UAV SILM logoUAV SILM
Student project
Lessons

What we learnt, and what went wrong

The analysis caught several problems before anything was printed, while they were still cheap to fix. These are the lessons we are taking into the build, and the troubles that taught them.

Weigh the airframe from the CAD

Our first balance sheet used placeholder masses for the fuselage, wings and V-tail. Taking volumes from the 3D model gave a 574 g airframe with its own CG at 435 mm, and put the whole aircraft's CG at 411 mm: behind the neutral point, so it would have been unstable.

Moving the wing beats adding weight

Moving the wing 20 mm aft shifted the neutral point 18.2 mm but the CG only about 5 mm. It did the job of about 80 g of nose weight, for no extra mass.

Balance and trim speed move together

Moving the CG aft lowers the stability margin and the trim speed at the same time: 21.2 m/s with the CG at 335 mm, 18.8 m/s at 345 mm. The tail incidence sets the rest.

Check the fit in cross-section

In side view everything fitted. Slicing the round fuselage showed three clashes, and that the battery can only sit on the centreline.

Each tool sees a different aircraft

XFLR5 models the wing and tail with partly laminar flow; OpenFOAM adds the fuselage and assumes turbulent flow everywhere. Lift agreed within 5%, while drag at 3° came out about 60% higher in CFD. Knowing what each tool includes tells us which number to use for what.

Heat is a design load for printed parts

LW-PLA softens at around 55 to 60 °C. The motor sits inside the tail cone and the ESC near the floor, so we are adding vents and printing trays and loaded parts in PETG.

Some choices are mechanical, not about balance

Putting the tail servos at x 520 or x 740 changes the CG by only 2.4 mm, so we will choose between them on pushrod stiffness, access and heat.

Our first CG estimate would have made the aircraft unstable

The CG came out at 411 mm, 47 mm behind the neutral point.

What we didRebuilt the mass model from CAD volumes, moved the electronics forward, planned a lighter V-tail and moved the wing 20 mm aft. The static margin is now about 20%.

Three components did not fit

The Raspberry Pi, the camera and the motor all clashed with the round fuselage or the narrowing tail cone.

What we didPi to the upper tray, camera behind a nose window, motor 19 mm further forward on its long shaft.

The CAD would not mesh directly for CFD

Some curved surfaces in the STEP export failed in the first meshing tool.

What we didFused the parts and triangulated them another way before snappyHexMesh could use the geometry.

Parallel CFD runs failed at first

Running the solver on two cores clashed with the installed MPI library.

What we didA library workaround, after which each case ran in parallel to convergence.

CFD drag disagreed with XFLR5

0.0367 against 0.0227 at 3°.

What we didTraced it to the fuselage and the fully turbulent boundary layer in CFD. We treat the CFD drag as an upper estimate until flight tests.

Our first failsafe settings assumed a radio we do not have

They were written for a pilot with an ExpressLRS radio who would land by hand. Without one, a radio failsafe would trigger at once, the aircraft would refuse to arm, and a return home would end in circling until the battery ran out. Several parameter names did not exist in ArduPlane at all.

What we didWrote a script that checks every parameter against the ArduPlane source, rebuilt the settings for hand launch and automatic landing with no radio, made every failsafe end in that landing, and wrote our own laptop mission planner that always adds the landing.

The camera speed estimate had its sign reversed

A test against an independent camera model showed forward flight reading as flying backwards. The estimate was switched off by default, so it never reached the autopilot.

What we didFixed the signs, added correction for pitch and roll, and now test it on rendered images of the ground.

The build started late

Airframe printing has not started; the print settings for the foaming LW-PLA still need tuning.

What we didPlanned test prints first (LW-PLA from 5 November, PETG from 13 November) and replaced the stage plan with dated tasks. See the progress report.

The second parts order slipped

Carbon tube sizes, wire gauge and the arming connector had to come from the CAD first.

What we didFinished the sizes and set the order for 4 November.
  • Build the mass model from the CAD on day one, and update it whenever a part changes.
  • Fit-check components in cross-section before choosing their positions.
  • Order the secondary parts together with the primary set.
  • Run a mesh-independence study before trusting drag values.
  • Leave room for cooling in every closed compartment.
Progress

Progress report

Interim report II covers 24 September to 2 November 2026: what we have done, what went well, what has been difficult, what comes next and how we are tracking against the schedule. We have added the work since then, up to 17 November.

