Drone Frame Design

Welcome back! Hopefully you enjoyed the introduction, so let’s get started. Just to recap:

Important attributes:

-         Toughness

-         Rigidity

-         Mass

And as per usual, they’re all diametrically opposed. However, 3D printed frames tend to have more issues in terms of rigidity than toughness- so we’ll primarily optimize for this. The most common mistake I see when looking at printed designs is that people follow waterjet cut carbon fibre designs- 2D plates bolted together. These designs are working around a very tightly constrained process- we paid for 3 dimensions on our printers, it’s best to use them.

Step 1 is mostly just going over basic fusion operations. If you’re comfortable with basic fusion operations, you can download the model here and skip over this step.

Step 1: Designing a basic limb

 

For this, we’ll be creating a basic drone arm with a lower rib to increase rigidity. This will double as a starting point for both our generative design and moulding. This part will be designed for a six part drone frame, with four separate arms sandwiched between an upper and lower plate. This is commonly found in larger drones, allowing easy replacement of a single broken arm in an expensive frame. Our carbon fibre frames will be one piece, for ease of production and weight reduction- but in this tutorial, it’s much quicker to focus on one arm.

 

Draw an outline for the motor mount

Complete the outline

Extrude the arm

Add a stiffening rib, like a T beam:

Step 2: Mould Design

Creating a mould is most easily done by first designing a part with moulding rules integrated, then creating a mould around it. The mould can also be designed directly, however this is less intuitive and some features may have unintended side effects. We’ll start by modifying the part with a “Draft angle”- a maximum wall steepness to ensure easy release from the mould. If two faces are parallel to one another, small defects in the mould (such as layer lines) may interlock with the part causing it to be impossible to remove. If a draft angle is unacceptable, a mould with more parts may instead be used- but this increases complexity of the process.

To create holes in the part, there are a few options. A large hole may be designed into the mould, but a smaller hole presents a problem. For something like an M1.5 bolt, there’s no chance of a printed pin surviving the process. To get around this, we can add a small nub to the mould in order to mark the location of the hole and then drill this later. This isn’t without issues, since it adds to the post processing work and also drilling can crack and weaken the part. Another option is to insert a steel dowel pin. The steel pin is so smooth that it can be extracted even with the parallel faces, leaving a very precise and strong hole. This does however increase the complexity of the mould more so than a nub.

 

Apply draft angle:

Create a base of the mould:

Create a “piston”. This is an extension above the volume of the part, which will be loaded with wet fibre. To consolidate the fibres this will then be compressed by the lid, squeezing out the excess resin and ensuring a solid, strong block of carbon fibre.

In order to ensure good clamping, we’ll add a lid which fits the mould. When the process completes, we’ll insert a screwdriver underneath the lid in order to lever it open.

Next, it’s time to analyse. In the following video, you’ll learn how to enhance your parts, zoom in on your parts, enhance your parts more and run it through forensics

The two tools used here are section analysis and draft analysis. Section analysis allows you to slice open the mould, and check for overlaps and gaps. An overlap may mean the mould won’t close, or it may mean that the part volume is smaller than expected. A void may mean that the opening is too large, or the part volume is bigger than expected.

If the part size is wrong, you may end up with a higher or lower fibre loading than intended. Too little fibre and the part will be weak, while excess fibre can make it hard for bubbles and voids to be squeezed out of the mould.

The piston is given zero tolerancing- this is important, since if there is too much space around the piston carbon fibre may be able to escape the mould as well as resin. If the piston is too tight, it may rupture the mould when clamped shut.

Next up, we’ll split the mould to make it easier to remove the part. We’ll add two M3 bolt holes for the main clamping force, as well as a smaller M2 bolt to prevent leakage near the bottom of the mould.

 

Last off, we’ll add tolerances.

Step 3: Generative Design

In rigidity critical applications such as this, it can be hard to know how useful your material is. Traditionally, this would be solved with FEA- a simulation to show stress in the part. Areas of low stress can be lightened, while high stress areas can be reinforced. Traditional engineering concepts like ribs and I beams can also be implemented to improve strength to weight.

However, We’re halfway through the year and I still haven’t mention AI so let’s change that. Generative design uses “AI” to do this for your, resulting in organic shapes that are highly optimized for you application.

We’ll get rid of all the mould parts, leaving only the part we wish to create:

 

Now cut in into 3 parts- the motor mount, the mount to the frame and the midsection of the arm:

Next, we need to make sure that the solver keeps material out of certain areas. Filling in the bolt holes would make the part much “better”, at the small cost of not working- so we’ll make some objects to block this:

Next, exit the modify space and assign usage to the bodies. Make the excludes red obstacle geometry, the mountings green preserve areas and the body of the arm a yellow starting suggestion (You don’t actually need a starting body, and can allow the solver to come up with something from scratch. A starting body can speed up the process and make the results better conform to your vision, however running it without a starting body can bring new ideas).

We’ll also need to add what loads are expected to be experienced. We’ll put the load down as 10N, even though it’s only going to be around 1.1N- the solvers will often generate odd results for heavily overbuilt parts such as this one.

 

The drone may experience unexpected load cases, such as lateral forces in a crash. Naturally improving strength in directions other than up will reduce strength-to-weight in the up direction, so we don’t want to overdo it. Generative design doesn’t support impact loading, and it’s also not practical to design a frame which will survive any conceivable crash unless you want to create a 50Kg tungsten brick. We’ll just eyeball it and call 7N good enough- if anyone questions this decision, tell them something about safety factors and hope they don’t listen too closely to the details.

Occasionally, I design something so wonderful that I decide I’m actually going to build it- like my 3D printed toothbrush holder. When this happens, having a design which is possible to manufacture comes in clutch- and the fever dreams of Autodesk’s compute clusters often don’t meet this goal. Having  a spread of 0.001mm spaced fibre lattice may be “optimal”, but actually creating it is another matter. Luckily, generative design allows for constraints on how it will be manufactured.

We’ll choose to use a handful of methods: 3 axis milling, 3D printing, 5 axis milling and unconstrained. 3 axis milling and 3D printing are generally the easiest to actually create, but compromise performance. Unrestricted will often make some really crazy stuff, with excellent performance but sometimes completely impractical to make. 5 axis milling serves a nice middle ground, but often creates geometry too fine for FDM or moulding.

Finally, set the study material. We’re gonna pick aluminium- and here’s a major deficiency of generative design. Generally it expects isotropic, linear deformation and does not account for crack propagation along layer lines for FDM. Forged carbon has similar properties to aluminium (for small loading), so it’s a good enough proxy for this. Overall, since we’re primarily looking for rigidity and not strength, this won’t pose too great an issue in practice.

In the video I set the mass a little high, you’ll probably get better results aiming for 4-5 grams. 8 grams is liable to give you a fat block.