Generative Design for Simple Lightweighting on an Electric Hypercar

PETER CHAMPNEYS: If I asked you to picture a design, which was created by an AI, what might you expect it to look like? I think often what we might expect to see would be something inspired by the highly optimized biological forms we see in nature, perhaps complex internal structures or organic forms which are highly tuned and optimized for their environment. Often, the designs we might expect an AI to come up with would perhaps be alien or at least characteristically unhuman in their design. Increasingly, however, with generative design technology, we're able to come up with design possibilities which build in good common sense engineering know-how and best practices to come up with design possibilities which build in all the advantages we'd expect from AI powered design, such as highly efficient optimized shapes or new ideas which might be completely different to anything we have thought of before, but which are increasingly simple, practical, and easy to produce using well-established manufacturing technology.

We might even describe them as increasingly human. Importantly, however, this technology isn't replacing the human designer. But instead, it augments and expands their design capabilities, helping them to create better design solutions more quickly than they ever have before. So today I'm here at Evolve, an engineering company in the UK, to take a look at how they have been exploring how this technology can help them to transform their design approach and add value to their design and manufacturing business. And I'm joined by Matt Hill, who is the founder and CEO of Evolve.

So over the next 30 minutes, we're going to be taking a really close look at how they have been applying generative design technology to one of their recent projects, which is to design one of the components within the assembly of the active aerodynamics of an electric hypercar. So we'll start by taking a closer look at the background of Evolve, their expertise, their capabilities, and experience. We'll introduce you a bit more closely to the project that they worked on.

And then we'll take you through the complete design to manufacturing process to produce this component inside a Fusion 360. And we'll explore in detail how the AI and the human designers are able to work together to create a highly optimized simple part which is very well suited for manufacturing on a CNC machine.

Matt, maybe a great place to start would be for you to give us a bit more background, tell us more about Evolve, your expertise, and what you do here.

MATT HILL: Yeah, sure. So back in 2014, I was working for a Formula One team and decided to start my own business. So I went about registering a company that today we know as Evolve. We originally decided to start offering services for R&D departments within Motorsport where we'd offer a complete turnkey solution that took away the pains I've experience from working in that type of environment.

We took on any project that had some kind of project resolution, anything from concept through to full delivery. So the business got off to a great start. We took on quite a few projects. And the projects got bigger, more complex. And, actually, we started to grow more customers.

We built trusting relationships with all of our customers because we were delivering on our promises and what we needed to give them to meet their performance goals. So in 2016, we decided to invest in some manufacturing equipment so that we could bring control of quality and delivery in-house. And that's the point at which we started to use Fusion 360.

Fusion 360 first was the way to start programming the machines. And for us, it was a big learning curve, as we'd never had any manufacturing experience before. But it was very easy to pick it up with Fusion 360.

So fast forward to today, we have a team of seven people. We have a purpose built facility here in Oxfordshire in the UK. And we provide still the R&D projects that we started out with. But we also provide anything from small design concepts all the way through to manufacture or complete turnkey projects that include all of that within.

We no longer solely focus on Motorsport. We have grown through recommendations. So we have customers in Motorsport, aerospace, automotive, medical, industrial machinery, and even clean technologies now. So our design ethos is always to provide the most elegant or efficient design solutions where possible to solve a customer's problems and whatever that problem might be.

So today we're going to talk you through a typical project of ours, which is involved in the active aerodynamic system of an electric hypercar. Unfortunately, the car is not yet in the public domain. So we can't show you the real car and components.

So what we've done in order to talk you through this next 30 minutes is to create an assembly that mimics the model constraints and the assembly within a car. And we've inserted that into a generic supercar model in order that we can show you our process today. So we were approached by our customer to get involved with the active aerodynamic system of this electric hypercar.

And for us, the areas of the system we're involved in were specifically the actuation systems to move the bodywork and also the components that hold everything in space and surround these components and form a subassembly in a subsystem essentially. So active aerodynamics is a way to move bodywork of a car in order to change its aerodynamic performance-- so whether that's downforce, or drag, or reduce drag. And essentially, we do that using hydraulic actuators.

