7 Essential CNC Machine Quality Control Tips to Avoid Costly Defects
This guide is for engineers and buyers who release machined parts to a supplier and then find out, at incoming inspection, that the run does not match the drawing. It covers the seven control points that decide whether a tight tolerance holds across a batch: probing, material, fixturing, cutting data, environment, calibration, and the supplier's quality system. Read it to judge where your own process is weakest.
Where scrap actually comes from
Most rejects are not caused by a worn tool. They are caused by a decision made hours before the first cut.
Measure during the cut, not after it
A CMM report written after the part leaves the machine tells you that you have scrap. It does not tell you when the error entered the process. In-process probing closes that gap: the machine checks a datum while the part is still clamped, so the correction happens on the next pass instead of on the next batch.
For a 5-axis aerospace bracket with a tight form tolerance, we probe the critical datums after each roughing pass. Roughing moves material in a way that is hard to predict, especially on thin walls, and the error accumulates if you only check at the end. Probing after roughing keeps the finishing allowance even across the part.
Not every job needs probing. On a simple 3-axis plate with a ±0.1 mm tolerance and one setup, a first-article check plus a final CMM run is usually cheaper and just as safe. Probing earns its place when the part has multiple setups, thin walls, or a datum that is hard to reach after finishing. The decision is about error accumulation, not about the price of the probe.
- 1Use probing whenMultiple setups, thin walls, tight datums, or a form tolerance you cannot recover later.
- 2Skip probing whenOne setup, open tolerances, and a stable fixture that repeats within a few microns.
- 3What to write downWhich datum is probed, at which operation, and what offset is applied if it drifts.
Verify the material before it reaches the spindle
Material variability is a quiet defect driver. Two bars of 7075-T6 from different heats can machine differently even when the certificate says the same grade. Hardness spread changes chip formation, and residual stress from the mill releases when you remove material, which shows up as bowing on a long part.
Incoming inspection is where you catch it. Every bar or plate gets a unique ID and stays traceable to the finished part. We run optical emission spectroscopy on samples before the material is released to the shop floor, which catches a wrong grade or a substitution before any spindle time is spent.
On titanium and stainless powder for SLM, traceability goes one level deeper: each batch's particle size distribution has to be documented, because a shift in distribution changes the melt pool and the density of the printed part. If your supplier cannot show that record, the post-machining tolerance is built on an unknown substrate.
The practical test for a buyer is simple. Ask for the heat number on the certificate and the same heat number on the inspection record for your part. If the two cannot be linked, the traceability claim is decoration.
- 1Check on receiptGrade, heat number, hardness, and a spectroscopy sample on critical alloys.
- 2Check before releaseResidual stress on long or thin parts; consider stress relief before finishing.
Fixturing decides whether the tolerance is reachable
A fixture that lets the part move by 0.02 mm under cutting force makes a ±0.005 mm tolerance meaningless. This is the most common reason a process that worked on the prototype fails in a 500-piece run: the prototype was held in a vise with light cuts, and the production job was not.
For complex geometry, the fixture should support the part at the points where the cutting force pushes hardest, and it should locate from a datum that survives all operations. On a thin-wall housing, that often means a custom soft jaw or a vacuum plate rather than a standard vise. It costs more up front and saves the run.
Clamping force is part of the same decision. Over-clamping a thin wall distorts it while it is held, so it measures correctly in the machine and springs back out of tolerance once it is released. If the part is measured in the fixture only, that error is invisible.
A useful rule: if the part can be pushed by hand and you can feel movement, the fixture is not ready for a finishing pass.
Set cutting data from simulation, then prove it on the first part
Speeds and feeds chosen by habit tend to sit in one of two bad places: too conservative, which burns cycle time and work-hardens stainless, or too aggressive, which shows up as chatter and tool wear rather than an obvious dimension error. Chatter is the harder problem, because the part can pass a dimensional check and still fail on surface finish or fatigue life.
Simulation helps before the first cut. Tool path simulation catches gouges, collisions, and excessive tool engagement on deep cavities where a holder can reach the wall. Force and deflection modeling gives a starting point for the radial depth of cut on long tools, which is where most chatter starts.
The simulation output is a starting point, not a certificate. The proof is the first article: measure it, listen to the cut, and look at the chip color and shape. Then adjust and lock the program. On a repeat order, that locked program plus a locked tool list is what keeps the run consistent.
Tool life needs a number too. Counting parts per edge and changing inserts on a schedule prevents the slow dimension drift that appears when a worn tool is left in the spindle for one more shift.
- 1Simulate forDeep cavities, long reach tools, 5-axis tilting, and any path near the fixture.
- 2Prove withFirst-article inspection plus chip and finish check before the run is released.
