FEA -> Flow Rate

Comment

Comment by Francois Dufrenoy on March 18, 2016 at 8:22am

I thank you Charles for your explanations. i like your approach.

I did not know Karamba.

Are you using large displacement behavior for the elements? (it's definitely needed)

One other point, According to the pictures of your experimentation, at the final stage some cells are collapsed.

Is it possible to introduce contact between strut in Karamba FEA? If not, experimentation and simulation will provide different result in case of contact (which will interact with adaptation phenomena).

When i write about material properties, i'm thinking to the rules which link stress and strain.

The relation takes form as here after in elasticity domain (in plasticity we need to introduce non linearity dependencies: more difficult. With an experimental approach you can validate firstly the elastic domain (until reversible deformations are visible) then you can investigate more complex plastic deformation (depending of possibilities of Karamba or other FEA code (i don't know if it is possible to run into grasshoppe3D a job for others numerical code like NASTRAN))

Concerning cells, I thank you for the link. If you take a look to Altair Optistruct(R) / lattice structure, you will find some examples. 

I'm not a specialist of Grasshopper3D and I'm not in capacity to answer to your question. I hope somebody else will read the post and will help us. My feeling is  to use a topological optimization like Millipede plug in and then to use density of material (which is the optimized solution for a continuous space) to perform an adaptive mesh of the useful volume of material. Then on the remaining mapped volume, we need to use a translation of dealing with (volume/density/strain/stresses/cell neighbors)  levels in order to  fit the adapted cell.

Such transformation may use your experimentation to suggest adequate solution for each cell.

Comment by Charles Fried on March 17, 2016 at 1:45pm

I'm starting by doing the FEA using Karamba, treating each line as a strut ( see below). This then informs the G-code for my print on which I finally conduct the physical test.

"KSH-1-15" in the graph I initially posted uses a technique whereby I increase flow in areas of compression and tension. However, it seems the results aren't as good the one illustrated below.

I'm not quite sure I understand the technique you describes in regards to taking material property into consideration through an iterative process. Would you mind elaborating? 

Regarding your last point, I agree it would be a very interesting route to go down. A similar methodology is described in this paper which uses a genetic algorithm to orient each cell. If you know of any other examples, please share. Do you know of a cell type that lends itself to this kind of transformation?

The Normal force I'm using:

Comment by Francois Dufrenoy on March 17, 2016 at 10:03am

Thanks Charles,

Objective : minimize displacement at iso load.

FEA is a numerical simulation. If i'm not wrong, Your original approach use mechanical test (by deducing internal loads, are you performing equilibrium of deflected section in order to extract loads, at each step of the general deflection, into each beam and finally to find loads into clamped sections?

It could be interesting to perform an identification of the mechanical properties of the material by optimization. Optimization loop can use iterative FEA workflow (in large displacement) to correlate simulation on experimental measurements.

With this workflow, you will identify mechanical properties (matrix 3x3) for the raw material characteristic between stress and strain. After it could be possible to process a general optimization and use dictionary of parametric cells families which could be adjusted via planar transformation (translation, rotation, size/density, thickness, continuity with adjacent cell...like a lattice structure) with the same mechanical properties (if we don't take car about 3D printed orientation)

Comment by Charles Fried on March 17, 2016 at 9:04am

Hi Francois, 

Sure, so I'm doing an FEA on the structure you see above where the load is applied to the end of each "arm" and the "feet" are fixed in all directions.

From this, I'm then extracting the normal force for each strut and mapping that to the extrusion flow of the G-code. Below you can see a few early results from the physical test where "H-17-1" is not optimised and "KTR-1-1S-2" is.

Comment by Francois Dufrenoy on March 17, 2016 at 8:55am

Hi Charles,

Can you explain?

It sounds very interesting.

I would like to know your objective, your workflow...

Can you share the project?

With regards