Cheap de-airing clay extruder auger

What's in the download

If you only want to print it, grab the STL. If you want to change it — different pitch, fatter shaft, bigger motor bore — open the F3D.

The original file is broken, and that's the real problem

Before any of the parametric stuff, there's a plainer issue. Load the original auger.stl in a slicer and look at the warnings:

That isn't a solid with a few flaws — it's 459 loose fragments, only four of which are actually closed, and the largest closed piece holds 5.44 cm³ of a part that should be about 7.1.

Slicers don't refuse a file like this. They quietly improvise a repair across thousands of holes, and they improvise differently depending on slicer, version and settings. That is the most likely reason the tip region has always been unpredictable — the slicer is inventing geometry there, and inventing it differently each time.

You can check this yourself in about thirty seconds: drop both STLs into PrusaSlicer and compare the warning icons.

Why rebuild instead of repairing the mesh?

You can run a mesh repair tool and get something watertight. But you'd still have a shell of triangles — no "pitch", no "shaft diameter", just a few hundred thousand flat facets pretending to be a curve. Every STL-to-CAD converter I know of gives you back a dumb solid: watertight, editable-ish, with no dimensions you can type a new number into.

I wanted the real thing: change flight_pitch from 8.86 to 10 mm, press enter, and have the screw rebuild itself.

The approach

1. Measure the mesh in Python and fit the silhouette to real geometry rather than eyeballing it. It turned out to be almost entirely lines and arcs: the S-shaped squeeze where the shaft narrows is two tangent arcs (R 97.3 and R 102.5 mm), the long middle is one very shallow arc (R 3108 mm), and the paddle edge is a circle of R 19.975 mm centred exactly on the origin. Every fit is better than 0.07 mm — a strong hint those are what the original designer actually drew.

2. Rebuild it natively in Fusion as sketches and features driven by named parameters. All 8 sketches are fully constrained, with no "Fix" constraints — the CAD equivalent of taping something in place.

3. Check it against the original mesh, in both directions. Final: 0.036 mm RMS, with 95% of the surface within 0.072 mm. Roughly a fifth of a typical layer line.

The bits that fought back

Fusion's API cannot create Coil features. The object appears in the documentation but creation was never wired up. So the flight is a profile swept along a mathematically exact helix, with a second helix used as a guide rail to stop the blade slowly rotating out of alignment across 17 turns.

A custom thread profile can't substitute either. Fusion's thread definition only controls diameters and a flank angle — blade width is structurally tied to pitch, so a 0.45 mm blade on an 8.86 mm pitch simply isn't expressible.

The blade is a wedge, not a constant thickness — about 2.4 mm where it meets the shaft, tapering to a near knife-edge at the rim.

The tip was the interesting part. It isn't sharp: it ends in a flat about 1.31 mm across, the cone starts at 160.6 mm, and the flight fattens as it blends into the point. I got the tip wrong twice before measuring it properly.

Measuring turned out to be harder than modelling. I measured a fin as 0.72 mm thick when it's really 0.498 — a 44% error from using a percentile spread instead of the median of each face — and it survived into the model until an independent review caught it. Two rounds of adversarial review found eight defects after I'd called the job finished. That loop is why the final figure is 0.036 mm and not 0.101.

What you can change

Overall length · flight outer diameter · flight pitch · blade thickness at the shaft and at the edge · shaft diameters along the length · paddle count, diameter and thickness · motor-shaft bore diameter and depth.

Each parameter carries a plain-English note in the F3D, so you don't have to reverse-engineer my reverse-engineering.

Printed and tested

Printed in PLA and run in the extruder with real clay. Nothing notable went wrong.

The one thing I can't claim yet is long-term tip behaviour. Fixing the original's tip problem was my whole motivation, and I can now prove the mesh side of it — there are no holes left for a slicer to guess at. Whether the tip geometry itself performs better over many hours, I haven't run enough to say.

Honest limits

This rebuild method works on machined-looking geometry — turned and prismatic parts made of planes, cylinders and arcs. It does not work on sculpted or organic models, or on noisy 3D scans. Those genuinely cannot be reduced to a handful of clean dimensions, and any tool claiming otherwise is guessing.

This is one part rebuilt so far. Treat it as a promising result rather than a proven pipeline, and if you print one I'd genuinely like to hear how it goes.