A Young practice breaking new grounds operating between old parallels in Spatial Design, Performance Arts, Computation Design and Contemporary Media Production. We contextualize a contemporary link, an echo between society, technology and architecture in design solutions, narrative interfaces and live performance backings. Located in Frankfurt am Main.
So Thanks to a friend giving some inspiration, this little exhibition-purpose pseudo code project has made some progress and in a couple of weeks it would be ready to be shown.
Site study/ Site modelling과 SD 과정을 연결하는 플랫폼, 전시용
이미지 오토 슬라이드쇼로 기능에서 3D로 전환해서 사용자가 직접 custom shader로 내비개이션도 가능.
Schematic Design 단계에서 많이 보이는 Exon Diagram Sequence들도 3d 로 구현 가능
이제 애니매이션은 실시간으로 GI가 가능해지기 때문에 frame당 Render하는 과정보다 많은 시간이 절약 됩니다.
vvvv같은 visual synthesizer (grasshopper 와 비슷한 개념의 소프트웨어이나 max/msp 계열이고 new media art나 엑스포에서 자주 보이는 디지털 터치 인터페이스 디자인에 쓰이는 소프트웨어)로 바로 구현할수 있어서 역시 gh/revit같이 실시간으로 geometry를 교환할수 있습니다.
little custom work on visualizing design iterations.
This little vvvv patch allows for setting up multiple videos in order to document design studies and help decision making in the process.
Helps you see an overview of iterations taken in vvvv.
Useful for documentation and presentation.
Next step includes either on screen metrics as part of the grasshopper animate screenshots
or imported texts/cvs/excel information inside this custom software.
Currently you can either play through the loop by loop speed, or use slider to see
Next feature might include being able to mark some slides - and show on screen which slides are marked.
Ideally, it should be possible to click on the numbers shown on screen and brings the viewer to that slide.
Here I am sharing the vvvv patch. you need to first install vvvv (it is not installed by an installer, if you need to know how to install vvvv, take a look here) I am running it with 31 x64 version.
Work in Progress : Architectural Visualization with Unity 4.5.
I have been mostly focused with vvvv for interactive framework for urban design review however with vvvv, there is a limitation that one has to be standing infront of the installation piece. So I have decided Unity would be aiming a whole different ball park. Maybe not as savy and information rich as a vvvv driven interactive media installtion, but more useful for showing people on the otherside of the continent.
I hope organizing a viewing experience this way can help me answer key 5 questions that a go-to-market strategy seeks to answer
WHO will we actively target within the market?
WHAT will be our product portfolio for target customers?
HOW MUCH will we charge for our products for different customers?
HOW will we promote our products to target customers?
WHERE will we promote and sell our products to target customers?
Its quoted from wikipedia :) http://en.wikipedia.org/wiki/Go_to_market
Here are some illustrations and design research work on a Mixed Use Development Concept for San Hose. It is in an early stage and I cannot say I have done the market research to really hone down on the ratio between residential, retail and hotel spaces.
And because of that, I have decided to have separate wings for programs so that they can be sized and scaled dynamically.
The demonstration of how this works will be shown on a separate set of videos.
In the videos it will outline the workflow from
1) a grasshopper definition to
2) a rhino file that archives numerous iterations, and
3) the new Rhynamo workflow that takes them into Revit as Conceptual Masses,
4) grasshopper environmental analysis to create layers of information to create the Typological Spandrel Mapping, and
5) based on the TSM method we will be able to detail facade components with Revit.
6) Then some Revit Python Shell examples will be shown for detailing facade components, and automation of shop drawings and lastly
7) example on how to use scheduling, material take off and cost estimation feature of revit.
Here in this particular design process, the described step 4 is one of the key design layers. It breaks down to two main category of analysis- indoor and ourdoor analysis.
The outdoor analysis is taken in a form of radiation analysis. The radiation analysis gives a layer of information that drives the articulation of vertical surfaces. Where are more accumulation of heat, there are more veins behind the facade system's rainscreen panels and in front of the load bearing columns and unitized curtain wall, and more folds and puffs that help dissipate the heat channeling the heat up along the load bearing columns in a way that draws natural ventilation across the floor plate. This facade system reduces the amount of cooling needed during summer and there is minimal maintenance cost.
During the winter each rainscreen become closed loop system and they act as thermal collector facade systems.
For the indoor analysis, one has to design windows and determine window to wall ratio for the segments of exterior wall. The ratio determines the amount of daylight inside, and the ratio needs to be sensitive with the respect to the orientation of walls. // 10.17.2014
More details on updates that are on the way including : CFD analysis to determine strategic locations for rains screen panels, Section drawings with natural ventilation analysis (to be done with Ambiens), Preliminary Energy Modelling, Prefabricated Unit layout and interior design features, GPU rendered promotion wheel (possibly with green screen actors composited), and last but not least, Unity 3d Online walk through.