So these actuators, or cylinders as we call them, we design them here evolve. They're quite complex assemblies. And they fit into this wider system assembly of other components that link the system together. So some of the key targets from the customer with this project are specifically weight reduction as well as performance.

And weight reduction is very important when we talk about electric cars because we have this trade-off in terms of mass of the vehicle and battery in terms of the performance and range that a vehicle can achieve. Battery technology at the moment means that to achieve the range and performance, you need very heavy battery modules in the vehicle. And therefore, there's even more importance on weight reduction of the components of the car that surround the battery in order to meet the vehicles whole performance specification.

So we decided to work with Fusion 360 and generative design to see if we could optimize some of the components in the system to really push towards this weight reduction target for the system components. One of the challenges we really didn't expect to find is actually finding the components that were suitable for the generative design process. So I personally have this preconceived idea that generative design is a complex process, very expensive, and creates organic solutions that are more suited to additive manufacturing than CNC machining. And this project really is focused around CNC machine parts for the production of this vehicle.

So the challenge was really to try and move that mindset into the fact that the tool could create something that was suited to CNC. And therefore, now we look at the component that we're going to show you today. It's a pretty obvious choice in hindsight. So the component we decided to use is the cross car mounting assembly plate.

This essentially ties the left and the right hand side of the assembly together with the actuators, the control manifold, the reservoir, and other electrical control items. Is essentially a flat plate. So really, the challenge for us then is to see how this process works, to see how we can come up with something highly optimized. We've really hit our weight reduction targets. We've still got high stiffness. But it's still manufacturable.

PETER CHAMPNEYS: Fantastic. So, you've got this design problem, essentially, of this cross car mounting bracket. You've got your objectives that you're trying to solve in terms of light weighting, optimization. So let's take a look at what that set up might look like for generative design inside a Fusion 360.

[MUSIC PLAYING]

So we've got our design problem that we are looking to solve. And that problem is going to consist of certain constraints, requirements for the design-- so things like it needs to fit within a certain area, and it needs to perform a certain function. So for this cross car plate that we're looking at, it needs to connect and support certain areas or components within the assembly. And we've also got design objectives.

And this is what makes one design better or worse than another. And, again, here we could think about things like the aesthetics of the design, where you want something that looks good but also looks in keeping with the style of the rest of the vehicle. And as Matt was explaining earlier, we want something that is as lightweight as possible.

So if we think about a typical design process, what we'd now start doing is maybe opening up a CAD program and start to model a solution or come up with solutions to the design problem. But when we use generative design, instead of starting to describe the design solution that we've come up with ourselves, instead what we're going to do is describe the design problem to the software and have it come up with solutions for us. Here we are inside a Fusion 360. And you can see the assembly that we're working on and the components that we are looking to connect.

And I can drop down into my generative design workspace. And this is where we're going to set up that problem definition that we're trying to solve. And you can see what that looks like here.

So the first time we see this, this might look a little bit strange. But I'm just going to talk you through the different steps that we go through to set up that problem description. So the first thing that we can see here are these red regions here. And we can see that these are really defining the wider assembly of the model.

And what these red regions represent, we can think about them as essentially keep out zones. And we're telling generative design, OK, the design that you come up with needs to avoid these areas. And the other thing that we can see are these green bodies here. And this is what we call preserve geometry. And this is, in some ways, kind of the opposite of those red regions.

And here we're telling the software, these are key elements of the design that we need in order for it to perform its core functionality. And for this cross car plate that we are designing, those are going to be the key connection regions to the wider assembly. Again, the primary purpose of this component is to support and provide stiffness to the rest of the assembly.

So between these two bits of geometry that we've defined, what we've essentially described is the space within which generative design can come up with solutions. And the next thing we're going to do is we're going to describe the requirements of our design in terms of strength, performance, and stiffness. And we can do that by describing different loading environments, different forces, which are going to be acting on this component during its lifetime.

