Accuracy verification

CNC 5axiscnc Accuracy: How to Check It on Real Parts

Machined surfaces look smooth. That tells you almost nothing about CNC 5axiscnc accuracy. This guide shows engineers and buyers how to measure what actually matters, step by step, using tolerances and inspection tools you already have.

Get Instant Quote Talk to an Engineer

±0.005 mm tolerance100% inspection16 five-axis centersISO 9001:2015

Custom auto spare parts checked for CNC 5axiscnc accuracy

On this page

Key takeaways

Measure the part, not the machineA machine that cuts a perfect test ball can still scrap your part if the fixture or the stock moves.

±0.005 mm is a process windowWe hold it on critical features, not on every surface of a 4,000 mm part.

Heat moves metal firstA 20 °C shop floor swing can shift a 300 mm aluminum part by more than the tolerance.

Report the feature, not the claimAsk for a CMM report keyed to drawing callouts, with the datum scheme you specified.

Why it matters

What CNC 5axiscnc accuracy actually means on a drawing

Accuracy is not one number. On a five-axis part it splits into three separate things: the machine's positioning error, the error added by the setup and fixture, and the error added by cutting forces and heat. The drawing only sees the sum of all three. That is why a machine that holds ±0.005 mm on a test artifact can still produce a part that misses a true position callout.

Positioning error is the machine's own limit. Rotary axes add error in a way three-axis machines do not: any tilt of the trunnion or the C-axis table is multiplied by the distance from the center of rotation to the cutting edge. A 0.01 mm tilt at the pivot can become 0.03 mm or more at the tool tip if the part sits far from center. Keep the work close to the rotary center when the tolerance is tight.

Setup error is usually the largest single contributor, and it is the one you control. A vise that lifts a corner by 0.02 mm, a soft jaw that springs open after clamping, or chips under a locating face will all show up as a flatness or parallelism failure. On five-axis work the part is re-datumed between operations, so any inconsistency in the locating scheme compounds.

Thermal error is slow and easy to miss. Aluminum expands roughly 23 μm per meter per degree Celsius. A 300 mm aluminum part that warms 5 °C from cutting grows about 0.035 mm. On a ±0.005 mm callout that is already seven times the tolerance. In-process gauging and stable coolant temperature matter more than most shops admit.

1
Machine geometryRotary axis tilt and backlash, multiplied by part distance from center
2
Fixture and setupClamp-induced distortion, chip entrapment, re-datum between ops
3
Thermal driftSpindle growth, coolant warming, shop floor temperature swings

Feature types

Which features carry the real tolerance

Not every surface deserves the same attention. On most parts, three feature families decide whether the part passes: mating bores and their true position, flat sealing faces, and angular features machined in a single tilted setup. Inspect those first and inspect them properly.

Bores and holes are checked for size, roundness, and position. Size is easy and it is the least interesting number. Roundness and true position are where five-axis work separates from three-axis work, because a bore drilled in a tilted orientation inherits any rotary error. Measure roundness at three depths, not just at the entrance.

Flat faces are checked for flatness and parallelism to a datum. A sealing face that is flat but not parallel to the mounting datum will leak under torque. On large parts, up to 4,000 mm, flatness over the full length is often specified looser than local flatness for exactly this reason.

Angular features, undercut pockets, and blended radii are the reason the part is on a five-axis machine at all. They should be cut in one continuous setup. If your supplier splits them across two operations, expect a witness line and a position shift. Ask how the part is held for each orientation.

1
Mating boresCheck diameter, roundness at three depths, true position
2
Sealing facesFlatness plus parallelism to the mounting datum
3
Tilted featuresConfirm they are cut in one setup, not two

Inspection

How to read an inspection report without getting fooled

A CMM report is only as good as the datum scheme behind it. If the report uses a different datum order than the drawing, the numbers are not comparable and a good part can look bad. Check that the primary, secondary, and tertiary datums match the drawing callouts before you read a single value.

Look at the distribution, not the maximum. A feature that reads 0.004 mm on five parts and 0.009 mm on the sixth is a process that is drifting, even if the average passes. Ask for the individual readings, not just the pass or fail column.

Confirm the measurement temperature. Metrology labs run at 20 °C. A part measured straight off the machine at 30 °C is not the same part you will assemble tomorrow. For tight work, ask that the part normalize before final inspection.

Finally, match the inspection method to the tolerance. A caliper resolves about 0.02 mm at best. For a ±0.005 mm callout you need a CMM, a micrometer with a proper stand, or a bore gauge with a known setting ring. Hand tools are for rough checks, not for sign-off.

