Being an Engineer
Being an Engineer
S5E39 Bradley Rothenberg | nTopology (nTop) & Computational Design
In this episode, host Aaron Moncur interviews Bradley Rothenberg about nTopology and its unique approach to computational design and engineering software. Bradley discusses the origins of nTopology and how it differs from traditional CAD tools. He explains how nTopology captures requirements in algorithms rather than geometry definitions. Bradley also covers nTopology’s applications in additive manufacturing and its tight integration with simulation tools.
Main Topics:
- The founding of Ntopology and Bradley's background
- Computational design approach vs traditional CAD modeling
- Implicit modeling technology and sign distance fields
- Applications in additive manufacturing and complex geometries
- Integration with simulation tools like ANSYS and LS-DYNA
- Process for exporting/importing models with other CAD systems
- Determining when Ntopology is the right solution
About the guest: Bradley Rothenberg is the founder and CEO of nTopology, a company pioneering next-generation engineering software for advanced manufacturing. Launched in 2015, nTopology enables engineers to create complex, optimized geometries, primarily for sectors like aerospace, automotive, and medical devices. With a background in architecture from Pratt Institute, Brad brings a unique perspective to computational design, bridging the gap between form and functionality in additive manufacturing.
Links:
Bradley Rothenberg - LinkedIn
nTopology Website
About Being An Engineer
The Being An Engineer podcast is a repository for industry knowledge and a tool through which engineers learn about and connect with relevant companies, technologies, people resources, and opportunities. We feature successful mechanical engineers and interview engineers who are passionate about their work and who made a great impact on the engineering community.
The Being An Engineer podcast is brought to you by Pipeline Design & Engineering. Pipeline partners with medical & other device engineering teams who need turnkey equipment such as cycle test machines, custom test fixtures, automation equipment, assembly jigs, inspection stations and more. You can find us on the web at www.teampipeline.us
Hi everyone. We've set up this being an engineer podcast as an industry knowledge repository, if you will. We hope it'll be a tool where engineers can learn about and connect with other companies, technologies, people, resources and opportunities. So make some connections and enjoy the show. You. Music.
Aaron Moncur:Hello and welcome to another wonderful episode of The being an engineer podcast today, we are privileged to speak with Bradley Rothenberg, who is the founder and CEO of nTopology, a company pioneering next generation engineering software for advanced manufacturing, launched in 2015. nTopology enables engineers to create complex optimized geometries, primarily for sectors Like aerospace, automotive and medical devices optimized for additive manufacturing. With a background in architecture from Pratt Institute, Bradley brings a unique perspective to computational design, bridging the gap between form and functionality in additive manufacturing. Bradley, thank you so much for joining us today.
Bradley Rothenberg:Yeah, thanks so much for having me on that podcast. It's awesome to be here.
Aaron Moncur:So I should have asked you before we started recording, but do you typically refer to your company as nTopology or just NtoP?
Bradley Rothenberg:So it's actually funny. I started the company's nTopology literally, the meaning means like N for any and then topology, which is like the mathematical structure of shape. And it was really because I wanted to make software that could allow you to make any shape. I didn't want shapes to be a bottleneck. And actually, about a year and a half ago, we acquired the four letter domain, nTop.com, after several years of negotiation. And once we acquired that domain, we shortened the name to NtoP, and I kind of like to say we lightweighted our name. Now we officially go by NtoP, but NtoP is short for Anthropology, so I'm okay, some of our social media handles are still nTopology.
Aaron Moncur:Got it? Okay, great, that's good history. So you have a background in architecture, tell us. How did you move from architecture to creating nToP?
