Extreme Optimization: An Adventure with Generative Design for Fusion 360

GAVIN BATH: Hi. Thanks for coming to my adventure in Fusion 360 generative design. My name is Gavin Bath. I work for CADPRO Systems in New Zealand. I currently run the software development team. But my background is in manufacturing and machine design.

And so this was a little personal project that I did a few years ago using Fusion 360 generative design to develop a skateboard truck for this electric longboard. Which has been nicknamed "Gavin's Gone" by a group of users on an e-skate, electric skateboard builder's community online.

So this story starts back in 2018 at IMTS in Chicago. Where some generous guys from DATRON, who manufacture CNC machines in Germany. They were making these skateboard decks on a trade show stand as a demo.

And I got chatting to one of the guys. And he happened to recognize me from Instagram. And said, I'll give you one. If you do something cool with it.

So I said, OK. Well, I'll build an electric skateboard, not knowing anything about electric skateboards at the time. And packaged it up and brought it back to New Zealand with me. And promptly got going ordering some components online.

You can see motor and belt drive on the back end of the longboard and then the enclosure underneath. So the enclosure underneath was CNC machined out of bamboo. And all the other components, except for the battery-- off the shelf. The battery's a custom made 18650 lithium cell pack.

I rode it like this for about 300 kilometers, mainly backwards and forwards to work. It does about 48 kilometers an hour. And it has about a 20 kilometer range. So it's quite a potent wee beast.

But I was having a few issues with belt misalignment. Which you'll see here. So for those of you that aren't skateboarders. The hanger, this piece that you can see, sort of t-shaped piece on the skateboard truck-- this is an off the shelf truck.

And I basically just manually machined a cylindrical section on the hanger to mount this motor mount plate. Which was an off the shelf item. But it never really worked too well, in terms of alignment.

So I decided I needed to make a custom built truck hanger. That would do a better job. And at the time, I'd recently gone through the generative design certification through Autodesk's online training on Coursera. And I thought I would put generative design to use in building the truck hanger.

So let's first start talking about what generative design is. There's a lot of buzz around generative design in industry. And there are quite a few different technologies out there and quite a few different styles.

Usually, an explanation-- they will mention something like it being a form of artificial intelligence. But you'll hear all sorts of commentary around what it's good for. At the moment, where you're seeing it a lot is in things like aerospace. Where lightweighting is obviously hugely important. And generative design has been used to design very efficient structures.

Now, part of the reason it's sort of being found a lot in industries like aerospace, is that it still tends to produce components that are quite costly to manufacture. But that is improving all the time. With the cost of additive manufacturing coming down. But also, the cost of technology for things like 5-Axis CNC machining coming down, the parts that you can produce with this-- especially, now that you can design them easier-- is starting to become a bit more accessible.

Now, a skateboard truck is probably not a great example of a component that would be a good candidate for generative design, considering the usual cost and volume of the components. But I had access to the technology. And this was a bit of a "because I can" sort of a project.

So let's start having a look at the different types of technology. Autodesk has two main areas for generative design in the manufacturing space. They do have other generative design in the buildings space, as well.

But the two main ones that you see in the manufacturing environments are what we'd probably generically call either topology optimization or generative design. Topology optimization was released in Inventor a few years back under the name Shape Generator.

And it tends to create mesh bodies that look something like this. If you look at the surface on it, it kind of has an almost rock kind of texture. That's the way I like to think about it. And they tend to be quite coarse.

But these measures effectively form a good guide for where to remove material to lightweight your component. I don't know of anyone using the mesh bodies directly to produce components. They tend to use them as just an indication of how to improve the topology of their component.

So this technology, if you contrast it to generative design, generative design-- and this is a bit of a silly example of a stool that I created with both technologies, just to do a comparison. You can see generative design creates a much more organic looking structure.

Now, I'm going to list a few points here on each technology, to kind of give a bit of a comparison between the two. So I like to think of topology optimization in Shape Generator as being a subtractive approach. You start with a volume. And the software will effectively run a finite element, stress analysis on the component. And it will iteratively remove material and keep testing and try and remove material from the least stressed areas of the component.

It runs inside Inventor. And that means it uses local resources. So if you have a powerful machine, you get results back pretty quick.