And, again, here we're thinking about ensuring we come up with a design that is going to be fit for purpose in its environment. So what we've now done is we've described the constraints and requirements of our design. And now we can start to think about giving generative design an idea of those design objectives-- so again, what makes a good design or what makes one design better than another.

And again, what we can do is we can describe our-- we've got this objectives panel. And in this case, I'm going to give generative design the target of coming up with a design that's as light as possible. So I'm going to say minimize mass. And you see that we've got a few different options here depending on the specific objectives for the kind of design problem that we're looking to solve.

So we've defined our objectives in terms of performance. But another key element of the design is to think about the manufacturing process. So whenever we use generative design, one of the key dialog boxes is going to be this manufacturing dialog box. And what this allows us to do is to put in specific parameters that are linked with the kind of manufacturing method that we're interested in.

So for example, you can see here, if I was interested in a additive manufacturing process such-- or 3D printing, then we might want to think about things like our build direction and orientation. And, again, for a lot of additive manufacturing processes, we want to think about the support material that might be required if we print the part in those different directions. And we want to come up with a design that will require as little support material as possible and therefore as little post-processing.

And what you see is that we can put in specific information like the overhang angle and minimum thickness that my specific printer can print at. And generative design is then going to create different design options for these different build directions that are going to give us different possibilities. So we can then say, OK, this is a perfect designed print a z plus build direction. That's going to use as little support material as possible and require as little post-processing as possible.

So in our case, we were interested in a CNC machining as our manufacturing method. So if I select this CNC milling as our manufacturing method, you can see that we've got a few different options for the type of CNC machine that we want to use. So you can see a 5-axis, 3-axis, or 2.5-axis.

So if you're at all familiar with CNC milling, then you should be reasonably familiar with terms like 3- or 5-axis. But I wanted to just explain a bit more about what this 2.5-axis manufacturing method means. So a 2.5-axis design will essentially consist of a series of parallel plates.

Generative design will add pockets and bosses to the design to come up with an efficient engineering shape while continuing to constrain the design to ensure that the design is easy to produce and manufacture using an end mill. So here you can see when I select 2.5-axis as my manufacturing method, I can select a tool direction and a minimum tool diameter. And that minimum tool diameter is going to be a really key parameter that drives and informs the design that generative design comes up with.

So if I select a larger tool diameter, then that is going to mean that the shapes that generative design produce have larger minimum fillets. And what that means is that we can produce the design with a larger cutting tool, remove material faster, and produce our shape more quickly and at lower cost, whereas if I select a smaller tool diameter, that's going to allow generative design to come up with shapes that have a smaller radii on the internal corners. So, this is going to allow it to come up with more complex shapes and potentially might be able to better meet those design objectives. So, in this case, we might come up with a lighter weight design.

However, we're going to need a smaller tool to produce the shape. And therefore, we're going to be removing material less quickly. Our part is going to often take longer to produce. And our cost is therefore going to be often higher. So Matt is now going to talk you a bit more about what some of those design from manufacturing constraints look like down at the CNC machine.

MATT HILL: So a key part of the design process here evolve is design for manufacture, or DFM as it's more commonly known. So design for a manufacturer is essentially understanding what the final manufacturing process is and applying that as part of your design process so that the component can actually be manufactured. So for example, if we consider an additive manufacturing process, it might be possible to achieve an internal void or some internal structure within the component that would be very difficult to achieve and subtractive methods like sensory machining.

So in this case, we're using a subtractive manufacturing method of CNC machining on this machine behind me. This is the typical type of tool we use in CNC machining. So this is an end mill.

And we effectively cut-- use the side of the tool to cut away the material. This is a 5 mil tool. And as a general rule, this type of tool comes in lengths that are five times the diameter of the tool So this, for example, has a 25 mil reach. So if we consider making a pocket such as this, 50 mil deep pocket-- this 25 mil reach tool is not going to reach the button.

So to achieve this 50 mil deep pocket, we know we at least need a tool at 10 mil diameter. There could be a reason that you could need to use a 5 mil tool. But imagine a 5 mil tool at 50 mil length.

