Autodesk Fusion 360 Generative to Improve a Large-Scale 3D Printer

HEATH HOUGHTON: Hi, I'm Heath Houghton, and we're going to talk about using generative for fluid paths to improve a large-scale 3D printer cooling nozzle. That's me, Heath Houghton. I've been in technical roles since I graduated college, first in industry for 10 years, then working for software companies. I've been helping our customers get the most out of the simulation tools for quite some time.

Years ago, we used to get the question from customers, can the software do flow optimization for me? And I would always say, well, that's like the Shangri-La of fluid flow simulation. And it has always been run the simulation, use expert judgment to make a geometry change, run the simulation again to see if it was better or worse, and rinse and repeat that cycle until the design is good enough or you run out of time on the project.

Now, we have finally entered into the age where the software can do that optimization. And it's up to us, the users, to still use our engineering judgment but just to pose the question in different ways to get unique and high-performance designs out of the software. I hope to showcase that today with a real-world example.

Before I get into specifics, since I am employed by Autodesk and we are a publicly traded company, I need to give the Safe Harbor Statement. Basically, the summary is some things I might say could be forward-looking statements, and they don't present any promises.

So let's talk about the large-scale printer. I'm going to give you some background on it. It's designed in Fusion 360. And from my perspective, it is quite a good showcase of Fusion 360 capabilities, particularly in the area of collaborative simulation of a simultaneous design. It was designed in two different locations simultaneously, at the University of Warwick and at Autodesk's Birmingham Technology Centre. This is just another image. I wanted to set up the size of this thing. This should set up-- or I'm sorry. This should be set up in the Exhibit Hall. So after this session, if you haven't already seen it, I encourage you to go take a look.

There are other sessions built around this printer. I'll just name them out right here. MFG502106 is a 60-minute case study meant to showcase this printer utilizing Fusion 360's overall capabilities. This class in particular highlights the end-to-end workflow from concept to design to validation of design all the way to documentation. It also highlights that real-time collaboration I spoke of. The link to this session is provided in the presentation. There also is a technical instruction session that goes in a deep dive on handling large assemblies within 360.

So let's dive into generative in general for Fusion 360. For those of the audience that may or may not be aware of Fusion 360 Generative, in the past, it has all been for structural components. And that is our largest set of users today that will use it for generating structural components. But now, we also have a technical preview for fluid paths as well. And the elevator pitch for generative for structural is that it is a manufacturing process aware geometry creation tool that takes into account multiple load cases, engineering goals, and constraints to generate multiple outcomes per study.

An outcome from Generative for Structural Components is a single body component, and it is the solid body. Now, to compare that with Generative for Fluid Paths, this is a tech preview. It's much newer. It also creates geometry via goals and constraints. It is not manufacturing aware at this point in time.

So you have a very similar setup as Generative for Structural but with focus from a fluid volume perspective. You are defining fluid volume conditions or fluid flow conditions, and the software is working on the fluid volume. It has less inputs. So due to that, we are limited to one load case per study, and the outcome for now is one outcome for study.

One thing to point out is today, the software requires we specify the desired final volume. This means if we give a starting shape shown in the yellow on the graphic on the screen, the software will optimize the fluid shape as it pares down to the target volume. This image shows the progression from 100% of the starting shape-- it's just moved the fluid path around, but it's the same volume as the starting shape down to 90%, down to 60%, and then finally down to the final target volume of 40% of the starting shape.

This is an important piece of knowledge to know that it is optimizing at each point, at each step as it goes from 100% to 40%, but it does further optimization once it reaches that final target volume. So that is your most optimized design at that percentage of volume.

The outcome is also a bit different from Generative for Fluid Paths when you compare it to Generative for Structural Components. It is not a solid part, but rather, it is the fluid volume. Now, I know I have some inside [INAUDIBLE] here, and I know we plan on implementing an automated enclosure volume around the fluid. But for now, we have to manually create the encasing solid. I will show the most efficient robust way to do this in Fusion towards the end of the session.

So let's just dive into the study. I shared some background of this large-format printer where we had some images of the real thing. And here, you're going to see some images of a piece of it in CAD. But the basic background is that we want to freeze that plastic as quickly as possible once it is laid down from the extruder itself. The faster it solidifies, the amount of deformation will be minimized. Bridging will be better. Just your total overall part [INAUDIBLE] better.

