0:01 Okay, it’s good to be here today. My name is Matt Sederberg, and I’ll be talking about how much more simulation could you complete if you didn’t have to mesh. Okay, so just to introduce myself, before I get started, my previous life, I had a company called T-lines, which introduced a new way to make CAD models to a number of industries, including automotive design, architecture, industrial design, or consumer products, jewelry. We commercialize this technology through plugins to Rino1, Solid Works, and then a little over a decade ago Autodesk acquired this, now the tels technology is available in Fusion 360, and then Inv Ventor and Dynamo. And I left Autodesk about eight years ago to start a new company with some co-founders, to introduce a new way to run finite element analysis, and that’s what I’d like to talk with you about today.

1:00 And before, I guess, to motivate that, I’d like to talk about two things that happened in 1965 that brought me here today. The first one was a book released by, ra, a young lawyer named Ralph Nater, called Unsafe at Any Speed, and in this he opened a lot of visibility into the lack of accountability for safety in the automotive industry. It opened with a memorable line, for over half a century the automobile has brought death, injury, and the most inestimable sorrow and deprivation to millions of people. That got the attention of the American public, this became a national bestseller, got the attention of Congress, Ralph Nater testified before Congress about what was happening in the automotive industry, and this led to the institution of the national highway, tra, traffic safety administration board, and eventually to the requirement of automotive crash tests before cars could be sold in the United States.

1:57 So this is what these crash tests looked like back then. We don’t like to smash these automobiles, but this is the only way to simulate the actual conditions which can establish the design criteria. So, believe it or not, that really helped, and today we’re, about the likelihood of us dying in a car crash is one5 what it was back in 1965. So, so for a long time that was the only way to improve safety. But interestingly, back in that same year, 198, 1965, the beginning of a new method that would eventually improve safety happened, and that was the development of the first mainstream finite element analysis code, Nastran.

2:45 And so what finite element allows you to do is to digitally test what you could only previously test physically, to enable better design, manufacturer of many products. So finite elements progressed to the point where, decades later now, it can be used to predict automotive safety. So here’s an example from LSD, which has been the gold standard in FAA simulation for automotive, you can see, you can actually predict much more with FAA, because you can instrument it, you can repeat, you can, it’s cheaper and faster. Just, FAA has really benefited a lot of industries, and now it’s A7 billion dollar a year industry.

3:26 But increasingly, over the last few decades, there’s been an emerging elephant in the room when it comes to the workful for finite element analysis, and that has come about because of the emergence of CAD, for computer AED design. And when FAA started to be used in the 1970s, there was no CAD, and so when you’re given a design to simulate, it was just part of doing business, you had to build an FEA model using the math that was used in all the FAA codes, these lran elements, you can see that those tend to be faceted, U, representations. Once CAD, and sorry, as as CAD came into play, then there began to be a much bigger pain in setting up the these finite element models.

4:17 So Core Form, for the last decade we’ve been interacting and interviewing hundreds of analysts, and we’ve come to realize that, in the terms of Ralph Nater, for over half a century, preparing models for FAA has brought pain, tedium, and inestimable sorrow to hundreds of thousands of analysts. So again, the the issue here is is the connection with CAD. And so when Autodesk, when AutoCAD was finally introduced in the 1980s, again, not so much pain there, because it’s just a 2D draft, you create the finite element model anyway. But beginning the mid 90s, when Solid Works came to the stage, in in Ktia, and NX and Proe were more mature, now the analyst, instead of being given a a paper drawing or a 2D drawing, he was given a fully detailed 3D model that some engineer had spent potentially months preparing. And his task was to undo a lot of that work, to remove a lot of details, remove a lot of fillets, change the math representation, that can take weeks in complex analyses for automotive and other industries.

5:21 No one really likes to do that. There’s a lot of engineers I’ve met that have started to enter the field of FAA and then left, because it’s really demotivating to do this job that has no business value. There’s no reason to spend a good chunk of your life just reworking data in another math representation.

So about 20 years ago, a group of leading FAA researchers decided that, again, given, given this current state status of of how FE is just not integrated into to CAD, could there be a better way, could there be a way of, rather than spending up to 90% of the overall time, dis, solution, in just reworking the data, and then running a simulation on something less accurate than the CAD, and then having it be difficult to go back to the CAD, because there’s no straightforward way to add the insights. And the whole point of running analysis anyway is to improve the design, but if it takes so long that there’s no time to improve the design, it’s just this past, fail, at the end of the product design process, we’re really underu, I, izing the value that could be driven from finite elements.

