CNC Machining Cpk Requirements

How do tool wear, thermal growth and fixture repeatability actually erode CNC capability between the warm-up part and the last part of a tool life?

Use the approved drawing, purchase order, customer quality agreement, Control Plan, and study method. Numerical values are program-specific. Common examples—only when specified—include:

• Cpk ≥ 1.67 for some initial or launch studies.
• Cpk ≥ 1.33 for some established-production studies.
• Cmk ≥ 1.67 for some controlled machine-acceptance studies.
• Tool-wear features may require trend monitoring or an alternative index such as Cpm.

CNC capability is dominated by drift, not random spread. Machine capability (Cm/Cmk) on short runs almost always looks excellent; process capability across a full tool life, thermal cycle and fixture reload is where reality appears. Customer targets are typically defined by the end-customer CSR (automotive, aerospace, medical) — CNC itself has no universal capability standard.

• Bore diameter on a hydraulic body — tracked per tool change.
• Face thickness — monitored across thermal warm-up.

1. Run a machine capability study (Cmk) after PM to isolate the machine from process noise.
2. Run a process capability study across a full tool life, including reloads and shift changes.
3. Sample at start, middle and end of tool life; do not sample only at peak condition.
4. Measure thermal state — record spindle and coolant temperature at each sample point.
5. Document tool-offset compensation rules in the Control Plan: interval, trigger, magnitude, operator authority.
6. Split data by fixture position and spindle before computing capability.
7. Verify measurement frequency covers the drift wavelength, not just per-shift snapshots.

• Cmk reported as Cpk to satisfy PPAP.
• Warm-up parts included in the capability dataset.
• Tool wear compensation applied silently mid-study.
• Fixture position pooled — one loose location masked by three good ones.
• Coolant temperature not recorded; drift attributed to "operator variation".
• Measurement frequency lower than the compensation cycle, so drift is invisible.

• Machine capability substituted for process capability is the single most common source of CNC PPAP failure downstream.
• An offset "adjusted when needed" rule guarantees the study is untraceable; formalize interval and magnitude.
• Thermal growth in the first 30–90 minutes is often larger than the entire tolerance window.
• Fixture repeatability rarely shows in Cpk because parts are pulled from the same fixture position.

• Do not use Cmk to demonstrate serial production capability.
• Do not pool multi-spindle, multi-fixture data before diagnosing stream differences.
• Do not compute capability across a compensation event without segmenting the data.

• Cmk vs Cpk: Cmk isolates the machine; Cpk includes tool life, fixture, thermal cycle.
• Cpk vs Ppk gap: Widens with slow drift — a diagnostic for compensation problems.
• Capability vs SPC: Xbar-R with trend rules catches drift long before Cpk does.

Use the free Process Capability Calculator to compute Cp, Cpk, Pp, Ppk, sigma level, and predicted PPM from your measurement data — no sign-up required.

• Never accept a CNC capability study without knowing where in tool life the parts came from.
• Never allow "automatic compensation" to hide manual offsets — both must be logged.
• Never approve a PPAP based only on the first cavity/spindle of a multi-station cell.

• ✓Cmk and Cpk reported separately, not conflated
• ✓Warm-up parts excluded and documented
• ✓Sample spans a full tool life
• ✓Data split by spindle and fixture position
• ✓Offset / compensation rule formalized in Control Plan
• ✓Thermal state recorded at each sample point
• ✓Measurement frequency finer than the compensation cycle

• Process condition: statistical stability is required.
• Distribution assumption: use a distribution model justified for the data.
• Confirm process stability and measurement-system adequacy before interpreting a capability index.
• Use a justified distribution model or non-normal method when the normal model is unsuitable.
• Numerical targets shown on industry pages are common examples, not universal requirements. The contract, drawing, customer-specific requirement, Control Plan, and validation protocol take precedence.