Project ALTius
Part IV: An extended essay to help see the big picture.
In the first three parts of this series the author described performance glider projects as complex with many repetitive operations. The latter can be addressed with CAD software and then automated in standalone SAD (software assisted design) apps and homemade HAM (hardware assisted manufacturing) apps. He then presented some same workflows to illustrate their use. Most recently, he covered working within both weight and financial constraints. — Ed.
Project ALTius is a long story of disappointments, failures, dead ends and above all resilience and hope. As any good drama this saga has its fair share of surprises and unexpected turns of events. Here are a few of them:
Failure №1
In the last part we concluded the easiest way to build this glider was to 3D print it. However the common FDM technology (AKA ‘squished molten plastic sausage’) is not adequate for this purpose and we need to consider printing with resin instead. I advocated this solution based on an analytical argument — taking into account weight estimations — but in reality the change of direction was due more to experimental results. When I got to the printing tests I found out that my DIY FDM 3D-printer was not adequate for the job: a wing printed with a 0.4mm nozzle was too heavy and printing with other nozzles took too much time and it was not structurally sound.
Failure №2
From my estimation a large part of the weight was due to the foil — that is, the surface of the wing — and I switched my focus on making the wing lighter by using Oracover® and some sort of holes and printing without infill. This was done by processing the 3D model in OpenSCAD and cutting some hexagon or square shapes. I also needed to make some important changes in xflrwing to separate the ‘solid’ parts, in the leading and the trailing edge and also the hinge area, from the rest.
Incidentally this was similar to Kraga’s Kodo but the design was different and my wingspan was not a mere 1.6m but a larger 3.8m–4.0m. The hexagon/honeycomb shapes were nice but not possible to print on my DIY 3D-printer without supports due to overhangs of 120 degrees.
I finally settled on a rhombic structure, with diagonals 20mm and 25mm that was easier to print. Good results in printing, bad results in covering with film: the PLA (polylactic acid) print filament and hot air are not very good friends. You can likely guess what happened to the trailing edge. ABS (acrylonitrile butadiene styrene) behaves a little bit better but I had no enclosure for my printer to contain the fumes and therefore I could not print with this material.
Failure №3
And the first U-turn in the project: I am quite stubborn and I decided it was nothing wrong with my idea, only the FDM printing technology was limiting me — so I had to print the wing with resin! The problem was that the resin printers are expensive — similar in cost to a competition glider! — so I decided to build my own resin 3D-printer.
I needed something able to print 250 mm chords: an iPad 10" retina screen was available on eBay as a spare part, some red Plexiglas® or Perspex® plates, some electronic parts and I was finally able to build the printer. It turned out that the non-UV resin was quite expensive — around €60/kg — and the iPad screen background light was quite weak and I needed longer exposure times. This can lead to all sorts of other problems like ‘bleeding’. When I switched to UV resin I found that the iPad screen filtered a lot of 405nm UV light and my LED array, made from flexible strips with 405nm UV LEDs, was not bright enough.
Another problem was with the slicer: the slicers for resin printing were a joke. You either printed it as a solid form — that is, 100% infill — or you had an option to hollow the model in an external app like MeshMixer and put a hole in the model to let the resin entrapped inside drain. Or — and I think this happened quite recently — there is the option of an internal structure but that’s quite heavy. None of these options was good enough for my needs.
The Core Problem
Rohacell® used for solid-core is very light: 30kg/m³ or the heavier version of 50kg/m³3. The resin is 1150kg/m³ and because of the density difference I had to use an infill of 2.5%–4.0%. In addition I also needed a surface foil printed with resin and in order to compensate for this additional weight — actually bigger than the lattice structure weight — I had to use only 1.0–2.0% infill.
With such a low infill percentage I definitely needed some extra strength and this was only possible with a composite structure such as carbon fibre and resin. I needed molds and plugs and I had to add this feature in xflrwing and also I had to figure out the easiest way to make these molds and plugs. I considered 3D-printing them or carving a plug in XPS and adding a mix of resin and mineral filler — but I needed a DIY CNC and probably I would have to write a program to generate the G-code for the CNC.
There was another problem with the infill: all the patterns had closed cells and I needed a ‘drain-proof infill pattern’. This light infill had to resist to all kind of loads during the flight. I used a specialised program — Element Free from nTop — and I designed a complex structure starting from a geodetic rib or a 3D disser and carving drain hole spawns. Modeling this kind of structure was quite CPU intensive and needed a lot of time. I changed the OpenSCAD programs I used before as a starting point to add these additional holes for draining resin.
