Case Study: a Broken Machine Gear — a New Part in 3 Days via 3D Scanning and Printing

On Monday at 9:40, Andrii — a maintenance engineer for a production line — walked into our shop on Bulvarno-Kudriavska street and put a bag with three fragments on the table. It used to be the film-feed drive gear of a packaging machine. The machine is German, mid-2000s; the manufacturer no longer exists, there is no documentation, and nobody supplies an equivalent. The line had been down since Friday.

A textbook reverse-engineering case: the part exists (almost), there are no drawings, and there is even less time. Here is the process step by step, with numbers and one honest screw-up in the middle.

What the client brought — and what we asked

The first thing we do in cases like this is not switch the scanner on but ask questions. The answers drive both the material and the geometry:

• Where does the part sit and what does it transmit? Film-feed drive: low torque, about 120 rpm, cyclic duty.
• Why did it fail? The film jammed and the drive took a shock load. So the cause was not years of fatigue but a one-off impact. That matters: the design can be reproduced, not redesigned from scratch.
• What was the original made of? Judging by the fragments — cast polyamide. Good news: a plastic gear replaced with a plastic gear does not change how the assembly behaves.
• What does downtime cost? We will not quote Andrii’s answer here, but it made the rest of the conversation very short.

There were three fragments: two torn-out teeth and the gear body cracked through the hub — right along the keyway, the classic weak spot.

Day one: scanning and geometry recovery

We fixed the fragments together with cyanoacrylate on a jig — for scanning the part only has to hold its shape. A thin coat of matting spray (polyamide is semi-transparent to structured light), then about forty minutes on a stationary structured-light scanner accurate to 0.05 mm. The output is a point cloud, and a couple of hours later — a watertight polygon mesh.

Reverse engineering: rebuild the part, don’t trace the scan

The most common mistake in restoring broken parts is tracing the scan with splines. That works for housings; for gears it does not: the tooth profile has to be a mathematically exact involute, or the new gear will chew up its steel neighbour.

So we did it differently. From the scan we took the base parameters: outside diameter 96 mm, 46 teeth — hence module 2 (96 = 2 × 48, the classic m(z+2) formula); pressure angle the standard 20°. The bore we measured separately with a bore gauge: Ø20H7, 6 mm keyway. Then the gear was rebuilt in CAD as a parametric model — the involute profile generated, not traced. The scan served as the check: the rebuilt model deviates from the surviving teeth by no more than 0.04 mm.

One thing we changed deliberately. The crack ran through the corner of the keyway — where the original had a sharp corner and no fillet. The new model got 0.6 mm radii in the keyway corners and a hub diameter increased from 36 to 42 mm: there was room in the assembly, and the stress concentrator was gone. That is the advantage of reverse engineering over copying — you can not only reproduce a part but reinforce it exactly where it broke.

Why SLS and PA12, not FDM

For printed spare parts that genuinely work under load, the technology matters more than the material. An FDM nylon gear would not live long here: in FDM the interlayer adhesion is weaker than the plastic itself, and the tooth root is exactly the zone where layers get torn apart.

SLS is a different story: the laser sinters powder into a nearly isotropic monolith, with no supports and no “layers” in the mechanical sense. We printed in PA12 on a Formlabs Fuse 1+ 30W, 110 µm layers. PA12 suits this assembly well: wear-resistant, low friction against steel, good under cyclic load. The honest caveat we gave Andrii as well: if this were a power gearbox with serious torque, we would recommend a CNC-machined metal gear, not a print. For a film-feed unit, PA12 is exactly right.

Printing, the fit test, and an honest screw-up

Overnight from Monday to Tuesday the part went into the chamber along with other orders (SLS prints in batches — parts are nested in the build volume; the gear was tilted 15° for even sintering). Print plus controlled chamber cool-down — and on Tuesday afternoon the gear came out of bead blasting: matte, grey, uniform surface.

Fit test on the shaft — and here is the screw-up of the day: the keyway came out ~0.08 mm tight. Powder shrinkage in that zone was slightly higher than the compensation we had allowed. Ten minutes with a needle file would have fixed it, but we do not do that on working parts: a “press fit after filing” is an unpredictable fit. A tolerance edit in CAD (+0.1 mm on the keyway), and the same night we printed version two — three pieces at once: one for the shaft and two spares for Andrii’s shelf.

Wednesday morning — the fit test: the key seats snugly, no play, no force. Mounted on the machine, test cycle, line up and running. From a bag of fragments to a working line — three working days, rework included.

The economics of the case

• Hobbing a single custom gear: the quotes Andrii found started at roughly UAH 12–15k with a 2–3 week queue — a one-off gear-cutting operation is expensive by definition.
• An original from the manufacturer: there is no manufacturer. Hunting catalogues for an analogue is a lottery with no guarantee on the fit dimensions.
• Our full cycle (scanning + reverse engineering + two print iterations + three finished gears): about UAH 8k and 3 days.

And the main point that does not fit into a table: the CAD model stays with the client. The next replacement is not a “case” anymore — it is just a print from a ready file: a day of time and the cost of the part itself.

If you have a broken part lying around too

We do this work all the time: prototypes and end-use parts printed in Kyiv — the full cycle from scanning and reverse engineering to printing and post-processing, from a single piece. Got a ready 3D model? Upload it to the calculator — a quote within 30 minutes during working hours. Got only the part (even in fragments) — call us: +38 073 661 68 66. The question “can this even be restored” costs one phone call.

Broken part and need a replacement? Order 3D printing in Kyiv — a quote in 30 minutes, a finished part from 24 hours.