PAVILION OF WAR AND PEACEMADE IN GRASSHOPPER

TU EINDHOVEN

7M842 - FREEFORM DESIGN

0736030 - FRANTISEK POUZAR

ERASMUS STUDENT FROM CTU PRAGUE

2010

1

CONTENTS

INTRODUCTION

AUTHOR

ASSIGNMENT DECISIONS

MAIN VOLUME DESIGN

PRIMARY THOUGHTS

CHOSEN GEOMETRY

VOLUME REFINEMENT

FUNCTION AND LOCATION DECISIONS

SPATIAL REQUIREMENTS HANDLING

TERRAIN AND SURROUNDINGS

SCRIPT DESCRIPTION

CANOPY DESIGN

CHOSEN GEOMETRY

DESIGN DESCRIPTION

SCRIPT DESCRIPTION

DESIGN SCHEME

FINAL RENDERS PREVIEW

CONCLUSION

2

2

2

3

3

3

4

4

5

7

8

10

10

10

11

14

15

17

2

INTRODUCTION

AUTHOR

I was born in 1985 in Budweis in the south of the Czech Republic and am now student of architecture and building structures at Faculty of Civil Engineering at CTU Prague. In Eindhoven, I was spending one semester, studying my 5th year. Next year, I should be graduating in Prague. My field of study is mostly concerned with designing at building scale, structures and building technologies.

I‘ve been interested in scripting since grammar school, where I did some stuff in Flash using Actionscript and started running my own website. At university I met another programming by doing some GDL objects in ArchiCAD.

ASSIGNMENT DECISIONS

I encountered Grasshopper for the first time just earlier this year. We did some things with it in our studio project group. As I mentioned above, scripting has always been appealing for me and therefore I wanted to try the Grasshopper as well. Though the graphic editing was not that comfortable in the beginning and I still can’t like it fully, it makes some processes easier and more comprehensible, keeping the code also well readable to some extent.

So I decided to work solely with this plugin in Rhino without any other form editing. I want to do all the modelling parts of the assignment in it. I’m aware that such a decision wouldn’t really be that smart in practice, as some simple parts could be just easier and faster when done in e.g. Rhino itself, but for learning purposes it works well while maintaining the filesize quite low (when there is no .3dm file needed) and especially keeping all the stuff editable with simple controls during the whole process.

Due to this decision, structure of my work maybe doesn’t correspond fully with the requirements, though I’ve tried to keep it organized in the suggested manner. Obviously skipping the Rhino refinement step, which I tried to compensate in the whole model and script complexity.

For the assignment, I used the built 0.6.0059 of Grasshopper (in Rhino 4.0).

1 Outlines of the final canopy fabric

3

MAIN VOLUME DESIGN

PRIMARY THOUGHTS

I started designing almost blank-minded, not many ideas were coming for such an indefinite task, which I even didn’t want to specify with decisions for a particular location or function. First I wanted to work with some mathematics and even did some Flash interactive stuff to investigate aesthetics of the Truchet tilling [2], which I meant to use somehow on facade. The Truchet tilling is a geometry consisting of randomly rotated squares, which are diagonally divided into two triangles of different colors. When superposed, it creates interesting pattern. Another mutation of this tilling results from squares being filled with two quarter-circles placed in opposite corners, when a maze-like pattern of quite disturbing curvy aesthetics originates.

I was thinking about making the building interactive, so that the tiles would change orientation according to some stimulus. In the end, these intentions didn’t seem good, weren’t bringing anything special and I left the ideas completely.

Following thought concerned making use of minimal surfaces, which I liked already in the past. Trying to grasp the actual geometrical definition broke down soon and I were only trying to simulate the look of minimal surfaces, which proved much easier with used software. The most interesting one in this way was for me so called Shoen GW surface [3], which I found at www.susqu.edu/brakke/evolver/examples/periodic/periodic.html and which I eventually partially used for the design of a canopy [22]. See the chapter CANOPY DESIGN to read how I did it in Grasshopper. Once more it made me think about resizable caps, which would interact with neighbourhood. Obviously that would be in reality unfeasible (which I still don’t like about many proposals, even when presented as a digital architecture) and I didn’t follow that further than to a fact that in some area, the caps are bigger than in another.

It’s not a secret that the canopy, or the complete scripted object mappable on any surface, was made even before the main pavilion volume and therefore influenced further designing in many ways.

At this point, freedom got restricted, as I needed to come up with the main volume which would be coherent with the proposed canopy, that I didn’t want to discard anymore.

CHOSEN GEOMETRY

I’ve spent many moments in Grasshopper searching for some interesting elements that might give me an idea what to do for a form of the pavilion, as for learning purposes it is easier to obey direct capabilities of the software than to try doing something what is hardly possible, moreover when I didn’t have any strong ideas.

