top of page


4.022, SPRING 2019


The ​brief for the second project, building a physics fabricator, is most simply translated as "build a thing that makes things." The object of the investigation is to discover analogue processes which translate apparently simple relationships into complex forms. Through subsequent drawing and documentation, we were to develop a systematic understanding of the forms, principles, forces on, and behaviours of the systems we designed. The physics fabricator demanded a full departure from the first project, which was exclusively two-dimensional, to a system with three-dimensional inputs and outputs. Questions we were given to consider were:


  • How does your fabricator demonstrate the characteristics/principles/behaviours of the process?

  • At what moment is the process manipulated or controlled, and how does that choice impact the material?

  • What moment in time or phase/state does the material represent in your process?

  • Can recursion or repetition play a role in the process? how can these be calibrated with each design iteration?

The final deliverables: diagrams and technical drawings of the construct; 3D material samples; catalogue/taxonomy of system inputs and outputs


My group of three chose to work on kerfing. Traditional kerfing uses incisions into a planar material to inform and enhance its bending behaviour. For instance, a series of parallel lines cut into the front of a sheet of wood can cause a stiff plank to bend backwards, away from the cuts. This is made possible by the reduction of material at key points, which, in a material like wood or plastic, increases flexibility. Because the cuts are artificial, the flexibility of the material can be precisely controlled. Since planar kerfing has been explored in great detail over the decades, we decided to look at kerfing in a more three-dimensional sense: was there a way we could apply the same principles to a block of wood as one would to a sheet? What degrees of freedom are permitted by a multiaxial system? Due to the complexity of the topic and the limited time, our focus was to create a structured taxonomy of different kerf patterns and combinations thereof.


Since this was a group project, we divided up the work among us. Once we had developed the ideas together through initial foam prototyping, I spent some time designing a grasshopper algorithm. This gave us a formalised way to perform rapid iteration, and study the effect of different variables on the behaviour of and final geometries of kerfed solids. I also spent a lot of time refining our documentation, creating drawings, and designing the final presentation.

Process and commentary here (below final presentation)




After brainstorming several initial ideas and migrating away from 2D kerfing to 3D kerfing, the first thing we did as a group was rapid prototyping. We did so with polystyrene foam. The semi-rigid material becomes quite flexible when cut to thin sheets, though it is also brittle, which makes it particularly risky to apply strong torsional forces to it. With a general idea of the fundamental cuts and combinations available to us, we moved onto materials like wood and HDPE plastic.

Wood, having a grain, opposed the kerfing very strongly. As you can see in the movie, it moves, but the force we had to apply was substantial compared to the HDPE or foam. The HDPE turned out to be an almost ideal material for our purposes, as it was dense enough to withstand various pressures while being 

flexible enough to accommodate bending. Our only qualm was the limitation

on fidelity our tools set on us. As we discovered over the course of several iterations, the higher the density of the cut patterns the more drastic the geometric change would be. Unfortunately, you can only make so precise a cut when using a bandsaw and a tough material. Sam and Seif did a fantastic job given the resources available to them, and we got some models which were quite remarkable, but it was clear that we had to figure out a new strategy.


Screenshot 2019-04-09 at 7.10.12 PM.png
Screenshot 2019-04-09 at 8.09.43 PM.png
Screenshot 2019-04-09 at 8.00.56 PM.png

The prototyping phase allowed us to identify the primary cut types we would use. Given our limited time, our goal was to document the behaviour of an initial constrained set of manipulations. We later diversified, but achieving a core understanding of 3D kerfing was essential. There are two categories of cuts we worked with, both based off unilateral lines.

The first (left) consisted of lines drawn from the edges to a central axis, the second (centre) was similar, but the lines extended beyond this axis. As you can see from a sample of our initial designs (right), there are several ways to diversify even these simple line types. These variations are parametric, achieved using a simple Grasshopper script, and allowed us to change the position of the axis, its shape, the angle of the lines, and their density.

The combinations of the cuts are what make 3D kerfing really interesting. Since we have more faces to work with, we can cut along multiple axes, combining behaviours to create something unexpected and new, like a torsional motion:


Initial Brainstorming



Artboard 4_2x.png
Artboard 3_2x.png
Artboard 5_2x.png
Artboard 2_2x.png
Artboard 6_2x.png
Artboard 8_2x.png
Artboard 7_2x.png






Following the guidelines we had set for ourselves, we developed a catalogue of more refined foam models.

Perspective with Elevations.jpg



Having performed initial protoyping and settled on the cut patterns we would investigate, we wanted to formalise our analysis. The best way to test several different variations on a similar theme was to first create a Grasshopper script which would create precise geometries we could print, and then design models which simulated the effects of kerfing on our blocks.

