You come up with some nice optimized design but then you remember that you have to make mould for it! it is 3D printing which is allowing us to do generative and optimization!
Mathematically speaking topology optimization is generative design... at least in the context of having basic design constraints. Since generative design from this definition from the video implies "starting from nothing" this fits with the mathematical definition of topology optimization rather than shape or size optimization. Commercially these people have beautifully exploited the keywords "generative design" "topology optimization" and "shape optimization"
Two people that don’t really understand what they’re talking about. Topology optimisation isn’t simply removing low stress areas. It’s defining a field where points can either have density 1 or density 0 and finding the optimum field subject to fitness functions and constraints.
Not exactly as one can remove material in either a top down assembly to component part, or a bottom up base part to assembly, the same for adding it. In short topology opt removes mass, or excess material in the cad-design phase, while on the other hand, gen design, adds material, to create the part. They can be combined in that the part you're generating is topologically optimized.
im still a bit confused here as the end goal seems the same. wether i start with a brick and put holes in it in softwere and then create the final part or, well im not even sure how i would even design a part in inventor using Generative Design. lets say i have a frame for a car and its a mettel shel thats very easy to model out in softwere. just seems like it makes more sence to use Topology Optimization on it. but it be nice to learn as much as i can
I'm curious about this too. I'm guessing it's a mix of a few factors: 1 Different software and experience levels of designers with a particular approach, i.e. a comfort factor. Perhaps some specific part applications that require one method due to operational or design constraints or maybe there are legal reasons too vis a vis safety? It would be nice to know if there is really a difference in approach or if you wouldn't arrive to the same shape regardless of method.
I am a mechanical engineering student and I have a product design project using Gen AI and I need information about the algorithms, model and libraries that I can use in this jste project to start my first step in the project and thank you very much
Yes - to the extent that generative design could be used to aim toward a part or structure with a particular mass as one of the conditions of the optimization. In theory, it seems plausible that generative design could also be used to target a specific mass distribution or center of mass. However, we are not aware of a case where this was done.
I am currently writing an essay partially about topology optimization and generative design. Does anyone know which methods are used for which technology ? I've read about mainly SIMP and SKO and because of their description I feel like : - SIMP is for topology optimization - SKO is for generative design Am I wrong ? Are they interchangeable ? Are there other mainly used methods or are these 2 the most prevalent ?
Here is what Eric had to say: SIMP: Solid Isotropic Material with Penalization looks at a continuous variation of density and stiffness. SKO: the Soft Kill Method is a great example of generative, instead of a hard kill, where you remove the stiffness. You change the stiffness and density. So, under my definition of Topological Optimization and Generative, both are generative. They both look at variations in stiffness and density across an object where TO is a constant stiffness/density but material is removed.
Nope, i would almost say topology optimalization or generative design has nothing to do with manufacturing category. Its all about reaching product which has best value ratio mostly in terms of weight and price/stiffness.
Couldn't resist counting the number of falsehoods in his explanation of topology optimization. I lost count. No clue how someone could argue topology optimization "removes" material. (unless you still use the highly outdated ESO methods) Number one misconception on the topic.
@@StijnKoppen I'm glad that you mention it. I don't think he really knows how TO works, otherwise you wouldn't say it's an easy process. Most of the time it is compliance-based TO, which doesn't have anything to do with stresses whatsoever, unless some kind of stress criterion is involved (often not the case). After years of research in TO, I still don't know what they mean with 'generative design'. It all comes back to what are the design variables, what's the objective, how do you constrain you optimization problem and what kind of optimization algorithm you choose. Apart from that, the second part of the video is better nuanced, discussing the use of an optimized design in practice. I agree on that.