At this point I realised it made sense not to over-engineer the geodetic structure and instead of 3D-printing it just cut some geodetic ribs made from balsa which is one-tenth the weight of resin. And I needed an app to draw the ribs. However the idea of this structure was still very appealing and I wanted to test it. The price of resin 3D-printers dropped and now I had a €500 Anycubic Photon. The problem was the screen was quite small — it enabled approximately 10cm (L) x 5cm (W) x 15cm (H) printing volume — and I had to print it at half scale. I did a couple of test prints and everything was fine — the structure was sound and light. But it was obvious that this was a dead end. Printing and assembling a puzzle of 100+ pieces was not something I was ready to take on.
An Important Trip to Slovenia
Around this time I had another change to the project — maybe the most important change — both in goal and in technology / method. I had the opportunity to meet one of the top composite builders in Slovenia for scale and competition models who was interested in the xflrwing app. When I was demoing the program — it was an early version — and explaining how I was doing everything he was very quiet. This made me uncomfortable, as what I was probably doing was very basic for him and it was a waste of his time. But he asked just one question:
“How do you model the hinges?”
This was very important because it was the trickiest part in modeling the wing. I explained about the 30% position for the hinge in the .dat files and how this is used for the long axis. I considered this as a fixed point for the washout rotation in order to keep the hinge on a straight line. He told me that I have a good solution — and it was his turn for explanation:
Creating the tooling for a glider is very expensive and time consuming. There is the aerodynamic part, of course, but then somebody has to take this and create the mechanical design — 3D model, molds and plugs. This is expensive and you likely will have to hire an engineer with expertise in using the CAD software such as CATIA or SolidWorks. Then you take this design and move to the CAM phase which involves machining the molds / plugs in CNC — and CNC time is even more expensive. Then you start polishing which is also very pricey and time consuming. The combined cost of the CAD project and the CAM part will run to thousands or even tens of thousands of euros.
The weakest link in this whole chain is the CAD project because it’s very hard if not impossible to verify the results unless you have the finished prototype in your hands. My program was a shortcut for this weak link.
I was quite shocked when he told me he wanted to be a customer for my program but he needed some additional features: IGES / STEP format instead of STL — the de facto standard for professional CAD and CAM programs — and some additional elements in molds and plugs. He showed me the molds for the wing of his next competition glider — it was hidden on a workbench under a soft cloth.
It was simultaneously a high and a low moment for me:
- High — I knew that my software was quite good, at least for the amateur builder. For pro builders there was some work to be done, but I was on the right path.
- Low — In that moment I knew for sure that I will not be able to manufacture these kinds of molds and plugs in a reasonable amount of time and with a reasonable amount of effort: a four-segment wing and two-segment fuselage would need something like 16 different molds.
So, a dead end for molds and plugs. I had to find another way to build my glider. That said, I never abandoned the mold/plug feature in the program and I later added the missing elements. But at least now I had a better understanding of why the competition models are so damned expensive: there is a huge effort and a big cost in getting them to the market.
Thoughts on ‘The Market’ and Its Needs
I also realised that the market for me was not a handful of composite builders interested in my simple app just to save some costs or the glider enthusiasts who can afford an expensive model. My real market was the tens or hundreds or maybe thousands of hobbyists who can’t afford an expensive model but are interested in building one. I had to create the simplest method to build the glider in the shortest time using the cheapest materials with a still acceptable final result. Not only this specific glider but any glider similar to this type with wing and tails defined in XFLR. In other words: my goal shifted from a single, commercial-quality competition glider — the market was already quite saturated anyway — to rapid prototyping builds.
After the meeting in Slovenia I worked in parallel on software and hardware. I started working on the fuselage app in order to automate design of a F5J type fuselage. I added a lot of features to the xflrwing app and finally when I considered it good enough I had my ‘IPO’ in RCGroups. I also had to rewrite the whole output section to organise the code better because the number of output files increased a lot and, candidly, the code was a mess. I also wrote the app for the geodetic balsa ribs. After the RCGroups post I was contacted by two builders who wanted to build the glider using my model — the classical path with CNCs.
Tug of War
The pandemic lockdown also had a big impact. No more business-related travel with a lot of time spent in airports, planes and hotels when I could work on the software. My focus now was finding the method and building the necessary hardware. It was clear to me that 3D-printing with resin was the solution but the method — MSLA (masked stereolithography apparatus) bottom-up, projecting the image through a transparent bottom of the vat — was wrong. It was good for printing small figurines or structural parts with 100% infill but not good for light structures.
