16:42 All right, so thank you very much. So my name is Peter Clausen, I’m from the company Dassault Systèmes. I’m here with my colleague Pascal. I so yes, my name is Pascal, I’m also working with system for K. Yeah, so I am from more the simulation side of things, and we wanted to have a little bit of another take on the computational design, which many people been working on for decades. So I’m originally from ol sigmon’s team, graduated back in 2005, and we want to kind of promote computational design also for conventional manufacturing, not only additive, right, and that’s kind of like the the the theme in this talk.

So the outline is, we’re going to give short presentation into what we call modsim on the 3D experience, then very shortly show what is possible today in additive manufacturing on the 3D experience and virtual prototyping and that stuff, then the main part of this talk is talking about conventional design, computational design for conventional manufacturing on 3D experience, and we’re just going to basically take two parts, so a sheet metal part and a casted part, and have a look how that actually is is going on from a computational design point of view, and at the end a conclusion, and a and a full small fun note, our Legacy.

So the so system started back in the 80s, actually Abacus, which is a find element solver, started in 19 78, so that’s the same year I was born, so we’re talking about technology with quite a bit of Legacy here, and not like some of the startups that that are probably only a decade or two old. Start out 3D design, then digital markup, where you basically have not just one part but assembly of Parts, like an airplane or car or something, then 3D PLM, which not Parts only, but also processes, so product life cycle management, then we’re moving into the 3D experience platform in the 201 12s. So here’s a a picture of of a of a city where you’re trying to to kind of Vis virtualize, I think it’s Singapore, Singapore, yeah, it is, Singapore is like a virtual prototype of Singapore, and we’re moving into like virtual twin experience for humans, so we bought medidata back in the in 2019 or something, and we’re moving towards the virtual twin experience for a sustainable world, I’m not sure what that means, but I hope this part is is actually on on the way to that.

The matsim technology is basically just combining modeling and simulation together in a common data model, and that’s in a single user experience for the 3D experience platform. So it’s basically for all of physics that we’re able to do in our simulation, we’re using mainly ktia Cad and simulation done by simia, so that would be Abacus or TSA, so that’s all what’s running here underneath. And the basis of this technology is that the the the old workflow is basically, you’re starting with some design, and then you usually change into other products to do meshing, and then you do execution, and then you postprocessing, and often that can even be multiple products at some companies, and that actually give takes quite a lot of time to do this whole process.

So the the goal of the mod Sim experience is, you start with a cad up here, and you select the cad, then you go into the CAE process, you do all of this, but all of that is on the 3D experience, and basically, when you go back, let’s say you’re doing optimization, so you do a CAE update, that forces then some changes on the cad, and then you can validate, and this Loop goes on and on, and we Target to reduce the time to simulation.

So now I’ll give the word to Pascal on additive. So we started generative design in the context of additive manufacturing and 3D printing, because of the shape that are produced by the topology ology optimization. So we started by launching topology optimization and giving some very nice shape, but very complicated to produce except by additive manufacturing, on which, by the way, we have the full workflow, including a ltis structure Creation, with, in the middle, some images of the mhes with solid element, shell elements, beam elements, as we saw this morning, but also with another brand from daso system, which is Delia, all the setup and the manufacturing simulation, until the the the printed part. But together with topology optimization, or for with other technique of optimization like shape optimization or bid optimization, we wanted, and we saw a lot of advantages to use that technique in other domains, like Peter said, for more conventional manufacturing processes, like casting, like meing, or bead for Sheet Metal, that we are going to show you now. Thank you very much.

So yeah, so we’re going to show these two examples, so it’s basically computational design for conventional manufacturing. So I’m first going to show bead optimization, so it’s basically just sheet metal model, and then a topology optimization of a solid part which is casted. So let’s say I’m in an automotive industry, and I’m probably not producing one of the the the just one of the cars, this is just a we call it blue car, it’s like a a generic car that we’re using for internal testing, and I have some design of some part, and the igon frequency for that is solow. The first thing I do is I do a naive bead design, so hopefully you can see here, there’s some small beads here, I do that’s an a good choice, but the ion frequency does not change that much. So what can I do again? Well, let’s run a bead optimization, I’m not going to go into much how that actually set up and work, but basically we’re ta changing the positions of the nodes here, we’re running the optimization, it takes like 20 minutes on a laptop, and we get this very curved design out of it. And actually, here you see the optimized design is a subdivision surface, so it is a cad part, but it’s not parametric in the classical part, the classical sense of the word. So the question is now, how do I actually get a parametric sheet metal part out of this, because that’s often what my other colleagues in the manufacturing actually wants. So Pascal, can you help me with that?

So we will now play a video that has been recorded from the the tool with. So we I hope we’ll play the video, so it’s it’s just at speed one, it has not been accelerated. So directly inside the optimization tool from, we can launch a construction, which is not to generate only a subdivision surface, but to automatically generate a cat mode, a c modelization, of the bit optimization result, and it should now run, but of course it doesn’t run, it used to run, I don’t know why it’s not going. Now it’s actually okay, sorry about. So we generate here a surface, a surface which is based on some ISO curve from the optimization, so you can choose your reference ISO ISO curve, and then we used the Sketcher capability from Kata in order to convert this curve into several type of sketches inputs. We have several strategy here, we are following a very complex complex geometry, but we can change that with several primitive. And afterwards we use all the C capability from the software, for example with the sweep, with the sweep managing the draft angle, we could have had fillets, and we finally have a part which is made thanks to the the result from the topology optimization, and which has full cat parameterization, so which is ready for design of experiment.