Bradley Rothenberg:So I actually think the origins of nToP go even before architecture school, when I was in high school and obsessed with our CAD program. And so I had this CAD teacher from freshman year to senior year of high school that was teaching, you know, we were in Fairfield County, Connecticut, which is also the home of GE, and so GE would like sponsor all the local schools. And, I mean, I went to public school, but we had this amazing CAD program where we were learning computer aided design, and we had a 2d pen plotter, and I would always like finish the CAD assignments early, and I taught myself how to program through writing add ons to AutoCAD at the Time, and basically hacking the pen plotter to do all sorts of crazy things. And so, you know, when I went to college, you know, I thought the architects were using more interesting software than the engineers in school, so I decided to study architecture, and it was in 2005 when I was at Pratt that we I was working in the lab to the digital manufacturing lab, and the professor, one of the heads of the lab, was like, hey, you know, we just got this new machine. It's called a 3d printer. It can consume a file, and it prints it in 3d and so, like, who wants to manage it during the summer? And I immediately raised my hands. I was like, This sounds like the coolest thing on the planet. You know, I was really obsessed with laser cutters and programming them to make certain shapes. And when the 3d printer came for me, that was kind of the first connection between a lot of the programming I was doing to write plugins and add ons to CAD systems and actually manufacturing. And to me, the 3d printer was like, this game changing step where it's like, okay, you can model something in 3d and it was the first, like, true way of manufacturing a 3d model that was created digitally. And so, you know, I very quickly found out that the way that the current tools represented 3d shapes and 3d geometries was a bottleneck, and was holding me personally back from making these type of shapes and producing these shapes. And, you know, it's funny, because computational design was a term that we used in architecture school, and we talk about computational design a lot now at NT, which is literally like capturing the requirements for an engineering problem into a set of like computer algorithms. So we can use computers to actually compute and produce shapes. And I've always been obsessed with how we can use computers to help us create new shapes and create new products. And so to me, architecture school was kind of one step in that direction where I was like, Okay, we have these computers that and now computers that can print and produce 3d products. And there's a gigantic bottleneck with which was the the kind of state of the art cad tools were developed on top of technology that was, you know, created in the late 70s, early 80s, that was still kind of built on the fundamentals of drawing and drafting. And so instead of capturing requirements in a set of algorithms, the cad tools, which are phenomenal, were built to kind of capture drawing operations in a set of sequential operations. And so whenever you're producing something in a CAD system, you're essentially producing a set of drawing operations. You know, you make a sketch in ly, you extrude that sketch. You have multiple bodies. You're joining bodies and providing bodies. You're clicking on edges, you're adding filets. And those are essentially, to me, it's kind of the same thing as drawing and drafting. It's just using the computer. Using the computer to draw and draft. And what I was really interested in is, you know, we have all of this amazing amounts of compute at our fingertips. You know, we have GPUs with parallel compute. And, you know, how can we use that compute to actually create the shapes for us? And in order to do that, we need to capture those requirements that drive the shape, and instead of algorithms. And that's kind of why, that's kind of why I created nTop, which is to better use computers to help us create new products.
Aaron Moncur:Give us a tactical example of how nTop differs from, say, SolidWorks or Creo or any of the other more traditional CAD programs out there? Yeah.
Bradley Rothenberg:So, I mean, I think, you know, if you look at the work that Siemens energy is doing in our software, they're designing, you know, they're being so the products that Siemens energy makes are, like the biggest jet engines on the planet. You know, like when you fly, you see these gigantic jet engines. Well, they make jet engines that are, like 10 times the size of them. They're like 18 feet high, the size of school busses. And they power cities. In fact, I think there's like four of them, you know, 14 blocks north of me in New York City, that's providing the power to my apartment. And these turbine engines are running off of, you know, fossil fuels today, and one of the goals is to convert those to hydrogen power. And hydrogen power burns much hotter, and you have to control the flame. That's a much harder. It produces a lot more heat. And so in order to redesign some of the burner components and the heat exchangers, these are extremely complex engineering problems where they need to model these these shapes that basically pull heat out of one fluid and transfer the heat into another fluid. And in order to do that, they create these hundreds of 1000s of tubes that kind of weave around in and out of each other, and one set of tubes has one fluid, and another set of tubes has another fluid, and there's metal in between them. And the goal is to transfer as much heat from one material to the other with as little pressure drop, because you have to have a pump to pump the fluids through, right? And so this is a an engineering problem that you could, you know, if you were to do this in a traditional tool like a SolidWorks or NX or CATIA, you would sit there and draw kind of all of these tubes. Or you might create one tube and array them in a way, but you create a set of manual operations to build up these tubes, and you'd have kind of one