After some experimenting with various forms which are not worth mentioning, I discovered the tool Facet Dome [4], that creates a facetted geometry based on a set of points. Given points are moved onto a sphere (dome, shown in the image on the left) and then tangential planes are assigned to each point. Inner intersections of these planes create polyline outlines of faces (facets). Realizing this principle helped me in following steps when I was implementing the given spatial needs.

The Facet Dome tool can be controlled also with radius of facets which might produce only circles of given size on each face. After some experiments I didn’t make use of this feature. Another quite

2 Truchet tilling experiments

3 Shoen GW minimal surface (images from www.susqu.edu)

4 Facet Dome tool in Grasshopper

4

important part of this element is the clipping box, which restricts the maximal size of the resulting geometry by simply set box-like borders. This particular feature was quite helpful. Have a look at the sub-chapter SCRIPT DESCRIPTION to read about some issues that occurred.

VOLUME REFINEMENT

Thinking about the previously proposed canopy, it took also some time to design the facade of the volume and actually many more or less different variants have been produced. I was sure about working with black and white materials once more. Especially the white fabric (e.g. some sort of ETFE foil) was an element I wanted to keep also for the main volume, because structures like this can in some scenario diffuse light and therefore are suitable for exhibition spaces.

After some considerations I came with idea of using the fabric to connect offset adjacent solid faces, what creates a sort of inverted structure effect [5, on the left], as one would usually assume the structure being on the edges. Though it doesn’t sound any smart, I didn’t feel any obstacles in doing so as the load-bearing structure doesn’t have to be placed so expectably and it wouldn’t bring many advantages when put into the edges in this case.

I like its appearance as it even slightly resembles cubist proposals [6], that are documenting the most unique Czech design and architecture era in the history which was remotely related to the aesthetics of well known cubist paintings. Though hereby I don’t dare claim relating to design history, just that the Czech cubist aesthetics might have influenced my taste and invention.

Another idea was closely related to the aesthetics of the canopy, spanning the fabric from previous idea from edges up to circle-shaped caps, which were centered on the faces and also offset. The direction of the offset was giving quite different impression and on the renders [5], both the normal vector and the center volume vector offsetting is depicted. The form in the middle of the picture is using the center volume vector, that is defined by the volumetric center-point of the facetted geometry and face centroid. Its definition in Grasshopper was based on BRep Area and BRep Volume elements and I’ve thought it looked better than the other, offset perpendicularly by normal vectors.

Especially in case of the variants with circular caps, I didn’t like the sharp volume corners. Quite complicated script has been developed to smoothen these projections. It’s elements and principle are described in sub-chapter SCRIPT DESCRIPTION.

Then, I was exploring some possibilities how to put these designs together or how to create a transition between these two and the canopy.

FUNCTION AND LOCATION DECISIONS

During development of the script, I determined purpose of the building for the first time. Once more, I’m aware it’s not the best way, it doesn’t work like that in real design processes, but here I wanted to keep some freedom as far as possible, so the definitions wouldn’t restrict the approach. Therefore the only requirements were concerned with the spatial needs stated on the course website.

Another decision concerned the building site, which I also preferred to keep undetermined in the beginning, though that disabled proper determining later on, as the building is obviously not designed to fit into some environment, it doesn’t really reflect its neighbourhood (‘just the glossy surfaces on the facade do’). Therefore it’s actually designed in the same way as some pavilions on world exhibitions,

5 Renders of some volume variants

5

where the architects don’t know the neighbouring objects neither.

Along with other aspects being held very general, I wanted to assign the building to some global problems. At some moment, while generating the main volume, form of a defensive hand grenade occurred to me and I decided to make the pavilion with some (anti-)militant content. Although the resemblance seems to be obsolete now, it influenced the rest of the designing significantly. So eventually, the design can be called “Pavilion of War and Peace“ and it might be most suitably located in a desert as shown in final renders, though the location doesn’t really matter, it’s definitely not a site-oriented architecture (I don’t like to say this).

The chosen function of pavilion is after all of course not solely based on the hand grenade resemblance but in following process, it has been quite intensively deliberated and hopefully more hints are obvious. The overall forceful contrast of the pavilion within itself as well as with the surroundings can also invoke the drama involved in the topic of the exhibition. Elements like terrain ‘impacted‘ by the object, disturbingly ruffled columns of the canopy not suggesting any way to the entrance or rugged volume of the whole building should impress in a discomforting yet interesting way, which is in my opinion coherent with the introduced topic. If that were not enough, the name could be actually understood as an honour for Tolstoy as well.