From just that statement, you would think that blasting with air is what is needed. But we also don't want to put a side force while it's still molten from blasting with air from a specific direction. So we would like the flow to be symmetrical pointed just below the nozzle. And those are our goals-- cool it as quickly as possible so we get good airflow, make sure we have symmetrical airflow.

Now, we have a traditionally created cooling manifold. It's symmetrical, and it has an internal baffle to try and promote that symmetrical uniform flow I just discussed. So this is a bottom-up view of that traditionally created part. One thing is to note that the supply for this ruling in this case is a centrifugal fan.

And you'll notice that the location is not symmetrical with the part. And that's just the way that the geometry lined up in this case. And we need to make sure what does that do to our designs as part of this whole case study. I mentioned an internal baffle. This is another view from the side. You can see those two internal baffles that are in the middle of the part that's in the bottom of the screen circled in red.

Well, we first, before we even start doing the generative, we would like to understand, do we even need to? Is the part we have traditionally created already doing the job we need, or can it be improved? And to do that, we take the traditionally created part, and to me, it actually looked pretty reasonable, and I thought it would have a pretty good performance, and we might have a tough time improving on it.

When I ran this through Autodesk CFD to get an in-depth look of its performance, we could quickly see that even with the baffle, we did get a bit of a jet on the outlet that is closest to the fan. We also get preferential flow. In fact, it appears a couple of the outlet ports don't get much, if any flow at all, and the direction of the flow for a couple of the other ports is not pointed exactly at the center. It's going off to the side a little bit. So we have not a perfect scenario with this design.

And as we take a deeper look looking from the bottom up, we see that the flow velocity at the outer ports is shown, and it's definitely not symmetrical. When you calculate out the actual flow from each one of the ports, you'll see a range from just a little less than 1 CFM all the way up to 14 CFM. So this is well outside the bounds of what we were hoping to see from a flow uniformity perspective. And as a matter of fact, our goal for this case study is to achieve something similar less than 20%, maybe within 15% variation from port to port. So there's room for improvement. That's good. It allows me to talk about Generative.

Now, we've laid the groundwork, that this design could use some help. Let's set it up in Generative for Fluid Paths. As a primer, I need to talk about just what you're going to see on the screen as I go through the presentation. There's a lot of information here. I've tried it to walk through live and describe everything in detail as I go through the pics and clicks.

There's just not enough time in these sessions to do so. So I've embedded videos at the end of each section of this presentation that actually show all the steps and/or walk through the geometry editing timelines for that section. You can view these afterwards if you want more specifics or a more instructional experience.

Well, let's just get back to the setup. The first part of setting up a Generative study is body assignments. This image shows our Generative model. The color coding in Generative is that bodies assigned as fluid preserves will be changed to a green color as you assign them. Solid obstacles will be red. We can also specify an offset to those obstacles that will be a transparent red.

If you assign a starting shape-- and I'd highly recommend to always assign one. It's best practice for Generative for Fluid Paths-- it will be yellow. Anything from the design mode that we don't use for the Generative setup will be gray. And you can see that for these extra pieces of the assembly of the extruder and other fans and so forth.

For an overview of the body assignments on the study, we have preserves where the float inlets and outlets are located. We have the starting shape, which can be much more blocky if you desire. This adds a little more taper to the shape. And it can even be the original design. But anyway, it's in yellow here. And how primitive or refined you make that starting shape is all up to you. and Most of my designs is a very primitive starting shape. But if you have an initial design and you want to really refine from there, feel free to take it for a spin in that manner.

In this case, we have the extruder and the build plate as obstacles as well as the fan housing. Now, the fan housing isn't really required. It's not in the flow path, per se. But it does help give reference to where we are in the assembly as we're looking through results and looking through the outcomes from this Generative study.

We also have included an obstacle offset from the extruder and the extruder nozzle. These are not required. As a matter of fact, a lot of the pieces aren't required. It just helps you give more definition to what you want the software to solve for. But with the offset, if you don't have one, the fluid part can actually touch right up against your obstacles. Since we're going to be building a solid part to enclose this fluid path, it's a good idea to give you an offset, which leaves us some room to fit that solid fluid encasing part.

