curved shapes if so requested.
In you setup you had both "increase element order" and "p:4" I don't really know how they come along. I would guess that additional nodes just hang along as zeros. You might get the time down by choosing only either. Still, the direct methods tend to be costly for higher order elements.
-Peter
Great that mesh convergence is achieved. That shows that codes are consistent!
To avoid confusion between higher order nodal or p-elements:
- Only either can be active at the same time
- Higher order nodal elements are created by the meshes and can have better geometry description
- p-elements always start from linear basis and use hierarchy of basis functions
- the "increase element order" was included inside ElmerSolver to be able to run larger cases where also "Mesh Levels" is used
- Both the internal increased element order and p-elements can follow some selected curved shapes if so requested.
In you setup you had both "increase element order" and "p:4" I don't really know how they come along. I would guess that additional nodes just hang along as zeros. You might get the time down by choosing only either. Still, the direct methods tend to be costly for higher order elements.
-Peter