09-10 Thousand Plateaus

Programming – 2D Diffusion

Methodology of the panelling system is broken down to each triangulated tile prefabricated, then brought to site. Each panel has a different variation with each drip and would be calculated through the use of computation. The substructure of the panels are also predetermined.

The drip creates a direct connection from exterior to interior as the roof is a continuous surface that merges into vertical elements and even floor.

The translation of the digital geometry to the actual is examined through prototypes of the drip in isolation of the site and any external influences.

With the use of a simple 2D diffusion the hyperbolic growth can be explored and modeled digitally to a low scale as an attempt to understand the spaces formed. The algorithm works by beginning with two grids that the user defines, next a point over the grid is determined by the user (this point is the location for the growth).

The area of growth is dependent on how many nodes will be fixed around the given fixed point chosen by the user, this is written within the script before hand. A loop then calculates and updates all other nodes according to the fixed nodes, as a result relaxing the grid. Changing the number of calculations the algorithm runs over can alter the curvature of the growth. Once the grid has completed its calculations the nodes are used to form the ribbed structure. A spline takes each point up/down and left/right. The splines are then copied downwards and joined to form a region that is then extruded. Panels are added to the nodes to add a surface to the structure. Within the algorithm panels are not added where the difference between two nodes reaches a certain value, an outcome was unexpected [show in image 00] whereby due to the nature of the liner grid the relaxation has formed a complete path along the hyperbolic growths allowing different routes to be taken over the surface. The final slide shows how vectors were added to the nodes. The base array of nodes become the agents and the upper array of nodes become the attractors for the agents. Each node is paired up with its corresponding point within the other gird and this will be its attractor. The agent is given a random direction to begin with and as the agent knows its attractor a value for the distance between both two points is taken.  New direction and movement made = heading + (direction – heading/distance) a line is traced and a new position is given for the next step in order to create a continuous line.