October 15, 2019
October 15, 2019
By Matthew Payor
This article is about the manufacturing processes we’ve developed with Wren V1, with focus on foam core positive layups.
Coming into the Wren build, Daniel Mann and I had a long list of nagging points from previous work that we wanted to try and nip in the bud. We wanted to build a composite skin airframe where the skin truly composed the majority of the structure, with embedded electronics and compartments for different objects.
Whilst the production of optimal and attractive composite skin sections is relatively easy by layup on negative moulds, this leaves a lot of work to follow. The production of an internal structure, the joining of the skin to this structure in a way that is both strong and light, and embedding hardware ends up being a slog. It also introduces many possible weak points that require additional reinforcing.
The way I see it, every joint is either wasted weight or a weak point. It’s work that relies heavily on having multiple independently produced elements having close tolerances, using epoxies, filling and hiding joints. All these aspects increase the difficulty of the construction, move the final product further away from the CAD / design, and add weight that is hard to predict and budget accurately.
I’ve previously created laser cut plywood skeletal internal structures with accommodation for mounting hardware and wire routing. These projects came across issues locating everything accurately, producing satisfactory sliding fits, etc. – nothing insurmountable with some time and care, but seemingly suboptimal for the applications we’re exploring. It is still certainly true that this method is exceptional for lightweight model aircraft with solar film skin.
So whilst laser cut internals provide a lot of control over balancing weight and strength at different areas, the job of adhering the internals to two or more composite skin sections is something we wanted to avoid.
With our growing CNC experience, the solution that first presented was to incorporate both the mould and internal structure as one – a positive foam mould with embedded electronics, spars, mounting points and other features.
We posited that a single positive layup of the entire skin that grips tightly to a core with adhesion along the full surface area would make best use of the benefits of a composite skin – although bound to produce an uglier surface. Furthermore, since there was to be a cargo bay in Wren, it seemed appropriate to lay onto the inside walls of this cargo bay to create strong vertical sections in the centre of the airframe for rigidity (important for the quadcopter aspect of the vehicle).
The sacrifice with this method is surface finish, accurate external dimensions, and forgoing the safety of making multiple smaller parts and putting them together – a positive layup over a core with embedded electronics and other hardware is essentially a one-shot process.
The objectively superior process is one that involves precise matching positive and negative moulds, as demonstrated in this build video:
However, the level of skill and expertise involved in producing two sided thin moulds is something we didn’t feel ready to tackle with this build. A positive or negative mould on its own with a vacuum bag providing the other side requires far less accuracy, which means less difficulty in design and manufacture.
We began by doing some small scale rough tests of positive layups over a core with some scraps of carbon just to practice the process and discover the difficulties in the process. Luke Jackson, Arfin Trisakti and I did a quick and dirty lay on a 3D printed scale Wren model, and a hastily hacked up piece of foam that included the characteristic challenging geometries we’d be seeing on Wren – a cutout cargo bay and a sharp trailing edge.
These tests were very promising, bringing up both the benefits and difficulties of positive layups over a core.
The strength of the test pieces was impressive, and I believe that’s because the compressive strength of foam is amazing for its weight, leaving the skin to handle tension and torsion. Being tightly wrapped by the vacuum, loads should transfer well through the structure.
One difficulty is in achieving a decent surface finish without a negative mould, which is not just about looks but also strength, since wrinkles will effect performance under tensile load. Also, doing one layup instead of multiple to produce a shape means that naturally the geometry will be more complicated to deal with, since you’ve taken on more at once. This means that having a good plan and the right tools for the layup is paramount. Especially when gravity is working against you on the downfacing side of the core.
The next step was to come up with methods to embed electronics, connectors, spars, and other internal components. By splitting the positive core into two halves about the approximate wing chord, I was able to machine cavities and aligning features for internals, and then flip the foam using aligning pins to machine the outside profile. For housing connectors and other things that needed to protrude from the skin, we glue all these into the foam and then plug them up for the layup. After release, the entirety of the post work is digging out these parts, filleting the boundaries, and doing finishing work on the carbon fibre.
Here’s some vids and pics from machining the foam core for the prototype.