OK, so this is-- I've talked about the starting shape for a little bit. This is a visual for the starting shape of the study. Again, starting shapes are not required, but they are highly recommended. If you don't give a starting shape, Generative Design will attempt to auto create one for you.

But it's in my experience at this point of development that the auto creation is not 100% successful, and you have a little less control over what you're starting off with. If you have really complicated inlet and outlet setup, you could get something that you really didn't intend. So I would always recommend doing a starting shape. It's not a lot of work to do that.

The rule for starting shapes is that it does have to be a single body. So if you create one and it's multibody and they're all touching, just do a Boolean operation to combine those bodies into one. And another note is that when you're in setup and putting your goals and constraints for the study, the goals of the study, your desired final volume is actually entered in as a percentage of the starting shape. So again, one more reason to recommend that you just give a starting shape all the time.

One last note is that starting shape and obstacles cannot interfere with each other. The intent is that the fluid never touches, never goes into an interference mode with those obstacles. So if you start off with an interference mode, you're kind of exceeding those constraints right off the bat.

And here's an image with some reference. You see these green bodies that are touching the starting shape. Those are the fluid preserves. And they are fluid volumes that remain part of the final outcome. Now, depending on your goals, the analysis, and things, you might see them get more material augmented around them, but they will not be reduced away from what they are. They're typically where you would assign your flow conditions, so your inlet and outlet flow conditions and pressures. But it is not required that they all have a flow condition assigned to the surfaces of the preserves.

You can have internal preserves, which guarantee you some kind of a functional property within your design. If you think of maybe a throat of a valve where you want to present a flat face for the valve stem to open and close, that might be where you utilize preserves and obstacles to make sure that happens.

In general, though, you will mostly see preserves on the periphery of the fluid for the purpose of assigning the boundary conditions. One other caveat or rule is that preserves cannot occupy the same space as obstacles. So again, they cannot interfere with obstacles. They can touch. They just can't interfere. It's the same rules as the relation geometrically between starting shape and obstacles.

So solid obstacle geometry are the keep out zones where we can't have the fluid path, where we don't want it to go. Our obstacles are shown here in red. It's the extruder, the fan housing, and the build plate. We don't want the flow volume touching that plate or the extruder itself.

So we can also use obstacle offsets to give additional buffer around the solid obstacles. Now, the offset can overlap with the starting shape. What it does is just tells the solver, let's try to avoid this region. There is no hard path for where it doesn't exactly touch. So you get a little more leeway.

But in our case, we want to place this buffer around our physical obstacle of the extruder, give ourself enough room for two to three millimeter plastic case around our fluid volume. So I use the obstacle offset to get clearance around the nozzle and extruder for that physical part. And that's the purposes that you will use it as well if you endeavor on doing some generative fluids setups.

We are doing a fluid flow optimization so we need to specify some flow conditions. Here, we select the face of the inlet preserve to specify 50 cubic feet per minute supplied by the centrifugal fan. And at those outlet ports, we specify a zero Pascal gauge impression. This is the outlet condition in Generative for Fluid Paths. And we have iconography and color coding, which shows up when the assignments are made.

Now moving on to objectives and limits, the options are intended to expand as capabilities expand for the software. For now, reducing pressure drop is the primary objective. And in the background, not even an option to specify but it does happen, if you have multiple outlets, the default is that the software will try and naturally balance the flow between all the outlets.

In other words, if we have 10 outlets and 100 CFM, the software will minimize pressure drop as it goes towards its final volume. But at the same time, it's going to try to shape the geometry such that each outlet gets exactly 10 CFM within some margin of error. So that's the way the software works on its objectives.

Let's talk about target volume. I mentioned it earlier, but this is our other design limit. Earlier, I stated that the solver is doing the optimization as it marches towards a final target volume. You're getting a more and more efficient shape, but the final, most efficient shape is at this final volume number that you have entered here. Now, it is entered as a percentage of the starting shape. If you don't have a starting shape, you're entering a percentage of the body that we're going to create. So you just need to pay attention to that actual volume number.

