CDFAM NYC 2023 · New York · 14–15 June 2023

Designing Material Properties with Advanced Toolpath Generation

Listen as a podcast

Abstract

In this comprehensive talk, Alex Roschli, a researcher at Oak Ridge National Laboratory’s Manufacturing Demonstration Facility (MDF), delves into the innovative world of advanced toolpath generation and material property design. With over 11 years of experience in large-scale manufacturing, Roschli shares insights into the development of toolpath generation software at ORNL, focusing on large-scale processes and the transition from polymers to metals.

The presentation covers a wide range of topics, including:

Infill Pathing: Roschli discusses the traditional infill pathing approach and introduces a more complex, stress-based infill design that adapts to the specific needs of the part.

Site-Specific Control: He explores how to control the toolpath on a site-specific level, allowing for different resolutions, speeds, and materials in various areas of a print.

Multi-Material Toolpathing: The talk delves into the design of specific materials in specific areas, blending materials, and functional grading.

Printing Faster with Continuous Passing: Roschli presents an innovative approach to continuous passing, using algorithms to optimize the path and significantly reduce print time.

Transcript

From YouTube’s automatic captions, lightly cleaned; expect some errors. Each timestamp opens the video at that moment.

Read the full transcript · 4,076 words

Foreign, get started. My name is Alex Roschli, I’m a researcher at Oak Ridge National Laboratory. I work at a facility called the MDF, which is the Manufacturing Demonstration Facility, if you’re not familiar. There’s a government research lab, the MDF is a DOE user facility, which means we’re kind of open to the public. We work directly with American industry and non-american companies as well, but primarily American industry, doing research and development on a variety of topics, mostly geared towards advanced manufacturing, with a heavy focus on added manufacturing, but we do have traditional composites work, machining, hybrid manufacturing, that sort of thing. Some of the other presenters kind of alluded to, added manufacturing is not the solution to everything, so we have all these traditional manufacturing methods, and we use added manufacturing to show how we can improve those processes and how we can work together to do more manufacturing and improve the entire landscape.

1:06 So some quick background about me, I’ve been working at the MDF for about 11 years now. Most of my work is on large-scale manufacturing, originally did a lot of polymer work, thermoplastics, got a little bit in the thermosets, transitioning more into large-scale metals, the DED wire-based processes. So here’s a variety of the projects that I’ve worked on. Now today most of my work has moved towards slicing and the the software that runs slicing, which is kind of the overarching thing between all of those projects you just saw pictures of, is that all of those are printed with the same software. It’s a tool path generation software that we have developed at Oak Ridge National Lab. We use it for, like I said, all these large scale processes, and we kind of keep growing it out into new new applications, new systems, new processes.

1:50 So for those that aren’t familiar, slicing, it’s kind of the step, the process to go from your solid body CAD model all the way out to your finished part. So it includes the creating your fasted model, be your STL or your 3MF, you know, we’re hopefully you’re going to get away from that triangulated mesh soon, Dewan’s leading the charge over there, but you know, move towards a better geometry representation and try to improve this process. Next step is going to be slicing, and then from there you’re going to fit your tool paths for the layers, print the layers, you know, if all that goes well, you’re going to have that finished part at the end that looks very similar or identical to your CAD model.

2:26 So a bit of background for this project, or for this presentation rather, you know, the the idea is always out there that we can 3D print anything, right, and and there’s a lot of truth that, but there’s a lot of failures to that as well. There’s a lot of things we can’t do, or a lot of things we’ve tried to do and haven’t succeeded on, and so you saw those projects I have done, how many of those worked the first time, how many of those worked the second time, how many of those, you know, took us a long time to do it, and how many things did we end up, you know, having to redo over and over and over again.

