*Process-based investigation developed and prototyped for MIT instructor Stanford Anderson*

*Download full research report here*

**DIGITAL DESIGN AND FABRICATION PROCESS**

For this project, I used a 2 1⁄2 axis CNC mill with a 4’ by 8’ bed and 2.5” vertical capacity. This machine can cut vector profiles and carve rela- tively shallow three-dimensional geometries; the machine manipu- lates a gantry that controls a spinning “endmill” (a special drill bit) which spins at high rates as the gantry moves it across the surface of the stock material. This machine has a steep learning curve, not due to its mechanical complexity but to its powerful interaction with the stock material it cuts; a keen sense of materials makes using the machine much less difficult (and dangerous).

The first step of any digital fabrication is the creation of a three- dimensional digital model of the geometry in question. For this proj- ect, the digital model had to be derived from photographs of the orna- ment, and is therefore fundamentally imperfect. This process is somewhat akin to the process by which craftsmen translated the Sullivan studio’s drawings into construction documents and, ultimately, the 3D relief ornament.

The process is very intensive, and requires a keen sense of modeling software and a spatial understanding capable of interpreting two- dimensional photographs. It proceeded in phases:

1. I digitally traced the head-on photograph of the ornament (found in an archival digital image collection of the University of Chicago). I used AutoCAD to draw vector lines to outline the myriad forms of the ornament’s pattern. Fortunately, this particular scheme is symmetrical in two dimensions, so only one-fourth of the fundamental unit had to be drawn.

2. This pattern was mirrored across the lines of symmetry.

3. The components then had to be connected to create a stack of surfaces the represent the pure forms in the pattern, which end up being truncated in the finished ornament as they overlap to create the total composition. This stack can then be manipulated in the third dimension to more accurately represent the three-dimensional composition of the final orna- ment. This step was carried out in Rhinoceros, another digital modeling software that is more commonly used for three- dimensional applications.

4. The stacked surfaces are then extruded into the base of the model, intersecting with each other and creating the void spaces necessary to complete the composition. These volumes are then connected into one contiguous hollow volume.

Once the digital model is assembled and “cleaned” of digital imper- fections (mismatched edges, displaced lines, etc.), it can be opened in the operational software which runs the CNC mill. The software allows the user to choose amongst the wide array of functions that the machine is capable of based on the particular geometry at hand. For this project, I chose the “horizontal” option, in which the gantry moves the endmill along the edges of solid volumetric elements of the geometry. It begins at the top of the geometry and traces the elements, plunging further into the material as the process continues. Each “pass” of the endmill is limited to a particular depth, as the machine removes constant-depth layers of stock material. The machine essentially removes all of the stock material occupying the negative space above and amongst the volumetric elements.

**PROCESS LIMITATIONS**

Immediately in this fabrication process the limitations of this technology becomes apparent. To mill surfaces larger than a square foot within a reasonable time span, the user must select an endmill of a diameter of at least 3/8” or 1/2” (1/2” in this case); this dimension becomes the “pixel” size that will be applied to the machine’s interpretation of the digital model. It will not be able to cut details that are at a finer resolution. In addition, this “pixel” is round so it cannot cut into angular concave insets; it can only approximate these elements as deep as the endmill will fit.

Highlighted in the drawing below are some of the many details of the pattern that defy this limitation.

**DIGITAL MODEL OPTIMIZATION**

To address these issues, the digital model was adjusted to better obey the limita- tions of the machine. Any fine geometry was magnified to at least the dimension of the diameter of the endmill (1/2”). All concave “insets” were filleted (rounded) at a radius of 1⁄4” to allow for smooth machining. Essentially, the optimized model was created so that the physical endmill could fit into any void space that it would need to carve out.

**RESULT**

After both versions of the digital model were completed (original and optimized), I milled one quadrant in pink insulation foam of each geometry (the limits of these models are indicated in the diagrams on the previous page, they had to be cropped to fit within the stock material). The original took 3.5 hours to mill, due mostly to the three-dimensional curvilinear faces of some of the key volumetric elements. The optimized version took 2.5 hours to mill. I also milled an approximately 1:1 scale version of the original, which I then detail sanded; for this model I used a smaller endmill but the pattern still proved too intri- cate for the machine. For reference, I also 3D printed the digital model at a smaller scale (due to the prohibitive costs of this technology); this machine has a much finer resolution and is therefore capable of creating very fine geometries.