Tetra Terrace Plaza - mixed use concept design studies from xsyam on Vimeo.
Who would like a definition that takes a mesh and let you choose a number of different surface types that approximate the mesh?
So Lets say your job is to rationalize a mesh surface that was created by someone else.
and you want to make a definition that takes a mesh and find the minimun number of unique parts for the least amount of deviation.
one way to visualize the curvature of a mesh is to use Mirco Becker's mesh curvature python script and have it analyze and show in colors.
or Yuxiao He's work on "four customized developed tools"
credit : Yuxiao He (link)
and an example of mesh curvature by color applied to my design work
Then the next question is, how do you recognize similar patterns within the field of color gradients? Just looking at the picture we can see there are quads that have red in the center but some have Yello around the edge and some blue and some white.
and still the harder question, how do we relate their custom shapes to their custom Performative role in relation to their whole? maybe I can to that question later on If I am lucky.
As been pointed out numerous times, in order to come up with a material take off and shop drawings in the end, it would be economical to evaluate different typologies of custom components. One of the key reasons I have spend months getting into Revit and interoperability is actually this - Revit has a cost estimate, scheduling and material takeoff feature that grasshopper and rhino lacks, out of the box. (and it is much easier and faster to generate architectural drawings in Revit)
Surface rationalization is a familiar topic for many.
Here you see an example done of surface rationalization with Maya MEL, copyrighted by Zaha Hadid architects
It seems to me that the technology described in the link above is one of the core methodologies that makes it possible for Zaha Hadid Architects to estimate the cost and the constructability of their designs (my speculation)
Considering work of Benjamin Felbrich, I remember he posted an interesting video about a script that takes pieces and puzzles together a dodecahedron. The video indicates that we might begin with the pieces only, without knowing about the overall shape that we would reference in order to check our progress. And this seems radically different from the quad panel rationalization, which has a reference surface in the beginning for the qual panels to approximate. In his example, he is working with an array of different shapes rather than one quad.
Inspired by the topic of Surface Rationalization, the automated puzzle solver and my own interest in outlining Advanced Performative Design - what I call Typological Spandrel Mapping, I have been looking into new ways to rationalize architectural surface.
Probably a big difference between the Maya MEL quad panel rationalization and my approach is that I am trying to preserve as much of the data structure of components in a form of tree between the different panels/ and types generated in rationalization surfaces.
The reason what I think it might be both interesting and useful is probably best illustrated with an analogy about how vray works, about an ant that tirelessly work in order to find a yellow leaf from a tree as shown in this link
This applies on at least two levels - if I have a surface made of many quads, I want to know which quads are not planar. so the quads can be organized in a binary way- planar or not planar and that distinction can be the first node in organizing the panels.
Secondly, lets say time flies and we have many iterations of quad panelization. Maybe each iterations can also be organized in a binary way - which iterations have more than 50 percent non planar quads and which have less, for example. And This would be a different way of evaluating iterations than, lets say the octopus plug in for grasshopper which is not a binary organization of possibilities but a multi variable one.
It should be possible to tag an instance of parametric relations, and other iterations and it should be possible to document them in way that can be easily carried into the next session. So that we can search among the design solutions and instances of specific parameters we spent time and energy like we search on google. I don't think google would be where they are if they had to retag and revaluate every websites everytime we google, and I don't know if we would be using google like we do. So there should be a method for quarrying objects in database of computed models, ideally.
We know we can do this either by baking into Rhino (we can even automate the baking as shown in designanalyze) or we can save the parameters in numbers and text files. With saving the geometry, we have the benefit of excessing the iterations directly from the outside rhino with rhynamo. The problem is that at the moment the import symbols generated by rhynamo read.rhino and import.instance are invisible to some features of Revit, such as scheduling so there needs to be addition work to make them visible as Revit Families, it seems.
Back to the earlier question about how to rethink the architectural surface and how to use computer to help us see patterns in a customized solution, I want to talk about something called space division algorithm.
Like I said earlier, I am trying to preserve all the necessary data structure of components in a form of tree between the different panels/ and types generated in rationalization surfaces.
One approach is that we take the vertices of the mesh and apply some kind of binary spatial partitioning algorithm and created bounding boxes out of the vertices.
The main purpose of this is so that we can sort the boxes into groups of similarity in terms of volume, dimensions, and so forth. So where there are a lot of vertices, we will get a lot of small boxes and where there are sparse, we get big boxes.