SPATIAL REQUIREMENTS HANDLING

I wanted to put the spatial requirements into the very core of the script, so it could influence the volume precisely. As well as in many other cases I eventually preferred to use the DotNET VB Script element to handle this data (I preferred coding in VB to C# because the tutorials in Grasshopper Primer don’t include the latter one). I believe that decision to include these classic scripts into the otherwise solely graphic interface of Grasshopper helped me in generating the form a lot as it was easier and more elegant solution in quite many cases thanks to my preceding knowledge of some scripting languages.

It seemed to me that the resulting form is quite interesting whatever points are used for the facetted geometry generation. But of course it’s all liable to some aesthetics laws and some refinement better be done, mostly with parameter adjustment and code variables editing.

Anyway, some experiments concerned also the volume division. I separated the form into three blocks, every of which included two spaces given in requirements. Blocks documented in following pictures were usually connected to form a golden angle (137.5°), albeit it didn’t help much in achieving a balanced composition. All of the following depictions are just exemplary, with properties like offset-depth or plate-size fully adjustable and were not much refined to look appealing.

One of the more complicated issues was the inter-block connection. Even after succeeding in selecting the right faces to be connected, the articulation of whole form was too evident and the ‘necks‘ didn’t look good. I’ve tried composing this form out of both previously introduced surfaces. The polygonal shape didn’t look that good this time [7], especially because of the strange connections, which I didn’t manage to make any smoother or more elegant. On the other hand, such a further composing worked quite well with the circular-caps-surface, where I kept the circular shape for the connecting tunnels as well [8].

Some experiments concerned also the colors and possible materiality of the object, though eventually I decided to keep the formerly chosen combination of white fabric and dark reflective solid plates. The inverted materiality is depicted with a screenshot [8].

7 One of the first volume division experiments

8 Volume division experiments with circular caps

6 Cubist box by Pavel Janák, 1911 and cubist box proposal by Vlastislav Hofman, 1914 (images from Czech book: ANDĚL, J.: Český kubismus 1909-1925. Praha: i3 CZ: Modernista, 2006)

6

As mentioned in previous sub-chapter, I was thinking about making a transition between the polygonal and circular caps [9], here in the way that the first and the last blocks were those extremes and the middle one tried to take something from both adjacent ones, though it’s not much obvious in the picture. The overall look resembled some disturbing insect annoyingly and that made me try something else.

From the linear connection of three blocks, I got the idea to connect also the both outer blocks to create a circle (triangle) [10], which could efficiently kill the insect resemblance. Though I didn’t manage to script that properly, probably also because of not being well motivated after the low quality of preceding results. Once more, different surface structures were meant to be used.

I believe the illustrations argue well for the reasons why I eventually decided not to follow such an articulated form. From now on, I wanted to keep the volume rather simple and united than literally split into the functional spaces, which though would seem quite objective expression of the requirements but wouldn’t really work well at this functional scale as far as I realized. I imagined the pavilion like a volume being divided inside while remaining compact when viewed from outside, therefore rather classic approach. Such a volume shouldn’t rather be just a simple blob as the depth of its form wouldn’t allow any reasonable functional use, moreover when exhibition spaces need quite a lot (day)light. An elongated form followed these thoughts [11]. It was meant to be divided by perpendicular partitions into needed spaces. The model was actually just an one-axis-scaled block from previous Grasshopper definitions with some more plates.

Fortunately after that, I got another idea - to do the previously invented triangle formation by introducing an atrium in the middle of a compact form [12], which was then both aesthetically and functionally enough convincing to keep it for the final form. I achieved the atrium space by creating second facetted geometry inside the first one and scaling it vertically. Then the atrium was cut by simple boolean difference. To get some better proportions, height of the whole object was decreased by another vertical scaling.

Outer volume was covered in structure with the polygonal plates and the walls of atrium got the circular cap surface. Such a decision seemed to me quite reasonable, not only to keep both of the designed structures in use but also because of possible demands for light inside an exhibition space, which can get more diffused light from an atrium than from outside.

Still an important challenge remained - to divide the form into the functional spaces of given areas. The DotNET VB Script element proved really advantageous when using this data, as I defined the points for facetted geometry generation in the way, that resulting facets were embracing various volumes in each corner, according to the given requirements. These points would make just an object consisting of solely vertical walls and therefore few more pseudo-random points were added to create the complex shape with variously oriented faces. The random points were kept in a range that doesn’t spoil or significantly influence the main facets defining the required rooms.

Physical division of the form into the functional spaces of given areas was naturally done by concentric partitions, which were oriented from the object centre to the points defining the outer facetted geometry, so actually perpendicularly to the main facets [13]. When finally working with actual required areas, I decided to come up also with division into floors [13], so each functional space was accompanied with an integer determining number of slabs. Eventually all of the rooms were divided into two floors, except for the ‘Entrance, Restroom and Cloakroom’, which got only one. Anyway, all of these controls were kept in working state in the Grasshopper script and can be still edited if needed.