version of the design that you would then run through simulation, maybe some CFD. You'd run kanji heat transfer and solve this kind of coupled thermal fluidic problem to understand the pressure drop and the temperature drop. And once you get your result, you go back to the drawing board, and you're like, wait a second, there's too much pressure drop, and so you have to start over again and kind of redraw all those tubes, or try and stretch them out a little bit. And, you know, each iteration essentially takes a set amount of time, and each step you're making a better heat exchanger. And I think, if you like, take a step back, engineering is like a problem solving process, right? And so you're you when you're engineering a new product, you know, you make a design, you test it, you learn something, and then you go back to the drawing board, and you make it better in nttc, you're capturing those requirements in a set of building blocks. And so we have this concept of building blocks in nTop. And so you might have a block in nTop that can take a 3d space and divide it up into multiple domains. And you might have some parameters for those domains, like number of tubes, the periodicity of a sine wave. Maybe. Or a or a curve going through the tube, or maybe you have some shells and some fins going through. And you have a, you have the definition of that 3d shape in a in what's called a block, and there's a set of parameters going into that block. And then you might have some parameters for some where some inlets go. You know, you might have two inlets for the hot fluid, two inlets for the cold fluid, and then two outlets for the hot fluid, two outlets for the cold fluid. And you might want to change that to three inlets, or four inlets or five inlets, and so you create this, let's say, like very parametric model, where all of the parameters or these variables going into the model kind of control or create output shapes, and the model is very robust to changing those parameters. So if you wanted to see, oh, what happens if I have 20,000 tubes instead of 10,000 tubes, that's a parameter that you change, the model just updates instantly, and then that model is also tied directly to the fluid simulation, so you can see and measure the fitness of each design point. And so, you know, if I use a more simple example, maybe it's even easier, like if you think about an airfoil, this is like the most simple thing, a wing right and a wing, NACA and NACA airfoil you have parameterized by its thickness and camber equations. And so in an nTop model, you can capture, you know, the equations for thickness, the equations for camber, and produce the output wing shape. And so you know, in a CAD system, you might draw the curve for that wing, and that represents one curve, but by capturing, like the entire requirements and the engineering problem in the end top model, you don't just capture kind of one version of that wing or one version of that heat exchanger. You capture all possible variations of that heat exchanger, or all possible variations of that wing, so that, you know, you can run a big design study automatically using computers to try and figure out what's the best performing wing, or what's the best performing heat exchanger, right? Because each design point or each set of parameters produces a design, and each design has some fitness, you know, is it better or worse? Is it better to have 20,000 tubes? Is it better to have 10,000 tubes? Is it better to use an airfoil with a lot of camber, or is it better to use a flat airfoil? Right? And so the power of these computational models is that you can iterate and make changes very, very quickly in real time. And instead of being limited by the amount of time it takes you to, like, manually redraw the shape, you're now limited by how fast your computer could compute the shape. And so we're trying to basically speed up the development time in order of magnitude faster than the legacy tools, just as the legacy tools sped up the iteration time in order of magnitude faster than like paper drawing and drafting.
Aaron Moncur:What kind of computer do you need to run NtoP effectively? Do you have to use a really high end computer, or is some of the computing done in the cloud. And it doesn't matter that much how high performance machine is.
Bradley Rothenberg:So I run and top on my laptop, and also I have a desktop at my desk that I sometimes remote into, or I'll work off of that. And like, my desktop, has a really fast GPU. It has, like, the newest eight or 6000 in it has a thread ripper and, like, 256 gigabytes of RAM. And my laptop is kind of just like a two year old normal laptop. And, you know, end top runs really well on both of them. Obviously, it runs a lot faster on the computer with multiple cores of CPU and the faster GPU. Yeah, what's what's interesting is, you know, these models in nTop are extremely lightweight, like, we have a para this, like, fully parametric model of a jet engine that's like a megabyte in size, or the heat exchanger. It's crazy. Yeah, it's totally insane. Like, a model like that, as a CAD model could gigabytes or gigabytes, you know, a heat exchanger that could be, like, eight gigabytes, and then top could be a megabyte. And the reason for that is our core modeling technology is based on what's called sign distance fields, or implicitly defined sign distance fields. And so instead of having to, like, store all of this surface data in memory and calculate all these intersections of surfaces, we basically store a really complex mathematical equation, and our software is really good at compiling that equation into machine code so that the computer could render and compute things about those shapes. Yes, and so that's what makes and that's what makes endop different and special and so fast is the core technology that we're built
Aaron Moncur:on. Have you ever considered or been approached about licensing that modeling technology to other CAD platforms? I bet that would be a huge advantage for like if SolidWorks could all of a sudden deliver files that were one megabyte in size.