- 3LockProgram revision, tool list, and insert change interval per operation.
Which control point applies to your part
Match the part geometry and tolerance to the control that actually reduces risk.
| Part condition | Main risk | Control that pays off |
|---|---|---|
| Thin wall, ±0.01 mm or tighter | Distortion from clamping and cutting | Soft jaws or vacuum plate, light finishing passes |
| Long shaft, L/D over 8 | Deflection and taper | Steady rest, reduced radial depth, tailstock support |
| Multiple setups, 5 axes | Datum shift between operations | In-process probing after each roughing pass |
| Deep cavity, reach over 5×D | Chatter and tool holder collision | Tool path simulation before the first cut |
| Aerospace or medical alloy | Wrong grade or untraceable heat | Incoming spectroscopy, heat number on record |
| High-volume run, 10,000+ parts | Slow dimension drift | Scheduled insert changes, SPC on key features |
Environment, calibration, and the quality system behind them
A shop floor that swings 10 °C between morning and afternoon will move a 500 mm aluminum part more than the tolerance you are trying to hold. Thermal control matters most on long parts and on tight tolerances, and it works both ways: the machine, the part, and the gauge should sit in the same conditions before measurement. Let a part cool to room temperature before final inspection, or you are measuring a shape that will not exist tomorrow.
Calibration and preventive maintenance keep the machine honest. A machine that has not been checked for squareness, spindle runout, and thermal drift will produce a part that is correct at the probe and wrong at the CMM. The schedule should be calendar-based, not breakdown-based, because a spindle that fails mid-run takes the batch with it.
The last control point is the supplier's quality system, and it is the one a buyer can actually audit. A certified system means documented incoming inspection, defined in-process checks, and a final inspection step that releases the part. Ask how many parts are inspected at final, who signs the report, and whether the report travels with the shipment.
GreatLight operates three plants covering 7,600 m² with 127 CNC machines, including 16 simultaneous 5-axis centers and a Ø400 mm rotary table. Tolerances run to ±0.005 mm (±0.0002 in) with finishes from Ra 0.2–0.8 μm on request. Every part is inspected before shipment, with raw material checks, in-process monitoring, and a final report available on request. The quality system is certified to ISO 9001:2015, IATF 16949:2016, ISO 13485:2016, and ISO 27001:2022.
Questions engineers ask before releasing a run
How do we decide which features need in-process probing?
Start with the features that are hard to recover after finishing: a datum that gets covered, a thin wall that moves, or a form tolerance that depends on an even allowance. If a feature can be measured at the end and reworked if needed, it does not need probing.
The other trigger is setup count. Two or more setups on a tight part almost always justify a probe after each roughing pass, because the error from setup one is carried into setup two and cannot be separated later.
What should incoming material inspection include?
Grade verification, heat number, and hardness on the sample, plus a spectroscopy check on critical alloys like 7075, 17-4PH, or Ti-6Al-4V. The heat number on the mill certificate should match the number on the part record.
For powder-based processes, add particle size distribution per batch. For long or thin parts, consider stress relief before the finishing operation so the part does not move after machining.
How often should a CNC machine be calibrated?
Set a calendar interval based on how tight the work is and how many shifts the machine runs. Squareness, spindle runout, and thermal drift are the checks that affect tolerance most. A machine producing ±0.005 mm work needs a tighter interval than one on ±0.1 mm work.
Preventive maintenance belongs on the same schedule. Replacing a worn ball screw or spindle bearing on a plan is far cheaper than scrapping a batch in the middle of a run.
Can a supplier hold ±0.005 mm across a large batch?
Yes, when the process is controlled: stable fixture, locked program, scheduled tool changes, and temperature-stable measurement. It is not a property of the machine alone.
The failure mode is drift, not a single bad cut. SPC on two or three key features catches drift before the parts go out of tolerance.
What documentation should ship with the parts?
At minimum, a final inspection report covering the drawing's critical dimensions, with the material heat number referenced. On regulated programs, add first-article inspection, SPC charts on key features, and traceability from raw material to finished part.
Ask for the report format before the order starts. Retrofitting a traceability record after the parts are packed is slow and sometimes impossible.
Does surface finish need its own control step?
Yes, because a part can pass a dimensional check and still fail on function. Finish affects fatigue life, sealing, and coating adhesion. Measure Ra on the features where it is called out, and check it on the first article and at intervals during the run.
Chatter and tool wear are the usual causes when finish drifts. If Ra moves, check the tool and the fixture before changing the program.
Send the drawing, get a manufacturability review
Quotation and free DFM analysis within 12 hours. Uploads are secure and confidential, and an NDA is available on request.
12-hour quote100% inspection before shipment±0.005 mm toleranceNo minimum order quantity