Some examples of box morphing are found here.
https://vimeo.com/54969747
http://zhangyixunya.blogspot.com/
Then we chop up the mesh up with the bounding boxes, and we superimpose all of them to see if they resemble each other. So I guess it would be like comparing a lot of meshes by their normals and gaussian curvatures. How to go about the comparison? Right now, I am not sure.
But what I know is that the benefit of having the bounding boxes is that we can morph them so if the split meshes are not matching, we can morph the boxes until the split meshes are more and more similar to each other.
So in order to break down the steps required, we need to
1) get a mesh object - a Standford Bunny and a custom mesh that has a nearly equal amount of triangles as with the bunny.
2) create a 3 dimentional kd tree from all the vertices. This step, obviously breaks down to many small steps, to be elaborated later. I am trying to outline the big strokes here. However, I do want to mention that one of the key question is how to draw the dividing planes between points - should it be xy plane, yz plane or xz plane? Here we want to bear in mind that whatever the dividing plane is, we want it perpendicular, not parallel to the dividing plane before the current division. Otherwise we lose the connectivity between all the branches. Ideally, we will end up with many closed brep boxes only.
There more issues but for now we move onto the next step.
3) once we have the boxes, we split up the mesh. This is a tricky process and probably is the bottleneck in the whole process. One way is to convert mesh face to surfaces and then to split the surface with mesh, group the pieces that are within the bounding boxes, join them and turn them back to mesh patches. The bottleneck is the fact that surface split operation in grasshopper is obviously very slow.
4) We compare the mesh patches. we take the naked edges and compare the profiles, take the center point and measure how far the corner points are, determine which points are the four furthest points, and etc. To be honest, I dont have it all figured out - how this could work and how useful would it prove to be.
5) We have to take some analysis result and group all of them into a given number (parameter) of lists. We want to bear in mind that we are not grouping by size, but by the similarity of surface curvature. This means that if there are a bunch of little pieces that could be all taken out of a one large surface because there are exact matches, then instead of producing little curvy pieces, we make one large curvy piece and cut the little pieces out of the large one (see illustration) - such fabrication technique was taken back in 2003 at sci arc for an installation by Marta Malé-Alemany
Imagine a patient who suffered a severe burn and the doctor will take a little skin out of the patients back in order to heal. Human skins can stretch, but we are assuming to be working with a rigid material.
6 ) now that we have groups we take the all smaller pieces of all groups and see if we can find a matching spot on the big surfaces with the little pieces. If we can within a given (parameter), then we are done. If there are left over pieces that can be overlaid within on one of the large pieces then we morph them using the box morph.
It is very possible we end up with a very leaky mesh when we put them back together after box morphing. And it is very possible that the smallest box is so small and the largest box is so large that a small patch of mesh have more than one matching instances out of one big mesh patch.
We may end up with a 3 dimentional jigsaw puzzle that in the end does not qualify as having a positive net present value in the eyes of developers and therefore does not need to be solved.
It should be clear that what I am trying to do here is a kind of optimization process where its goal is to reducing the number of unique pieces required for one production (project), to make the production simpler and close to initial input mesh. But I imagine there to be far more powerful application of such process/ algorithm, esp for analyzing a very large terrain and identify similar patches of earth.
The goal isnt necessarily making a geomtry more simple, but to find an optimal number of typology to describe a shape with satisfying result.
As I understand, mainly BSP-Trees are used for collision detection and handling in computer graphics. Its a pity I still lack the skills to code this myself but its an idea if anyone thinks this might be a worthy approach maybe there is a possibility of creating a plugin, if there isn't plug in that does something similar already.
I have been spending a lot of time and energy working the interoperability with Revit and my impression is that, with the recent development with rhynamo, it has become very easy to push Brep and surfaces into Revit.
Then grasshopper can mature even more into more geometric definition, pattern recognition and multi objective optimization. Then the speed of computation and the documentation of iteration becomes really important as well, to keep the process light and to the point.
A Pavilion converting an underutilized plaza on a main business throughfare in frankfurt. While farmers markets are common and weekly events that are very well intergrated to the life of frankfurt, frankfurt as a city lacks spaces for cultural activities and exhibition spaces for thosewho seek to create opportunities that are yet to exist.
This pavilion creates bufferzones between the casual and the business of Frankfurt, with changing exhibitions that are all temporal and designed to engage urban issues of Frankfurt. At the same time it seeks to act as an interface between business-casual Ideas and venture capitalism.
One example is an exhibition of interior designers that seek to innovate office space and housing space. Model houses are to be docked on the plaza above an undulating canopy, creating new visions to explore while acting as a public patio.
blakekathryn
kianamaiart
gummygunk
tangentwithtanner
snowghoul
jbbartram-illu
ellievsbear