Now, water is the default fluid material. So in this case, we're using air, and we need to change it to air for this case study. And you can also specify custom materials if and when needed. We get one outcome per study. Each study optimizes towards that target volume.

So in my mind, it's best practice to clone that one study multiple times and modify that target volume limit, that way you get multiple different-sized outcomes that are fully optimized for that size. This way, you can pick and choose, and you have the most benefit for setting up one study. You get multiple different designs. It makes so you can pick the best one basically. Otherwise, you will get one outcome.

So my best practice is enter in a low percentage of design volume as your first study, clone it, bump it up 10%, clone it, bump up another 10%. I would do this for like five or six. And then therefore, you get a plethora of designs to choose from. They'll all have similar characteristics. They're just optimized for that one volume.

Now, to solve in the cloud, you just select all the applicable studies, and they will solve concurrently. This is all the same as Generative for Structural Components. Like I stated previously, if you want to see the steps summarized in this section that I just summarized in this section, if you want to see them in more detail, you can watch the embedded video.

There's just not enough time in the session to do so. But if you can download the PowerPoint, this video is included with it. It's at the end of each section. So it shows going, in this case, through all the different steps of Generative, assigning the body assignments, preserves, obstacles, and starting shape, and the [INAUDIBLE] conditions.

So we've gone through how you would set up this problem for this case study. Let's look at the results, shall we? A little education on results, I'm going to do a little education on results doing in Generative for Fluid Paths. You have the Explore Environment and the Outcome Environment. So there's two separate environments where you're looking at the results of what Generative is giving you.

The Explore allows you to filter, sort, and really peruse all of the outcomes at once. So this is a great way to really find which ones you're wanting to look at in more detail and toss out the ones that you're really not that interested in. It might not be performative. It might not be the right shape that you're looking for, just whatever the reason is. And we can sort by those things, sort by the pressure drop, sort by the amount of volume that it has, sort by visual similarity. There's just a ton of things to sort through in the Explore Environment.

One of the things that I like to do is look at the graph view, which is very telling for Generative Fluid Paths in particular. If you plot the pressure on the y-axis and the volume on the x-axis, you can actually see the shape optimizing as it reaches for that target volume. Now, the pressure might be going up because it's reducing in volume so you're restricting the flow path size. But it's not going up in a random manner. It's reducing in a manner that is optimal at each step, not fully optimized, but more optimal than the shape before.

But what you'll see is as it reaches the target volume, you'll see all the sudden the pressure just start dropping along a constant volume. And it's staying at that volume, moving stuff. Instead of moving and removing, it's just moving the fluid around to give you a more and more optimized design. And you'll see that in my imaging. You see if you follow the graph from the right to left, there's two different outcomes here. And you can see them both hit a target volume and then reduce in pressure. I think it's a really nice way to look at how well the optimization ran.

So we want to look at the Outcome Environment. Now, the Outcome Environment is where you see one outcome in more visual detail or even compare multiple outcomes side by side where you can pan, zoom, and rotate. And like I said, more visual detail, so you could look at flow lines. You can look at the pressure contours.

You can look at it with contextual items on overlay, so the obstacles, the starting shape, the preserves. You can see all of those or turn them on and off as you really examine the model or the outcome. You can even look at previous iterations. So if it's just a little bit too small, what did it look like at iteration 21 instead of 23 where it might be a tenth of a liter larger or something like that?

So the other thing that you do in this environment is you look at the outcomes as a CAD geometry. So the outcomes become CAD geometry when you create the outcome or you select Create Outcome from the iteration. This environment allows you to preview what that might look like.

So you get a little more crisp of a view because when we do convert, your preserves become exactly back to what they were in the original geometry instead of a meshed form. And you get to look at that here. But this is also where you specify which outcome you want to have created as CAD geometry.

So you can view a decent amount of detail in the Outcome Environment. There's a lot of hard data. You see flow lines. You can see the pressure contours. But if you want to take a more deep dive look, my experience is that I take that into Autodesk CFD. It's made for validation work instead of creation work.

