0:25 Good morning, everyone. Rik Baruah with Intact Solutions. Today, I’m going to talk about how we need to change how we deploy physics across the engineering life cycle. But first, let’s start with this UAS. So, as we’re as we change manufacturing processes, we’re changing the entire engineering life cycle, right? And that starts with a design. So, say we have a built-up assembly with CNC machine parts, with hundreds and fasteners, and lots of labor, and tons of stuff are built in there, right?

1:02 And then we’re changing to a cast part. So, there are some tradeoffs that goes with that. It’s kind of hard to see on the left, but then what he’s saying here is that like every disruption, a closed foundry, a resourcing different vendor, changing up the material, it regenerates this entire new design space. You have to re-qualify, you have to run different variations. Ultimately, we’re running 5,760 simulations.

1:33 And typically, what happens is that you’re running one design, one simulation. So, how do we run these 5,760 simulations, right? So, you can brute force it, where it’s a very manual process. We can train an AI surrogate as well, but then it’s still bottlenecks at this running 5,760 simulations in this in this in in this case. So, having said that, this is why we believe that physics must become an infrastructure.

2:10 Right? It needs to be on demand, inexpensive to run, but we cannot compromise on accuracy and trust. So, traditional CAE typically lacks automation. Right? You’re not able to go and you go No, we have to go through all the manual pre-processing steps to get to your result. And then, it’s also not jump diagnostic. Right? As the as the as the manufacturing as as you evolve across the engineering life cycle, you’re changing your representation or representation is changing.

2:42 It goes from CAD to say CT scan data for as manufactured part. And then, on the right side, we have physics AI view. The challenge one of the challenges there is really convergence, right? And we see physics AI as a customer of the physics infrastructure. Right? Because you need traceability, you need reproducibility. And then, and the baseline here is that this is table stakes, right? The It has to be API first.

3:18 It has to be scalable. It has to be auditable. And it needs to be deployed anywhere and everywhere. So, looking at this, like we see a couple of shifts. Right? So, before I get into the shifts, with traditional CAE, like you want to run hundreds or thousands of simulations, it needs to be fully automated. Right? If you have to go and simplify your geometry, repair the model, and then deal with like non-conventional CAD models like an implicit or G-code or scan, it’s not able to handle that.

3:55 That’s why we have to shift in how we approach this. We have to decouple the geometry from the analysis. Right? For traditional CAE, it’s the the the your mesh is tightly conformed to the body, right? And because of that, it brings a lot of a lot of challenges like clean up, simplification, and and and so forth. But, on the right side, this is how Intact operates, right?

4:23 It is first principles. There is no mesh generation. And we can match the accuracies of Abaqus’s and and Nastran’s that’s out there. So, the second shift that we have to make is how do we how do we accurately capture stresses? Right? To to do this, we have to you know, it it needs to be automated. It needs to be an automated way to capture these stresses because otherwise, it’s not just about running one simulation, right?

5:00 For a convergence study, you’re running multiple simulations. And this is how Intact is able to get capture these accurate stresses without having to go through the deep you know, convergence studies and all that. So, let’s talk about some real-world applications. So, one is geometry being independent, right? So, we you know, tackling small features, small holes, you have a large scale model with a lot of tiny holes or lattices, or there’s evolving geometries, or even there’s bad geometry involved, right?

And and and tackling the physics side of it as well, such as large deformations and non-linear materials, conjugate heat transfer. And then when we decouple this, mesh or this analysis from geometry, we do introduce some, technical challenges as well. Like how do we account for interfaces? This this is we have we have solutions for this. That’s what I, you know, getting to. But, let’s talk about some workflows and how we deploy Intact today.

So, just couple of videos here. So, with Intact, it really about generating the high-quality data for, you know, training the AI models, exploration exploring large design spaces and optimization. On the left, we see this evolving structure, that we’re running thousands of simulations, but it’s all scalable, reproducible. We are not there’s no human in the loop. But, it’s also, auditable as well. On the right side is the same example, but it’s non-linear.

7:00 The same way it’s like we’re capturing this metamaterial expansion, but then we we’re capturing the non-linear behavior of it as well. In this example, this kind of a a closed loop orchestration, right? So, this is an intelligent way of, where to simulate because when you have loads of parameters, you want to simulate every single one of them. It, becomes a combinatorial explosion. So, we can use machine learning tools to identify which parameters simulate because we can offer the sensitivities as well.