12

10

9

11

Variants of volume division.

In chronological order, with 12

being the final one.

7

During whole process, atrium form was being taken into account and it’s in quite complicated way subtracted from the needed volumes, though this step doesn’t bring much accuracy. Two-step area check has been included in the script to show that the resulting slabs have proper area. Though after introducing more floors, the areas decreased, because the slabs are cropped by walls which are not more purely vertical on that level [13] (formerly the slabs were on 0-level where there are only the main facets giving shape). I decided not to operate further with this issue, as the volume, which is not checked, is definitely sufficient anyway.

The partitions and slabs and therefore also the actual volume division is in the end not that much visible in the final renders [32,33], but can be easily viewed with the Grasshopper script. Not revealing this was part of the intention described above, I wanted to keep the volume simple and compact.

Although any load-bearing structure hasn’t been devised properly, it would rather be based on the structure of these partitions and slabs than on an outer structural shell. Depending on materiality, some more elements would apparently have to be added. For instance especially in the most projecting corner [13, on the right side]. Most of the partitions reach on the ground while tapering themselves, which might serve as quite classic columns connected to foundations.

TERRAIN AND SURROUNDINGS

When the main form was done, some parterre considerations have been made. Quite soon, I got the idea of slightly submerged volume surrounded by a ditch and therefore even no variants can be shown here. The idea refers to some military related symbols, like a crater made by a bomb, battlefield trenches or even an old castle / fortress design. Although any of those relations is sufficiently obvious, some dramatic and conflict suggesting impression is still achieved in my opinion.

The only experiments here concerned depth of the ditch and amount of water in it. Decision not to put any much water seemed to be reasonable. It offers some reflections above all. The ground shaping is assumed to be made with some solid material, most probably concrete or large plates of stone. Some finish should rather be done on the surface to deal better with effects of water, though the surface was not supposed to be glossy. The proposed building is not meant to comply with natural processes attacking the water surroundings and should be kept in clean and technical look.

I supplemented the form with an entrance platform, which spans over the ditch in front of the rooms dedicated for ‘Entrance, Restroom, Cloakroom’ and ‘Restaurant, Kitchen, Store, Buffet’. Entrance obviously needed to be made like this and extension of the slab in front of the restaurant was meant to serve as a terrace. Whole platform is covered with the previously designed canopy including its slender ruffled columns and circular caps of various size. Now, when entering the object, visitors can see the same structure above himself in front of the entrance and then in atrium, which is accessible on a protrusion of the entrance platform.

Base geometry for the canopy has been generated with some sinus-like deformations, to let the columns ruffle because they are perpendicular to the surface at their tip. The columns stab the entrance slab and reach the ground and hence might be structural support for both the canopy and the platform.

As mentioned in FUNCTION AND LOCATION DECISIONS, I didn’t come up with a complex location and therefore no actual connections of the object on existing or assumed roads or walkways have been made. The parterre is solved just in immediate surroundings, also to keep the object more or less fitting into almost any site.

13 Partitions to divide the volume into required rooms and slabs made inside the object

8

SCRIPT DESCRIPTION

I’ve tried to maintain the script in well readable condition during whole process, as even for me as the author it is sometimes hard to realize performance of the code I did few days ago. So there is the cyan color indicating control and settings elements, magenta for resulting final geometries (divided into solid preview and geometry to be baked), yellow for some informative generation check and default blue for grouping of functional parts of the code. The arrangement is kept also for the layout, where the controls are on the left and results on the right. Receivers were sometimes used instead of classic wires, especially when long, to decrease the involution.

The upper part of my Grasshopper script is dedicated to the main volume design [14]. It consists of 206 elements, though the number is kept fairly low thanks to the DotNET VB Script element mentioned already on previous pages, which helped me in achieving quite complicated tasks within single objects. However probably still a lot of optimization could have been done.

In this report, I don’t want to explain the whole script because I believe that most of its parts are quite well understandable to an experienced Grasshopper user even without description.

The control panel for this part consists of the given spatial requirements, numbers of floors, atrium size, properties of the surface as well as some terrain settings. I decided to use only the spatial requirements expressed with floor area.

Yellow panels are placed next to the controls and show preliminary results of floor areas to assure user about proper handling of provided inputs. The last yellow panel shows the size of volume corner filleting, the larger it is the smoother form should be generated. But the number can’t be simply increased as it depends on too many numbers, even the pseudo-random ones inside the code.

The largest and also the most important is the ‘tg points generation, central script‘ element, where the given numbers are proceeded and points defining further geometry are generated. Outputs for the polar points coordinates of both facetted surfaces as well as for partitions, slabs and some additional usage are calculated. I didn’t think that pasting the 200 lines code into this report would be helpful.

