Within one week each of us had the task to design a free standing “Parametric Paravent” as a room divider for a new conference room for the chair of CAAD.
The structure had to be made from CNC-produced components that connect via a simple, rule based system etc.
The jury, consisting of the CAAD and Mathias Kohler from DFAB, then selected two projects for further development. This project was one of them.
For the remaining two weeks I was happy to working intensively with Mihye An, Nikola Marincic and Aleksandar Lalovic to realise the proposed design and produce a prototype of a “Physical Paravent” .
During this period various structural and material tests were necessary to prove the design ideas and methodologies. Programming informed this process and made the final design realisable within the limited time available.
As the initial designer I would have to operate as project leader and the one responsible for coordinating the rest of the team.
Video showing progress of the last week
Using the materials properties
Given 6 mm MDF plates as the main material to use for the construction I went to test its physical properties in the wood workshop.
Using the materials properties of bending/folding was inspirational and I decided to use it as the main principle to guide the development of the design.
From the model you could start to investigate different properties such as:
- the size of the opening according to view though it and
- the angle of intersection in relation to how light is reflected on the surface
But the connection of only two plates into a component made up a closed form that was difficult to propagate further.
It was obvious from both drawings and physical test, that it was not possible to fix the bending of a plate connected to only two different plates.
So to control the shape of the MDF plate it should have at least 3 connections to other bent plates
This network of slightly bent plates allows for the forces to travel in a network without a primary structure having a direct connection to the ground.
In other words everything is connected with each other and nothing can be taken out without the structure changing.
Potential
Using material properties to guide the development of the design I wanted to give each element the desire to be a designer.
Potentially working with the rule of connectivity without defining a limited number of ways of connecting/relating parts to each other.
Meaning the topology is set free in the way that the bent shapes relate to each other.
So patterns are not predetermined but come into being as the relations evolve.
Resulting in a pattern with different properties that can be used for different purposes such as controlling structural strength/depth, light transmittance, view etc
Ultimately finding questions instead of giving solution as what the chair has set out to do
For the proposed design this is however not present, but the potential is still there to be investigated in the future.
Plan
Given the location for the room divider to enclose the conference room of the CAAD Chair I investigated the conditions that would possibly influence the design.
In general people are working on both sides of the proposed space and people will be approaching from different angles.
Together with the list of components to be considered in the right part of the image this set out the space for design
Section
An analysis of the section shows that there are two very different orientations.
One longitudinal direction where the space is perceived from a seated position
So you want to create differentiated workspaces for the individuals but without separating the two wings of the floor plan.
One the transversal direction the space is perceived from a standing position on different levels.
So you want to create connections to the surrounding.
Twisted cylinder
This leads us the idea of a differentiated shape that responds to these to principal directions.
The pictures below are early conceptual diagrams to give an idea of the intended perception.
A simple twist of a cylinder embracing the two columns turns the inside out and outside in, creating
- an entrance through a covered intermediate space
- an opening to the facade for a view
- while at the same time creating a distinctive space closing of to the workplaces around it while at the same time providing a view through.
from the approach of the building it creates
- a distinctive gesture that identifies the chair
- and from the staircase it gives approaching students and guests a sensation of something to be discovered. As you can see into the room you will also know if a lecture or meeting has started when arriving.
Shape as topology
If you instead of perceiving this shape as geometry but instead think of it as topology/connectivity it is simply a twisted cylinder.
It’s a shape you can twist, fold and bend infinitely.
Whereas the topology is fixed at the scale of the overall shape, the topology has the potential to be set free in the pattern on the surface where connectivity between neighbours potentially can change.
But for now just the pattern showed earlier.
As the overall shape is a developable surface of a cylinder it is easy to map the curves of the pattern onto the surface. This also allows for the edge of the structure to be controlled in an easy way, and make the connection to the ground straightforward.
During the second week a digital model of this twisted cylinder was investigated to be able to control the process precisely.
Link to grasshopper forum for discussion on developable surface generation.
Structural considerations
As tested with the paper strip models the shape in itself is quite strong when you apply load on top.
And if you fix one point at the ground the self load of the strip will define the second point of support at the ground.
Simplified I assume that the forces travel like they do in an arch.
