Watch your process part by part, and see what Cp, Cpk, Pp and Ppk really mean.
Keep the crimp example, or pick one close to your own work. The example also sets how the part is checked.
Start with one question: how many parts can you sensibly measure together at each check?
| Chart | Use when |
|---|---|
| I-MR n = 1 | Only one reading makes sense per check: one value per batch or heat (bath pH, furnace temperature, hardness per heat-treat load), destructive or expensive tests (pull force, burst test), slow processes (a few parts per shift). |
| X̄–R n = 2 to 9 | Parts are quick and cheap to measure and come off the machine one after another: machining, moulding, stamping, crimping. The shop-floor standard; usually n = 3 to 5 consecutive parts. |
| X̄–S n ≥ 10 | Large subgroups, or the calculation is done by software or an automatic gauge. S uses every reading; R only uses the largest and smallest, so it loses information as n grows. |
Rule for any subgroup: the parts in one check must be made under the same conditions (same machine, cavity or spindle, consecutive parts). Mixing cavities, spindles or batches in one subgroup hides real problems.
Continuous processes (steel, aluminium, chemicals, paper): most data is one value per heat, cast, coil or pot, so I-MR is the everyday chart. Two traps: readings taken seconds apart from an online gauge are related to each other, which makes limits too tight; and if you subgroup readings within a coil, the X̄ chart will flag every coil-to-coil difference. Chart one value per coil (I-MR), or use an I-MR-R chart.
Counting instead of measuring? (good/bad, number of defects) Use attribute charts: p or np for defectives, c or u for defects. They are not covered in this simulator.
Look for: how wide the pile of dots is, and red points on the control chart below it. A red point means something in the process has changed.
| Pattern | Usually means |
|---|---|
| Point beyond a limit | A sudden change: offset, tool, lot, operator. |
| 7 in a row on one side | The average has shifted. |
| 7 in a row up or down | A drift: tool wear, warm-up. |
| Extra patterns ("All 8") | |
| 2 of 3 beyond 2σ | Early warning of a shift (outer third of the chart). |
| 4 of 5 beyond 1σ | Early warning of a smaller shift. |
| 15 in a row within 1σ | Points hug the centre: mixed streams in a subgroup, or out-of-date limits. |
| 8 in a row outside 1σ | Points avoid the centre: two sources mixed (machines, cavities, operators). |
| 14 alternating up/down | Saw-tooth: over-adjustment, or two alternating sources. |
More rules catch problems earlier but also raise more false alarms. Many plants use the basic three plus one or two extras. A single pattern signal on a long run can be chance; a point beyond a limit rarely is.
Look for: where the pile sits between the red spec lines. Capability is judged only after the control chart shows a stable process (step 2).
| Index | Data | σ used |
|---|---|---|
| Cm / Cmk | About 50 consecutive parts: one machine, one setup, one operator, one lot. Used for machine acceptance. | Std dev of all readings |
| Cp / Cpk | Subgroups collected over time (e.g. 25 × 5). Cpk is only valid when the control chart is stable. Used for ongoing production. | σ within: R̄/d2, s̄/c4 or MR̄/1.128 |
| Pp / Ppk | The same subgroups as Cp/Cpk. Used for PPAP / initial studies, and whenever the process is not stable. | σ overall: std dev of all readings |
Cp vs Pp: same data, different calculation. Cm vs Pp: same calculation, different data. Cm is high because its data is "clean", not because the formula is special.
Cm is not defined in the AIAG SPC manual; it comes from customer and OEM procedures, and details vary. Typical: Cm ≥ 2.0, Cmk ≥ 1.67. Some customers divide by 8s instead of 6s. Always check the customer's requirement.