8 / 8
primary components received
V11b
SILM Monedula airframe in FreeCAD, 69 print-ready parts, CFD-checked
14 Nov
first motor thrust test
35 days
to the presentation
28 SepFuselage modelled with the tail boom and carbon rod.CAD
13 OctWing V4 edge-cut test for splitting the wing into printable sections.CAD
19 OctFirst thrust rig parts printed and assembled on the base plate.Test rig
20 OctAll eight primary components ordered and received: flight controller, ESC, GPS and compass, airspeed sensor, 4G module, camera, battery and motor.Procurement
20 OctSecondary component list started: servos, inserts, filament, carbon tubes, connectors.Procurement
by 29 OctAirframe CFD run in OpenFOAM on cleaned-up CAD, adding to the XFLR5 results.Analysis
by 29 OctServo selection closed (SG90, 2 to 4× torque margin); tail and ruddervators modelled and sectioned.Design
by 29 OctMotor mount for the thrust test finished; the rig design went through sketches up to version 5.Test rig
31 Oct to 2 NovWiring diagram for the flight controller, sensors, radio and servos updated using the H743 pinout reference.Electronics
1 NovHollow tail with connectors, fuselage with nose mounts and the full-aircraft assembly V4.CAD
2 NovThrust rig documented; day-by-day plan to the presentation agreed and assigned.Planning
3 NovOpenFOAM CFD of the full V4 aircraft at 0°, 3° and 6°: lift within 5% of XFLR5, neutral point at 363.6 mm.Analysis
3 NovCentre of gravity redone from the 3D model: the old sheet layout put the CG behind the neutral point.Analysis
3 NovWing moved 20 mm aft in CAD and checked in OpenFOAM: neutral point from 363.6 to 381.8 mm, lift and drag unchanged.CAD, analysis
3 NovInternal layout fit-checked against fuselage cross-sections and three clashes fixed. CG 342.7 mm, static margin about 20%.Design
3 NovWebsite rebuilt with interactive 3D models of the airframe, the internal layout and the CFD results.Website
3 NovTwo top hatches modelled, for the battery and for the Raspberry Pi and flight controller. Each has a tongue under the skin at one end and two 6×3 mm magnets at the other, with clearances set for foamed LW-PLA.CAD
9 NovNose cone printed in PolyLite and weighed: 29.6 g, against 43 g predicted by the slicer. The masses of all printed parts are now calibrated on it.Build
9 NovV9 CAD: overlapping joints between the fuselage sections. The forward section's spigot slides 10 to 12 mm into the next one with 0.2 mm clearance, and stays clear of the hatches and components.CAD
9 NovV9 CAD: lightweight battery tray with a strap channel, cooling vents for the ESC and motor, and SG90 aileron servo pockets with screw-on caps held by M2 heat-set inserts.CAD
9 NovMass and balance redone for V9: about 1.70 kg all-up, CG 341.7 mm, static margin about 21%. The onboard computer is now a Raspberry Pi 4.Analysis
10 NovFull SolidWorks assembly built from the V9 parts, with the second wing mirrored, ready for fit checks.CAD
15 NovWebsite reorganised into separate pages. The current airframe is now called SILM Buteo, and SILM Monedula is the next airframe.Website
15 NovSILM Monedula started as a parametric model in FreeCAD, so a change rebuilds and rechecks the whole airframe. 1.5 m wing, 79 × 81 mm fuselage.CAD
16 NovFirst OpenFOAM run on Monedula. The first mesh used the print files (gaps, seams) and gave L/D 11.6; we discarded it and meshed a clean outer surface with boundary-layer cells instead: L/D 18.Analysis
16 NovThe same CFD showed the layout was unstable in pitch: static margin −3.4 % instead of the +14 % our design tool predicted. The tool ignores the wing's downwash on the tail; a vortex-lattice check agreed with the CFD.Analysis
17 NovMonedula V11: fuselage lengthened by 68 mm and the V-tail enlarged. A second CFD run confirmed a 13.7 % static margin, chosen over 18 % to limit trim drag.Design, analysis
17 NovV11 connectors: PETG couplers between all six fuselage sections, Flightory-style PETG wing root ribs and sockets, PETG spar carrier, TPU flex hinges, PETG horns and servo frames, screw-on TPU sleds. The central payload bay was removed; payloads go on the nose.CAD
17 NovAll 69 parts checked for clashes and fit; STEP files labelled with their print material. Assembly guide and a full materials list written: about €870 in parts and 1.79 kg modelled.Documentation
18 NovMonedula V11b: smoother fuselage, stronger tail, motor inside the tail; interactive 3D model on the site.CAD
✓

CFD caught a real problem before printing

Running OpenFOAM on Monedula showed it would have been unstable in pitch. We found the cause, changed the design and confirmed the fix with a second run, all before printing anything.

