PUMA CNC Production Accuracy: Where the Microns Come From

• 1
  Key takeaways
• 2
  What PUMA CNC production accuracy actually measures
• 3
  Thermal drift: the largest single error source
• 4
  Axis geometry and the five-axis error chain
• 5
  Tool load, deflection, and surface finish
• 6
  Probing and inspection: closing the loop
• 7
  Five checks that hold PUMA CNC production accuracy
• 8
  When five-axis helps accuracy, and when it does not

A PUMA machining center is a turning or mill-turn platform with a rigid bed and a moving column. The name refers to the machine family, not to a tolerance grade. Two shops running the same model can hold very different results, because accuracy is produced by how the machine is set up, warmed, fixtured, and measured.

Production accuracy means the same feature lands in the same place on part one and part five thousand. That is different from positioning accuracy, which is a one-time measurement with a laser interferometer on an empty table. Under cutting load, with chips and coolant in play, the two numbers diverge.

At GreatLight we quote ±0.005 mm (±0.0002 in) on the features that allow it, and we inspect 100% of parts before shipment. The number is a floor we can defend on specific geometry, not a blanket promise across every part that enters the shop.

• 1
  Positioning accuracyHow close the axis reaches a commanded point, unloaded.
• 2
  RepeatabilityHow close it returns to the same point, over and over.
• 3
  Production accuracyWhat the finished part measures after cutting, fixturing, and release.

Cast iron grows about 11 μm per meter for every 1 °C rise. A ball screw grows similar amounts. When a spindle starts from cold and runs at 12,000 rpm, the front bearing housing can rise 8–15 °C within the first hour. That pushes the tool tip forward by tens of microns and moves every dimension you cut.

The fix is boring but effective. Machines run a warm-up cycle before the first part. Coolant temperature is held near ambient with a chiller, not left to follow the shop. For tight work, an unattended run at low load for 30–60 minutes brings the structure into equilibrium before the first inspection cut.

In a Dongguan summer, shop air can swing 6–8 °C between morning and afternoon. A part cut at 08:00 and a part cut at 15:00 will not match if the machine is not temperature-controlled. We schedule ±0.005 mm work into stable windows and check the first article against the same conditions as the last.

On a three-axis machine, error sources are squareness, straightness, and scale. On a five-axis machine, two rotary axes add their own center offsets, and any misalignment multiplies as the tool moves away from the pivot. A 10 μm pivot offset can become 40 μm of position error at 200 mm from the rotary center.

That is why five-axis work depends on calibration, not just purchase price. Rotary axes need their center points measured and compensated. A ballbar or laser check confirms the linear axes before the rotary axes are trusted. On our 16 simultaneous five-axis centers, this is a scheduled task, not a repair response.

The gain is real. A part with undercut pockets, angled ports, or compound curves can often be cut in one setup instead of three or four. Each eliminated setup removes a re-clamping error and a datum transfer. Fewer setups usually matter more than a tighter machine spec.

A 12 mm end mill hanging 60 mm out of the holder bends under load. The deflection shows up as wall taper on a deep pocket, and it grows as the tool wears. Feed and depth of cut are chosen to keep the cutting force stable across the whole path, not just at the start.

For finishing passes we typically hold Ra 0.8–1.6 μm on aluminum and steel, and Ra 0.2–0.8 μm when the application calls for it. Reaching the lower number depends on tool condition, stepover, and spindle speed. Worn tools smear the surface instead of cutting it.

Chatter is the visible symptom of a load problem. It shows as a rippled wall or a whistle that changes pitch. The response is to shorten the tool, reduce radial engagement, or change spindle speed. Adding more clamping pressure is not the answer on thin parts.

A machine that measures itself can correct itself. Spindle probes and tool setters let the control check a datum, adjust a work offset, and verify a critical feature before the next part starts. This catches thermal drift during a long run instead of discovering it at final inspection.

In-process probing is not a substitute for a CMM. It is a fast, in-machine check that keeps the process centered. Final inspection on a coordinate measuring machine confirms the geometry independently, and reports are available on request.

GreatLight runs raw material checks, in-process monitoring, and final inspection on every order. The qualification rate across production is 99.99%. That number comes from catching drift early, not from inspecting harder at the end.