The base facetted geometry [15] originates when points coordinates are grouped into points with Point Polar element and then plugged into the Facet Dome, which is also fed by a Box 2Pt element defining its borders. Solid Intersection puts the box and facetted geometry together, as otherwise there would be holes in places where the box border is reached. As mentioned before, one-axis scaling then transforms the proportions and the atrium is cut from the outer volume with Solid Difference.

Facetted surface is afterwards split into single faces and its outlines help determining the maximum possible filleting distance. For future corner filleting, all the curves have to have the same blunt corners and therefore half of the shortest of all edges is taken as the distance into Fillet Distance thanks to Sort List, List Item and Multiplication element as shown on [16].

Obtained rounded outlines are in next step exploded into their component parts - lines and arcs [17]. Only arcs are selected with a DotNET VB Script, which simply checks the string form of the list for “Arc-like Curve” and puts a boolean list into Cull Pattern element which lets through only the arcs. Then with Evaluate Length, the midpoints of arcs are obtained and through Closest Point neighbouring arcs are determined. Now, there are always three arcs creating a closed curve due to the non-symmetric facetted geometry in each volume corner. Selection of proper arcs is ensured by a series of those Closest Point, Cull Index and List Item elements. These threesomes of arcs are then put into another

14 Preview of the script that is dedicated to main volume generation

15 Facetted surface generation

16 Filleting of facetted geometry curves

17 Filleting of facetted geometry corners

9

DotNET VB Script which simply creates a ‘nurbs surface’ out of its outlines. This element is then amply used in the canopy design.

Process of offsetting of the polygonal plates of outer volume is shown on [18]. The facetted geometry is exploded into individual faces and their outlines are further used to define planes, at which the faces are scaled, and normal vectors, along which the scaled faces are moved (offset). I decided to use Move rather than surface Offset because the script was eventually more simple due to some issues occurring by Offset caused by normals of the surface. Meanwhile, scaled faces are filleted with Fillet Distance element, where the distance is set from the control panel. Normal vectors are multiplied by a number from the control panel as well to adjust the offsetting distance. Same set of planes and vectors serves also for the atrium surface generation - for circular caps. Those are made with Circle CNR from normal vector and radius determined with Curve CP (closest point on outline curve to the center point gained with Brep Area element) and then multiplied by given number expressing percentage of maximal size. Circles are moved afterwards in the same manner as the polygonal plates. Actually both structures are generated for whole surface of the facetted geometry and just then separated for subsequent use.

The division into atrium and outer surface is made with few other elements [19]. Where the single atrium facetted geometry is exploded and its faces are evaluated for normal vectors, x-coordinates of which are then compared with the normal vectors of already generated surface. It is done inside another DotNET VB Script element called ‘curves for lofts‘, which in addition orders the available curves into pairs for further lofting and also selects curves that then simply form Planar Srf of dark solid plates and caps. Only x-coordinates were used, because I didn’t manage to properly compare the vectors themselves and one coordinate proved being sufficient for flawless determination.

Partitions dividing the volume into required rooms are created from few outputs of the ‘tg points generation, central script’ by feeding elements of Plane and then Plane Srf [20 up]. The planes are extruded along the normal vector multiplied in vector Multiply element by set wall thickness (same as slab thickness). Another vector (half of the unit vector) then moves the walls in the way to set them centred on the dividing plane. The partitions are in the end cut by the shell of facetted surface so they don’t project out of the building.

Another part of the script is the generation of slabs [20 under]. It takes some outputs from the main VB script to create a 4Point Surface which is the actual slab plane already. The plane is extruded afterwards by Unit Z vector multiplied by given numbers from control panel. The resulting slabs are eventually also cropped by the facetted geometry with Solid Intersection function. These solids proceed then through a series of list editing and calculations which result in the second area-check panel.

Three elements are used for the preview settings of each part of the resulting structure [21]. Not only to separate it for baking (exporting the geometry into Rhino) but to set different colors for each object and therefore make the perception of the design easier. The geometry of given part is put into a renamed Brep which leads into a Custom Preview element, color of which is set by ARGB. The Brep is meant to make the baking process easier and faster, the Custom Preview just displays the geometry in better way than the standard preview.

19 Division into atrium and outer surface

21 Preview elements

20 Generation of partitions and slabs

18 Geometry of offset polygonal solid plates

10

CANOPY DESIGN

CHOSEN GEOMETRY

I decided to elaborate the canopy as a detail needed for the 3rd part of the assignment. This part of proposal actually preceded the main volume design. I’ve tried to describe my current designing attitude in the first chapter, I didn’t follow many ideas from the beginning, it was rather just playing around. Therefore when the resulting structure seemed to be obviously suitable for use as a canopy or some sort of roof or facade, I’ve pursued the project in this way. Though it was clear that single canopy can’t make whole pavilion.

