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

MfAM Walks so DfAM Can Run!

Listen as a podcast

Abstract

In this talk, Chelsea Cummings of The Barnes Global Advisors explores the critical relationship between Manufacturing for Additive Manufacturing (MfAM) and Design for Additive Manufacturing (DfAM). Drawing from her extensive experience in industrializing additive manufacturing, Cummings emphasizes the importance of considering both MfAM and DfAM in harmony to fully leverage the benefits of 3D printing.

Understanding MfAM and DfAM: Cummings defines the differences between MfAM and DfAM and explains why both are essential for successful additive manufacturing.

The AM Maturity Model: A visualization of skills versus applications in additive manufacturing, illustrating the progression from pre-production to fully optimized design and manufacturing.

Real-World Examples: Cummings provides examples of designs that either focus too much on light-weighting without considering the process or fail to leverage the full potential of additive manufacturing.

Digital Workflow and Design Tools: A look at the digital workflow and the need for dynamic MfAM inputs, material libraries, and comprehensive digital tools that combine various loops of the design process.

The Importance of Process: Emphasizing that the optimal design must be buildable, Cummings argues for a more comprehensive approach that acknowledges process constraints early in the design phase.

Transcript

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

Read the full transcript · 2,943 words

Foreign, good morning, everyone. Thanks, Tim, for teeing up some of the concepts I’m going to talk about, hopefully they weren’t necessarily new when Tim brought them up, they certainly shouldn’t be now that he’s spoken of them. We didn’t even coordinate that much, we’re just that in sync, we’re, we’re on it.

0:26 So I’m part of the Barnes Global Advisors, we’re a small consulting firm, we help industrialize additive manufacturing, we’re experts in areas other than additive manufacturing though, so if it’s not the right answer, it’s not the only answer that we offer when people come to us for help. These are some of the the customers we worked with, or or companies we’ve partnered with to deliver things.

0:52 And when I’m not giving training with Tim, my primary role is to help customers redesign their parts. Sometimes it’s a clean sheet design that they want to build with AM, they know AM is important and relevant, but they don’t know where to begin, so I will help them with design, I’ll help them understand the process.

1:10 I’m mainly working in metals, rather, rather than polymers, polymers is a pretty autonomous of a process now, I think, the the press and go is there, I would say. It’s not so for metals, which is where I come in to demonstrate and explain all the challenges. Tim tells you the art of the possible, and then I come in and tell you how difficult it’s going to be. So let’s get into it, the, yeah, okay.

So MFAM walks, so DFAM can run. Now that I have you all, you brilliant people gathered in front of me in a room, you’re going to get the user feedback that you didn’t know you wanted. A couple of people have alluded to the engineer that is toiling in the computational design fields of AM, and it’s me, I’m that engineer.

2:06 So what I would like, and what more than anything I want you to walk away with, the mission to make me obsolete, just take me out of the loop, Chat GPT me out of the loop, my job is harder than it should be. I’m resilient, I think I could figure out other useful things to to do with my time.

2:30 Alright, so let’s get into it, I’m gonna, I’m going to further define what MFAM is, how is it different than dfim, where the organizational approaches that utilize both, why they need each other, how it relates to the adoption of AM. This is a really big one, this affects culture, plays a lot into this, and it affects the uptake of the technology. The effects of the design cycle, on the design cycle, from both perspectives, and my personal user wish list.

2:58 So, I, I like movies, so a couple references: we don’t want to be like Forrest Gump, he’s running without a purpose, spinning his wheels. This is what I would like in, to jumping into DfAM without the M-FAM aspect: sometimes, like this before, these are design guides, but they’re really telling you how to modify for the process, rather than how to design something that truly takes advantage of everything you get out of out of manufacturing.

3:29 This is just telling you how to overcome the process constraints and the different barriers we’ve all seen this before this, and these guides are fairly static, so we have to keep in mind that OEMs are making improvements all the time, material is a huge factor, you could have two different types of metal on the same powder bed machine, and they’re going to behave pretty differently, so that’s not really acknowledged by these guides.

3:57 So how, how did the the two interplay with each other, when we’re modifying for additive, we’re looking at the process, very process focused, this is our foundation for success, though, and this is kind of the paradigm shift. I think Ryan had mentioned how often stuff gets thrown over the fence, it goes to analysis, it goes to manufacturing, and it’s someone else’s problem, it’s we don’t have to work that way anymore, we are a team, we live in a very digitally communicative environment, so it doesn’t have to be that way.

4:28 We should consider the manufacturing first, lay that foundation, and then comes our design freedom and all those benefits that we can leverage on the DfAM side. So we’re designing, when we, when I say DFAM I’m thinking of a design that can only be made with additive manufacturing, and if you don’t go in the sequence, if you only consider the product outcome, the product design first and not the manufacturing process, you’ll, you end up in, and I’ll get to a chart later, where you’re, you get into these, within the design cycle you get into these iterative loops that basically stall the project entirely.

