visualizing wind vectors (kangaroo)

Is it possible to visualize wind vectors in GH? Im curious if kangaroo could give some visual or even better computational feedback on the effects of wind acting on a building form or panel.

I did see this example which is pretty cool but it would be great to get some local feedback from GH or build logic that responds to it.

Replies to This Discussion

Permalink Reply by Daniel Piker on October 3, 2011 at 3:33pm

Unfortunately the way the wind force currently works in Kangaroo is one-way only.

So you can create a wind vector field which can affect a mesh, but the mesh does not affect the wind back.

I'd love to incorporate some fluid dynamics, but it's not trivial stuff, and not something I see myself tackling on my own any time soon.

I'd be very happy if someone developing a 3D fluid solver does want to work together to make this happen though.

Maybe it would even be possible to set up a feedback loop between Geco and Kangaroo via HoopSnake...

Permalink Reply by Charles Aweida on October 3, 2011 at 4:26pm

Thanks for the Reply Daniel, I was looking at your wind example and notice just what you described. Maybe if ecotect has a fluid dynamics plug in you could bring that data in via Geco. It would be pretty cool seeing all these plugins that are plugins for plugins talk to each other...

Permalink Reply by Kris Hammerberg on November 2, 2012 at 2:45pm

Hey Daniel,

Anyone take you up on that offer of developing a 3D fluid solver? Or have you heard of any other development in that area for GH?

Permalink Reply by Daniel Piker on November 2, 2012 at 4:09pm

You're a brave man!

As I say, I've always kept CFD at arms length during my physics simulation adventures. Solids and elasticity certainly pose their challenges, but they seem relatively well behaved and understood compared to the swirling turbulent maelstrom of fluid dynamics.
So I can only offer fairly general pointers, but I'd be very happy to do what I can to help.

I do recall seeing various little demos on here of basic 2D fluid solvers in, or linked to gh, but don't know of anyone who has taken it that far or is currently developing it.

I believe much of the fast fluid simulation you see in games and animation can be traced back to Jos Stam's seminal 'stable fluids' siggraph paper. It is relatively simple, fast, stable even with fairly large time steps, and looks nice and 'swirly'. However, it is far from realistic, and I'm not sure what it is useful for in practical design terms.

Realistic wind simulation in 3D, at architectural scales and at interactive speeds on consumer machines is I think still a way off from current state of the art, even using the gpu.

That's not to say there aren't interesting and useful things that could be done with CFD in design, but I think it is a challenge to come up with appropriate applications and figure out what aspects can be ignored or compromised, while still keeping it relevant.

Even working only with laminar flow might be worthwhile...

And this is some very interesting work using SPH:
http://www.physics.umu.se/english/research/statistical-physics-and-...
(I'm a big fan of the work of this Swedish group, and have been looking a lot recently at their approach to solids, so it could integrate well with that)

Permalink Reply by Kris Hammerberg on November 3, 2012 at 2:49am

Hmm. I'm not so familiar with mesh free techniques, but those videos are appealing. To be honest (or at least more specific) about my intentions, I'm more interested in implementing a meshing tool for unstructured grids with a PDE solver for heat transfer. So for me, Smooth Particle Hydrodynamics isn't so applicable as it isn't actually "fluid", per se, that I'm interested in. Of course a PDE solver for Navier-Stokes could be used for a lot of great things...

On the other hand, the great thing about their SPH model is that it looks like it can integrate very nicely with solids and rigid body physics as you said.

Permalink Reply by Daniel Piker on November 3, 2012 at 8:46am

I think an unstructured mesh generation tool could be great (and not just for fluids).

When I was looking at dynamic remeshing, I found a few interesting examples in CFD

(http://www.youtube.com/watch?v=JabyDtAn6pg).

Are you thinking 2D or 3D, and triangles or quads / tetrahedra or hexahedra ?

Permalink Reply by Kris Hammerberg on November 4, 2012 at 6:11am

3d would be ideal, obviously, but for my own sanity and as a first step, I'd probably try to tackle the 2d realm with a hybrid quad/tri (for a well behaved mesh close to boundaries). So an advancing layer of quads, followed by triangulation.

Although hybrid situations can be tricky to make general (also, I'd guess, slower: problematic for dynamic remeshing). Maybe voronoi discretization with some adjustment to assure it is as orthogonal and regular as possible near boundaries.

Thoughts?

Permalink Reply by Daniel Piker on November 4, 2012 at 7:06am

For incremental remeshing approaches I've only ever seen quad-dominant meshes generated rather than all quads.

Check out "An Incremental Approach to Feature Aligned Quad Dominant Remeshing"

and also "Quad-Dominant Mesh Adaptation Using

Specialized Simplicial Optimization"

Eppstein and Bern also have a nice relatively simple approach based on circle packing : http://arxiv.org/pdf/cs/9908016v1.pdf

Permalink Reply by Daniel Hambleton on November 3, 2012 at 8:22am

I'd be interested in working on this.

A while ago, I wrote a basic SPH simulation for Unity (but the code would port pretty easily) that incorporates the non-linear stress tensor, constraint functions, and the various smoothing functions. *DISCLAIMER* It is not even remotely close to the current state of the art, but it is a start...here it is.

I think one of the biggest technical challenges would be the scale of a realtime simulation. GH is not that fast - and even with fancy neighbour searches there is still a fair bit of calculation to do. Maybe applying the recent remeshing techniques to 3d....?

If there is any interest, I'd be happy to start a Git project.

Permalink Reply by Daniel Piker on November 3, 2012 at 9:04am

Nice!

What language do you use for scripting Unity ? I see C# JavaScript and Python are options. Do you solve the constraint functions using projected Gauss-Seidel or something else ? Are you making use of any solver libraries ?

and I agree - Remeshing in 3D could be very interesting as an alternative way of maintaining proximity information through just iterative local updates.

Permalink Reply by Andrew Kudless on November 3, 2012 at 4:13pm

Team Daniel + Daniel!!!

This sounds interesting. I wish I had a brain more inclined towards these higher level maths so that I could help. Looking forward to your collaboration on this.

Permalink Reply by Roland Verheul on November 12, 2014 at 7:36am

Hi, first post here..

So i built this model of a surf kite. It's all spring elements, set to work in tension only and not in compression as you'd expect from a fabric. Pretty stoked with Kangaroo, solves about 4000 elements in 1/10th real time on a 3yr old laptop..

Next step is to validate some elements and have an improved aerodynamic model. I used the standard wind mesh for the current model, but don't think it will suffice for what i want to do (flying the kite around on long lines).

I've been reading about aero-elastic solutions in this discussion, but i might want to take a simpler approach. This would be to take the original kite shape and define a number of chords and their initial airfoil shape. Then use this initial shape for calculating the aero forces during flight, depending on the (local) apparent wind speed and angle of attack. Of course the airfoil shape distorts during flight but basically you ignore this (but only take the airfoil start and end point as a reference).

You could use either a vortex lattice method or an analytic approach, like a CL/alpha slope. (This would require a feedback loop).

Does this make sense? I'm a pretty bad programmer so hope it will work. Anyways great tool!

Roland