Should I Use Automated Modeling or Generative Design? Both!

HEATH HOUGHTON: Hi. I'm Heath Houghton, a Professional Services Consultant with Autodesk. And I'll be presenting today my AU session, which is, should I use automated modeling or generative design? Well, I say both. This is class CP602678. So we'll go ahead and get started here.

Quickly, the first thing I need to do at all AU presentations is give the safe harbor statement, which means that I am an Autodesk employee, and I may make forward looking statements regarding future developments or planned future developments. Just understand that anything I say that has any kind of future development efforts may change without notice, and should make any purchasing decisions based on anything I say. There's also this presentation contains information and opinions that I may have personally about the software based on my own experiences, and also, that the content is proprietary.

Having said all of the legal statements, we'll go ahead and get started about what is automated modeling, what is generative design, when should I use one or the other, or both in my design process and my design making decisions. First, I'll just start with the new kid on the block. And that's automated modeling. And we want to understand what is it.

By the way, I'll be doing live demonstration in addition to presentation. So don't tune me out because you see present PowerPoint. It's the quickest way to get all of my ideas in front of you. And then I'll share everything live. So we'll use that format today.

But basically, automated modeling, the way I've heard it described by many people is sort of a super loft. So it quickly generates design alternatives just to connect existing geometry. It doesn't require a line of sight like a regular loft. So that's why it's called a super loft by many people.

You can select a bunch of surfaces, or faces actually, within the model that also tell the feature where you want to avoid other geometry. And it will create feature or geometry alternatives for you to use as a geometrical shape for your design. It doesn't require token consumption. It's a feature within the design workspace inside of Fusion 360. And the product is an editable geometry. It's in the timeline feature, and it creates a foreign body so you can then edit this piece of geometry that's created through automated modeling.

We'll contrast that with what is generative design. So generative design, a lot of people call it optimization, but I would say that it's a lot more than optimization. Technically, I can give it a real concise description of what it is. It's a multi-objective design exploration tool. It creates multiple designs. And actually, the more complex of a setup you give it, it can create dozens upon dozens of designs that meet your manufacturing requirements, your performance requirements, and even your cost requirements.

There are so many options that it actually is in its own separate workspace. It does require a lot of computation in order to do the optimization, or the multi-objective design exploration. And therefore, it does require token consumption or a subscription to the generative design extension. But there are a lot of benefits. It has manufacturable outcomes. You can do quicker prototyping, lightweighting, cost reductions all through the use of generative design.

So I'm going to do a lot of live demonstration. And we'll use a tutorial model actually from the Fusion 360 sample set. It's actually a tutorial model that is in both Inventor and Fusion 360. It's the front loader model.

It is as most real world of a model that I can use that doesn't have customer proprietary data from my consulting engagements. So this actually is a very good example of what the differences are, or shows a good example of what the differences are between generative design and automated modeling. And from the image on the screen, the part highlighted in yellow is what we'll use to show the differences. This is a structural part that uses the control, the tilt of the bucket. And that'll be the focus of our talk and demonstration.

But my own personal interjection here is, when would I use this automated modeling versus generative design? Is that for me, if I need something really quick, and just to occupy space, or just to show the concept of what a part might look like without any functional requirements, without any manufacturing requirements, then that's when I would typically use automated modeling. It's very quick and easy to create the connected part, if you will, the connected the faces with a form of a part. And you can use that as inspiration or as a final part depending on how the requirements of its function are.

But if you have detailed design definition where you have the loads, constraints, the functional requirements, the manufacturing methods that you want to use, and you want to explore a lot of design options that are basically prevalidated for your inputs that are more than just occupy space, then that's when I would use generative design. Just realize that there's more input so it requires more time on your part less than the traditional method of design something, then use simulation, or a physical testing to validate, less time than that, but obviously, more time than pick a couple of surfaces and get a form from automated modeling. So that's the two differences. And I'll demonstrate that here today.