It was like a tug of war: ‘team adhesion on the lifting plate’ versus ‘team adhesion on the vat flexible film’ pulling the ‘light resin structure’ rope. The game is rigged because ‘lifting plate’ has an advantage — increased exposure time for first layers. But it’s not a sure thing: sometimes ‘flexible film’ wins. There are also cases when nobody wins and the resin structure is simply torn apart. With the kind of structures I had in mind this was high risk. It was time for another change in direction.
‘Bottom-up’ was changed to ‘top-down’ — this solved the adhesion problem on the building plate as there was no more flexible film. MSLA was changed to ‘moving laser spot’ — which solved a lot of other problems. However, I still had two very big issues:
- The complex 3D model of the structure and…
- There was no slicer available I could use.
In fact it was quite a regular structure and I could bypass the software to generate this complex 3D model and the slicer and just write G0 / G1 / laser on / laser off commands. The extrusion control is the major part for slicers and it was of no use to me. I actually wrote the app to control the laser for printing this structure. But somehow the structure looked very sparse. I started investigating alternative lattice structures.
Thank You Mother Nature
I started with what Mother Nature already had on offer: carbon, in two variants with one being very strong — diamond — and one being weaker — graphite. The diamond lattice is regular with beams starting in four directions from a central point and repeating in a recursive manner. The graphite lattice is with beams starting in six directions and is weaker but it has parallelism between planes and much easier to code. This led me to a simpler structure with a lattice based on cubes. It was very easy to compute and even easier to ‘slice’ — that is, to write code for the 3D resin printer which is actually closer to a laser cutter. We’ll talk about the HAM part later in this series.
The infill factor formula was similar with the infill for balsa: (3n — 2) / n³ with an approximation 3 / n². When I was researching graphite and graphene I found an MIT research paper about “a new 3D material with five percent the density of steel and ten times the strength, making it one of the strongest lightweight materials known”. While I was not able to replicate some fancy nano-material what really got my attention was “you can replace the material itself with anything, the geometry is the dominant factor. It’s something that has the potential to transfer to many things”. And they had tested it on 3D-printed models! The gyroid surface was the magical solution to my problems and it was time to abandon other lines of research and use it as infill. Not really a new failure but another change of direction.
I had to create the algorithm to fill the wing with the gyroid surface. It can be approximated by a simple symmetric equation (sin x cos y + sin y cos z + sin z cos x = 0) and it was symmetric in three directions — good for strength — and also full of holes also in three directions — good for draining the resin. It was not something new — it was first published in a NASA paper from the 1970s — and one of the properties was a “triply periodic minimal surface”. Also good, “periodic” and “minimal” was just what I was looking for!
It had another remarkable property: it divided the space in two equal exclusive regions which gave me an idea about how I could use it to create the algorithm for an infill of 1–2%. It was quite simple: let’s consider the typical laser ‘nozzle’ of 0.1mm — which is actually even smaller, around 0.06–0.07 mm. For a 1% infill you need to have a structure with a period of 0.1 mm x 100 = 1cm The infill line is repeated after 1cm, but this looked dense enough to me. I had to figure out how to draw the laser inside a 1cm x 1 cm x 1 cm cube and then repeat it in every direction needed. I used a 3D matrix of 100 x 100 x 100 cells (later 200 x 200 x 200) and computed the values of the function f(x,y,z)= sin x cos y + sin y cos z + sin z cos x . It sounds complicated and time consuming but it’s not — you do it just once and it’s fast. If two cells close to each other — either horizontally, vertically or in diagonal — had values of opposite signs then between them there is a point where the function was zero — so the point was on the gyroid surface — and I could compute the coordinates of this point with a good approximation using interpolation.
I was doing the final adjustments and optimisations, for continuous lines and for compacting the number of travel moves and for limiting them inside the wing, when I had to stop and abandon work. Why? Because I was not the only one interested in using gyroid as an infill and somebody else crossed the finishing line before me, soon enough all major open source slicers had gyroid as the new infill. No need to reinvent the wheel. Another failure? Not at all, it was just a confirmation I was on the right path.
From now on I could use an open-source well-tested slicer. And not only me — every builder could use these free slicers with ease. Almost all of the software pieces were in place and I could focus on the hardware part.
