Week 7: Fusion 360 Shape Optimization + Generative Design

This week we were introduced to Fusion 360, and were tasked with exploring topology optimization and generative design to design a mounting bracket for an existing object. I chose a dichroic LED strip light that I often use in my work, and attempted to make an interesting mounting bracket utilizing generative design.

• Fusion 360
• Ender-3 Pro
• Cura

I attempted to follow the Topology Optimization tutorial using the example hook file, but I was unable to get a successful result. I downloaded the Fusion 360 hook example file, and attempted to follow the tutorial step by step, but when I got to the "solve" step, I kept getting error messages. My first guess was that the settings were already all set in the example file, and by following the steps myself I was adding redundancies that threw off the settings. I tried to reimport the example file and "solve" it without making any changes, but that failed as well. I attempt to resolve this a few times, but feeling the deadline looming, I decided to move onto the generative design portion of the assignment.

The LED strip light comes with a mounting bracket, so I had a template for my modeling. Initially, I scanned the profile of the mounting bracket on a flatbed scanner, imported the image into Illustrator, and exported an SVG of the profile. I first mocked up a fit test by extruding the SVG in Blender, slicing the result in Cura, and printing the result. I was quite pleased that this initial test fit perfectly.

After that initial test, I tried to model the mounting bracket in Fusion 360. It was my first time modeling in Fusion 360, so I was a bit slow. I kept checking the press fit dimensions of the model with my real world template. I then attempted to model a circular base with a 45˚ taper. I added cylindrical forms to be an obstacle. I forgot to place obstacle geometry at the top where the light would go, so my first test was super blocky. I then added a cube obstacle to represent the light. I ran into some issues with the load parameter, but I kept using the error messages to resolve the issues through trial and error.

The first results that appeared were super chunky, and I was afraid I had set the force too high (10lbs). I didn't realize the machine learning algorithm would keep going progressively through iterations of reduction on each test model, so that was a big learning lesson. The results kept getting progressively more interesting, but I wasn't sure how long it would take to finish. I exported one of these early iterations as a mesh, just to test whether the press fit would actually work. I sunk the mesh down on the z-axis into the print bed in Cura, so I could get a quick test of just the press fit bracket, without having to wait for the entire print. This test fit, but was a little looser than my initial models. I attempted to bend the bracket in for more tension, but it immediately snapped. I didn't add any force values on the walls of the bracket, so perhaps that made it too weak. I also printed it with infill in Cura, so perhaps I need to print it solid with no infill. I did a second test with 50% cubic infill, which seemed stronger.

Once I started to get multiple results, I began to use the "compare" viewer function. It was interesting to see how much the interface lined up with the Dream Lens paper, and it was interesting to see how that research gets folded into the actual software suite. It was also very satisfying to watch the time lapse of each model as it progressed through multiple iterations.

I decided to export the mesh of the first converged result, sliced it in Cura, and began printing. All of the other iterations were taking a while, so I figured I should print something while I waited for the set to converge. I will see if any other results are worth printing. I included some of the wonky glitches because the generated supports are interesting (or comical), though the bracket would need to be cleaned up in post production.

Often when I'm frustrated by new software or new techniques, I explore the same problem with something I am more comfortable with while I wait for something to render or print. I decided to try to use a "generative" approach in Cura by using the infill patterns and supports. I added a bracket and a base, and then explored Cura's slicer settings to generate support. This wouldn't be practical, as the part is meant to easily separate from the supports, but it did help me to visualize and ideate. I didn't think the result was worth printing, so I just took screen caps.

I wound up using tree supports for my final print, and it was interesting to see how the Cura generated tree supports echo the Fusion 360 generated design.

I also included a test I did earlier in the week using the Lightning infill setting in Cura as a generative structure within the interior of an icosahedron. I stopped the print at this level to expose the infill pattern. I plan to pour clear resin into this form, and then remove the PLA walls of the print once the resin has cured. If the resin will release from the PLA, I think this could nicely encapsulate the lightning infill, and the fluorescent PLA should glow on its edge.