And I took one of the outcomes that I saw in the Generative stuff that we-- that were created from the Generative setup that we went over and ran this inside of Autodesk CFD. And here, we can see the side view and the top down view of the flow lines. And right off the bat, you can see that flow lines go to each outlet. They have a pretty good outlet flow direction. It may not be perfect, but it's really good.

But one thing I noticed through this deep dive is that we didn't get perfect flow balance. So we didn't hit our target of 15% variance. But it was much better than the original manually created design. One thing that I noted as I went through the Explore results was that it was telling me that these outcomes were completed yet not converged. That is what is cluing us into that we weren't going to reach that manifold balancing objective.

So looking at the bottom-up view, we can see the outlets and see the velocity. We see it looks OK, actually pretty good, but not equal. And the flow rates vary from 5.6 to 12 CFM. If we contrast that with what we got from the manually created design, it was just under 1 to over 14. So we're much improved, but it still falls short of my lofty goals of less than 15% variance, which is good.

That allows us to do some more exploration, do a further iteration, change the starting shape a little bit, and we'll work through that. But again, if you desire to see a bunch of pics and clicks going through the Explore Environment and the Outcome Environment and what you might do within Generative, I've embedded a video at the end of the section. So feel free to download and take a look.

Now, we had a really good but not perfect initial study. It was better than the original design by quite a bit. But I'd like it better. We will utilize the clone and modify workflow to accomplish that. So I haven't talked about the Edit Model Workspace, but in short, in Fusion 360, you have different workspaces.

You have the Design Workspace where you would create your assemblies and parts for your physical models. You have a Simulation Workspace where you might take pieces of those or whole assemblies in and do structural analysis or thermal analysis and so forth on pieces of your design assembly.

Well, in Generative, we don't want to work directly on the Edit Model because we don't want to change your design and documentation, but we want to be able to have that available to us. So we have this Edit Model Workspace where it is associative to the design model, but it's downstream.

So if you make edits, add stuff, it doesn't work upstream back into your design. It is literally right here within the Generative space as a method to modify and create things you need in order to run Generative. So for instance, in your Design Workspace, you probably wouldn't have a starting shape for a fluid or a fluid preserve. It's not a physical part. But we can create those within the Edit Model Workspace.

With that as a primer, we're going to take advantage of this workspace to make a modification to the port geometry, the fluid port geometry, those preserves, and see how that affects the design performance, see what the outcome that's created from a slightly altered starting and constraint requirements gives us in the performance.

So if you're in the Edit Model Workspace, you'll have always Generative Model 1. That is the direct reflection without any edits of your design model. And then on the timeline, you'll see additional stuff. If you're editing the model, creating additional features, you'll see it down in your timeline.

But if we want to make variations, we don't want to keep messing with the same model and overwriting our previous work, we'll go ahead and right mouse button on whatever-- for instance, in this case, Generative Model 1, and then select to clone that Generative model.

If you've already set up a study, which we have in this case, it'll ask you do you want to copy over or clone the studies as well. And the answer is Yes because you're reutilizing your previous work. So when you go back and you look at the setup, for what is still applicable to your now altered design, we'll have it associatively still working instead of starting from scratch.

So this will preserve our setup for how it applies to the new model, and we'll just modify this port geometry. I don't want to go through because we don't have enough time to go through all the pics and clicks that it takes to get to this model variant. But like I said, what we're trying to do is make the ports smaller in cross-section.

So we think that's going to give it a little bit better exit directionality centered towards the actual nozzle. I think it also gives it a little bit better options for the flow variants between ports. Let's the software have just a little more to work with. At least that's the theory, and we'll see how it turns out.

So we doubled the number of ports and decreased their size. We also decided to give the software a little bit more leeway and a more blocky starting shape. So you see just an actual brick right there instead of that tapered starting shape. So there's a video going through all the pics and clicks after cloning and then walking through the timeline to show how he made that model variant. Again, there's just not enough time to go through it in its entirety. So feel free to look through it at your leisure.

So let's see how that change in port size and number of ports affected the outcome that Generative gives us and how it affected performance. So here's an image of-- again, this is in the Outcome Environment. And what we have are four quadrants or four quadrants of the image. The bottom right is that original six-port design where we took the original part as far as the port, its exact location and size for the outlet ports, and let Generative tell us what it thinks would be the best design for our goals.

