Generative Design Result Errors

[joes14159]

Hi all,

I have been attempting to run a generative design study on the following body and have been encountering issues with the study creating solutions. I have tried several variations of preservation and obstacle geometry with all the studies failing. The main error I receive I have screenshot below.

The plan for this design is to apply a force to the left end face of the body, with fixed geometry on the angled face. I don't want the external profile altering, so I have created an external obstacle body using surfaces and thicken. Similarly I require a minimum wall thickness of 1 mm, so using the same method, created a preservation body. I wish generative design to infill the void of the body to a safety factor of 1.5, and a manufacturing method of ALM. The design is to minimise weight

Has anyone any suggestions on the issue, or possible ways to get this to work? I have attached the 

[Ben-Weiss]

Hi @joes14159 !

Thanks for your thorough description of what you're trying to do and for attaching the f3d. I downloaded your setup and took a look around, but I don't see anything obviously wrong. I'll run it on my end and see if I can reproduce the error.

Ben

[Ben-Weiss · Accepted solution]

OK I was able to reproduce the error on my end, and you've hit a subtle issue with starting shapes. When you use a starting shape, the generative system checks to make sure your starting shape includes at least some additional geometry beyond your preserves, because generative design typically creates the final shape by removing material from the starting geometry (though this isn't a requirement like it is for topology optimization). To keep users from wasting money on bad solves, we check that (volume of preserves) < (volume of starting shape). In your setup, the starting shape exactly overlaps with the preserve bodies, so the above check is failing. We still have some work to do to get that error information back to you (instead of just saying, "we failed, call tech support"!).

The fix is to add a little bit of geometry to your starting shape so it has a volume greater than the preserves. If I add a ball inside the tunnel like this, the solve runs fine. Unfortunately the solves finish pretty quickly because the factor of safety of the initial design was less than the target (solves work better when the initial design meets the factor of safety target). The results are not very interesting. Here's the X+ additive outcome:

X+ additive outcome

Note that the additive constraint isn't very robust right now, and honestly the results are less than satisfactory for lots of problems (I'm working on improving it).

Free, unasked-for advice: I think you should consider creating a starting shape which is heavier/thicker (so the design is stronger than it needs to be at first). In addition, because almost the entire outside of your part is preserve body, you need to provide generative with some exposed surface area on the inside where it can make changes to your design, so don't just make a fully solid starting shape. Think of filling the interior of your starting shape with a lattice of crisscross bars to provide strength and surface area simultaneously. Note though that generative doesn't accept fully-enclosed voids in starting shapes (there has to be a path to the outside). If I were working on this problem, I would probably not include the long thin tube in the generative solve since it likely won't play much of a structural role beyond the first few centimeters and we know how to easily fill tubes with material so they are self-supporting after the fact (e.g. with a tear-drop shape). Not only will that get you a higher resolution solve, but it will make it much easier to assess the design from the Explore environment (which doesn't have a section view tool) instead of having to pay 100 CC just to look at your design.

[joes14159]

Hi Ben,

Thanks a lot for your input and information it is much appreciated! I did as you suggested, placing an internal structure to act as 'seed' points for the generative design, and increased the design strength beyond the preservation zones. See images below for results. We have a weight saving of 28% using generative design!

One further question I have; the generative design window provided this solution with a safety factor of 1.5, however, running FEA on the body post download we can only get 1.2. This isn't so much of an issue with this part, just a query as to why this may be?

Thanks again for your help on this project.

[Ben-Weiss]

Sweet! That design also looks really cool...I personally haven't seen too many people optimizing internal structures with generative, and this is a nice example.

As for the factor of safety mismatch, this happens quite a bit in generative results. There are three reasons:

1. The generative solver and the Simulate workspace use different meshes when performing FEA. Because the stress maximum is typically an infinitely small point on the body, it is only imperfectly estimated by any simulator, and the accuracy depends on the type and quality of the mesh used. Every time you change meshes, you see small variations in the maximum stress detected.
2. The generative results you get when you export your design have been converted back to a T-Spline body. Small shifts in the shape that occur during this translation can create stress raisers at corners, typically where a preserve body interfaces with the organic T-Spline, and these can change the perceived maximum stress and factor of safety.
3. There are some minor numerical differences in the way factor of safety is computed inside generative and inside the FEA workspace. We do a little filtering inside of generative to help the system produce more robust outcomes. Sometimes this shows up as a discrepancy between the values reported in Explore and validation simulations.

Hope that helps explain things,

Ben