5:07 And a huge draw, seemingly, from our customers to additive is they see the the schedule benefits they can get, I mean, it’s certainly faster than going and getting a new casting, molding, for a legacy part or something like that, but the longer that we toil in in these iterative loops, we, we start to lose that value proposition. So you have wasted engineering labor, your schedule then goes into a risk state, and then ultimately, if, if this goes on for too long, then the the vision of the value proposition might be completely lost, and then there’s no adoption of the technology anymore.

5:46 So this is a little bit of that, or more of a demonstration of that, so looking at how organizations kind of enter the AM space. At the bottom, the mass, masses are generally aware, this is what I call the witnessing of the 3D printing propaganda, that it’s usually just a pure DfAM part that’s like been a poster child, but it will never be meant for production, the manufacturing and and the economics weren’t totally considered, but this is what has caught people’s interests, so they’re excited about it.

6:22 At the specialist level, maybe they’ve gained some education through training or a workshop, or they’ve experienced it themselves via trial and error, they’re starting to gain some process understanding, some of the MFAM skills, but the problem here is they look back on what they became aware of, and they’re like, okay, this isn’t quite a simple press and go straightforward as, as I was sold, and if this happens for too long within an organization, they might just completely abandon AM altogether, and we don’t want that.

6:56 At the top, we have our masters of process and part design, they’re using MFAM and DFAM and harmony culturally, the technology has been accepted, and that’s all goodness. So we call this the AM maturity model, this is basically a demonstration, or or a visualization, of skills versus what you’re actually doing with AM, so it’s just a construct, a way we like to bucket the the different ideas, so that people can sort of absorb and understand where they’re at, a little bit better, on their AM journey.

7:31 So at level zero, we have our pre-production, we’re doing some testing, maybe some prototyping, we’re becoming vaguely aware of what materials and machines are at our disposal. At level one, we’re starting to influence production, maybe we’re making tooling and shop aids, we’re starting to use perhaps our own printer that we have on site.

7:53 We encounter this a lot at tbga, so customers come, and it’s that story, they have a legacy part that they can’t get any other way, they need it quicker than the casting can turn around, but again, it’s, you’re often working in, Tim mention this, you’re working with a part that was designed for a completely different manufacturing process, and so you’re trying to force fit it, and this is kind of like Death Valley of the maturity model, because you’re not truly leveraging the benefits, but hopefully your skills are growing, and a lot of times this is done for a Pathfinder part that’s getting qualified for production.

8:29 So at level three, you’re becoming a little more functional, you’re in your design, you’re utilizing things you can only get from AM, you’re homogenizing the part, perhaps unitizing it, you’re eliminating joints, things like that. At level four, we know everything about the process, we know what our constraints are, they’re considered, we’re utilizing all the amazing tools that that this group has developed for our defim, and we’re getting the performance enhancement and just everything we’ve ever dreamed of.

8:59 So I’m gonna go through an example of DFAM without the M-FAM considerations, and then M-FAM without the DfAM considerations, and hopefully that illustrates why both are so important to use together. So this design is a robot arm, this was just an open source thing that we got from the internet, but we did the the top up ourselves, just to demonstrate a couple concepts.

9:22 So there are three parts in here that we replaced with top opt parts, so this is what I would consider, like, level two, the part for part, or sorry, this is a little, this is a little bit higher than level two, we’re lightweighting, so we’re, we’re doing something, but as you can see these, these build setups are, aren’t great, like the optimized part has more support than the original part.

9:50 And as Tim had shown, the grizzly scene of archaically busting out supports, it’s, it’s true, and it’s still like that, it’s, it’s very hard to remove metal supports, especially from power bed parts, and they degrade the surfaces and introduce quality issues that we don’t like to see. So I’m kind of picking on this example, these are optimized for light weighting, but they were not optimized for the process, these probably aren’t that viable for production, gonna end up with bottlenecks at the support removal stage for the poor soul that is having to do that by hand, so we want to avoid that all together.

10:30 Now, here’s an example, this was a Pathfinder part that was used on the v22 by navair, and they, they just wanted to prove that they could qualify production parts and and do it with AM, and in a way that could be qualified and compete with the original part. So this, in the back, we see what the final part looked like, and then in the front we see what the as built part looked like, and what role did AM play here, like, practically none.

11:00 This is, this isn’t, this we’re not getting a lot out of AM. As someone mentioned yesterday, it is an expensive manufacturing process, so if you’re gonna use it, you know, make it count. They, they did it for the purposes of of path finding, and and that’s fine, but I mean, yeah, you’re not getting a lot. As you can see, they, they went through their MFAM principles, they’ve extruded down to the plate, maybe they’re doing some wire EDM to eliminate a machining operation and just cutting it off that way, they have their their guiding cones, truncated cones, as drill guides, everything is chamfered, so there’s nothing, no, no large overhangs, so they did their M-FAM work, but did they really leverage AM?