2:55 And so one of the more famous ones of those is the the first three print card, which is at the the top, and I’ll show a picture of it again later, but before we printed that car, that you know, that looks great, and it kind of traveled around the world, was a rather successful project for us, we made this car. And this is obviously, you know, significantly ugly, there’s got cracks in it everywhere, there’s huge gaps, the armrests have holes in them large enough I can put my arm through, the armrest, the the fenders can be pulled apart with your hands, you know, that this is a printed part that’s supposed to be able to survive like an impact, but you can just go peel it apart.

3:27 So this was the first car, this was seven days of printing around the clock, seven days of people at the machine maintaining it, all for for this part that, you know, really isn’t a success. And how could we have improved that, how could we have had better software, better computational design, better tools to understand what we were going to do before we did it, and to try to prevent that failure. And so this right here, this project, the car, was finished in 2014, this print right here was late June, early July of 2014. Like I said, it took seven days, by the time we made the actual one we were able to print the whole thing in about two days, and so printing faster was part of the solution, but redoing the design, better understanding of the tool path, better modeling of the whole printing processing of the car is really what helped us get there.

4:14 And so I’m going to step through a few various aspects of the software design, specifically towards tool pathing. The first of those is looking at infill pathing, for, and this applies no, mostly to polymer systems, but you do see it in fill on some of the larger metal processes as well. And so for those who aren’t familiar, the traditional infill passing approach is going to be, you project a pattern, lines is the most common, over your object for a layer, you clip it to the boundary, and then you join it all together, and now you’ve got this pattern. Pretty simple, works great, but it’s all uniform, and how many of our parts are going to get used in a uniform way? This is a square, if you make a cube, that it’s a pretty good example, you use that pretty uniformly, but to get to larger objects, more complex shapes, they’re all going to have different loading, certain areas you’re going to see more load than others, loading from the side might be more than from the top, and so how do you accommodate that, how do you design for that, how do you simulate that?

5:11 And so that’s kind of what we’ve been working on, the the left side of this is what you just saw, the conventional method of generating infill, where it’s completely uniform, and the right side is looking at how to model and simulate that, how do we iterate through that infill until we get an infill pattern and an infill design that’s going to work for the actual end use application of the part. And so looking at areas of high stress, and so doing tighter infill, doing these a honeycomb pattern, but doing a much smaller, tightly packed honeycomb than areas of low stress, trying to spread that out, use less material, print faster.

5:44 And then, as you see, there’s a gonna be a connectivity issue, if you’re just putting those two, if you’re just drawing those uniformly and putting them next to each other, that doesn’t quite solve the problems, you have to iterate through that, and you have to be willing to modify it and kind of change the the structure and the layout of all those to make them blend together. And so here’s how that kind of works for us, the simulation for this is all done through Abacus, so you load in your your object, in this example we used a like a very basic structure for an aircraft wing, simulate the the loading on that, find the areas it’s going to have high stress, areas of low stress, and then start doing an infill patterning over that whole process, or over that whole shape, specific to the the stress analysis.

6:25 And then once you’ve got that infill pattern, go back in and simulate it, simulate what the infill pathing is going to look like, and kind of iterate through that until you have a solution that looks workable, and then the last step of all that is trying to validate, do some testing. And so the patterning algorithm is primarily done with circle packing, other algorithms, so filling the whole space with circles, big circles for areas of low stress, small circles for areas of high stress, trying to pack that infill more tightly, and then looking at that grid, trying to understand where you need, like I said, the large circles versus the small structures, or large circles versus small circles, to fill that adequately.

7:05 And so looking at all of those as they apply to the actual design, the very first one on the far left is going to be your uniform infill pattern, so all those hexagon shapes are the exact same, it’s just like the lines, it’s a pattern projected across the shape, and it’s cut at the boundaries and everything’s connected together. The next one is looking at the iterative circle packing, so putting circles in there and changing the size of the circle to get the optimal fill in there. The next step, sectioning, trying to look at the areas of high stress and low stress, and so you can see a very distinct area that’s going to be the high stress, it’s got those tightly packed hexagons, and then the last one there trying to combine those approaches, so now we have a variety of hexagon sizes and they kind of morph together to give you this more complete infill pattern across the whole whole structure.