1
Datum schemeMust match the drawing order, or the report is meaningless
2
Spread over averageAsk for individual readings across the run
3
Temperature at measurement20 °C reference for tight tolerances

Method

Step by step: verifying CNC 5axiscnc accuracy on a delivered part

1
Let the part normalizeLeave it on a stable surface at 20 °C for at least 2 hours before measuring. Skip this and every number you take is suspect. A warm part measures large on bores and long on length.
2
Clean and deburr the datum facesWipe datum A, B, and C with lint-free cloth and check for burrs with a fine stone. A 0.01 mm burr under a datum face will tilt the whole part in the CMM and shift every result.
3
Set up the datum scheme exactly as drawnEstablish primary, secondary, tertiary in the drawing order. Do not let the CMM software pick a best-fit alignment unless the drawing says best-fit. Record which surfaces were touched.
4
Measure size first, then form, then positionStart with diameters and lengths using a micrometer or bore gauge. Then check roundness and flatness on the CMM. Then true position. This order isolates a bad feature fast.
5
Check roundness at three depthsMeasure each critical bore at the top, middle, and bottom. Taper or bell-mouth shows up here and is invisible on a single-depth check. Watch for more than 0.005 mm variation between depths.
6
Verify tilted and blended featuresScan the angular face and the blend radius in one continuous pass. Look for a step, a witness line, or a sudden change in surface finish. These indicate the feature was cut in two setups.
7
Compare against the drawing, not the sampleWrite each measured value next to its callout. Flag anything above 70% of the tolerance band. A part at 90% of tolerance today will be out of tolerance after plating or anodizing.
8
Document and close the loopRecord the ambient temperature, the instruments used, and their calibration dates. Send the data back to the supplier so the process can be corrected before the next run.

Selection guide

Which accuracy check to use, and when

Match the method to the feature and the tolerance band

Feature	Best method	Practical resolution	When it is not enough
Bore diameter	Micrometer or bore gauge	About 0.001 mm with a setting ring	You also need roundness and taper
Roundness	CMM or roundness tester	About 0.001 mm	Deep bores need a dedicated form tester
Flatness	CMM scan or surface plate	0.002 mm on a granite plate	Large parts need a laser or level
True position	CMM with the drawn datum scheme	About 0.002 mm	Wrong datum order invalidates the result
Surface finish	Portable profilometer	Ra 0.05 μm steps	Curved or hidden faces need a skidless probe
Angular features	CMM scan in one pass	About 0.003 mm	A witness line means two setups were used

The honest verdict

CNC 5axiscnc accuracy is a process result, not a machine spec. Fix the setup, control the temperature, and measure the features that matter. If a supplier cannot explain their datum scheme and setup plan, the tolerance on the drawing is a wish, not a promise.

Get Instant Quote Talk to an Engineer

FAQs

Questions engineers ask about CNC 5axiscnc accuracy

Can a five-axis machine hold ±0.005 mm on a part 500 mm long?

Yes, on specific features, with the right setup and temperature control. The limit is not the machine alone. It is the fixture, the thermal state, and how far the cutting edge sits from the rotary center.

On a 500 mm part we would expect tight tolerances on bores and mating faces close to the rotary center, and looser, realistic tolerances on features at the far end of the part. Put the tight callouts where the process can actually hold them.

Why does my part measure good at the shop but fail at assembly?

Most often it is temperature or datum mismatch. A part measured warm will shrink as it cools. A part measured against a best-fit alignment may not match the assembly fixture that uses the drawing datums.

Check both before blaming the machining. Ask for the measurement temperature and the exact datum scheme used on the report.

Does surface finish affect dimensional accuracy?

It does, indirectly. A coarse surface leaves peaks that a micrometer sits on, so the measured diameter reads slightly large. Anodizing and plating also build thickness, typically 5 to 25 μm per surface depending on the process.

For tight bores, specify the finish and the coating together, and ask the supplier to machine to the pre-plate size.

How do I check accuracy without a CMM?

You can cover a lot with hand tools if you accept their limits. Use a micrometer with a stand for diameters, a granite plate and dial indicator for flatness, and gauge pins for hole position relative to a machined edge.

Hand tools resolve roughly 0.01 to 0.02 mm. That is fine for general machining but not for a ±0.005 mm callout. For those features, ask for a CMM report.

What causes a witness line on a five-axis part?

A witness line means the surface was cut in two orientations, usually because the part was repositioned or the tool could not reach in one pass. The two cuts rarely blend perfectly, so a step appears.

If the drawing does not allow a visible step, the feature must be cut in one continuous setup. Ask your supplier to confirm the setup plan before cutting.

How much does thermal drift really move a part?

Aluminum expands about 23 μm per meter per degree Celsius, steel about 12 μm. A 300 mm aluminum part that warms 5 °C grows roughly 0.035 mm. That is several times a ±0.005 mm tolerance.

This is why coolant temperature, spindle warm-up, and letting the part normalize before inspection are not optional on tight work.

Send us the drawing and we will tell you what is holdable

Upload your part and tolerances. We return a quotation and a free DFM analysis within 12 hours, with a clear note on which callouts the process can hold and which need a different approach.

12-hour quote100% inspection±0.005 mm toleranceNDA on request

Get Instant Quote Talk to an Engineer