Bradley Rothenberg:Oh, 100%. We actually have a product for that, also called nTop core. And so nTop core is a library that we've released for partners to basically read and write nTop data directly. And we haven't done it yet, but it's something we're considering, which is opening up more of the modeling API into nTop so that partners can start to build more complex applications on top of nTopcore like today. So like, the most common use case of nTop core today is for like, you know, materialize with magics incorporated nTop core into their software so that they could consume an end top implicit and then slice it for 3d printing in version 28 Autodesk fusion also just loaded. We just released a plugin together with Autodesk to read and top implicits into Fusion for the same purpose, build simulation and build processing to D printing. Soon we'll have cam and other types of CAD, like operations, what's,
Aaron Moncur:what's the modeling like in top is it similar at all to traditional CAD programs? Or is it completely different? So
Bradley Rothenberg:it's definitely not similar. You know, the way I like to think about it is you're constructing or building a set of relationships that produce the geometry, rather than defining the geometry itself. So like, when I use the term implicit, it's like, instead of defining and drawing a set of edges, you're kind of drawing a set of primitive shapes, and the edges happen at the intersections, and you could add filets and rounds at those intersections. But the modeling itself is like, it's, it's, it's different. Like, if you've used CAD for 25 years, you're probably going to open up end top and be like, Okay, this is really hard. But for like, for on the flip side, you know, we had a high school intern with us all summer, and he's used fusion before in his Robotics Club, FIRST Robotics Club. But, um, he picked up and top in two days, and he modeled and printed completely an end top a tri copter drone that he threed printed and assembled, and it didn't quite take off yet, but it's if he this was in six weeks of an internship, and I think if he had eight weeks, he would get it to take off. And that actually wasn't because of the mechanical system. It was because of the electronics, and getting those to work was more complicated that
Aaron Moncur:is super impressive. Wow. So it's not really like a traditional modeling system. It's more define the constraints and let the software build the model,
Bradley Rothenberg:yeah, or set up the set of relationships, okay, and, and what,
Aaron Moncur:what different physics does this work with? I mean, there's like, loads, vibration, heat, fluid, like, what are the core physics that your engine works with?
Bradley Rothenberg:So internally, we have our own stress, linear, static, stress solver, vibration and thermal and we've those are all like using traditional fine elements. So the end top model, you take the end top model, you mesh it, and then you create a volumetric mesh automatically, and end top and then set up like a traditional find element analysis. We recently launched a few integrations directly from entop where you could run mesh list simulations. So instead of having to mesh the implicit, the implicit is itself is sent to the solver, and then the solver does some interesting things with that, like puts points inside of it, or surrounds it with a grid, and can solve the physics over that type of data structure. And we did that with a company called intact solutions for their stress solver. And then we also did that with a company out of Germany called Cloud fluid for their their computational fluid dynamics, their fluid solved, their Lattice Boltzmann fluid solver. And then you could also do pre processing to set up models for other tools as well. And so, you know, the integrations, like, one of the big steps that has kind of moved the needle for nTop in the last year to two years compared to when we launched five years ago, is how integrated nTop has become into our customers digital thread. Because, like, we're in around 400 Accounts right now, and our customers use our software, kind of next to, or with all the mainstream CAD systems, from Catia to nx to Creo SolidWorks. And they're using nTop models and solving the physics in inside of nTopfor the kind of design experimentation loop, but then also in like star CCM and fluent on the fluid side, they're using us with ls Dyna on the non linear side. And so I think engineering software like it's not really just like a winner take all kind of market like our customers are using the best tools for the job always. And you know the way that nTop implicits can integrate into those tools has been really critical to us, to our customers, really seeing the value of of what being able to iterate like an order of magnitude faster can do for them.