It does, however, only generate one solution at a time. Now, contrast that with generative design. Which is more of a--I like to think of it as an additive approach. I don't mean additive manufacturing. I just mean the way it approaches the problem.

So it will add material. And you can see in this case, I basically created a seat and four feet. And it grew the structure in between. Whereas on the left hand side, I defined this as a rectangular box. And it removed all the material where it wasn't needed. So quite different approaches.

It's hosted by Fusion 360. But it solves the problem in the cloud. So you can take advantage of mass parallelization there. Which allows you to create a number of outcomes for different design problems or different design criteria, at the same time.

It's also manufacturing aware. And this is probably the biggest difference between this and topology optimization. Is that it will only generate structures that can be manufactured using the methods that you've chosen. And it can do multiple solutions in parallel.

So if you happen to have access to 3D printing, CNC machining and profile cutting. You can put those different manufacturing methods in as criteria. And it will generate multiple outcomes that take into account the different materials that you've selected, as well as those different manufacturing methods.

So let's have a look at the workflow. First of all, we start off with some starting geometry. Now, on the skateboard truck, there are a few key interfaces between the hangar and other components of the skateboard.

Obviously, the wheels have to attach to something. So I've got these representative stub axles on the outsides. These cylindrical areas, I left them there to give clearance for the wheel. In case it was going to grow some structure that might foul on the wheel. And I needed space on this side, on what is currently on the right, for the belt drive. So these are effectively just an area where I need space.

And in the disc is my motor mount flange. So that's where the plate is going to attach. You'll see the plate later that the motor attaches to for the belt drive assembly.

So those bits of geometry I need to remain there. That's what I wanted to attach the structure to. I have this bushing seat in the middle. Which is where the urethane bushes that support the truck-- they act as springs, effectively. That's where they mount. Well, that's where they seat.

And you'll notice that there's some clearance in the middle for the kingpin. So that the kingpin can rock backwards and forwards, side to side. And then you have this pivot at the top. Which goes into a polymer cup. And that provides the axis of rotation for the truck.

So I model these components. I actually ended up using the model that I found on GrabCAD of the hanger for the Paris truck hanger. Which is the standard off the shelf part. And then, I just used the modeling tools inside Fusion to sketch and cut away the material that I didn't want.

Once we got that, we need to define what's called obstacle geometry. And this is how we effectively guide the process to grow material in roughly the areas that we want and nowhere else.

So you can see the translucent, red bodies here. These are just fusion solids that have had a Boolean subtraction performed on them to remove the starting geometry. And that leaves me with areas where the software has to avoid.

So if you imagine that on these [INAUDIBLE] on the left and right, those faces are open, as well as the cylindrical face around the bushing seat. And the face on the end of the pin-- the structure that the software grows can attach to those, wherever it likes. And it can grow anywhere where there's no red volumes. So that starts to define your envelope. That the software is allowed to create its optimized structure to suit the design challenge.

We then go in to what's called the study definition. Now, this is the first look so far at the generative design workspace in Fusion 360. And this is the area where you configure your study for prior to solving.

Now, you can see here that there are some additional bodies created. That you didn't see previously in the starting geometry. There's this big rectangular block. This was a later addition.

And this is something that comes up quite a lot in the generative design process. Is that you may not think of all the scenarios that might occur when you first set up the study. And when you start to process the outcomes, you can quickly see if it's growing structures that aren't going to be suitable. And you might come back to Fusion, change your study definition, and then resubmit the job.

Now, that can be an expensive process. So you do need to take some time to make sure you've got as much of it correct as you think you can. Before you send the solve up to the cloud.

And you will when you get into the software and start to play with it. You'll notice that there's a pre-check and also a preview button in the Generate tab in the generative design workspace. And that allows you to see a very coarse idea of what the structure might look like. So it's a good way to just rule out, just do some sanity checks before you start to process the job.

So you'll see later, when I show you the different outcomes. But I was having a problem with ground clearance. And so I built this box around. Which has a slope face on the bottom, to give me more ground clearance. And so if you imagine all that area in amongst the red bodies. That's where it's constrained to growing this form.

So to work through this process, generally, you're moving from left to right across the top of the toolbar. And this class is obviously not a hands on demo. So I'll move through fairly rapidly. But I'm happy to answer any questions on this in the Q&A later.