7:50 So then looking at what happens when you load it, when you put the weight on there, and see how does it deflect across all these different designs. And so the picture at the top there shows just the the basic uniform infill, and then the very last one the iterative plus the the sectioning approach, and then here’s kind of all that data moved together, ultimately showing that with the the inner perception, iterative plus sectioning approach to that infill pattern generation, you get about 50 higher stress for that same shape.

8:26 Next thing I kind of move into is looking at how you control the tool path thing on a more site-specific level, how do you do different things in different areas, and how can the user design this from the software, maybe you don’t have the ability to simulate what you want to do, but you know, close enough to what your end use of the application, the part is going to be. And so this example you see at the bottom, I’m looking at the mold for an Ironman mask, so you’ve got the the mask, you know, protruding from the front, and so we want to do something different just in the area where we’ve got the mass, but the rest of the structure we don’t need to be as high resolution, maybe we can print it faster, maybe we can use a different material, so trying to identify that area.

9:02 And so, you know, the second image there shows in orange, this is the area we want to do something different, and how do we want to control that? And so in a lot of our applications it’s just as simple as changing the print speed in that area, so printing it really slow, trying to do a higher resolution, you can change the layer height, you can change the bead with, so you can put really thick beads there, if you’re going to machine it, gives you more material to cut into. And so the user can just design all of this, and it takes about the same amount of time in the slicing approach, still, you know, five seconds to slice this object once the user sets it up.

9:35 The last object, last image there, showing you can make the whole part hollow with a slicer, one of the speakers mentioned it yesterday, kind of the issue with CAD and the offsetting process, that doesn’t, not a problem when it’s in slicing, because you’ve got it at the polygon level, the offset to very easy, whereas CAD for these complex shapes, doing the offset, trying to make the exact thickness you need can be a lot more challenging. So just a quick look at some of these, as how they would have been built, the one on the left has been machined to give you that nice solid surface.

10:07 Another use for this, looking at site-specific control, this is on a large scale metal process, if you, you print that bottom right object, it’s just a, you know, a plus shape, design the way the bead, the weld bead is going to move around those corners, you’re going to have a porosity in the middle, you have that, you know, that void space. So what if we can keep the same pathing, but we just modify the welding parameters when we get to the center of that, so we get a more solid object, but we can keep the print speed the same, we just change the power of the lasers. This is a made out of titanium using a laser wire system, but the user can go in and define that central area, we need to do something different here.

10:44 Another application of that is what we call embossing, so being able to change the the parameters again in a specific area, in this instance, you know, you can project an image or project another object onto the surface and then just change the parameters in that specific area. So again on the bottom, that’s the same system, it’s a titanium wire projecting the ORNL leaf on there, and then just changing the weld pool parameters while it’s going through that area to emboss a shape.

11:12 So here’s the car again, as you can see, that it’s a printed one, big monolithic structure, and then seats have been added in, but you know, with as software has progressed and the machine has progressed a lot as well, what if we could print those seats in, how would we go about doing that? If we wanted to do, you know, more comfortable material, or the original one is just a solid bench with like a divider down the middle between the two seats, but how would we tell the software we want a specific material there, or we want to, you know, print that area differently, so the settings readings are part of that.

11:43 And that kind of leads into doing multi-material tool passing, so being able to do specific materials in specific areas, this is useful for, you know, mixing rigid and flexible materials, but also mixing lightweight materials, low cost materials. So back to the Iron Man mask, if you want to do the the tool surface out of a very high expensive material, if you’re printing the whole thing out of metal, you want to use tool steel for that surface, but the rest of it you want to be cheap, you could use mild steel, until you just designate that surface as one material and the rest of it design it as the other. And so this can all be designed by the user through the software.

