CNC Machining Transfer Guide

• 1
  Key takeaways
• 2
  What a CNC machining transfer actually changes
• 3
  Datum strategy: the part of the transfer that fails first
• 4
  Tolerance stack: adding the numbers before you commit
• 5
  Feature access and machine travel
• 6
  Material behavior after a transfer
• 7
  First article inspection closes the loop
• 8
  Which transfer path fits which part

A transfer is any move where the cutting operations for a part change hands. That can mean moving a job from a 3-axis shop to a 5-axis shop, moving from a vertical mill to a mill-turn center, or shifting a part from a casting or die-casting process onto machined stock. The geometry on the print does not change. What changes is the chain of surfaces, fixtures, and machine errors that position the tool against that geometry.

Every machining process holds a part against something. A vise jaw, a chuck, a fixture plate, a soft jaw, a vacuum table. That contact surface becomes the datum in practice, even when the drawing calls a different datum. When the part moves, the new shop builds a new fixture from the same drawing. If the drawing's datum scheme does not match how the part was held before, the new fixture references a different surface. Small deviations on that reference surface get pushed into the features that matter.

The second change is the error budget. On the original machine, the operator may have dialed in an offset to correct a known drift. A new setup has no such history. Positioning accuracy, thermal growth, tool deflection, and fixture repeatability all reset. A transfer is not a copy of the old process. It is a new process that has to meet the same print.

The third change is access. A feature that was reachable on a 3-axis machine with a long reach tool may be blocked on a mill-turn center, or the opposite. Tool length and diameter set the limit. So does the angle between the feature axis and the spindle. This is why two shops can both quote a part and only one can hold it.

Most transfer failures start at datum selection, not at the machine. If the print uses datum A as a large flat face and datum B as a hole, the shop needs a way to seat the face and clock the hole in one setup. If that face was a casting surface before and is a machined surface now, the flatness changed. The new fixture still seats it, but the seating is different.

Ask one question before awarding the job: which surfaces will be cut in the same setup as the datum features? Features cut in the same setup share the same zero point. Features cut in a second setup inherit the re-fixturing error. On a transfer, the number of setups often goes down, which is good, but the first setup has to establish more datums.

Here is a practical test. Take the three features with the tightest position tolerance and trace them back to the datum callout. If two of them are in different setups and the print gives them a shared datum, the shop will need a fixture that can locate both. That fixture may not exist yet, and it may cost more than the machining.

For parts that were previously cast or forged and are now machined from solid, the datum picture changes completely. A casting has draft and parting lines. Machined stock has square edges and known flatness. The transfer often improves repeatability, but the first operation has to remove enough material to clean up the uneven surface without moving the datums.

A useful rule: cut the datums first, then cut everything else from them. If the datums are cut last, the part has already moved twice.

A print tolerance is the total allowed variation. The machine's positioning accuracy is only one term in that total. The others are fixture repeatability, thermal drift, tool wear, and measurement uncertainty. On a transfer, all of them are new.

Take a position tolerance of 0.02 mm on a bolt circle. A machining center that holds ±0.005 mm positioning leaves 0.015 mm for everything else. A fixture with 0.01 mm of repeatability eats two thirds of that. Add thermal drift over a 4-hour run and the budget is gone. This is arithmetic, not opinion, and it is why a shop that holds ±0.005 mm on one part may not hold the same callout on a transferred one.

Surface finish follows the same logic. Ra 0.8–1.6 μm is a normal machined finish with a sharp tool and a rigid setup. Transfer the part to a longer tool or a thinner wall and the finish degrades through chatter, not through a change in cutting parameters. The fix is usually support and a shorter tool, not a slower feed.

Measure the stack before quoting, not after. If the numbers do not add up, the honest answer is that the print needs a change, or the fixture needs a redesign.

Machine travel is the easy limit to check and the least common reason a transfer fails. A 4,000 mm maximum processing size covers large frames, rails, and housings. Most transferred parts are far smaller. The real limit is whether the tool can reach the feature at the right angle.

Consider a part with a side hole whose axis is perpendicular to the main bore. On a 3-axis machine, that hole needs a second setup, which adds a re-fixturing error and a second datum. On a 5-axis machine with a Ø400 mm rotary table, the table tilts and the hole is cut without re-fixturing. The transfer from 3-axis to 5-axis usually improves position tolerance because it removes a setup.

The trade is stiffness. A 5-axis setup with the part held off the table at an angle loses rigidity compared to a flat 3-axis setup. Deep pockets and thin walls may chatter where they did not before. That is the boundary: 5-axis wins on access, loses on rigidity for heavy cuts.

Long parts are a separate case. A 750 × 1,150 × 550 mm envelope suits large plates and frames, but the overhang at the ends deflects under cutting load. Support them or accept a looser flatness callout.

Check access with the actual tool, not a nominal diameter. A Ø6 mm end mill needs a holder that is often wider than the cutter. The holder is what hits the wall.

Aluminium 6061 and 7075 are the common transfer materials. 6061 machines cleanly and holds a fine finish. 7075 gives higher strength but is more prone to distortion when a lot of material is removed from one side. If the original process left more stock, the transfer may remove less and the distortion picture changes.

Stainless 304 and 316 work-harden. A transfer that changes the depth of cut or the tool path can push the surface into a harder layer, and the next pass cuts the hardened skin. 17-4PH in the H900 condition is different again, and heat treat after machining moves the part. Plan the sequence so that critical tolerances are cut after any thermal operation.

Titanium TC4 (Ti-6Al-4V) and Inconel are low thermal conductivity alloys. Heat stays in the cutting zone, the tool wears faster, and the same program that ran on steel will not run the same way. On a transfer, tool life and thermal growth have to be re-measured, not assumed.

Plastics behave differently. POM and PEEK move with temperature and clamp pressure. A fixture that held a steel part firmly will deform these materials. Light clamping and sharp tooling matter more than machine accuracy.

A transfer is not proven until a first article is measured against the print. Pick the features with the tightest tolerances and the features that depend on the new datum scheme. Measure them on a coordinate measuring machine, not with a caliper. Record the values, not just pass or fail.

Compare the first article to the previous supplier's data if it exists. A shift in one axis across several features usually points to the fixture, not the machine. A random spread across features points to thermal or tool issues. Both are fixable, and both are cheaper to fix on one part than on a batch.

Reports can be supplied on request. Our inspection covers raw material check, in-process monitoring, and final inspection, with 100% inspection before shipment. That is the standard flow, and it applies to transferred work the same way it applies to new work.

If the first article is out of tolerance, change one variable at a time. Move the fixture, not the program and the fixture together. Otherwise you will not know what fixed it, and the next batch may drift back.