✓

The design is now fully parametric

Monedula is a parametric FreeCAD model. Lengthening the fuselage moved the wing, bay, tail and every joint together, and the clash checks ran again automatically on all 69 parts.

✓

Problems were found on paper

The balance and fit checks found an unstable CG and three component clashes before any part was printed, while they were still cheap to fix.

✓

Main components arrived on schedule

We planned procurement for 6 to 19 October and had all eight primary components in hand on 20 October, so the electronics work is no longer waiting on deliveries.

✓

The aerodynamic design is complete

XFLR5 and OpenFOAM agree on lift and stability, and with the wing moved aft the aircraft has a static margin of about 20%. Aerodynamics is no longer holding up the build.

✓

CAD reached a full-aircraft assembly

Version 4 brings the fuselage, wing, hollow tail with connectors and nose mounts together in one model, which we now use for sectioning, part design and the balance calculation.

✓

We built our own thrust rig

Rather than relying on manufacturer data, we designed and built a stand that measures static thrust for any motor and propeller combination. It will let us check the propulsion choice before flight.

✓

Clear split of work

A shared task sheet with priorities and owners, and a dated plan to 8 December, means each of us knows what is due and when.

!

Our design tools disagreed

The quick aerodynamic tool we used to size Monedula gave a 14 % static margin; CFD gave −3.4 %. We now check stability with a method that includes downwash and confirm it in CFD.

!

Joints and hinges are still unproven

The PETG couplers, root ribs and TPU hinges have only been checked in CAD. Bond strength and hinge life need printed test pieces before we trust them in flight.

!

The build phase started late

Airframe printing has not started yet. LW-PLA is a foaming filament whose weight and strength depend strongly on temperature and flow, so we need test prints to dial it in (5 to 10 November), followed by PETG settings (13 to 16 November).

!

The second component order slipped

The secondary list has been open since 20 October, and several items still needed sizing from the CAD: carbon tube sizes, wire gauge and the battery arming connector. We will order on 4 November.

!

Joining printed sections

The wing and fuselage are printed in sections joined with overlapping edges, carbon-filament rivets, bolts and PETG root gaskets. Hole tolerances have to be proven with test pieces before we commit to full parts.

!

One structural decision is still open

We have not yet decided whether the 12 mm main spar joins at mid-fuselage or runs the full span. It affects transport, assembly and wing removal.

!

Software work is behind the hardware

Flight-controller programming and datalink research have not started. The internal layout is now fit-checked, but the trays and mounts still need to be modelled. These now sit on the critical path for ground testing.

!

The thrust rig took longer than expected

We sketched it up to version 5 before building, and it is still being finalised. It has not yet been run with a battery.

3 NovTop hatch modellingRomet
3 NovBody sectioning and shellingLudvig
3 NovElectronics wiring schematicKarl-Robin
3 NovWebsite update and interim reportRomet
4 NovCamera mountLudvig
4 NovBody connectors modellingRomet
4 NovSecondary component list finalised and orderedRomet
5 NovWing main carbon rod and ruddervator rod holes, connectors and researchLudvig
5 to 10 NovLW-PLA print settings dialled inRomet
10 NovMotor adapter: PETG slab to LW-PLA fuselageRomet
13 NovBattery holderRomet
13 NovServo compartmentsLudvig
13 NovWing connectors: overlapping sections, filament rivets, bolt holes, PETG root gasketLudvig
13 to 16 NovPETG print settings dialled inRomet
14 NovMotor thrust test, if the rig is readyRomet
18 NovFlight controller and Raspberry Pi mountLudvig
24 NovMotor thrust test (latest date)Romet
30 NovFinished electronics circuitKarl-Robin
8 DecPresentation dayTeam

From our working plan of 2 November. Printing and assembly of the airframe sections follow as soon as each part model and the print settings are ready. The battery holder and the flight controller mount will follow the tray plan in Inside the airframe.

