Accuracy of CNC Machine Tools: What Engineers Should Actually Measure

• 1
  Understanding accuracy of CNC machine tools beyond the spec sheet
• 2
  Where the error actually comes from in the accuracy of CNC machine tools
• 3
  How machine class changes the accuracy of CNC machine tools
• 4
  How to verify the accuracy of CNC machine tools before you place an order
• 5
  When the accuracy of CNC machine tools stops being the limiting factor
• 6
  Matching machine class to part tolerance

Two numbers get quoted when people discuss the accuracy of CNC machine tools. One is positioning accuracy: how close the tool center point arrives to the commanded coordinate. The other is repeatability: how tightly the machine returns to the same point when it moves there again. A machine can be repeatable without being accurate, but it can rarely be accurate without being repeatable.

On a typical vertical machining center, positioning accuracy sits around ±0.005 mm under ISO 230-2 test conditions. Repeatability is usually tighter, often half that value. That gap matters. If your bore spacing tolerance is ±0.02 mm, positioning error is a small share of the budget. If you are chasing ±0.005 mm on a 300 mm bolt circle, every source of error counts, including spindle thermal growth.

The spec sheet is measured in a temperature-controlled room with a light cut or no cut at all. Your production part is cut after the spindle has run for hours, with coolant flowing and chips evacuating. Those conditions change the number. A machine rated at ±0.005 mm may hold ±0.010 mm on a hot afternoon if the shop has no thermal management plan.

So the real question is not "how accurate is this machine?" It is "how much error does this process add, and can my drawing absorb it?" That reframing is what separates a useful supplier conversation from a spec-sheet argument.

• 1
  Positioning accuracyHow close the axis arrives to the commanded point.
• 2
  RepeatabilityHow consistently it returns to the same point.
• 3
  Thermal driftError that grows as the spindle and ballscrews warm up.
• 4
  Process capabilityThe combined error from machine, tool, fixture and material.

Machine geometry is the first source. Squareness between X, Y and Z, straightness of each axis, and the pitch error of the ballscrew all push the tool off the commanded path. A machine that has been crashed or has worn way surfaces will show this as a growing circularity error on a test cut. Ballbar testing catches it in minutes.

Thermal growth is the second and usually the largest source. A spindle running at 12,000 rpm for two hours can grow 20 to 40 μm in Z. Ballscrews warm unevenly along their length, so the error changes with axis position. Shops that hold tight tolerances warm up the machine before the first cut, keep the shop at a stable temperature, and sometimes use glass scales to close the loop on the actual slide position.

Tool deflection and wear are the third. A 6 mm end mill at 3× diameter depth of cut will deflect under cutting force. The deflection shows up as a taper in the wall or a size drift over the run. Tool wear adds a slow, monotonic drift. An operator who measures the first part and the twentieth part will see it.

Fixture and material effects close the list. A part held on three points will move when the clamps release. A thin wall will spring back after the cut. Residual stress in a 7075 block will distort it after every material removal pass. None of these are machine errors, but they all appear in the final measured dimension.

• 1
  GeometrySquareness, straightness, ballscrew pitch error.
• 2
  ThermalSpindle and ballscrew growth during the shift.
• 3
  ToolDeflection under load and progressive wear.
• 4
  WorkholdingClamp release and residual stress movement.

Not every part needs a five-axis center with glass scales. A three-axis mill with a good operator holds ±0.02 mm all day on a bracket. A five-axis simultaneous machine with thermal compensation holds ±0.005 mm on a complex aerospace fitting. The cost difference is real, and matching the machine to the part is the whole game.

Three-axis machines suit prismatic parts with features reachable from one or two setups. They are the workhorses for brackets, plates, housings and covers. Four-axis machines add a rotary table for parts with features on multiple faces, like a manifold or a gearbox cover. Five-axis machines handle contoured surfaces, undercut features and parts that would otherwise need five separate fixtures.

The accuracy of CNC machine tools in the five-axis class depends heavily on the rotary axes. A tilting head or trunnion table adds two more error sources: rotary positioning and the offset between the rotary centerline and the spindle. Shops that do five-axis work measure these with a ballbar and a test sphere, not with a dial indicator on a flat face.

Mill-turn centers blur the line. They turn and mill in one setup, which removes the re-fixturing error that plagues parts with both turned and milled features. For a shaft with cross-holes and milled flats, a mill-turn center often beats a two-machine process on true position, even if the individual machines are less accurate.

Ask for a capability study, not a spec sheet. A Cpk study on a real feature from a real setup tells you more than an ISO 230-2 certificate. If the shop cannot produce one, ask them to cut a test part from your drawing and measure it on a CMM. That is the only number that matters for your order.

Ask which machine will run your part. A shop with one five-axis center and twenty three-axis machines may quote your five-axis job on the right machine, or it may quote it on a three-axis machine with a tilting fixture. The quote should name the machine class. If it does not, ask.

Ask about thermal management. Does the shop run a warm-up cycle? Is the inspection room temperature-controlled? Are the CMM and the machine in the same building? A part measured in a 20 °C inspection room after being cut in a 30 °C shop will show a size change that is not a machine error at all.

Ask about in-process measurement. A shop that measures only the first part and the last part will miss drift in the middle of the run. A shop that checks every tenth part, or uses an in-machine probe, catches drift before it becomes scrap. That habit, more than the machine brand, decides whether your tolerance holds.

• 1
  Capability studyCpk on a real feature, not a spec sheet.
• 2
  Named machineThe quote should state the machine class.
• 3
  Thermal planWarm-up cycle and controlled inspection room.
• 4
  In-process checksProbing or sampling through the run.

Below about ±0.005 mm, the machine is rarely the problem. The part itself becomes the limit. A 0.8 mm wall on an aluminum housing will move when you unclamp it, no matter how accurate the machine is. A 300 mm long shaft will bend under its own weight during turning. A titanium impeller will distort from residual stress after each pass.

Surface finish has a similar ceiling. A fine Ra 0.2–0.8 μm finish needs a sharp tool, a rigid setup and a stable spindle. If the tool is dull or the part is ringing, the machine's positioning accuracy will not save the finish. The best shops match the tool and the strategy to the finish, not just the machine.

Feature size matters too. A Ø2 mm hole in a deep pocket is harder to hold than a Ø20 mm bore on an open face, even on the same machine. The tool is more slender, the coolant is harder to reach, and chip evacuation is worse. Tolerance on the drawing should reflect that difference.

For most production work, the practical floor for a well-run shop is ±0.005 mm on a small part with a rigid setup and a controlled environment. Push much below that and you are paying for inspection time and scrap risk, not for a better machine. The engineering question is whether the extra cost buys a function.