Difference between revisions of "Contrib:KeesWouters/staticshell"

From CAELinuxWiki
Jump to: navigation, search
(''Comparison to solid elements'')
(''Comparison to solid elements'')
Line 64: Line 64:
 
[[Image:Kw_convergence.png]]
 
[[Image:Kw_convergence.png]]
  
[[Media:Kw_convergence.png]]
+
So it is fair to say that linear elements will much too stiff for normal cases. Therefor the displacements of the structure will be predicted much too low.
 +
 
 +
(I will add some more Salome and comm files later on. Going for a break now)
 +
 
 +
--[[User:Keeswouters|kwarky]] 14:44, 7 May 2009 (CEST)

Revision as of 13:44, 7 May 2009

Static behaviour of rectangular plate

The dynamics of a rectangular plate can be predicted quite well with shell (coque_3d) elements. I also liked to know the accuracy of the displacements for this kind of structures. A comparison between solid elements and shell elements has been made.

Dimensions of the plate The length and width of the plate are 1*1 m. The thickness of the shell elements has been increased to 0.02 m to enable meshing with solid elements (tetrahedrons and hexahedrons).

Boundary conditions and load Two types of boundary conditions have been given:

  1. clamped at one side (all displacements and rotations are suppressed)
  2. clamped at the four sides (all displacements and rotations are suppressed)

The load of the structure is a pressure of 0.1 bar on the surface of the plate

Material properties The The Young's module and poisson ratio for steel are set to 210 GPa and 0.28.

Displacement at the tip using beam theory

According to the shell theory the displacement of a beam the tip under uniform pressure is:

  • w = q*L^4/(8*D)
  • uniform pressure q = 10 kPa
  • length of plate L = 1 m
  • Youngs' module E = 210 GPa
  • poisson ratio nu = 0.28
  • bending stiffness D = Eh^3/(12*(1-nu*nu)) = 151910 Nm and
  • displacement at the tip w = 8.2286 mm

Salome model of the rectangular plate

Shell elements

The geometry and meshing follows the same procedure as for the dynamics: geometry of the plate and generating the mesh.

Boundary conditions

For these calculations all the degree of freedoms (dofs) of one edge of the plate have been suppressed: three displacements and three rotations. This has been done in the corresponding comm file with the fillowing rule:

clamped=AFFE_CHAR_MECA(MODELE=modelc,
                       DDL_IMPO=(_F(GROUP_MA=('Linex0',),
                                   DX=0.0,
                                   DY=0.0,
                                   DZ=0.0,
                                   DRX=0.0,
                                   DRY=0.0,
                                   DRZ=0.0,),),
                        FORCE_COQUE=_F(GROUP_MA='shell',
                                   PRES=10.0e3,),);

Also the applied load has been given here in the last two lines. The group Linex0, of course, defines one edge of the plate.

For this load case the calculated displacement at the tip of the plate is

w_ca = 8.460 mm for a mesh with 40 nodes (3*3 quad8 elements) and
w_ca = 8.473 mm for a mesh with 2300 nodes.

In both cases this is slightly more than the theoretical value based on the shell theory: +2.8 % and +3.0 % for 40 and 2300 nodes respectively.

Media:dsadasfdas.jpg

Comparison to solid elements

So it may be noted that the shell elements converge fairly rapidly to the correct value (here they even overestimate the displacement, can that be explained?). We expect that for solid elements convergence will be much slower.

In the graph below the results of some calculations are shown. We have used tetrahedron and hexahedron elements, both linear and quadratic.


Kw convergence.png

So it is fair to say that linear elements will much too stiff for normal cases. Therefor the displacements of the structure will be predicted much too low.

(I will add some more Salome and comm files later on. Going for a break now)

--kwarky 14:44, 7 May 2009 (CEST)