StagePlannedWhat happenedStatusRevised
Design and analysisto 5 OctAerodynamics finished on time; balance redone and wing moved on 3 Nov. Detailed part CAD (mounts, connectors, compartments) continues alongside the build.On trackDetail CAD to 18 Nov
Procurement6 to 19 OctPrimary set received 20 Oct. Secondary order delayed while sizes were finalised.Partly lateSecondary order 4 Nov
Build20 Oct to 9 NovThrust rig built. Airframe printing not started; print settings still to tune.About 2 weeks behind5 to 30 Nov
Testing10 to 27 NovNot started.Moved14 Nov to 7 Dec
Presentation28 Nov to 8 DecNo change.On track28 Nov to 8 Dec
We kept to the plan for the aerodynamic design and the main component order. The build is about two weeks behind, mainly because the secondary parts list and print settings took longer than expected. We have replaced the stage plan with a dated task list for each of us, moved the thrust test to 14 November (24 November at the latest), and made the first flight a stretch goal: it happens before 8 December only if the airframe and electronics are complete in time, otherwise right after the presentation. The revised dates are shown in the timeline below.

Our first interim report (23 September) recorded where the project stood before this period.

Done by 23 Sep
Hover or tap for detail
  • CFD analysis in XFLR5 (VLM1 and VLM2); MH32 airfoil selected over RG15 and SD7037.
  • Full-aircraft geometry set: 1.40 m span, 0.263 m² wing, aspect ratio 7.46, high wing and V-tail.
  • Mass and CG modelling: centre of gravity 0.335 m, static margin 18.5% at 1.58 to 1.60 kg.
  • Aileron and ruddervator response evaluated.
  • Electronics selected and sourcing links collected; cost target under $500.
Planned next, 23 Sep
Hover or tap for detail
  • Finalise V-tail geometry and construction details. Done
  • Order and receive components. Primary set done
  • 3D-print the airframe and wing. Moved to November
  • Assemble electronics and test the link. Circuit due 30 Nov
  • First ground and flight tests. Thrust test 14 Nov
Timeline

Timeline, revised 2 November

Five stages working back from the presentation on 8 December 2026. Struck-through dates are the original plan from September.

01
Doneto 5 Octaerodynamics done; detail CAD to 18 Nov

Design and analysis

Hover or tap for detail
  • Airfoil selection and XFLR5 analysis of the full aircraft
  • OpenFOAM airframe CFD, then a second run with the wing moved 20 mm aft (3 Nov)
  • CG from the 3D model and internal layout fit check (3 Nov)
  • Full-aircraft CAD assembly V4; part models continue
  • SILM Monedula V11: parametric FreeCAD airframe, CFD-checked 14 % static margin, 69 print-ready parts (17 Nov)
  • SILM Monedula V11b and its 3D model (18 Nov)
02
Partly done6 to 19 Octprimary 20 Oct; secondary 4 Nov

Procurement

Hover or tap for detail
  • Primary electronics received 20 Oct
  • Secondary order (servos, inserts, filament, carbon tubes, connectors) on 4 Nov
03
In progress20 Oct to 9 Nov5 to 30 Nov

Build

Hover or tap for detail
  • LW-PLA settings 5 to 10 Nov, PETG settings 13 to 16 Nov
  • Printing the fuselage and wing sections; joining with rivets, bolts and gaskets
  • PETG trays for the battery and electronics
  • Electronics circuit finished by 30 Nov
  • 4G and WireGuard link added after the core system works
04
Ahead10 to 27 Nov14 Nov to 7 Dec

Testing

Hover or tap for detail
  • Static thrust test 14 Nov (24 Nov at the latest)
  • Weighing the printed parts and checking the CG
  • Ground tests of control surfaces, link and failsafe after 30 Nov
  • First hand-launch as a stretch goal before 8 Dec
  • Endurance and range validation after the presentation
05
On track28 Nov to 8 Dec

Presentation and documentation

Hover or tap for detail
  • Results summary and final website update
  • Presentation day: 8 December 2026

UAV SILM is an ongoing student STEAM project, not a finished aircraft. Every figure on this page (mass, endurance, cost, balance, control response) is a theoretical estimate from CFD analysis and hand calculation, not from flight-test data. The 3D models show our CAD and simulation results, not a built aircraft.

We have not held formal discussions with Kaitseliit or the Estonian Defence Forces; we present the platform as a design study and a longer-term aim toward a working, versatile aircraft.