Okay, good morning, I think, is it’s still morning, just barely. I’m Jeremy, I’m from Sandia National Labs, at, in Albuquerque, New Mexico, and I am the manager who drew the short straw, so I’m up here to tell you about the work of a lot of other people, some of whom, or many of whom, are are here today. So if you have more questions, or want to dig into the details, or find out all the ways in which I misspoke and misrepresented their work, talk, talk to them.

0:44 So okay, I have a a organizational responsibility for a group who does modeling and simulation, as well as a lot of soft materials development and advanced manufacturing processes, but also programmatic responsibility for production simulation activities, with one, within one of our advanced scientific computing programs, at at at the lab.

1:05 So okay, if you’re not familiar with Sandia National Labs, perhaps you’re familiar with Los Alamos National Labs, or perhaps the Manhattan Project, that’s our genesis. We’ just celebrated our 75th anniversary. We are an FFRDC, a federally funded research and development center, so basically what that means is contractor, government owned but contractor operated. In t, a wholly owned subsid, subsidiary of of of Honeywell has the contract to run Sandia currently. And as an FFRDC, basically our job is to be long-term strategic partners with the federal government, and we operate in the public interest with objectivity and independence, and maintaining some core competencies in missions of national significance. Obviously the strategic nuclear deterrence is our genesis, but we have a number of missions across national security areas these days, and one in particular that’s relevant to the crew here is advanced science and technology, especially computational and manufactur, advanced manufacturing technology.

2:08 But Sandia has made contributions in the area, I mean, as as an FFRDC within the the indd mission space, we are a design agency and a system integrator, but we have much broader impact in other missions, even in energy, energy security. So we’ve made contributions in fuel cell technology, and help field the first fuel cell powered ferry, also wind turbine design, more efficient wind turbine design, unmanned vehicle sensing technology. In the manufacturing space we pioneered clean room technology quite a long, long time ago, which enabled much of what we’re talking about today. We have a long history in h, high performance computing, both software and architecture, and large scale, you know, high powerered modeling and simulation activities, that sort of goes all the way back to those early days, around, you know, precision microelectronics fabrication.

3:11 But in terms of the other impacts that we’ve had in the manufacturing space, actually for advanced or additive manufacturing, Sandia has had 30 plus years of technology development and commercialization. So you’ve seen a a lot of those represented today, from any, anything from what we used to call casting, nowadays we call it direct ink, right, so extrusion based processes of soft materials, like polymers, paste inks, those kinds of things, whether they end up being fired and and made hard like ceramics, or pads and and cushions and encapsulant, also lens, lens processes, laser engineered net shaping, I think, is what that refers to, and and thermal spray processes, just to name a few. So s India has has been involved in in this space for for quite a while, and developed some seminal technologies in the added manufacturing space.

4:08 More relevant to this talk right now is our activities in the computational space. So historically we’ve been, we’ve developed a number of tools within the advanced simulation and computing program, focused mainly on qualification, U, activities, detailed high fidelity physics, high performance calculations to understand the function of systems and and components, as well as to quantify whether they will perform, or qualify whether they perform as as intended. So that’s kind of our our history.

4:47 More in more recent time we’ve, we’ve moved sort of to the left, if you think of the qualification activities on on on the move to the right, on the far left. Of the qualification activities centered around our our Sierra tool suite, which is a finite element tool suite that houses everything from thermal fluids analysis to solid mechanics and structural dynamics, arrow, arom mechanics, as well as a lot of com, underlying foundational computational technologies, that’s sort of been in the qualification space. Moving all the way to the right, more recently, with the development of the Plato tool suite, we’ve moved into a design optimization, and and orchestrating workflows around gradient-based optimization processes, topology optimization, sh, shape optimization, and those are two the things that I’ll talk to you about today, some of those tools, Plato being open source, and and other open source tools that that we use as well.

5:44 But more recently we’re trying to integrate across the entire workflow, to bring about both a design to qualification centered around the actual production process, manufacturing activities, a a a a a full digital workflow of our computational, integrated computational capabilities. So and then trying to instantiate that, not a very good projection up here, instantiate that in something like a product realization workflow.

6:22 Okay, so looking at the the manufacturing activities that we have in the computational domain, a simple sort of step through of how we might go from an idea to a product, we could lay out in a nice linear box diagram like this, just to highlight some of the activities and and some of the tools. So there might be some idea for for a part, or some design that we might want to optimize, maybe that’s a to topology optimization type process, but we’re not necessarily limited to that. Typical additive manufacturing, you are kind of, you’re building the material at the same time you’re building the part, and that has its challenges, especially when it comes to variability in the process, and and uncertainties that that brings into the actual material structure, and the material properties, and then the performance of that material.

