• aditya mehrotra.

helicopter updates: trying some generative design with the helicopter [updates]

Yes so you heard right. We're playing with generative design a little and using a tube-frame chassis to do it. This tube-frame chassis was built for the helicopter drone we have, chili pepper, that we can hopefully get working by the end of the summer. The frame is a landing gear system and we're going to use generative design with aluminum to see if this thing will work. Here's the initial design:

This will go under the helicopter as landing gear. But first we want to optimize it and we'll do that by setting up a generative design study.

the first thing we're going to do is create a "ground" that this thing can sit on for the purposes of the design study. We're going to be pretending the design is plastic 3D printed and we're going to be putting (if the weight of the helicopter is 4.2kgs that's 42N and if the helicopter's maximum vertical acceleration is 500 cm/s^2 https://ardupilot.org/copter/docs/parameters.html we don't want it to really break). So that's 42*5m/s^2, 200N it need to withstand and let's put in a 2X safety factor so that's 400N.

Let's make our mass target a reasonable 0.5kgs for a set of 3D printed landing gear (and iPhone 11 is about 200g).

Let's make the mass target 0.5kgs (maximum mass we can add) and the safety factor 2 and we're trying to maximize stiffness. Because we don't want this thing to break. The mass isn't so much a concern because 0.5kgs in the helicopter with such a large thrust to weight ratio won't matter much. The extra weight will make it fly smoother. We'll do this optimization for both Aluminum and Plastic and we won't restrict the manufacturing methods this time.

Then we put 400N on the vertical face as that's where the load would be applied. And then we FIX the ground in the XYZ direction.

We will not be using the plastic when it comes to milling because that can't be calculated with Fusion as the error says, and then the only other error is the fact that this structure may be too large to 3D print but that's OK because our printer can handle it.

And since we have unlimited cloud credits because of Autodesk education :))) let's do this on the cloud and see what's what.

Now we wait...

And the outcomes will appear in this window and we shall go through them one by one as they come up and finish. We're bound to get a lot of outcomes as we now have unrestricted manufacturing. We might need to re-run and increase the weight limit thanks to the fact partially we're using aluminum but we will see. I think this will likely take the rest of the evening to generate so we have time.

So it looks like this thing is going to give us 9 outcomes and we already know that - again we need to wait and see. It'll be interesting to see how the outcomes looks and evolve. I also do want to note where this is even going so let me put a picture there:

That's where this thing is going to go exactly. It's going to replace our really janky version of landing gear we have now with the new design that's 3D printed on the printer. This will hopefully boost our stability on the ground - and once we print and install that we plan to redo all the software and setup of the Heli especially in AUTO mode and start taking it out for test flights. We might add a body as well depending on how things are going.

Also important to note that before this thing gets printed we will be deleting the ground but then we will combine that plate we saw with the actual structure because it'll be printed with the structure. I estimate this will take at least 1/2 a roll of filament so we'll have to decide to use wood or CF we really don't know at this point. It might be wood to give it some design and to preserve the filament but if we have CF that's definitely a stronger solution.

Here's a detailed view of what the outcomes are testing. We have (3x) aluminum additive manufacturing units, we have (1x) aluminum milling, (1x) unrestricted aluminum, (1x) unrestricted plastic, and (3x) additive plastic. We're going to look at and consider all of them. Unrestricted (for those wondering) includes CASTING.

It's also entirely possible we'll have to re-run this without the GROUND and make the ground not just a large block but individual "feet." Just because right now we're getting results that look like this:

Who knows what that is but we're leaving it running just for kicks since it could actually come up with something rather cool. Who knows. Let's try it. We should also likely FEA our design first next time. Just to see if it can even withstand these loading conditions. It also might be completely abstracting the shape at first and then going towards a solution that will eventually look like what we were imagining. Who knows, let's see what happens when the thing is DONE. It's also possible the load conditions here are way too high so we really should look at dropping the load to 200N and letting the safety factor be in the generative design itself.

Actually I just realized that we can do all the changes we're talking about in a new design while this one is running.

Also, I think we may have forgotten to define the initial shape of the body.. yea. So let's run with the SAME conditions real quick. With the new definitions:

Okay so now we have two studies running on the cloud at the same time. So I'm going to say this one will likely give us much better results.

Ah yes, this is the design output from the first trial. Good. Let's rerun it defining a starting geometry because yea... we'll re-run this at some point. But we're waiting really for the second study we created to give its outcomes that's going to be a fun one.

So for the second study things seem to be failing and Fusion says this is usually because the loads are too low or the material strength is too high. So let's just make it support 5000N and go with that for now and just see what it comes up with. We're really doing this more to play with generative design.

Soooooooo we now wait until the system gets finished with these design outcomes before we can run the rest.