Simultaneously… 3D Printing
Whilst I was working on the foam experiments, Daniel Mann was investigating the promising features 3D printing could bring to a core material. The notion of an object being comprised of a skin and a minimally infilled core that is optimised to support the skin seems like it could outperform foam as a core, since building in strength only where it is required should result in a lighter and stronger item than having a constant substrate. The additional things 3d printing could bring to the table would be endless possibilities in terms of inbuilt complex connector housings, wiring channels and other things you would be unable to machine without great difficulty. So we decided to have a foam centre section and 3d printed wing stubs for our first prototype of Wren, to put these thoughts to test. Things wing stubs would include strong mounting points for the quadcopter booms, which we wanted to be bulletproof.
Here’s an example of a gyroid style infill with nylon filament.
Dan looked into the strength and weight of 3d prints of different materials, skin layers, infill style and percentage. Nylon proved to be very strong, but an impractical core material due to its compressibility – it was hard to break but easy to deform. It is also rather difficult to print due to its sensitivity to humidity and temperature, and its high shrinkage. PLA and its many variants were weaker but more rigid and much easier to print. All we really required was something that would survive the vacuum during layup, so this seemed to be the way to go. Dan ended up settling on a single skin print and a low infill that seemed to make the best compromise between strength and weight. Unfortunately, this still was about twice to three times the weight of an equivalent foam part. The main issue is that with a fixed nozzle size and extrusion characteristics, lower infill means scaling up of the infill pattern and thus larger areas of unsupported skin. Ideally, there would be some way to achieve a lower infill weight without greater spacing of the infill, by thinning the walls. Perhaps this can be achieved with a constant nozzle size by fiddling with extrusion parameters, but with typical slicing software we could not find a good solution.
We did our second positive test layup on a 3d printed wing stub section, this time using fresh fibreglass instead of offcuts and trying to achieve a good surface finish. This wing stub section had hollowed areas for batteries, spars, and connectors. The battery hollow crushed under the vacuum bag, and the trailing edge softened and smushed due to the heat produced by the resin cure and the low melting point of the 3d printed core. This taught us to plug up large cavities for future layups, and to keep temperatures down during the lay. The surface also had dimples between infill areas. Despite these issues the surface finish of the fibreglass was great, although the geometry was very simple to work with.
Some scientific stress testing…
So we concluded that plastic FDM 3D printed cores were an inferior alternative to foam cores for our specific application. Where they still excelled for us was in situations where we needed small components stronger than foam in the core, with complicated features across multiple axes. Mounting points were a prime example. The D-sub connectors we used for the wings screwed into small 3D printed housings that were glued to the foam. Where we wanted to bolt through the wing we inserted strong 3D printed sections to handle the compressive load of a bolt, whilst still maching the 3D contours of the wings. It is also worth noting that a 3D printed wing, without being reinforced with a composite skin, could outshine foam wings in terms of easily incorporating internal features and reinforcements, and providing a more durable skin than foam. Wings typically dont see compressive loads so there’s strictly nothing wrong with the weaknesses of 3D prints besides the weight (which could probably be dealt with by custom infill and smaller nozzle width) for making something that flies.
So all these elements came together in a first prototype / proof of concept that incorporated a CF tube skeleton, foam centre section, 3d printed PLA wing stubs and a 3d printed nylon forward flight motor firewall. The internal CF skeleton was designed by Rowan Whiteman and Arfin Trisakti with the intention of being bulletproof for the inevitable crashes we’d face during flight testing, and the completed core prior to reinforcement was already really strong. We were unsure at the time of how much weight and strength the positive layup would contribute so we were playing it safe there.
The gluing of all the parts together proved difficult. A significant problem was that with the compound angles of the skeleton tubes, shrinkage in the 3Dprints took us out of tolerance and we needed to really force some things to fit. Despite attempting to account for shrinkage in the modelling of the 3D prints, the nylon motor mount section fit especially poorly due to the high shrinkage of that particular filament.
We started by prepping the carbon tubes by drilling some small holes and heavily scratching around the joints to other tubes, and scuffing up the other surfaces that would be contacting 3D prints and foam. We then applied Scotch-Weld to the tube joints, and then a PU woodworking adhesive to the foam and inside the 3D prints, and then fitted everything together.
We weren’t prepared and had to improvise with the clamping. I sprayed some canola cooking oil on carboard and wood scraps and rigged up a monstrous concoction of clamps.
