0:01 What’s up? Oh, what’s that? That was a super impressive building demo right before. My background is actually, I studied architecture here, just down the street at Pratt Institute, so that stuff always wows me. But thank you guys for having me. This is g to be a little bit of a different presentation. Just for a little bit of background, how many of you guys have heard of Enop before? We used to be called Anthopology, by the way, we lightweighted our name recently. So today I’m going to talk a little bit about our view on computational design, and then I have Andrew Kappers from Seaman Energy joining me to talk a little bit about their work in computational design as well. But for for the conference today, like first off, you know, why computational design, and like what are we actually seeing in the market across our customer base? And you know, when when I go on site with our customers and when I talk to them, you know, basically that we’re we’re hearing kind of the same thing across all of our customers, which is there’s never been more pressure to deliver products to market faster. And this is across aerospace, medical, automotive, industrial, consumer, and these are actually all products that our software is used on today across these segments, and I’ll talk a little bit about them.

1:16 But you know, when when we look into kind of what this bottleneck is to delivering products to market faster, the bottleneck that we’re addressing with End Toop very directly is that design iteration isn’t fast enough. And so you know, when you’re solving an engineering problem today, like just from like first principles, you know, you’re going to make something, test it, learn from it, and then you’re going to make it again and hopefully make it better, and you’re going to keep iterating until you get to a product that kind of meets the requirements that you want to, you want to get. And the bottleneck today is like how fast does it take to kind of turn around a new model, right? Like, because today, once when you make a model and then you learn something and you have to remake it, the the design process is basically limited by how fast you can recad that model, or redraw it, or recalculate that b based model.

2:08 And when you look at the kind of constraint here, right, like the constraint is how is manual iteration, and it’s really the kind of traditional tools and processes that constrain these type of engineering processes, and there’s really three three parts of that. First off, most traditional tools require like manual modeling of low-l features, and so this is a a screen recording of one of our customers making an airframe structure manually, like drawing a box, clicking on edges to add fillets, and the models recalculating. And like even if you could programmatically generate this grid, the the core modeling technology here, a boundary based model like we saw in the last one, if the if the curved segment was curved and fileted, it almost never regenerated because of geometric errors that tends to happen here.

Also, the second constraint is that it’s really difficult to set up the types of relationships between like physics driven problems. So this happens to be a manifold by one of our customers that makes rocket components, and it’s a manifold that goes from like a circle to a dis, and what I would want to do is I I would want to basically set up a function that controls the size of that cylinder that takes into account the topological change from the from the circle to the disc coming out, such that for some inlet pressure at the kind of inlet here I could basically size that tube to get some targeted exit pressure. Setting that up is really difficult, and then if you, you know, what I would do in a CAD system is like draw a curve, add sketch planes along the curve, have some circles, and you know, drive the circles with some parameter that I have, and then try and loft them together and just like pray and hope that that surface comes out, and most of the time it’s not going to.

3:50 And then lastly, the physics themselves are difficult to capture, right? Like the parts and components that our customers are building, they’re operating in environments that you, you know, are hot or really cold. There’s coupled thermal fluidic stress problems, electromagnetic problems. And so, you know, being at the Computational Design for Additive Manufacturing Summit, you know, the solution here is computational design. And and the way we define computational, I guess kind of obvious, right, but the way that we define computational design is literally a process of creating algorithms that capture design logic. And so instead of, you know, manually drawing the design logic in shape, we can use, you know, we can capture that design logic in some computer understandable algorithm so the computer can process the shape.

4:38 And so you know, let’s look at a much like oversimplified example: is like a NAA air foil, right? In NAA air foil, you have it parameterized, there’s a thickness field for the air foil and there’s a camber field, and there’s some equations that define that. So an Ent Toop, you know, we can capture those equations in the notebook, and then Ent Toop will compile that into machine code and compute all the different solutions of that of that air foil. And so instead of manually drawing a component like this, the computer is computing that, it can compute all possible air foil shapes. And so you know, with computational design, instead of, you know, being bottlenecked by manual iteration, the new bottleneck is like how fast your computer is and how many computers you have, and we know computers are really fast now and running in parallel.

