Can you reverse engineer a worn part?
Yes, but the worn surface is evidence, not automatically the original design. A replacement has to be reconstructed from function, mating parts, wear pattern, surviving geometry, standard fits and any available historical data.
Why is copying the measured size risky?
A shaft, bore, tooth, seal surface or guide can lose material in service. If the measured worn size is copied into CAD, the new part may reproduce the failure state rather than the intended fit.
What evidence can recover the nominal geometry?
Mating components, unworn reference surfaces, symmetry, standard bearing or seal sizes, fits, assembly constraints, old manuals, photographs and failure history can constrain the original design. The confidence of each reconstructed feature should be explicit.
What else must be recovered?
Material, heat treatment, surface finish, hardness, critical tolerances, load path and inspection criteria may matter as much as geometry.
When should the part be redesigned instead?
If the wear pattern exposes a systematic weakness, reproducing the original may only reproduce the same failure. That is the point to compare like-for-like replacement with a controlled redesign.