As already mentioned before, my first intention was to use some mathematics in developing the model, after few unsuccessful experiments, I’ve gained some inspiration from minimal surfaces. Especially so called Shoen GW surface [3] has been appealing for me and I tried to make a geometry in Grasshopper that would at least resemble it. At first I wanted to make the geometry completely looking like the model, but then I couldn’t come up with reasonable structure that would need both shells of the mentioned minimal surface and therefore only one side has been kept, though addition of the other wouldn’t have been much more complicated.

Coming up with final version of the canopy was quite long process as I were searching for ways how to do something all the time.

Few following experiments concerned development of some supporting system. At first, I imagined the structure being directly stretched on a space frame. Its script wasn’t any complicated as I’ve done similar one in Grasshopper already before, though I still can’t say I know Grasshopper well now. I were trying to map it on a vault-like surface [23] but the results weren’t much convincing and on a large flat surface neither. So the investigations went on and I got the idea with many slender columns, one for each cap, which afterwards became ruffled [22].

The canopy shown here was mapped on a surface defined by series of varying sinus-like curves that were lofted and so a sort of waved surface originated. It was necessary to make every next curve different, so the columns would ruffle quite randomly. I eventually fairly liked this design, though neither the idea of ruffled columns nor the one of Shoen GW surface inspired canopy is original as I later discovered. Little different canopy, but still following the idea of high and low columns was part of an installation during events held in Hiroshima by recent WWII attack anniversary.

DESIGN DESCRIPTION

The white fabric stretched between the circular caps is supposed to be some plastic- or paper-based foil, which should let quite a lot of light through. The light diffusion capability appeared good in relation to exhibition spaces from the very beginning. The fabric is also supposed to be able to support somewhat structurally the columns when providing mutual tension between them.

The columns are designed in dark black color, preferably not glossy. Surface of natural carbon fiber tubes would be one of the favorite also what concerns the structural abilities. The shown diameter of columns corresponds to 50 mm. Their height is obviously various, but the canopy is held in minimal height of about 2.5 m above ground. Columns were supposed to be attached to some unspecified sort of simple foundation, but after implementing them into the final design along with the main volume,

22 The designed canopy structure

11

they are piercing the entrance slab and founded in some concrete below.

Caps are circular, presumably made of the same material and with the same finish as the columns. Some variants were investigating shape of the caps, though eventually simple planar one seemed best. When using the carbon fibre, columns could be probably advantageously manufactured with caps in one piece.

Around the caps, there was meant to be a rim-like belt, which would help fixing the fabric onto the caps. It’s shown in Grasshopper under the name ‘frame‘.

Some ideas primarily enriched the design of the canopy, but when introducing it to the whole project, ideas were slightly changed as they concerned rather separate use of the canopy.

To make the design more sustainable-looking, a grass field was proposed beneath the canopy. There would be probably sufficient amount of light under a properly chosen fabric and the water would be brought through the columns into some layer of soil included in the base of the canopy. Rainwater is naturally led into the lower parts of the fabric, where the caps work like inlets, being also cone-shaped (unlike the aesthetic decision mentioned above) and are taking the water into hollow columns and therefore could water the grass. Effect of this design would be augmented when used on a place where there is no grass, like on city squares.

Evidently, there is no grass field in front of the entrance of the final building, but water draining capability has remained as the water is led through shorter hollow columns into the ditch. Also the cone-shaped lower caps are visible in final renders [32, 33].

Another idea concerns providing light during evenings. The upper caps, not used for water draining, are provided with PV panels and some electricity accumulation system is also included. The bottom side of these caps incorporates e.g. some LEDs which start shining by nightfall. The shape of adjacent fabric serves as light reflector and refractor and can prevent dazzling unlike when used for the lower caps. Another variant was to substitute this system with a luminiscent material which would accumulate some heat energy during the day and release it then in form of light. So in the end, all of the caps have some special usage.

SCRIPT DESCRIPTION

This part of the script consists of 313 functional elements, though during development of the geometry I used more than 600 of them before realizing possibilities how to reduce it significantly. I’ve definitely learned a lot thanks to this assignment. Mostly problematic for me was handling of lists and trees of data which caused some most time-consuming issues.

As usually, the script starts with the cyan control panel on the left, where UV surface division, diameter of caps, thicknesses and heights are set.

Generation of base surface of the canopy is part of the previous part of the script - for main volume, it’s called ‘umbrella’ in there. This geometry is taken with a data receiver ‘surface to be mapped‘ and proceeds through surface division containing Divide Domain2, Isotrim and Brep Components elements.

