Banana Katana 3D Printing: What It Takes to Print One Clean

The shape is a single curved body roughly 700–1,000 mm long, with a 30–45 mm wide blade section and a grip that tapers down to about 25 mm. Most hobby printers cap out at 220–350 mm in Z, so the part has to be split into three or four segments. Each joint becomes a weak line where bending stress concentrates. A straight sword can sit flat on the plate and use gravity in its favor. A banana curve cannot, because the center of mass hangs off the build surface and the blade wants to lift at the ends.

Curvature also changes how the nozzle lays down plastic. On the outer face of the bend, the toolpath stretches; on the inner face, it compresses. Slicers handle this by adjusting extrusion width, but the effect shows up as thin walls on the inside of the curve and slight over-extrusion on the outside. On a 2 mm wall that is a 10–15% thickness variation, which is enough to start a crack under repeated handling.

There is a second issue: the tip. A katana-style point narrows to 1–2 mm over the last 40 mm. Printed vertically, that tip is a series of tiny layers with almost no cross-section, and it will snap off in a bag or a costume rig. Printed horizontally, the tip needs support that leaves a rough surface. Neither option is free.

• 1
  Split at low-stress pointsCut at the grip or a flat blade section, never mid-curve.
• 2
  Blunt the tipA 3 mm rounded tip prints clean and survives handling.
• 3
  Watch wall variationKeep walls at 2.4 mm or more on the inside of the bend.

FDM parts are weak between layers. Tensile strength along the extrusion direction is often 2–3× higher than across layer lines. On a long prop, the highest bending moment sits right where the blade meets the handle. Print the model standing up with the grip at the bottom and every bending load pulls directly on the layer bonds. That is the wrong choice.

Laying the part flat on its side puts layer lines perpendicular to the bending load, which is better. The trade-off is support material along the full underside and a visible seam on the curved face. On a painted prop that seam disappears under filler and primer. On an unpainted print it is obvious.

A third option works well for display pieces: print in two halves along the centerline, then glue. Each half is a shallow shell, so there is less support and less warp. The glue joint runs down the middle of the blade, which is a low-stress line in bending. For a 900 mm part this is often the cleanest route.

Warp starts at the corners. A 900 mm part has a lot of corner. Use a brim of 8–10 mm and raise the first-layer temperature by 5 °C over the rest of the print. Bed adhesion is not the goal here; it is holding the ends down while the middle cools.

Three perimeters at 0.4 mm gives a 1.2 mm shell, which is not enough on its own. Set the shell to four or five perimeters so the wall reaches 1.6–2 mm, then drop infill to 10–15% gyroid. The shell carries the bending load; the infill only stops the top surfaces from collapsing. More infill adds weight and print time without adding much stiffness.

Cooling needs to be moderate, not maximum. Blasting a long part with 100% fan cools the top layers fast, which increases the temperature difference between layers and drives warp. On PLA, 50–70% is usually enough. On ABS or ASA, run 20–30% and keep the chamber closed.

Print speed at the outer walls should drop to 30–40 mm/s on the curved sections. Faster than that and the extruder cannot keep the wall thickness consistent through the bend, which shows up later as a thin spot that cracks.

• 1
  Brim 8–10 mmHolds long ends down while the middle cools.
• 2
  Four to five perimetersReaches 1.6–2 mm shell for bending stiffness.
• 3
  Infill 10–15% gyroidEnough to support top surfaces, not more.
• 4
  Outer wall 30–40 mm/sKeeps wall thickness even through the curve.

Segment joints need more than glue. A printed tongue-and-groove or a short internal dowel adds shear area and keeps the two halves aligned while the adhesive cures. Cyanoacrylate works on PLA and PETG but is brittle. Two-part epoxy gives a little more flexibility and fills small gaps. For a part that will be swung around, add a 6 mm carbon or aluminum rod through the grip and into the first blade segment.

Filler goes on before primer. Sand the joint to 220 grit, apply a thin layer of epoxy filler, let it cure, then sand to 400. Two thin coats beat one thick coat every time. Thick filler shrinks and shrinks unevenly, which reintroduces the curve you just spent hours printing straight.

Primer and paint add 0.1–0.3 mm of thickness. That is fine on a prop but it matters on the grip if the model has tight finger grooves. Mask the grip before painting or leave 0.3 mm of clearance in the CAD model.

If the part is going to be handled at a convention for three days, a clear coat helps. Two light coats of matte clear keep the paint from rubbing off at the grip and the tip. Skip it and the paint wears through in a weekend.

FDM is the right answer for a prop, a costume piece, or a display model. It is the wrong answer when the part has to hold an edge, take a real impact, or fit into a mechanism with tight tolerances. A printed blade cannot cut anything, and it should not pretend to.

For metal versions, the geometry changes. A machined or metal-printed blade can be thinner because the material carries more load per unit of cross-section. The trade-off is cost and weight. A 900 mm aluminum blade is heavier than the printed one, which changes how the grip has to be designed.

Where a functional metal part is needed, we run 5-axis CNC machining to ±0.005 mm with surface finish down to Ra 0.2–0.8 μm. That is a different class of part from a printed prop. Both are useful. Mixing them up is how projects get expensive.

If the design has a real mechanical function, send the CAD file. We quote and return a DFM analysis within 12 hours, and production can start within 24 hours on approved drawings.