So a quick overview of what I'll do in automated modeling and what it is, it's one slide. That's how quick and easy it is to utilize the feature within the design space. So it's geometry absent of any structural calculations. You simply select the faces that you want to connect. You select your obstacle geometry. And then it'll give you six design alternatives. You can change the thickness of each one, then you select OK, and it shows up as a feature group in your timeline. It's that simple. So I'll take you through that real quick in Fusion.

So here we are in Fusion 360. And we are in the design space. I've already run one of these guys, but that's a finished outcome. We'll start from scratch. So here we are. This is that part-- remember, the part that was in yellow was connecting these two connection points on the assembly. I don't have the whole assembly loaded here just to make it easier to focus on the task at hand. But we're wanting to basically connect these three faces and have a functional part.

So in the middle of your feature toolbar, or in your solids works toolbar, and the design workspace is the automated modeling button. So if you select that, you'll get your automated modeling dialog, right? So the first thing you do is select your faces to connect.

So I've selected the three faces that we would need to collect, or connect, and the three bodies, or the bodies we want to avoid within the assembly. If you have the entire assembly, you can start selecting more and more. But here, I'll just select the three bodies here so we don't have it give us a shape that interferes with something else in the assembly.

We turn the opacity on and off so it's easier to see for our really complex assemblies. Here, it's really not necessary, but there's just one little feature within the toolbar and within the dialog. So your options are to make a new component or a new body, just like pretty much every other feature that creates geometry within Fusion. Here, we'll just create the new body. And once you click Generate Shapes, it takes about one to three minutes to start showing up with design alternatives.

So here we can see it's preparing a few. It actually does go to the cloud and do some quick solving in order to create this geometry. So it takes just a minute or two to start showing up with some design alternatives. So this is fairly quick, but it isn't instant. It's about as quick as you can get to instant.

You can see now I have some showing up. So if I could start looking at them, I can see even as it's generating the final shape, I can see the different shapes that are showing up. And again, these don't have any real structural details in them. There's no stress calculations being done in the background, because there's no loads that we specified. They're just different alternatives.

One thing you should notice on the feature itself is that you'll see these little icons here, this has more of a organic shape on this icon. And here, you have a little-- it's hard to see, and it's maybe not hard to see, but it's subtle as you'll see an icon that shows maybe harder edges. So that means that there are sharp connections versus smooth connections. So if I were to select one of the top three alternatives, those give me smooth connections where everything's rounded from the faces that I've selected to connect. Whereas, the bottom three would be sharp connections.

Here, I actually like the sharp connection for this type of design. We have the square face that we're connecting. And it just looks better to me. I might alter the thickness.

If I go too thick, you start seeing where it's not interfering. But as this articulates, you would get interference. So maybe that's a bad idea overall to show too thick of a part in this instance.

So we'll pick this guy. And we hit OK. So we have here the part that was created. And what we see instantly is that it was inserted into the timeline as a feature group. So you see this group for the automated modeling feature, and you get inserting a new body, a form geometry, and then a fill. So that is-- and then a Booleans for all your obstacles. So it does go through a full set of feature creation within that feature group.

What's nice about it though, is that we can edit the form geometry if we desire. So I can edit, and push, and pull, and do things that I would typically do when I'm in the form environment of-- or the form toolbar set of the design environment. So I'm not going to do anything here to alter the shape right now. But you have that ability.

And in addition, it is in the timeline. So let's say for whatever reason I decided I wanted one of the organic connections, or the smooth connections, or one of the other alternatives, make it thicker, make it thinner, it is editable as a feature. So I can go straight into here and pick. Without changing my selections, I still have all these other alternatives instantly available to me. So if I want this one instead, and maybe a little thicker, I can do that. And then it becomes part of-- as it updates the feature set, it becomes a part of my geometry.

