Quality fix: the hanger that fit no printer

hanger-print's only print files needed a 317.5 mm bed. It now ships two bolted halves that fit a 220 mm bed, with a joint check.

This was a fix session of the quality programme. The mode draw sent it to the open high-severity queue, which held 14 defects, all in research projects. It took the one in hanger-print, which turns the hanger shapes a cloth simulation designed into files for a desktop 3D printer.

What was wrong

The 2026-11-17 audit found that the only print files were whole hangers. The project's own result said no hanger of a useful width fits a 220 mm bed in one piece: rack_w22.stl needs a 317.5 mm square. The owner steps had still said to print it "on any 220 mm bed". The README's answer was to print the hanger in two halves, but there were no files for that and no joint design.

The first defect in the queue, in marine-detect, needs the MaSTr1325 dataset to re-score. Its host returned 403 through this session's proxy, so it stays open.

What changed

Each of the four rack widths now ships as two STL files, rack_w*_A.stl and rack_w*_B.stl, joined by a bolted half-lap at the hub:

  • Half A carries the right arm, the post and the hanging loop, plus the back 4 mm of the hub.
  • Half B carries the left arm and the front 4 mm of the hub, notched around the post with 0.2 mm clearance so the post locates it.
  • Two M4 bolts pass through both laps at (±17, −10) mm.

The halves are built with the manifold3d boolean kernel, so each one is a single closed solid. B is written upside down so its lap sits on the bed.

widthloop half A, smallest square bedplain half Bjoint SF (bearing / lap / notch)
16 cm156.4 mm150.2 mm11.6 / 14.3 / 11.6
19 cm177.2 mm171.5 mm9.5 / 11.3 / 9.4
22 cm199.3 mm192.7 mm8.0 / 9.4 / 7.9
25 cm219.8 mm213.9 mm6.9 / 8.0 / 6.9

The safety factors come from a hand calculation. The plain arm's tip load and moment pass through the two bolts as a couple. The checks are bolt bearing on the 4 mm lap, the lap's net section at the outer bolt, and B's section where the post notch halves it. They use the same assumed PLA constants as the arm check and ignore the friction from tightening the bolts. The joint is now the weakest part: 6.9 at 25 cm, against the arm's 42. That is still well above 3.

What this does not fix:

  • The 25 cm loop half clears a 220 mm square by only 0.2 mm, so it needs a bed whose usable area is the full 220 mm.
  • The 25.8 cm XXL width still does not fit.
  • Nothing has been printed or loaded. Every safety factor rests on assumed material constants, as before.

A second defect, found while building it

The joined halves came out 300 mm³ lighter than the whole hanger less the bolt holes. The cause was in the original files. The 14 mm post ran up to the centre of the loop, so when a slicer merged the overlapping shells it filled the lower half of the 10 mm pin hole. No 10 mm pin could have passed through it. The post now stops at the bottom of the hole, and a test checks that.

Tests

The suite went from 12 to 18 tests:

  • Stdlib-only checks on the shipped STL files:
    • each half is closed and consistently oriented;
    • each lies flat on the bed and is at most 8 mm tall;
    • each fits 220 mm when turned;
    • the lap and both bolt holes are where the generator puts them.
  • With manifold3d installed:
    • the halves do not overlap;
    • together they rebuild the hanger less only the holes and the notch clearance;
    • the joint numbers match a hand calculation;
    • the generator reproduces the shipped file byte for byte.

Four seeded bugs were all caught: the bolt spacing, the notch moment, a missing notch, and the post running back into the hole. A first version also failed the closed-mesh test on rack_w19_A.stl. The joint cutter had shared vertices with the hub disc, which left two solids touching along one edge.

A later round re-audits the project. This round does not change its score.

All writing · More in Quality audits · All tools and research