Düzce Teknopark´s parametrics

 

For the initial renderings of the freeform DÜZCE TEKNOPARK, the main goal was to quickly achieve a Grasshopper definition that divides the base surface into a number of separate quads with full, interactive control over the quad’s size and aspect ratio. As we were dealing with an untrimmed surface, we initially used the UV values of the surface to panelize the surface. However, the surface contained both areas of high curvature as well as rather flat areas, especially towards the sides – therefore, a regular pattern of UV curves led to an irregular panelization of the surface. We mostly circumvented this problem by parametrically defining two “master curves” in U and V direction, which were then divided into equally spaced segments. The resulting curve parameters were then again mapped onto the surface, resulting a relatively uniform spacing, especially in areas close to the intersection of the two master curves. For the rendering, the master curves were dynamically adjusted to always face the chosen camera position.

In addition to the panelization we also introduced functions that analyzed the uniformity of the panels as well as their planarity. This quickly proved that a solution found for rendering is in no way suitable for fabrication. Not surprisingly, the main issues were:

- UV panelization always results in highly variable panel sizes, as e.g. the outer radius of a torus gets divided into the same amount of points as the inner radius of a torus. And yes, we do see the similarity between a donut and the TEKNOPARK.

- The UV curves that would later make up the primary structure of the TEKNOPARK weren’t planar curves, but twisting in space.

In the few weeks during our 2011 summer workshop in Bodrum we therefore focused on developing a new system that allows us to create relatively uniformly sized quads and a planar primary structure. As you can see on the two screenshots comparing the “visualization” definition with the “production” definition, implementing these constraints is not a minor task. What we have now is a highly customized definition that provides about 15 single- and multidimensional sliders to finetune the panelization of the TEKNOPARK. It is by far not a fire-and-forget solution where you enter a geometry and you receive a panelization, but whenever someone decides to nibble at a corner of the base surface, you don’t have to start from scratch, but just redefine the surface and tweak the sliders until you reach an aesthetic and efficient solution.

Parametric modelling is described as a tool for non-regular systems. What we are doing now, is using parametric modelling to make a system regular – or as regular as possible – again. But even with the most efficient tools and brightest minds, it will not be possible to reduce the complicated base surface to a regular system with e.g. four different panel sizes. While we aren’t there yet, we do know that this will highly complicate the final steps of the project, fabrication and assembly.

Computer Aided Manufacturing (CAM) is of course the key to fabricating these systems, as – in theory – it takes the same time to e.g. mill a hundred individual wooden beams as it takes to mill hundred identical beams. However, in practice there are various workflow problems involving the batch processing of geometries. When using Photoshop, the user can perform the same set of recorded actions to a whole folder of images, e.g. auto-adjusting levels, switching to black and white, adding a watermark and then outputting both a small JPEG and a high-resolution TIFF. Using generic CAM software, especially for geometries that were created outside the software – as is often the case in architecture – this batch processing just isn’t possible at the moment.

Finding solutions for these problems is one of our research topics over at the Association for Robots in Architecture, where we have introduced a plugin that integrates the NC-code generation for robots into the parametric model. At the Düzce TEKNOPARK, we aren’t yet far enough that we would have to think about such things – but it feels good knowing that at least in theory we should be able to solve these upcoming problems.