5:27 And so this is kind of like an over simplified example, but you know, we could do something more complex, like this arrow structure, like a systems example, right? Like I have, you know, the outer mold line of this airframe is parameterized, I have a turbine location, some payload, some structure, some primary structure, a secondary isog grid, and I can leverage, you know, I could, you know, run through the different iterations of this rocket model, and for each design point calculate the fitness of that design point. This happens to be using the Intact Solution stuff that we saw presented beforehand, and then I can find, okay, well, what’s my model that actually meets the requirements that I need to solve the mission that I need to do, okay, that’s the design point that I can get. And so the computer then could allow us to iterate much much more rapidly.

6:13 And you know, I I see this as kind of like a a a new process for our customers, right? It’s not something that our customers are used to doing in traditional CAD tools, where, you know, they’re when you’re building a computational model, you kind of have to encode that logic first, right? So you need to capture what are the inputs, you need to encode the set of operations, and then you have an output, a step file, a solid model, maybe it’s some data coming out. Once you have that model built, you can then run that model and run through different design points, maybe you train up an AI surrogate model like we just saw in the last presentation, or some of our customers are doing that as well, and then that optimal design needs to be integrated back into the overall system.

7:00 And so what we do at End Toop is provide tools purpose built, or we provide the best tools for computational design, and we have three pieces of software. So we have nTop, which is our desktop software, and how many have, how many you guys have used nTop also, okay, handful, more than the last conference I was at, I was at the Seaman Conference last week, okay, yeah. So End Toop, you know, built upon sign distance fields, we have Automate, which, and how many of you guys are running Automate, probably not as many, just a handful, okay. So Automate is another product that we have, you could think of Automate as like the End Toop agents, or it’s robotic version of nTop that can run the models that you build in nTop, and then we have nTop Core, which is for our partners to integrate implicits, and I’ll talk about this a little bit after Andrew’s presentation.

7:49 And you know, what makes End Toop purpose built for computational design is our core modeling technology, and it’s, we call it Sequoia internally, and we basically represent shapes using their sign distance fields, and we basically compile the kind of analytical function that represents that shape into machine code, so that we can very quickly do evaluations on that shape inside of End Toop. And there’s really, there’s three differentiators, right? So it’s the implicit model at the core, which enables the model to be kind of super robust and not break, so those booleans don’t fail, but then there’s field driven design, so being able to control those parameters in the model with the physics or with the fields, and then there’s the process automation aspect of the block system in End Toop.

8:30 And so you know, you could build up models that are fairly complex systems, like a, you know, a concept model generator of a jet engine that has some parameters for the number of stages of the compressor, the number of stages of the turbine, and this, this is screen recorded from in real time, so you can modify these points and the model just updates and updates and updates. We also just launched some adaptive rendering, so that when you’re viewing these models and zooming in, you can represent multiscale features on the model. So in the combuster chamber you might have a texture on the model, and the jet engine model was running on a machine with like a really beefy GPU, like in Ada series RTX Ada 6000, this is just running on a laptop.

9:13 And so you can kind of see the evaluation updating, and again, you have control over that texture, and so you can take that whole model of the system and run CFD through the whole model to kind of understand how your, how your fluid’s flowing to the model, and then leverage that CFD data to control like the texture on the combuster or things like that. And this is using the new kind of Cloud Fluid integration, which I’ll also talk about after, but where we’re at today, we’re being used at a a little over 400 customers, and luckily we have one here. Andrew Kaer is from Seaman Energy, who can talk a little bit more about the work that he’s doing, so I’ll invite Andrew up, and then I’ll kind of come back up. So welcome.

10:00 Okay, hi, my name is Andrew Kappers. I’m a mechanical engineer at Seaman Energy, specifically in the additive manufacturing department. Do you hear that, okay, I will not touch a microphone. So just a quick blurb about Sean Energy and who we are. So in 2020, Sean spun off its gas and power division, and we became Seaman Energy. So we are a publicly traded company that is traded separately from Seaman AG on the Frankfurt Stock Exchange. So I just wanted to mention that, because there is a little bit of confusion, because our names are very similar, but we are different companies. And to further add confusion, we we both operate in the additive manufacturing space. Seaman is more on the the software side.