So you have a study definition on the left hand side. Where you can create multiple studies in your browser, down the left hand side. And they can be different design scenarios.

So you can see I tried all sorts of things-- a low torque version, higher torque. I ended up realizing that the torque was actually a very-- the torque from the motor itself and the braking was actually a very minimal-- it wasn't much of a determining factor, in terms of the geometry. So I moved on to some other things to do with ground clearance.

And then, you can see at the bottom, the one that's selected at the moment is 5-Axis. Axis. Because initially, I was going down a 3-Axis machining route. And I ended up having a great opportunity to do a 5 axis version in collaboration with John Saunders from NYC CNC. Which you'll see a little bit later.

So the selected study is the 5-Axis version. Now, you can keep creating as many of those studies as you like. And you can solve multiple studies at the same time and then explore the outcomes together.

But there is also another button. I'm sorry, another control nested under study. Which is synthesis resolution.

Now, synthesis resolution is one to be aware of. If you're getting unpredictable results. If you're getting very noisy structures or structures that seem too coarse. You can adjust the little synthesis resolution slider.

And an analogy or an easy way to think about the synthesis resolution is like the coarseness of the mesh in a finite element analysis study. So the finer the mesh, the longer it's going to take to solve. But the higher the resolution, effectively, of the underlying structure.

So it pays to start off-- just to avoid too much solve time and then not being happy with the results. It pays to start off with that slider on the coarser end and then gradually refine that, bring it to the finer end and see what it does to your results. If you really want to optimize things and come up with the best aesthetic body at the end.

Edit Model allows you to enter a modeling environment within the generative design work space. Which allows you to do basic solid modeling features. So you can create bodies to represent fasteners. Or just create envelopes, those kinds of modeling techniques, without modifying your original design model.

So Autodesk is starting to put these Edit Model sort of work environments into the various workspaces. Which apply just to that workspace. So it's quite a nice way to work. Where you can have machinist or CAM programmers creating geometry just for CAM programming. And the generative design use creating bodies just for generative design. It's quite a nice way to work.

The next step, and probably the most specific to generative design tools that are in here, are these next two in the design space. So that's where you select the bodies that you want to represent your starting geometry, your obstacle geometry, and optionally, your starting shape.

So the starting geometry that we created in the modeling environment is called preserve geometry in generative design. And that's these green bodies. Those are the things I want to keep. They must be part of the end result.

The obstacle geometry is the red stuff. And so I've used that tool there to select the bodies that I want to represent the obstacles.

Now, this particular study, I didn't use a starting volume. Later, I started to play around with starting volumes. Where you can model a body. It works a little bit more like the topology optimization. Where you model a starting body. And then you remove or it removes material from that. But it can also add to that, if it needs to. So it just allows you to guide it a little bit more.

Once you've defined those bodies. The next step is to define your loading conditions. And this is exactly the same as a finite element analysis, stress analysis study. You define structural constraints and loads.

So you can apply things like pinned constraints, fixed constraints, frictionless constraints, and then loads-- pressures, forces. I think you can do remote loads now, as well. So there's a whole bunch of options there.

You can just see them there. There's a couple of little gray icons. Which represent some of the loads that are placed on this, as well as some constraints-- you know, pinning and fixing certain bits of geometry.

Once you've got that defined, the next stage is to start thinking now about what you want to solve. So you've got some design criteria options. You can choose whether you're focusing on light weighting or optimizing stiffness of the component. You can set a design mass. Or you can allow to just refine it, as much as it can.

And then the manufacturing methods. And this is arguably, probably the most interesting component to generative design or the most unique to the Autodesk solution, at least. Is that you can see here on the right hand side and the manufacturing panel. I can specify a whole lot of options for manufacturing.

So I've currently got that set to milling 5-Axis. But I could easily add milling 3-Axis and milling 2 and 1/2 axis. And then it would generate outcomes for all of those different types of milling machines. Additive for 3D printing, unrestricted, allows it to just create whatever structure it wants. And then two access cutting and die casting down the bottom.

You'll notice also in the milling panel. That it actually takes into account some of the tooling information. So things like the tool diameter. So it won't generate pockets that have radii that are smaller than the tool diameter. That I can't get to. So it allows for clearance and makes sure that this thing is actually going to be machinable.