12:18 And so again, Foams is another great one, looking at variable density, and so here’s kind of a more look at that, this is on a large scale thermoplastic system, the the top right is the same structure you saw from the infill, rather than using the same material for the whole thing, but if we did PLA for the areas that we needed high stiffness, but out in the low stiffness areas we mixed in a foam, so it prints, it’s even lower cost, less weight, but still one machine, and it’s all blended through. And so we can functionally grade those materials to the software by designating, we can load in the the loading applications into the slicer.

12:55 And so apply those, the bottom left there is kind of looking at a different process, this is what I was talking about earlier, combining edge manufacturing with traditional manufacturing, and so printing material directly into a mold, and then as soon as you turn off the extruder, you throw that mold into a stamping press. So we’ve got a 510 press, and so you can compress the object, all those voids and deformities, but you can still have a good solid object, and you can do localized material control. So we mixed orange and black material there, that the black is going to be your carbon fiber ABS, so it’s much much stiffer, but much more expensive, and you can use it just in the areas that you need it.

13:37 Just a quick, on the Foams, we’ve got two types of Foams we can use, one to be a syntactic foam, so trying to blend in your actual resin material with a foaming agent to give you a material that’s still pretty stiff, but it’s lightweight, so looking at density of less than one gram per cubic centimeter, so something’s going to float. And then if you want to go really low density, a really inexpensive material that you’re just trying to fill an interior structure, looking at expandable microsphere, so Foams that when they come out of extruder they really expand, and so designing that into an object.

14:12 This right here is a tool for precast concrete, I’ll show it on the next slide, but it’s actually a building that’s been built here in Brooklyn, but the the important tooling surface, make it out of your expensive material or carbon fiber, but the rest of the interior structure we just need a little bit of support, use your syntactic foam to design the material properties and to lower that cost. So on the right there’s just a quick look at what that image of that building looks like, it’s, I think it’s on the south waterfront, but the whole building needed about 200 different of these precast concrete molds to make all the exterior architecture pieces, all those concrete molds that go around the windows, it’s meant to kind of deflect the light differently, no matter where you look at it, to shine like a sugar cube, it’s built on the location of an old Domino Sugar Factory.

14:59 And just a look at the system for how that was done, it’s a thermoplastic system, Cincinnati Bam, it’s got dual material infeeds, and so it can switch those materials right at the end feed, the extruder, but there’s also a blending unit, so that’s how you do the functional and graded materials, if you want to go from that PLA to the PLA foam, you can blend those materials as you feed them out to the extruder, so you can change that density. And so look at how the software works, this is again a simple square shape, but simulate your material, similarly your object, in Abacus or whatever you prefer, then you can load it, take out that data, and you can load it into the slicing process, and then apply it to your infill, and so that’s what this is showing here, is that every step of that infill is going to have a different material recipe applied, and so you can grade the material through that infill structure.

15:46 Last thing I want to talk about is, I was trying to figure out how to print faster, how to do different passing structures, so basic figure eight shape here, the way this is going to be traditionally printed is three closed loop pass, you got one for the outside, and then one for the around each of the openings in the center, so in a large scale system, at the end of a path you have to slow down, as you go into the end of the path you have to wipe the tip, you have to lift up, you have to travel to the next path, you have to lower down, and you have to start it again. It’s a lot of time, it’s a lot of wasted time, a lot of wasted machine time where you’re not printing.

16:16 So is there a way we can improve that, where we can kind of combine all this together and print it faster? So that’s looking at a single passing approach, and so a really easy way to do it is just go in and cut your CAD model, but that takes a lot of time, it’s a lot of manual labor to go in and modify the object and cut it, but if you have the software go in there and optimize it, you can cut it in these different locations, then you can move those cuts layer to layer, and that’s what you see in these images. The one on the left is kind of showing really narrow tool pathing, so you can better see the gaps, and one on the right it’s actually showing the real beadlets, and when you print that in reality you’re going to overlap them, so you still have that strength there.

