CNC Processing Welfare Guide: What Actually Pays Off in a Shop

• 1
  Why setup count drives cost more than cycle time
• 2
  How tolerance stacks across datums and setups
• 3
  Tool access and surface finish on curved geometry
• 4
  Material behavior changes the whole calculation
• 5
  Inspection closes the loop on process capability
• 6
  When to choose three-axis, four-axis or five-axis

Most cost arguments about machining start with spindle speed. In practice, the number of times a part is unclamped and re-referenced decides the price more often than the feed rate does. Every extra setup adds a fixture, a zero-point, an operator decision and a chance for a datum shift.

A three-axis machine reaches one face per setup. A block with features on five sides needs five setups, or three if you accept two angled fixtures. Each refixture re-establishes the relationship between features, and that relationship is where the tolerance is actually spent.

When the same block is cut on a simultaneous five-axis center, the workpiece stays clamped once and the tool is oriented instead. The geometric relationship between features is set by the machine kinematics, not by how well a fixture was seated the second time.

That is the first rule of this CNC processing welfare guide: count setups before you compare hourly rates. A shop with a lower hourly rate but two extra setups is usually the more expensive supplier.

• 1
  Rule of thumbEach additional setup typically adds one fixture and one datum transfer.
• 2
  Where it bitesHoles on opposing faces, slots crossing a corner, angled bosses.
• 3
  Where it does notFlat plates with features on one face and one edge.

A drawing that calls out ±0.005 mm on a single feature is easy to hold. The same callout between two features cut in different setups is much harder, because the error of the second setup adds to the error of the first. Machinists call this a stack, and it is arithmetic, not opinion.

Take a housing with a bore on face A and a matching bore on face B. If each setup contributes ±0.005 mm of positional error, the true center-to-center relationship can drift by roughly twice that. On a single-setup five-axis process, only one positioning event contributes.

This is why we ask for the datum scheme before quoting a tight part. A drawing that dimensions everything from one origin is cheaper to make than one that chains dimensions across three faces. Same part, same tolerance numbers, very different risk.

For parts that need it, we hold ±0.005 mm (±0.0002 in) and report the measured values. What we cannot do is hold a stacked callout as if it were a single feature. That is a drawing problem, not a machine problem.

• 1
  One originDimension from a single datum and the stack collapses.
• 2
  Chained dimensionsEach link adds its own error to the total.

On a three-axis machine, the tool axis stays vertical. Cutting a curved surface means stepping the tool over in small increments, and the leftover scallop height is what you see as surface roughness. Finer steps give a better finish and a longer cycle.

Five-axis control lets the tool tilt so it stays closer to normal to the surface, or approaches tangentially. The stepover can be wider for the same scallop height. The finish comes off the machine closer to final, which reduces hand polishing on curved molds and impellers.

Tool length matters too. A deep pocket cut with a long, thin tool deflects and chatters. Tilting the part or the head lets a shorter, stiffer tool reach the same floor at an angle. Less deflection means better dimensional control and longer tool life.

Where the geometry is flat and open, none of this helps. A plate with through holes is not a five-axis job, and quoting it as one adds cost for nothing.

• 1
  Good fitImpellers, mold cores, angled ports, deep pockets.
• 2
  Poor fitPrismatic plates, simple bushings, flat brackets.
• 3
  Finish targetRa 1.6–3.2 μm as machined, Ra 0.8–1.6 μm with a finishing pass.

Aluminium 6061 and 7075 cut fast and hold a good finish, but 7075 moves more after roughing because of internal stress. A part with thin walls can distort between roughing and finishing. Leaving uniform stock and taking a light finish pass controls this better than cutting to size in one go.

Stainless 316 and 17-4PH work-harden. If the tool rubs instead of cutting, the surface gets harder and the next pass is worse. That means a rigid setup, a sharp tool and a feed that stays above the rubbing threshold. Five-axis access helps here because a shorter tool can be used.

Titanium Ti-6Al-4V and Inconel generate heat at the cutting edge and conduct it poorly. Tool life drops quickly if the contact angle is wrong. Orienting the tool to control the contact arc is one of the few real levers a machinist has on these alloys.

Plastics and carbon fibre behave differently again. POM and PEEK need sharp tools and air blast rather than flood coolant. Carbon fibre dust is abrasive, so tool wear is a cost line, not a surprise.

• 1
  Stress reliefRough, stress relieve, then finish on thin aluminium walls.
• 2
  Work hardeningNever let the tool dwell in 316 stainless.
• 3
  HeatTi-6Al-4V needs high pressure coolant and controlled contact.

A process is only as good as what you can measure. If a feature cannot be reached by a probe or a CMM stylus, its tolerance is a claim rather than a fact. We plan the inspection method at the same time as the toolpath, not after the parts are cut.

For tight work, in-process probing checks the datum and the first critical feature before the part leaves the machine. That catches a fixture shift while there is still material to correct it. Final inspection is done on all parts before shipment.

Reports are available on request, including dimensional results and material certificates. For medical and automotive work, the paperwork trail matters as much as the numbers, which is why we hold ISO 9001:2015, IATF 16949:2016 and ISO 13485:2016.

The honest limit: a tolerance you cannot verify is a tolerance you should not specify. If the drawing demands it, say how it will be measured, or accept a slightly looser value that can be proven.

• 1
  In-processProbe the datum and first critical feature on the machine.
• 2
  Final100% inspection before shipment; reports on request.