Right up the top, there's also a cost estimation tick box. Where I can put my volume in there. And the software will use a third party service. That I believe is called a priori to calculate some sort of rough costings for the component, based on your production volume.

Once you've set those up. The materials tab is simply just choosing from a list of materials that you want as options for it to be manufactured from. And this is also a really handy tool. Because if I'm not already set on what I want to make it from. I can specify a range of materials. And then it will create outcomes for all the permutations of those materials and manufacturing methods. And allow me to later choose which is the best overall solution.

With the material set, that's the study defined. And then on the Generate panel, that's basically getting the outcomes ready to solve. So the pre-check will quickly tell me if there's any problems with the way I've defined it. If I've over constrained the geometry. Or if I haven't constrained it enough.

I can do the shape preview to quickly see the generated body. And then when I'm ready to generate the results, I can hit Generate. And it sends it to the cloud.

And almost immediately, I'd say within a minute on the Explore dropdown. If I click Explore Outcomes, it will bring up the outcome view. So let's have a quick look at what that looks like.

And these bodies-- I mean, this is at the end of the process. But these bodies will preview, as they're generating.

So a lot of these have been through a number of iterations to get to what you see on the screen there. But as they're going through those iterations, you can actually view the bodies in the outcome viewer. As they get generated.

So that's quite nice to see early on. If it's going way off course. And if it's going way off course, you might choose to cancel that study and start a new one.

So you can see here that there's a bunch of different options-- different materials, different manufacturing methods. We've got 5-Axis milling there. We've got unrestricted, additive, three axis milling.

And so you can very quickly start to see what you like the look of. See what you like, in terms of performance, in terms of mass. There's a lot of information to allow you to compare studies.

You can sort of create a short list pretty quickly. And then for the ones that are looking like good candidates. You can then start to look at them in detail.

And you can select any of them and spin the model around, zoom in and out. You can look at a stress plot to see the stresses on that component, based on the loading conditions that you have. And you can start to make a more informed decision about which one you want to manufacture.

When you've done that. And when you've chosen the one that you want to manufacture. You can export the design.

And you have two options on the Create tab. You can export a solid or a mesh. Now, if you're going to be 3D printing this, sometimes the mesh may be the preferable option there.

But if you're going to be doing anything else with it. Meshes tend to be quite hard to post-process compared to solids. So I tend to prefer the solid option, which is the little blue cube there.

Now, the nice thing about that solid download. Is when it comes down into Fusion, as a new design ready for cleaning up, it's a solid where the preserve geometry that you defined uses the solid features that you created it with. But the structure that it's grown in between is a free form body.

Now, if you haven't used free form bodies in Fusion 360, they're well worth a look. It's very easy comparatively. It's a very easy to manipulate geometry type using T-splines technology. And that allows you to clean up the body for manufacturing quite nicely.

So you'll notice here that on the left hand side, when I downloaded the geometry, I had a bunch of sort of extra faces that I didn't want. There was a hole in the back. And I wanted to clean those up.

So these extra faces can easily just be deleted by selecting them and hitting the Delete key. And then to patch that hole, I use the tool in the free form environment. I think it's called Delete and Fill. And basically just selected the faces around the hole. And it removed the hole and filled it up. So you can clean the geometry up quite nicely.

I did end up spending quite a lot of time on this geometry. You'll notice that on the front end near the pin, I created beveled edges. And I really tidied up a lot of the T-spline surface to make it look good once it was machined-- so adding in some flatter faces and things like that.

So you can take this to whatever level you want, in terms of tidying it up. Obviously, you don't want to make changes that are too drastic. Because then you start to affect the structure of the component that has been optimized, from a stress point of view. So these are fine tweaks, smoothing surfaces out, and things like that. Just minor adjustments for aesthetic reasons.

You can always run the component through another stress analysis, if you want to. Once you've done your clean ups. I didn't bother in this case. Because the tweaks were so minor. But that's the beauty of having that free form body.

Now, something to just a trap to watch out for. If you tweak the geometry where it intersects with the solids too much. If it breaks any connection between the free-form body and the solid. It will fail to solve the geometry.

So you need to be careful around these interfaces. And make sure that you leave enough intersection between the free form and the solid. That it doesn't break your free foreign body later and prevent it from converting to a solid.

