Shape (?) Optimization for 1D Elements
Hey guys,
I was wondering if it is possible to setup the following optimization problem.
Lets say I have a set of nodes and bar elements. This framework has some loads and boundary conditions.
I am not sure if the node position of each node is optimal regarding compliance.
I would like to define 'shapes' for the nodes where I can set some pertubations e.g. Node x can move in a sphere of radius y.
I have not seen this kind of optimization although I think it must be possible from an algorithmic point of view. Any suggestions?
Thanks!
Answers
-
Hi @Merula This sounds interesting. I am going to try this and I will update you soon.
0 -
Much appreciated!
0 -
Some update:
I tried with a simple beam but the node movement is not following the spherical/cylindrical coordinate system.
I will share my latest findings soon.
0 -
It could also be a cube, if the spherical movement is the problem.
I just can't see how I can implement the optimization setup (i.e. the design variables )
Best regards,
Merula
0 -
Trick here is I am realizing a 1D beam to 3D shape. So I am using 3D element grids for shape optimization.
0 -
@Prakash Pagadala thanks. Can you show me how to realize this or send me a model file where I can see this?
Thanks in advance
/emoticons/default_smile.png' srcset='/emoticons/smile@2x.png 2x' title=':)' width='20' />
0 -
Do I have to use morph volumes for this? I would like to just simply define that each node can freely move in x y z within a certain range. Minimizing the stress is the objective. I share a model file, where I have a simple beam setup where I would guess that a symetric design would be optimal.
I stuck at defining the shapes.... I've learned that free shape is not possible because I have no 2d/3d elements
Best regards,
Merula
0 -
-
Sorry for a late reply,
You are right, shape optimization is not possible with 1D elements. You need to model as either 3D or 2D. So I created a 3D model with a layer of 2D skin and used shape optimization to accommodate all Dofs,
Unfortunately, I don't have the mode with me.
0 -
I digged on this a little deeper and found a way to optimize 1d elements....
it is possible if you choose freehand under morphing and save the shapes after having changed the node positions.
Best regards,
Merula
0 -
I agree, but whatever you get is a concept and this needs to be fine-tuned with gauge optimization, maybe?
Apart from that, the results need to be validated with 2D/3D elements, as the number of Dofs will be different.
The shape on optimization level is not supported for 1D elements.
0 -
Yes, this is a concept as it does not change the shape of the actual 1d elements but the location of the linking points, but nevertheless important when you come from a topology optimization and want to fine tune the reconstructed results (when using simple bars for example).
A gauge optimization would be the final step.
Best regards,
Merula
0 -
Hey guys,
what do you think of the result I got for the CClip.
I saw it is not a symmetric result although the input and loadcase is symmetric. Ignore the buckling constraint for now.
My question is: Is this procedure correct? The compliance is better than with topology optimization.
I attached the following results:
1) Input Deck of the topology optimization (modified version of the cclip tutorial): cclip1.fem
2) Result of the topology optimization (modified version of the cclip tutorial): cclip1_des.h3d
3) Input Deck of the Shape Optimization ShapeOptimization.fem
4) Result of the Shape Optimization ShapeOptimization_des.h3d
Best regards,
Merula
0 -
Hi,
This looks pretty good. I see the results are better.
0
Altair Employee