The other three are the 12-port designs. And like I said, we took those, used the same exact setup, 50 CFM at the inlet, zero pressure at all the outlets, and we went through and said, OK, 50% target volume, 25% target volume, 35% target volume, and that's what those are reflecting. So it's giving you the optimized shape at each one of those volumes.

And just looking at this in general, we can already see just from the flow lines that we have better outlet profile characteristics. When we looked at it with the pressure contours and with the data that we're given, we can also see that it does have a slight sacrifice of pressure drop. We have smaller ports. They're going to have a little bit more restriction, a little more pressure drop. So we're engineering give and take. But it is the most optimized shape for those size ports is what we're getting out of Generative.

Again, we looked at these in depth. We did a validation. So I took the 15% starting shape, 15% final volume outcome and ran it through Autodesk CFD. In Generative, I can tell that the pressure drop was not much lower. It was very little as far as a change between the 25% or 35% of target volume. So for me, let's just go smaller.

And we ran this one through Autodesk CFD. And you can immediately tell that from the flow lines, they're all pointing towards the center of the extruder nozzle. That looks great. Looks like it's pretty evenly distributed. It does have a slightly higher pressure drop. But again, it was well within our goals.

And what we see is that variance is actually really small. The highest outlet flow rate was 4.4 CFM, and the lowest was 3.9. So we have really done a good job of having this symmetrical flow aimed right at the cooling the molten material after it is laid out. So this is a fantastic result.

And if you look at it, there is a video again of comparing the results of the 6-port design versus the 12-port design or 6-port outcome versus the 12-port outcome and the other variants of the 12-port outcome with different desired final volumes. But again, we don't have time to go through all of that. But if you want to look and see how I would typically look at results, feel free to peruse the video after the session.

So we have that fluid volume, right? It had great results. What we need next is to get the solid encasing volume. As I stated earlier, the outcomes of fluid volume, we need to figure out a way to get the solid encasing volume. In the future, I expect that the solid encasing volume will be part of the just automated process. It isn't today. So I think this is really crucial for utilizing this great outcome is to show people the best way to get that solid shape in today's Fusion 360. So I'll go step by step here because it's crucial.

So as of late summer 2022, this is how you would do it. In the future, again, it might be automated. Once you pinpoint a design that you really like-- in my case, it was that 15% final fluid volume design with the 12 ports-- what you'll do is click the Create Design From Outcome button in the Outcome Environment.

It will take it a little bit. It does all the work on the cloud. And you can do multiple outcomes. So you have it give you multiples and open up each individually. But for our case, we knew which one we wanted. We honed in on one. When the conversion is done, you'll get some notifications.

Most importantly, you'll see a little green banner in the upper left-hand corner of your Outcome view. And if you click on that banner, the one with the green checkmark, it opens that outcome design, that outcome geometry as a new unsaved design. This is the actual [INAUDIBLE] of that fluid volume. It's [INAUDIBLE] body. So it's actual editable CAD.

To create the enclosing solid part, we'll utilize that fluid volume. Now, you could do traditional CAD operations like shell it out or offset. But in my experience, those operations on really complex geometry like you're going to get from fluids or from Generative for Fluid Paths, those operations can be really finicky in any CAD system.

So we're going to do a more robust workflow that we have-- I just kind of labeled it BRep, which is your physical geometry, to Mesh. And then you're offsetting or you're showing, and then back into BRep. So we're going to go from BRep to Mesh back to BRep, solid CAD. It's a very robust way.

There's a slight sacrifice in the accuracy of the wall thickness that you specify, whereas if you do CAD operations and you say three millimeters, it's always exactly three millimeters. So that's why it's so finicky. If you do it in this Mesh view, you'll get 3 millimeters on average, and you'll get a little bit of variance off that three millimeters depending on the mesh density, right? It's not even noticeable to the naked eye. So technically, it's less accurate. Just thought I'd put it out there, but it is very robust. And so that's the method I use.

