Conversion from Calculix to Deal ii.

[Andrew Robinson]

Hello,

I'm learning FEA, and have worked my way through the first six lessons of the DEAL II tutorial.  Nice. :)  I am not seeing, though, how to do mixed dimensional problems.

The simplest example I can think of: I want to model a piece of netting woven in a certain way.   The cloth/net itself is a 2D object.  But the problem I am interested in solving is 3D,

From the Deal II tutorials, I get the impression that Deal II is based on a paradigm that the dimensionality of the object determines the space that the object is simulated in.  So, this is a little puzzling to me.

Before Deal II, I was attempting this problem using Calculix CCX.

The project's been stalled for four years because of bugs in Calculix.

SInce the problem is large ( A typical piece of cloth has a half million nodes); I wrote a python script to create a piece of 2D netting using a low computational cost three point beam element:

https://web.mit.edu/calculix_v2.7/CalculiX/ccx_2.7/doc/ccx/node53.html

The python script handled node numbering, degrees of freedom, and all of that to create a piece of cloth with hexagon type weave and fixed at end two nodes.  (Essentially a Hammock, hanging between two fixed points. )  I only wanted to do a static analysis of how such a hammock would SAG under weight/gravity.

I'm not seeing examples of how to do similar things in DEAL II; and Calculix crashes so I'm trying other FEA systems.

.

I've been able to complile DealII from source, and paraview, and install and run the demos.  (I'm good at Linux and problem solving, and post before now!!!)  Although my laptop is not a supercluster, just an 8 core Ryzen, the performance is plenty good. 

These are problems that I can let run a week or a month to solve, if necessary.

To give a very clear example of what I was trying to do, here is the output from the python program, truncated, so the problem's outline can be seen easily.

Comments, suggestions, links to tutorial(s) I missed, appreciated.

Thanks.

--Andrew Robinson.

[ START CALCULIX CCX PROGRAM ]

 *HEADING
Model: hammock   Date 24-Jul-2019
*NODE
1     ,     0.0,     0.0,   661.9
2     ,   399.7,  -140.3,   431.3
3     ,   400.3,  -138.6,   430.6
4     ,   399.7,  -136.8,   431.3

... SNIP ...

353661,  3017.6,   142.0,   431.7
353662,  3424.0,     0.0,   661.9
*NSET , NSET=Nall, GENERATE
1,353662
*NSET, NSET=FIX
 1, 353662
*ELEMENT, TYPE=B31 , ELSET=Estring
1 , 1 , 2
2 , 1 , 164
246 , 2 , 165

... SNIP ...

*ELEMENT, TYPE=B32 , ELSET=Ebar
101072 , 54544 , 54945 , 54944
466785 , 298353 , 298754 , 298753
*ELSET,ELSET=Eall
Estring,Enet,Ebar
*BOUNDARY
FIX,1
FIX,2
FIX,3
** Polyester netting material defintion, simple.
*MATERIAL,  NAME=PolyEster
** N/mm^2 Youngs,  Poisson, (warp) diagonal 1/3 that value other dirs
*ELASTIC, TYPE=ISO
3600, 0.16
** Density 1.385g/cc (1385kg/m^3)
*DENSITY
1385
** BEAM SECTION ... ; thickness local tangent/x, thickness local tangent/y.
** Thickness from square estimate= ( 0.37mm * 0.37mm / pi )**0.5 * 2. ~= 0.42mm
*BEAM SECTION, MATERIAL=PolyEster, ELSET=Enet, SECTION=RECT
0.37,0.74
*BEAM SECTION, MATERIAL=PolyEster, ELSET=Estring, SECTION=RECT
2.0,2.0
*MATERIAL,  NAME=Aluminum
** N/mm^2(AKA MPa) Youngs,  Poisson ratio,  ?? (warp) diagonal 1/3 ??
*ELASTIC, TYPE=ISO
100000, 0.32
**DENSITY
*DENSITY
2710
** BEAM SECTION ... ; thickness local tangent/x, thickness local tangent/y.
** Thickness-- Assume 5/8 inch. (~16mm)
*BEAM SECTION, MATERIAL=Aluminum, ELSET=Ebar, SECTION=CIRC
16.,16.
******************************************************
*STEP
*STATIC
*DLOAD
Eall,GRAV,9.81,0,0,-1
**DLOAD
**LoadNodes,GRAV,9.81,0,0,-1
*NODE PRINT, NSET=Nall
U
*EL PRINT,ELSET=Eall
S
*NODE FILE
U
*EL FILE
S
*END STEP

[END PROGRAM]

[Wolfgang Bangerth]

On 12/9/22 21:01, Andrew Robinson wrote:
>
> From the Deal II tutorials, I get the impression that Deal II is based on a
> paradigm that the dimensionality of the object determines the space that the
> object is simulated in.  So, this is a little puzzling to me.

Andrew:
No, that's just the default :-) We can solve problems where the dimensionality
of the domain is less than the dimensionality of the space in which the domain
is located -- we call these dim and spacedim, and dim can be smaller than
spacedim.

You might want to look at steps 34, 38, 54, 60, and several others.

Best
W.
--
------------------------------------------------------------------------
Wolfgang Bangerth email: bang...@colostate.edu
www: http://www.math.colostate.edu/~bangerth/