CNC Machine Parts Fit: Seven Proven Steps to Get It Right

• 1
  Why In-Tolerance Parts Still Fail to Assemble
• 2
  Choosing Materials for Matched Assemblies
• 3
  Setups, Fixturing and Workholding Tradeoffs
• 4
  Prototypes Through Production Without Drift
• 5
  Inspection for CNC Machine Parts Fit
• 6
  Cost Drivers and Sourcing Coordination
• 7
  Seven Proven Steps for CNC Machine Parts Fit
• 8
  Which Fit to Specify for Common Joints

A shaft measures 24.995 mm on the CMM. The housing bore measures 25.010 mm. Both sit inside their individual tolerances, and the clearance is theoretically 15 μm. Then the shaft refuses to seat. Nothing is out of spec. The problem is that the requirement that matters never appeared on either drawing.

That hidden requirement is the relationship between features, not the size of one feature. A print can tell you a hole is Ø8.00 ±0.05 mm. It does not tell you the hole has to sit 0.03 mm from a shoulder so a dowel lines up on the far side. The second number is what separates a conforming part from one that actually goes together.

Stack-up is where this shows up. Machine a bore to the middle of its band. Machine the mating boss to the middle of its band. In isolation both are fine. In the assembly the accumulated variation can push the fit past the clearance you designed. Repeat that across a bolted joint, a bearing seat and a dowel pattern, and a build that worked in prototype starts to drift in production.

So the goal is not tighter tolerances everywhere. Banding every dimension at ±0.005 mm raises cost fast and fixes nothing if the datums are wrong. The goal is the right tolerance in the right place, measured the right way. The rest of this guide covers the seven decisions that get you there.

Mating parts should not always be the same material. Two 304 stainless surfaces in sliding contact are prone to galling. Pair 304 with a bronze bushing, or specify 316L on one side and a hard-coated 17-4PH on the other. The material choice is a wear decision, not just a strength decision.

Thermal expansion matters more than most engineers expect. Aluminium 6061 expands about 23 × 10⁻⁶ per °C. Steel 4140 expands about 12 × 10⁻⁶ per °C. A 200 mm aluminium housing with a steel shaft grows roughly 0.046 mm over a 10 °C shift. If your clearance is 20 μm, that shift alone closes the fit.

For bearing seats and press fits, hardness drives the interference you can hold. A 6061-T6 housing at H7 is fine for light loads. For a repeated press fit, move to 7075 or steel. The bore will hold size longer and resist brinelling at the seat edge.

Corrosion pairing is the last check. Aluminium and stainless in contact with moisture form a galvanic couple. Anodize the aluminium or add a barrier washer. It is a cheap fix at design time and an expensive one after the parts are in the field.

• 1
  Same material, sliding contactRisk of galling. Break it with a coating or a dissimilar pair.
• 2
  Mixed expansionCalculate the fit at the highest and lowest service temperature.
• 3
  Press-fit boresHarder material holds the seat and resists edge damage.
• 4
  Galvanic pairsAnodize, plate or isolate. Do not rely on paint.

Every re-fixture adds error. A part that runs in two ops has two datum transfers, and each transfer adds its own position error. On a 5-axis machine with a tombstone, you can often finish three or four faces in one setup. For a mating part, that is usually worth the slower cycle.

Soft jaws machined in place beat a standard vise for second ops. Bore the jaws to the actual part diameter at the actual clamping pressure. A 6061 part will deform under a vise load that a steel part shrugs off. If the bore is measured while clamped and then released, the reading is wrong.

Thin-wall parts need support, not more clamping. Add a plug, a wax fill or a sacrificial web. Removing the support after machining releases the residual stress, and the bore moves. Plan the release before you cut the final pass, not after.

Temperature is the quiet variable. A 300 mm aluminium part grows about 0.07 mm over a 10 °C shop swing. If you machine in the morning and inspect in the afternoon, the numbers will not match. For large or tight parts, let the part stabilize and record the temperature.

The prototype build is a proof of concept, not a process. It usually runs on the most capable machine, with the most experienced operator, at a relaxed pace. Production runs on the machine that is free, at the takt time the schedule allows. If the datum scheme depends on the operator's judgment, it will drift.

Capture the process when it works. Which machine, which fixture, which tool, which offsets, which inspection method. Write it down in a traveler that follows the part. This is the cheapest insurance against a build that worked once and fails on the tenth run.

For volumes from one prototype to 10,000+ parts, the same rules apply. The difference is that at volume you can afford a dedicated fixture and a functional gage. At low volume you cannot, so the print has to carry more of the intent. Make the drawing unambiguous.

Incoming material is part of the stack. A change in bar stock supplier can change the machinability and the residual stress. If a mating part is sensitive, keep the material source fixed and note it. A cheaper bar that warps after machining is not cheaper.

Inspection should answer one question: will this part assemble and function? That is not the same as checking every dimension on the print. Start with the datum features, then the features that control the fit, then the rest.

For position, use a functional gage where the volume justifies it. A gage that mimics the mating part catches stack errors that a CMM report can miss, because the CMM measures each feature in isolation. For low volume, a best-fit alignment in the CMM software is the next best thing.

Surface finish belongs in the inspection plan for any sliding or sealing surface. A bore at the right size but the wrong finish will leak or wear. Ra 0.8–1.6 μm is a common target for bearing seats and hydraulic bores. Ra 0.2–0.8 μm where a seal runs.

Record the results, not just pass or fail. A trend of bores drifting toward the high limit over a run tells you the tool is wearing. That data lets you change the tool before the parts go out of spec, not after.

Tolerance is the biggest cost lever on a mating part, and it is the one engineers control most directly. Going from ±0.05 mm to ±0.01 mm on a bore can double the cycle time and add a finishing op. Going from ±0.05 mm to ±0.1 mm often saves nothing, because the setup cost dominates.

The second lever is the number of setups. Each additional op adds fixture cost, handling and a datum transfer. Designing the part so it can be finished in one or two setups usually saves more than loosening a tolerance. It also improves the fit, because there are fewer transfers.

Sourcing coordination matters when mating parts come from different suppliers. If the housing and the shaft are made in different shops, the datum schemes have to be compatible. Send the mating drawing with the RFQ. A supplier who cannot see the mating part is guessing at the fit.

Lead time is part of the fit decision. If a prototype has to be in hand in three to five days, the design has to be simple enough to machine in that window. Complex datums and tight stacks take longer to plan and inspect. Build the schedule into the tolerance choices.