7:17 One of the things we would like to do is sort of invert the current state of of of technology, which is you live with what you print, we would rather print what we want. So using the fact that we’re making the micr structure at the same time we’re making the part, can we tailor that micr structure in in some way that enhances the function, or or optimizes the performance in particular regions of interest. Then we might want to take that design, which is pushed both topology optimized, let’s say, in an ideal world, at the part scale, all the way down to micr structure, and make it process aware. So actually now that we have a part, simulate the manufacturing process, process, and optimize that process to minimize maybe residual stresses or distortions, subject to some engineering requirements.

8:06 And then at some point in the process there’s a handoff, between what has been talked about, as so far, as a, as a computational virtual workflow, to the actual physical process of of making the part. But that, in itself, as well, has become very digitized, in terms of the tool, tools and the techniques and how they function, as well as the the sensing and the inent you monitoring, that can be tracked and and analyzed. And at some point we need to fuse these two things, in the digital world, the the physical part, and all the information and data that we can collect about the process, and and and the materials, and the part as it’s being built, with our our virtual prediction.

8:52 So that on on on the one side we might have the the phys, physical part, and running in real time, either as a check, or perhaps in a closed loop setting, being able to monitor that process, correct that process, or inform our calculations as a result. So really at the core of what we’re, we’re trying to get at, when we integrate these computational tools, is a level to design both at the part and microstructure scale that’s process aware, and then integrated digitally with all of the information that we can extract from the actual process, as as the part is being being manufactured.

9:36 So with that as a back, backdrop, I’ll talk a little bit of some of the comp, about some of the computational tools that that we use at at different phases of of these workflows. So the first one I mentioned was Plato. Plato is really, u, a a tool suite for integration and orchestration of of optimization based analyses, particularly shape and topology optimization. So Plato, at its core, has, it has its own physics analysis engine, which is highly optimized and built around automatic differentiation, which exposes those gradients and derivatives of of key properties for gradient-based optimization, and and orchestrates all those workflows. In addition, Plato can also interface with your, our favorite finite element analysis tools, which we’ve spent a long time developing, and and making very robust and very performant, in terms of the high performance computing architectures, as well as validated material models, our Sierra tool suite. So Plato is able to integrate with all of these, and and sort of orchestrate these workflows, for for some key sort of bread and butter activities.

10:56 One of those, as I already mentioned, centered around topology optimization, but there’s a lot of, and and shape optimization, but there’s a lot of other research activities that are going on in this space around this this tool suite, and I’ll, I’ll go into those a little bit. But one of the, one of the items that I want to point out is optimization under uncertainty. So if you have very well-known and well described loadings, you might take a a structure such as which in the center here, and optimized to a very nice, neat, almost elegant geometry. But if there’s any uncertainty around those loadings, you might need to build in some some robustness, and so some of the research activities in the past have centered around optimization under uncertainty, us, using Play-Doh.

11:44 More recently, however, there have been a couple activities, I mentioned already the process of wear design for for AM. So the the movie down there, in in the lower left, is optimization of the stiffness of this, of this part, at the same time you’re trying to minimize the residual stress, as it’s being built up in a laser powder bed metal AM process. So controlling the heat, and and and the stresses that build up in in the layer, as you’re, at the same time you’re optimizing for this, for the stiffness of that part. So this is one example of sort of bringing these computational tools into multi objec optimization, to a process aware situation, for metal AM activities.

Another optimization example here is around cellular design. So if you have a lattice structure, and this is where 3MF comes into play, so topology optimized design, it’s based on a homogenation approach, but there’s a way of mapping the carefully between the homogenized continuum method, that’s being used to op, to do a, again, a multi-objective optimization here, sheer forces and thermal gradients, on this double ringed sort of structure, which is printed, actually, and over on the table if you want to take a look at it. So multiobjective optimization for this lattice structure, I think Josh Robins is responsible for this work, he was just telling me during the break that three ords of magnitude is a reduction in file size, for the, to go to the, from a STL representation to the the 3MF representation for this particular geometry. So a significant advantage with going with that file formatting capability, loaded easily, Brandon told me it loaded easily into the machine, and we’re able to print it, Ti64, I believe, is what the metal one is, sitting over there, pretty, pretty seamlessly. So powerful tools for lattice structures and and and process aware optimization.

