What Can Be Learned From CNC Machine

• 1
  Key takeaways
• 2
  Tolerance Is a Budget, Not a Number
• 3
  What Tool Wear Teaches About Process Control
• 4
  Setup Count and Fixturing Decide Real Cost
• 5
  Surface Finish Is a Cutting Parameter, Not a Promise
• 6
  Inspection Data Closes the Loop
• 7
  What Each Signal Tells You

A drawing calls out ±0.005 mm. The machine does not deliver that number by itself. It divides the budget across spindle runout, fixture stiffness, thermal growth, and the cutting edge. When a feature lands out of spec, the cause is almost always one item in that chain, not the control system.

Watch a 5-axis center hold a bore for two hours. The first 20 parts sit near nominal. By part 60 the dimension drifts a few microns. That drift is thermal: the spindle and ballscrews warm up, and the machine grows. A shop that understands this will warm up, measure a master part, and adjust offsets before the run starts.

The lesson for a designer is simple. Do not spend the whole tolerance on one feature. If a bore is ±0.005 mm and the mating shaft is ±0.005 mm, the assembly has no room for surface finish or plating thickness. Leave clearance somewhere. Tolerance is a system property, not a single callout.

Bilateral tolerances cost more than unilateral ones. A ±0.005 mm callout forces the shop to control both sides of nominal. A one-sided tolerance of +0.010 / −0.000 mm lets the operator aim at a single target. On a 10,000-part run, that shift alone can cut cycle time and scrap.

• 1
  Thermal driftWarm up the spindle, then re-zero offsets before critical cuts.
• 2
  Fixture stiffnessA flexible workholding setup adds error the control cannot correct.
• 3
  Stack-upTwo tight parts can still fail if the assembly budget was never checked.

A carbide end mill does not fail suddenly. It wears. Flank wear grows, cutting forces rise, and the finished dimension creeps. An operator who checks parts every 20 cycles sees the trend before the part is out of tolerance. The lesson is that process control is a trend, not a pass/fail gate.

Aluminium 6061 cuts cleanly at 3,000–8,000 rpm with high feed rates. Titanium Ti-6Al-4V does not. It work-hardens, holds heat at the edge, and wears tools fast. The same program that runs 200 aluminium parts may only run 15 titanium parts before a tool change. That difference is a planning input, not a surprise.

In-process probing changes the picture. A probe measures a datum after each tool change and updates the work offset. The machine compensates for wear before the next part. On a 4,000 mm part, that correction keeps a long bore straight where a static offset would drift.

The engineering meaning is clear. Tool life data belongs in the quote. A shop that tracks wear and probe results can hold ±0.005 mm across a run. A shop that does not will hand you a chart of drifting dimensions and call it normal.

• 1
  Log tool life by materialAluminium, steel and titanium have very different edge life.
• 2
  Probe after tool changeUpdate work offsets instead of trusting a static setup.
• 3
  Watch the trend, not one partA slow drift is a warning you can still correct.

Machine hour rate is the number buyers ask about. Setup count is the number that actually moves the price. A part that needs three fixtures on three machines carries three datum transfers, three chances for error, and three queues. A part that fits one 5-axis setup avoids all of that.

This is where DFM work pays. Move a hole so a tool can reach it from the same side. Add a boss that gives the vise something to grip. Open a corner radius from 1 mm to 3 mm so a standard cutter can clear it. None of these change function. All of them remove a setup.

The travel envelope matters too. GreatLight runs machines with travel from 500 × 310 × 200 mm up to 4,000 × 400 × 150 mm, plus a Ø400 mm rotary table. A part that fits the envelope in one orientation is cheaper than one that must be re-fixtured to reach the back side.

A useful rule: count the faces that carry toleranced features. If they sit on two or more sides, expect a second setup unless the part fits a 5-axis or mill-turn platform. Ask for the setup plan before you approve the quote.

• 1
  Datum transferEvery new fixture adds one more place for error to enter.
• 2
  Tool accessDesign corners and pockets around the cutters that will machine them.
• 3
  Envelope fitCheck the part against real machine travel before committing.

Ra 0.8–1.6 μm is a normal machined finish. Ra 0.2–0.8 μm takes a finer step-over, a sharper edge, and often a second pass. Ra 1.6–3.2 μm is as-machined and cheaper. The finish you specify changes cycle time before it changes anything else.

A 0.4 mm step-over with a 6 mm ball nose leaves visible scallops. Drop to 0.1 mm and the scallops nearly vanish, but the path length grows four times. That is the trade. When a drawing calls Ra 0.4 μm across a large face, the shop must decide between a long finishing pass and a secondary process like lapping or polishing.

Some features cannot be finished by the cutter at all. A deep slot with a 2 mm width and 20 mm depth cannot take a finishing pass with a rigid tool. The finish will be rough, and the drawing should say so. Otherwise the shop will quote a process that cannot deliver the callout.

The lesson is to separate cosmetic finish from functional finish. A sealing face needs a controlled Ra. A bracket that nobody sees does not. Specify the functional ones, and let the rest run as-machined.

• 1
  Match finish to functionSealing and bearing surfaces earn a tight Ra; covers do not.
• 2
  Watch step-overHalving step-over roughly doubles the finishing path length.
• 3
  Flag deep slotsNarrow, deep features often cannot be finished in one pass.

A part that passes inspection is one data point. A run that passes inspection with a recorded trend is a process. Raw material check, in-process monitoring, and final inspection together form the loop that keeps the next order on target. 100% inspection before shipment is the last gate, not the only one.

First article inspection matters most on a new program. It confirms that the offsets, the fixture, and the program match the drawing. If the first article is good but part 50 is not, the problem is wear or thermal drift, not the setup. The data tells you which.

Documented inspection supports certification work. ISO 9001:2015 and IATF 16949:2016 both expect traceable records. For medical work, ISO 13485:2016 adds stricter control over process validation. A shop that already runs these systems treats inspection reports as routine output, not a special request.

Ask for the inspection plan during quoting. Which features are measured, on what instrument, and how often. If the answer is vague, the ±0.005 mm callout is a hope rather than a plan.

• 1
  First article firstConfirm setup and program before the run scales up.
• 2
  Trend, not snapshotPeriodic checks catch drift that a single end-of-run check misses.
• 3
  Records travel with the partReports on request support audits and traceability.