So people have tried to come up with new cat representations for various reasons. The interesting thing is, though, that of all of the new cater presentations that have come up in the last recent years, I mean, ENT Toops would been one that’s been mentioned today as well, other, as well as others, no one has picked finite element math as the basis for creating new cat software. And the reason is because finite element math is good for really nothing besides finite elements.

6:57 So there’s, to to pick your poison with how you represent finite elements, you can either use hex meshes, which are the most accurate and performant and robust, but the only problem is it’s a very manual process to create these, at best it’s semi-automated. And and so that’s kind of on the one side. The other side you can make use tetrahedra, which can be automatically generated, but they can be very expensive to compute because of the element quality. So to get the automated ability to add these T element elements to fill up a volume, you sacrifice the the aspect ratios, you, the jacobians are are much more poor, and that just leads to a less predictable and a longer solve time.

7:41 So since no one’s picking FAA math to make CAD, what if we flipped it, and said, well, what if an FAA code could be created based on the mathematics of CAD. So that’s what we’re doing at Core Form. This is part of a larger field called isogeometric analysis, that these papers were first published about 20 years ago, there’s been over 4,000 papers now published exploring the idea of using CAD math as a basis of FAA. And it turns out that these higher order smooth spline basis functions that are used to define CAD are actually more accurate, more robust, and better behaved than the traditional Lrange elements used for FEA. And so it, it’s really creating a new, a new solver, that’s based on the super set of the technology that was first implemented in n in 1965.

So my business partners in cor, at Coreform, and I have created this new solver called Coreform Flex, and this drastically accelerate simulation workflows. And so we just take the CAD part, we prepare it for simulation through Core Form Flex, and then we run the simulation on the fully detailed CAD geometry. The way that we do this is we leverage the trimming capabilities inherent in the CAD. So what trimming is, this is the basis of modern CAD, CADs comprised of analytic surfaces, or nerve surfaces, that are rectangular, so to create something that’s not rectangular, you just kind of cut it out, like cutting out a piece of paper, and what you have left is what you use as that part of the CAD. So to create a more complex CAD model, you just cut out a lot of pieces of paper and stick them together, and and that’s CAD.

9:19 But the one thing is, even though this is called solid CAD, it’s actually hollow inside, and so you can’t use that for simulation. So if you want to run a simulation, how are you going to parameterize the volume? And so we just extend this concept of trimming to three dimensions. And so we take the CAD, we we immerse it in a volumetric, re, rectal linear, high order smooth spline mesh, and we just trim out that mesh with the CAD, and then we have an interior parameterization that we can use to simulate on. And this really, what this nets, is just a reduced human cost of simulation.

9:58 So there’s three key benefit, benefits that are available now through Coreform Flux. First is the ability to have fully automatic model generation, second is to be compatible with all CAD data, and then the the kicker is that this also gives you superior accuracy even for advanced nonlinear physics. I mentioned earlier that with simulation, part of the pain is when you’re running nonlinear simulation, especially with the automotive and defense customers that we work with, the vast majority of their time is spent just preparing data to actually run the simulation. That makes it difficult to automate simulation, makes it hard to do design iterations.

10:36 And so by driving all of that data prep, which again has no business value, when you come in from the outside it’s completely surprising that so much of time is just spent representing something from one math type to another in a non-automated way, by driving that down to zero, then simulation can be much more accessible and integrated into the design process. So anecdotally, as we’ve, as we’ve done problems together with customers, we save about 50% cost savings for project, and that huge savings is in this upfront labor time, where we just drive that mesh creation time to zero.

11:13 This is the version one solver right now, and so there’s parts that we’re still optimizing, for instance, our model setup, we don’t have two decades worth of pre-processing polish we’ve added yet, so that’s a little bit more expensive today. Our simulation code isn’t optimized yet, so that’s a little bit more expensive. But even so, that time saved that you don’t need to mesh is so significant that even today there’s significant time savings.

11:37 Given the chance to write a new solver, we wanted to avoid the mistakes that came in the past with people not talking to each other. And we’ve generalized the solver so that it can work with any CAD type, where you can define it, where you can run an inside outside test for a given point. So B reps aren’t watertight, but within a tolerance you can tell if any point is inside or outside, that the same can be said with implicit definitions, surface meshes. And so today we’ve been focusing on on this, running, working for traditional CAD with v-ups, we have completed prototypes with implicit definitions and surface meshes, and one of the reasons that I’m here is to talk to this crowd and see if we can accelerate, get some more customer demand, to push those those parts of the workflow forward as well.