10:58 Yeah, so Seamans is more on the software side, like NX, myself, Seamans Energy, we focus on the serial production of additive manufacturing components for our core port portfolio products, which are wind turbin, gas turbin, and electric generators. So just a fun fact, our products are in power plants, and we, they’re used to produce one sixth of the world’s electricity. So we’re also, we’ve been doing out of manufacturing for for many many years, we’re both a user of AM and we’re also a service provider. We primar, we do Wham, which is why arc outed manufacturing, but our we primarily do laser powder bed fusion. We have 50, over 50 machines worldwide, many in the UK, Sweden, Germany, Charlotte, Orlando.

13:10 And what I like to highlight here is that we have over a 100 parts that are atively manufactured in serial production. Okay, so I’m here at CDAM, we have many applications that take advantage of implicit modeling. We have heat exchangers, air filters, containment vessels, expansion droids, sealing structures, print supports, and lightweighting features on on many components, and I tried to include a wide variety here just to show the breadth of applications across our product portfolio portfolio, and include real pictures where where I can show them, because I just want to emphasize, you know, implicit modeling is, you know, it’s out of the R&D space now, it’s it’s beyond that, it’s tested, it’s in serial production, and it’s in service.

Okay, so implicit modeling, a lot of people think TPMS, and then they automatically go to heat exchangers. This was a really cool design generated by a colleague of mine, Marcus Lumpy, in Germany. The unit sale on these TPMS in the heat exchang of core are really really small, and what I want to show with this example is that while the heat exchanger is relatively small, just the the number of fine features here to have the resolution to go to print, the STL file was greater than 5 gigabytes, and while that’s doable from a print standpoint, it’s really not doable, right, it’s it’s very cumbersome. And so the the print from an implicit that is offered by OV, it’s really just a it’s a game changer, the implicit file is six megabytes, and you know, using that and the the interop with the the EOS EOS print, you get that perfect resolution in the print, there’s no fating, the file sizes are manageable, and it’s just, you save a lot of headache from the from the engineering standpoint.

14:04 And like I said, this heat exchanger was relatively small, it was already pushing the boundaries of what we can send to the printer. The image on the right is a much larger heat exchanger of similar complexity, similar unit cell size, in fact you see it there on a build plate, that’s an entire, it fills up the entire M29, EOS M 290 build plate, so it’s quite large, and it’s just, you know, I’ll just say it’s not not possible to to print this from u, from mes geometry, so no se, not sure if 3MF could do it, this is, I still said 3MF, so do I get a cup, you didn’t say the context, but yeah, in this case, the what you see on the right is is actually a printed part.

14:55 There’s a white paper out there with End Toop where we use the the implicit interop functionality to make this happen with meshless geometry, and one other thing I’ll say about the the heat exchanger on the on the left with the the colors, this design was entirely done in End Toop using field driven design, so the it’s a, there’s no dedicated CAD to create the baffles, the baffles were created by fields that varied the wall thickness and the wall offset to generate those baffles, and also the fields were used to bias the walls to change the the cross-sectional area of the flow as it goes from inlet to outlet to maximize the heat transfer, minimize the pressure loss, as the the temperature and the density, the fluid is changing as it goes from inlet to outlet. So it’s a really cool, really cool part, entirely done with End Toop, and really only possible with field driven design.

16:04 I’ll go pretty fast through this one, this is just a, this is a different heat exchanger, and in this cross-section I just wanted to show a different way we’ve used field driven design. So in this case, in the in the in the gold is basically the the heat exchanger wall with that’s unmodified, so there’s no no incre local increases to the wall thickness, no fillets added, it’s just for comparison. So in the gray you see the actual heat exchanger geometry where I used fields to modify the locally the wall thickness as it intersects the casing, so on the sides and the bottom it’s locally increased so that you don’t have thin to thick transitions, and then at the top again field was used to further increase the thickness so that you could you could bridge the gap to aid in the manufacturing of this part.