This just interferes with my thoughts, while trying to understand why audi has managed to design a fundamentally flawed gearbox mount which cannot last over 2000 miles... I think that they are using numerical design and predictive simulation to shrink the robustness margin of the product, and I'm still completely convinced and aware that they are able to design a robust part that lasts possibly over 120000miles, but the sad truth is that, the power of fea run on hpc platforms is often used to limit the product functionality and robustness to a mediocre design that sells more frequently and piles up more profit... these optimization loops run with only two parameters in mind: 1// lower the cost 2// shrink the product lifespan to a minimalistic acceptable period.
Duncan, it depends on the thing you want to optimize on. If it can be captured as a number that changes with changes in shape and topology, then yes. Most packages are simple and only optimize on a few things like minimize weight, get stress in this range, etc... But more advanced packages let you get under the hood a bit and calculate your own optimization goals. The key is that changes in the value of the goal need to be correlated to changes in geometry/topology. You mention surface finish. Generative design does not change that. It is driven by manufacturing choices. Fatique is something that advanced tools should be able to address, although you may have to jump through a few hoops to get there.
Generative design is a fancy word for topology optimization. Is just something as simple as changing the objective function, the constraints and the algorithm based on energy principles. Generative design is a buzz word: generative design is topology optimization. Nothing new.
The passion and fundamental understanding in this guy is what moves technology.
You come up with some nice optimized design but then you remember that you have to make mould for it! it is 3D printing which is allowing us to do generative and optimization!
Mathematically speaking topology optimization is generative design... at least in the context of having basic design constraints. Since generative design from this definition from the video implies "starting from nothing" this fits with the mathematical definition of topology optimization rather than shape or size optimization. Commercially these people have beautifully exploited the keywords "generative design" "topology optimization" and "shape optimization"
This video was very helpful as a mechanical engineer student
Two people that don’t really understand what they’re talking about. Topology optimisation isn’t simply removing low stress areas. It’s defining a field where points can either have density 1 or density 0 and finding the optimum field subject to fitness functions and constraints.
Yes, very strange to describe it like that.
so basically topology opt is top-down approach and generative design is bottom-up approach?
Not exactly as one can remove material in either a top down assembly to component part, or a bottom up base part to assembly, the same for adding it. In short topology opt removes mass, or excess material in the cad-design phase, while on the other hand, gen design, adds material, to create the part. They can be combined in that the part you're generating is topologically optimized.
This has implications in the automotive realm, especially in motorsports.
yup, already in use
@@janzugic6798 examples?
@@canobenitez Czinger 21c
im still a bit confused here as the end goal seems the same. wether i start with a brick and put holes in it in softwere and then create the final part or, well im not even sure how i would even design a part in inventor using Generative Design. lets say i have a frame for a car and its a mettel shel thats very easy to model out in softwere. just seems like it makes more sence to use Topology Optimization on it. but it be nice to learn as much as i can
I'm curious about this too. I'm guessing it's a mix of a few factors: 1 Different software and experience levels of designers with a particular approach, i.e. a comfort factor. Perhaps some specific part applications that require one method due to operational or design constraints or maybe there are legal reasons too vis a vis safety? It would be nice to know if there is really a difference in approach or if you wouldn't arrive to the same shape regardless of method.
I am a mechanical engineering student and I have a product design project using Gen AI and I need information about the algorithms, model and libraries that I can use in this jste project to start my first step in the project and thank you very much
I wish people didn't care about aesthetics in engineering design. Especially for internal components.
people that care about efficency won't care, I'm sure.
People say that it's beautiful a lot in this field but really we just do what is efficient then say it's beautiful in its efficiency.
The way I've viewed it is topological optimization is for walls and generative design is for infills
Hi, thanks for the video. Do you know if generative design could be used for natural frequency optimization?
Yes - to the extent that generative design could be used to aim toward a part or structure with a particular mass as one of the conditions of the optimization. In theory, it seems plausible that generative design could also be used to target a specific mass distribution or center of mass. However, we are not aware of a case where this was done.
I am currently writing an essay partially about topology optimization and generative design.
Does anyone know which methods are used for which technology ?