Pattern for initial competition
I used repetitive patterns, in the first step to figure out how to propagate the local rule, of bent plates with three connections to other bent plates, into a larger network.
Reciprocal structures
I started to look at reciprocal structures to find a way to propagate it infinitely
I made a test of one of the proposed structures, but as you can see the curvature of the boards are not preserved even though you have 3 or more connections.
So the border condition became a field of study
Triangles
Instead of defining a “component” with connections predetermined I tried to make a basic component of three bent boards without connections.
To preserve the curvature I connected similar components in a systematic way and arrived at a pattern you could simplify to consisting of:
- triangles of different size or
- three shifted layers of the same triangular pattern.
Again the physical model showed that the border condition had to be worked out.
So to do so I made circular components to make up a closed loop for each triangle
But actually you don’t need to close up every triangle, you can choose to close up the border, or every component within the border.
This component could be of another material to introduce differentiation between bent and planar surfaces.
Ongoing material investigations
Prototype 1
As expected the scale and proportions of the initial pattern, made from cardboard in the competition, was not directly transferable to MDF, as the material to be used for the final prototype.
But to get an idea in which direction to go, the first test with MDF was made with the same proportions and way of relating parts to each other as the cardboard model.
And it was quickly realised that the 6 mm MDF was too rigid for the proposed dimensions and furthermore the way the plates intersected each other weakened the material and caused it to break at these lines, when bent into shape.
Prototype 2
Still working with the same typological pattern but now with 3 mm MDF the plates could slide into position and create the pattern, but now it was too flexible, which also caused it to break.
The intersections between the parts was still weakening the plates, but also the design of the earlier proposed discs did not work structural.
Instead this prototype evolved to use bolts and zip ties to fix the plates into position. However this was not sufficient as the zip ties only made up for the tension in the joint and not compression.
Prototype 3
To make the pattern stable and structural a new kind of joint that could withstand both compression and tension was required. Like another way of connecting crossing elements was needed, not to weaken the material too much.
This lead to a new design of the pattern, where the plates were not intersecting by crossing but rather by meeting edge to surface. Resulting in a pattern where a combination of adjoining bent plates and the newly designed joint kept each plate into its bent position creating a network of structural bent plates working together.
This design only required 3 small cut outs along the centreline of the plate, not weakening the material as the earlier proposed waffle connections.
Prototype 4+5
Now when a solution for the material system was found, the next two prototypes investigated how to use it to make up a curved surface.
Prototype 4 investigated concave and convex formations on a otherwise flat base surface
Prototype 5 applied the pattern to a double curved surface
Final prototype
The final prototype used the same basic principle as prototype 5, to apply the pattern to the proposed twisted cylinder. But the pattern was scaled to 80% and the material changed to 4 mm MDF to gain more strength and visual appearance.
Just as the joints are closed off to gain a clearer visual impression of the surface, when perceived orthogonal.
The initial idea of a double sided surfaces that turns the inside out and vice versa, was also pursued through the project. After many studies and considerations, a solution that left one side flat and the other more pointy informed the assembly process as well as it distinguished the two sides visually by only changing geometry.
Scripting
A combination of mainly Rhinoscripting with some additional use of Grasshopper have been used continuously through the project.
Grasshopper was mainly used to gain an understanding of the typology of the pattern applied and develop it further.
Whereas Rhinoscripting informed the actual process of assembly in 1:1 by scripting of intersecting plates and distinguishing layers in the structure.
In the final stages Rhinoscripting was also used to automate the preparation of drawings for the Kuka robot and naming each element for ease of assembly
Steps in Rhino
- Twisting cylinder and giving it a thickness as two surfaces
- Generating 2d pattern with 3 layers in grasshopper
- Mapping 2d pattern to twisted cylinder by “flow along surface”
- Lofting lines on two surfaces in Rhinoscripting
- Orienting the planar joints onto the surface with Rhinoscripting
- Using 3 layers to cut plates according to planar joints with Rhinoscripting
- Unrolling bent plates column by column and naming them for final assembly with Rhinoscripting
- Creating cutting paths for 6 mm drill
- Exporting to machine code through Rhinoscripting
Jesper Thøger 