7:40 And so, this, this is some interesting work we’ve been doing with with Ansys for Ansys we can read their or work with their dot implicit file directly. And this allows us to couple with the fluid solvers that they have. We can also couple with external fluid solvers as well. But, in this scenario, everything is handled implicitly. There is no meshing of the geometry when you need to do the sculpting.

8:09 So, we can do fluid structure interaction without going through these steps. This example here is something we’ve been working on. Exciting to see is you know developing transient capabilities and dampening and and so forth. This is really for our aerospace customers. Without having to go through, you know, again these technical manual steps. I want to shift gears a little bit now and talk about these agentic workflows.

8:46 So, today in design CAD and CAE tools you have a co-pilot that sits inside the tool, right? But, what we see the future is going to this agentic workflow where you have a deal we have a domain experts, you have subject matter experts, but these are agents. They’re working together, talking to each other. But, at the bottom, you still need a robust callable physics infrastructure. Right, that we can go through during the process.

9:21 Here’s an example of the workbench example where we have I mean you’ve probably seen these MCP examples or demos quite a bit, but this one, you know, cloud code is calling the intact MCP to do a bad simulation, run through like 10 or 12 variants, and identify the best best one depending on the criteria that we set. And in cloud, it’s also calling out external tools like ParaView here to kind of visualize the the results.

9:55 Right, so you can have these as as independently calling these tools to provide you the complete workflow. And then this example this is a kind of a sneak peek for Rujix’s discussion later today. They’re developing this agentic system where you have a design agent, a manufacturing agent, a simulation agent. The simulation agent calls intact in the background to go through that workflow. And I’ll let I won’t steal his thunder.

10:25 He’ll you know, he’ll go go through and explain this much better than I can. But these are capable tools today. Right, but how do we deploy intact today? So we have a few ways. We’re inside the design tools such as Ansys and Rhino and Senario and Onshape. But the power here is about automation. Right, we have a Python native APIs called PyIntact. It’s deployed it’s deployable anywhere.

10:56 We’ve got customers deploying it in their own on-prem HPC clusters. And it’s a Python native so you can use it with other Python tools as well to create this entire workflow for automated simulation. And of course we have an SDK as well that we license to other software vendors. I want to go back to that to the UAS. Because it’s never just one thing. Right, when you’re changing the manufacturing process, we talked about the design change.

11:27 But then you’re also changing tolerances. You have to account for that. In in the design, right? There is changes in defects of the process of the manufacturing. So, you can account for these things. So, building on this we are last year yeah, it was end of last year we kicked off this project with with DARPA STO office. Kind of developing these tools to be able to analyze the the GD&T how GD&T impacts analysis and design and redesign and what are the trade-offs, right?

12:06 So, this is kind of legwork, but it really is building to this right here. And this is actually pretty hot off the presses because I just got approval to share it from from DARPA folks. This is our COMMANDER project. It’s a composable manufacturing digital twin that we’re working on. So, idea here is that we need to address the supply chain, right? We need We need tools. We need capabilities where we can evaluate these different manufacturing methods at run time rapidly, right?

12:38 Because the world is changing. The economics are changing. So, we take in the the the manufacturing process and then compose that with you know, different solvers cuz we’re not necessarily in the business of developing our own solvers, right? We want to interface with what’s existing today and take all these different data and really predict an output what what we want to know is that is this a treatable system good enough?

13:05 Can it Can it do what it needs to do? Right? We don’t So, we don’t have to go through the entire recertification process. So, this is kind of the building block that we’ve been working on this this year and this will be a ongoing project. And the point here is it’s a fully automated. It’s a rapidly developed and deployed the physics as infrastructure. And I’m going to close on on this slide.

13:33 Intact is not just a a research project. And we we have we have real we have customers who are deploying this today across aerospace and advanced manufacturing and computational design. But we’re always looking for partners and collaborators to do the technology transfer. Intact, we’ve been around since 1999. We’ve We’ve been around for a long time. But just maybe 3 years ago we started commercializing these capabilities. And we’re adding more capabilities and then and working with government partners who are aligned with the needs that they have and aligned with the needs that the DIB has as well. And we’re also looking always looking to transfer the technology to you folks. So, I will end at that. Thank you for your time.