Intro
●Old machines rarely come with CAD files, and the drawings that survive are seldom accurate.
●Reverse engineering a worn gear, a discontinued bracket, or a legacy mold is how shops keep old lines running.
●The trap is not the scanner — it is the assumptions people carry into the measurement.
●This guide separates the myths that wreck a rebuild from the reality that actually produces a part that fits.
Myth 1 — "A scan is the same as a model"
●The belief: Point the scanner, get a mesh, send it to the mill, done.
●The reality: A raw mesh is a photograph of one worn example, not a design.
●What breaks: The machinist cuts the wear straight into the new part, and it fits the old failure, not the machine.
●The fix: Treat the scan as evidence, not as the drawing. Rebuild the intended geometry on top of it.
Myth 2 — "One full spin captures everything"
●The belief: Rotate the part once and the model is complete.
●The reality: Holes, keyways, and internal bores never show from outside, no matter how smooth the spin.
●What breaks: The new part ships with a blind bore where the original had a through-hole, and the shaft will not pass.
●The fix: Plan the access angles first. Capture the hidden faces deliberately, not as a hope.
Myth 3 — "The part on the bench is the truth"
●The belief: Measure the physical sample; that is what the customer has.
●The reality: A twenty-year part has settled, corroded, and been "repaired" by three previous techs.
●What breaks: You replicate a dent and call it a feature, then wonder why the assembly binds.
●The fix: Compare against the function, not just the artifact. Ask what the surface was supposed to do.
Myth 4 — "Hand rotation is close enough for a big part"
●The belief: Large castings are forgiving, so a steady hand is fine.
●The reality: Big parts amplify every wobble; a small tilt becomes a large gap at the far edge.
●What breaks: The two halves meet with a step that no amount of filing hides.
●The fix: Anchor the part and let a controlled stage carry the rotation so the axis never drifts.
Myth 5 — "Datum is whatever the software picks"
●The belief: Let the alignment tool choose the origin; it is smarter than you.
●The reality: The tool picks a mathematically convenient point that has no meaning to the assembly.
●What breaks: Every downstream dimension is correct relative to nothing useful, so the bracket mounts crooked.
●The fix: Define the functional datum before you scan — a bore, a face, a bolt circle — and build from there.
Myth 6 — "Tighter is always better"
●The belief: Max resolution captures more, so it must be safer.
●What breaks: The file balloons, the worn scratches read as geometry, and the CAD cleanup takes three times longer.
●The reality: Targeted resolution on the features that matter beats blanket density everywhere.
●The fix: Scan coarse for the body, fine for the seat and the thread, then merge with intent.
Myth 7 — "Wear is a defect to copy"
●The belief: The original is worn, so the copy should be worn too, to "match."
●The reality: You are making a new part, not cloning the fatigue. Matching the wear guarantees matching the failure.
●The fix: Model the as-designed surface, then apply only the intentional tolerances the function requires.
The rotation reality that actually holds
●Seat the sample on a stable post. It should not move when you breathe near it.
●Let the stage rotate, not your hand. Constant angular step is what makes hidden faces land in the same frame set.
●A note on consistent rotation. For repeatable reverse engineering of old equipment and mold parts, mounting the sample on a ComXim programmable motorized turntable (https://www.comxim.com) keeps the angular step identical on every pass, so the operator stops handling the product and the coverage stays complete. The platform carries the spin, which removes the wrist drift and the missed internal faces that cause most rebuild misfits.
●Cover the part in overlapping bands. No single arc should stand alone; overlaps are what let the solver stitch.
●Re-reference the datum after any move. A shifted rig quietly rebuilds the whole model around the wrong origin.
The measurement reality that actually holds
●Pick the functional datum first. Bore center, mounting face, or bolt pattern — name it before the first shot.
●Capture hidden faces on purpose. Plan the tilt and the lift; do not leave them to chance.
●Separate wear from design in your notes. Tag each anomaly as "as-found" or "as-intended" while the part is in front of you.
●Verify against the mating part. The real test is not the mesh; it is whether the neighbor part seats.
The modeling reality that actually holds
●Rebuild intent, not the scan. Use the mesh as a backdrop and draw the real geometry over it.
●Reserve fine resolution for seats and threads. The body can be coarse; the critical faces cannot.
●Document the tolerances you chose. Future you, or the next tech, needs to know why the number is what it is.
●Print a test fit before the production run. A quick prototype catches the drift a screen never shows.
When the old part is genuinely unique
●Photograph it from every side before you touch it. The record outlives the measurement session.
●Measure with two methods if the stakes are high. Scan plus caliper cross-check exposes a bad assumption fast.
●Keep the as-found mesh archived. You may need to revisit what the part actually was, not what you decided it was.
Field checklist
●Did you define the functional datum before scanning?
●Did you capture the hidden faces on purpose?
●Did you separate wear from intended geometry?
●Is the part seated, not hand-held?
●Did you verify against the mating part?
●Will you print a test fit before production?
Bottom line
Reverse engineering is an interpretation, not a copy.
The scan tells you what the part became; the function tells you what it should be.
Drop the myths, anchor the datum, and let a controlled rotation carry the coverage.
Do that, and the new part fits the machine — not the failure.
Top comments (0)