So I now have again, smoother connection to the faces that I connected it to and one of the different design alternatives. It just one of the nice-- it is actually like any other geometry feature other than the fact that it automates greatly the creation of the geometry. So that is basically everything about automated modeling. Like I said, there's no structural constraints placed upon the geometry, so it doesn't give you a how much is my safety factor, or any particular manufacturing method, but is a very quick and easy way to go through and get geometry to fill some space and give you a placeholder for either using this as an inspiration, completely redesigning something later on, or using generative to create a part as you get more design requirements, and a better understanding of what manufacturing method you might want to choose, items like that.

So yeah, that is automated modeling. And next thing we'll do is look at generative design, and the differences in what we would do for generative design. So with generative design, the differences here are that we have obstacles and preserves very similar to connected faces and geometry to avoid, but a little bit more detailed. And we have load cases.

So basically, you create and/or select your preserve and obstacle geometry. You set up multiple load cases. So this is going to have articulation when you're lifting the bucket up versus tilting the bucket up or down. So there's multiple load cases in the real function of this piece of geometry. And so the basic up and down forces on the linkages, right?

And we could even get even more detailed. There might be lateral forces that we might want to include as additional load cases for generative to take into account. I'll go through just the basic up and down forces in this demonstration.

The other thing that you have are your design criteria. So are you going to minimize mass or maximize stiffness with a target mass that you give the software? Are you going to look at how much production volume you might have in order to do a costing estimation through the software?

In addition, we have different materials for each manufacturing method. So you would select your material options for each manufacturing method that you want to explore, and the appropriate options for each of those manufacturing methods. So if you're using a two axis cutting, which axis are you using? For three axis milling, which are the degrees of freedom that you have for your cutting operations, the size of the tooling that you might utilize, all of that ARE options within the software. In addition, certain materials will give you the icon to show if they do have the costing information built-in.

So I will do a quick demonstration of the full setup of this model from scratch to the end. And then we'll talk about the differences and demonstrate to you why it might be easier to use automated modeling for initial just design ideas. And then as you get more again, when you get more information to utilize a generative design.

So back in Fusion, I'm going to go ahead and just set this up from scratch. So we're in generative. We're in the design environment. Like I said-- sorry-- we're in Fusion, and we're in the design environment, which is the default workspace for Fusion. To run generative, we would utilize the generative workspace.

Now, generative has a couple of things, structural being the one we want to utilize. We'll create a study for structural. And you see all of your parts that were in that assembly, including the automated modeling. Now, in this case, we're not going to utilize that part for any of the setup for generative. It's not required.

So the first thing you do in generative, you start specifying your preserved geometry. So we have a few bodies here that we'll select. And this is all the linkages where they connect, we need that geometry in order to put loads and constraints. And we also want that geometry to be part of our final outcome. So we would select those three bodies.

And in addition, we want to make sure we don't interfere again with the other bodies in the assembly. This outcome, we don't even need. So things that we don't need as part of the generative setup, we can just hide that as unassigned geometry. So that is basically the first bit of generative is just telling it what's to be kept, what's to be avoided.

And then the next step in the process is doing your loads and constraints. So for there, we apply those to our preserves. I'm going to go ahead and hide our obstacle geometry. So the first thing we want to do is give it the basic up and down forces that we might see at the two extremes of this. And we'll do that in one of the load cases. So go ahead and put the constraints on the faces here of where this linkage is, and make it fully fixed.

And then similarly, we're going to go ahead and put the load, the forces on the other end of the assembly. So these two guys will both get 1,600 newtons, yeah, or minus 1,600, so they're going to get a down force. And that's one load case. So we could design something based off this one load case, but this has multiple different types of forces that it would see in operation.

So we'll go ahead and make use of instead of redefining everything, of cloning. So I'll go ahead and clone that load case, and just alter the force here to be in the positive direction so it changes direction to positive. And that's a second load case.

Then we'll move even further along, and we'll give a force where we have a lot of load and it's actually being felt by the center preserve. So we'll make the load case three the cloning. We make it active, right? And then this force we're going to make it substantially higher. That's going to be in minus 10,000 newtons instead of 1,600. And like I said, we were going to change the constraint from this face, so these two faces on this linkage to the center component, or the center linkage area.