The project was about a 3.6m–4m wingspan glider. For practical reasons — the T8 lead screws I had at the time for the Z-axis were 100 cm and I also had some 110cm 1204 ball nut screws with a 100cm active travel — I had to settle to a travel of maximum 96 cm on Z-axis. Scratch off the 4m variant for now. I wanted to print a 90cm or 96cm segment in a reasonable amount of time, let’s say during weekend days in daylight: 15–16 hour outside the window of eight hours sleep. Fifteen hours for 90 cm, 16 hours for 96 cm, that’s 6 cm per hour => 1 mm per minute => a layer of 0.1 mm in 6 seconds.
I needed a vat big enough for the wing and maybe for printed molds and plugs. A minimum 30cm x 5cm x 100cm but better to make it bigger in order to use some tricks to do some ‘mirror printing’ — printing a part and the symmetric mirror which is to say the left and the right segment in the same time. In the end I settled for 30 x 15 cm and the whole volume was 50L. Filling it with resin was out of the question — again, the cost was similar to a competition glider! — so we need resin to float on a liquid with higher density — resin is 1.15g/cm³. A saturated salt solution was out of the question — salts in solution are very corrosive. There was also the possibility of using sugar — I did some experiments and was not completely satisfied. Glycerine (1.26g/cm³) was a good choice but at €2/L means €100 in the budget for a 1m tall vat which would OK for the moment.
As with liquid, the resin has a significant surface tension. This means that after printing out a layer a ‘small dip / descent’ of only one layer height is not enough and we need a ‘big dip’ of several millimeters, similar to the pop-up in the bottom-up printers but this time with less drama and no ‘tug of war’ game followed by a ‘big rise / ascent’. From the six seconds per layer we need to provision at least one second for this ‘dip and raise’ movement. Better make it two seconds and do it in a smooth move and let the resin rest after this movement — only four seconds left to print a layer. This is like drawing with laser the two walls (that’s 4 x 25 cm = 100 cm) and the infill lines — at least two times some very curvy lines.
Final Figures
We need a printer able to ‘draw with laser’ two meters in four seconds — let’s call it a ‘500mm/s class printer’ — we need this printer to be robust to sustain 16 hour print sessions, to be easy to build and above all to be cheap because we have a very limited budget. Tough task! In the last part we had left only €200 in the budget for tooling including this 3D-printer. Is it even possible?
“The best part is no part. The best process is no process. It weighs nothing. Costs nothing.” — Elon Musk
We will use the same principle and simplify everything. No tooling means no complex molds or plugs. We will use the good old vacuum bagging method but with some improvements: household domestic use products like vacuum bags for clothes and using the vacuum cleaner for suction. Even better, use 28cm vacuum film rolls used for food and the associated vacuum sealing machine. Instead of the traditional mylar film we will use a cheap elastic release film made from FEP (fluorinated ethylene propylene), the same polymer used in the flexible transparent bottom of the resin vat in MSLA bottom-up printers. And in order to keep everything in place we will use some simplified 3D-printed molds: a simple contour of the surface with a width of, let’s say, 2mm. I call this simplified mold a ‘bark’ — as in tree, not the sound dogs make. In detail we will vacuum bag a sandwich of:
breather — ‘bark’ — FEP release film — TeXtreme® — core 3D -printed with resin — TeXtreme® — FEP release film — ‘bark’ — breather
I know it sounds crazy but when you think about it — I did think about it for several years! — and you eliminate all that is not really needed and leave ‘just the bare necessities’ … it feels like the right solution so it must be the right solution. The puzzle is almost complete after the bombshells “the easy way to build a performance glider is to print it with resin using a gyroid infill” and “no special tooling needed, just build a cheap 3D resin printer and borrow the food vacuum sealer from your kitchen”. All the pieces are in place except an essential one: the 3D-printer. But not for long.
Let’s Pause for a Moment for All Recap
In fact this whole part of this series is an extended pause to help you see the big picture. Project ALTius is not about the ALTius glider, as in my glider — it’s about the method and the set of software and hardware tools you need to build a glider, as in your glider. And by choice I’ve set up a set of hard constraints for:
- Software — open-source, complex yet easy-to-use, multi-platform
- Tools — cheap laser cutter and 3D-printers that are easy to build
- Materials — no Rohacell®, CarboLine® or other exotic stuff but rather just UV resin / balsa and affordable spread carbon cloth
- Time of Execution — all these constraints are derived from my initial target of “€500 glider build in a week”
However, let’s be candid: we will never get the perfect looks of a commercial competition glider — at least I know I am not able at this time. But we can challenge them in other areas: we can get a similar weight and a similar flight behaviour but with a faster time of execution and much less expensive. When your bird will be up in the sky you will not be able to see its ‘less than perfect’ wing surface. And if we get this right we will pass the ‘affordable’ mark — maybe we will not reach ‘disposable’ yet but we will definitely be near “if I crash, it’s not the end of the world, I can fly it repaired next week or a brand new model in two weeks”. An interesting dilemma because repairing the crash will cost €50 to €100 and building a new one €100 to €200 in materials.