Aaron Moncur:That's great. I'm going to take a very short break here and share with the listeners that the being an engineer podcast is brought to you by pipeline design and engineering, where we don't design pipelines, but we do help companies develop advanced manufacturing processes, automated machines and custom fixtures, complemented with product design and R D services. Learn more at Team pipeline.us The podcast is also sponsored by the wave, an online platform of free tools, education and community for engineers. Learn more at the wave dot engineer and we are privileged to be speaking with Bradley Rothenberg today talk a little bit about the handoff between nTop and other CAD programs. So for example, if I'm using SolidWorks for most of my CAD modeling, and I'm using NTP, how do those two applications shake hands? How do they trade files between each other?
Bradley Rothenberg:So let's say you're making you're working on an assembly in SolidWorks. Maybe you're making a, you know, maybe you're like our customer, Ocado, and you're designing a rogue factory robot, and you have the kind of assembly that you're you're doing your kind of assembly itself in SolidWorks, and you might want to do some optimization on some of the parts. So you might do some D like, the most common application of nTop today is, you know, detailed level part design for 3d printed parts. And so you might have your, you build your assembly and kind of set up, okay, here's where I want a motor to sit. Here's where I have a gearbox. Here's where I have some tubes coming together. And in SolidWorks, you might have a region that you want to opt you want to create a part in, and you don't necessarily know what you want that part to look like, but you know it has to fit within a certain area, and you have kind of keep out zones and stuff like that. So you might export that as a STEP file, or save the SolidWorks part file for that component, and you'll import that into nTop. You'll set up your problem in nTop, set up your loads, your boundary conditions, and you'll run your problem in nTopand produce a component or a part, and that part you can then export as a STEP file and bring that back into SolidWorks as part of the assembly. Got it. Now, you might also do the analysis from the nTop model itself as well. Now, our customers that are kind of using like PLM environment, like wind chill or team center next to Catia or next to NX, you know, they'll manage that end top file directly in the PLM system, and for each change that happens upstream event top that will kind of automatically get fed into nTop and that can be updated with the design process. We also have a newer set of applications, not at the detailed part design level, but much earlier in the process, in the kind of low fidelity concept design where customers are starting in NtoP and creating a model. It's more like a low fidelity systems model in NtoP, where they can iterate really fast. And what things I'm thinking of are like, you know, drones, unmanned vehicles, smaller robots and stuff like that, where a model is built in NtoP that could take in some of the initial requirements. Maybe you have some payloads or gas tanks, and you have an engine, and you want to see what happens if I change where the engine is, what happens if I move the wings around and very quickly, you can iterate through like 1000s of different variations of a kind of low fidelity concept model. Once that model is honed in and meets the requirements that you want, you can then make it a STEP file and bring that STEP file into SolidWorks or NX or Catia or. Have that as a starting point to then do all of your downstream detail level design from. And so that's kind of a new area, but, like our bread and butter, has been in the detailed part design. Okay?
Aaron Moncur:And you mentioned that the biggest use case for NtoP right now is in additive manufacturing, 3d printed parts. And I'm assuming that's because I'm no expert at all when it comes to end top, I've seen parts that have been made there, and they seem to have this very organic shape and structure to them, almost, I feel like the word anatomical almost applies. It's almost like this, like network of vasculature, different veins and lofted surfaces, weaving throughout and so I'm guessing the Reason 3d printing has been found to be such a successful application of the technology is because you can make these crazy shapes in printing that you just can't make if you're injection molding something or or machining something. Is that a true statement?
Bradley Rothenberg:Yeah, I mean, I think that's, you know, for those type of design problems, NT is kind of the only technology that can produce those type of shapes in a robust, fast and easy way. And so, you know, early on in our deployment, we saw a ton of those type of components, and a ton of, like, heat exchangers, anywhere where, you know, if you look at these type of models and think about having to, like, draw them in a CAD system, it would drive you kind of nuts, yeah? But there's no, actually, yeah. There's no way, like a heat exchanger with, you know, hundreds of 1000s of tubes, or trying to optimize where the composite, you know the orientation of composites through automated fiber placement machine or an automated tape layout machine, and so things like that where you're dealing with like very complex geometries and top core technologies really, really good at solving
Aaron Moncur:so If I understand correctly, if we look at an example where nTop is solving a structural problem, you've got some part that is bearing loads on, on, let's say a top, the top, on the bottom. And so you apply these boundary conditions within nTop, you've got loads here, you've got a ground there. And top starts doing the the FEA, the solving, and it identifies areas in this block of material, this theoretical block of material with which you start that just aren't needed for the structural support of that component. And so it strips those areas away and kind of just leaves the areas that are required to support those loads, which is why, at least these models can end up looking like, I'm using the word anatomical, I don't know. I've always had that thought that they look like, like anatomical models because they're so organic in their their shape. Is that, more or less how it's working.
