Continuous Spiral Pavilion Part 1 – learning from Zipshape

This is a project a while ago, collaboration by Victor Leung (me), Apple Chan and Wendy Huang in HKU.

The project is about a conceptual building method, that is by using a digital fabrication method, we are able to produce bent timber piece which could be non-identical without the use of moulds or formwork. The fabrication process is originated from Schindler and Salmeron from Zurich, they have developed an algorithm called zipshape. The process is tightly linked to the precision offered by CNC machines, to be able to cut straight timber into splines with teeth, that a matching pair would fit together like a zip, and forms a predetermined curvature due to precisely calculated teeth geometry.

The advantage of putting this technique into architectural realm, is to assist the fabrication process of non-identical curved timber pieces. Particularly to reduce the wastage of formwork needed for steam bending or glue lamination when facing non-identical forms.

The logic described in this sketch are very much owned to Schindler and Salmeron in their development of zipshape. I have a tendency to develop upon their research based on the teeth idea, but in later development, I would avoid refering the teeth to zip, as it is a bit misleading towards what the teeths are doing here. (Many people would associate the words zip to a mechanical device, which could open or close repeatedly. Which is completely not our case here) Of course, I would also want to avoid the use of the word “zip” because the term has already been well associated with the work by Schindler and Salmeron and that I would not want people to confuse.

The project starts with an attempt to replicate the cool zipshape project, making use of the laser cutter in HKU, I made a Grasshopper definition that takes in a Spline in Rhino and create the cut file automatically.

The curved wood shown in these images are straight when cut. (In case you don’t believe) We laser cut the two spline from a sheet material, for model demo purpose. In real construction where a very thick curved beam is needed, I would expect a straight timber being cut with a CNC router or water jet cutter to archive the precise teeth geometry.

Different from the timber panels being bent to curved sheets in zipshape, we have abstracted the idea to a linear element being bent. The sectional side of the piece where the teeth are shown is the most interesting feature of the piece as it shows the process of assembly with the relationship of geometry.

We did a series of testing models to verify the script’s ability to produce physical bending result, that is geometrically equal to the spline from the input. To explain a little bit if you have not read through Schindler and Salmeron’s work: The thin part of the timber is actually bending, while the teeth part is not, the desired bending amount are archived (simply speaking) when a bigger teeth is forced  into a gap with less space, the gap is forced to open with a certain amount. Thus the overall bending of the linear rib are archived through segments of smaller local bending.

The difficult part of the scripting process is to predict the physical behavior of the material when it is bending. Consider a hypothetical rectangular block (material here: MDF) being bend by force, we found out that the inner curve of the rectangular block do not reduce in length very much, but instead, the outer edge increase in length. We are not material scientist, but I would expect this behaviour to be linked to how fibers are organized in the material, and the different spring coefficient of the material when in tension or compression. This is only a very small increase or decrease in length, but we found out that this effect had a very crucial impact on the accuracy of the model, and I have programmed this factor into the Grasshopper definition.

We have also found out that the dovetail teeth is much easier to assemble than the tapered teeth because clamps are not needed when gluing. (middle piece in the above image)

We also found that the linear spline could potentially be joined by providing a key at the join of segments (above image). The key could be as much as half of the length of the segment.

Failure piece, where material expansion is not taken into account, and the proportion (thickness and width) is a bit off.

This piece utilize a more sophisticated algorithm, where the spline thickness is increased where curvature is higher (where later tests actually showed no help in the bending precision).

But this piece used a script that caters the material expansion, thus the accuracy is higher. ie: The gap is smaller in the ring. (To explain more: The curve I fed into the script is a closed curve thus I would also expect the bent piece is a closed loop. However, due to the material behaviour explained above, the actual curvature of the timber will be less then that in the calculation. Thus the gap you see, indicates the inaccuracy of the script.)

We also founded that the density of the teeth could be varied along the curve depending again on the curvature. This proved to be adding an extra level of complexity to the aesthetic of the piece, also seems to show a response to the geometry of the input and increase the chance to convince the audience that this is really high tech.

The next step we would be looking at is how this fabrication method could be applied to small scale architectural use, by designing a pavilion with it. That will be part 2. Further development to a 2-direction bent timber that is able to create a 3D curve will be in part 3.

Before I end, I must again credit Schindler and Salmeron’s work on zipshape that inspire this project.

(Updated on 2011-02-10)

Here is the very basic script about making the bending with teeth.  I’m working on sharing other scripts too, but give me some time tidying them up.



I’m happy to share the work with the Creative Common License (Attribution-ShareAlike 3.0)

One thought on “Continuous Spiral Pavilion Part 1 – learning from Zipshape

Comments are closed.