Points as boundary

The graphical user interface of Elmer
Post Reply
saxonfahrer
Posts: 38
Joined: 27 Jan 2011, 13:47

Points as boundary

Post by saxonfahrer »

Hello,
i'm trying to simulate the bending of a plate, which is placed on thin pins (bending by pressure). I assumed "elastic plate" as an easy and fast solution method.
To keep the meshing effort low i tried to use only points as boundarys instead of small circles (fix deflection to zero at this Points).

Therefore i used gmsh to create the geometry, then i marked the points as "physical points" (and embedded in mesh) and created the mesh.

Now the problem: I can't see the points in Elmer.
Is it possible to have points as boundarys in Elmer GUI?
I also tried points with the Elmer Mesher (IN2D-files), but here i also saw nothing in Elmer GUI. Is this possible with IN2D-files?

Thanks a lot in advance.

Best regards,
saxonfahrer
saxonfahrer
Posts: 38
Joined: 27 Jan 2011, 13:47

Re: Points as boundary

Post by saxonfahrer »

Hello,
unfortunatley no response to the question up to now.

Did anybody used Points as single objects (physical point of view). Or in other words: did anybody uses Points as a Boundary?
And did this in Elmer GUI?

Thanks in Advance.

saxonfahrer
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Points as boundary

Post by raback »

Hi

ElmerGUI does not support points. Also philosophically we prefer BCs usually to be just one smaller than the leading dimension. Still, ElmerSolver supports nodes as elementtype 101 but that may be of little comfort if the meshing does not support it. Sometimes you just need to set one point. For example, to fix the potential level of a Poisson equation one point needs to be fixed. For that purpose there are two alternative keywords replacing "Target Boundaries" that define nodes

Code: Select all

Target Nodes(n) = ind1 ind2 ... indn
Target Coordinates(n,dim) = x1 y1 z1 .... xn yn zn
The latter chooses the nodes that have the shortest distance to the defined point. Look ElmerSolver manual for more details.

-Peter
saxonfahrer
Posts: 38
Joined: 27 Jan 2011, 13:47

Re: Points as boundary

Post by saxonfahrer »

Hi,
thanks for the feedback: that's excatly what i was looking for.
Thanks a lot.

Best regards,
saxonfahrer
saxonfahrer
Posts: 38
Joined: 27 Jan 2011, 13:47

Re: Points as boundary

Post by saxonfahrer »

Hi,
within an elastic linear plate Model i used the "Target Nodes" but now i got a new problem (probably i do something wrong):
From a certain number of nodes on (~150-200) i get the solver reaction "ERROR:: ReadAndTrim: Output length exeeded." while reading the model.

I had a look to the doxygen-server and there it's not mentioned, that the ReadAndTrim function do have a max. number of elements.

Could anyone give me a hint what to do? I need to fix ~15.000 nodes (maximum) and i would like to have a proper solution for this.

Thanks a lot in advance.
saxonfahrer
Posts: 38
Joined: 27 Jan 2011, 13:47

Re: Points as boundary

Post by saxonfahrer »

Hello,
i'd like to bring this topic up again.
(short summary: i'd like to use nodes as BCs (as springs in elastic plate solver). The problem: i like to use up to 15.000 nodes!)

Meanwhile i've read some comments @ other threads (related topics).
What i understood:
- the "target node" option is only suitable for small number of nodes (~100 ?)
- i should go for "scalar fields" to apply the condition to the nodes

To be honest: i've no idea how to do the second advice (scalar fields).

It would be very nice i've someone could give me a hint.
I've already looked in to documentation but may be overread the important notes.

Thank you in advance.

Best regards,
saxonfahrer
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Points as boundary

Post by raback »

Hi

The plate solver has a keyword "Spring" that might give you what you need. In SI system the unit of this spring is Pa/m. The nodal spring would be N/m that does not exist. Generally Elmer works more frequently with fields rather than nodal quantities.

-Peter
saxonfahrer
Posts: 38
Joined: 27 Jan 2011, 13:47

Re: Points as boundary

Post by saxonfahrer »

Hi Peter,
thanks for the quick response.
I already using the keyword "spring" for standard face BCs.

Could you give me an idea or better example (or hint were to find one) to create a field including up to 15k areas with a certain value for spring and everywhere else extrem low (but >0)?

So far i only have seen "tables" to e.g. give temperature dependent coefficient of expansion.

Thank you very much

Best regards,
saxonfahrer
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Points as boundary

Post by raback »

Hi

So you want to really have the spring at certain nodes. So 15 k nodes with spring randomly distributed? Or how do you determine which nodes should have spring and which ones not?

There is a solver that could create a field of distrbuted data (DataToFieldSolver) but still your data should be smooth so that it makes sense to evaluate it using FE basis.

-Peter
Post Reply