Bradley Rothenberg:I think that's yes for when you run the kind of density based topology optimization. Okay, and so there's the algorithm that's in n top, called the density based topology optimization. And basically what you're doing there is you're taking, like a block of material, applying some loads to that block of material, and the algorithm is removing material where it's not needed. So essentially, what you get is like a kind of a tracing of where the load paths are, right? Yes, yeah. And so that's a fairly common use case of nTop, and our implicit modeling tech is really good for that. But you could also build up models that you kind of build in, like, one of the things with those is, you know, if you can build this free form shade, can you actually manufacture it with the traditional manufacturing process, not just 3d printing and so what we also you can build in some constraints to the topology optimization, so that it's, you know, you make sure that if you're machining it, the machine could always the tool could always, like, get at the part in each direction. But you could also build up a model from, kind of, from, from the beginning that takes into account the manufacturing constraints. Like, if you're building an injection molded component, you know that injection molded component can? You know, you can make sure all the walls are always the same thickness, and you're not changing cross sectional area too much. And you have a set of ribs that are in certain areas, and you have a set of stiffeners. You might have an A side surface that you want to maintain, but the B side surface doesn't matter, but you just can't hit other certain regions. And so you can build those in model that's not necessarily just a running a traditional, topology optimized model. And so I would call that kind of like a parametric model in a top that's very robust and very quick. Okay.
Aaron Moncur:Yes, that's very cool. I didn't realize that it had that capability. I thought it was always this, like, very organic shape that got exported. So does that? Does that infer, then that you can set up your model so that the export, if you need to bring it into another platform, like SolidWorks or Creo or whatever, you can export models that have basically flat faces on them so that you can do additional modeling operations if you need to on, you know, nice, clean flat faces.
Bradley Rothenberg:Ah, you can absolutely. And some of our like mesh to step filing is getting really good at identifying those flat faces. But my question always is like, what additional modeling Do you want to do in the CAD system? Like, why would you not do that in n top itself?
Aaron Moncur:The example that came to mind was like, like, draft on a part. And maybe you can just add the draft in nTop, and you don't even have to worry about that as a post operation.
Bradley Rothenberg:Yeah, draft is a block and end top that you can just like, apply. You know, you have your plane that you want to drop from, okay, sponsor, and it's actually all it's it's pretty automatic. So that's kind of a nice that's awesome, and the nice aspect of that, but I think there's this, a lot of people who use CAD historically have certain things that they know work, and they have certain ways of, like, kind of hacking the CAD model to get what they want. And so a lot of times when I'm talking to customers, their first thought is like, Okay, I want to bring this model back into CAD so I could do add draft, or I can add a whole here. And they don't realize that actually adding that still wallets and implicit might actually be faster and better. Yeah, and, and if you're adding draft, then, like, you know, you can do the analysis on the drafted part you don't need. You're not doing dry. You're not doing that as, like, a post processing operation.
Aaron Moncur:If you wanted to. Are there ways that you can add discrete features, like, if I wanted to add a hole that was point 782, inches in diameter and point 3325, inches deep. Is there a way to do that?
Bradley Rothenberg:Oh, absolutely. I mean, you everything in the model is defined through a set of parameters. So you can make a you can make a cutting tool that has those parameters and cut that out from your part. Very good. So, yeah, like all, all of those type of things you could do in traditional mechanical parts like you and you're making a mechanical part, you can, you can set that up and but the cool thing is that hole that's point 332, inches with, you know, plus or minus whatever tolerance you define. You know, if you run the analysis and realize, wait a second, there's too much load at the bolted connection there, maybe you want to make a change and try a different size hole. You can just update that hole, and the whole system will just write, similar to a parametric model. But a lot of times, when you build a parametric model in a CAD system, you make an update and the model won't update, or it takes a long time to update an end top that updating is like instant. I
Aaron Moncur:feel like FEA is typically considered a separate process than CAD modeling, right? We might create our model in CAD and then maybe you've got solid work simulation, so you're at least in the same environment, or maybe you don't have that, and you export it and run something in ANSYS, but it's like a separate operation. I get the sense that in nTop, it's not really a separate operation to run the analyzes. It's more of a baked in, like, just part of the workflow. Is that the right way to think about it?