Vertices of the exploded geometry are then sorted in series of quite complicated steps [25], because the geometry of Shoen GW surface is based on specific points in triangular grid. At first, with List Item the single subsequent corners are obtained which then undergo further selection. There are

23 One of the unsuccessful structural variants

24 Preview of the script that is dedicated to canopy generation

12

more ways how to do it, I eventually decided to use DotNET VB Script to create a boolean pattern for culling of the lists of points with Cull Pattern element. There is one VB script for sorting in rows (U) and then another one for doing so in lines (V). Through this process, none of redundant points can pass, including the ones that are discarded when defining the triangular grid on the former quadrangular one. With this complicated sorting, branched structure is obtained, which can be put into three sorts of points (upper, middle and lower points of canopy) and two groups (odd and even parts of the fabric), which go separated into the next step where the geometry is actually generated.

The further definition became quite complicated and hard to understand, it creates fabric of the canopy [26]. There are many vectors, planes and additional points defined to achieve the look resembling the minimal surface. Two of these scripts are made, one for even [26, light surface] and second for odd [26, dark surface] parts of the fabric. I didn’t manage to unite these, though the scripts are almost identical, they vary just in some vector directions.

At first, there are the vectors between given points obtained with Vector 2Pt and also normal vectors of the surface for each point are gained with Surface CP and Evaluate Surface tools. Then some points are moved in proper direction along a normal vector - to create the lower and upper tips of the surface, where there the caps are placed later. For this, a function of three variables F3 worked well, while taking coordinates of former point and adding it to dimensions of normal vector multiplied by depth of the canopy set in control panel.

Attractor-like behaviour is defined in group of elements called ‘attractor behaviour‘ [27], where function of three variables calculates the proper size of caps according to their distance to some attracting points and two extreme values that are set in control panel. After some considerations and variants, the formula looks like y+z*x/35 where x stands for the distance from attractor point, y for minimal cap radius and z is the maximal cap radius. I avoided using another variable for the denominator, the number 35 stands for the approximately largest distance in the set of evaluated points. Eventually two of the base surface vertices are used as attractors, their definition is right next from the definition of the surface in the upper part of the whole script.

The tip points then undergo another translation, thus in a way perpendicular to the normal vector of surface to place themselves on the edges of future caps [26, points 0, 3]. Definition of this direction is fairly complex and includes elements of Plane, Plane Normal, Plane / Plane and vector Multiplication. The first plane defined by vector between 2 base points and the normal vector is intersected with plane defined only by normal vector and the resulting line is used as translation vector after being multiplied by adequate radius of cap from the ‘attractor behaviour’. Then the points are moved with simple Move tool. This procedure is repeated four times - two times for lower tip, two times for upper tip as the geometry part is triangular.

The third point of this triangle is still unmoved [26, point 1], being the midpoint. Further on, there are some vectors around this point defined to bring some smoothness between individual parts of the canopy. The process looks the same as in previous paragraph, just the point is not moved, only the vectors in normal plane are set.

Similar calculations are made for the point opposite in triangle to the middle point - between the two tip points [26, point 2]. Groups of elements called ‘average direction in the middle‘ and ‘proper vector for the middle‘ were made for this. They proceed some existing hints from other vectors to create its own proper normal plane, which should be the same for the neighbouring part of the fabric.

26 Scheme of the canopy fabric geometry (light surface is even part, dark one is odd part of the fabric)

25 Half of the points selection script (for even fabric parts)

13

Output of all previous steps are vectors. Some of them are reversed and another ones are obtained with Cross Product tool. All of these vectors specify start- and end-directions of curves, which are hereby feeding Bezier Span tools [1]. Each triangle is actually split into two triangles, because the NURBS surface that is created afterwards can’t have more than 4 outline curves. Whole triangle would otherwise have 5 curves [26, curves 0-1, 1-0’, 0’-3’, 3’-3, 3-0], as on both lower and upper tips, there are arcs of the caps present.

The DotNET VB Script called ‘nurbs surface’ [28] was already mentioned in previous chapter. It executes the 4 given curves [26, curves 0-1, 1-2, 2-3, 3-0] with GetNurbForm and RhUtil.RhinoCreateEdgeSrf commands and outputs a NURBS surface. Two of these scripts are used, one for each sub-triangle [26, both upper and lower surface].

When the described script is repeated once more for odd fabric parts, generation of additional elements of the canopy can proceed [30]. Out of the arc-like curves, surrounding the caps in fabric, the actual caps could be made with simple Join Curves and Planar Srf if there weren’t some issues further on, presumably caused by imperfections in individual curves linking - not all circles behaved fully closed. Moreover the caps on edges of the canopy wouldn’t be circular without another introduced script. Therefore the circular curves are analyzed to obtain normal planes and radii, which helps also in generating columns, cones and frames around caps afterwards.