11:50 So this is a typical digital workflow, where rapid designer, hopefully rapid, not always rapid, design iterations occur throughout the digital workflow, and this is just to demonstrate the many loops that we can get stuck in in engineering. So you might be generating, or top opting, your part to improve the geometry, then you go analyze that, you can go back and forth with that for a while, you prepare AM build with its many supports, and you, you improve that build, you simulate the build to make sure it won’t fail, maybe that forces you to improve the geometry.

12:32 Use an MFAM to DfAM progression, look like, when it’s working, it might look something like this. So this is a part, this is just purely unfam, they’ve added some chamfers to the tabs to make them build, this is kind of like a part for part substitution. At the next stage, we’ve reduced some some weight, some overall length that we just determined we didn’t need, we have some organic type of structures, this should build fine.

13:01 And then finally, at, if we’re at level four, we are truly only using the material and the places that we need it, we’ve determined that this will build, we can fit more of these on a build plate certainly than these, if you’re crafty maybe you can find a way to stack them. So this is like a normal progression that that an organization might go through, but ideally this happens, you know, before the build trials ever begin. If, if we have our MFAM understanding and our objectives to to get something built are clear, then we should be able to skip over this basically.

13:37 So what does a good solution for a a design tool look like? Is it taking the design rules for 3D printing and then programming them into the generative algorithms, does that make a perfect design tool? Probably not. So at this point you ask, what is it that I want? If you’ve ever seen this scene, maybe it’s resonating right now, it’s not that simple, maybe it is that simple, I don’t know, you guys tell me.

14:07 I, I think I want dynamic M-FAM inputs, these are very static, they don’t consider material or the progress that OEMs are making. You know, machines like Velo can almost go support free, certainly depends on the material. I think, a much like the polymers, a material library that’s defined, perhaps by, if I can go as specific as machine, even specifications that I’m building and qualifying to, maybe it goes as fine as build parameters, laser scan speed, and stuff like that.

14:40 I was thinking about Andre’s presentation, so I, part of what I do is test out software to see what digital tools are out there. If anyone has software that I have not tested, or I haven’t tested in a while and has drastically changed, please send me a trial license, I would love to do that.

15:00 So you may be saying, hey, we have considered some of this, this is old news, and I, and and I’ve tested some of them, but I think, from what I’ve seen, many of them, the algorithm is still very much swayed in the way of that peak solution. So the the process factors are not weighty enough to actually change the outcome, because I’ve turned some of these knobs and they still ultimately end up with that defined, that optimal solution.

15:27 So I think I want to approach closer, yeah, to the plateau, I want something somewhere on the plateau that I can actually build, because if the optimal design won’t build, the final product won’t be the optimal design. I know that sounds intuitive, but again, if I’m changing these knobs and they don’t have enough weight in the algorithm to actually change the the design output, I’m still kind of in in square one, I’m back to square one.

16:01 So, if I just want to emphasize again how important process is, and hopefully all of you people will, will make it a priority, because once, let’s say we, we find that peak design, that one solution, we try and force fit it through the process, the design integrity is ultimately going to be lost, and and so you don’t end up with the optimal design anyway, because we’re going to have to make so many modifications to actually make it build, adding those supports that degrade the surface, all those many things.

16:33 So a more comprehensive digital tool may combine several of these loops, and we’ve seen this in Paramatters, Carbon, I know the the FEA is being incorporated into one step, and this is certainly helpful, for me, again, just program me out of the loop, make me obsolete. In court, it would be good to incorporate real data into the generative process as well, and completely close the design loop. I have not really seen this in a in a product on the market.

17:08 So, but all, all the option, these are other people’s examples that I’m just using to drive home the point, but MVM and D fam Unite. This is going to build no problem, this pine cone imitation that reduces noise with air bleed valves, we can cram a bunch of those in in a build and get that high yield and and lower cost per part. The oil cooler that was optimized at Morph 3D with entropology, you can build this as it is, we want to reduce post processing.

17:43 So, in conclusion, MFAM should be acknowledged early in the design phase, we want to acknowledge process constraints early, it matters a lot. M fam and dfan play different roles, but both are necessary, as you can see we don’t get a lot out of AM without both of them. The organization should consider them early, this can be still, it’ll still probably be iterative throughout the design cycle, but it should be, it should shorten it. Implementing both increases the adoption of the technology, the more success that we have earlier matters a lot to leadership. So with that, thank you everyone, appreciate your attention, thank you.

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

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

Designing Material Properties with Advanced Toolpath Generation

Designing Material Properties with Advanced Toolpath Generation

Alex Roschli · Oak Ridge National Laboratory

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.