So that is giving us force for the structural load in one direction with the linkage seeing the force down. We'll make one more. It's again, another clone with force in the opposite direction, so again, making use of cloning, and then just editing that one load. So just change the direction. So we have force up and down being felt by the end linkage. And then force much higher magnitude up and down being constrained from the middle face here.

And then one more would be both ends seeing some force, or force down on this end with constraining the far end on the left. So here, we can just make a brand new load case and do the same things we've done before, so which is to give a structural constraint to the far end on the left side, and then a load to this guy so this is again, going to be fairly small in this case. We'll just do minus 1,600 again in the y direction. So that is all of the structural constraints.

Now, you can see already it's a little bit more involved than what you would see with automated modeling. You do need to know, or at least have an idea of the approximate forces that your geometry is going to be seeing. So for our purposes, it makes a lot of sense to use automated modeling when we don't have any idea of the forces, and then go into and start setting up your generative modeling when you do have an idea. At least that's my opinion.

OK. So we've gone through and applied the constraints and loads, which is what I've seen a little bit of understanding, a little better understanding of your design constraints is needed. The next thing is not the design constraint, but really, how am I going to go about looking at objectives? So do I want to minimize the mass? Or do I have a mass target goal.

So quickly, if I tell it to maximize stiffness, it's going to have a limit of safety factor, but give a mass target and make this thing as strong as it can for a certain target. So if you already have a pre-existing design, and let's say you want to just shed some weight by like 20%, then this is what you could use is maximize stiffness. That's one way that we have seen this done.

But if you really just know the safety factor and you want to get as light as possible of a design, and still make that a safety factor, then minimize mass is the option or the objective that I would typically choose. It's the most common that I've seen used. So we'll go ahead and utilize that here. And we'll increase that safety factor to four. And that's our design objective for any outcomes that we would get from generative design.

The next thing is, this is the part where generative has just a lot of capability. And that's in the manufacturing awareness that it has. So by default, we have unrestricted and additive. I'm going to go ahead and unselect both of those. And we're going to concentrate on milling and two axis cutting as our options.

So for milling, we're going to look at two and 1/2 axis. And here, we will actually even change the minimum tool diameter from the defaults. In this case, we're going to go ahead and go with five millimeters as the minimum. And minimum wall thickness of five is good. Going to change the head diameter to 40 millimeter for these designs. So that is one configuration. That's the two and 1/2 axis. And we're going to say the tool direction is in the Z direction.

We're going to add another configuration where we do three axis. We're going to include all directions. And then again, the tool diameter is going to be five. And the head diameter is going to be 40. Select that.

So that's for three axis. We'll go ahead and bump it up a notch and go to five axis, and have the same tooling diameter requirements, or tooling requirements. And in addition, we'll choose two axis cutting, and we'll use the Z direction as the cutting direction. And we'll utilize the five millimeters minimum wall thickness again to keep it consistent between the different manufacturing techniques.

In addition, let's go ahead and do one more. Let's do some die casting, right? So we'll do a die casting with the ejection. Let's go ahead and say it's in the x direction in this case. And for this, we're going to go ahead and give the minimal thickness of 10 just to keep it consistent yet again, between the different manufacturing techniques.

So we've gone through and specified different milling options, two access cutting, a die casting. So those are our manufacturing techniques. And that is setting up everything up into materials. And generative will have default materials, but let's go ahead and specify the materials. So by default, you can choose between all the different methods the different materials for each individual method.

For all methods, I'm going to go ahead and make a slight change. Yeah, so we'll go ahead and start adding the different materials. So for all of them, we'll go ahead and add stainless steel. We'll also look at some of our models, right? So we'll go with stainless steel 316, 316L, so go down on the list. And start making our selections.

