Generative Design - The future or fashionable?

[jwelihin]

Wondering what everyone thinks of Generative Design (you know, those bikes that look like two wheels are playing cat's cradle...)

I saw a demonstration of how it can be used to create 100s of designs that would be a surface modelling nightmare, and curious if you think it will actually take off or not?

[matt]

Yes, I think some companies and some designs will be able to make good use of generative. We now have the ability to manufacture those shapes through 3d printing, and optimization of stress/material is always a good thing.

These designs will be manipulated as mesh data, so surfacing is less of an issue.

[jwelihin]

Yes, I think some companies and some designs will be able to make good use of generative. We now have the ability to manufacture those shapes through 3d printing, and optimization of stress/material is always a good thing.

These designs will be manipulated as mesh data, so surfacing is less of an issue.

I was shocked and thought this is the future. Have AI save costs or find the right configuration of things.

[matt]

Well, it's going to take off in the same way that configurators took off. Not like how iPhones took off, or like how Anna Nicole Smith took it off either, but that's another story. People who have the vision and a need to do it and a product that fits the strengths of generative will do it. It's not going to be consumer products that benefit most. It's going to be big industrial stuff where there is currently massive over engineering and wasted material and low volume, they just design that way because manufacturing methods had not caught up to design tools.

[SPerman]

I'm not sure who else is playing in that space, but my understanding is F360 is fairly powerful. You can even tell it how you plan to manufacture the part. I don't think they are targeting big industrial designs, but then again, I'm not sure who ADesk is targeting with that product.

https://www.autodesk.com/solutions/gene ... ufacturing

[bnemec]

All the marketing pictures come to mind of the organic shapes. I was tempted to search the term then paste images but no need, everyone has already seen them. They make pretty fliers and get peoples attention in the "What's new fliers" Problem is, Marketing shows something it can do but is mostly useless rubbish. Software still hasn't replaced human reason and intuition in Engineering principles. There are a lot of great tools out there, don't get me wrong, and we should be using them as much as is helpful and possible. Most of the software is made and marketed to make profit for the company developing it not for the customer, regardless of what they say.

What is "Generative Design"? I don't know for sure, it's a word that some group came up with. What's it doing? I THINK it's automating the design, simulate, adjust iterative process. So it is completely dependent on correct design inputs, do you know them? Do we always know the exact force vectors and moments on the faces of the geometry? Ok, lets say we do and move on. Simulation, I haven't done much FEA but in what little I have done the results must be interpreted, the color orange could mean 10ksi or 10000ksi, so those pretty pictures are out. The first question is usually, "let me see the mesh" which tells me the meshing is very important. Some claim they have excellent automatic meshers that will adjust size as needed depending on stress gradients in a localized area.... It goes on and on, the point is the generated shape is dependent on many things and the user mush understand or else the output is likely no better than what can be done with hand calculations and good modeling. I'm old fashioned I guess, Engineers should have a strong understanding of mechanics of materials and statics and dynamics etc. and use tools to make them more efficient.

[matt]

All the marketing pictures come to mind of the organic shapes...

Generative design essentially is reverse stress analysis. You give it a block of material, and the connections/load points, and the software figures out where in that block the stress goes, and then eliminates all the unstressed material. It tries to even out the stress - optimize. The shapes aren't limited to the straight lines and arcs that we manufacture in bulk. Stress analysis uses meshed geometry, which is also used for subd modeling. So it's not that far to travel between a model meshed for analysis and a model meshed for subd. It's no accident that these things show up close to one another in the development cycle.

Generative is made possible because we can now print those organic shapes, hollow shapes, etc. Printing organic shapes doesn't cost any more than printing straight lines. We all know that sharp corners are stress concentrations. Bridges were the first examples of shapes being used for structure where we used to use rectilinear shapes. You've seen the Gaudi spans made 100 years ago? Now we're just extending it into 3 dimensions with new manufacturing technology.

These shapes won't work for every project, but for some, they save a lot of weight and money.

