Introduction

Preview

As a reminder, here is my basic building form, with a neighboring landmark (Coit Tower) visible in the background.

Evaluation Metrics

Construction Costs

Calculating construction costs given the assume linear gradient based on height was a fairly simple matter of arithmetic using GH nodes.

(note: I’m realizing now that I put the total building height of 480’ where the given max of 750’ should have been used. Thus, my resulting construction costs are significantly inflated, although since I ended up normalizing later, this shouldn’t change my comparative choices.)

Direct Sunlight

Using Ladybug’s nodes: SunPath and DirectSunHours together, I obtained a cumulative yearly count of hours of direct sun hitting a number of panelized surfaces on the building. We want this metric maximized in the assumption that more direct sun could provide either renewable energy or natural daylighting. The following images show a heat map of which parts of the building get more sun over the year and the section of the GH graph to produce this analysis (note: the roof and sides were analyzed separately since the offset distance directionality worked differently for those two forms).

View to Coit Tower

Coit Tower is a famous San Francisco landmark that is likely visible from many high rises in downtown. To practice calculating directness of view, I built a replica small cylindrical Coit Tower to see how well each building design could view the landmark through its windows. To come up with one simple metric for this, I divided my main building into panels (with U = number of floors, and V = a reasonable divisions, such that this grid could actually represent windows). Then comparing (via dot product) the normals of those panels to the vector from the center of my tower to the center of Coit tower, I simply figured out what percentage of windows had a dot product greater than or equal to 0.5 (figuring that having Coit within about 45 degrees of the view straight out would be desirable, and that wide angle views less so).

Below you can see the two towers in relation to one another, my main tower with all panel normals highlighted, and the associated graph section.

Results

Raw Results

Here, we see all 12 building designs, with my three chosen inputs and six total outputs (three old ones and three new ones). Shading helps to see where greater and lesser values are. Of note: construction costs range from about $1.3Bn to $1.6Bn (*see note above on the inflation here), Direct Sun Hours vary by about 11% from low to high, and the Percentage of Windows with Good Views of Coit spans from about 45% to 50%.

Normalizing and plotting these results helps to visually draw a few other takeaways from the data and the relationships between factors.

Single-Objective Optimization Scheme

A simple weighting scheme was employed to come up with one simple score by which to compare buildings. As a simple starting point, I weighted my three factors as follows:

$Score = (-0.5*Construction Cost) + (0.25*DirectSunHours)+(0.25*PercWindowsGoodView)$

Construction cost is made negative as higher values are less desirable, whereas for the other two metrics the opposite is true.

Normalized values of inputs and outputs are shown below, along with the final resulting scores on the far right (multiplied by 1000 for easier comparison):

The three best scores are highlighted. We see that with my scoring scheme, the design with the lowest construction cost (which was weighted twice as strongly as the other two metrics) easily won out. So, looking closer at the 2nd and 3rd options, we might be tempted to compromise. Examining the line plot provided earlier, we see that Design #6 might be appealing in that it offers a steeper price drop from its predecessor while retaining relatively high values for sun and views (see green highlights):

Thus, Design #6 was chosen for the final steps of panelization and rendering.

Conclusion

Learnings

Some learnings of note included more lessons about workflow, plug-ins, and rendering.

Workflow: As this assignment added complexity to an already heavy graph, working efficiently and avoiding crashes and slowdowns was essential. To do so, any looping was executed with just 2 or 3 runs before spanning the whole design space, and was generally avoided until the last minute when everything else was in place. I found it helpful to practice and learn new nodes on separate clean graphs with simpler inputs. And on this assignment, I learned (from Dodo) how to ‘disable the solver’ which is the equivalent to Dynamo’s ‘run in manual [not automatic] mode’.

Plug-ins: Dealing with various versions of Ladybug proved to be a bit of a headache. I’d already had the latest version (1.4) but tutorial videos demonstrated functionalities in older versions, forcing me to uninstall and reinstall multiple times. The main node of Ladybug that I wanted to use was Radiation Analysis, but this required an additional software (Radiance), the exact version of which was not available for proper download on GitHub. Anyway, this was a good (low stakes) exposure into the challenges involved with open source software and having to figure out workarounds or sometimes just take another take (for me this meant using Direct Sun Hours instead).

Rendering: Once I chose my design and panelized the wall surfaces into windows with frames, I had some fun changing colors and then figuring out how to change materials in Rhino. I also experimented with the many options for rendering (some of which took quite a bit of computation) before settling on a look I liked that didn’t take too much time to load.

Next Steps

Solar: I hope to figure out the software/version issues regarding Ladybug and Radiance so that I can run all nodes of Ladybug successfully and enjoy full functionality. Doing a radiation study and getting kWh would be useful, for example, for a PV analysis; if you have insolation values in kWh and can assume an efficiency for your PVs (say 30-40%), you could directly calculate how much electricity you could generate from rooftop PVs (although these probably would be less appealing on a tall skinny skyscraper than on a flatter building). Combining this PV potential with a rough energy model (kWh used) could be a useful starting point towards designing a Net-Zero building.

Views: My view exercise with Coit Tower was functional but rough. It could be further refined by drawing Coit to a more accurate scale, locating it at a more appropriate altitude (it’s on a hill), and lastly including additional surrounding buildings as context (since I pretended like the view would be unobstructed).

Energy Modeling: one of my main goals going forward is to learn how to use Honeybee in Grasshopper for energy modeling. I’ve identified a useful video series that outlines the steps involved and look forward to working through these when I have time (the first video recommends becoming highly familiar with Ladybug fundamentals [incl. weather, sun, analysis, etc.] before diving into the energy modeling with Honeybee).