So we're still getting a lot of "this thing the loads are too low or the material strength is too high and parts aren't load bearing."So now I think we're using generative design a little in-correctly. I think this structure itself is not a very load-bearing structure and to ask it to then generate a design from it doesn't work because the structure itself is not load bearing but. I want to try something. What if I do remove that starting body, select minimize mass with a safety factor of 2 and a load of 5000N and see what it comes up with? In the meantime it might be worth doing some actual math to design one of these things.

So the reason we're running that study isn't to get a final design. It's to get an idea of with this geometry which are the parts that are actually load-bearing. And those are the places we can put the pipes and then re-run the simulations. There may be a much better solution we aren't thinking of or can't really CAD.

So we should learn a little more about how fusion 360 den-design works:

These are both really good video that help us understand a little better how to use generative in processes. Knowing the loading conditions and defining obstacles is essential to fusion's generative design. What we're going to try next is defining an obstacle body as the fuselage of the helicopter and then define the loads more properly because I don't think they've been defined really well. Right now we're locking the face of the feet and we should be locking the bottom so let's try that so here are the changes we need to make:

(1) figure out how to calculate the force of landing of a vehicle like a helicopter such as ours
(2) re-define the constraints in the model so this works a little more properly 
(3) define an obstacle body that will help us generate more interesting and useful designs

So let's first make sure our loads are correct. Then we're going to define new constraints and a new OBSTACLE BODY.

So we defined new constraints as the bottom of the panels and we've defined the old design as an obstacle body so hopefully the design stay within that or goes around to form something interesting. Remember for generative purpose we'll also likely use the more accurate ABS material and not aluminum or whatever we were using. We're trying die-casting just for fun but using aluminum in the calculation.

Now it's time to calculate what force we really should be using to generate this design.

At a minimum we know the force the gear must withstand on most normal operations is the mg of the helicopter and that's like 5kgs*9.8 ~50N that's really nothing. But it also would need to withstand a side force of the helicopter rubbing against the ground or trying to tip over and etc like a torque from one side and a side force. So maybe let's add those in. And then there's a torque about the structure. There are a whole lot of conditions we are not currently simulating that we should be so let's grab a paper and actually figure out what they are. This could be our largest problem.  


Here's the math that resulted from looking at this for a little while:

So in conclusion: 

The Thunder Tiger e550 current setup...
(1) Hovers are about 0.244 deg plate pitch at full throttle
(2) Cd of main blades is at max around 0.008 
(3) Max thrust is around 2200 N
(4) Max torque on the frame is around 53 Nm

So this landing gear needs to survive a downwards force of 2200 N and a torque of 53Nm about the frame in the right hand direction with z-axis pointing up. So we have a 60Nm torque acting on the frame and a 2200 N force acting on it. And we're using ABS plastic. And we changed the base to be better constrained. Let's see what this generative design comes up with for additive manufacturing of ABS. Remember, we're minimizing mass.

The good news is, all the calculations wee just did will really help us in the setup of the helicopter. Because we know know what the hover pitch is and things like that.

The reason we're simulating that 2200 N downwards force is we want to be able to withstand if we leave the blades at full negative pitch - the forces on the landing gear. And the torque because in-case we have the rudder rotor at full opposite lock we need to withstand 2*60 Nm of torque so maybe we'll re-run this whole thing with those configurations - twice the torque to one side you know. I also think our torque might be in the wrong direction but we will see.

So the torque is in the correct direction but half the magnitude we want it to be. So we'll up that to 120Nm of torque maybe even though we really never should experience that because those conditions mean the helicopter is likely just going to fall over and crash. The stabilizer likely wouldn't let that happen.

It's already starting to look a lot better :)))) yay! This one I'm excited for look at the wonderful shapes we are getting and how it'll likely also withstand the forces we want it to! Yay we're actually designing things now :)))

It's starting to look a little strange hopefully it'll get cleaned up as it goes along but this thing is also telling us, so far, that those large vertical walls weren't doing anything to help sustain the forces oops.

I will also note that since the torque is only going in one direction we will likely figure out what side is "beefier" using FEA which has more force on it and etc. Then extrude cut the other side and mirror the stronger side to finish the body.

These shapes are starting to look particularly interesting. The factors of safety for aluminum are really high but the plastic ABS is down to like 5.69 which is good it means we're designing not just adding a crap ton of extra material.

I think I do, however, want to compare this thing to an FEA. So I think I'll run that in a separate design. We're running the whole FEA as a single part as that how it would be printed and out of ABS. We're going to see if the original design would work as well. We're using the same constraints and the same loads. And yes, of course, we're solving this on the cloud!

Wooooohoooooo, safety factor of 0.15 YAY so there are a lot of parts that need supporting if we decide to make this thing out of ABS. Just curiously, I want to see what would happen if it was aluminum. We're going to use regular 6061 just because that's what we're familiar with. 7075 is stronger but eh we're just playing right now it'll be made of plastic anyways.

