What can be recreated from a sample

• Recovers reliably: overall geometry, interfaces, thread forms, mounting features, anything that has not worn.
• Recovers with judgement: worn diameters and faces, original surface finish, and the tolerances the designer intended — all reconstructed from the unworn geometry and from what the part has to do.
• Does not recover from measurement: material grade, heat treatment, case depth and surface treatment. These need testing, or an assumption based on the application that is stated as such.

The instinct is that a damaged part is a poor reference. In practice, wear is confined to the working surfaces, and the rest of the part is untouched. Mounting faces, threads, shoulders and overall proportions usually survive intact, and those are what establish the part's relationship to everything around it.

Wear patterns carry their own information. An unevenly worn bore says something about alignment; a fretted mounting face says something about clamping. That is worth capturing before the part is cleaned up.

The risk in reverse engineering is rarely geometric. It is that a part is reproduced faithfully in the wrong material condition, and fails early in a way that looks like a manufacturing defect.

This is why material should be treated as an explicit decision rather than a detail. Where the original specification is unknown, the options are to test the sample, or to select a material and treatment appropriate to the duty and record that choice on the drawing so it is visible.

If a part failed early, copying it exactly reproduces the failure. It is worth asking whether the design contributed — a sharp corner where a radius belonged, a material that was marginal for the duty, a feature that concentrated load.

Improvements should be deliberate and documented. An unannounced change is indistinguishable from an error when the next person compares the new part to the old one.

• Assess the sample: what is worn, what is intact, what is missing entirely
• Measure the surviving geometry and reconstruct the worn features from it
• Establish material and treatment, by test or by stated assumption
• Issue a drawing for your approval, with assumptions marked as assumptions
• Manufacture a first-off and, where practical, trial it in the machine
• Release the approved drawing, which then belongs to you for future orders

A reverse engineering exercise should come with a statement of confidence, feature by feature where it matters. Some dimensions will be certain; others will be a considered reconstruction.

You should also expect to be told when a part is a poor candidate — because too little survives, because the material is critical and untested, or because the failure suggests the original design was the problem.