12:24 And lastly, the way that Corin Flex differentiates from other new simulation code is is the superior accuracy in automation. For as long as I’ve been in the industry, there’s been different codes that have come up, focused on having a better simulation workflow for designers, to push simulation more up front, to give you insight early in the design process, and there’s absolutely a place for those in product design workflows. But where those fall down is, then, when the analyst needs to verify that everything’s actually correct, analysts don’t trust those tools, and so these have to reun everything in Abacus, or or Diner Anis, anyway. So by coming up with a code that is a true finite element method, that’s been vetted by thousands of research papers, that we can demonstrate.

13:12 So here’s just one example, where, with this, with this simple C bracket simulation, where we can press down on a ball bearing, the, we can converge faster than traditional P1 linear that are still the standard for nonlinear, linear dynamics. We get higher accuracy PR degree freedom, the robustness is superior. So we get all those benefits you usually see what the hex elements that I showed with finite elements, but also the speed and automation that everyone uses Ted elements for. And so finally, we do have a path forward where we can get both the accuracy and the automation in a single element definition.

So we’ve, with that background, the question is, well, how much more simulation could you complete now, with with Core Form Flex, if you didn’t need to go through all that manual preparation time each time you wanted to run a simulation. So today, so Coreform Flex is a a solver, we also have a pre-processor, and the available physics commercially is linear static today, but we have a lot of beta nonlinear capabilities that our customers are exercising as well, and I’ll just share some examples with you about about what you can do today.

14:24 So to start out with a linear static example, the examples I’ll, I’ll be sharing are more geometrically complex, because that’s where Core Phone Flex really shines, the the higher geometric complexity is just as easy for us as the low geometric complexity. So unlike with meshing, where the more small details you have, it, the mesh can just explode with degrees of freedom, with with Core Form Flex that’s not an issue. And the way that we can control how accurate, or we can get insights in the different quantities of interest, is just by the resolution of the background grid that we immerse this in, so that can, that’s defined on the part level, and so with different parts in the in the assembly, you can determine which you want more resolution in.

15:09 For more serious application, we can run contact, en large information, this is the trailer hitch used to trans, transport missiles, and so it’s it’s important that this behave as intended. And again, with with the Core Form Flex, we, this is a project with in con, in conjunction with C Master Lab, we’re able to have high accuracy contact and large deformation, even with these complex geometries.

15:35 A couple of examples for addtive manufacturing for this crowd, we’ve been working with RTX on demonstrating being able to determine accurate stresses on additively manufactured finstock chambers, and, as well as, here’s an implicit dynamic simulation with contact. Then, together with Honey Wolf, Honey Wolf, for direct incite pads, the issue here is that even something that seems like simple geometry, when you try to put a hex mesh on this, that can take hours of work, because you have to partition it to get a nicely behaved hex mesh, the minute you change the diameter, you repeat all that work. And Honey Well was given the chance to design these, the direct inkrite pads, but you can’t, they could only design things that they could simulate, and so if you can’t simulate that really narrows the design space. And with corn form flx, that, it, that enables the full automation of exploring many types of designs.

16:34 An example of large deformation, Ser, is a fux cable bending, this was used in in a missile, and when you, when you put a fux cable in, that you want it to be good for decades, to not malfunction, and so they need, they need to accurately understand whether anything was going to delaminate after they bent it. This is a surprisingly complex hex meshing exercise, this is for simplification, we had it be this thick, but there’s many elements to the thickness that can really explode in the, in how expensive that is to create and compute. And with Coreform Flex, even with these thin, these thin wall structures, we can have, we we can have less than one element through the thickness in our background grid. So it’s a very efficient way, even with things that you traditionally do with shells, with Cor Flex you can do those as a solid as well, without needing to get a mid surface.

17:26 The last example I’ll show, just the yielding of metal parts, again with with Coreform Flex we can do all sorts of linear, nonlinear plasticity, if we haven’t implemented it yet, it’s because the customer hasn’t asked for it yet. So I’ I’d welcome any any requests from from any of you in the crowd. The reason to come to places like this is is to meet and to understand what applications there may be for for what software you might be creating, and so I’ welcome the chance to interact with any of you, it feels like, again, there’s never enough chance to talk with everyone in the hall afterwards. I’d love to talk to you even a offline, so if you want to scan this QR code, we can figure out a time in the next week or so, we can find my schedule, we can meet for 15 minutes. But I’m excited to get you access to Coreform Fols for you to have a chance to see, I mean, what, what could you simulate without needing to mesh. And thank you very much.