Reverse Engineering: The Part Exists, the Supplier Doesn't

The most common reason someone contacts us about reverse engineering isn't curiosity about the competition – usually it's about parts that are needed but for which no manufacturer exists anymore.

Without a dataset, you're stuck. You can't change anything, can't optimise anything, and can't ask a second supplier for a quote. You can't compare an offer either, because you have nothing you could send. The part belongs to you, the design doesn't.

Usually you don't need a scanner

Most mechanical engineering parts can be described with a handful of design operations: extrude, cut, drill, chamfer, fillet. The entire geometry sits in a handful of numbers. Whoever measures these numbers has the part.

A scan, by contrast, delivers millions of points, from which someone then has to extract that same handful of numbers. For a milled part, that's a detour.

Free-form surfaces are the exception – housings, flow channels, formed sheet metal. There, there's no way around the scan, and in that case we buy it in. For everything else, measuring is the right approach, because it gets closer to the original design than a scan does.

Measuring is the easy part

Anyone can measure a component. The work starts afterwards.

What's on the table is a single, manufactured specimen. It carries dimensions that come from the function, and dimensions that come from manufacturing. A wall thickness of 3.2 mm might be a deliberate design choice or simply what the tool produced. A radius might be a relief notch or just the milling cutter that happened to be in the machine. Add wear, add tolerances, and sometimes a repair from twelve years ago.

Anyone who copies everything one to one builds the errors into the design as well. That's why reverse engineering isn't about measuring a part, but about deciding what each dimension actually means.

How we proceed

First, we disassemble wherever there's something to disassemble. For an assembly, only separating the parts reveals how they work together: what carries load, what guides, what seals, what merely holds. Material, surface finish and joining technique are recorded at the same time.

Then we measure and classify. Functional dimensions get a tolerance suited to the task, not to the specimen found. Manufacturing artefacts are discarded.

The result is a parametric CAD model – a design in which the dimensions are linked: change one value, and the rest follows. At the end stand the drawings and the bill of materials. A bare 3D model could already be manufactured, but only the drawing defines which tolerances and which material secure the function – and only with it can the part be reliably put out to tender and offers compared.

What this isn't

Replicating is rarely the purpose. More often it's about a spare part for which no drawing exists anymore. About the basis for further development. Or about understanding a competitor's product before comparing it to your own.

Such projects are sensitive, even when it's only about your own part: third-party protective rights may still apply, and whatever is on the table remains confidential. We work under NDA.

What drives the effort

Two factors.

The geometry. A milled part with flat surfaces and bores can be captured in a few hours. A formed sheet metal part or an organically grown housing cannot – there, the bought-in scan comes into play.

And how much you know about the part. If you know the function, the material, the critical points, half the decisions have already been made. If the part arrives without any context, we have to make them ourselves – conservatively if in doubt, which affects manufacturing costs.

We don't quote fixed prices for this. They arise from the specific part, and any figure given beforehand would be a guess.

What you end up with

You don't just get a spare part, you get a dataset that allows you to change, adapt or improve the part. And it becomes easier to find a manufacturer for it.

The plant runs again. And if the same part breaks again in two years, it's no longer an emergency.