As suspected we can see a few things with the design. The front right lower beams have an increased amount of stress due to the torque that's being generated. A lot of the small corners in the design have a lot of stress just from the load itself. So the good news is even with a 0.5 or 0.7 safety factor we're probably never going to be seeing a 2200N load. Also, we're not likely to use this design. We'll likely use one of the generative designs that are being developed online because they will actually pass FEA with 2.0 safety factors. After we edit them a little.

The aluminum ones are looking pretty good and even with their reduced material their high factor of safety might actually make them viable. We'll FEA them later out of ABS and not aluminum to see.

Now that it's been some time, we can start noticing some trends.

The first trend is on the right-front under-side of the design where the system seems to be generating a wider leg on the right with some stick like supports coming from the larger under-belly.

And then in the left-back we consistently see on that component some small amount of material being taken away from that area leaving some pipe like structure and holes. This is interesting. What we'll likely do is take the front right leg, and the back left. And mirror and connect them to create a body.

This one is also looking a little interesting from the structure its defining. It's only been through 18 iterations but this is lookin less thin that a lot of the other ones and the worry about the others being too thin isn't necessarily un-founded.

This one also isn't looking too terrible. It'll be interesting to see what this one turns out to be especially since it is in its later stages.

The aluminum are interesting but are also well past the point where the designs were useful. We'll need to go back and check those out if we want to use them they're too thin for ABS now.

Who even knows what's going on with this thing... if I'm being honest thought, I'm pleasantly surprised this isn't going terribly now that we've done some math to try to make it work. Some of them are getting really strange since Fusion is trying to optimize out every single bit.

Outcome 5 is now looking to be the most promising because it's looking like one of the fuller ones out of all of them with a lower factor of safety. And the design is interesting looking!

But outcome 6 is done! Let's take this and process it and see exactly what we'll be able to do with it - I have a feeling we'll be able to make something really pretty out of it after we fix some of the dumb parts.

Outcome 5 is going back to starting to look a little interesting. Let's see where this goes I think it's almost converged since the FS is at 2 at this point. This also means we'll likely not be able to edit most of it.

I think I want to run one more generative design study after this but one that has the center as a place to avoid which would be really interesting because right now all of the solutions go to the center.

The final study we are going to run is this one. Where we have added these random obstacles in the way just to see what the algorithm will do to minimize the mass. And we've defined the old design as starting geometry. So let's see! This study will be running at the same time as the other one. The reasons we're re-running is is because I really want some strange strange geometry some art-like geometry but one that also passes the constraints we have given it. So starting with that tube chassis as a base shape and the constrains for not having just direct legs going, let's see what fusion does!

Study 1's outcomes are becoming slightly terrible so thats why we are making a totally new solution over here. And we'll likely run one more after this that doesn't define the starting body as that frame and maybe one more that does but the pipes start much thicker? Who knows. So let's run that third study not identifying a starting configuration but with identifying the obstacles so let's see what happens if we do that. I have a feeling study 3 might actually give us something we want.

Ooh it's starting to look like our evil plan might actually work... maybe? Who knows we will find out later. It's looking promising though.

Part of me does say though looking at this side view we need to add a few more obstacles because otherwise it just makes a strange looking box at the bottom.

You know, looking at this, I'm starting to understand how bones evolve. These have very bone-like shapes they're strange... I'm also curious about what would happen if we added both the frame and the bars together as obstacles. Let's try that too!

Both study 3 and 4 are now going and I think 4 will give us the most interesting of the shapes. We're literally forcing it not to make the thing it wants to - I'm not saying that's good engineering but it's a lot of fun to see all the wack things it comes up with! I'm also not going to view study 3's results right now because they're not great for what we're trying to do. See the study we're running, study 4, is kind of like "so we don't want to just have a tripod that would be boring looking and maybe annoying to print and we don't want the frame design we had before because most of it doesn't do anything in terms of support it's bad, so I need to you find the in-between."

So it's literally like "find what's in-between study 1 and study 3 because I want a FUN design as well."

I also want to note, at the same time, let's try generating a suspension design just because I really want to see how this generative design works you know. Let's see the possible outcomes we can have.

Yes so we are trying to generate this kind of design and see if it works. And I'll put here how that goes but... but more importantly. LOOK AT THIS:

Nowwww it's starting to take those interesting shapes we wanted to see! I can't wait to see how this study finishes up that'll be a lot of fun. Part of the reason I'm playing with generative design is because of the solar car suspension system. I want to see if we'll be able to make a suspension that can fit our wheels and stuff, one we can actually make with generative design!

So we ran all the generative design studies and if I'm honest none of them turned out right. So for the design we're really imagining what we're going to do is go to a T-Spline entry.

So from the FEA, we know the rear area is trouble and the underside needs a little support so let's start in those two places... we're going to do this by strategically bridging sections of the thing.

We'll show the results of our different approach later.


0 views0 comments
© copyright 2019 | aditya mehrotra