The stiffness of the tool becomes a problem. And you start to have tool deflection, which results in poor surface finish and difficult to control tolerances. So once we know we need an end mil of 10 mil diameter to reach the bottom of this pocket, that starts to define our internal radius. So this internal radius here would need to be a minimum of 5 millimeters for a 10 mil tool to be able to get into the corners.

And one thing we want to try and avoid when CNC machining is a change of direction of the tool, which will result in poor surface finish in the corners. So therefore what we do is we increase the radiuses by, say, another millimeter where possible so that the tool rolls around the corner of the pocket and produces a really nice surface finish.

PETER CHAMPNEYS: So we've now set up our problem description inside of generative design. So we can now generate outcomes. And generative design is going to generate multiple design possibilities simultaneously on the cloud.

And here we can see some of the designs that generative design came up with. And you can see that each one of these is a full CAD model. And I can actually see data about how these different designs compare against each other.

So here, for example, I can see the manufacturing method that this was designed for, the material, the weight of the design, cost estimate, and things like the factor of safety. And another great way to visualize this information is we actually provide graphs with these different design possibilities. So each one of these dots again represents a full CAD model. And I can open up any one of these if I want to get more information.

We can also change the axes of these graphs to compare different relevant pieces of information. So here, for example, I'm comparing maximum displacement against mass. So if we want a really lightweight stiff design, then we know that we're going to be looking for designs in this kind of bottom left corner. And again, if I open up one of these designs to look at it further, we can see that we've got these iterations along the bottom.

And I can actually click and drag this. And we can see how the generative solver has moved iteratively to come up with this final design. And what we can actually do is if we want any one of these previous iterations, these are actually all available for us to download and use if we want to. Another great tool that we've got is cost comparison.

So in this case, we can see that this bar associated with each CAD model is essentially an upper and lower prediction for what the cost might be. And this is powered by our partner aPriori, who use a machine learning tool to predict different costs. And, again, if I'm interested in a design that's as lightweight as possible but which also costs as little as possible, then this kind of view can be really, really helpful.

So I'm comparing mass against cost. And I can quickly see which designs might make the most sense. So once we've zeroed in on a design that we're happy with, we can go ahead and create that design. And that's just going to open up that design in Fusion 360. And you can see what that looks like here.

So this is our kind of raw output from the generative design process. But we don't have to just take this design and produce it, but actually we give this to you in this really editable format that lets you go in and really kind of put your own stamp on this design. So Matt's going to now talk a little bit more about what design for manufacturing looks like for Evolve when they're designing for CNC machinery.

MATT HILL: So now we've selected our preferred solution from the generative design set of solutions. And the form we have is-- it's created something that was very interesting, something that perhaps we wouldn't have created if we'd have done this from a traditional design method. And so the task now is to take that model and to modify it, to just clean some of those surfaces up and clean the edges, and add a human element to it to make this finished and polished product ready for manufacturing.

So now, as a designer, there are certain elements of design changes that I want to make this part. And different designers would do that in a different way. But for me, there are certain elements that I wanted to attack with this component. So firstly, within generative design, we could have selected symmetry for the component. And we decided not to.

But for the final solution, we're going to make the part symmetrical. So I'm going to focus on the one side of the design that we preferred, the one that has a bigger pocket, and therefore could end up with a more lightweight component. And the next thing for us to do is to clean up these outside edges.

So in this area of the component here, this is where it mounts to the vehicle and has a cylinder mounted to it. So we know that this area is going to be sandwiched between two of the components. What we want to do is clean the outside edge so that the profile matches the other components around it and looks very neat and pleasing on the human eye.

So once we've done that, there is a few other tweaks we need to make. And, literally, these are based on things like tangency. And this is more for the manufacturing side.

So we could make the component in the state it's in. But if we consider, again, the tool rolling around the outside of the component, tangency means that the tool will have a smooth motion throughout. And we'll get a much more consistent finish around some of these curves around the outside of the part.