24:56 And the the other thing I’ll say about this design is pretty cool, is that the unit cell was actually not created in nTop, it was created by a partner using a one of their proprietary algorithms, so it’s it’s a custom unit cell, kind of looks like like a brain, it’s very twisty and windy in a very random and non-uniform pattern, and I I had the task of populating the full heat exchange core with this custom unit cell, and I found no other way to do it, to pattern this unit cell, other than to bring it into End Toop, convert it to an implicit, and then I was able to pattern that implicit, kind of clean up the geometry at the interface between the unit cells, and then I was able to mesh it, mesh the fluid domain, and take to Star CCM for a conjugate heat transfer.

So this is probably the most simple design, but it’s my favorite. So this is a, this is an air filter where it’s it’s actually a legacy design, so there was no CAD, it was a 2D drawing that basically just, you know, gave a sketch that was, you know, would be revolved, different cross-sections, and then an area where 30,000 very small holes would be populated equally spaced apart. Traditionally it’s been manufactured through conventional methods, we had the idea of possibly printing this using LPBF or binder jet, but to do that you have to have have 3D geometry. So I had the task of creating this, some colleagues of mine tried to do this using traditional VRE CAD, I took it over, quickly regretted that.

So yeah, you can see the table at the at the top right, to generate the 30,000 holes took three to four hours, and it’s, there’s nothing fancy about it, right, it’s just you create a sketch of one hole, then you pattern basically along a curve 30,000 times, and it takes three to four hours to do that, and then to save the file took one hour, and then if you close it, God forbid you have to reopen it, it takes another hour. So I ended up doing this, I brought it into nTop, Ent Toop, your software, Brad, saved my sanity that day. So I brought in the, I brought in the the basically the blank 3D sleeve and did the whole perforation within End Toop, and it took 45 seconds to create these 30,000 holes equally spaced apart, save the file, 15 seconds, load it back up, 30, and again, like it’s it’s lightweight, so the file size is three megabytes versus 198 megabytes, and then for good measure I just, yeah, I meshed it and did a stress simulation with an Anop.

So this is an example of another heat exchanger, this is actually from a from an external company, they asked us to create, using additive manufacturing, a heat heat exchanger for them. It looks an awful lot like a conventional heat exchanger, you see at the bottom left, but there were enough tweaks and modifications to it that it warranted non-traditional manufacturing methods, so we we set out to sort of optimize this design, design again. My colleague Marcus Lumpy in Germany had this project, and here he took a submodel and he had four parameters to vary, and he used End Toop to, he created a sub submodel entirely in nTop, and had the functionality or sorry, had this automated workflow between nTop, Star CCM, and and Heeds, to was basically the wrapper to do the optimization, but the robust geometry creation and Ontop allow this to be done without many geometry failures.

Okay, so going back to, I am missing, that’s okay though. So this is something I’ve been thinking about, I haven’t actually done other than create some some fun images and animations, but another application we have is containment structures, where we will perform explicit dynamics calculations of to simulate like a like a rotor burst, and these structures have to, you know, contain that high amount of energy that’s being released. So I have, there’s not a slide here, but there’s another project I was going to show where we used, we basically used, if you imagine cylindrical casing, and we had, we in End Toop we added reinforcement webbing to it, and then we meshed it, took it to Ansis, ran an explicit dynamics calculation in in using Anis, autod, extracted from that a damage field or a damage point cloud, brought that back into End Toop, created a field of that damage, and then morphed that reinforcing webbing using that field.

So that is an example of basically adding material to improve your containment capability, and U, I’ve I’ve been wondering if, you know, it’s possible to instead of adding material intelligently where you need it, is it possible to subtract material and still have your containment capability. The answer is I don’t know, it’s TBD, but you know, it’s, you know, it’s it’s been postulated, I guess, that you know, you could, you have some structures, your diamond TPMS, or honeycomb, or lattice, or something like that, and possibly your specimen is, or your thickness of the casing is is increased, but your density is still less, so it could be a lightweighting, a lightweight containment structure. This is something I want to explore using some of these automated process chains that we’ve developed and explored, use between Antop and Ansis, it’s really the only way to explore this design space in a robust manner.