Obtaining the circle parameters was achieved in quite unusual way [29], by getting two points on each curve with Evaluate Length tool (at 0 and 0.66 of length) and creating Plane Normal perpendicular to the curve at given point. The two planes then produced an intersection line with Plane / Plane tool and the result was put into Curve CP to determine radius (as the shortest distance between the line and start point of the curve.

The actual caps as well as columns have been done with Circle fed by acquired parameters (in case of columns the radius was set from control panel), Planar Srf and Extrude. Another issue concerns the normals of these planar surfaces as the direction of extrusion has to be set properly and as it differs for caps and columns, so pairs of extrusion directions have been made. Moreover the former normal directions of a set of caps didn’t match (according to the character of curve which were the parameters obtained from) and therefore a Cull Pattern was introduced, which was getting the boolean pattern from a Closed element determining the character of the base curve. Its inverted output from Gate Nor tool lead into another Cull Pattern element. The Dispatch element could have been used with the same result and whole script could have been little simplified as well.

The definition of caps could be also done directly from the selected points and ‘attractor behaviour’, but the way I used surprisingly proved less number-of-elements-demanding.

Same set of data has been used to generate so called ‘frames’ around the caps that are supposed to work for the fabric attachment [30]. Circles defining the caps were put into the Pipe element to create the needed rim, diameter of which is the same as the thickness of caps set in control panel.

One of the least important parts of the design are cones (described as water inlets above) [30]. They were made with simple Cone tool fed by similar data as the caps while getting their height as a half of the base circle radius. Cones were done for both lower and upper caps, though the upper ones are disabled in the end.

30 Canopy structure close-up with visible fabric, columns, cones as well as the rim around caps

29 Obtaining parameters of unclosed curves of caps28 Script for the NURBS surface generation

27 Attractor behaviour script for each fabric part

14

DESIGN SCHEME

FABRIC OF THE MAIN OBJECT

stretched between solid elementsM: ETFE foil or paper-based foil

FABRIC OF THE CANOPY

stretched between circular capsM: ETFE foil or paper-base foil

POLYGONAL AND CIRCULAR CAPS OF THE MAIN OBJECT

solid elements witch structural performance, attached to supporting structureM: carbon fiber, metal, dimming glass or glass

CAPS OF THE CANOPY

consist of caps, cones and rims for fabric attachmentability to shine during evenings from the bottom side, PV panels on the top, water collectingM: carbon fiber or metal, PV panels, LEDs

SUPPORTING STRUCTURE, WALLS, PARTITIONS

partitions divide the volume into required roomson the lower side are the walls tapered and in form of columns connected to the foundationsM: concrete, metal or carbon fiber

COLUMNS OF THE CANOPY

pseudo-randomly ruffled to create dramatic impression of the entrancecolumns supporting the lower caps are hollow, to drain the collected waterM: carbon fiber or metal

ENTRANCE PLATFORM

connected to the supporting structure, spanning from the ground to the object over the ditchM: concrete, metal or carbon fiber

DITCH AND TERRAIN

filled with water to some extentsupports symbolically the military function and helps in HVAC systemslocation undeterminedM: concrete or large stone plates

31

15

FFIIINNNAAAALLLL RRREEENNDDDDEEERRRSSS PPRRREEEVVVVIIIEEEEEWWWWWWWW

32

16

33

17

CONCLUSION

I’ve taken this assignment quite seriously. Not only because it is my last one in this semester but I had a lot of time to work on it and learned therefore a lot in Grasshopper, what was quite pleasant experience. The Freeform Design little substitutes my studio project this summer as I wanted to practise designing of complex objects, what wasn’t actually aim of my master project in Eindhoven. Though as you can read on previous lines, my approach wasn’t really architectural, I’ve rather played with the software and followed interesting ideas which rather originated by themselves.

Grasshopper proved to be an impressive tool offering many unknown designing opportunities, but I wasn’t impressed any further as the code becomes too unintelligible when it comes to some complex composed geometries, there I would prefer some classic written code. Though apparently that’s not what is Grasshopper made for and I’m aware of absurdity of my approach.

Although I liked the visuals of some scripted architecture designs, I wasn’t big fan of digital architecture, because many of such proposals seemed like humble experimenting with available technologies, where the form doesn’t follow function that much (unlike usually stated in project description) and the designs often give impression of certain unfeasibility as the authors have just made maximally unexpected forms.

But generative design can be obviously really valuable tool and especially the possibilities concerning manufacturing of some scripted structural elements are something I’d like to give some more attention in the future.

PAVILION OF WAR AND PEACE

MADE IN GRASSHOPPER

University: TU Eindhoven

Part of assignment for: 7M842 - Freeform design

Author: Frantisek Pouzar

TUe student number: 0736030

Erasmus exchange student from: CTU Prague

Contact email: frantisek.pouzar@gmail.com

Homepage: http://dasti.net

Date: August, 2010