There we go. So we'll just drag and make it part of all the methods. And then we have maybe we saw AISI 304, and maybe get a little exotic and go with titanium. Might be costly, but we'll see what the performance differences it gives us. So we'll go and choose titanium. So these are materials that we're adding to all methods. As a matter of fact, let's go ahead and go with an aluminum as well. So aluminum 6061 typically a pretty good choice. We'll add that.

Now, in addition for the milling operation, you might want to add a couple of different materials as well. So let's add in addition to those, we'll add a state at 10.5 390, and maybe an additional steel, maybe 1008 91 HR. Be good. And a couple of just generic, so steel, maybe some galvanized steel as well.

Now, for die casting, let's remove everything except for-- so we'll just start moving. We'll move everything except for a certain type of aluminum that we want to use for die casting. So we'll use a 356 aluminum. To access cutting, it has again, everything that we had on the rest. So it has aluminum. It has stainless steel, the 304, it has a titanium, so we're good. So that's basically declaring all the different materials for the different manufacturing techniques or methods that we might want to use.

So after that, you're ready. So we've gone through. And again, it's a little more involved, but we've gone through and give it our workspace, which is where are we going to preserve geometry, where are we going to avoid geometry, where are we going to put our loads and constraints, and different load cases. And then what are our objectives? What's the manufacturing techniques that we're going to use, and what materials we'll use for those.

If you're wanting to make sure that you have your setup correct, you can actually do a quick preview here. Now, this isn't running the actual simulation that runs in the background that gives you your optimized geometry for your outcomes. But this will show you your different-- potentially, if you have everything set up correctly. In your settings, you can have a low level of detail, a high level of detail just to get a better idea.

But basically, what I see here is, yes, it's going to link these three parts together, and we're going to get some kind of geometry. This is not an indication of the actual geometry itself. Just to give you an idea, if it is totally different than what you expect, then you might have issues.

So at this point, you would go to generate, and it will generate the design studies that you have active, or that you select. In this case, we've done one study with multiple load cases. And it would generate them in a cloud. Now again, this takes a little while. It doesn't take just the one to three minutes that we would see inside of automated modeling.

For the purposes of the class, I've already done that outcome generation. And so we'll go through the explore environment. And one of the really big things about generative is that it's going to create a lot of outcomes. And you could have multiple studies, or just one study. But with all of the different outcomes of different manufacturing methods, different objective ranges, you can see that we could have a plethora of options to choose from. The big thing about generative is that a lot of time and effort has gone into allowing us to filter through all of the different outcomes to find our best, what we as an engineer decide is the best outcome for us.

So yeah, it's going to give some recommended outcomes based on some predefined settings of what's important, which is for me, be low mass, making sure I achieve my safety factor, low cost, low neutral. If I wrote rise up, low cost, it might change my settings. But basically right now, achieving a low mass and a high safety factor is also important. So it's giving me give me a few recommendations right off the bat.

Users, obviously, AI. We have visual similarity we can look at between them to sort through all of the noise I shouldn't say noise, all the different design ideas. I look at the data in different ways. So that's just a pure thumbnail view. I can look at it in a table view and sort by what it thinks is the recommendations, or maybe by material manufacturing method, my cost, so I can find the lowest cost in the items.

The way to sort the data is very extensive. For me, what I would actually typically do is turn off and on some of the filters, or the properties that we look at, and then in addition, start filtering through things. So I don't want the really expensive parts. Might filter out some of the really expensive parts. And you might have seen my screen start to change a little bit.

I also might not want to look at parts where I have high displacements, right? So maybe I filter all the way down to where displacement is less than 10 millimeters. That's good. And then also I don't want really heavy parts. So filter down here.

So these parts didn't change. There are probably other parts in the background that did. So it looks like if I am sorting my mass, the lowest mass is a five, you know, sorting by mass, my lowest mass is a five axis milling part. Also, might be a little more expensive to make than the other. So if I were to by my cost, the lowest cost is going to start becoming the two and 1/2 axis milling options.