All right, so my last slide, this is a a turbine vein that we’re LPBF printing. You know, ideally you would put it on the build plate directly, but in this case it has to be oriented in such a manner, approximately 45 degrees, so it’s got to be supported, you see that at the bottom left. The easiest way to do it is just to create solid supports, extrude it down to the build plate, but all those supports, you know, the blue, the orange, and the purple, that’s just wasted material, that’s wasted time, and that’s that’s cost, that’s just going out out the door. So ideally you would, you would want your supports to be have all the functionality, but to be to minimize the mass, so when we think about minimizing mass, we usually, we typically, people go to a topology optimization, and if you’re, if you do a topology optimization considering structural, you you’ll think of something like you see the right, which is an idea we had and we printed.

So here the objective is to, there’s an objective to minimize the compliance with some addtive manufacturing constraints put on it, but it really doesn’t consider thermals at all, and if you think about the LPBF process right, it’s a welding process, so you’re injecting heat at each layer, and the layers below it are cooling, so there’s a thermal component to the the sports, they have to function as a as a means of dissipating the heat down to the build plate. And the the picture you see at the right, this particular support structure, while it did reduce the the mass of the supports by 50%, it really didn’t consider an entire functionality that the supports have. So the image in the middle, this was, we used Antop and Anis, there was, this is a thermal driven lattice optimization that was completed in Ansis, and then the geometry, this geometry here in the middle was created in Antop using this optimized LCE density field extracted from the Anis optimization, so it’s it’s a TPMS structure that’s uniform, but it has this density that came out of the optimizer, and it again has about 50% U of the support material.

25:46 So kind of extrapolating on that, we merged the two methods, and the image on the right, we ran a combined thermoelastic structural compliance objective optimization, and that’s what the image on the right is included, and we got an additional 8% mass out of the supports, and that’s all I have, thank you.

26:13 Andrew, super impressive. So by the way, if you guys don’t know, on 14th Street, if you if you’re driving up the FDR and you see the four large towers at the base of each of those towers is a large gas turbine engine that is made by Seen Energy, and they’re gigantic, they’re like the size of trolley buses or or subway cars. But you know, looking forward with nTop, you know, we launched nTop 5 June in 2024, you know, I looked at the first phase of End Toop really as targeting like point solution part components for 3D printing. With End Toop five, we re architected the core solid modeling technology to be in order of magnitude more precise, so we can handle not only larger scale engineering problems, but also we can connect more to partners.

26:53 And so in terms of our partner ecosystem, we’ve really expanded that, as you saw with Intact Solution Solutions and Cloud Fluid, but we’ve also released some integrations with CAD tools, C tools, and Cam, and then we also have the file compatibility, but in terms of implicit de cam, something I’m super excited about, and you can see over on the materialized table is a part by another company that makes really large engines, Villa, these happen to be diesel engines that power ships, and they produced the the cylinder head component in End Toop, which is really only possible for them to print now, because of the direct implicit connection to Materialize Magics, this is in version 28 of Magics for the SLM NXG machines, the large scale machines, and it’s going to be released next year as well, but we have a beta program that’s just recently started.

27:42 Implicit to CAD, this just shipped in July, in End Toop into Fusion, so you could export a do implicit file from End Toop, open it in Fusion, we had the 3MF committee meeting at our office the last two days, where we approved the volumetric and spec and the implicit spec, and so we should soon be able to do this via 3MF as well. And then we also just announced a strategic partnership with Nvidia, and so we have a demo version of this integration with Nvidia Omniverse, and so you know, the End Toop model can do a real-time implicit live link into an assembly model in Omniverse, so you can make changes to that model in End Toop, and then in real time kind of see that model update in the Omniverse. And so you know, you could basically, this is a prototype of how we could connect into other type of viewers, team viewer, or the PLM viewers and stuff like that, or you might want to do other stuff downstream, and Omniverse, other type of simulators and factory simulation stuff like that. And so with that said, Duann, thank you for having us, and appreciate it, thank you guys.