13:56 Other emerging activ ities in this space really have to do with nonlinear mechanics and non-equilibrium mechanics. So if you want to optimize for these foam or lattice like structures for soft materials, you want to optimize, minim, energy minimization, or stiffness, or these snap through sort of geometries, which are kind of neat and cute for various, for various applications. Again, the tools that are being developed in the Plato suite, or, again, dynamics, if you’re interested in the modal response of of of parts, and you want to optimize those those modal response for structural behaviors, again, these nonlinear capabilities and solvers that are built in, being built into Plato, are are are handy for that, again, open source package.

Okay, so, so moving on from Plato, going back to Sierra. So Sierra, as I said, was a find an element package that handles a lot of of the thermal, thermal fluids capability, as well as structural mechanics, structural dynamics, very robust, highly performant. And we’re developing very detailed meltpool models, and validating those models against physically realizable processes and materials, and then upscaling, quantifying uncertainty, and the predictions of those models based on machine uncertainty, for pick your favorite metal AM, let say, machine, and understand what the variabilities of the laser scan speed, and and power, and all those fun things are, and propagate uncertainty all the way through the the the the the full physics weld pool, and and understand what is the variability, and do your simulations fall within or bound the experiment, experimental variability that that you might expect.

15:50 So coupling these full physics, full physics models, with full up verification and validation, uncertainty quantification, so that we can then scale up to sort of laser scan strategy, and and optimization procedures, to do more detailed and more rigorous residual stress and distortion compensation activities, as well as predictions for micr structure, and the design of scan strategies, to get specific types of microstructure in specific locations. So coupling these activities, from the meltpool scale to the part scale, thermal, fluid, mechanical, micr structural, and passing uncertainties through all of these modeling and simulation, is is is something that we’re very keen on, and have spent a lot of time developing that that that rigor to be able to do that.

16:51 Okay, direct INR is another application. So you see a movie here for lattice cushions and and pads, or you’re crushing, kind of like foam or encapsulant. Standard workflow might be you come up with a, some kind of geometry, then you slice that and print that, and then you test it, does it give me the response that I want, and you iterate on that. We’re developing computational workflow flows, based on our tools, both meshing and geometry representation, and our finite element, element analysis tools, to automate that, to explore virtually what is the optimal, or or what are the different lattice structures, and built up the library, from computational perspective, and then export those directly to the printer, as opposed, and and and optimize that sort of design loop on the virtual side before we actually go and and print.

17:45 We’re also developing the the computational tools to to model the printing process itself, for, again, defect detection, insitu process, monitor, monitoring, and and perform production optimization as well. Some newer capabilities that were just starting to get up to speed, also at, like the lattice cellular structure, sort of mesoscale, these are at powder scale simulations. Again, the resolution is not great in the projector, but what you would see here is individual powder particles resolved, for something like a binder jet, or other processes that are centering based, we really want to understand how the part is shrinking, and the burnout process, process happens, at the sort of powder scale, to be able to predict that and control that, using our open source, actually molecular dynamics code, but it has a granular powder modeling, LAMPS package, to to be able to develop these tools for these types of processes.

18:48 Finally, what we’d like to do, as I said, is kind of take all this capability, in terms of validated and verified, uncertainty quantified, part scale, process scale, microstructural scale, insitu diagnostics and monitoring, and bring that all together into a a framework that allows us, perhaps, a generative design, U, kind of an approach, or, at a minimum, some U inent you process monitoring, defect detection, and perhaps, you know, feedforward control and automation of the processes of themselves. So for us, generating tons of this data and operating on this data is something that, within the hyper poring computing activities at Sandy, is something we are very keen on doing.

19:37 And and have the capabilities to do so. In the end, again, hopefully you got a taste for some of the physics based modeling and simulation tools that we’re advancing, and sort of these credible simulation approaches we’re integrate them, integrating them, at at at part design and process optimization scales for CDAM, they’re mul scale, from meso and micr structure. Also the the validation, validation and verification of these capabilities is is robust, along with the uncertainty quantification and uncertainty propagation, to the coupling of these different, different methods, is something that that we have an expertise in. Going forward, ultimately, what we’d like to do is develop approaches for digital integration of these modsim workflows, with the Inu data collection activities, and this is something that we’d like to do, for the ability to have sort of on-machine inspection, and born qualified parts, and ultimately to accelerate our product delivery time cycles. So with that, thank you for your time, appreciate it.