So now we can compare before and after. So before we made the changes, this was the outcome from generative. And we can see that these profiles here continue to match the original component from the solution. And here we can see this is where we've added these changes just to make things a little bit neater for manufacturing.

So once we've got a finished design, we can now look at creating a jig for manufacture and consider how we actually hold this component when on the machine. So the first operation will be done using two Makro-Grip Lang vices in the machine on raw stock material. So this will take out most of the material in operation one.

Again, this is all programmed in the same file within Fusion 360. So, it means that we can carry that stuff from operation one to operation two. And by simulating this operation one, we can see that the tool paths are clear of the vices, that we're not going to have any collisions, and that we end up with a finished operation one ready for operation two with effective material removal.

So for operation two, this is where we need a dedicated jig. So operation one has created the first set of form of the component. And so we now need to hold that form in order to remove the remaining stock material and finish off the part. So we designed a jig within Fusion 360.

The jig is held unanchored so that we can do a quick swap in and out of the machine with that one. And the component is held by these side clamps, which are very low profile so that we can avoid them when using the end mills in the machine. So with the machine simulation in Fusion 360, we can now simulate the component machining with the remaining stock from operation one and see the machine movement itself in order to achieve the finished component.

[MUSIC PLAYING]

PETER CHAMPNEYS: So here we are now with the final part. It's been fully machined. Manufacturing process is finished. And I think what's really exciting about this component is that it represents a couple of ideas.

And I think first one of those is the convergence of the AI and the human designer. So what we see represented here is a design which was partly driven by the generative design process but also has the stamp of the human designer working together with the AI process. And what we've ended up with is a design which is better than either one of those two could have come up with working by themselves and has enabled them to come up with a design much more quickly than would otherwise have been possible.

The other idea that we've got represented here is the convergence of design and manufacturing. So typically when we think about design and manufacturing, we come up with our design concepts. And we then figure out, OK, how on Earth am I actually going to turn that design into reality? But with generative design, what we do is we describe those manufacturing process parameters up front.

And we actually build that manufacturing intelligence into the design right from the beginning. And we can really see that represented on this final component in many different ways. And, obviously, we spoke earlier about how that enabled us to produce a design very quickly on the CNC machine and very efficiently.

And, obviously, this component is relatively simple. It's a small part of a wider assembly. And we were able to see some fantastic weight savings with this component. So from the starting plate, we were able to reduce this down to almost 50% of that starting mass. And if we were to scale that up, obviously, we could take this design process, and we can apply that to many other components within the assembly, to many other projects that Evolve work on, and then the performance gains in terms of the efficiency savings, the performance, the fuel savings on assembly, cars-- all those industries that you work in-- can really be very significant.

So, Matt, you've obviously got many years of experience designing components like this. So what were your thoughts about how the generative design process differed to kind of your traditional design methods?

MATT HILL: So for me, there was really two main advantages that we found from using this process. The first is the speed. So the speed to get to a solution has been mindblowing, almost. The number of iterations that we might go through to get this level of optimization-- so design iteration followed by FE model followed by another design iteration-- and if you consider that, you can send that with all your constraints to the cloud and get a solution like this in such a short space of time is a huge benefit to us in terms of resource time.

The second, really, is that surprisingly for me, we've created something that's very aesthetically pleasing. The AI has created something that perhaps we wouldn't have created going through the traditional method. We might have kept more straight edges and not had this really pleasing outer profile. So the combination of the AI solution and the traditional design method to pull it all together has created a part that actually looks like fantastic. It looks really good.

And so for me, actually, the biggest take home we've got from all this is the change in our expectations of what generative design is. As we mentioned previously, we had this preconceived idea of these organic structures that the process creates, really only suited to additive manufacturing, when actually we've created a component like this by using the manufacturing constraints that works, is very easy to make, and produce a design solution that we're really happy with. I mean, for a company our size to be able to have access to this type of technology through an extension of the Fusion 360 has been fantastic. So we're really looking forward to having this as a tool in our toolbox for the future to really make our design process more efficient and add value to our customers going forward.