So I might want to look at-- I can see that now I have 2.5 axis milling different materials. Then three axis milling start showing up. And it's not that much more expensive. Instead of having to look through all of these, I might actually look at a scatter plot. And I might look at these part costs, and look at the mass and then materials.

So what I'm starting to see is there's a range here of low mass and lower cost. You might want to look at those in a different way. I might want to look at to see if they have a high enough factor of safety. So again, I'll look at maybe the minimum factor of safety is much higher than my prescribed for pretty much everything.

If I want, I can take a quicker look, and I can see again, this is five axis milling. This is doing sorting by materials. Let's look at the manufacturing method. So I can see three axis milling in the same mass region as I see for the five axis milling. So that looks interesting to me that we can get a three axis milling that's all on the same cost structure.

So I might start making comparisons, right? Let's look at these three and compare them, and look at them side by side. So these are very similar. It's starting to look at different materials that are used. It's basically the distinct differences.

The overall shape is very similar. So their mass is, or the amount of volume they have is pretty much the same. Their mass is going to change just by the density of the part itself. So they're all within that 130 kilogram range.

That's interesting. I've actually gone through here already and looked at some different items. Go back to my explorer environment, I've looked at some different outcomes. And I'm going to go ahead and do a quick sort. So I created an out-- look, this one. And this is a two and 1/2 axis option. So it's the least expensive of all of the milling operations.

And it also met my requirements of safety factor. The displacement was pretty good. And so what I decided to do was to create a design from the outcome. And this is where we're going to show the difference and what you get from generative design as opposed to automated modeling.

So I've had this software create an outcome. And you can open the design, and it becomes its own individual part. It has its own history timeline. And with this two and 1/2 axis option, depending on the manufacturing method, whether it's a little more organic of a manufacturing, like additive, or if it is very I guess, easy to create features-- I shouldn't say easy-- but more applicable to be creating features where you're having sketches and extrusions-- will actually create those for you as part of the outcome creation. So check this out.

That's that two and 1/2 axis design. It's basically pre-validated for my loads. I can actually-- if I need to do more load cases than I entered in in generative, I can enter this into the simulation workspace, and it will transfer all of my existing load cases to the simulation environment for me. So I can just add extra, not recreate my already defined work. So that's an option.

But that's basically the outcome that we would get from the generative design. So again, more functional requirements, cost requirements, materials are all considered as it's creating the design, as opposed to automated modeling, where you just select a few quick things and you get a general idea of the shape that you might see when you're connecting parts in an assembly. And so what we have here is a good showing of what you get from generative design.

On the left is me sorting through, or the ability of the software to sort through all of the different outcomes and give you a view, different views of the stress locations, and your design preview. But the ability to create your design outcome as an actual part, and on the right is that outcome as a part. And again, what we had shown was that you have your timeline with the different features that make the part.

If it is an organic manufacturing method, or something that's a little bit more an organic nature, you would see some form bodies and other features. And in this case, we have two and 1/2 axis, we're actually getting sketches, or profiles, and extrusions, which is really nice. You can contrast that with what we get from automated modeling, right? This whole discussion is, when should I use automated modeling versus generative design?

And for me, my personal experience is, if I don't have full detail on the functional requirements, but I do want to have something in the part, or in the assembly to take up some space, or give me an idea of what the shape might look like, then I would use automated modeling, because it's quick, it's easy, it's like a super loft as we said earlier. On the left is automated modeling output. On the right is my two and 1/2 axis milling. Certain, I think it's steel material. And with all of my load cases, for the tilting of the bucket and the lifting of the bucket in taking into account for the structural integrity of the piece.

So that's a good discussion of when we might use automated modeling versus generative design. And hopefully, you saw the difference in the amount of effort level required. But based on the effort level, the type of output you get from both of the different technologies within Fusion 360. I want to thank you for your time. And I didn't get to meet you in person if you're watching this on video, but I hope you enjoy your virtual AU experience. Have a great day.