Bradley Rothenberg:Yeah, I mean, I think of simulation. And this is like a lot of organizations are set up where they have two different teams, like they have a draft, a drafting team of draftsmen and a team of analysts, yeah. And so the way I think about simulation is it's just one step in the engineering problem that you need to solve, and you need that very tight digital thread or very tightly coupled simulation to the model. So like, if I make it like, ideally, what I want in the models, I want to make a change and see, okay, what happens if I pick in these little ribs, or what if I use, you know, five degrees of draft instead of six degrees of draft? How does that impact the performance of my part? I'd love to just make a change and see it in real time. Now, we're not there yet, because the physics still takes time to run, but I think we'll be there within the next five years or so, but I don't, I don't know if that I know it's, it's, it's, that's it. Getting it in real time is going to be tricky, but I think some of the getting it in real time and accurate is going to be tricky. Um, at least being directionally accurate. I think at that stage of design is probably good enough. And then you can freeze the model and let it run like I have a model on my before our sound cut out. I had a stick I was talking about. I had a model on the desktop, on my desktop machine that's running, and I'm using end top automate to run through 100 different design options where I'm just doing a parameter sweep on the wall thickness of an injection mold. Part. And I'm trying to understand for it's for a customer of ours running one of their models where, like, what's the impact of the wall thickness to the weight of the overall component and the max stress throughout the component. And trying to hone out, hone in on what the wall thickness is. My computer's doing that work right now, and I'll get back and I'll go see that. It'll probably be done in a couple hours. And the thing that takes time there is meshing the geometry to prepare it for simulation. And it's automatic, but it's still the computer has to crunch away and figure out all these little elements, and then it's solving the actual partial differential equations over each of those elements to understand where the stress is in the part. And so if we could, if we could speed that up and make it accurate, I think that's kind of, you know, where I where I want to go with this. And I think we're close. And I think I don't think of the simulation as something separate that's just done to, like, validate the part like the simulation. It's a way of understanding the fitness of one version of the design. And in a world where you're constantly making changes and constantly updating your design to get the performance that you want out of it, each time you make a change, you should see, okay, is it getting better or worse? Yeah,
Aaron Moncur:in your case, in n tops case, the simulation directly leads to and is directly linked to the the final geometry of your part. Whereas in traditional modeling, you you model your part, you know, with a rough engineering understanding of what it should look like and how big it should be, etc, but you're, you're often not optimizing that part based on any sort of feas. It's usually just kind of engineering judgment, unless it's some kind of truly structural critical part. And then, of course, you could get into FEA. So maybe, maybe an interesting question to ask would be, what, what kinds of operations or actions should you not do an end top within the context of modeling and creating part geometries.
Bradley Rothenberg:Today, actually, setting up assemblies in nTop is really not so easy to do. Like, we don't have all the constraints and all that stuff and then sketch we don't have a good sketcher yet. So, like, actually, I still think it's too hard to like define the geometry itself, like you could set like for certain models where you know you need to iterate faster, you know you need to reuse them a lot. So like, heat exchangers are a good example, where I do recommend, like modeling the whole thing from the beginning and end top. But if you're modeling like a simple bracket or, like, something that a sheet metal part, like, that's better for a CAD system. Like, if you're doing, you know, if you're drawing where the tubes should go in a part, maybe automated pipe routing makes sense for nTop. But if you're just, like, going and drawing some tubes, that doesn't make sense. And a lot of the mechanical design, probably like 80 or 90% of the mechanical design that's done today still makes sense for a CAD system. But there's certain examples, like designing the bulkheads of an aircraft where it's like the detailed design, or kind of the high fidelity structural layout where the bulkhead should go in an aircraft system, which are very good problems for nTop.
