Watch Band Link Topology Optimization
THE PROJECT
My goal with this was to redevelop the apple watch band to include topology optimzations and develop a product fit for eventual production/sale. The rendering to the left is one of 12-13 links that would mount to each other on either side of the apple watch, connecting with a clasp at the bottom and connecting to the apple watch with a proprietary lug that matches Apple's dimensions and tolerances.
Generative has been my largest design interest for the past couple years since finding out about it at IMTS 2018. Leaning hard into computational design had me constantly thinking of how I could bring this technology to consumers. Teaming up with a friend I'm commited now to designing a series of products with the hopes of eventual release. The first of these designs is a watch band link.
Going through the design process for this link, the largest benefit was developing an ideal workflow. A further goal with the entirety of this project is to delve into purely functional artifacts such as bicycle frame lugs, athletic accessories, or architectural features. I chose Ntop for this specific reason of deriving form from something so inherently functional. Bringing this technology to artifacts such as a watch band link which doesn't necessarily need the functional benefits is part of the appeal.
The link you see to the left is one of the final iterations of a workflow I developed for this task using Autodesk Inventor and NTopology. A portion of my Ntop nodal map is shown below.
PRODUCTION
The goal is to produce this system of links and forthcoming claps/connecting pieces largely with metal additive manufacturing solutions. I'm in contact with the CAMDI lab at UT Austin that has two DMLS machines available for use. The plan is to make use of these machines for early prototypes and later pursue a more established service overseas for higher scale production. The nature of optimized topology expels the use of traditional subtractive manufacturing processes. Part of the appeal with this design is to bring novel manufacturing techniques and materials to consumer products, providing a level of novelty while also promoting principles of material/environmental efficiency, pure design optimization, and form as a deriviative of function.
There are some incompetencies with the current capabilities of metal AM, however, which deal with costs as well as accuracy for our level of application. The holes in this link are 1.5mm in diameter, for example. The mounting solution cannot be M1.5 threading that is included into the design because most Metal AM machines are not capable of tasks below a required accuracy of 200 microns. Solutions I'm thinking through are either tapping the parts post production or integrating an additive composite based nut/bolt system.
Above are two representations of slicing the artifact for additive manufacturing, the visualization for which I thought was particularly beautiful.
THE NODAL MODEL
To the right is a portion of the nodal workflow from raw cad file, to implicit body, to objective and load cases, to topology optimization, to union, subtraction, and translation of initial cad bodies, and a couple more operations to get the look I wanted.
Much of my iterative process dealt with rearrangement of the forces and fixed displacement regions associated with the load cases that I defined for each of the three structural compliance responses shown below.
My method of validation during the iterative process was larely aesthetic. If a design feature developed that looked 'good', I retained the inputs. The reason for this is that forces on a watch band link are largely arbitrary. The task itself is somewhat trivial due to the small scale at which a watch band link exists/operates. Regardless, the result is something I'm very satisfied with
FORCES
A raw representation of some of the forces is shown to the left
ISSUES
A large part of our planned marketing for this product deals with photorealistic renderings and animations with accurate materials and textures. The issue we dealt with was outputting to complex of a mesh for the blender render environment. The image you see directly below is an initial version of the mesh that has around 100k faces. When rendering with complex geometry in blender it becomes imperative to decimate the number of the faces in the object prior to applying materials, texturing, and then rendering an image. When trying to decimate the object in the image below we were met with odd points in the mesh where polygons seemed to converege and effect the smoothed shape of our mesh. Rendered view would then have this converging geometry show up, which wasn't ideal.
The issue persists, although to a much lesser extent after adjusting the mesh output within Ntop, the final result which you can see in the second image below. Remeshing, quadrangulating, and then refining the mesh gave a much more uniform polygon structure which decimates within blender in a more predictable way. There are still problematic sections, however, which we are working on solving. A way to selectively refine the mesh and increase the variability of selection would be very ideal. Experimenting with Ntop Automate and delving into the code to manually adjust this variable might be the solution.
NEXT STEPS
Coming up with an effective hardware solutions is imperative, the options for which I outlines brielfy above. Additionally, I'm working on designing the clasp mechanism and lug linkage that mounts directly into the Apple watch. Having found the necessary tolerance documentation online, finishing the product is a matter of perfecting the design, prototyping on all ends, and finding something that we're comfortable with putting out.
I also plan to experiment with a mechanical watch movement at some point, implementing the generative link above or perhaps something similar.