16:53 And this is just looking at some more complex shapes, the one on the left to really exaggerated the cut distances so you can see where where they are, and basically turning those paths around, and this is fully automated, so you load in your object just as you would normally design it, and then the software will design all these cuts, in it’ll cut it, and significant time savings. The object in the top center there is the cross section of a boat hole, the time savings on that is about 30, and then the object on the right is what it would look like if you wanted to leave all of your outer contours as solid, you don’t want to cut any of those, so around all of your holes you have a solid contour, and then you combine all the interior pathing to speed that up.

17:31 The way this works on infill, this is going to tie back to what I talked about first, is using a continuous passing algorithm for your interior structures, and so using the Chinese Postman problem, once you’ve got that circle packing done, and you’ve got each of those circles as a node, how do you travel through each of those nodes optimally, how do you try not to replicate any of those, don’t want to go to the same one twice, so how do you move through the whole thing optimally, and how do you have the software design that and simulate that and then print that.

17:58 And so this is what it looks like, the one on the left is your traditional pathing, so it’s going to do one big outside path where it kind of meets the contour, and then all the interior ones are all these individual closed loop pass, you can see the blue highlight, search, we’re going to stop and have to travel and move to the next circle, because every single one of those is going to be printed individually, whereas on the right, you know, pretty much faster to combine all those together to one homogeneous path. And so the time savings on this object, kind of an extreme case, is about 50 percent.

18:31 And then here’s what it looks like if you put it on small scale, so on a small scale printer the time savings is not as significant, there’s a lot less of that wiping the tip, slowing down, when you have a filament system, because you can pull that filament out to cause retraction and stop the material flow, but on thermoplastic systems when you’ve got pellets and things like that, you can’t just stop the flow as easily, because you can’t spin it backwards, you’re going to pull molten material up into an area where you shouldn’t have it. And so the the small scale systems definitely can go a little bit faster and have less of a need, but still, still very applicable both. That, thank you, I’ll be around the rest of the afternoon if anybody’s questions.

More from CDFAM NYC 2023

Design Computation Human

Design Computation Human

Onur Yüce Gün · New Balance

Mud y Robots: Keynote Presentation

Mud y Robots: Keynote Presentation

Ronald Rael · UC Berkeley / Emerging Objects

Roadmap to Replicators

Roadmap to Replicators

Neil Gershenfeld · MIT Center for Bits and Atoms

Scaling & Standardizing Patient Specific Design

Scaling & Standardizing Patient Specific Design

Adam Wentworth · Mayo Clinic

Computational Design Ops

Computational Design Ops

Ronnie Parsons · Mode Lab

Overcoming the Design Dilemma

Overcoming the Design Dilemma

Ryan Carter · Johns Hopkins APL

Generative Design & Digital Manufacturing at NASA Goddard

Generative Design & Digital Manufacturing at NASA Goddard

Ryan McClelland · NASA Goddard Space Flight Center

Digital Materials – Geometry and Data Combined

Digital Materials – Geometry and Data Combined

Marek Moffett · General Lattice

Ten Years Teaching Additive

Ten Years Teaching Additive

John Hart · MIT

10 Years of Applying DFAM

10 Years of Applying DFAM

Tim Simpson · Penn State University

MfAM Walks so DfAM Can Run!

MfAM Walks so DfAM Can Run!

Chelsea Cummings · The Barnes Global Advisors

Informatics and AI for Materials Development

Informatics and AI for Materials Development

Wes John Alder · Citrine Informatics

Predicting Fatigue Life in AM Metal Parts Using DigitalClone

Predicting Fatigue Life in AM Metal Parts Using DigitalClone

Behrooz Jalalahmadi · Sentient Science

Design Automation Using Machine Learning & Computational Design

Design Automation Using Machine Learning & Computational Design

Rhushik Matroja · Cognitive Design Systems

Learning to Generate Shapes

Learning to Generate Shapes

Karl Willis · Autodesk Research

Design for Artificial Intelligence

Design for Artificial Intelligence

Chris McComb · Carnegie Mellon University

Register for Updates and Discounts on CDFAM events.