Intro

Preview

Here’s a preview of my building’s general shape. As you can see, the base shape is a triangle (polygon with (3) sides and fileted corners). That base shape is lofted upwards with a certain degree of twist.

The basic construction of this building form can be seen in the graph section screenshotted below (Figure 1). Therein, you can see that the base shape is lofted up to a ‘Mid’ profile (0.5 * [Building Height]) and then up again to a ‘Top’ profile (full ‘Building Height’). You can also see here that the Top section gets a full rotation, while the Mid section gets half the rotation to make for an even twist throughout.

Process

Boundary Testing

Fitting my particular base shape within the lot boundaries (300’ * 450’) was not immediately intuitive or obvious. Having my base centered on the origin and the size of the shape dictated by the ‘polygon radius’, not to mention the fileted corners, meant that simple geometry with pencil on paper didn’t make me confident that I’d found my maximum base size. So, I made a separate Grasshopper file to visually test the arrangements to find the maximum radius. Figures 2 and 3 show the Grasshopper graph and resulting Rhino geometry respectively, and this video shows this simple visual test in action.

Making this separate graph was a simple but encouraging example of how Grasshopper can make some calculation or test much easier (simpler, quicker, AND more visually clear). Also, building this in a blank new file allowed the rendering to work quickly and smoothly (as opposed to running this within the larger graph, which was running more slowly).

Design Constraints

Meeting the design constraints for floor area (between 1.2 to 1.5 million square feet) was more difficult than I anticipated. As mentioned above, I chose to vary three input parameters: Building Height, Triangle Radius, and Twist. At first, I varied each of these parameters across wide value ranges resulting in buildings of all sorts of shapes (short/skinny/twisty, tall/stout/straight, and every variation therein). I also set up these value ranges and then shuffled them in Excel to get randomized combinations. Naturally, very few of these building shapes actually fell within the required range of floor area.

So, instead, I played with limit states, finding a combination of values that would result in ~1.2M SF, and another combination of values that would result in ~1.5M SF. Increasing Building Height and Triangle Radius obviously led to more floor area, but in this process, I discovered that with the way Rhino/GH was lofting my three profile shapes, increasing the Twist actually reduced floor area with the other two parameters held constant. I had originally thought that each floor should have the exact same area, but in the way that the software lofted the three profiles and then subsequently divided that form into floors and found areas of those intersections (floor planes with building shell), some floors (those furthest away from the specified profiles) actually shrunk in size by about 10%.

After some trial and error, and starting this exploratory process from Excel, I found that plugging in sliders to temporarily replace the Excel data was a much quicker way to test scenarios and find appropriate limit states (see orange lines in Figure 4). Organizationally, adding the relay nodes on the right made this process quicker and easier to do many times as needed (rerouting / replacing notes).

All this experimentation led to a design space of (12) forms, shown in Figure 5.

The resulting range of building forms spanned linearly from short/stout/twisty (Figure 6), to tall/skinny/less-twisty (Figure 7). This video shows a run through of all the building forms.

Implementation

Figure 8 shows a full picture of my final graph with areas of focus boxed in orange.

Much of the graph followed the example given in the Module 5 videos on Canvas. Of note, however, is the extraction of (3) instead of (2) design parameters from Excel (Figure 9), and the efficient / automatic transfer of collected data back into Excel (Figure 10). (The process of perfecting the workflow of the ‘Excel Write’ node was much more difficult than anticipated, and was finally made clear with some quick help from Donatien).

Results

Assessing these (12) building forms by the given metrics (Volume, Surface Area, and Floor Area) yielded some interesting results. These results are summarized in Figure 11.

Cell shading helps to blatantly distinguish inputs from outputs and to indicate the direction in which values were increasing (with maximums in the darkest shade).

Normalizing and plotting these values on a line chart revealed some interesting non-linear and irregular relationships between our inputs and our resulting outputs (see Figure 12).

Conclusion

Learnings

This assignment forced me to think and deal with workflow more so than previous assignments did. The complexity of the computation made my program run much slower and actually caused my Rhino/GH to freeze a few times. Even successfully executing a series of 12 runs of my design space took a full minute or two of time during which adjustments could not be made. To account for this, extra care was warranted to make a run ‘worth it’, and workarounds were sought to temporarily reroute nodes and avoid unnecessary computation and automatic looping. Simple things like setting up an extra watch panel with large font as an index counter helped to avoid getting lost in the runs and make sure the computer was on track. Separately, I found that working on this assignment in multiple sittings was helpful - each sitting often ended in a dead end or spiral of being stuck, but coming back fresh on another day always had a beneficial effect.

Next Steps

Given more time, I would refine this project in a few ways. For one, my building form was not particularly interesting or attractive (in my opinion). It took long enough to find shapes that lofted properly that once I got that functionality running, I simply moved on to the next steps. But, I would like to try more complex and more attractive shapes and forms, like incorporating a taper, making the profile shapes different from one another, or drawing a custom base shape (even perhaps in Rhino) and lofting that. Edward’s take on the Shanghai Tower (with the rounded shape with a square corner notched out) would be a fun challenge to try.

Aside from aesthetics, I would have also liked to have varied the design space of options more. Given that a run of (12) forms took about a minute and my early trials in randomization put me outside of the floor area requirements, I felt somewhat limited in my experimentation. But, given more time, I would like to perfect the process of expanding the ranges of input values of parameters, randomizing combinations, and assessing the results. This approach would likely require developing many many more runs - say 100 or a 1,000 - since my initial trials with 30 runs provided only ~1-2 acceptable floor areas if any.

This was another fun and challenging assignment whose discoveries awoke even more curiosities about what might be possible with these tools and methods.