An Analogy
Here’s a familiar car analogy to compare to F5J competitions: right now on the track there is an exclusive club of Lamborghini / Ferrari / McLaren / Porsche owners. But you can buy (or pimp up your own) Toyota / Hyundai / Ford / Skoda / Renault / Fiat / Dacia / Tata with carbon parts and a super-motor for a tenth of a super-car price. And the brand of the car will not matter anymore but the skills of the person behind the wheel. That person is not just the pilot, they are also the designer of the car and the builder and the mechanic as well.
Makes sense, no?
And a Confession
I’m not a pilot nor a builder — due to my travel and job constraints I could not spend much time in the field or in the shop. I see myself more like a ‘technology enabler’. I used my time trying to solve some software or hardware problems. And in the weekends I like to spend time with kids in the local club teaching them CAD or recently writing these articles. There is a famous quote by Ralph Waldo Emerson:
“It’s the not the destination, it’s the journey.”
For Project ALTius it’s both, I’ve set up a desirable but distant destination and I’m prepared for a long but also interesting journey. It’s up to you if you decide just to read a travel blog or start to prepare for your own journey.
Next month, it’s back to the nuts-and-bolts. For now, if you have any questions feel free to add them in the Responses section below. You get there by clicking the little 💬 below. Thanks for reading. Until next time, best of luck with your project.
©2023 Tiberiu Atudorei
Resources
- CATIA from Dassault Systèmes. — “delivers the unique ability not only to model any product, but to do so in the context of its real-life behavior: design in the age of experience…”
- Carboline® — “a global manufacturer of coatings, linings, and fireproofing with offices and manufacturing facilities worldwide…”
- Element Free from nTop. — “The nTop Ed community is bringing together educators, researchers and students … how to access your free non-commercial license of nTop…”
- G-code on Wikipedia. — “the most widely-used computer numerical control (CNC) programming language. It is used mainly in computer-aided manufacturing to control automated machine tools, as well as from a 3D-printing slicer app…”
- Kodo from Kraga. — “Kodo is a proof of concept that 3D printing can be used for building RC planes. It was designed as a multipurpose glider that does it all…”
- Meshmixer — “state-of-the-art software for working with triangle meshes…”
- Photon from Anycubic. — “LCD-based SLA 3D-Printer … supreme accuracy for highly detailed prints…”
- OpenSCAD — “software for creating solid 3D CAD models. It is free software…”
- Oracover® from Lanitz-Prena Folien GmbH. — “our leading product for covering RC model airplanes is patented worldwide … permits re-positioning without fear of colour-layer separation…”
- Perspex® — “suitable for glazing, furniture and display cases to name a few applications…”
- Plexiglas® — “the first cast sheet made of polymethyl methacrylate … stable, transparent and impact-resistant polymer…”
- RCGroups thread for Project ALTius. — “altius, citius, fortius — sounds familiar? That’s the Olympic motto where ‘altius’ means ‘higher’. But the spelling (ALTius) is related also to my initials — Atudorei Lucian Tiberiu…”
- Rohacell® — “For 50 years, Evonik’s ROHACELL® structural foam has been offering the aerospace and automotive industries, medical technology, and other markets boundless possibilities for lightweight construction…”
- SolidWorks® — “SOLIDWORKS® and the 3DEXPERIENCE® Works portfolio unite your entire ecosystem…”
- TeXtreme® — “spread tow reinforcements are a uniquely adaptable, safe and ultra light supportive solution for your carbon fiber composites…”
- XFLR5 — “an analysis tool for airfoils, wings and planes operating at low Reynolds Numbers…”
- xflrwing — “STL generator for an XFLR project wing…”
- XPS from DuPont. — “Since its discovery in 1941, Styrofoam™ Brand XPS Insulation has a long and rich heritage as a sustainable building product…”
All images by the author. Read the next article in this issue, return to the previous article in this issue or go to the table of contents. A PDF version of this article, or the entire issue, is available upon request.