So what we'll do in step one is use offset surface for the entire part. Now, it gives us another body that is just a surface body that is the exterior of that fluid, or if you think of it in the other sense, it's the interior of the walls of your solid part that you want. The next step is to go to the Mesh tab and tesselate that surface body. This is where we're starting to get into the Mesh process. So you'll hit the icon in the Mesh tab of design in Generative-- sorry, in Fusion 360 to tesselate that surface body. And surface body is what you've [INAUDIBLE], and just use the defaults for the tessellation parameters.

So from there, you will get this multicolored mesh face that it perfectly conforms to that surface body where what was that surface body is now a tesselated body. The default behavior for that tessellation process is to make face groups, but we don't actually want them. So the next step would be to combine the face groups. You select the Combine Face Groups icon and just box select everything.

What you end up with is now a singular face group tesselated body. At this point, we're going to click on the Repair Mesh icon within the Mesh tab. And the options we will choose are for the Rebuild Repair Type and the Accurate Rebuild Type. And here, we'll also enter in-- it starts off with a zero offset. We're going to enter in our desired solid part wall thickness. In this case, I chose two millimeters. And if you hit Preview here, it will actually show you what that looks like. It'll look a little bloated, of course, because it just got thickened, right? So you will see the actual preview of your solid mesh body.

So it's now a mesh. It's thicker than it was, or it's offset by two millimeters. We need to convert it back to solid geometry because we want to use it in actual assembly and not as a mesh part. So we will use the Convert Mesh operation within the Mesh toolbar of Fusion 360.

The options we will use are the parametric, and we will choose the organic method. So you'll have a little drop down for the type of operation for conversion, and we're going to use parametric. And the icon just below it first to the right is the organic method. And what that does is gives us a T-spline body.

And this is what that T-spline body will look like via the upper image here. It's a thickened solid, but it is fully solid. So we'll need to do a little bit of finishing work. We will do a Boolean cut with the original fluid part. It's still there. We haven't deleted it anywhere. It's still in the design space as one of the bodies. We'll do a Boolean cut to remove that fluid interior.

So now we have the shelled-out part, but then our ports are still closed off. So we'll do some Boolean cuts at the inlets and outlets to give us the exact inlet and outlet profiles now that it's a shelled-out part. So you can see what it looks like on the image on the lower portion of the screen. That is the completely shelled-out solid body that encases that fluid. And it perfectly conforms to that fluid so it's exactly what you want.

That is the best method, most robust. It doesn't take-- it's like six steps to go ahead and create the solid encasing part. Again, in the future, I expect this to be an automated process where there's zero steps. You just tell the software what thickness as you're generating your outcomes, and it would give you this part. That's my expectation, at least. It's not a promise, but that's what's in discussion now.

So here it is, that solid in context of the original assembly. You'll notice it doesn't interfere with anything. It matches up perfect to the centrifugal fan. It's pretty cool looking. It's nothing I would generate on my own in order to create a performance design. I don't know how many iterations manually I would go through before I arrived at something even close to looking like this.

But you saw from the performance, it does exactly what our goals are. And that's the whole point of Generative for Fluid Paths. Again, there's a video that walks through that whole from fluid part to solid encasing part that I just walked through. It's all the stuff embedded in the video.

One thing that I want to say is I know-- thank you for sticking around me. That was quite a journey, a lot of things I went over. But I hope you can see how we improved the design for the centrifugal fan-- for-- I'm sorry-- not for the fan itself. We improved the design for this outlet nozzle, the cooling supply when a centrifugal fan was our supply.

So we went from a design that we thought looked good. When you did a deep dive, not so great. Had an initial study, much improved, still didn't meet our lofty expectations. Did another round from what we learned from Generative in the first round and come up with a really performant design.

Now, in the end, if you go and look at this large format printer in the Exhibit Hall, you'll notice that it doesn't have the one I just showed on the screen prior. I actually has what's shown on this screen. There were some other design changes that led to using supply hoses instead of a centrifugal fan for the cooling air. So if you see it, you'll see this other part.

But I do want to note, Generative for Fluid Paths was also used to optimize this design. It was just a little more nuanced, and it wouldn't show well as far as really explaining and going through the whole process of seeing an initial part. So we chose to use the initial studies in order to do this session. I want to thank you for attending. It's been my pleasure. And please feel free to reach out to me with any questions.