Quality Assurance of CNC Services: What to Check Before You Order

• 1
  Key takeaways
• 2
  What Each Quality Claim Actually Tells You
• 3
  Why Quality Assurance of CNC Services Starts at DFM
• 4
  Incoming Material and Setup Control
• 5
  In-Process Monitoring and SPC
• 6
  Inspection Equipment and Report Content
• 7
  Finishing, Marking and Final Release
• 8
  Five Steps to Qualify a CNC Supplier

Most dimensional failures are designed in before a chip is cut. A wall too thin to hold without chatter, a deep pocket that needs a long reach tool, a datum that only exists on the drawing and not on the part. When we review a model, the first pass is not a price. It is a list of features that will be hard to hold.

A useful DFM review names the problem and the fix. For example, a 0.8 mm wall in 6061 will deflect under finishing passes; moving it to 1.5 mm lets us hold ±0.005 mm without a support fixture. A Ø6 mm hole 60 mm deep needs a pilot drill and a reamer, and the straightness will still be looser than a shorter hole.

Ask the supplier what they changed and why. If the answer is only about cost, they are quoting, not reviewing. Good DFM notes also flag which tolerances drive the price. Tightening one bore from ±0.05 mm to ±0.005 mm may add a separate finishing operation on a second machine.

The output of this step should be a revised model or a marked-up drawing, plus a short list of dimensions that need a decision. That list becomes the basis of the inspection plan later, so nothing gets measured that nobody specified.

Quality assurance of CNC services fails quietly when material is wrong. A 6061-T6 bar and a 6061-T4 bar machine differently, and the second one will move after machining. We check mill certificates against the drawing before the job is released, and we mark the heat number on the traveler so a finished part can be traced back to a lot.

For stainless, the difference between 304 and 316L matters for corrosion and for magnetic response. For titanium, TC4 (Ti-6Al-4V) and commercially pure TA2 behave very differently at the same feed. A supplier who treats all aluminum as one line item on a quote is telling you something about their process.

Setup control is where datum strategy gets locked in. On a 3-axis machine, a part with features on five faces needs multiple repositions, and each one adds stack-up. On a simultaneous 5-axis center, the same part can often be finished in two setups, with the rotary table holding position between operations.

That is the practical difference between the 27 three-axis machines and the 16 simultaneous 5-axis centers on our floor. Neither is better in the abstract. The part decides.

Checking parts only at the end is expensive. If the first 40 parts drift out of tolerance, you scrap 40 parts and lose the day. In-process monitoring catches drift while there is still material to cut. On a long run we measure a defined feature every set number of parts, plot the values, and adjust offsets before the trend crosses the limit.

What to ask for depends on volume. A one-off prototype needs a first article inspection and a final check. A 10,000-part run needs a control plan with sample size, frequency and reaction plan. The reaction plan matters most: who stops the machine, and what happens to parts made since the last good check.

Temperature is an underrated variable. Aluminum expands about 23 × 10⁻⁶ per °C. A 100 mm aluminum bore measured at 30 °C will read roughly 0.012 mm larger than the same bore at 20 °C. For ±0.005 mm work, either the shop controls temperature or the report states the measurement temperature.

We keep 100% inspection before shipment as the last gate. It is the backstop, not the plan.

A report with numbers but no method is not useful. A CMM report should list the datum reference frame, the software and probe tip used, and the nominal, actual and deviation for each dimension. It should also state the measurement uncertainty, or at least the equipment accuracy, so you can judge whether the ±0.005 mm callout is even measurable.

Hand tools have limits. A caliper reads to 0.02 mm on a good day, and a micrometer to 0.001 mm on a clean, round, room-temperature feature. Neither can verify position or profile. If your drawing has true position callouts, you need CMM or optical data, not a caliper sheet.

Surface finish is often specified but rarely reported. Ra 0.8–1.6 μm is a typical machined finish; Ra 0.2–0.8 μm needs a finer pass and sometimes a different tool path. Ra 1.6–3.2 μm is acceptable for many brackets but not for a sealing face.

Ask for a sample report before you place the order. It takes two minutes to read and it tells you whether the shop understands your drawing.

Finishing is functional, not cosmetic. Anodizing adds a few micrometres of build-up on every surface, which can close a tight bore or change a thread fit. Hardcoat anodizing builds more than clear anodizing. If a masked area is required for electrical contact, that must be on the drawing before the parts go to the line.

Heat treatment moves parts. A stress-relief step before final finishing is often cheaper than scrapping a batch after distortion. For thin parts, we plan the sequence so the last operation removes the least material.

Laser marking has a practical floor. Minimum character height is 1.5 mm with our setup; smaller text becomes hard to read and may not survive bead blasting. If traceability requires a data matrix code, its size and contrast need to be agreed before marking, not after.

Final release is a signed check against the purchase order: quantity, revision, finish, marking and report package. When the release step is written down, missing paperwork stops being a surprise.