16:42 All right, so here are the the result validation. So basically, we have first the initial design, we have the let’s say the naive design, we have the direct result, which is is kind of wild and and let’s say bumpy. U, but then we have this really nice beat design, actually I I added here, we added here also some fillets, to to make it look nice, and and almost manufacturable. There’s probably going to be some manufacturing people who’s going to say this is going to be really hard to manufacture, so that’s definitely a discussion wor worth having, but just to showing you that that there’s a lot of possibilities here with conventional design as well. So that was beat up to my stage. Oh, and also I should probably mention that we’re losing a couple of Hertz in the IG frequency, which is quite typical, so when we’re going from, let’s say, the optimize design into the paramet parametric version of it, we are going to lose something, but it’s at least a a lot better than our initial design, and also much better than the naive design.

16:42 All right, so the next one is this casted part. So this is basically just an engine mount, it’s also a little bit of a generic model, you should imagine that on the end of this there’s there’s like an electric engine, and this Mount was was done, designed by someone in the in the company, and we’re looking at it and we fig figure out it’s not stiff enough. So what we do, well, we can just run to poly optimization, we want it to be cast, right, so we have many manufacturing constraints in in our software, so you can do ribs, different casting, and and and so on, maximum, minimum member size, Symmetry, and so on so forth. So here we’re choosing this rib constraint, so what you have here is basically this topology optimization, again, ran quite quickly on a laptop, so only like 9 minutes, 10 minutes, is probably should have used a much finer mesh to get finer ribs and stuff, but this is actually not the point of the talk. The point I want to make here is that, since many years, actually 2016 I think, we were one of the first companies to produce these Advanced subdivision surfaces from a topology optimization, which is nice and cool, and it’s a cad surface, I can actually do stuff with that, however it’s not parametric in the sense of what we usually think about parametric. So again, Pascal, what do what do we do with this?

So same ear video that I open will launch correctly that time. So here, all the idea is first to set the the constraint directly in the topology optimization, and then, in the Reconstruction afterwards, to reuse the cat cap capabilities from Kata to create a new kind of of CAD, which is not made of subdivision surface, but on classical modelization features, like sketches. So here you see the same kind of methodology, with strategy to approximate the the the iso surface, and some traditional pocket feature, like like Pockets or Boolean operation, just to provide a finished part to the end user, that he can freely edit, thanks to the the cat parameters and all the parameters that are been automatically defined for him.

All right, good stuff. So now we have a parametric model, and we can now do validation, and a couple of talks points here to make. So we have the initial designs, about 5 kilos, the maximum displacement is 7 mm. So when we do the Topol optimization, we’re actually gaining a bit of mass, the reason for this it’s a little bit hard to see, but, due to the big flat structure there, the subdivis I surface is actually adding a little bit of Mass on the whole middle section of the structure, which shouldn’t be there, so that’s why that thing is a bit heavier. When we then go into the Reconstruction part, we are down at the the the good Mass again, so we’re below the initial mass, and we see that our displacement has decreased a lot.

So the whole point of this talk was not to go into too much details about all the simulation and optimization stuff, in the so systems we can do much more complicated simulations, optimizations, we also have the capability to combine parametric optimization together with topology and others in in in single workflows, and we’re working actually also together with the the CST, who who got acquired by the so systems, to do Electro electrical magnetic optimization on like electric drives and stuff like that. We don’t have time to show all that, so unfortunately we are trying to stay within the time limits of today, and get to the conclusion, which is, I hope I gave you just a little taste of what mod Sim, so combining mod modeling and simulation in a single experience, or a single platform, and that it’s possible to do computational design not only for additive, but very importantly for conventional manufacturing, which is really essential for any big Industries building, like let’s say cars, planes, whatever things with large numbers of Parts. It’s really important for the industrial applications to have end to-end workflows, I think a lot of the colleagues from the other companies showed that as well.

So in my mind, the the result of an optimization is a new CAD, it’s new CAD geometry, it’s not just a density field or an ISO surface, it should be the thing that actually goes into your product, your car, your iPhone, whatever the thing is. And then I think non-parametric methods are really powerful, so topology, shape, bead, sizing, whatever, they can do a lot of cool stuff, but you need to have the right manufacturing constraints, and you need to have these cat reconstruction methods to be able to use them in an industrial setting. And then, at the last, I’m just going to give the word to Pascal, for yes, thank you.

So I would like to end by two points. So first is to thanks plan for setting up the workshop, the conference here, in this nice building from in Berlin, but also with a disclaimer. So this building has been built and designed by Fran G, which is a user of Kya since decades, and he used Kya to design his U is building so this one, but also theim Museum in bilbow, or the foundation Louis Von in in Paris. So be careful using that kind of software, you may also be inspired and design some very nice product like this one. All right, so thank you very much.