I've read about mainly SIMP and SKO and because of their description I feel like :
- SIMP is for topology optimization
- SKO is for generative design
Am I wrong ? Are they interchangeable ? Are there other mainly used methods or are these 2 the most prevalent ?
Here is what Eric had to say:
SIMP: Solid Isotropic Material with Penalization looks at a continuous variation of density and stiffness.
SKO: the Soft Kill Method is a great example of generative, instead of a hard kill, where you remove the stiffness. You change the stiffness and density.
So, under my definition of Topological Optimization and Generative, both are generative. They both look at variations in stiffness and density across an object where TO is a constant stiffness/density but material is removed.
In blunt words, forgive me to tell you that such a description is a bit misleading because of the mixed conceptions.
So, generally speaking, T.O. would be for milling, and GD is for additive manufacturing?
With the correct constraint, you can fabricate topology optimized parts various ways. It can be printed, cast, milled, you name it
Nope, i would almost say topology optimalization or generative design has nothing to do with manufacturing category. Its all about reaching product which has best value ratio mostly in terms of weight and price/stiffness.
This is however a good rule of thumb even if its not always the case, yes.
Topology optimization is a simple process? I just did a year of it and it is not simple
Simple to use the software
Couldn't resist counting the number of falsehoods in his explanation of topology optimization. I lost count. No clue how someone could argue topology optimization "removes" material. (unless you still use the highly outdated ESO methods) Number one misconception on the topic.
@@lolgamez9171 garbage in -> garbage out
@@StijnKoppen elaborate. the part at 4:57 looks lighter and looks to me it has less mass than its counterpart
@@StijnKoppen I'm glad that you mention it. I don't think he really knows how TO works, otherwise you wouldn't say it's an easy process. Most of the time it is compliance-based TO, which doesn't have anything to do with stresses whatsoever, unless some kind of stress criterion is involved (often not the case). After years of research in TO, I still don't know what they mean with 'generative design'. It all comes back to what are the design variables, what's the objective, how do you constrain you optimization problem and what kind of optimization algorithm you choose.
Apart from that, the second part of the video is better nuanced, discussing the use of an optimized design in practice. I agree on that.
This just interferes with my thoughts, while trying to understand why audi has managed to design a fundamentally flawed gearbox mount which cannot last over 2000 miles... I think that they are using numerical design and predictive simulation to shrink the robustness margin of the product, and I'm still completely convinced and aware that they are able to design a robust part that lasts possibly over 120000miles, but the sad truth is that, the power of fea run on hpc platforms is often used to limit the product functionality and robustness to a mediocre design that sells more frequently and piles up more profit... these optimization loops run with only two parameters in mind: 1// lower the cost 2// shrink the product lifespan to a minimalistic acceptable period.
Are there any software packages that assess fatigue, surface finish etc requirements as well?
Duncan, it depends on the thing you want to optimize on. If it can be captured as a number that changes with changes in shape and topology, then yes. Most packages are simple and only optimize on a few things like minimize weight, get stress in this range, etc... But more advanced packages let you get under the hood a bit and calculate your own optimization goals. The key is that changes in the value of the goal need to be correlated to changes in geometry/topology. You mention surface finish. Generative design does not change that. It is driven by manufacturing choices. Fatique is something that advanced tools should be able to address, although you may have to jump through a few hoops to get there.
Yes; Solid Works, Catia, PTC Pro Engineer, Autodesk Inventor are the ones that come to mind, though there are also others.
Back to watch this dope conversation yet again
So very very well explained, thank you.
This is fantastic!
-Excellent engineering designs are going to be coming out now..!
Wish there was open source software, designs would increase exponentially
Mathematically developed inverse design will become the most efficient and reliable way to automate part design...soon^{TM}
How this topology works for thin plastic parts?
same as for everything....
thank you
Generative design is a fancy word for topology optimization. Is just something as simple as changing the objective function, the constraints and the algorithm based on energy principles.
Generative design is a buzz word: generative design is topology optimization. Nothing new.
Awesome
lol teleprompter