Then it was time to layup. In being frugal, and not knowing how things would turn out, we decided to use standard weave fibreglass for the prototype, two layers of roughly 65gsm glass. This also meant that we were diving in the more difficult end with the standard weave, and using twill in future lays would be easier to hit the geometry.
The layup proceeded pretty much perfectly, and simultaneously Daniel Wong and Arfin Trisakti whipped up some solar film & laser cut plywood outboard wings. There was one issue that occured after the prototype centre section went under vacuum – three of the four 3D printed wing stubs were out of a slightly different PLA (coloured black) to the last piece (coloured grey), we had printed all four in this filament and one of the prints failed, so we used a different PLA filament for the last piece. This one grey piece, which was printed with the exact same settings as the others, crushed under the vacuum. This wasn’t an insurmountable issue, but it was annoying because the CF VTOL booms were going to sit in adapting plates along the bottom of the wing stubs, and since one section was crushed, it was both weaker and also would seat the adapting plate at a different angle. What was a purely cosmetic issue was that the nylon 3D print squished under the vacuum. It did not break, but the nylon filament is quite flexible and it just compressed. It became super clear after this point that all foam was the way to go next time.
The post work on the centre section after the layup was to cut off the excess fibreglass, and make the necessary repairs to the crushed 3D printed section. I didn’t want to add too much weight in the repair, so I built up an area to roughly the original wing profile on either side of the stub using PU woodworking glue and various ratios of water as a kind of rigid casting foam, and masking tape as a form. I also built up a small section in the middle of the stub on the underside to seat the boom adapter correctly. Then I solar filmed over the section, gluing to the built up edges to create a fairing that hid the crushed section (but just had air underneath it). As an afterthought we decided to spray paint the section lightly with some leftover cans – it did not look great close up, but did the job.
Not really the topic matter of this post, but here’s some pictures of the prototype flight testing.
So what did we learn from this initial prototype in terms of manufacturing? We decided to primarily use foam for the core and keep 3D prints to small, high infill, low volume parts selectively reinforcing or providing mounting points at specific areas. We validated our assumptions about producing a structural skin over a core via a positive layup – the centre section was incredibly strong, to the point where Daniel Mann and I tried jumping and stomping on it and couldn’t do any damage without point loads of impact like a heel. This meant that we were happy to strip down the internal structure, which frees up more room and weight. Something we hadn’t explored was testing embedding electronics and plugs in the positive and laying over these things, but we definitely wanted to do this for the next build. So with all these things in mind, it was time to get started on a carbon fibre airframe.
The first parts we made were the wings, since there was less time and cost invested in making them and we could test out the last few little things on these without huge consequence if mistakes were made. I machined the foam cores with wiring channels, cutouts for connector housings, and electronics like servos. Daniel Mann tried using a 3D printer filament consisting of nylon with chopped carbon fibre reinforcement to make connector housings and other fixings for the wings, and they turned out amazing – the strength of the nylon paired with the rigidity of the fibre reinforcment to create things that could almost mimick an injection moulded part in terms of properties, we were very impressed with the filament. It was extremely temperamental to moisture and other environmental factors during printing though. I also machined end plates for the sections out of 2mm carbon fibre plate. The layup of these was super simple because purely wing geometries require almost no stretch in the weave when forming to the core. Vineeth Rao and I were able to assemble the wings including electronics and lay them up across two afternoons. The resulting wings were equal in weight to within 2%, and well within tolerance.
Next was the centre section. After machining the foam, we potted all the electronics and connectors into the bottom half of the fuselage. Then we glued the foam sections together using thin aluminium sheet and whatever else we had on hand to spread the load of clamps and weight. Then we glued on the end plates, plugged up the large voids like the battery cutouts, and laid over the core.
The finishing work required was cutting off the excess fibre from the layup, adding in the externally mounted elements such as the wing latches, and painting.
The wing latches that Daniel Mann designed connected both structurally and electronically the wings to the centre section. They were a major win from this build. Incredibly satisfying to use. The only problem was that they were quite heavy, and we’ll be looking to do something equally cool at less cost to our weight budget in future.
Here’s a few pictures of the completed Wren V1. The wingtips were routed carbon plate with some 3D printed housings for the radios, and the nose cone was a machined HDPE (with calcium carbonate filler) mounting plate for the hardware with a vacuum formed PET fairing.
ADD FUSION AND RHINO SCREENSHOTS + timelapse