UAV Demonstrator

Goal and context

I wanted to start a UAV team at my university to compete in the AIAA or IMechE flying competitions. Before pitching that, a friend and I wanted to prove to ourselves we could actually take an aircraft from blank canvas to flying. I worked through relevant chapters of Sadraey's Aircraft Design: A Systems Engineering Approach and ran the project in a way that aligns with my university’s systems engineering focus: requirements, conceptual design, then analysis, with focus on fast decision making and functionality over perfection.

I owned all of the aerodynamic, structural, and mechanical design and manufacturing, and specced the motor and electronics. My friend (a computer science student) selected the flight controller and radio system, configured ArduPilot, and handled some of the soldering.

Requirements

We set high-level requirements that best served our goals of a functional good-looking aircraft to prove capability to our school and to be a marketing tool in recruitment. The UAV shall:

Take off and fly under its own power

Look sleek - the aircraft is also a potential recruitment tool so it should be polished visually

Be quick to build - as a self-funded project in the middle of a semester, and with short timeframes to get a potential team set up for competitions

Fit in my car - the whole airframe had to fit in a small hatchback, an unfortunate constraint that capped the wingspan at 1300mm and required a disassemble-able tail

Landing wasn't a priority beyond getting clean photos of it flying - a scoping decision that eventually came back to bite me. From these requirements I derived specific technical metrics to drive the initial sizing and design.

Design and methods

I sized the aircraft with matching-line analysis to set the wing loading and thrust-to-weight from the requirements, then designed the wing with lifting-line theory, all with python scripts and approaches from the textbook. The 1200mm span limit pushed me toward a large (300mm) chord and low aspect ratio, with a high lift low Reynolds aerofoil (albeit not a Selig for manufacturing reasons), giving a low ~40 km/h stall speed that would be forgiving to fly. It's a single-motor taildragger with a V-tail; the V-tail was purely an aesthetic call for this prototype, in service of the "look sleek" requirement rather than for efficiency.

Motor and propeller testing to provide data for the matching line thrust was conducted through a rough thrust testing jig I put together – using timber, hinges and a luggage scale as a force sensor

Before designing the plywood structure, I validated the manufacturing method by skinning 3d printed test spars with clear plastic film – to test spar spacing distance and wing skin droop. I decided to switch to fiberglass as the preferred skinning material due to its stiffness to reduce concave droops between spars, as well as composites being the preferred design material for future competitive drones.

I sized the laser-cut plywood wing spars using hand calculations against manoeuvring-load accelerations characteristic of UAS competition regulations (+-3g flight manoeuvring, 5g landing), using the lowest stress grade of plywood in AS1720 as a conservative baseline (at some point I got curious how they engineered timber structures, so I read into AS1720). For the tail I modelled it in XFLR5, simply because hand calculations weren't enough to capture the stability response and the downwash from the low aspect ratio wing.

To be clear, I chose to stick with hand calculations and rough numerical modelling rather than finite element methods for stress and aerodynamic analysis. It is firstly unnecessary for the goal of getting a functional drone into the air over a short design period; and secondly, I know I am already capable of the FEA and CFD processes for this design, from my time in Solar Car – any more analysis would just be for optimisation, which is beyond the scope of this prototype.

Manufacturing

I intentionally designed a mixed set of manufacturing methods, as the aircraft is a capability showcase, and fully 3D-printing it would have demonstrated far less. They were:

Wing: laser-cut plywood spars and ribs, bonded, then skinned with a fiberglass sheet

Nose and tail cones: fibreglass moulded over 3D-printed, wax-released moulds

Wingtips and tail surfaces: 3D-printed, mostly single-perimeter skins at very low infill, with extra perimeters and infill added locally with slicer modifiers only where the part carries load

In hindsight, even though I was familiar with all these methods, developing and testing with all of them was tedious and slowed me down - the wet fiberglass layups especially. I would 3D-print more of the components next time, or go full in with a composite structure. Knowing which process is worth the effort is part of what the build taught me.

Flight test

This first prototype was always intended to be a fail-fast fail-first project, and the flight test went about as expected. The aircraft took off under its own power, rotating at roughly the speed my lift calculations predicted (we had no airspeed sensor, so this is estimated against the ~40 km/h stall), which was evidence that my wing sizing and lifting-line work were sound.

Two faults then showed up, also mine. First, a hard landing shortly after its first high speed taxi and liftoff snapped the 3mm plywood base on which the landing gear was mounted. Even though the landing gear strut was analysed to 5g loads and designed with ductile materials to fail safely, the mounting plate was not reinforced to the same robustness. As such, we switched to hand-launches afterwards.

The deeper problem was the static instability. A pre-flight balance check put the CG at about 30% of the chord, essentially on the neutral point for a static margin near zero (possibly slightly negative, depending on the NACA 6412's aerodynamic centre and its high pitching moment), and that was after I had already poured 0.5kg of mortar as ballast into the nose. As inexperienced pilots we couldn't correct the resulting pitch-up tendency, and neither did we have a gyroscopic closed loop control system working at the time of the first flight. Both hand-launched flights pitched up, stalled immediately, and came down, the last one hard enough to retire the airframe.

Failure analysis and what's next

The root cause can be traced back to an engineering sequencing mistake: I committed the fibreglass body geometry too early, before fully locking weight and balance, which pulled the CG too far aft to recover even with the 500g of ballast I ended up adding. To reach a sane ~15% static margin, the CG needs to move about 50mm forward. The blessing here is that the body is the only thing that is irrevocably broken – the very part that needs redesign, so the wing, tail, and electronics can be taken off and reused in a new body. I will be redesigning the body with proper weight and balance calculations beforehand, to ensure the next prototype does not encounter the same issue.

This development loop – move quick, fail fast, is what I believe a student team should be fundamentally and is the culture I want to set for a potential university UAV team.