Ducted frame generator for solidworks / build123d

the build

I made a 13 inch "ducted" quad that I designed and printed myself: four 380 mm rings, 4006 380 kV motors on 6S, three carbon arms per hub, half lap joints at mid height with two M3 each, and a pair of 22.3 mm carbon tubes running the length of the fuselage as well as carbon fiber wrap/epoxy finish. Its not much of a "ducted" quad as much as it is a guard for the props, but it still does better for the thrust than an open air frame.

flight was okay but it was seriously lacking on thrust. A ducted rotor only pays for itself if the inlet stays attached, and static thrust guidance for a hover duct puts the lip radius at 5 to 10 percent of prop diameter. Mine was 0.7 percent. The chord was 40 mm, 12 percent of diameter against a usual 30 to 50. The prop disc sat 4 mm below the lip, inside the lips own radius, so the blade tips were working in the separated region instead of downstream of a reattached one.

Three ratios, all wrong in the same direction, all compounding. this is definetely something I should have gone over before I printed the frame, and it is the thing ductgen now does automatically.

dimensions

Before anything could be generated it had to be measured, so the reading i made initially then the whole airframe was exported as a single mesh and I had to take it apart in software.

A STL reader gives bounding box, surface area and signed volume. A component splitter welds vertices onto a tolerance grid then runs union find over shared vertices, which turned one 157,000 triangle file into 19 separate labelled bodies. A plane slicer reports radius and angle clusters about any axis.

this way i was able to produce the references needed for the base generation: throat 340.0, OD 380.0, chord 40.0, motor spacing 475.45, lip radius 2.3, a 30.5 mm flight controller pattern. Each one is recorded with how it was made, because a few numbers could not be measured at all. The prop plane depth is one of those.

geometry

The inputs reach the geometry through the aerodynamics, and argues back when the numbers do not add up.

prop diameter sets the duct ID at whatever tip clearance, throat, chord, lip radius and prop plane depth scaled. kV x cells gives rpm gives tip Mach, and it tells you when you cross 0.6. Diffuser angle gives expansion ratio, which = the ideal static thrust gain against an open rotor. Section plus printer to mass, per ring and for four.

For example, a 6 percent bellmouth needs 36 mm of radial room before there is any structure left, so asking for it inside a 20 mm wall makes the section self intersect. Instead of letting the revolve fail, the duct OD grows to accommodate it and the report says the ring got wider. Every rule prints their status with the number and the target beside it.

Cutting a 417 mm ring onto a 256 mm bed

The ring has to become printable pieces, and where the cuts land matters structurally. By hand this is a lot of work. The ring is seperated into arcs and 2 middle connectors, and the arcs are phased to keep the joints off the strut roots.

the arc sector is tested at every in plane rotation, because a 256 mm bed will swallow a much longer part across the diagonal. Joints are then phased to keep them off the strut roots, so no glue line sits where the motor loads are. All of the segments carry a half lap tab so consecutive arcs overlap rather than butt together, and the connector between each duct pair is cut as the exact complement of that lap, which closes the ring by construction.

The carbon rods are the awkward part. A perimeter rod crosses the duct wall as a chord, not a radial hole, and on this frame that chord across about 36 degrees of ring, roughly a whole segment. So it almost always continues past a joint into the neighbour, and both parts need drilling. The first version treated the bore as a point, gave the hole to the nearest segment only, and left up to 60 mm of solid wall sitting exactly where the tube had to pass.

Holes that sit at angles no reference plane can reach are handled by rotating the solid body under a fixed plane, cutting, and rotating it back. solidworks really would fight to not build a plane through the vertical axis, so the part moves instead of the plane.

whats next

im going to continue working on my fiber optic uart transciever, see if i can make something worth using for long ranges.