Consistent CNC Machining Is Accurate: How Repeatability Works

• 1
  What Consistent CNC Machining Is Accurate Actually Means
• 2
  Thermal Stability and Machine Geometry Set the Baseline
• 3
  Tool Wear Is the Largest Single Drift Source
• 4
  Clamping Force Moves Thin Parts Between Operations
• 5
  You Cannot Hold What You Cannot Measure
• 6
  Material Behavior Changes the Recipe
• 7
  Where Consistency Is Hard and When It Is Not Worth Chasing
• 8
  How to Prove a Process Is Repeatable Before Mass Production
• 9
  What Drives Variation and How It Is Held

Accuracy is the distance between the cut and the nominal dimension. It is measured on one feature, on one part, at one moment. Consistency is the spread of that distance across many parts and many hours. A shop can hit ±0.005 mm on a single cavity and still drift 0.04 mm by the end of a 500-part run. Those are two different numbers, and they come from two different sources.

The distinction matters when you assemble. A locating bore at Ø12 H7 with a tight mean but a wide spread will still jam dowel pins, because the fit depends on the worst pair, not the average pair. The same logic applies to a shaft that carries two bearings, or to a housing that stacks four shims.

So when we say consistent CNC machining is accurate, we mean the process holds its position. Spindle growth, tool wear, chip load and fixture stiffness all move the cut. Consistency is the rate at which those movements are predicted, compensated and contained. The rest of this page is about that control loop.

Cast iron and polymer concrete beds absorb vibration, but they still expand. A spindle that rises 8 °C during a roughing cycle can grow 15–25 μm in Z. If you finish the part at 9 a.m. and the next one at 3 p.m., the operator sees a shift that looks random and is not.

On our simultaneous 5-axis centers we run spindle chillers and monitor bearing temperature. The rotary table is Ø400 mm and the axes are calibrated against a ballbar on a fixed schedule. Geometry error shows up as a taper or a bowed face, and it grows slowly, so it is easy to miss on a short sample.

Warm-up matters more than most programmers expect. A 20-minute warm-up cycle before the first finish pass costs less than scrapping the first two parts of a titanium batch. On long runs, we prefer to keep the machine running through breaks rather than let it cool to ambient and re-heat.

• 1
  Chilled spindleHolds bearing temperature inside a narrow band during long finishing cuts.
• 2
  Scheduled ballbar checksCatches geometry drift before it becomes a taper on the part.
• 3
  Warm-up cycleStabilizes Z growth before the first tolerance-critical pass.

A carbide end mill loses edge sharpness as it cuts. Flank wear raises cutting force, which pushes the tool away from the wall, so the slot gets narrower and the floor gets higher. In aluminum at moderate speed, the change is small over one part. In 17-4PH stainless or Ti-6Al-4V, it can reach tens of microns within a few parts.

The practical answer is not to sharpen more often. It is to measure the tool and offset it. In-process probing on the machine, or an offline presetter, gives a real number for length and diameter. The controller applies that number on the next part. Wear becomes a known, compensated quantity instead of a surprise.

Tool life also depends on the path. A constant chip load keeps wear even along the flute. A path that leaves 0.2 mm for the finish pass on one wall and 0.8 mm on another will wear the tool unevenly, and the two walls will not match each other even on the first part.

• 1
  Probe the tool, not the part onlyTool offsets correct length and diameter before the cut.
• 2
  Even stock allowanceUniform finishing stock keeps flank wear balanced along the flute.
• 3
  Log wear by materialTitanium and stainless consume edges faster than 6061 aluminum.

A thin wall that is straight in the vise can spring back after unclamping. The measured dimension on the machine is not the dimension the customer receives. The same happens on ring-shaped parts and on any pocket floor under 1.5 mm thick.

We treat clamping as a variable, not a constant. Soft jaws machined to the part profile, torque-controlled bolts and light finishing passes reduce the stored stress. On some parts we cut the finishing pass with the clamps relaxed, holding the part on a vacuum plate or a low-melt fixture instead.

Fixtures wear too. A locating pin that has run 5,000 cycles has a smaller diameter than when it was ground. If the pin is the datum, the part moves. Replaceable wear pads and periodic pin inspection keep the datum honest. This is unglamorous work, and it is where a lot of batch variation is born.

A CMM sitting in a warm inspection room and a shop-floor caliper do not agree. Before blaming the machine, check the measurement. Gauge R&R on a critical feature tells you how much of the spread comes from the gauge itself. If the gauge contributes 30% of the tolerance, the process number is not trustworthy.

Calibration intervals, probe stylus condition and part temperature all feed in. Aluminum parts pulled straight off the machine are warmer than the inspection room, and a 100 mm aluminum part shrinks about 5 μm over a 5 °C drop. Let it sit, or measure at the same temperature the drawing assumes.

We inspect 100% of parts before shipment, with raw material check, in-process monitoring and a final inspection. Reports are available on request. For a first article, we prefer to run a capability study on the critical features rather than signing off on one good part.

• 1
  Gauge R&R firstSeparate gauge error from process error before you change the program.
• 2
  Same temperatureLet parts stabilize before final measurement on tight features.
• 3
  Capability studyCpk on 30 parts says more than a single first article.

6061-T6 cuts clean and holds size well, which is why it is the default for prototypes and fixtures. 7075 is stronger and more notch-sensitive, so a sharp edge matters more. On thin 7075 ribs, tool pressure can deflect the wall far more than thermal drift ever will.

Stainless grades work-harden. A dwell in the cut raises local hardness, and the next pass cuts a harder surface than the last one. Constant feed and no dwelling on the surface keeps the cut predictable. 17-4PH in the H900 condition is harder again, and we usually plan for more finishing passes rather than a heavy one.

Titanium and Inconel push heat into the tool instead of the chip. Cutting speed drops, and the wear rate climbs. For a consistent titanium batch we budget tool changes by time and part count, not by the look of the edge. PEEK and other engineering plastics move with moisture and temperature, so they are measured after stabilization, not immediately.

• 1
  6061-T6Stable and predictable, good for first articles and tight batches.
• 2
  Stainless and 17-4PHNo dwell in the cut, plan extra finishing passes.
• 3
  Titanium and InconelChange tools on a schedule, not on visual inspection.
• 4
  PEEK and PCLet the part stabilize before final measurement.

Consistency costs money. Probing cycles, tool presetting, capability studies and softer fixtures all add time. On a decorative bracket with a ±0.25 mm profile tolerance, none of it changes the outcome. Spending there is waste.

It also has hard limits. A part with a 0.5 mm wall in 7075 will move no matter how good the machine is, because the material itself is unstable after machining. A deep 8 mm slot in a 3 mm cutter will deflect. At that point the answer is a design change, not a tighter process.

The realistic ceiling for a controlled batch process is around ±0.005 mm on features that can be probed and compensated. Features that depend on clamping, thin walls or long tool reach will be looser. Knowing which category a feature falls into is the difference between a workable quote and an argument at incoming inspection.