So you'll actually notice in the timeline down the bottom, I've got a little red combine failure down the bottom. And that's because when I tweaked the geometry here, it broke away and created a hole. The surface was no longer watertight. And so it couldn't solve the solid.

But that's quite an easy fix when you know what to do. You come back in here and find out where the hole is and just pull the faces. So that they intersect the solid again.

All right, now, just a quick look at some of the outcomes that were generated. This is starting with the standard Paris 180mm reverse kingpin hanger. That's what it looked like to start with. This was the first outcome that I generated.

And I quickly realized that I had a problem with ground clearance. There was way too much material down the bottom here. You'll notice with a few tweaks, you can see the cone or the influence of that cone. That I talked about in the obstacle geometry. The cone allows for the truck to tilt. And for the kingpin to have clearance to move around.

And you can see where it hasn't generated geometry where that cone was. That's a nice visual example of how the obstacle geometry affects the outcome. It's still too chunky for my liking, didn't quite look like what I wanted.

This is starting to look a bit more interesting. You can see there's a really big hole at the bottom here. Where it connects to the boss. And that made me a little bit nervous, from a fatigue point. Right where the axle mounts.

So I tried a few other tweaks and revised that ground clearance geometry. And ended up with this structure here. Now, you can see, that this one after-- I just said before, that I didn't like that hole at the bottom.

I noticed that the next iteration had even more material removed on both sides. And you can see it through a slight change in the loading conditions. It added the structure up the top to support the pin.

So this was starting to look a lot better at the top half just based on, I guess, a little bit on aesthetics, but mainly on-- I guess it's an intuitive thing about for the loading that I was applying.

This geometry started to look a bit more like what I felt was right. But I didn't actually get as far as doing a stress analysis on the final result. So I only looked at the stress analysis plot in the generative design environment.

Now, I'm not an engineering analyst. So I would at some point like somebody who's skilled with FEA to have a look at the actual component. And tell me what they think about whether it's suitable. Because at 48 kilometers an hour, I'm not really excited about a fatigue failure. So I'm going to have to get that done at some point.

So with that design selected, the next thing to do was to make it. Now, I had that fantastic offer from John Saunders at NYC CNC, as I mentioned. He said to me, if you want me to machine it in 5-Axis. Let's do a collaboration.

So I spent a ton of time toolpathing this geometry. And this is another thing to be aware of. Is that while this geometry is manufacturable, I wouldn't say it's easy to program, from a manufacturing perspective for CNC machining.

It tends to be quite a lot of work. As good as the CNC tools are inside Fusion 360, there are a lot of little details to a complex, organic surface like that. That do take a lot of time to program.

So it turned out to be a big program with a fairly big cycle time, too. But John took it over from me. We used the collaboration features of Fusion 360 to link up.

And he got access to the design. And he refined the CAM. He did a lot of simulation. And he proved it out on his machine. And he put a huge effort in to get the final part manufactured. And he's got a really good video on YouTube going through the process.

With that, we got into machining. Now, this is jumping back slightly in time. This was the first version I did, as a three axis machined example. And I did this on a friend's HAAS VF2. So you can see it there after the roughing stages.

And then, you can see it. I ended up using my dad's manual lathe to turn the ends down. So I did all the milling on the VF2 and then manually turned the ends.

And that was actually a decent part. And it would have performed the job fine. But then, I had the opportunity to do the 5-Axis version. Which looked a lot more interesting. And so I decided to go down that route.

So this is not the actual final component on the left. But it's an earlier iteration. You can see it doesn't look quite as interesting or as clean.

But I took the 5 axis version that you've seen. And I machined this 2 and 1/2 axis motor mount plate that you see on the right manually. And then I waited for the component to come back from the US from John.

And this is what we ended up with. So there's a video on YouTube that goes through the entire generative design process of this component. If you're interested in more detail.

But this is it mounted in place on the board. And you can see, it's a really nice looking sort of organic structure. I ended up sandblasting it or bead blasting it to even the surface out. But it looks really good in place.

And so far, it hasn't broken. But like I say, being aluminum, I'm a little bit worried about fatigue over time.

So thank you very much for coming to the class and listening to this little adventure in generative design. And now, I'd be happy to get into more detail on anything that you found interesting and answer any questions about the topic.