How to Judge the Accuracy of a CNC Machining Center

• 1
  Key takeaways
• 2
  What accuracy of a CNC machining center really means
• 3
  Check axis geometry before you cut anything
• 4
  Spindle runout and thermal growth
• 5
  Ballscrew backlash and servo tuning
• 6
  Thermal drift over a full shift
• 7
  Test part and CMM report close the case
• 8
  How to judge the accuracy of a CNC machining center: step by step
• 9
  Accuracy check values and what they tell you

Machine builders quote positioning accuracy, repeatability, and sometimes volumetric accuracy. Those numbers come from a laser interferometer run on a cold machine in a controlled room. In a real shop, the machine sits on a floor that moves, a spindle that heats, and a fixture that clamps. So the accuracy you care about is the accuracy of the finished part, not the accuracy of the brochure.

For most work we see, the practical target is ±0.005 mm on critical features and Ra 0.8–1.6 μm on sealing faces. That is achievable on a well-maintained three-axis or five-axis machine, but only when geometry, spindle condition, and thermal behavior are all in range. One bad axis ruins the whole part.

The order matters. Check geometry before you check anything else. A machine with a bent axis will show good repeatability on a short move and still cut a tapered bore. Repeatability tells you the machine can return to a point; accuracy tells you the point is in the right place.

• 1
  Repeatability vs accuracyRepeatability is scatter; accuracy is offset from nominal.
• 2
  Static vs dynamicA cold check misses spindle growth and servo lag under load.
• 3
  Part accuracyThe only number your customer sees is the CMM report.

Start with a granite square and a dial indicator on a magnetic base. Sweep the square along X and Y to measure squareness between the two axes. On a typical vertical machining center, we want squareness within 0.010 mm over 300 mm. Anything larger will show up as a taper or a skewed bolt pattern.

Next, check parallelism between the table surface and the X-Y plane. Set the indicator on the table and sweep it across the full travel. A reading above 0.015 mm over 500 mm means the machine needs leveling or way adjustment before it can hold tight tolerances.

Then check Z-axis squareness to the table. Mount the indicator on the spindle and sweep a precision square standing on the table. This is the check that catches a machine that was crashed and re-leveled without re-scraping.

If any of these fail, stop. No amount of tool compensation will fix bad geometry. Fix the machine first, then re-run the checks.

• 1
  Squareness X-YTarget within 0.010 mm over 300 mm.
• 2
  Table parallelismTarget within 0.015 mm over 500 mm.
• 3
  Z to tableSweep a precision square; look for consistent reading.

Spindle runout is the most common reason a bore comes out oval or oversized. Put a 0.002 mm indicator on a clean test bar and rotate the spindle by hand. TIR should stay under 0.005 mm at the gauge line. Above that, check the taper for chips or fretting before you blame the bearings.

Thermal growth is harder to see. A spindle running at 10,000 rpm can grow 20–40 μm in Z within the first hour. On a batch of 50 parts, that shift alone can push the last parts out of tolerance. Run the spindle for 30 minutes, then re-zero your tool offsets.

Watch the spindle load too. If a face mill pulls more than 70% of rated load on a light cut, the bearings are worn or the tool is unbalanced. Either one will show up as chatter and poor surface finish.

• 1
  TIR limitUnder 0.005 mm at the gauge line.
• 2
  Warm-up30 minutes before first cut on tight-tolerance work.
• 3
  LoadKeep face milling under 70% of rated spindle load.

Backlash hides at direction changes. Cut a 200 mm slot in one direction, then cut the same slot in the opposite direction and measure the difference. More than 0.010 mm of mismatch means the thrust bearings or the nut need attention.

Servo tuning matters on contouring work. A machine with soft tuning will overshoot corners and leave a witness mark. Cut a 50 mm square with a 10 mm end mill at 0.2 mm radial depth and check the corners under a toolmaker's microscope. A visible step over 0.010 mm is a tuning or mechanical problem.

On a five-axis machine, add a check on the rotary table. Indicate the Ø400 mm table at four points. Runout above 0.010 mm will show up as a tilted feature on any part that gets indexed.

• 1
  Backlash testBidirectional slot; mismatch under 0.010 mm.
• 2
  Corner test50 mm square, 0.2 mm radial depth, step under 0.010 mm.
• 3
  Rotary tableØ400 mm table runout under 0.010 mm.

A machine that is accurate at 8 a.m. may not be accurate at 4 p.m. The room temperature, the coolant, and the spindle all add heat. Measure a reference feature every two hours on a single part held in the fixture, and log the numbers.

A drift under 0.010 mm over eight hours is acceptable for most work. Beyond that, look at the coolant chiller, the shop HVAC, and whether the machine is sitting in direct sunlight. We have seen a 0.030 mm drift on a machine near a loading door in summer.

For tight-tolerance runs, we keep the shop at 20 ±2 °C and run a warm-up cycle before the first part. That is not a luxury; it is the difference between a stable process and a chasing game.

• 1
  Log every 2 hoursSame part, same feature, same operator.
• 2
  Acceptable driftUnder 0.010 mm over an eight-hour shift.
• 3
  Shop control20 ±2 °C, no direct sun, no open doors near the machine.

Geometry, spindle, and thermal checks tell you the machine is healthy. The test part tells you what it can actually produce. Cut a part with the same material, tool, and fixture you will use in production. Do not use a soft aluminum coupon if the real part is 17-4PH stainless.

Measure the part on a CMM and compare to the drawing. Look at the features that matter: bore diameters, true position, flatness, and surface finish. A machine that holds ±0.005 mm on a test bar may only hold ±0.015 mm on a real part with a long tool and a thin wall.

Keep the test part and the report. When a customer asks how you know the machine is accurate, a dated CMM report on a real part is the only answer that holds up.

• 1
  Same setupMaterial, tool, fixture, and coolant as production.
• 2
  CMM reportOne report per machine per quarter, dated.
• 3
  Keep the partPhysical evidence beats a spec sheet in any audit.