English Wheel and Generative Design

[TheCADWhisperer]

I am involved in a project in which I would like to apply Generative Design tools.

We have a significant Automotive Restoration program down the hall from my office.

They have requested help in re-designing and fabricating an ultra-rigid English Wheel to improve upon our existing tool. 

Now of course I could simply throw a lot of steel at this design, but thought it would make a good exercise for Generative Design tools (and perhaps a published paper).

I attended AU this past Nov to do a bit of research on the state of Generative Design in Fusion 360.

I am going to use this video as reference for the rest of this inquiry. (unfortunately the AU website is not working at all for me today - but that is another issue)

Timestamp 47:35

Here is an example presented in the Manufacturing Keynote of traditional part design.

Timestamp 48:34

Here is generative design solution presented.  It does not appear to be a manufacturable design.

From what I saw at AU - this is typical output from running the Generative Design tools in Fusion 360.

Timestamp 49:09

If I understood correctly, this image was reported to be Autodesk Inventor design based on the Generative Design result.  My takeaway from this is that Generative Design in will, in effect, take place of "napkin sketch" that will then be used to guide the CAD operator in creating a *manufacturable parametric part.  *(Keep in mind that wherever I use the term manufacturable - I am considering Design for Manufacturability constraints like processes, materials AND economics - everything that would be covered in a 15-week DfM class.)

The next example presented in the Keynote was a redesign of this wheelchair frame.

Timestamp 55:20

Timestamp 59:45

And an image of the impressive number of possible iterations generated within the defined constraints.  This doesn't make a very informative image for those who have not run a Generative Design.  Give it a try, I post a link to appropriate AU classes once the AU website is running correctly again.

Timestamp 1:02:08

And this was presented as the selected solution from the iterations.

First thing I notice is that it appears perfectly symmetrical and aesthetically different than Generative Design results.

Timestamp 1:01:23

Then t-Splines editing is demonstrated on the geometry.

My question:

How do I get from a Generative Design "solution" to a perfectly symmetrical, aesthetically pleasing t-Spline or traditional parametric (and manufacturable/functional) solution for my English Wheel?

[TheCADWhisperer]

Actually, I an not concerned with perfect symmetry - if the analysis indicates asymmetry is appropriate - I was just trying to figure out how the wheelchair frame analysis came out symmetrical as presented.

Edit 2:  I forgot to ping  @MikeSmell_ADSK

[I_Forge_KC]

You are correct about the napkin sketch concepting. The results coming out of AGD at this time are organic and while they can technically be machined (based on the manufacturing constraints added), they are still 3D curves with very high finishing times. 2.5D milling is on the horizon but not available in the application currently. As a decent approximation, you can use a method like that employed by Slicer to create thick slabs representing your machining depth. Similarly, you can leverage other machining constraints (short/wide tool) for creating castings and forgings.

Symmetry is not naturally attainable with the current paradigm because of solver sensitivity and voxel distribution. With that said, all the symmetrical results you see have been cut and mirrored with likely a smoothing done to the mirror interface. A symmetry constraint will be along sooner or later (it exists in Dreamcathcer).

In order to obtain a high degree of result potentials like you see in the wheelchair example, you must employ exotic starting shapes as well as multiple materials. The intent being that AGD provides you with a large number of pre-validated concepts that you can then engineer around. The solver is highly sensitive to local minima in the design space, so exotic seeds will yield different than end shapes. Level set methods and holes have a funky relationship, too, so punching a bunch of holes in your starting shapes can dramatically change outcomes as well.

The results out are true BRep with a beautiful parametric recipe where your interface geometry are true to form and the organic growth is a TSpline surface between them. A little sculpt/boundary fill magic and you end up with a nice BRep. That TSpline is fully editable like any other form body.

PM me for help if you'd like it directly, JD.

[Anonymous]

I dont use generative design but of all of the topology optimization packages I've used, they offer a symmetry plane option. 

If you're just using the output as a "back of the napkin", I still dont see the value of generative design over topology optimization in it's current form. The real value of GD (at least to a design and analysis engineer) is in its ability to explore a broad range of materials and manufacturing methods. If you already want to design for a specific manufacturing process (my assumption based on your comments) just run it as a topology optimization with a few cloned studies in the materials you want.

Now if the proposed features of GD are ever released that were mentioned at AU, that will probably be a game changer in the design world.