Siemens/solid edge are doing this as well.

[MJuric]

All the marketing pictures come to mind of the organic shapes. I was tempted to search the term then paste images but no need, everyone has already seen them. They make pretty fliers and get peoples attention in the "What's new fliers" Problem is, Marketing shows something it can do but is mostly useless rubbish. Software still hasn't replaced human reason and intuition in Engineering principles. There are a lot of great tools out there, don't get me wrong, and we should be using them as much as is helpful and possible.

The real point here is that "Software hasn't replaced human reasoning....yet".

This is true even for much less complex things like scheduling and other "Decision making" softwares. People don't realize how many decisions are made in the human brain on even simple things like "What part do we run next". "What machine does it run best on"...well that depends on material, who's running the machine, machine capabilities and on and on. Oh but oh, wait, this job is due later but we actually want to make sure it's done on time so it should be run before that one and on and on and on.

The biggest part of having any of these softwares working correctly is to set it up correctly. Software is completely incapable of walking out into the world and understanding the nuances of design, company, resources etc etc. The end result is that ALL of that has to be done for the software and that means TONS of setup time. Otherwise all you get is garbage out even with good stuff going in.

So until such a point that the software can also "Set itself up" there is going to be a WHOLE lot of up front time and that investment only works for people that are going to use it alot....and a good portion of designers won't use it very often.

[bnemec]

The real point here is that "Software hasn't replaced human reasoning....yet".

This is true even for much less complex things like scheduling and other "Decision making" softwares. People don't realize how many decisions are made in the human brain on even simple things like "What part do we run next". "What machine does it run best on"...well that depends on material, who's running the machine, machine capabilities and on and on. Oh but oh, wait, this job is due later but we actually want to make sure it's done on time so it should be run before that one and on and on and on.

The biggest part of having any of these softwares working correctly is to set it up correctly. Software is completely incapable of walking out into the world and understanding the nuances of design, company, resources etc etc. The end result is that ALL of that has to be done for the software and that means TONS of setup time. Otherwise all you get is garbage out even with good stuff going in.

So until such a point that the software can also "Set itself up" there is going to be a WHOLE lot of up front time and that investment only works for people that are going to use it alot....and a good portion of designers won't use it very often.

Thank you for stating that more clearly for me.

[SPerman]

I don't think this has to be an "either / or" option. The engineer can still look at the results and make changes, or add new limitations to the solve parameters. But it could save the engineer hours / days of design iterations and results analysis.

[MJuric]

Thank you for stating that more clearly for me.

This is one of my current projects, attempting to properly set up our planning system in our ERP. I spent an our on the phone with our vendor and he said two things. "70% of the planning module is setup, set it up wrong and it will spit out garbage" and "Most customers hire us to set it up".

Now this is a "simple", in comparison to a design, scheduling piece of software and in essence 70% of the work is setup and you should hire an expert to do it. That's where we are with true "AI" intelligence at this point.

[Lucas]

Not sure if Generative Design is very useful for conventional manufacturing, since removing material and adding complex curves increases the processing time; but for Additive Manufacturing it will truly shine, because in AM customization is pretty much free. It is more about getting the initial cost lower to become more popular.

Also it will definitely go fashion as SubD integrates with it. lol
The future is awesome

[mike miller]

Not sure if Generative Design is very useful for conventional manufacturing, since removing material and adding complex curves increases the processing time; but for Additive Manufacturing it will truly shine, because in AM customization is pretty much free. It is more about getting the initial cost lower to become more popular.

image.png

Also it will definitely go fashion as SubD integrates with it. lol
The future is awesome

The real irony: sub-D is a lot like clay modeling. So it's back to the future..................again!

[Lucas]

The real irony: sub-D is a lot like clay modeling. So it's back to the future..................again!

Cant wait to work listening to Unchained Melody

[Frederick_Law]

https://www.axios.com/2024/05/08/autode ... ls-bernini

Exclusive: Autodesk's AI turns text or still images into 3D models