Physical prototyping 

Prototype A
Prototype B

The first scripts were expressly used to create objects that could be printed. Since our process demanded a high fidelity inachievable with a bandsaw, this was the natural progression of our prototyping phase. The use of Grasshopper gave us flexibility in determining which variables we wished to test.

Comparison of pattern generator to 3D-printed prototypes. We intended to get cubes but the printer bailed on us, so the end result was about 1/4 of a cube. That was fine too. I really loved the plastic we used — it clicked in a deeply satisfying manner when we bent it.

Initial Kerfing Simulator

Having made those models, I was curious whether I could generate something that would allow me to predict the motion of the blocks before we even printed them. My initial attempts were relatively unsuccessful. While a bending motion was achieved, I realised that I did not need the whole geometry to bend, just the central axis. From there, I could add appendages.

While this is a cool effect, it's not quite what I was looking for...


Screenshot 2019-04-14 at 12.36.10 PM.png
Screenshot 2019-04-14 at 12.36.04 PM.png

Armed with this knowledge I set out to make a new model. Albeit, I took a roundabout way getting there. One of the things that has fascinated me about Grasshopper from the start is Kangaroo (a physics simulator which operates on the basis of springs and forces, acting within the framework provided by Grasshopper. It is an absolutely amazing tool, and getting to learn it is a fun challenge.) I first made a flexible beam which behaved the way that the axis of the kerfed blocks would, then retrospectively added the kerfing.

Demonstration of different variables

9x9 assemblage of different kerf patterns

Since the code was pretty flexible, I was able to output geometries, intersect them, and send multi-axial cut shapes to the 3D printers. Only one actually printed (a whole 20% success rate. It particularly shocked us that the printer anticipated spending 53 hours on one of our prints. Consequently, we didn't have that many models to show for the final review, though we did have a lot of cool diagrams.)

Screenshot 2019-04-09 at 7.01.03 PM.png
Screenshot 2019-04-09 at 7.02.29 PM.png
Screenshot 2019-04-10 at 8.06.10 AM.png
Screenshot 2019-04-09 at 7.08.42 PM.png
Screenshot 2019-04-09 at 7.09.37 PM.png
Further Exploration


I would love to carry on studying this topic, particularly since we did not get quite as far as we'd hoped in producing actually programmable distortions. That said, we did achieve our initial goals, to create a catalogue of different kerf behaviours in 3D. Other avenues that we had begun to explore but ultimately abandoned due to time constraints were metamaterials and auxetic materials. I had gone so far as to replicate the Karlsruhe Institute of Technology study on twisting materials. Even though we did not print it, this mini-project influenced my own understanding of what we were doing greatly — the objective, at least when it came to torsion, was to transform a linear force into a rotational force by changing the geometry of the lines along which said force travelled. I would love to be able to print this someday.

2017_169_Science Metamaterial mit Drehef
Screenshot 2019-03-27 at 8.35.25 PM.png

Grasshopper replica (can change diameter and thickness)

Screenshot 2019-03-31 at 10.23.21 AM.png

STL file, 10x10 array



Finally, I want to mention some of the feedback that we received during the final crit, so that it might inform your reading of our project. The project was very poorly received. A couple of the reasons stated were: despite our clearly having done a lot of work on it, the project was poorly organised; we only succeeded in creating a taxonomy/catalogue of the different kerf behaviours, rather than working towards a particular end goal; our models were not sufficiently refined. In some ways, I hope I implicitly defended against that in the preceding paragraphs, but I respect the critic's final judgement and will take those lessons to heart going into future projects.

What I enjoyed the most about this project was working in my team, the other members of which are two absolutely fantastic people, and learning more about the software and materials I will be using in my future classes. One of these is Grasshopper. This project allowed me to spend a lot of time messing with the Kangaroo physics engine. I have also begun to formalise my design methods, and I hope that I am developing a better intuition for the coding process. On a superficial level, my code has become much better organised and more aesthetically appealing to look at since I began using grouping and invisible wires

properly. On a deeper level, I have realised that I am completely fascinated by parametric design and am gradually growing more adept at discovering efficient solutions to specific problems.


In short... this was super fun. I am glad we chose such a challenging topic, although the disappointment at the end soured my week somewhat. I still have the singular successful 3D-printed object on my desk, and look on it fondly. It sits right next to the first object I ever welded. The one is a reminder never to shy back from trying new things, the other reminds me that sometimes failure happens even when you've done your very best. Most importantly, it's a great fidgit toy!