Prismatic Skin

Replies to This Discussion

Permalink Reply by Julian Buhagiar on March 4, 2017 at 5:16am

For some reason I'm not getting notification of replies?

Yours is a very good solution and use of scripting is fantastic. I'm only just beginning to tap into the power of the SDK, but for the time being whilst I'm still getting used to certain concepts. I try to mock up what I want with the native components and cluster them, that's my recommendation for anyone new to grasshopper, but definitely move onto scripting when you find performance of the cluster too slow or you need to tap into different software libraries.

Permalink Reply by peter fotiadis on March 4, 2017 at 8:18am

I do hope that you haven't misinterpret my meaning: it's a challenge just for fun (given the fact that you have the touch) and nothing else.

Other than that the use of a suitable programming language is the only way to "bind" the vast amount of applications/requirements that coexist in a contemporary AEC practice (or in some practice that has some chances to survive).

Since time flies the usual argument: someone should start the easy way and then "advance" ... conflicts with the fact that ...er ... time flies. The typical result is that the door is closed before even is opened (and unemployment or under payment is the added bonus). All that whilst we are at the dawn of the parametric era.   

Permalink Reply by Julian Buhagiar on March 5, 2017 at 12:19am

Thanks for the complement. I love a challenge, I'll get back to you once I've begun looking at it.

Permalink Reply by peter fotiadis on March 5, 2017 at 12:32am

The main thingy: the  struts clash thingy.

Imagine a node in a mesh (or a "flat" graph) with some valence. We can use a classic PointToPoint connectivity Tree and get the neighbor points. But in order to check adjacent neighbor pairs VS the node (for finding the angle) we need to order ("sort") the neighbors. At first one can assume that's piece of cake: 

Fitted plane intersects with fitted sphere > circle > to nurbs curve > sort points along curve.

But ... this (very slow approach) doesn't work IF the node belongs to some truss with W depth. 

Tip: the orthodox way to do this is using indices Lists intersections.

Permalink Reply by Julian Buhagiar on March 5, 2017 at 12:41am

Right, I think I understand the problem. Will have a play and let you know, I have a few ideas I wish to try.

Permalink Reply by Julian Buhagiar on March 5, 2017 at 1:07am

In the absence of a GH to work on and armed only with a mobile phone, it forces me to visualise the problem before putting capsule to canvas...

My first thoughts are either dot product of the edges associated to a node which can be filtered by index or my other thoughts are related to triangular mapping and whether the angles can be derived on a face basis rather than node basis since the tree topography would reflect this (I think) (all we actually need to know is the location of the new central node relative to the original face verts and its perpendicular displacement to then work out new angles created from the original meshes face (to which we can get angles quite easily.)

Permalink Reply by peter fotiadis on March 5, 2017 at 4:20am

Well ... forget slats/ribbons/stripes and stick to the core of the matter.

Do:

A prox 3d graph (lines connecting random points within a cube for instance), then use Sandbox for the connectivity stuff ... and then start thinking ways to check clash matters between "truss like" struts (i.e. some offset for the ball [in real life: ball + sleeve] + a cone + a tube + a cone + same offset for the ball [in real life: ball + sleeve]).

Expect: 

A way to deal with items in collections on a per item basis. That's impossible to do with components thus I'll provide you with the means to do it AFTER locating potential issues in the nodes. Of course raising the ball radius is not the way to address this (aesthetic disaster).

Expect:

A way to place gazillions of some (the same) thing - cones for instance - in real time using instance definitions: the holly grail of engineering.

Permalink Reply by Julian Buhagiar on March 5, 2017 at 4:26pm

Getting there...I've got things organised, just up to the challenging bit now. Hopefully what I've done so far is good enough and robust enough to cope.

Permalink Reply by peter fotiadis on March 5, 2017 at 11:45pm

Here's a thingy for the end of the walk: puts stuff (instance definitions) in 3d space using points or planes or regions (and random points). Using the region (boundary) option compare the speed of drilling 35 holes VS placing 666 instance definitions ... do I need to say more?

Get familiar with it and when time arrives we can use a (slightly) modified version to place all the truss components as instance definitions  (an other C# would create the instances).

Meaning that the only output that we expect from a proper "truss maker" is a collection of coordinate systems.

Why may you ask using an approach like this? Because an "average" truss of a certain size may contain 20-30++K "items".

Moral: long is the path (and hilly)

Attachments:

• Instances_place_entry_level_V2.3dm, 2.9 MB
• Instances_place_entry_level_V2.gh, 155 KB

Permalink Reply by karelovic on March 24, 2017 at 11:48am

Do you know the way to chose the orientations of the slats, i mean not randomly? I would like to make slats with specific orientation on all slats with the face orientated to east, same to all slats orientated to west etc...

Thanks

Permalink Reply by peter fotiadis on March 24, 2017 at 12:34pm

Well ... you can use Dot/Cross products in order to test Normals and/or X/Y plane vectors against some "orienting" vector.

Post what you have done insofar and we'll see ways to elaborate more.

But have always in mind that using moving "louvers" (meaning service, cleaning, failures and maintenance) in most of real-life cases it proves just another chimera. Very rarely solutions of that type yield some meaningful break-even point ... meaning that you start aiming to save the polar bears and you end into another rabbit hole.

Google "smart glass" > future is there.

Permalink Reply by karelovic on March 20, 2017 at 6:31am

Hey! Thanks a lot again!

Finally i would like to have 4 different patterns for each facade because of the sun path. The light needed inside is not the same everywhere.

Is it possible to have one facade with the same triangulation pattern but only horizontal and vertical slats (no diagonal or random orientations)? The aim is to have horizontal slats on all the triangle faces wich are orientated to the sky (up), and vertical slats on those wich are orientated to ground (down).

One more thing i would like to do is to make larger triangles faces on both sides of the facade (left and right). Just at the end, to connect with others larger pattern facades.

Do you think its possible? Im don't know if its clear or not...

• ‹ Previous
• 1
• 2
• 3
• Next ›
• Page