Czech CNC Machining: Accuracy of Each Part

• 1
  What czech cnc machining accuracy actually measures
• 2
  Which machine holds which accuracy level
• 3
  How material behavior moves the achievable number
• 4
  How we prove accuracy on each part
• 5
  Accuracy targets by feature type

Accuracy is not one number. On a machined part it splits into three separate behaviors: dimensional accuracy, which is how close a feature lands to its nominal size; geometric accuracy, which covers flatness, parallelism, perpendicularity and roundness; and surface accuracy, which is the Ra value left by the tool. A part can hit ±0.005 mm on a bore diameter and still fail because the bore axis is 0.03 mm out of parallel to the datum face. Engineers who treat accuracy as a single score usually get surprised at first article inspection.

The machine sets a ceiling. A 3-axis vertical mill with 500 × 500 × 450 mm travel behaves differently from a 5-axis center cutting a 4,000 mm frame. Thermal growth in the ballscrew, spindle runout, and the stiffness of the fixture all move the tool tip. On a 100 mm aluminum part, a 2 °C shop temperature swing can shift a critical dimension by roughly 0.002 mm before the cutter touches metal.

So when a drawing says ±0.005 mm, the shop is not promising that every feature on the part will land inside that band. It is promising that features called out at that tolerance can be held, given the right material, geometry, and inspection method. Features left at general tolerance follow a looser rule, often ±0.1 mm on machined edges.

The practical takeaway: accuracy lives per feature. Write the tolerance on the dimensions that control fit and function. Leave the rest alone, and the part gets cheaper without getting worse.

We run 127 high-precision CNC machines across three wholly-owned plants. The split matters because each machine class has a natural accuracy window. 16 simultaneous 5-axis machining centers handle contoured surfaces, angled holes, and parts that need five faces in one setup. 12 four-axis mills cover rotational work such as shafts with cross-drilled holes. 27 three-axis machines take flat plates, pockets, and simple prismatic parts. 16 mill-turn centers combine turning and milling so a single chucking holds concentricity.

Setup count drives accuracy more than spindle specs. Every re-clamp adds a locating error, typically 0.01–0.03 mm depending on fixture quality. A part machined on three faces in three setups accumulates that error three times. The same part cut on a 5-axis center in one setup keeps all three faces tied to one datum. That is why we push contoured or multi-face parts toward 5-axis even when a 3-axis machine could reach the geometry.

Size changes the picture too. Our largest travel is 4,000 × 400 × 150 mm. Long parts bend under their own weight and under cutting force, so a 4,000 mm frame needs support every 600–800 mm and light finishing passes. Medium travels of 750 × 1,150 × 550 mm and 600 × 600 × 600 mm cover most enclosure and manifold work. Compact travels of 500 × 500 × 450 mm and 500 × 310 × 200 mm are the most rigid, and that rigidity is why small parts often hold the tightest numbers.

A Ø400 mm rotary table extends 4-axis work to round parts that would otherwise need a second op. If your part is round with off-axis features, say so at quote time. It changes the process plan.

Aluminum is the forgiving choice. 6061-T6, 7075, and 6082 cut clean, hold ±0.005 mm on a well-fixtured feature, and reach Ra 0.8–1.6 μm without special tools. 7075 is stronger but more prone to distortion on thin walls. If you have a 1 mm wall in 7075, expect to rough, stress-relieve, then finish. Skipping that step is the most common cause of a wall that bows after machining.

Stainless 303 and 304 machine with more heat and more tool wear. 316L and 17-4PH work-harden, so a dwell at the cut line raises local hardness and pushes the next pass off size. We manage this with constant feed, sharp inserts, and coolant directed at the cut zone. Tight tolerances in 17-4PH are achievable but the process window is narrower, and we often add an in-process check after the semi-finish pass.

Titanium TC4 (Ti-6Al-4V) and Inconel are the hard cases. Both hold heat at the cutting edge, so thermal growth in the workpiece is real. A 200 mm titanium rib can move 0.01–0.02 mm during a heavy roughing pass and come back as it cools. The fix is sequence: rough, let it normalize, semi-finish, measure, then finish. Magnesium AZ31B and AZ91D cut fast but need chip control because fine magnesium chips ignite easily.

Plastics behave differently again. POM and PEEK hold good dimensional stability, ABS and PP move with temperature, and carbon fibre eats tool edges. In carbon fibre, accuracy is limited by delamination at the cut exit more than by the machine. Tell us the fibre direction if it matters.

A tolerance claim is only worth the measurement behind it. Our flow is raw material check, in-process monitoring, then final inspection. Every part gets 100% inspection before shipment, and reports are available on request. That does not mean every dimension is measured on a CMM. It means the features that matter are measured, and the method is matched to the tolerance.

Calipers and micrometers handle general dimensions down to about ±0.02 mm. Below that, we move to bore gauges, height gauges, and a CMM. Surface finish needs its own instrument; a visual check cannot confirm Ra 0.2–0.8 μm. For hard-to-reach features such as deep bores or internal radii, we cut a sample or use a profile projector.

In-process monitoring is where most scrap gets caught. Operators check the first part, then at set intervals through the run. If a dimension drifts toward the limit, the offset is corrected before parts leave the window. This is also why our qualification rate sits at 99.99%.

For regulated programs, the paper trail matters as much as the number. We hold ISO 9001:2015, IATF 16949:2016, ISO 13485:2016, and ISO 27001:2022. The first three cover quality and medical device process control. ISO 27001 covers how your drawings and CAD files are handled, which is a separate risk from whether the part is round.