HELP please please! Two stiffnesses in dynamic relaxation

Hi everyone,

I am doing a dynamic relaxation for a gridshell structure used for a library but I have some areas with a too small height and others with too high height. I was thinking of playing with the stiffness of the lines of the mesh: increasing it where I want it to be less deformed and decreasing it in the opposite case. I tried to do so by just adding a second spring box kangaroo component but it does not work...

I already posted twice this message but I am trying again hoping someone could help. I know it is possible cause I saw a discussion here talking about it but the complete script was not available. Yeah, I am desperate :(

I've joined the rhino file with the mesh and the two ghx files: the original one with one stiffness value and my attempt of modification with two values.

In the pictures below, this is a test with a stiffness value of 1000 and a gravity value of about 120. This is too high in fact, but this is for now the only way to have a min height everywhere of 2.2 m. I would like to change the rigidity of the elements by areas to have the desired height values (approximatively).

Tnx in advance!!

top view of the mesh before D R :

perspective view after D R with one value of stiffness (volume in grey is to have a reference height of 2.2 m)

Attachments:

Replies to This Discussion

Permalink Reply by Daniel Piker on April 5, 2013 at 6:07pm

Hi Ikram,

I think part of the question here is what kind of shell are you after ?

A typical gridshell has equal length curves between the nodes, and forms a regular square grid when laid flat. When pushed up into shape, the boundaries are fixed and diagonals braced, but the length of the curves from one node to the next remains very close to equal. While you could adjust the heights of your form by varying the stiffnesses across the surface, the result will be uneven edge lengths, so not buildable using that system.

If you do want to keep equal edge lengths, one way you can approach it is to model a target surface, then drape a regular grid over it. The result will be similar to the Chebyshev net approach used for the Savill garden gridshell described here (though that is a purely geometrical construction). You can even use this approach to generate an initial form, then fix the boundaries and do a further step of hanging-chain type form-finding to optimize for compression (though in continuous lath gridshells the bending is also important, and can be good to take into account in the form-finding).

If it is not necessarily an actual gridshell you are after, but a more general compression shell, then you have even more freedom (though there are some subtleties to the weight distribution when lumping masses at the nodes like this).

Permalink Reply by Ikram Ch on April 7, 2013 at 1:50pm

Hi Daniel,

In fact I am working – master thesis - on the optimization of a gridshell on Matlab with an ant colony algorithm but before optimizing, I need the shape and I thought of D R as the form-finding process (I tried other things as force density with a software called Easy but it did not work due to the complex geometry and boundaries).

So I don’t have the constraints of equal length, of course the length should be reasonable (maximum 2 m for example). They just should be in the same order of magnitude (0.5 to 2m). My only real constraint is that the areas around the boundaries can be used for the library (so a minimum of 2.2 m at least). And of course, I assume the maximal height should not be up to 10 m.

In the file ‘two stiffness’, there s my attempt of the script with two stiffnesses but it doesn’t work. I also tried on a simple flat meshgrid by selecting half of the lines (after baking it) with one stiffness value and the other half with another value but still there is an error. Can you take a look at it or do you already have a script for this issue?

Btw, I already thank you very much for your answer!

Ikram

RSS