Aaron Moncur:What are some questions that an engineer should ask him or herself when trying to determine whether nTop is the right solution for their particular design problem?
Bradley Rothenberg:So I think the good types of questions to ask, are like, is this the type of problem where, if I spend a little bit more time upfront, defining the sets of relationships that go into this design, it will actually produce a better design longer term by exploring what those relationships do? Or like, you know, is this the type of model that requires lots of different changes in the structure or the topology to find the right design, and those changes are hard in the CAD system. You know, if I make a change in a CAD system, I have to wait 30 minutes every time I make a change for the model to update, and there's like a 50% chance that model will fail when it updates. That's probably a good problem for NtoP, you know, is this something that has, you know, hundreds of 1000s of surfaces that are all repeated, and they're all a little bit different, but they're kind of the same. That's probably a good problem for an implicit we defined sign, distance, field based model. But I think the main thing is, like, do I have through the design process? Am I going to be making a lot of changes, and those changes are, are hard to do in a CAD system and very manual. That probably seems like a good problem for for end top, great to to, to be to, for you to do an end top. Yeah.
Aaron Moncur:Okay, great. Well, Bradley, I think we'll wrap things up here just a second before we do. Are there any other questions or topics that we should hit on that we haven't talked about yet?
Bradley Rothenberg:I think we probably could talk for like, the whole day on computational design in general, and like, how you how you build these models, the type of models that you build. I mean, there's a lot of, like images, and we're engineering, so it's good to kind of see things. And so there's a, there's a video on YouTube that we made recently called the powers of nTop that we made with the launch of N top five. And I, you know, I see n Top Five is kind of opening up this new realm of problems that computational design can solve, because what we did is we re architected our core model to be an order of magnitude more precise and an order of magnitude faster, so that we can be useful for much more ambitious, larger scale problems. And I think this video kind of demonstrates that with the kind of drone model, the concept design for a drone, and then, and then an engine system as well. And there's other types of like GIFs and videos of these type of models that I think kind of demonstrate the power and robustness of these implicitly defined models. So you can, like, make changes really quickly and learn from those changes and and then I think that the following us on social media also is obviously where you can find us and see what we're up to.
Aaron Moncur:Terrific, terrific. You're the CEO of the company, so you don't want every Tom Dick and Harry reaching out directly to you, but if people have legitimate causes to do so, what's, what's the best way to get in touch with you and and also the best way to reach out and learn more, in general, about nTop?
Bradley Rothenberg:Well actually, you can reach out to me on LinkedIn via messages or shoot me an email. Brad@ntop.com It's easy. It's four letters at four letters. So it's brad@ntop.com and I'm pretty responsive as well. I'm on Instagram as well. If you do, you know, send a DM, usually I'll, I'll get back to you. And, yeah, I mean, I think we're pretty, like, open, you know, we're a new piece of technology, and one of the things that's really important for me is to spend as much time as possible with our the cut, with people that are using our technology. And so I'm always on the road, and so hopefully I'll we'll have some more kind of meetups as we're traveling again and and visiting our customers across the world. And so the other place to go is ntop.com and you can sign up for a demo. You could see our learning materials or support pages. Something I just want to make sure the listeners know too is that you know we're free for students to use. And so if you're a student, you can go to ntop.com go to the edu tab, and you can sign up to get n top for free. And that's super important for us, because that's where you learn new software, right? Like, if you're working for Honeywell, Honeywell is not paying you to learn new tools, and so think it's important to start with nTop in college.
Aaron Moncur:Phenomenal. All right. Bradley, thank you so much for your time. Congratulations on starting NtoP all those years ago, on the success that you and your team have seen, and thank you for being on the podcast today.
Bradley Rothenberg:Yeah, thanks so much for having me, and we'll talk soon.
Aaron Moncur:I'm Aaron Moncur, founder of pipeline design and engineering. If you liked what you heard today, please share the episode to learn how your team can leverage our team's expertise developing advanced manufacturing processes, automated machines and custom fixtures, complemented with product design and R D services. Visit us at Team pipeline.us. To join a vibrant community of engineers online visit the wave. Dot, engineer, thank you for listening. You.
